Confidential: For Review Only · 2010-10-17 · Confidential: For Review Only ABSTRACT Objectives: To investigate associations compared to warfarin between direct oral anticoagulants

Confidential: For Review Only

Risk and benefits of direct oral anticoagulants versus

warfarin in real world setting: cohort studies using two primary care databases

Journal: BMJ

Manuscript ID BMJ.2017.041857

Article Type: Research

BMJ Journal: BMJ

Date Submitted by the Author: 10-Oct-2017

Complete List of Authors: Vinogradova, Yana; University of Nottingham, Primary Care Coupland, Carol; University of Nottingham, Division of Primary Care Hill, Trevor; University of Nottingham School of Medicine, Primary Care Hippisley, Julia; University of Nottingham School of Medicine, Primary Care

Keywords: apixaban, dabigatran, rivaroxaban, warfarin, bleeding, stroke, mortality, primary care

https://mc.manuscriptcentral.com/bmj

BMJ

Confidential: For Review OnlyRisk and benefits of direct oral anticoagulants versus warfarin in real

world setting: cohort studies using two primary care databases

Yana Vinogradova, Carol Coupland, Trevor Hill, Julia Hippisley-Cox

Authors

Yana Vinogradova Research Fellow in Medical Statistics1

Carol Coupland Professor in Medical Statistics in Primary Care1

Trevor Hill Research Statistician1

Julia Hippisley-Cox Professor of Clinical Epidemiology & General Practice1

Institutions

1 Division of Primary Care, 13

th floor, Tower Building, University Park, University of Nottingham,

Nottingham, NG2 7RD.

Correspondence to: Y Vinogradova

Email: [email protected]

Telephone: 0115 8466939

Fax: 0115 8466904

Keywords: apixaban, dabigatran, rivaroxaban, warfarin, bleeding, ischaemic stroke, observational

study, primary care.

Word count: 4932

Page 1 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Confidential: For Review OnlyABSTRACT

Objectives: To investigate associations compared to warfarin between direct oral anticoagulants

(DOACs) and risks of bleeding, ischaemic stroke, venous thromboembolism and mortality.

Design: Prospective open cohort of incident oral anticoagulant users between 2011 and 2016.

Setting: UK general practices contributing to QResearch and Clinical Practice Research Datalink

linked to hospital, mortality and social deprivation data.

Participants: 132,231 warfarin, 7,744 dabigatran, 37,863 rivaroxaban and 18,223 apixaban users

without anticoagulant prescriptions for 12 months before study entry, sub-grouped into with atrial

fibrillation (AF) and without (non-AF).

Intervention: Initiated DOAC or warfarin therapy.

Main outcome measures: Primary outcome: major bleed leading to hospital admission or death.

Secondary outcomes included ischaemic stroke and venous thromboembolism as efficacy measures.

All-cause mortality was also studied. Analyses were adjusted for demographics, life style,

comorbidities, and other prescribed medications.

Results: Compared to warfarin, apixaban was associated with a decreased risk of major bleed (HR

0.62, 95%CI 0.53 to 0.72), including intracranial (0.49, 0.35 to 0.68) and gastro-intestinal (0.67, 0.53

to 0.83). Dabigatran and rivaroxaban were associated with decreased risks of intracranial bleeding

(0.42, 0.27 to 0.66; 0.72, 0.58 to 0.90). Findings were consistent for AF and non-AF sub-cohorts.

Rivaroxaban was associated with increased risk of all-cause mortality (1.35, 1.27 to 1.43), consistent

for both sub-cohorts. A small increased all-cause mortality risk for apixaban (1.15, 1.06 to 1.25) was

statistically significant in the non-AF sub-cohort (1.21, 1.06 to 1.38) but not in the AF sub-cohort

(1.11, 1.00 to 1.23).

Page 2 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Confidential: For Review OnlyOver six months, the lowest number needed to treat (avoid extra major bleed) was for apixaban

(157, 95%CI 127 to 214), the lowest number needed to harm (observe extra death) was for

rivaroxaban (100, 80 to 129).

Conclusions: Overall, apixaban was found to be the safest drug, particularly for patients with

increased risk of intracranial or gastro-intestinal bleeding. There was a decreased risk of major

bleeding events associated with the use of apixaban when compared with warfarin. Both apixaban

and rivaroxaban were associated with increased all-cause mortality risk when compared with

warfarin but the increased risk was significantly higher for rivaroxaban than for apixaban. This

deserves further study especially given the increasing use of these drugs.

Page 3 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Confidential: For Review OnlyWhat is already known on this subject

Randomised controlled trials on patients with non-valvular atrial fibrillation have demonstrated non-

inferiority anticoagulating qualities for DOACs compared with warfarin. Such trials are, however,

based on carefully selected cohorts and are closely monitored. Observational studies of this topic,

aiming to show outcomes in a more real-world environment, have, however, been inconsistent both

in terms of patient selection and study design, and most have studied only patients with non-

valvular atrial fibrillation.

What this study adds

Our study demonstrated a decreased risk of major bleeding events associated with the use of

apixaban when compared with warfarin. This included any gastrointestinal bleed, upper

gastrointestinal bleeds and intracranial bleed. Dabigatran and rivaroxaban were not associated with

a decreased risk of major bleeding events but were associated with a significantly decreased risk of

intracranial bleed compared with warfarin. Both apixaban and rivaroxaban were associated with

increased all-cause mortality risk when compared with warfarin but the increased risk was

significantly higher for rivaroxaban than for apixaban.

Page 4 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Confidential: For Review OnlyINTRODUCTION

Anticoagulants are used for prevention and treatment of venous thromboembolism1 and for the

reduction of stroke risk2 in patients either with atrial fibrillation3 or after acute pulmonary embolism,

deep vein thrombosis, hip or knee replacement surgery or acute coronary syndrome.4 Warfarin has

been used for six decades but in the last eight years its use has been gradually replaced by a novel

class of direct-acting oral anticoagulants (DOACs). Unlike warfarin these drugs have set doses and

do not generally require regular international normalisation ratio (INR) blood test monitoring.5 They

also have faster onset and offset of action. There are, however, some concerns regarding the safety

of DOACs with respect to bleeding because there is an absence or a limited choice of antidotes,

which can be expensive.6 7

Atrial fibrillation is the most common condition requiring anticoagulants, and most clinical trial

evidence has been based on this group of patients. These trials have established the non-inferiority

in anticoagulating qualities of DOACs compared to warfarin in controlled trial settings,8-10 but there

are residual concerns regarding their safety, particularly in more real-world settings where they are

prescribed to a broad range of patients. A recent meta-analysis has shown that apixaban has

advantages over warfarin, providing a better balance between efficacy and safety.11

The included

studies were, however, differently designed, and none provided data for all DOACs. These findings,

therefore, represent only indirect comparisons between different types of DOACs derived from

network meta-analysis techniques.

The majority of well-powered observational studies have also focused on patients with atrial

fibrillation.12-24

Only two have provided data for the wider population,13 15

only one of which

presented results for the group without atrial fibrillation.13

Both studies were based on

commercially insured patient data, containing billing-related information, and were conducted a few

years ago, since when the profile of anticoagulant prescribing has changed, with a growing

preference for DOACs. Our study aims, for all incident users of anticoagulants, to compare the risks

Page 5 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Confidential: For Review Only(major bleeding and mortality) and benefits (reduced ischaemic stroke and venous

thromboembolism) associated with the three commonest types of DOACs when compared with

warfarin. We also provide separate results for the group with atrial fibrillation and for the group

prescribed the drugs because of other conditions.

Page 6 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Confidential: For Review OnlyMETHODS

Data sources

Two UK primary care databases QResearch (www.qresearch.org) and CPRD (www.cprd.com) were

used. Each is representative of the national population in terms of contributing practices and of

patients.25 Both have been widely validated against other sources of information and used in a wide

range of clinical studies.26

All 1457 QResearch (version 42) and 357 CPRD (November 2016) practices

were linked at the patient level to hospital admissions data, which provided dates and diagnoses for

hospitalisation (HES, http://content.digital.nhs.uk/hes). These practices were also linked to

mortality data supplied by the Office for National Statistics, which include diagnoses and dates of

death (ONS, www.ons.gov.uk). Most patients in linked practices also had information on their level

of deprivation based on quintiles of Townsend score and provided by Census 2011

(www.ons.gov.uk/census/2011census).27

We used READ codes to extract the information from

general practices and ICD-10 codes for HES and ONS data (eTable1).

Study design

We used a new user design to captures all events occurring after starting treatment and reduce the

impact of confounding.28

For a study period from January 2011 to the latest date of HES link data

(October 2016 for QResearch and March 2016 for CPRD), patients prescribed the oral anticoagulants

warfarin, dabigatran, rivaroxaban and apixaban, and aged 21 to 99 years at study entry date, formed

the cohort. Entry date was defined as the date of the first prescription of any of the anticoagulant

drugs. To facilitate a direct comparison between new-users of DOACs against new-users of warfarin,

and to help minimise the impact of indication bias, patients were excluded if they had any

anticoagulant prescription in the last 12 months before the entry date. To ensure the quality of

data, patients were also excluded if they had either fewer than 12 months of records prior to entry

or had no valid Townsend score.

Page 7 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Confidential: For Review OnlyPatients were followed from their first prescription of an anticoagulant until they experienced an

outcome of interest or were censored. Users were censored for stopping or suspending treatment

(at 30 days after the expected end date of any prescription where the gap between the expected

end date and the start date of any subsequent prescription was more than 30 days); if they switched

treatment (at the day before the prescription start for a different anticoagulant); if they left a

practice (at the day of deregistration) or the end of the study period.

Outcomes

To assess the scale of unintended side-effects of anticoagulant treatment, the primary outcome was

a major bleed after study entry leading to hospital admission or death, based on linked hospital or

mortality records. The first occurrence was used in the analyses of specific outcomes including

intracranial bleed, haematuria, haemoptysis, and gastrointestinal bleed (also separated into upper

and lower, where recorded), because these were identified as possibly preventable and potentially

life-threatening or life changing.

To assess the efficacy of anticoagulant treatments, secondary outcomes were ischaemic stroke,

venous thromboembolism and all-cause mortality, with outcome date being the earliest record after

study entry from GP, hospital and mortality data records. For ischaemic stroke and venous

thromboembolism outcomes patients with relevant records of a diagnosis prior to the start of DOAC

treatment were excluded from the analysis because of difficulties in distinguishing between events

and event reviews in the primary care records.

Exposure to anticoagulants

Three DOACs – dabigatran, rivaroxaban and apixaban – were compared to warfarin. The newest

anticoagulant, edoxaban, was not included because it was licensed for use in the UK only at the end

of 2015. Two other oral anticoagulants, acenocoumarol and phenindione, were also not included

because they were rarely prescribed in the UK.

Page 8 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Confidential: For Review OnlyExtracted data for prescriptions contained the preparation details, number of days and number of

tablets per day. The daily dose was estimated for each prescription and categorised as reduced and

standard: for dabigatran, <300mg and ≥300mg; for rivaroxaban, <20mg and ≥20mg; for apixaban,

<10mg and ≥10mg. Precise dosages for warfarin were not available because they vary according to

INR measurement and are not consistently recorded in general practice.

Confounding factors

It is possible that patients at higher risk of bleeding may be preferentially prescribed DOACs rather

than warfarin, so all analyses were adjusted for demographic and clinical variables, either because

they may have been used as indicators for prescribing a specific anticoagulant or because they have

possible associations with increased risk of stroke, venous thromboembolism (VTE) or bleeding. We

similarly adjusted for co-morbidities, previous events and medications also used as indicators or

associated with increased risks.29

The covariates were assessed at the date when the anticoagulant

was first prescribed.

Demographic and life-style variables, included because they affect the risk of bleeding, ischemic

stroke or VTE, were: gender; age at study entry;30

self-assigned ethnicity; smoking status; alcohol

use;31

deprivation.30 32

All categories are available from Table 1. Clinical values included for the

same reason were body mass index and systolic blood pressure.

Co-morbidities were included if recorded before the therapy start were: alcohol dependence, atrial

fibrillation or flutter; bleeding disorders; cancer (the 12 most-commonly occurring types); chronic

liver disease or pancreatitis;31

congestive cardiac failure; chronic obstructive pulmonary disease;

chronic renal disease;31

coronary heart disease; diabetes; dyspepsia or heartburn; treated

hypertension;31

previous stroke/transient ischaemic attack; oesophageal varices; peptic ulcer;

valvular heart disease; VTE; previous bleed (including intracranial, haematuria, haemoptysis or

gastrointestinal). If recorded in the 6 months prior to the start of anticoagulant treatment, falls or

hip fractures and hip- or knee-replacement operations were both included in the analysis.

Page 9 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Confidential: For Review OnlyRecent and concurrent medications, included in the analysis because they may affect bleeding risk or

interact with anticoagulants, were: proton pump inhibitors; macrolide antibiotics; antiplatelets;31

antidepressants;33

anticonvulsants (phenytoin or carbamazepine); NSAIDS; corticosteroids; statins.

For women, hormonal treatment included hormone replacement therapy and oral contraceptives

and was also added to the analysis of VTE outcome because they may increase VTE risk.

Finally, year of study entry was included as a confounder because of changes in recorded rates of

outcomes over the study period and because the balance of prescribing between different

anticoagulants was changing. Specifically, rates of bleeding, ischaemic stroke and VTE were

changing in the general population, while, at the beginning of the study, warfarin was

overwhelmingly the most common anticoagulant prescription but, by the end of the study,

combined prescription rates for DOACs were considerably higher than for warfarin.

Statistical analysis

The baseline characteristics for each group of patients and anticoagulant of interest were described

as percentages, means (standard deviations) or medians (interquartile ranges). Incidence rates for

each outcome were calculated based on the numbers with the outcome and the person-years of

follow-up and were age-standardised for each drug. To estimate risks associated with each type of

anticoagulant, an outcome-specific Cox model containing all confounding factors was used, with

warfarin as a primary reference. To quantify differences between apixaban and other DOACs an

additional analysis was run with apixaban as a reference.

To account for a log-normal distribution, logarithm of body mass index was used. Age was included

using fractional polynomials. Patients with missing ethnicity were categorised as white. Missing

values for body mass index, smoking status, alcohol consumption and systolic blood pressure were

assumed as missing at random and imputed using chained equations.34

The imputation model used,

was outcome-specific including outcome, length of follow-up, all confounders, anticoagulant type

and prescribed dose. The results obtained from QResearch and CPRD were pooled using a fixed

Page 10 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Confidential: For Review Onlyeffect model with inverse variance weights. Where any heterogeneity was detected, the results

were combined using a random effect model.35

Our main results were based on the cohort of all patients who started anticoagulants in the study

period. We also included an additional analysis to facilitate comparison of our results with findings

from other studies. In this, the group of patients with atrial fibrillation was selected and analysed

separately from the remaining sub-cohort of patients with other indications for anticoagulant

prescription.

To estimate the absolute magnitude of outcome risks associated with different DOACs when

compared with warfarin, we calculated numbers needed to treat/harm using the adjusted hazard

ratios and baseline rates for warfarin.36

This was estimated by weighting rates from QResearch and

CPRD. We calculated the numbers for six months after treatment commenced and at 12, 18 and 24

months.

Three sensitivity analyses were run. Hospitalisation for bleeding, ischaemic stroke or VTE may result

in a change of anticoagulant used without any subsequent GP records of this. So, in the first

sensitivity analysis, patients hospitalised for one of these were censored at the time of the

hospitalisation in the analysis of other outcomes. The second sensitivity analysis was run only on

patients with complete data to assess the validity of the assumption that missing data were missing

at random.

A final sensitivity analysis, using propensity score weighting,37

was run on the sub-cohort with

complete data. Three separate propensity scores were developed, the first to predict the use of

dabigatran among dabigatran and warfarin users, the second to predict use of rivaroxaban among

rivaroxaban and warfarin users, and the third to predict use of apixaban among apixaban and

warfarin users. All available variables described as confounding factors were included in the

Page 11 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Confidential: For Review Onlydevelopment of the propensity scores. Three separate Cox models were then run, where the use of

each DOAC in turn was adjusted for the relevant propensity score.

Patient involvement

Patient representatives from the QResearch Advisory Board have written the information for

patients on the QResearch website about the use of the database for research. Patients were not

involved in setting the research question, the outcome measures, the design or implementation of

this study. Lay people and patient representatives were involved in the writing and approving of lay

summaries during the bid process, and have been and will be involved in the writing and approving

of lay summaries being prepared for the dissemination phase. Patient groups, which will include

people prescribed the oral anticoagulants warfarin, dabigatran, rivaroxaban and apixaban will also

advise on dissemination, including the use of lay summaries describing the research and its results.

[Note: We are currently working with patients on the dissemination phase. By the time we have

addressed the reviewers’ comments, this patient involvement subsection will have been updated.]

Page 12 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


RESULTS

Cohort characteristics

156,005 patients from QResearch and 40,056 from CPRD, who started or restarted anticoagulants

between 2011 and 2016 were eligible for inclusion. Overall 53% were diagnosed with atrial

fibrillation (81,251 in QResearch and 22,019 in CPRD) leaving 47% of patients prescribed

anticoagulants for other indications (74,754 in QResearch, 18,037 in CPRD)[fig1, eTable 1].

Overall there were 132,231 patients (67%) on warfarin, 7,744 (4%) on dabigatran, 37,863 (19%) on

rivaroxaban and 18,223 (9%) on apixaban. Although 67% of patients were prescribed warfarin, its

proportion declined during the study period from 98% in 2011 to 23% in 2016 over both databases.

By contrast DOAC use had risen, from 1% to 42% for rivaroxaban and from 0% to 31% for apixaban.

Dabigatran reached a peak in 2013 (10%) and dropped to 3% in 2016[fig2].

Tables 1 and 2, which describe the included patients, show consistency between the cohorts derived

from the two databases. Patients were exposed to warfarin for longer than to DOACs, with a

median exposure of 8 months across the databases. By comparison, rivaroxaban had the shortest

median duration of 4 months, dabigatran had six months and apixaban five. Patients on different

anticoagulants differed by age, with the group on apixaban having an average age of 76 years and

the group on warfarin having an average age of 71 years. The dabigatran group had the highest

proportion of patients (70%) with atrial fibrillation, the rivaroxaban group the lowest (43%). There

were more patients with previous VTE in the warfarin (31%) and rivaroxaban (25%) groups than in

the dabigatran (5%) and apixaban (10%) groups. There were more patients with ischaemic stroke in

the dabigatran (22%) and apixaban (23%) groups than in the warfarin (15%) and rivaroxaban (13%)

groups.

Page 13 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Confidential: For Review OnlyIncidence rates

Follow-up time and the number of events are presented in Table 3, demonstrating similar age-sex

standardised rates in the two databases. Gastrointestinal bleeding was the most common bleeding

outcome, with the highest rates for rivaroxaban (13.4 per 1000py in QResearch and 15.8 per 1000py

in CPRD) and the lowest for apixaban (8.5 per 1000py in QResearch and 9.7 per 1000py in CPRD).

Haematuria was the second most common (from4.5 to 10 per 1000py across all drugs and both

databases). Haemoptysis and lower GI bleed had many fewer events.

Associations with anticoagulant exposure

Adjusted hazard ratios were similar between the databases across all outcomes [eTable 3].

Apixaban was associated with a significantly lower risk of major bleed than warfarin (adjusted hazard

ratio 0.62, 95% confidence interval 0.53 to 0.72 in combined analysis) [Figure 3] or the other two

DOACs [eTable 4]. All DOACs had significantly lower risks of an intracranial bleed than warfarin, but

only apixaban had a lower risk for gastrointestinal bleeds.

The risk of stroke did not differ for any of the anticoagulants studied. Compared with warfarin,

efficacy with respect to reducing VTE was significantly lower for rivaroxaban (1.56, 1.40 to 1.75) but

higher for dabigatran (0.26, 0.17 to 0.40) and apixaban (0.46, 0.37 to 0.57). [Figure 3]

Overall, rivaroxaban and apixaban were associated with increased risks of all-cause mortality (1.35,

1.27 to 1.43 and 1.15, 1.06 to 1.25 respectively) compared with warfarin. Most of these deaths

were, however, due to causes other than bleeding, stroke or VTE (91% in QResearch and 88% in

CPRD).[eTable 5]

The measure of relative benefits or disbenefits of DOACs in comparison with warfarin were also

presented in terms of number needed to treat or number needed to harm. Over six months, the

lowest number needed to treat to avoid one extra major bleed was for apixaban (157, 95%CI 127 to

214). The lowest number needed to treat to avoid one extra VTE event in patients without previous

Page 14 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Confidential: For Review OnlyVTE was for dabigatran (111, 99 to 136). The lowest number needed to harm (to observe one extra

death) over six months was for rivaroxaban (100, 80 to 129) [Table 4].

Dose analysis

Overall, patients on reduced doses were on average 10 years older (80 years old vs. 70 across the

databases), more likely to be women (59% vs.43%) and to have more age-associated morbidities,

with diagnoses of hypertension (61% vs. 49%), congestive cardiac failure (13% vs. 7%), coronary

heart disease (25% vs. 17%), valvular heart disease (10% vs. 6%) and chronic kidney disease (4% vs.

1%). Patients in this group were also more likely to have had falls or hip fracture (14% vs. 5%), and

hip or knee replacement operations (11% vs. 2%)[eTables 6 and 7].

Age-sex-standardised rates for patients on reduced and standard doses overlapped are shown

[eTable 8]. In the adjusted analysis [Figure 3, eTable 9], although standard doses were associated

with lower risks than reduced doses, the confidence intervals overlapped for most outcomes and

drugs. The exception was risk of mortality for apixaban; the adjusted hazard ratio for standard dose

was 0.98 (0.87 to 1.10) and for reduced dose was 1.29 (1.17 to 1.42).

Subgroup analyses

Patients diagnosed with atrial fibrillation were on average 8 years older than patients without atrial

fibrillation, more likely to be ex-smokers and less likely to be current smokers [eTable 10]. Over both

databases, similar proportions of patients with AF and without AF were exposed to warfarin.

Patients with AF were more likely to have been prescribed dabigatran (6.0% vs. 2.1% without AF) or

apixaban (11.5% vs. 7.3%) and less likely to have been prescribed rivaroxaban (18.0% vs. 20.5%) than

patients without AF [eTable 10].

In the AF sub-cohort compared with non-AF sub-cohort, there were more patients with heart-

related diseases such as congestive cardiac failure (13% against 7% across the databases), coronary

heart disease (25% against 17%), treated hypertension (62% against 42%), with previous ischaemic

Page 15 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Confidential: For Review Onlystroke (19% against 13%) and with diabetes (19% against 15%) [eTable 11]. There were fewer

patients with AF who previously had a VTE event (6% against 50%). Age-sex-standardised rates were

similar or slightly lower for the AF sub-cohort for most of the outcomes. The largest difference was

for mortality rates; between 40 and 61 per 1000py across drugs and databases for the AF sub-

cohort, and between 58 and 108 per 1000py in the sub-cohort with other indications [eTables 12

and 13].

For the primary outcome of major bleeding, the adjusted hazard ratios were similar in the two sub-

cohorts.[Figure 4] Risk of VTE was also similar but it was significant only for patients without atrial

fibrillation. Risk of all-cause mortality was significantly increased in rivaroxaban users compared

with warfarin in both sub-cohorts (1.20, 1.11 to 1.31 and 1.51, 1.38 to 1.65) and significantly

increased in apixaban users only in the non-AF sub-cohort (1.21, 1.06 to 1.38).[Figure 4, eTables 14

and 15].

Mortality risks in both sub-cohorts were higher for reduced doses than for standard doses. Similarly

for both sub-cohorts, the mortality risks were statistically higher for both reduced and standard

rivaroxaban doses compared to warfarin, but for apixaban, the mortality was significantly higher

only for reduced dose.[Figure 5, eTable 16]

Sensitivity analyses

Re-analysis of the whole cohort, but with patients censored if hospitalised for bleeding, stroke or

VTE, gave results which were the same as or very similar to the main analysis for all

outcomes.[eTable 17] Results from the complete case analysis were comparable to the main

analysis [eTable 16]. Analyses adjusted with propensity scores also resulted in similar hazard ratios

compared to the complete case analysis [eTable18].

Page 16 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Confidential: For Review OnlyDISCUSSION

Statement of principal findings

Our study demonstrated a decreased risk of major bleeding events associated with the use of

apixaban when compared with warfarin. This included any gastrointestinal bleed, upper

gastrointestinal bleeds and intracranial bleed. Dabigatran and rivaroxaban were not associated with

a decreased risk of major bleeding events but were associated with a significantly decreased risk of

intracranial bleed compared with warfarin. Risk of primary VTE was higher in patients with

rivaroxaban and lower for dabigatran and apixaban, but it was significant only in patients without

atrial fibrillation. Both apixaban and rivaroxaban were associated with increased all-cause mortality

risk when compared with warfarin but the increased risk was significantly higher for rivaroxaban

than for apixaban. Significantly increased risk for apixaban was only in patients on reduced doses,

but for rivaroxaban, it was for both standard and reduced doses.

Strengths and weaknesses of the study

This was a comprehensive study using the most recent data, so one of the study strengths is its

representativeness in terms of new users (or restarters) of anticoagulant therapy. All data were

routinely collected and included, not only comorbidities and any medications, but information on

life-style factors such as smoking and alcohol not commonly available to other studies.12 15 17 18 21 22

The general practice records were linked to hospital and mortality data, so all the outcomes were

identified. The study used the two largest primary care databases in the UK to deliver statistical

power, represents a significant additional contribution to evidence from other major studies.

Consistency in records of comorbidities, life-style and prescribing across the databases also

facilitated combination of results from each, so delivering narrower confidence intervals for our

estimations.

Page 17 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Confidential: For Review OnlyAn important limitation for our study and all earlier observational studies is the lack of information

on patient adherence to their prescribed medications, which may have led to possible

misclassifications of exposure. In analyses of bleeding, rates of bleeding would tend towards the

baseline level, shifting hazard ratios towards unity for patients who did not adhere to prescribed

dosages or discontinued anticoagulant treatment. This effect of non-adherence on bleeding rates

has been shown using commercial insurance data.38

Non-adherence of any kind may, however, contribute to underestimation of the efficacy of drugs in

preventing ischaemic stroke or VTE. Warfarin has been shown to have the highest non-persistence

and apixaban and rivaroxaban the lowest.39

In our study, apixaban and rivaroxaban were prescribed

to an older population group, which would have had a higher baseline all-cause mortality rate than

the group on warfarin or dabigatran. Being part of a treatment group associated with higher

persistence, a greater proportion of the older patients on apixaban and rivaroxaban may have died

while still using anticoagulants, but from age-related causes other than ischaemic stroke or VTE.

Non-adherence may also be linked in more complex ways to differences in apparent efficacy

between drugs, but this would need a study with access to adherence information for all treatments.

Exposure in our study was also based only on GP records, without information from other possible

sources of anticoagulants such as anticoagulant clinics or hospital stays. There is also some

uncertainty surrounding VTE diagnoses in QResearch and CPRD because the results of diagnostic

tests are not available in primary care records. This might lead to a misclassification of the outcome

and a slightly increased rate of VTE. This may, however, happen to patients using any anticoagulant

drug and we are not aware of any systematic differences between the prescribing of these drugs,

but we accept a possible shift in results towards unity.

These uncertainties could have affected our results in a number of ways. We may have included

some patients who had had exposure to anticoagulants in the 12 months prior to their entry.

Included patients hospitalised for bleeding events might also have stopped anticoagulant therapy

Page 18 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Confidential: For Review Onlyand then suffered a stroke, developed VTE or died, so leading to their misclassification as

anticoagulant users. Our sensitivity analysis censoring such patients did not, however, require

alterations to our conclusions. We also lacked information about over-the-counter purchases of

other medications such as NSAID’s or aspirin, but this is likely to have affected only a small number

of patients.

Atrial fibrillation is one of the most common indications for anticoagulant prescribing so almost all

observational studies provide evidence for this restricted group. Approximately the same numbers

of patients without AF are, however, also prescribed anticoagulants, creating a gap in knowledge

about the effects of these drugs. Such patients are different in their comorbidities and indications

for prescribing, so the risks of stroke, VTE and mortality are unlikely to be the same. We believe that

our findings for the whole cohort provide more generalisable evidence than findings based only on

the subset of patients with AF. This is because bleeding risk is likely to be more dependent on the

factors relating to the medication itself, such as which drug is used and at which dose, rather than

the indication for prescribing a particular drug. Baseline characteristics for the cohorts with and

without atrial fibrillation were broadly similar which supports this view. Presenting only aggregated

results for the sub-cohort without AF, however, highlights risks associated with DOAC drugs without

being able to be more specific about the underlying associations between different drugs and

different conditions.

To facilitate comparison with other studies, our study offers analyses separately for patients with

and without AF and for patients on different DOAC doses. Although we used a proportional hazard

model adjusting for all available confounding factors, we also undertook a sensitivity analysis using

the propensity score method and obtained very similar results.

Strengths and weaknesses in relation to other studies; important differences in results

Incidence rates of outcomes in general for patients on anticoagulants depend on a number of study

design factors. One is inclusion criteria, with incidence rates being lower for cohorts excluding

Page 19 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Confidential: For Review Onlypatients with previous events. Another, the duration of the grace period after a prescription ends

but when the patient is still considered exposed, may result in incidence rates being lower in studies

with a shorter grace period. Grace periods were not consistent across the studies, ranging from 3 to

30 days, with studies in Denmark assuming continuous treatment.18 22

Our rates were much higher

than rates from the Danish studies and from studies using the US insurance data.12 15 16 20

Although patients with valvular heart disease were excluded from some trials and observational

studies for patients with AF, a meta-analysis has demonstrated that DOAC risks compared with

warfarin for bleeding, stroke or systemic embolism and for death were similar for AF patients with or

without valvular heart disease.40

For the main outcome of major bleeding, results from our study for

the sub-cohort with AF were consistent with existing evidence from RCT’s.11

Apixaban appeared to

be associated with lowest risk of major bleeding in most of the larger studies.12 14 18 20 21

The risk of

mortality in our AF sub-cohort was similar for warfarin, dabigatran and apixaban but elevated for

rivaroxaban. Like the Danish study,22

our mortality risk in this sub-cohort was elevated for patients

on reduced doses of apixaban and rivaroxaban. The other Danish study of standard dosage showed

decreased mortality for apixaban,18

but our findings demonstrated equivalent risk to warfarin for

such patients. As for rivaroxaban, even the standard dose was associated with an increased

mortality risk, not confirming the Danish findings of an equivalent mortality risk.

The risk of stroke associated with DOACs in our sub-cohort with AF patients was equivalent to

warfarin, which is in line with the latest meta-analysis for stroke prevention11

and both Danish

studies.18 22

Similarly, we did not show any different risks of VTE for any DOACs compared with

warfarin, which is also in line with the relevant findings from the latest meta-analysis.11

Meaning of the study: possible explanations and implications for clinicians and

policymakers

Anticoagulants are prescribed for a wide range of indications but the side-effects have been studied

mostly in patients with AF.12-24

Our study has shown that the risk of major bleeding is lower in

Page 20 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Confidential: For Review Onlyapixaban users regardless of the reason for prescribing. This was most pronounced for

gastrointestinal bleeding – in particular upper-gastrointestinal bleeding. Our study has also shown

that the risk of intracranial bleeding is lower for all DOAC users and this was consistent between the

AF and non-AF patients. Increased all-cause mortality risk was found in rivaroxaban users for both

AF and non-AF patients. Although apixaban was also associated with all-cause mortality increased

risk, this was observed only in the non-AF patient cohort. Overall, apixaban was found to be the

safest drug, particularly for patients with increased risk of intracranial or gastro-intestinal bleeding.

Unanswered questions and future research

Use of DOACs in patients with AF has been extensively studied but this group represents only half of

anticoagulant users. Our study provides the evidence for this group and highlights increased risks

for stroke and all-cause mortality in the group with non-AF indications for anticoagulant prescribing.

This group includes patients undergoing preventative treatment for VTE or stroke after hip or knee-

replacements, fractures or other operations. Studying this group in detail would require further

splitting, particularly to assessing risks for stroke, VTE or mortality. We were also unable to

investigate the risks for stroke and VTE in patients who had already experienced a prior event

because it is difficult to distinguish new events from reviews of existing events in electronic health

records.

CONCLUSION

This study, based on a general population in the primary care setting, provides reassurance about

the safety of DOACs as an alternative to warfarin across all new incident users. In particular,

apixaban was found to be associated with a decreased risk of major bleeding, particularly for

intracranial and gastrointestinal bleeds and this was consistent for AF and non-AF patients.

Page 21 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Confidential: For Review OnlyOur results also give an initial, reassuring, indication of the risk patterns for all patients taking

anticoagulants. The findings for treatment efficacy with respect to stroke, VTE and all-cause

mortality, however, indicate the need for more detailed study including adherence assessment.

Page 22 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Confidential: For Review OnlyFIGURES

Figure 1 Flow of the included patients for QResearch and CPRD analysis.

Figure 2 Proportion of patients prescribed different anticoagulants in each year by database.

Figure 3 Adjusted Cox hazard ratios (95% CI) for outcomes associated with exposure to study drugs

overall and by prescribed dose compared with warfarin.

#) adjusted for age, sex, ethnicity, smoking, alcohol, Townsend quintile, BMI, SPB, falls&hipfracture, hip/knee ops,

comorbidities (alcoholism, AF, treated hypertension, CKD, COPD, liver disease, CHD, CCF, any cancer, valvular, peptic ulcer),

previous events (bleed, VTE, ischaemic stroke), medications at the baseline (macrolides, antiplatelets, anticonvulsant,

corticosteroids, NSAIDs, statin, hormones), study year

*) P-value<0.01

Figure 4 Adjusted Cox hazard ratios (95% CI) for outcomes associated with exposure to study drugs

compared with warfarin, in patients with and without atrial fibrillation.


comorbidities (alcoholism, treated hypertension, CKD, COPD, liver disease, CHD, CCF, any cancer, valvular, peptic ulcer),



*) P-value<0.01

Figure 5 Adjusted Cox hazard ratios (95% CI) for major bleed and all-cause mortality associated with

exposure to study drugs by prescribed dose compared with warfarin, in patients with and without

atrial fibrillation.





*) P-value<0.01

Page 23 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


TABLES

Table 1 Characteristics of patients in the QResearch and CPRD cohorts, percentage (numbers).

QResearch CPRD

warfarin dabigatran rivaroxaban apixaban warfarin dabigatran rivaroxaban apixaban

% (n) % (n) % (n) % (n) % (n) % (n) % (n) % (n)

Total N of patients 101,252 6,402 32,020 16,331 30,979 1,342 5,843 1,892

Median (interquartile range)

for days of treatment 240 (127, 548) 195 (60, 534) 134 (58, 327) 168 (70, 346) 231 (130, 479) 177 (67, 434) 132 (62, 274) 152 (77, 296)

Gender

Men 54.6 (55290) 56.6 (3621) 51.1 (16376) 51.8 (8453) 54.5 (16893) 59.4 (797) 50.7 (2964) 53.6 (1015)

Women 45.4 (45962) 43.4 (2781) 48.9 (15644) 48.2 (7878) 45.5 (14086) 40.6 (545) 49.3 (2879) 46.4 (877)

Age at baseline

20-29 0.8 (810) 0.2 (12) 0.8 (256) 0.2 (40) 1.1 (332) 0 1.0 (61) 0.3 (5)

30-39 2.4 (2476) 0.7 (48) 2.6 (844) 0.9 (142) 2.4 (740) 0.6 (8) 2.5 (148) 0.6 (11)

40-49 5.6 (5666) 2.5 (162) 5.5 (1764) 2.5 (412) 5.4 (1682) 2.2 (30) 5.5 (322) 1.8 (34)

50-59 9.6 (9733) 7.1 (455) 9.9 (3183) 6.8 (1108) 9.6 (2975) 9.1 (122) 9.5 (555) 6.0 (114)

60-69 20.4 (20632) 21.4 (1369) 20.0 (6398) 17.4 (2843) 20.2 (6245) 20.4 (274) 19.9 (1160) 16.6 (315)

70-79 31.4 (31798) 34.4 (2204) 28.5 (9110) 30.6 (4993) 31.4 (9719) 34.9 (468) 28.4 (1658) 31.7 (599)

80-89 25.7 (25989) 27.8 (1779) 25.8 (8246) 31.8 (5199) 26.0 (8057) 26.8 (360) 26.1 (1524) 32.6 (616)

90 and over 4.1 (4148) 5.8 (373) 6.9 (2219) 9.8 (1594) 4.0 (1228) 6.0 (80) 7.1 (415) 10.5 (198)

Mean age at baseline 70.92 (13.75) 73.83 (11.46) 71.39 (14.37) 75.36 (12.21) 70.9 (13.9) 73.7 (11.3) 71.4 (14.6) 76.1 (11.7)

Ethnicity recorded 83.6 (84635) 81.8 (5235) 80.5 (25787) 83.1 (13568) 96.7 (29951) 94.1 (1263) 94.9 (5547) 95.6 (1808)

White or not recorded (UK

or European) 95.8 (97043) 98.0 (6272) 96.4 (30878) 97.0 (15836) 97.0 (30064) 98.5 (1322) 97.7 (5708) 98.0 (1854)

Indian 0.8 (840) 0.5 (34) 0.7 (237) 0.7 (108) 0.7 (202) 0.4 (5) 0.5 (31) 0.4 (7)

Pakistani 0.5 (459) 0.2 (12) 0.4 (128) 0.4 (72) 0.2 (71) <5 <10 <5

Bangladeshi 0.2 (227) 0.1 (7) 0.1 (34) 0.2 (32) 0.1 (16) 0 <5 <5

Other Asian 0.4 (365) 0.3 (17) 0.3 (85) 0.2 (33) 0.4 (118) <5 0.3 (17) <5

Caribbean 0.9 (882) 0.5 (29) 0.8 (247) 0.5 (89) 0.4 (129) <5 0.4 (24) 0.4 (7)

Black African 0.5 (478) 0.1 (6) 0.5 (150) 0.2 (35) 0.4 (113) <5 0.3 (20) 0.3 (5)

Other 0.9 (958) 0.4 (25) 0.8 (261) 0.8 (126) 0.9 (266) 0.5 (7) 0.6 (34) 0.5 (10)

Page 24 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


BMI recorded 95.9 (97125) 95.9 (6137) 94.7 (30330) 95.6 (15614) 95.9 (29723) 95.6 (1283) 95.5 (5579) 95.8 (1812)

Mean BMI in kg/m2 (SD) 28.75 (5.73) 28.45 (5.56) 28.71 (5.82) 28.33 (5.67) 28.6 (6.0) 28.2 (5.8) 28.4 (6.0) 28.1 (5.7)

SBP recorded 99.8 (101027) 99.9 (6395) 99.6 (31888) 99.8 (16305) 99.4 (30796) 99.8 (1339) 99.3 (5804) 99.9 (1890)

Systolic blood pressure in

mmHg (SD) 131.1 (17.6) 132.0 (17.4) 131.5 (17.2) 131.1 (17.6) 132.0 (17.1) 132.3 (16.0) 131.8 (16.6) 132.7 (17.1)

Smoking

Recorded 99.7 (100998) 99.9 (6395) 99.7 (31931) 99.8 (16298) 99.9 (30953) 100.0 (1342) 99.8 (5832) 99.7 (1887)

Non-smoker 50.2 (50860) 51.4 (3293) 52.3 (16756) 52.0 (8499) 50.1 (15529) 52.5 (705) 51.9 (3035) 51.6 (976)

Ex-smoker 38.7 (39183) 40.3 (2577) 36.5 (11684) 38.9 (6346) 39.9 (12375) 40.3 (541) 37.0 (2160) 41.3 (782)

Current smoker 10.8 (10955) 8.2 (525) 10.9 (3491) 8.9 (1453) 9.8 (3049) 7.2 (96) 10.9 (637) 6.8 (129)

Current light smoker 5.7 (5757) 4.4 (284) 5.9 (1885) 5.1 (836) 5.6 (1734) 5.0 (67) 6.4 (372) 5.0 (95)

Current moderate smoker 3.1 (3153) 2.3 (149) 3.0 (961) 2.2 (359) 2.7 (830) 1.4 (19) 2.8 (165) 1.2 (23)

Current heavy smoker 2.0 (2045) 1.4 (92) 2.0 (645) 1.6 (258) 1.6 (485) 0.7 (10) 1.7 (100) 0.6 (11)

Not recorded 0.3 (254) 0.1 (7) 0.3 (89) 0.2 (33) 0.1 (26) 0 0.2 (11) 0.3 (5)

Alcohol consumption

Recorded 93.2 (94384) 93.5 (5989) 92.3 (29568) 94.3 (15394) 93.5 (28977) 92.8 (1245) 92.1 (5381) 93.4 (1767)

Non-drinker 34.6 (34991) 30.0 (1922) 34.2 (10937) 37.6 (6137) 32.9 (10199) 29.8 (400) 33.4 (1950) 33.1 (626)

Trivial (< 1 unit per day) 28.3 (28605) 27.6 (1770) 27.3 (8730) 26.0 (4247) 32.4 (10049) 30.4 (408) 31.0 (1809) 31.1 (588)

Light (1-2 units per day) 13.2 (13330) 14.2 (909) 13.0 (4149) 13.1 (2135) 16.8 (5201) 18.1 (243) 15.7 (915) 16.6 (314)

Moderate (3-6 units/day) 15.1 (15326) 19.0 (1214) 15.6 (5004) 15.1 (2470) 6.1 (1899) 7.5 (100) 6.4 (374) 6.7 (127)

Heavy (7-9 units/day) 1.3 (1300) 1.7 (107) 1.4 (453) 1.5 (249) 3.1 (953) 4.0 (54) 3.8 (220) 3.8 (71)

Very heavy (>9 units/day) 0.8 (832) 1.0 (67) 0.9 (295) 1.0 (156) 2.2 (676) 3.0 (40) 1.9 (113) 2.2 (41)

Not recorded 6.8 (6868) 6.5 (413) 7.7 (2452) 5.7 (937) 6.5 (2002) 7.2 (97) 7.9 (462) 6.6 (125)

Townsend quintiles

1 (Affluent) 20.3 (20571) 25.0 (1602) 20.2 (6455) 21.3 (3482) 24.4 (7558) 28.3 (380) 25.9 (1515) 29.1 (551)

2 20.7 (20918) 23.4 (1496) 21.2 (6780) 21.6 (3524) 24.7 (7656) 28.6 (384) 26.0 (1522) 24.8 (470)

3 20.2 (20458) 21.0 (1346) 21.6 (6930) 20.3 (3314) 21.7 (6707) 20.3 (272) 20.8 (1217) 20.9 (395)

4 19.1 (19354) 17.2 (1101) 20.2 (6464) 19.4 (3174) 18.4 (5685) 13.9 (186) 17.5 (1022) 16.3 (308)

5 (Deprived) 19.7 (19951) 13.4 (857) 16.8 (5391) 17.4 (2837) 10.8 (3355) 8.9 (120) 9.7 (566) 8.9 (168)

Page 25 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Table 2 Comorbidities, previous events and other medications at baseline for QResearch and CPRD cohorts, percentage (numbers).

QResearch CPRD

warfarin dabigatran rivaroxaban apixaban warfarin dabigatran rivaroxaban apixaban

% (n) % (n) % (n) % (n) % (n) % (n) % (n) % (n)

Total N of patients 101252 6402 32020 16331 30979 1342 5843 1892

Comorbidities at baseline

Alcohol dependence 2.7 (2754) 3.0 (192) 3.1 (987) 3.2 (522) 2.3 (698) 2.7 (36) 3.0 (177) 3.5 (67)

Atrial fibrillation 53.3 (53921) 70.8 (4534) 42.5 (13597) 56.3 (9199) 53.8 (16664) 74.7 (1003) 50.5 (2950) 74.1 (1402)

Bleeding disorders 1.1 (1127) 0.8 (52) 1.1 (357) 1.1 (175) 1.4 (424) 0.9 (12) 1.7 (97) 1.3 (25)

Cancer (any) 12.8 (12935) 12.0 (769) 13.2 (4214) 13.2 (2154) 13.1 (4044) 11.4 (153) 13.3 (778) 13.1 (247)

Chronic liver disease or

pancreatitis 1.3 (1281) 1.3 (82) 1.3 (416) 1.3 (217) 1.2 (383) 1.3 (18) 1.5 (89) 1.5 (28)

Chronic obstructive

pulmonary disease 9.1 (9254) 8.2 (528) 8.7 (2771) 10.0 (1639) 9.0 (2791) 8.2 (110) 9.1 (529) 8.3 (157)

Chronic renal disease 2.9 (2953) 1.0 (63) 1.5 (470) 2.2 (353) 3.0 (936) 1.3 (18) 2.0 (115) 2.0 (38)

Congestive cardiac failure 11.0 (11135) 9.8 (626) 7.7 (2472) 11.0 (1798) 10.3 (3191) 9.8 (131) 8.3 (486) 12.9 (244)

Coronary heart disease 21.8 (22096) 21.1 (1350) 17.2 (5493) 23.5 (3839) 22.0 (6811) 21.9 (294) 17.5 (1023) 25.4 (481)

Diabetes 17.2 (17398) 17.2 (1102) 16.3 (5215) 19.6 (3197) 16.1 (4986) 15.9 (213) 16.0 (934) 19.1 (362)

Dyspepsia 17.6 (17850) 17.7 (1133) 17.9 (5716) 18.6 (3031) 25.4 (7857) 24.5 (329) 26.1 (1525) 25.2 (477)

Falls or hip fracture (within

last 180 days) 7.4 (7529) 10.0 (639) 8.0 (2553) 7.5 (1228) 5.9 (1816) 5.3 (71) 6.3 (366) 6.6 (124)

Hip or knee operation

(within last 180 days) 1.7 (1764) 7.3 (470) 4.5 (1449) 2.0 (320) 3.2 (984) 8.6 (116) 4.9 (288) 3.1 (58)

Hypertension 52.1 (52739) 57.9 (3705) 48.1 (15416) 56.5 (9227) 53.5 (16583) 57.1 (766) 52.4 (3059) 60.8 (1150)

Ischaemic stroke* 15.2 (15413) 21.6 (1383) 13.2 (4227) 22.8 (3722) 15.1 (4664) 23.4 (314) 15.2 (891) 29.7 (561)

Oesophageal varices 0.2 (188) 0.2 (10) 0.1 (29) 0.1 (17) 0.1 (38) <5 0.1 (5) <5

Peptic ulcer 7.1 (7162) 7.3 (467) 6.3 (2011) 7.9 (1285) 7.6 (2359) 7.3 (98) 7.8 (455) 9.6 (181)

Valvular heart disease 10.6 (10686) 8.4 (536) 6.2 (1982) 9.1 (1481) 8.4 (2612) 7.0 (94) 5.1 (299) 8.4 (159)

Venous thromboembolism* 30.5 (30914) 4.6 (297) 25.1 (8052) 10.0 (1633) 33.2 (10283) 6.8 (91) 26.4 (1545) 9.9 (188)

Previous bleed (any)* 23.1 (23400) 25.0 (1600) 24.5 (7842) 25.7 (4190) 26.9 (8346) 28.5 (382) 27.9 (1632) 30.2 (572)

Previous Intracranial bleed* 1.0 (969) 1.2 (75) 1.1 (353) 1.4 (230) 1.2 (368) 1.9 (25) 1.5 (87) 2.2 (41)

Previous Haematuria 10.1 (10260) 11.9 (764) 10.7 (3440) 11.2 (1831) 11.1 (3449) 11.5 (155) 10.5 (615) 11.5 (218)

Page 26 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


*) Based on GP and HES records #) In the last 6 months prior to the drug start date

Previous Haemoptysis* 2.7 (2733) 2.3 (150) 2.7 (850) 2.6 (426) 3.5 (1073) 3.4 (46) 3.6 (212) 3.6 (68)

Previous All GI bleed* 12.3 (12501) 13.2 (848) 13.4 (4278) 14.2 (2315) 15.3 (4734) 15.9 (214) 16.3 (953) 17.9 (339)

Previous upper GI bleed* 4.2 (4255) 4.3 (277) 4.4 (1394) 5.0 (821) 4.9 (1531) 5.4 (73) 5.9 (347) 6.4 (121)

Previous lower GI bleed* 9.1 (9253) 10.1 (649) 10.1 (3250) 10.4 (1698) 11.8 (3643) 11.7 (157) 12.4 (723) 13.4 (254)

Other medications

PPI 43.1 (43634) 44.4 (2841) 40.7 (13027) 43.4 (7095) 40.9 (12671) 42.2 (566) 41.5 (2424) 43.4 (822)

Antibiotics# 10.5 (10682) 8.7 (558) 8.1 (2591) 6.2 (1019) 9.5 (2947) 7.2 (97) 8.7 (508) 5.6 (106)

Antiplatelet 25.6 (25932) 22.8 (1458) 17.8 (5708) 17.6 (2875) 32.0 (9921) 36.2 (486) 24.3 (1422) 32.2 (610)

Antidepressants 18.6 (18796) 16.2 (1039) 19.3 (6189) 18.1 (2961) 17.5 (5412) 14.6 (196) 20.1 (1174) 18.6 (352)

Anticonvulsants 1.1 (1109) 0.5 (35) 0.9 (302) 0.9 (141) 1.0 (321) 0.7 (10) 1.1 (65) 0.7 (14)

NSAIDs 9.3 (9431) 11.0 (707) 10.7 (3428) 5.7 (929) 9.1 (2828) 12.1 (163) 9.5 (556) 6.5 (123)

Corticosteroids 12.9 (13040) 10.6 (678) 10.6 (3384) 9.7 (1584) 11.8 (3646) 9.2 (124) 10.8 (630) 9.2 (175)

Statins 47.9 (48489) 52.2 (3345) 42.1 (13479) 52.8 (8630) 46.0 (14237) 52.2 (700) 42.6 (2489) 55.1 (1043)

Hormones (women) 2.0 (916) 2.3 (63) 2.2 (350) 1.5 (115) 4.6 (648) 3.1 (17) 5.9 (170) 3.0 (26)

Page 27 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Table 3 All patients: Age-sex-adjusted incidence rates per 1000py of outcomes in QResearch and CPRD

QResearch CPRD

Person-

years

N of

events

Age-sex adjusted rate

per 1000py (95%CI)

Person-

years

N of

events


per 1000py (95%CI)

Major bleeding

Warfarin 111,823 2945 26.5 (25.6 to 27.5) 30,427 896 29.6 (27.6 to 31.5)

Dabigatran 6,117 140 22.9 (19.0 to 26.9) 1,117 23 19.4 (11.4 to 27.3)

Rivaroxaban 20,581 576 27.8 (25.5 to 30.1) 3,245 107 33.2 (26.8 to 39.6)

Apixaban 10,744 190 16.6 (14.1 to 19.1) 1,099 25 20.0 (11.9 to 28.2)

Intracranial bleed

Warfarin 113,705 692 6.2 (5.8 to 6.7) 30,891 193 6.3 (5.4 to 7.2)

Dabigatran 6,218 18 2.9 (1.5 to 4.2) 1,126 <5 1.6 (0.0 to 3.9)


Apixaban 10,805 41 3.4 (2.3 to 4.4) 1,107 <5 3.1 (0.0 to 6.2)

Haematuria

Warfarin 112,791 936 8.3 (7.8 to 8.8) 30,702 267 8.7 (7.7 to 9.8)

Dabigatran 6,173 41 6.6 (4.5 to 8.6) 1,122 8 6.7 (2.0 to 11.4)


Apixaban 10,789 54 4.5 (3.3 to 5.8) 1,104 9 6.5 (2.0 to 11.0)

Haemoptysis

Warfarin 113,667 172 1.5 (1.3 to 1.7) 30,882 51 1.6 (1.2 to 2.1)

Dabigatran 6,204 10 1.5 (0.6 to 2.4) 1,126 <5 0.9 (0.0 to 2.7)


Apixaban 10,811 5 0.5 (0.1 to 0.9) 1,108 <5 0.7 (0.0 to 2.2)

All GI bleed

Warfarin 113,044 1176 10.5 (9.9 to 11.1) 30,716 395 12.9 (11.6 to 14.2)

Dabigatran 6,180 73 12.2 (9.2 to 15.2) 1,122 12 10.0 (4.3 to 15.7)


Apixaban 10,774 93 8.5 (6.6 to 10.3) 1,104 11 9.7 (3.8 to 15.5)

Upper GI bleed

Warfarin 113,143 1048 9.4 (8.8 to 9.9) 30,740 356 11.6 (10.4 to 12.8)

Dabigatran 6,181 69 11.6 (8.7 to 14.5) 1,122 11 8.9 (3.6 to 14.2)


Apixaban 10,779 87 7.9 (6.1 to 9.7) 1,105 10 8.7 (3.2 to 14.2)

Rectal bleed

Warfarin 113,685 140 1.2 (1.0 to 1.4) 30,894 43 1.4 (1.0 to 1.8)

Dabigatran 6,218 5 0.7 (0.1 to 1.3) 1,126 <5 1.1 (0.0 to 3.2)

Rivaroxaban 20,826 18 0.8 (0.5 to 1.2) 3,285 <5 0.3 (0.0 to 0.9)

Apixaban 10,808 7 0.6 (0.1 to 1.1) 1,106 <5 1.0 (0.0 to 2.9)

Ischaemic stroke

Warfarin 93,464 1165 12.7 (12.0 to 13.4) 25,517 335 13.2 (11.8 to 14.7)

Dabigatran 4,499 77 16.3 (12.6 to 19.9) 793 10 12.2 (4.5 to 20.0)


Apixaban 7,883 130 15.2 (12.5 to 17.9) 733 12 14.4 (6.2 to 22.7)

VTE

Warfarin 88,065 981 11.0 (10.3 to 11.7) 22,856 250 10.9 (9.6 to 12.3)

Dabigatran 5,976 16 2.6 (1.3 to 3.9) 1,057 7 6.7 (1.7 to 11.6)


Apixaban 9,978 108 12.3 (9.9 to 14.8) 1,015 8 7.4 (2.1 to 12.6)

Mortality

Warfarin 113,799 5409 49.3 (48.0 to 50.6) 30,808 1434 47.6 (45.2 to 50.1)

Dabigatran 6,220 287 47.4 (39.1 to 55.8) 1,122 52 43.2 (31.3 to 55.1)


Apixaban 10,812 784 59.4 (55.0 to 63.8) 1,101 86 64.4 (49.1 to 79.7)

Page 28 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Table 4 Number needed to treat/ harm (95% confidence interval) compared to warfarin

Outcome At 6 months At 12 months At 18 months At 24 months

Numbers needed to treat

Major bleeding

Apixaban 157 (127 to 214) 92 (75 to 126) 68 (55 to 93) 54 (43 to 73)

Intracranial bleed

Dabigatran 497 (394 to 839) 258 (204 to 435) 183 (145 to 310) 143 (114 to 242)

Intracranial bleed

Rivaroxaban 1049 (693 to 2884) 544 (360 to 1496) 387 (256 to 1066) 303 (200 to 833)

Intracranial bleed


All GI bleed


Upper GI bleed


VTE#

dabigatran 111 (99 to 136) 98 (88 to 120) 90 (80 to 110) 82 (73 to 101)

VTE#

apixaban 152 (131 to 190) 134 (116 to 167) 123 (106 to 153) 113 (97 to 140)

Numbers needed to harm

VTE#

rivaroxaban 148 (111 to 209) 130 (98 to 184) 119 (90 to 168) 110 (83 to 155)

Mortality

Rivaroxaban 100 (80 to 129) 60 (48 to 77) 44 (36 to 57) 36 (29 to 47)

Mortality


Note: The calculations are based on the hazard ratios derived from combined analysis. Only statistically significant

associations between the exposure and outcome are included. #) Based on patients without VTE prior to the start of anticoagulant

Page 29 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Confidential: For Review OnlySTATEMENTS

Ethics and dissemination

The protocol for QResearch has been published in eprints and was reviewed in accordance with the

requirements for the East Midlands Derby Research Ethic Committee [ref 03/4/021].41

The protocol

for CPRD has been approved by Independent Scientific Advisory Committee for MHRA Database

Research (N 16_284R).

Contributorship statement

JHC initiated the study, undertook the original literature review, prepared the grant application,

drafted the study protocol, organised the extraction of the QResearch data, advised on clinical

aspects of the study, interpretation of the results and critically reviewed the paper. CC contributed

to the development of the idea and the study design and advised on the analysis. TH did initial

exploration of the data. YV reviewed the literature, contributed to the grant application and the

study design, organised the extraction of CPRD data, did the analysis on both datasets and wrote the

draft of the manuscript. JHC, CC and TH critically reviewed the paper. YV is the guarantor of the

study. All authors have approved the submitted version.

Acknowledgements

We acknowledge the contribution of EMIS practices who contribute to the QResearch® and EMIS

and the University of Nottingham for expertise in establishing, developing and supporting the

QResearch database. The HES data used in this analysis are re-used by permission from the NHS

Digital who retain the copyright. We thank the Office of National Statistics for providing the

mortality data. ONS and NHS Digital bear no responsibility for the analysis or interpretation of the

data. QResearch acknowledges the infrastructure funding from the NIHR funded Nottingham

Biomedical Research Centre.

Page 30 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Confidential: For Review OnlyCompeting interests

All authors have completed the Unified Competing Interest form at

www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and declare

no support from any additional organisation for the submitted work. JHC is professor of clinical

epidemiology at the University of Nottingham and unpaid director of QResearch®, a not-for-profit

organisation which is a joint partnership between the University of Nottingham and EMIS

(commercial IT supplier for 60% of general practices in the UK). JHC is also a paid director of ClinRisk

Limited, which produces open and closed source software to ensure the reliable and updatable

implementation of clinical risk algorithms (including QRISK2) within clinical computer systems to

help improve patient care. There have been no other relationships or activities that could appear to

have influenced the submitted work.

Transparency declaration

The lead author affirms that the manuscript is an honest, accurate, and transparent account of the

study being reported; that no important aspects of the study have been omitted and that any

discrepancies from the study as planned have been explained.

An exclusive licence

The Corresponding Author has the right to grant on behalf of all authors and does grant on behalf of

all authors, an exclusive licence on a worldwide basis to the BMJ Publishing Group Ltd and its

Licensees to permit this article (if accepted) to be published in BMJ editions and any other BMJPGL

products and sublicences to exploit all subsidiary rights.

Funding

This work has been supported by NIHR with a SPCR round 11 grant [reference number 304]. This

report is independent research by the National Institute for Health Research. The views expressed in

Page 31 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Confidential: For Review Onlythis publication are those of the author(s) and not necessarily those of the NHS, the National

Institute for Health Research or the Department of Health. No specific grant from any commercial or

not-for-profit sectors has been received. Only the authors are responsible for analysis,

interpretation of the data and writing the report for publication.

Data sharing statement

To guarantee the confidentiality of personal and health information only the authors have had

access to the data during the study in accordance with the relevant licence agreements. Access to

the QResearch data is according to the information on the QResearch website (www.qresearch.org).

CPRD linked data will not be available because it will have to be erased on 28 April 2018 according to

the CPRD licence.

Page 32 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Confidential: For Review OnlyREFERENCES

1. National Institute for Clinical Excellence. Venous thromboembolism in adults admitted to hospital:

reducing the risk. In: NICE clinical guideline 92, ed. London: NICE, 2010.

2. Hart RG, Pearce LA, Aguilar MI. Meta-analysis: antithrombotic therapy to prevent stroke in

patients who have nonvalvular atrial fibrillation. Ann Intern Med 2007;146(12):857-67.

3. Mant J, Hobbs FDR, Fletcher K, et al. Warfarin versus aspirin for stroke prevention in an elderly

community population with atrial fibrillation (the Birmingham Atrial Fibrillation Treatment of

the Aged Study, BAFTA): a randomised controlled trial. The Lancet 2007;370(9586):493-503.

4. National Institute for Clinical Excellence. Anticoagulants, including non-vitamin K antagonist oral

anticoagulants (NOACs). Key theraputic topic 2016;KTT16

5. Lip GY. Atrial fibrillation in 2011: Stroke prevention in AF. Nat Rev Cardiol 2012;9(2):71-3. doi:

10.1038/nrcardio.2011.203

6. Holster IL, Valkhoff VE, Kuipers EJ, et al. New oral anticoagulants increase risk for gastrointestinal

bleeding: a systematic review and meta-analysis. Gastroenterology 2013;145(1):105-12 e15.

doi: 10.1053/j.gastro.2013.02.041

7. Zheng Y, Sorensen SV, Gonschior A-K, et al. Comparison of the Cost-effectiveness of New Oral

Anticoagulants for the Prevention of Stroke and Systemic Embolism in Atrial Fibrillation in a

UK Setting. Clinical Therapeutics 2014;36(12):2015-28.e2. doi:

https://doi.org/10.1016/j.clinthera.2014.09.015

8. Connolly SJ, Ezekowitz MD, Yusuf S, et al. Dabigatran versus Warfarin in Patients with Atrial

Fibrillation. New England Journal of Medicine 2009;361(12):1139-51.

9. Patel MR, Mahaffey KW, Garg J, et al. Rivaroxaban versus warfarin in nonvalvular atrial fibrillation.

N Engl J Med 2011;365(10):883-91. doi: 10.1056/NEJMoa1009638

10. Ruff CT, Giugliano RP, Braunwald E, et al. Comparison of the efficacy and safety of new oral

anticoagulants with warfarin in patients with atrial fibrillation: a meta-analysis of

randomised trials. Lancet 2014;383(9921):955-62. doi: 10.1016/S0140-6736(13)62343-0

11. Sterne J, Bodalia P, Bryden P, et al. Oral anticoagulants for primary prevention, treatment and

secondary prevention of venous thromboembolic disease, and for prevention of stroke in

atrial fibrillation: systematic review, network meta-analysis and cost-effectiveness analysis.

Health Technol Assess 2017;21(9) doi: 10.3310/hta21090

12. Abraham NS, Noseworthy PA, Yao X, et al. Gastrointestinal Safety of Direct Oral Anticoagulants:

A Large Population-Based Study. Gastroenterology 2017;152(5):1014-22.e1. doi:

https://doi.org/10.1053/j.gastro.2016.12.018

13. Abraham NS, Singh S, Alexander GC, et al. Comparative risk of gastrointestinal bleeding with

dabigatran, rivaroxaban, and warfarin: population based cohort study. BMJ 2015;350:h1857.

14. Amin A, Keshishian A, Trocio J, et al. Risk of stroke/systemic embolism, major bleeding and

associated costs in non-valvular atrial fibrillation patients who initiated apixaban, dabigatran

or rivaroxaban compared with warfarin in the United States Medicare population. Current

Medical Research and Opinion 2017;33(9):1595-604. doi: 10.1080/03007995.2017.1345729

15. Chang H-Y, Zhou M, Tang W, et al. Risk of gastrointestinal bleeding associated with oral

anticoagulants: population based retrospective cohort study. BMJ 2015;350

16. Graham DJ, Reichman ME, Wernecke M, et al. Cardiovascular, Bleeding, and Mortality Risks in

Elderly Medicare Patients Treated With Dabigatran or Warfarin for Nonvalvular Atrial

Fibrillation. Circulation 2015;131(2):157-64.

17. Hernandez I, Baik S, Piñera A, et al. Risk of bleeding with dabigatran in atrial fibrillation. JAMA

Internal Medicine 2015;175(1):18-24.

18. Larsen TB, Skjøth F, Nielsen PB, et al. Comparative effectiveness and safety of non-vitamin K

antagonist oral anticoagulants and warfarin in patients with atrial fibrillation: propensity

weighted nationwide cohort study. BMJ 2016;353

Page 33 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Confidential: For Review Only19. Lauffenburger JC, Farley JF, Gehi AK, et al. Effectiveness and Safety of Dabigatran and Warfarin in

Real-World US Patients With Non-Valvular Atrial Fibrillation: A Retrospective Cohort Study.

Journal of the American Heart Association 2015;4(4)

20. Li X, Deitelzweig S, Keshishian A, et al. Effectiveness and safety of apixaban versus warfarin in

non-valvular atrial fibrillation patients in “real-world” clinical practice. A propensity-matched

analysis of 76,940 patients. Thrombosis and Haemostasis 2017;117(6):1072-82. doi:

10.1160/TH17-01-0068

21. Lip GH, Skjøth F, Nielsen P, et al. Effectiveness and safety of standard-dose nonvitamin k

antagonist oral anticoagulants and warfarin among patients with atrial fibrillation with a

single stroke risk factor: A nationwide cohort study. JAMA Cardiology 2017 doi:

10.1001/jamacardio.2017.1883

22. Nielsen PB, Skjøth F, Søgaard M, et al. Effectiveness and safety of reduced dose non-vitamin K

antagonist oral anticoagulants and warfarin in patients with atrial fibrillation: propensity

weighted nationwide cohort study. BMJ 2017;356

23. Sørensen R, Gislason G, Torp-Pedersen C, et al. Dabigatran use in Danish atrial fibrillation

patients in 2011: a nationwide study. BMJ Open 2013;3(5):e002758.

24. Shun K, Tatsunori M, Naoko I, et al. Bleeding risk of apixaban, dabigatran, and low-dose

rivaroxaban compared with warfarin in Japanese patients with non-valvular atrial fibrillation:

a propensity matched analysis of administrative claims data. Curr Med Res Opin

2017;3(11):1955-63. doi: 10.6084/m9.figshare.5363980.v1

25. Hippisley-Cox J, Vinogradova Y, Coupland C, et al. Comparison of key practice characteristics

between general practices in England and Wales and general practices in the QRESEARCH

data. Report to the Health and Social Care Information Centre.: University of Nottingham,

2005.

26. Jick SS, Kaye JA, Vasilakis-Scaramozza C, et al. Validity of the General Practice Research Database.

Pharmacotherapy 2003;23(5):686-89.

27. Townsend P. Deprivation. Journal Social Policy 1987;16:125-46.

28. Johnson ES, Bartman BA, Briesacher BA, et al. The incident user design in comparative

effectiveness research. Pharmacoepidemiol Drug Saf 2013;22(1):1-6. doi: 10.1002/pds.3334

29. Hippisley-Cox J, Coupland C. Predicting risk of upper gastrointestinal bleed and intracranial bleed

with anticoagulants: cohort study to derive and validate the QBleed scores. BMJ

2014;349:g4606. doi: 10.1136/bmj.g4606

30. Button L, Roberts S, Evans P. Hospitalized incidence and case fatality for upper gastrointestinal

bleeding from 1999 to 2007: a record linkage study. Aliment Pharmacol Ther 2011;33(1):64 -

76.

31. Pisters R, Lane DA, Nieuwlaat R, et al. A novel user-friendly score (HAS-BLED) to assess 1-year risk

of major bleeding in patients with atrial fibrillation: the Euro Heart Survey. Chest

2010;138(5):1093-100. doi: 10.1378/chest.10-0134

32. Crooks C, West J, Card T. Upper gastrointestinal haemorrhage and deprivation: a nationwide

cohort study of health inequality in hospital admissions. Gut 2011;61:514 - 20.

33. Tata LJ, Fortun PJ, Hubbard RB, et al. Does concurrent prescription of selective serotonin

reuptake inhibitors and non-steroidal anti-inflammatory drugs substantially increase the risk

of upper gastrointestinal bleeding? Aliment Pharmacol Ther 2005;22(3):175-81.

34. Royston P. Multiple imputation of missing values. Stata J 2004;4(3):227-41.

35. Borenstein M, Hedges L, Higgins J, et al. A basic introduction to fixed-effect and random-effects

models for meta-analysis. Res Syn Meth 2010;1:97-111.

36. Altman DG, Andersen PK. Calculating the number needed to treat for trials where the outcome is

time to an event. BMJ : British Medical Journal 1999;319(7223):1492-95.

37. Sturmer T, Wyss R, Glynn R, et al. Propensity scores for confounder adjustment when assessing

the effects of medical interventions using nonexperimental study designs. J Intern Med

2014;275(6):570-80.

Page 34 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Confidential: For Review Only38. Yao X, Abraham NS, Alexander GC, et al. Effect of Adherence to Oral Anticoagulants on Risk of

Stroke and Major Bleeding Among Patients With Atrial Fibrillation. Journal of the American

Heart Association 2016;5(2)

39. Lamberts M, Staerk L, Olesen JB, et al. Major Bleeding Complications and Persistence With Oral

Anticoagulation in Non-Valvular Atrial Fibrillation: Contemporary Findings in Real-Life Danish

Patients. Journal of the American Heart Association 2017;6(2) doi: 10.1161/jaha.116.004517

40. Pan KL, Singer DE, Ovbiagele B, et al. Effects of Non–Vitamin K Antagonist Oral Anticoagulants

Versus Warfarin in Patients With Atrial Fibrillation and Valvular Heart Disease: A Systematic

Review and Meta-Analysis. Journal of the American Heart Association 2017;6(7) doi:

10.1161/jaha.117.005835

41. Hippisley-Cox J, Coupland C. Protocol to assess the effectiveness and safety of novel oral

anticoagulants (NOAC) vs warfarin in real world settings: cohort study in primary care.

eprints 2016. http://eprints.nottingham.ac.uk/35133/.

Page 35 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Flow of the included patients for QResearch and CPRD analysis.

254x190mm (96 x 96 DPI)

Page 36 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Proportion of patients prescribed different anticoagulants in each year by database.

248x180mm (72 x 72 DPI)

Page 37 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Caption: Adjusted Cox hazard ratios (95% CI) for outcomes associated with exposure to study drugs overall and by prescribed dose compared with warfarin.

Legend:

#) adjusted for age, sex, ethnicity, smoking, alcohol, Townsend quintile, BMI, SPB, falls&hipfracture, hip/knee ops, comorbidities (alcoholism, AF, treated hypertension, CKD, COPD, liver disease, CHD, CCF, any cancer, valvular, peptic ulcer), previous events (bleed, VTE, ischaemic stroke), medications at the baseline

(macrolides, antiplatelets, anticonvulsant, corticosteroids, NSAIDs, statin, hormones), study year *) P-value<0.01

254x190mm (96 x 96 DPI)

Page 38 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Caption: Adjusted Cox hazard ratios (95% CI) for outcomes associated with exposure to study drugs compared with warfarin, in patients with and without atrial fibrillation.

Legend:

#) adjusted for age, sex, ethnicity, smoking, alcohol, Townsend quintile, BMI, SPB, falls&hipfracture, hip/knee ops, comorbidities (alcoholism, AF, treated hypertension, CKD, COPD, liver disease, CHD, CCF, any cancer, valvular, peptic ulcer), previous events (bleed, VTE, ischaemic stroke), medications at the baseline

(macrolides, antiplatelets, anticonvulsant, corticosteroids, NSAIDs, statin, hormones), study year *) P-value<0.01

254x190mm (96 x 96 DPI)

Page 39 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Caption: Adjusted Cox hazard ratios (95% CI) for major bleed and all-cause mortality associated with exposure to study drugs by prescribed dose compared with warfarin, in patients with and without atrial

fibrillation.

Legend: #) adjusted for age, sex, ethnicity, smoking, alcohol, Townsend quintile, BMI, SPB, falls&hipfracture, hip/knee ops, comorbidities (alcoholism, treated hypertension, CKD, COPD, liver disease, CHD, CCF, any cancer, valvular, peptic ulcer), previous events (bleed, VTE, ischaemic stroke), medications at the baseline (macrolides, antiplatelets, anticonvulsant, corticosteroids, NSAIDs, statin, hormones), study

year *) P-value<0.01

254x190mm (96 x 96 DPI)

Page 40 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


DOAC paper, Supplementary data, eTables v 3

Re: Risk and benefits of direct oral anticoagulants versus warfarin in real world setting: cohort

studies using two primary care databases

Supplementary data

Content

eTable 1 ICD-10 codes for the outcomes ................................................................................................ 2

eTable 2 Supplementary data for Figure2: Number of patients starting anticoagulants by study year 2

eTable 3 All patients: Adjusted hazard ratios (95% CI) for outcomes associated with exposure to

study drugs compared with warfarin by database. ................................................................................ 3


study drugs compared with apixaban ..................................................................................................... 4

eTable 5 Cause of death by drug and database, proportion of deceased patients on the drug ............ 5

eTable 6 Cohort characteristics by database, percentage (number) of patients on reduced and

standard doses for DOAC's ..................................................................................................................... 6

eTable 7 Comorbidities and other medications, percentage (number) of patients on reduced and

standard doses for DOAC's ..................................................................................................................... 8

eTable 8 Rates per 1000py of outcomes by drug and dose for NOAC’s in QResearch and CPRD in all

patients ................................................................................................................................................... 9


different study drug doses compared with warfarin. ........................................................................... 11

eTable 10 Cohort characteristics and exposure by database, percentage (number) of patients with

and without atrial fibrillation ................................................................................................................ 13

eTable 11 Comorbidities, previous events and other medications, percentage (number) of patients

with and without atrial fibrillation ........................................................................................................ 15

eTable 12 Rates per 1000py of outcomes in QResearch and CPRD in patients with atrial fibrillation 16

eTable 13 Rates per 1000py of outcomes by database in patients without atrial fibrillation ............. 17

eTable 14 Patients with atrial fibrillation: Adjusted hazard ratios (95% CI) for outcomes associated

with exposure to study drugs compared with warfarin, by database .................................................. 18

eTable 15 Patients without atrial fibrillation: Adjusted hazard ratios (95% CI) for outcomes associated

with exposure to study drugs compared with warfarin, by database .................................................. 19

eTable 16 Patients with and without atrial fibrillation: Adjusted hazard ratios (95% CI) for outcomes

associated with exposure to study drug different doses compared with warfarin. ............................. 20

eTable 17 Sensitivity analysis, censoring patients at the time of hospitalisation related to other

outcome. Adjusted hazard ratios (95% CI) for outcomes associated with exposure to study drugs

compared with warfarin. ...................................................................................................................... 21

eTable 18 Complete case analysis and propensity score adjustment by database. Adjusted hazard

ratios (95% CI) for outcomes associated with exposure to study drugs compared with warfarin....... 22

Page 41 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



eTable 1 ICD-10 codes for the outcomes

outcome ICD10-code

Intracranial bleed

I60, I600, I601, I602, I603, I604, I605, I606, I607, I608, I609, I61, I610, I611, I612,

I613, I614, I615, I616, I618, I619, I62, I620, I621, I629, S065, S066

Haematuria R31, R31X

Haemoptysis R04, R041, R042, R048, R049

Upper GI bleed

I850, K226, K250, K251, K252, K254, K255, K256, K260, K261, K262, K264, K265,

K266, K270, K271, K272, K274, K275, K276, K279, K280, K281, K282, K284, K285,

K286, K290, K920, K921, K922

Lower GI bleed K625

Ischaemic stroke

G45, G450, G451, G452, G453, G454, G458, G459, I63, I630, I631, I632, I633, I634,

I635, I636, I638, I639, I64, I64X

Venous

thromboembolism I26, I260, I269, I81, I81X, I82, I820, I822, I823, I828, I829

eTable 2 Supplementary data for Figure2: Number of patients starting anticoagulants by study year

QResearch CPRD

Year Warfarin Dabigatra

n

Rivaroxab

an

Apixaban Warfarin Dabigatra

n

Rivaroxab

an

Apixaban

2011 21807 167 190 7841 54 42

2012 23608 1055 607 13 7843 246 46 <5

2013 20538 1606 2611 336 6940 412 719 78*

2014 16622 1446 6109 1802 5161 352 1733 482

2015 12573 1204 11342 5701 2836 237 2715 1046

2016 6104 924 11161 8479 358 41 588 286

*) This figure includes patients for 2012 and 2013

Page 42 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



eTable 3 All patients: Adjusted hazard ratios (95% CI) for outcomes associated with exposure to study drugs compared with

warfarin by database.

QResearch CPRD

Comparison

with warfarin

Adjusted HR (95%CI) P-value Adjusted HR (95%CI) P-value

Major bleeding

Dabigatran 0.91 (0.77 to 1.08) 0.3 0.73 (0.48 to 1.11) 0.1

Rivaroxaban 1.03 (0.93 to 1.14) 0.6 1.13 (0.90 to 1.42) 0.3

Apixaban 0.61 (0.52 to 0.71) <0.001 0.73 (0.48 to 1.11) 0.1

Intracranial bleed

Dabigatran 0.44 (0.27 to 0.71) <0.001 0.27 (0.07 to 1.10) 0.07


Apixaban 0.48 (0.34 to 0.68) <0.001 0.56 (0.20 to 1.57) 0.3

Haematuria

Dabigatran 0.89 (0.65 to 1.23) 0.5 0.89 (0.43 to 1.82) 0.7


Apixaban 0.69 (0.51 to 0.94) 0.02 1.15 (0.56 to 2.36) 0.7

Haemoptysis

Dabigatran 1.18 (0.62 to 2.27) 0.6 0.70 (0.09 to 5.15) 0.7


Apixaban 0.32 (0.13 to 0.82) 0.02 0.84 (0.11 to 6.69) 0.9

All GI bleed

Dabigatran 1.20 (0.94 to 1.52) 0.1 0.87 (0.48 to 1.55) 0.6


Apixaban 0.68 (0.54 to 0.85) 0.001 0.59 (0.32 to 1.12) 0.1

Upper GI bleed

Dabigatran 1.24 (0.97 to 1.59) 0.09 0.88 (0.48 to 1.62) 0.7


Apixaban 0.68 (0.53 to 0.87) 0.002 0.59 (0.30 to 1.14) 0.1

Rectal bleed

Dabigatran 0.86 (0.35 to 2.12) 0.7 0.70 (0.09 to 5.21) 0.7


Apixaban 0.74 (0.32 to 1.73) 0.5 0.62 (0.08 to 5.12) 0.7

Ischaemic stroke

Dabigatran 1.28 (1.01 to 1.62) 0.04 0.87 (0.46 to 1.65) 0.7


Apixaban 1.14 (0.93 to 1.41) 0.2 0.98 (0.53 to 1.82) 1

VTE

Dabigatran 0.20 (0.12 to 0.33) <0.001 0.49 (0.23 to 1.05) 0.07

Rivaroxaban 1.51 (1.34 to 1.71) <0.001 1.85 (1.40 to 2.43) <0.001

Apixaban 0.46 (0.37 to 0.57) <0.001 0.48 (0.23 to 1.00) 0.05

Mortality

Dabigatran 1.02 (0.91 to 1.15) 0.7 1.01 (0.77 to 1.33) 0.9


Apixaban 1.12 (1.03 to 1.23) 0.01 1.38 (1.10 to 1.74) 0.006

*) adjusted for age, sex, ethnicity, smoking, alcohol, Townsend quintile, BMI, SPB, falls&hipfracture, hip/knee ops,




Page 43 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



eTable 4 All patients: Adjusted hazard ratios (95% CI) for outcomes associated with exposure to study drugs compared with

apixaban

QResearch CPRD Both databases

Comparison

with apixaban

Adjusted HR

(95%CI)

P-

value

Adjusted HR

(95%CI)

P-

value

Combined analysis

HR (95%CI)

P-

value

Major bleeding

Dabigatran 1.50 (1.19 to 1.87) <0.001 1.00 (0.56 to 1.79) 1 1.42 (1.15 to 1.75) 0.001

Rivaroxaban 1.69 (1.43 to 2.00) <0.001 1.55 (1.00 to 2.40) 0.05 1.67 (1.43 to 1.96) <0.001

Warfarin 1.64 (1.40 to 1.93) <0.001 1.37 (0.90 to 2.09) 0.1 1.61 (1.38 to 1.87) <0.001

Intracranial bleed

Dabigatran 0.91 (0.52 to 1.61) 0.8 0.49 (0.09 to 2.70) 0.4 0.86 (0.50 to 1.47) 0.6

Rivaroxaban 1.41 (0.98 to 2.05) 0.07 1.74 (0.59 to 5.16) 0.3 1.45 (1.02 to 2.05) 0.04

Warfarin 2.07 (1.47 to 2.93) <0.001 1.79 (0.64 to 5.06) 0.3 2.04 (1.47 to 2.83) <0.001

Haematuria



Warfarin 1.44 (1.07 to 1.96) 0.02 0.87 (0.42 to 1.80) 0.7 1.34 (1.01 to 1.77) 0.04

Haemoptysis



Warfarin 3.10 (1.22 to 7.89) 0.02 1.19 (0.15 to 9.52) 0.9 2.64 (1.12 to 6.19) 0.03

All GI bleed

Dabigatran 1.77 (1.29 to 2.43) <0.001 1.46 (0.63 to 3.35) 0.4 1.72 (1.28 to 2.32) <0.001


Warfarin 1.48 (1.17 to 1.87) 0.001 1.68 (0.90 to 3.16) 0.1 1.50 (1.21 to 1.87) <0.001

Upper GI bleed



Warfarin 1.47 (1.15 to 1.87) 0.002 1.70 (0.88 to 3.29) 0.1 1.50 (1.19 to 1.88) <0.001

Rectal bleed



Warfarin 1.35 (0.58 to 3.14) 0.5 1.61 (0.20 to 13.22) 0.7 1.38 (0.63 to 3.02) 0.4

Ischaemic stroke



Warfarin 0.78 (0.66 to 0.93) 0.006 1.02 (0.55 to 1.89) 1 0.80 (0.68 to 0.95) 0.009

VTE

Dabigatran 0.66 (0.25 to 1.75) 0.4 1.02 (0.36 to 2.85) 1 0.81 (0.40 to 1.65) 0.6

Rivaroxaban 1.44 (0.92 to 2.26) 0.1 3.85 (1.87 to 7.91) <0.001 1.90 (1.30 to 2.78) <0.001

Warfarin 1.47 (0.94 to 2.30) 0.09 2.09 (1.00 to 4.34) 0.05 1.62 (1.10 to 2.37) 0.01

Mortality



Warfarin 0.89 (0.82 to 0.97) 0.01 0.72 (0.57 to 0.91) 0.006 0.87 (0.80 to 0.94) <0.001





Page 44 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



eTable 5 Cause of death by drug and database, proportion of deceased patients on the drug

QResearch CPRD

Cause of

death

Bleed

Cause of

death

Stroke/ VTE

Cause of

death

Other causes

Cause of

death

Bleed

Cause of

death

Stroke/ VTE

Cause of

death

Other causes

All patients

Warfarin 5% 5% 89% 4% 8% 88%

Dabigatran 2% 6% 92% 0 10% 90%

Rivaroxaban 2% 5% 94% 4% 8% 88%

Apixaban 1% 4% 95% 3% 13% 84%

Patients with atrial fibrillation

Warfarin 5% 5% 90% 4% 5% 92%

Dabigatran 3% 7% 91% 0 8% 92%

Rivaroxaban 2% 5% 92% 5% 8% 87%

Apixaban 1% 4% 95% 5% 13% 82%

Patients without atrial fibrillation

Warfarin 5% 6% 89% 4% 12% 84%

Dabigatran 3% 4% 95% 0 14% 86%

Rivaroxaban 1% 4% 95% 3% 8% 89%

Apixaban 1% 4% 95% 0% 13% 88%

Note: The percentage do not always make 100% because a few patients had records of both bleeding and

stroke/VTE as death cause and also because of rounded numbers. Because of the low numbers of deaths only

percentage is provided due to Data Protection Act.

Page 45 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



eTable 6 Cohort characteristics by database, percentage (number) of patients on reduced and standard doses for DOAC's

QResearch CPRD

Reduced dose Standard dose Reduced dose Standard dose

Total Number of patients 15,636 39,117 2,896 6,181

Anticoagulant

Dabigatran 23.4 (3655) 7.0 (2747) 27.1 (785) 9.0 (557)

Rivaroxaban 40.0 (6245) 65.9 (25775) 47.7 (1382) 72.2 (4461)

Apixaban 36.7 (5736) 27.1 (10595) 25.2 (729) 18.8 (1163)


for days of treatment 149 (60; 377) 150 (59; 342) 122 (60; 300) 146 (75; 296)

Gender

Men 40.9 (6399) 56.4 (22051) 43.9 (1271) 56.7 (3505)

Women 59.1 (9237) 43.6 (17066) 56.1 (1625) 43.3 (2676)

Age at baseline

20-29 0.2 (30) 0.7 (278) 0.4 (12) 0.9 (54)

30-39 0.7 (104) 2.4 (930) 0.9 (27) 2.3 (140)

40-49 1.6 (255) 5.3 (2083) 2.3 (68) 5.1 (318)

50-59 3.4 (531) 10.8 (4215) 4.8 (140) 10.5 (651)

60-69 9.0 (1414) 23.5 (9196) 10.4 (302) 23.4 (1447)

70-79 21.6 (3383) 33.0 (12924) 23.7 (685) 33.0 (2040)

80-89 46.4 (7259) 20.4 (7965) 42.1 (1220) 20.7 (1280)

90 and over 17.0 (2660) 3.9 (1526) 15.3 (442) 4.1 (251)

Mean age at baseline 80.1 (11.3) 70.0 (13.3) 78.4 (12.4) 70.1 (13.4)

Ethnicity recorded 81.7 (12767) 81.4 (31823) 95.6 (2770) 94.6 (5848)

White (UK or European) 96.7 (15116) 96.8 (37870) 98.0 (2838) 97.8 (6046)

Indian 0.9 (146) 0.6 (233) 0.4 (12) 0.5 (31)

Pakistani 0.3 (54) 0.4 (158) 0.2 (5) 0.1 (7)

Bangladeshi 0.1 (21) 0.1 (52) <5 <5

Other Asian 0.2 (25) 0.3 (110) 0.3 (10) 0.2 (15)

Caribbean 0.8 (130) 0.6 (235) 0.3 (10) 0.4 (23)

Black African 0.2 (36) 0.4 (155) <5 0.4 (24)

Other 0.6 (93) 0.7 (270) 0.6 (18) 0.5 (33)

BMI recorded 95.1 (14874) 95.1 (37207) 95.6 (2770) 95.5 (5904)

Mean BMI in kg/m2 (SD) 27.3 (5.5) 29.1 (5.8) 27.2 (5.7) 28.8 (5.9)

SBP recorded 99.9 (15617) 99.6 (38971) 99.7 (2886) 99.4 (6147)


mmHg (SD) 131.5 (17.9) 131.3 (17.1) 132.3 (16.6) 131.9 (16.7)

Smoking

Recorded 99.8 (15602) 99.8 (39022) 99.9 (2892) 99.8 (6169)

Non-smoker 55.9 (8746) 50.6 (19802) 55.5 (1608) 50.3 (3108)

Ex-smoker 37.2 (5811) 37.8 (14796) 37.2 (1078) 38.9 (2405)

Current smoker 6.7 (1045) 11.3 (4424) 7.1 (206) 10.6 (656)

Current light smoker 4.0 (622) 6.1 (2383) 4.3 (125) 6.6 (409)

Current moderate smoker 1.6 (248) 3.1 (1221) 1.7 (50) 2.5 (157)

Current heavy smoker 1.1 (175) 2.1 (820) 1.1 (31) 1.5 (90)

Not recorded 0.2 (34) 0.2 (95) <5 0.2 (12)

Alcohol consumption

Recorded 93.2 (14572) 93.0 (36379) 92.7 (2684) 92.4 (5709)

Non-drinker 42.0 (6570) 31.8 (12426) 39.2 (1134) 29.8 (1842)

Page 46 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



Trivial (< 1 unit per day) 27.4 (4291) 26.7 (10456) 31.3 (907) 30.7 (1898)

Light (1-2 units per day) 11.4 (1787) 13.8 (5406) 13.8 (400) 17.3 (1072)

Moderate (3-6 units/day) 11.2 (1747) 17.7 (6941) 4.8 (140) 7.5 (461)

Heavy (7-9 units/day) 0.7 (103) 1.8 (706) 2.3 (67) 4.5 (278)

Very heavy (>9 units/day) 0.5 (74) 1.1 (444) 1.2 (36) 2.6 (158)

Not recorded 6.8 (1064) 7.0 (2738) 7.3 (212) 7.6 (472)

Townsend quintiles

1 (Affluent) 20.3 (3176) 21.4 (8363) 27.0 (781) 26.9 (1665)

2 20.7 (3241) 21.9 (8559) 25.3 (733) 26.6 (1643)

3 21.4 (3348) 21.1 (8242) 20.6 (598) 20.8 (1286)

4 20.2 (3153) 19.4 (7586) 17.0 (492) 16.6 (1024)

5 (Deprived) 17.4 (2718) 16.3 (6367) 10.1 (292) 9.1 (562)

Page 47 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



eTable 7 Comorbidities and other medications, percentage (number) of patients on reduced and standard doses for DOAC's

QResearch CPRD

Reduced dose Standard dose Reduced dose Standard dose

Total N 15,636 39,117 2,896 6,181


Alcohol dependence 2.1 (322) 3.5 (1379) 2.0 (57) 3.6 (223)

Atrial fibrillation 55.3 (8653) 47.7 (18677) 59.0 (1709) 59.0 (3646)

Bleeding disorders 1.2 (194) 1.0 (390) 1.7 (50) 1.4 (84)

Cancer (any) 15.1 (2358) 12.2 (4779) 14.3 (415) 12.3 (763)


pancreatitis 1.4 (222) 1.3 (493) 1.3 (38) 1.6 (97)

Chronic obstructive pulmonary

disease 10.1 (1585) 8.6 (3353) 9.4 (273) 8.5 (523)

Chronic renal disease 3.8 (587) 0.8 (299) 4.1 (120) 0.8 (51)

Congestive cardiac failure 13.4 (2102) 7.1 (2794) 13.2 (382) 7.7 (479)

Coronary heart disease 25.4 (3966) 17.2 (6716) 24.4 (707) 17.7 (1091)

Diabetes 19.3 (3025) 16.6 (6489) 18.8 (545) 15.6 (964)

Dyspepsia 18.8 (2942) 17.7 (6938) 27.8 (804) 24.7 (1527)

Falls or hip fracture (within 180

days) 14.7 (2302) 5.4 (2118) 9.0 (260) 4.9 (301)

Hip or knee operation (within

180 days) 10.4 (1624) 1.6 (615) 11.1 (322) 2.3 (140)

Hypertension 60.4 (9438) 48.3 (18910) 62.6 (1813) 51.2 (3162)

Ischaemic stroke* 22.9 (3588) 14.7 (5744) 23.9 (692) 17.4 (1074)

Oesophageal varices 0.1 (16) 0.1 (40) <5 0.1 (6)

Peptic ulcer 8.6 (1341) 6.2 (2422) 9.5 (275) 7.4 (459)

Valvular heart disease 10.2 (1595) 6.1 (2404) 8.2 (238) 5.1 (314)

Venous thromboembolism* 9.6 (1507) 21.7 (8475) 13.3 (386) 23.3 (1438)

Previous bleed (any)* 27.6 (4311) 23.8 (9321) 30.7 (890) 27.4 (1696)

Previous Intracranial bleed* 1.7 (262) 1.0 (396) 2.2 (64) 1.4 (89)

Previous Haematuria 12.2 (1910) 10.5 (4125) 12.6 (365) 10.1 (623)

Previous Haemoptysis* 2.5 (384) 2.7 (1042) 3.2 (92) 3.8 (234)

Previous All GI bleed* 15.2 (2374) 13.0 (5067) 17.8 (515) 16.0 (991)

Previous upper GI bleed* 5.7 (899) 4.1 (1593) 6.7 (195) 5.6 (346)

Previous Rectal bleed* 10.9 (1701) 10.0 (3896) 13.2 (382) 12.2 (752)

Other medications

PPI 48.0 (7509) 39.5 (15454) 45.9 (1330) 40.2 (2482)

Antibiotics# 9.0 (1401) 7.1 (2767) 8.1 (235) 7.7 (476)

Antiplatelet 22.6 (3529) 16.6 (6512) 33.9 (981) 24.9 (1537)

Antidepressants 19.5 (3051) 18.2 (7138) 20.1 (583) 18.4 (1139)

Anticonvulsants 0.8 (125) 0.9 (353) 0.7 (20) 1.1 (69)

NSAIDs 9.9 (1547) 9.0 (3517) 10.4 (302) 8.7 (540)

Corticosteroids 11.7 (1832) 9.8 (3814) 10.4 (302) 10.1 (627)

Statins 49.1 (7683) 45.4 (17771) 49.0 (1418) 45.5 (2814)

Hormones (women) 1.5 (139) 2.3 (392) 2.8 (45) 6.3 (168)

*) Based on GP and HES records #) In the last 6 months prior the drug start date

Page 48 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



eTable 8 Rates per 1000py of outcomes by drug and dose for NOAC’s in QResearch and CPRD in all patients

QResearch CPRD

Person-

years

N of

event

s


per 1000py (95%CI)

Person-

years

N of

event

s


per 1000py (95%CI)

Major bleeding

Dabigatran <300 3,241 96 26.8 (19.5 to 34.0) 655 13 20.1 (5.6 to 34.6)

Dabigatran 300 2,876 44 28.1 (5.5 to 50.8) 461 10 56.3 (0.0 to 120.4)

Rivaroxaban <20 4,208 156 34.5 (27.6 to 41.5) 703 37 48.0 (29.6 to 66.3)

Rivaroxaban 20+ 16,374 420 26.4 (23.9 to 29.0) 2,543 70 29.4 (22.5 to 36.4)

Apixaban 5 3,774 97 26.5 (17.0 to 36.0) 400 9 9.3 (2.9 to 15.7)

Apixaban 10 6,970 93 13.5 (10.6 to 16.3) 699 16 23.8 (11.1 to 36.5)

Intracranial bleed

Dabigatran <300 3,299 <15 3.0 (1.3 to 4.7) 661 <5 0.8 (0.0 to 2.4)

Dabigatran 300 2,919 <5 5.6 (0.0 to 14.7) 465 <5 1.1 (0.0 to 3.2)



Apixaban 5 3,803 22 3.6 (1.4 to 5.8) 403 0 -

Apixaban 10 7,002 19 2.8 (1.5 to 4.1) 705 <5 6.9 (0.0 to 14.8)

Haematuria

Dabigatran <300 3,270 24 6.5 (3.6 to 9.5) 658 5 11.8 (0.0 to 25.1)

Dabigatran 300 2,904 17 8.2 (0.0 to 17.4) 464 <5 3.1 (0.0 to 6.6)



Apixaban 5 3,795 30 9.2 (3.4 to 14.9) 402 <5 2.3 (0.0 to 5.4)

Apixaban 10 6,994 24 3.4 (2.0 to 4.7) 702 6 8.4 (1.3 to 15.5)

Haemoptysis

Dabigatran <300 3,296 <5 1.6 (0.0 to 3.5) 661 <5 1.6 (0.0 to 4.8)

Dabigatran 300 2,907 <10 1.4 (0.2 to 2.5) 465 0 -

Rivaroxaban <20 4,270 6 1.7 (0.0 to 3.6) 715 <5 2.1 (0.0 to 6.3)


Apixaban 5 3,806 <5 2.7 (0.0 to 6.4) 403 <5 1.1 (0.0 to 3.4)

Apixaban 10 7,006 <5 0.1 (0.0 to 0.3) 705 0 -

All GI bleed

Dabigatran <300 3,277 55 15.5 (9.5 to 21.6) 659 6 5.7 (0.9 to 10.6)

Dabigatran 300 2,903 18 13.2 (0.0 to 31.6) 462 6 52.4 (0.0 to 116.9)



Apixaban 5 3,789 41 10.8 (4.8 to 16.7) 402 5 5.8 (0.7 to 10.9)

Apixaban 10 6,985 52 7.6 (5.4 to 9.8) 702 6 8.5 (1.5 to 15.5)

Upper GI bleed

Dabigatran <300 3,278 53 15.0 (9.0 to 21.0) 659 6 5.7 (0.9 to 10.6)

Dabigatran 300 2,903 16 12.9 (0.0 to 31.3) 463 5 50.8 (0.0 to 115.1)



Apixaban 5 3,790 38 9.9 (4.2 to 15.7) 402 5 5.8 (0.7 to 10.8)

Apixaban 10 6,989 49 7.2 (5.1 to 9.4) 704 5 7.2 (0.7 to 13.8)

Rectal bleed

Dabigatran <300 3,299 <5 0.8 (0.0 to 1.8) 662 0 -

Dabigatran 300 2,919 <5 0.3 (0.0 to 0.8) 464 <5 1.6 (0.0 to 4.8)

Rivaroxaban <20 4,271 <5 0.4 (0.0 to 0.8) 716 0 -

Rivaroxaban 20+ 16,555 <15 0.8 (0.4 to 1.2) 2,569 <5 0.5 (0.0 to 1.4)

Apixaban 5 3,805 <5 1.0 (0.0 to 2.3) 403 0 -

Apixaban 10 7,002 <5 0.4 (0.0 to 0.8) 704 <5 1.2 (0.0 to 3.7)

Page 49 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



Ischaemic stroke

Dabigatran <300 2,257 43 14.0 (9.3 to 18.6) 433 7 11.9 (1.9 to 21.8)

Dabigatran 300 2,242 34 16.6 (4.8 to 28.4) 360 <5 5.2 (0.0 to 11.1)



Apixaban 5 2,621 58 16.1 (9.1 to 23.1) 256 5 7.9 (0.7 to 15.1)

Apixaban 10 5,262 72 12.9 (9.9 to 16.0) 477 7 18.2 (3.2 to 33.2)

VTE

Dabigatran <300 3,148 10 3.2 (0.8 to 5.5) 622 <5 6.4 (0.0 to 14.0)

Dabigatran 300 2,828 6 1.3 (0.2 to 2.3) 435 <5 5.8 (0.0 to 12.6)



Apixaban 5 3,496 22 11.4 (3.3 to 19.5) 370 <5 8.6 (0.0 to 19.3)

Apixaban 10 6,482 86 15.0 (11.6 to 18.3) 646 <5 6.6 (0.0 to 13.5)

Mortality

Dabigatran <300 3,300 233 51.7 (41.0 to 62.5) 659 39 54.5 (33.4 to 75.6)

Dabigatran 300 2,919 54 66.6 (26.1 to 107.1) 464 13 21.5 (9.3 to 33.7)


Rivaroxaban 20+ 16,562 1,006 65.1 (61.0 to 69.1) 2,558 165 67.0 (56.7 to 77.3)

Apixaban 5 3,806 487 94.8 (79.5 to 110.0) 398 58 103.3 (53.6 to 153.1)

Apixaban 10 7,006 297 47.3 (41.7 to 52.9) 703 28 46.7 (27.3 to 66.0)

Page 50 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



eTable 9 All patients: Adjusted hazard ratios (95% CI) for outcomes associated with exposure to different study drug doses

compared with warfarin.

QResearch CPRD

Comparison with

warfarin


Major bleeding

Dabigatran <300 1.01 (0.82 to 1.24) 0.9 0.61 (0.35 to 1.06) 0.08

Dabigatran 300 0.75 (0.55 to 1.01) 0.06 0.99 (0.53 to 1.86) 1

Rivaroxaban <20 1.15 (0.97 to 1.36) 0.1 1.47 (1.04 to 2.09) 0.03

Rivaroxaban 20+ 0.99 (0.88 to 1.11) 0.9 1.00 (0.76 to 1.31) 1

Apixaban 5 0.75 (0.61 to 0.93) 0.01 0.58 (0.30 to 1.14) 0.1

Apixaban 10 0.51 (0.41 to 0.63) <0.001 0.85 (0.51 to 1.41) 0.5

Intracranial bleed

Dabigatran <300 0.50 (0.29 to 0.85) 0.01 0.18 (0.03 to 1.29) 0.09

Dabigatran 300 0.32 (0.12 to 0.85) 0.02 0.54 (0.08 to 3.91) 0.5


Rivaroxaban 20+ 0.64 (0.49 to 0.85) 0.002 0.83 (0.43 to 1.60) 0.6

Apixaban 5 0.54 (0.35 to 0.85) 0.008 n/a#

Apixaban 10 0.43 (0.27 to 0.69) <0.001 1.16 (0.41 to 3.28) 0.8

Haematuria

Dabigatran <300 0.92 (0.61 to 1.39) 0.7 0.88 (0.36 to 2.15) 0.8

Dabigatran 300 0.86 (0.53 to 1.39) 0.5 0.91 (0.29 to 2.87) 0.9



Apixaban 5 1.11 (0.75 to 1.64) 0.6 0.91 (0.28 to 2.93) 0.9

Apixaban 10 0.47 (0.31 to 0.72) <0.001 1.29 (0.55 to 3.06) 0.6

Haemoptysis

Dabigatran <300 0.95 (0.35 to 2.59) 0.9 1.20 (0.16 to 8.93) 0.9

Dabigatran 300 1.41 (0.61 to 3.25) 0.4 n/a#

Rivaroxaban <20 0.98 (0.42 to 2.28) 1 1.25 (0.16 to 9.62) 0.8


Apixaban 5 0.83 (0.29 to 2.34) 0.7 2.56 (0.32 to 20.76) 0.4

Apixaban 10 0.09 (0.01 to 0.69) 0.02 n/a#

All GI bleed

Dabigatran <300 1.42 (1.08 to 1.88) 0.01 0.63 (0.28 to 1.41) 0.3

Dabigatran 300 0.80 (0.50 to 1.29) 0.4 1.40 (0.62 to 3.17) 0.4


Rivaroxaban >=20 1.15 (0.97 to 1.35) 0.1 0.97 (0.66 to 1.41) 0.9

Apixaban <=5 0.68 (0.49 to 0.95) 0.02 0.59 (0.24 to 1.45) 0.2

Apixaban >=10 0.67 (0.50 to 0.91) 0.009 0.60 (0.26 to 1.38) 0.2

Upper GI bleed

Dabigatran <300 1.50 (1.13 to 1.99) 0.005 0.70 (0.31 to 1.57) 0.4

Dabigatran 300 0.79 (0.48 to 1.31) 0.4 1.29 (0.53 to 3.16) 0.6



Apixaban <=5 0.67 (0.48 to 0.95) 0.02 0.63 (0.26 to 1.57) 0.3

Apixaban >=10 0.69 (0.51 to 0.93) 0.02 0.55 (0.22 to 1.36) 0.2

Lower GI bleed

Dabigatran <300 0.87 (0.27 to 2.76) 0.8 n/a# 0

Dabigatran 300 0.84 (0.21 to 3.44) 0.8 2.26 (0.30 to 17.33) 0.4

Rivaroxaban <20 0.92 (0.33 to 2.57) 0.9 n/a#


Apixaban <=5 1.09 (0.38 to 3.17) 0.9 n/a#

Apixaban >=10 0.52 (0.16 to 1.73) 0.3 1.04 (0.13 to 8.63) 1

Page 51 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



Ischaemic stroke

Dabigatran <300 1.02 (0.78 to 1.33) 0.9 0.91 (0.43 to 1.93) 0.8

Dabigatran 300 1.26 (0.88 to 1.83) 0.2 0.80 (0.25 to 2.51) 0.7



Apixaban <=5 1.31 (1.05 to 1.63) 0.02 0.88 (0.35 to 2.20) 0.8

Apixaban >=10 1.24 (0.99 to 1.55) 0.06 1.07 (0.49 to 2.34) 0.9

VTE

Dabigatran <300 0.32 (0.08 to 1.28) 0.1 0.37 (0.13 to 1.01) 0.05

Dabigatran 300 0.62 (0.20 to 1.93) 0.4 0.65 (0.21 to 2.03) 0.5


Rivaroxaban >=20 1.01 (0.80 to 1.29) 0.9 2.54 (1.90 to 3.40) <0.001

Apixaban <=5 0.46 (0.20 to 1.06) 0.07 0.65 (0.24 to 1.78) 0.4

Apixaban >=10 0.81 (0.49 to 1.34) 0.4 0.40 (0.15 to 1.11) 0.08

Mortality

Dabigatran <300 1.11 (0.97 to 1.27) 0.1 0.98 (0.72 to 1.34) 0.9

Dabigatran 300 0.77 (0.59 to 1.01) 0.06 1.11 (0.64 to 1.93) 0.7

Rivaroxaban <20 1.34 (1.21 to 1.48) <0.001 1.74 (1.42 to 2.15) <0.001

Rivaroxaban >=20 1.32 (1.22 to 1.42) <0.001 1.34 (1.13 to 1.60) 0.001

Apixaban <=5 1.24 (1.12 to 1.38) <0.001 1.68 (1.27 to 2.21) <0.001

Apixaban >=10 0.97 (0.86 to 1.10) 0.6 1.02 (0.71 to 1.48) 0.9




corticosteroids, NSAIDs, statin, hormones), study year #) no events available

Page 52 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



eTable 10 Cohort characteristics and exposure by database, percentage (number) of patients with and without atrial

fibrillation

QResearch CPRD

AF group Non-AF group AF group Non-AF group

Total N 81251 74,754 22,019 18,037


for days of treatment 312 (130; 621) 160 (81; 275) 264 (126; 538) 180 (97; 292)

Gender

Men 55.0 (44697) 52.2 (39043) 55.7 (12260) 52.2 (9409)

Women 45.0 (36554) 47.8 (35711) 44.3 (9759) 47.8 (8628)

Age at baseline

20-29 0.1 (43) 1.4 (1075) 0.1 (12) 2.1 (386)

30-39 0.3 (258) 4.4 (3252) 0.3 (70) 4.6 (837)

40-49 1.8 (1425) 8.8 (6579) 1.7 (375) 9.4 (1693)

50-59 5.9 (4794) 13.0 (9685) 6.0 (1319) 13.6 (2447)

60-69 19.3 (15686) 20.8 (15556) 19.4 (4264) 20.7 (3730)

70-79 35.1 (28551) 26.2 (19554) 35.3 (7776) 25.9 (4668)

80-89 31.4 (25537) 21.0 (15676) 31.8 (7003) 19.7 (3554)

90 and over 6.1 (4957) 4.5 (3377) 5.4 (1200) 4.0 (721)

Mean age at baseline 75.1 (10.6) 67.8 (15.6) 75.0 (10.5) 66.8 (16.0)

Ethnicity recorded 83.4 (67788) 82.2 (61437) 96.0 (21140) 96.6 (17429)

White (UK or European) 97.2 (78949) 95.1 (71080) 98.0 (21583) 96.3 (17365)

Indian 0.7 (529) 0.9 (690) 0.5 (113) 0.7 (132)

Pakistani 0.3 (256) 0.6 (415) 0.1 (27) 0.3 (56)

Bangladeshi 0.2 (135) 0.2 (165) 0.0 (6) 0.1 (13)

Other Asian 0.3 (223) 0.4 (277) 0.3 (62) 0.4 (81)

Caribbean 0.5 (435) 1.1 (812) 0.3 (63) 0.5 (99)

Black African 0.2 (181) 0.7 (488) 0.1 (33) 0.6 (106)

Other 0.6 (456) 1.0 (760) 0.6 (132) 1.0 (185)

BMI recorded 96.6 (78486) 94.6 (70720) 96.7 (21289) 94.8 (17108)

Mean BMI in kg/m2 (SD) 28.6 (5.7) 28.8 (5.8) 28.4 (5.8) 28.7 (6.1)

SBP recorded 100.0 (81235) 99.5 (74380) 99.8 (21984) 98.9 (17845)


mmHg (SD) 131.2 (17.4) 131.1 (17.7) 132.2 (17.1) 131.7 (16.9)

Smoking

Recorded 99.9 (81151) 99.6 (74471) 99.9 (22003) 99.9 (18011)

Non-smoker 51.3 (41718) 50.4 (37690) 51.3 (11296) 49.6 (8949)

Ex-smoker 40.9 (33196) 35.6 (26594) 41.8 (9213) 36.8 (6645)

Current smoker 7.7 (6237) 13.6 (10187) 6.8 (1494) 13.4 (2417)

Current light smoker 4.3 (3525) 7.0 (5237) 4.2 (932) 7.4 (1336)

Current moderate smoker 2.0 (1664) 4.0 (2958) 1.6 (363) 3.7 (674)

Current heavy smoker 1.3 (1048) 2.7 (1992) 0.9 (199) 2.3 (407)

Not recorded 0.1 (100) 0.4 (283) 0.1 (16) 0.1 (26)

Alcohol consumption

Recorded 94.6 (76824) 91.6 (68511) 94.6 (20835) 91.7 (16535)

Non-drinker 34.1 (27737) 35.1 (26250) 32.4 (7131) 33.5 (6044)

Trivial (< 1 unit per day) 27.9 (22637) 27.7 (20715) 32.4 (7129) 31.7 (5725)

Light (1-2 units per day) 13.8 (11178) 12.5 (9345) 17.6 (3877) 15.5 (2796)

Moderate (3-6 units/day) 16.6 (13492) 14.1 (10522) 6.4 (1420) 6.0 (1080)

Heavy (7-9 units/day) 1.4 (1162) 1.3 (947) 3.6 (782) 2.9 (516)

Very heavy (>9 units/day) 0.8 (618) 1.0 (732) 2.3 (496) 2.1 (374)

Not recorded 5.4 (4427) 8.4 (6243) 5.4 (1184) 8.3 (1502)

Page 53 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



Townsend quintiles

1 (Affluent) 21.5 (17487) 19.6 (14623) 26.1 (5747) 23.6 (4257)

2 21.7 (17610) 20.2 (15108) 26.0 (5720) 23.9 (4312)

3 20.6 (16747) 20.5 (15301) 21.5 (4731) 21.4 (3860)

4 18.7 (15222) 19.9 (14871) 17.0 (3733) 19.2 (3468)

5 (Deprived) 17.5 (14185) 19.9 (14851) 9.4 (2080) 11.8 (2129)

Exposure

Warfarin 65.7 (47368) 64.2 (53884) 75.2 (15034) 79.5 (15945)

Dabigatran 6.3 (4534) 2.2 (1868) 5 (1003) 1.7 (339)

Rivaroxaban 18.9 (13597) 22 (18423) 14.8 (2950) 14.4 (2893)

Apixaban 12.8 (9199) 8.5 (7132) 7 (1402) 2.4 (490)

Page 54 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



eTable 11 Comorbidities, previous events and other medications, percentage (number) of patients with and without atrial

fibrillation

QResearch CPRD

AF group Non-AF group AF group Non-AF group

Total N 81251 74,754 22,019 18,037


Alcohol dependence 2.6 (2102) 3.1 (2353) 2.4 (525) 2.5 (453)

Atrial fibrillation 100.0 (81251) 0.0 (0) 100.0 (22019) 0.0 (0)

Bleeding disorders 0.9 (762) 1.3 (949) 1.2 (266) 1.6 (292)

Cancer (any) 12.4 (10108) 13.3 (9964) 12.5 (2746) 13.7 (2476)


pancreatitis 1.2 (951) 1.4 (1045) 1.0 (231) 1.6 (287)

Chronic obstructive pulmonary

disease 9.8 (7971) 8.3 (6221) 9.5 (2083) 8.3 (1504)

Chronic renal disease 2.4 (1951) 2.5 (1888) 2.7 (585) 2.9 (522)

Congestive cardiac failure 13.3 (10823) 7.0 (5208) 13.0 (2871) 6.5 (1181)

Coronary heart disease 24.4 (19861) 17.3 (12917) 24.8 (5456) 17.5 (3153)

Diabetes 18.8 (15246) 15.6 (11666) 17.9 (3943) 14.1 (2552)

Dyspepsia 18.2 (14783) 17.3 (12947) 26.1 (5753) 24.6 (4435)

Falls or hip fracture (within 180

days) 7.7 (6217) 7.7 (5732) 6.0 (1332) 5.8 (1045)

Hip or knee operation (within

180 days) 0.7 (549) 4.6 (3454) 1.5 (334) 6.2 (1112)

Hypertension 61.4 (49867) 41.8 (31220) 62.2 (13698) 43.6 (7860)

Ischaemic stroke* 18.6 (15132) 12.9 (9613) 19.1 (4215) 12.3 (2215)

Oesophageal varices 0.1 (79) 0.2 (165) 0.1 (12) 0.2 (35)

Peptic ulcer 7.6 (6144) 6.4 (4781) 8.4 (1854) 6.9 (1239)

Valvular heart disease 11.3 (9147) 7.4 (5538) 9.2 (2025) 6.3 (1139)

Venous thromboembolism* 5.9 (4802) 48.3 (36094) 7.4 (1636) 58.1 (10471)

Previous bleed (any)* 24.7 (20058) 22.7 (16974) 28.4 (6256) 25.9 (4676)

Previous Intracranial bleed* 1.0 (778) 1.1 (849) 1.3 (281) 1.3 (240)

Previous Haematuria 11.4 (9241) 9.4 (7054) 12.0 (2652) 9.9 (1785)

Previous Haemoptysis* 2.6 (2145) 2.7 (2014) 3.5 (775) 3.5 (624)

Previous All GI bleed* 13.1 (10638) 12.4 (9304) 16.1 (3538) 15.0 (2702)

Previous upper GI bleed* 4.4 (3574) 4.2 (3173) 5.3 (1163) 5.0 (909)

Previous Rectal bleed* 9.8 (7937) 9.2 (6913) 12.3 (2709) 11.5 (2068)

Other medications

PPI 43.1 (35021) 42.2 (31576) 41.5 (9145) 40.7 (7338)

Antibiotics# 9.2 (7493) 9.8 (7357) 8.5 (1866) 9.9 (1792)

Antiplatelet 26.4 (21486) 19.4 (14487) 38.5 (8480) 21.9 (3959)

Antidepressants 15.7 (12758) 21.7 (16227) 15.0 (3299) 21.3 (3835)

Anticonvulsants 0.7 (587) 1.3 (1000) 0.8 (169) 1.3 (241)

NSAIDs 6.9 (5612) 11.9 (8883) 6.7 (1468) 12.2 (2202)

Corticosteroids 11.7 (9503) 12.3 (9183) 10.6 (2333) 12.4 (2242)

Statins 54.3 (44138) 39.9 (29805) 53.3 (11729) 37.4 (6740)

Hormones (women) 1.5 (563) 2.5 (881) 2.7 (265) 6.9 (596)

*) Based on GP and HES records #) In the last 6 months prior the drug start date

Page 55 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



eTable 12 Rates per 1000py of outcomes in QResearch and CPRD in patients with atrial fibrillation

QResearch CPRD

Person-

years


per 1000py (95%CI)

Person-

years


per 1000py (95%CI)

Major bleeding

Warfarin 72,487 25.1 (24.0 to 26.3) 19,331 26.7 (24.4 to 29.0)

Dabigatran 4,988 21.8 (17.7 to 26.0) 925 18.3 (9.5 to 27.1)

Rivaroxaban 12,515 26.5 (23.7 to 29.4) 2,005 33.4 (25.2 to 41.5)

Apixaban 7,471 15.4 (12.6 to 18.3) 852 26.1 (15.0 to 37.3)

Intracranial bleed

Warfarin 73,776 6.2 (5.6 to 6.7) 19,631 5.8 (4.7 to 6.8)

Dabigatran 5,082 3.0 (1.4 to 4.6) 933 1.0 (0.0 to 2.9)


Apixaban 7,508 2.6 (1.4 to 3.7) 859 4.5 (0.0 to 9.1)

Haematuria

Warfarin 73,105 8.0 (7.3 to 8.6) 19,497 8.1 (6.8 to 9.4)

Dabigatran 5,040 6.4 (4.2 to 8.6) 929 7.0 (1.7 to 12.3)


Apixaban 7,498 4.4 (2.9 to 5.9) 857 7.8 (1.9 to 13.7)

Haemoptysis

Warfarin 73,755 1.4 (1.2 to 1.7) 19,622 1.4 (0.9 to 1.9)

Dabigatran 5,067 1.4 (0.4 to 2.5) 933 1.2 (0.0 to 3.7)


Apixaban 7,511 0.5 (0.0 to 1.1) 860 1.1 (0.0 to 3.2)

All GI bleed

Warfarin 73,360 9.5 (8.8 to 10.2) 19,525 11.5 (10.0 to 13.0)

Dabigatran 5,047 11.2 (8.2 to 14.2) 930 9.0 (2.7 to 15.2)


Apixaban 7,489 8.2 (6.1 to 10.2) 856 12.7 (4.7 to 20.6)

Upper GI bleed

Warfarin 73,424 8.5 (7.8 to 9.1) 19,536 10.4 (9.0 to 11.9)

Dabigatran 5,047 11.0 (8.0 to 14.0) 930 7.7 (2.0 to 13.4)


Apixaban 7,491 7.6 (5.6 to 9.7) 857 11.3 (3.8 to 18.7)

Rectal bleed

Warfarin 73,769 1.1 (0.8 to 1.3) 19,634 1.1 (0.7 to 1.6)

Dabigatran 5,082 0.3 (0.0 to 0.8) 933 1.3 (0.0 to 3.8)


Apixaban 7,509 0.6 (0.1 to 1.2) 858 1.4 (0.0 to 4.0)

Ischaemic stroke

Warfarin 59,343 13.5 (12.6 to 14.5) 15,797 14.3 (12.4 to 16.2)

Dabigatran 3,744 15.9 (11.8 to 20.1) 671 11.0 (2.6 to 19.4)


Apixaban 5,573 15.2 (11.9 to 18.5) 594 14.7 (4.9 to 24.5)

VTE

Warfarin 69,569 3.1 (2.7 to 3.5) 18,188 3.7 (2.8 to 4.6)

Dabigatran 4,921 1.2 (0.2 to 2.2) 884 1.3 (0.0 to 3.7)


Apixaban 7,230 2.5 (1.3 to 3.6) 805 6.2 (0.5 to 11.8)

Mortality

Warfarin 73,839 44.6 (43.0 to 46.1) 19,584 41.8 (38.9 to 44.8)

Dabigatran 5,083 43.1 (37.3 to 49.0) 930 40.3 (27.3 to 53.3)


Apixaban 7,511 53.5 (48.4 to 58.5) 855 61.0 (45.1 to 76.9)

Page 56 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



eTable 13 Rates per 1000py of outcomes by database in patients without atrial fibrillation

QResearch CPRD

Person-

years


per 1000py (95%CI)

Person-years Age-sex adjusted rate

per 1000py (95%CI)

Major bleeding

Warfarin 39335 29.2 (27.5 to 30.9) 11096 34.5 (31.1 to 38.0)

Dabigatran 1129 31.0 (18.8 to 43.1) 192 27.2 (4.1 to 50.4)

Rivaroxaban 8066 29.4 (25.6 to 33.1) 1240 32.1 (22.1 to 42.1)

Apixaban 3273 18.3 (13.6 to 23.1) 247 5.3 (0.0 to 11.7)

Intracranial bleed

Warfarin 39929 6.3 (5.5 to 7.1) 11260 7.2 (5.7 to 8.8)

Dabigatran 1137 2.9 (0.0 to 5.8) 193 3.3 (0.0 to 9.9)


Apixaban 3297 5.2 (2.7 to 7.7) 248 -

Haematuria

Warfarin 39685 8.9 (8.0 to 9.8) 11206 9.7 (7.9 to 11.6)

Dabigatran 1133 7.9 (2.0 to 13.8) 193 7.9 (0.0 to 23.4)


Apixaban 3291 4.3 (2.4 to 6.1) 247 2.7 (0.0 to 6.6)

Haemoptysis

Warfarin 39912 1.6 (1.2 to 2.0) 11260 2.1 (1.3 to 3.0)

Dabigatran 1137 2.3 (0.0 to 5.5) 193 -


Apixaban 3300 0.3 (0.0 to 0.8) 248 -

All GI bleed

Warfarin 39684 12.4 (11.3 to 13.5) 11191 15.4 (13.1 to 17.8)

Dabigatran 1133 17.7 (8.1 to 27.4) 192 15.9 (0.1 to 31.8)


Apixaban 3286 8.8 (5.2 to 12.3) 248 2.6 (0.0 to 7.6)

Upper GI bleed

Warfarin 39719 11.1 (10.0 to 12.1) 11203 13.7 (11.5 to 15.9)

Dabigatran 1134 15.6 (6.3 to 24.9) 192 15.9 (0.1 to 31.8)


Apixaban 3288 8.0 (4.6 to 11.4) 248 2.6 (0.0 to 7.6)

Rectal bleed

Warfarin 39917 1.6 (1.2 to 2.0) 11259 1.9 (1.1 to 2.7)

Dabigatran 1136 2.1 (0.0 to 4.6) 193 -

Rivaroxaban 8155 1.1 (0.4 to 1.9) 1255 -

Apixaban 3298 0.7 (0.0 to 1.8) 248 -

Ischaemic stroke

Warfarin 34121 11.2 (10.1 to 12.4) 9719 11.4 (9.3 to 13.6)

Dabigatran 755 20.8 (10.7 to 30.9) 123 20.5 (0.0 to 43.7)


Apixaban 2311 15.4 (10.6 to 20.3) 139 15.3 (0.0 to 32.6)

VTE

Warfarin 18496 41.0 (38.1 to 44.0) 4668 38.8 (33.1 to 44.4)

Dabigatran 1055 9.7 (3.5 to 15.9) 173 33.4 (5.0 to 61.8)


Apixaban 2748 44.0 (33.4 to 54.7) 210 10.4 (0.0 to 22.3)

Mortality

Warfarin 39960 58.4 (56.0 to 60.8) 11224 57.7 (53.2 to 62.2)

Dabigatran 1137 67.4 (41.7 to 93.0) 192 57.7 (26.0 to 89.5)


Apixaban 3301 72.8 (63.9 to 81.7) 246 93.6 (35.0 to 152.2)

Page 57 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



eTable 14 Patients with atrial fibrillation: Adjusted hazard ratios (95% CI) for outcomes associated with exposure to study

drugs compared with warfarin, by database

QResearch CPRD

Comparison with

warfarin


Major bleeding

Dabigatran 0.90 (0.74 to 1.10) 0.3 0.62 (0.36 to 1.07) 0.09


Apixaban 0.62 (0.50 to 0.75) <0.001 0.88 (0.54 to 1.45) 0.6

Intracranial bleed

Dabigatran 0.45 (0.26 to 0.77) 0.004 0.19 (0.03 to 1.39) 0.1


Apixaban 0.40 (0.25 to 0.64) <0.001 0.73 (0.21 to 2.49) 0.6

Haematuria

Dabigatran 0.94 (0.66 to 1.35) 0.7 0.92 (0.40 to 2.15) 0.9


Apixaban 0.68 (0.46 to 1.00) 0.05 1.19 (0.49 to 2.94) 0.7

All GI bleed

Dabigatran 1.16 (0.87 to 1.53) 0.3 0.64 (0.28 to 1.46) 0.3


Apixaban 0.76 (0.57 to 1.02) 0.07 0.91 (0.46 to 1.83) 0.8

Upper GI bleed

Dabigatran 1.25 (0.94 to 1.67) 0.1 0.59 (0.24 to 1.46) 0.3


Apixaban 0.77 (0.57 to 1.04) 0.09 0.88 (0.42 to 1.82) 0.7

Ischaemic stroke

Dabigatran 1.18 (0.90 to 1.55) 0.2 0.74 (0.35 to 1.59) 0.4


Apixaban 1.15 (0.90 to 1.49) 0.3 0.94 (0.46 to 1.92) 0.9

VTE#

Dabigatran 0.41 (0.18 to 0.93) 0.03 n/a#

Rivaroxaban 1.34 (0.94 to 1.91) 0.1 n/a#

Apixaban 0.82 (0.48 to 1.39) 0.5 n/a#

Mortality

Dabigatran 0.98 (0.85 to 1.13) 0.8 0.87 (0.62 to 1.24) 0.4

Rivaroxaban 1.19 (1.09 to 1.30) <0.001 1.27 (1.03 to 1.58) 0.03

Apixaban 1.10 (0.98 to 1.23) 0.1 1.19 (0.88 to 1.61) 0.3

*) adjusted for age, sex, ethnicity, smoking, alcohol, Townsend quintile, BMI, SPB, falls & hip fracture, hip/knee ops,



corticosteroids, NSAIDs, statin, hormones), study year #) insufficient data for CPRD analysis

Page 58 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



eTable 15 Patients without atrial fibrillation: Adjusted hazard ratios (95% CI) for outcomes associated with exposure to

study drugs compared with warfarin, by database

QResearch CPRD

Comparison with

warfarin


Major bleeding

Dabigatran 0.97 (0.68 to 1.37) 0.8 1.03 (0.52 to 2.01) 0.9


Apixaban 0.60 (0.46 to 0.79) <0.001 0.48 (0.21 to 1.11) 0.09

Intracranial bleed

Dabigatran 0.44 (0.16 to 1.19) 0.1 0.49 (0.07 to 3.61) 0.5


Apixaban 0.63 (0.37 to 1.08) 0.09 0.30 (0.04 to 2.24) 0.2

Haematuria

Dabigatran 0.78 (0.38 to 1.59) 0.5 0.80 (0.19 to 3.32) 0.8


Apixaban 0.73 (0.45 to 1.20) 0.2 1.12 (0.33 to 3.82) 0.9

All GI bleed

Dabigatran 1.36 (0.85 to 2.18) 0.2 1.41 (0.61 to 3.25) 0.4


Apixaban 0.55 (0.37 to 0.83) 0.004 0.14 (0.02 to 1.04) 0.05

Upper GI bleed

Dabigatran 1.25 (0.75 to 2.08) 0.4 1.55 (0.67 to 3.58) 0.3

Rivaroxaban 1.03 (0.81 to 1.30) 0.8 1.00 (0.62 to 1.62) 1

Apixaban 0.55 (0.36 to 0.83) 0.005 0.16 (0.02 to 1.16) 0.07

Ischaemic stroke

Dabigatran 1.76 (1.09 to 2.83) 0.02 1.77 (0.55 to 5.76) 0.3


Apixaban 1.16 (0.80 to 1.68) 0.4 1.13 (0.33 to 3.86) 0.8

VTE#

Dabigatran 0.15 (0.08 to 0.29) <0.001 n/a#

Rivaroxaban 1.43 (1.25 to 1.63) <0.001 n/a#

Apixaban 0.42 (0.33 to 0.53) <0.001 n/a#

Mortality

Dabigatran 1.10 (0.87 to 1.39) 0.4 1.34 (0.85 to 2.11) 0.2


Apixaban 1.15 (1.00 to 1.33) 0.05 1.65 (1.15 to 2.37) 0.006

*) adjusted for age, sex, ethnicity, smoking, alcohol, Townsend quintile, BMI, SPB, falls & hipfracture, hip/knee ops,



corticosteroids, NSAIDs, statin, hormones), study year #) insufficient data for CPRD analysis

Page 59 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



eTable 16 Patients with and without atrial fibrillation: Adjusted hazard ratios (95% CI) for outcomes associated with

exposure to study drug different doses compared with warfarin.

QResearch CPRD

Comparison with

warfarin

Adjusted HR

(95%CI)

P-value Adjusted HR (95%CI) P-value

AF patients

Major bleeding

Dabigatran <300 0.99 (0.78 to 1.25) 0.9 0.63 (0.33 to 1.19) 0.2

Dabigatran 300 0.77 (0.55 to 1.08) 0.1 0.61 (0.23 to 1.65) 0.3



Apixaban 5 0.72 (0.54 to 0.95) 0.02 0.59 (0.26 to 1.36) 0.2

Apixaban 10 0.54 (0.42 to 0.71) <0.001 1.13 (0.63 to 2.03) 0.7

Mortality

Dabigatran <300 1.03 (0.88 to 1.21) 0.7 0.80 (0.54 to 1.20) 0.3

Dabigatran 300 0.82 (0.60 to 1.13) 0.2 1.16 (0.60 to 2.26) 0.7



Apixaban 5 1.20 (1.05 to 1.37) 0.008 1.35 (0.93 to 1.95) 0.1

Apixaban 10 0.97 (0.82 to 1.14) 0.7 0.98 (0.62 to 1.57) 0.9

Non AF patients

Major bleeding

Dabigatran <300 1.12 (0.74 to 1.70) 0.6 0.52 (0.16 to 1.62) 0.3

Dabigatran 300 0.71 (0.37 to 1.37) 0.3 2.04 (0.90 to 4.63) 0.09

Rivaroxaban <20 0.99 (0.73 to 1.34) 1 1.49 (0.91 to 2.45) 0.1


Apixaban 5 0.82 (0.58 to 1.15) 0.3 0.52 (0.16 to 1.66) 0.3

Apixaban 10 0.45 (0.31 to 0.65) <0.001 0.45 (0.14 to 1.41) 0.2

Mortality

Dabigatran <300 1.24 (0.95 to 1.60) 0.1 1.43 (0.87 to 2.38) 0.2

Dabigatran 300 0.75 (0.44 to 1.28) 0.3 1.08 (0.40 to 2.89) 0.9

Rivaroxaban <20 1.55 (1.33 to 1.81) <0.001 1.90 (1.40 to 2.58) <0.001

Rivaroxaban 20+ 1.44 (1.29 to 1.61) <0.001 1.57 (1.24 to 1.99) <0.001

Apixaban 5 1.30 (1.09 to 1.54) 0.003 2.21 (1.46 to 3.34) <0.001

Apixaban 10 1.00 (0.82 to 1.21) 1 1.01 (0.55 to 1.88) 1





Page 60 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



eTable 17 Sensitivity analysis, censoring patients at the time of hospitalisation related to other outcome. Adjusted hazard

ratios (95% CI) for outcomes associated with exposure to study drugs compared with warfarin.

QResearch CPRD Both databases

Comparison

with warfarin

Adjusted HR

(95%CI)

P-

value

Adjusted HR

(95%CI)

P-

value

Combined analysis

HR (95%CI)

P-value

Major bleeding



Apixaban 0.60 (0.51 to 0.71) <0.001 0.73 (0.48 to 1.11) 0.1 0.62 (0.53 to 0.72) <0.001

Intracranial bleed



Apixaban 0.46 (0.32 to 0.66) <0.001 0.56 (0.20 to 1.57) 0.3 0.47 (0.34 to 0.66) <0.001

Haematuria



Apixaban 0.67 (0.49 to 0.91) 0.01 1.14 (0.55 to 2.35) 0.7 0.73 (0.55 to 0.97) 0.03

Haemoptysis



Apixaban 0.33 (0.13 to 0.85) 0.02 0.86 (0.11 to 6.86) 0.9 0.86 (0.11 to 6.86) 0.9

All GI bleed



Apixaban 0.67 (0.53 to 0.85) 0.001 0.59 (0.32 to 1.12) 0.1 0.66 (0.53 to 0.83) <0.001

Upper GI bleed



Apixaban 0.67 (0.52 to 0.85) 0.001 0.59 (0.30 to 1.14) 0.1 0.66 (0.52 to 0.83) <0.001

Rectal bleed

Dabigatran 0.98 (0.39 to 2.42) 1 0.69 (0.09 to 5.16) 0.7 0.69 (0.09 to 5.16) 0.7


Apixaban 0.87 (0.37 to 2.04) 0.7 0.61 (0.07 to 5.02) 0.6 0.61 (0.07 to 5.02) 0.6

Ischaemic stroke



Apixaban 1.16 (0.94 to 1.43) 0.2 0.98 (0.53 to 1.82) 1 1.14 (0.94 to 1.39) 0.2

VTE


Rivaroxaban 1.50 (1.33 to 1.70) <0.001 1.84 (1.40 to 2.43) <0.001 1.55 (1.39 to 1.74) <0.001

Apixaban 0.46 (0.37 to 0.57) <0.001 0.48 (0.23 to 1.00) 0.05 0.46 (0.37 to 0.57) <0.001

Mortality


Rivaroxaban 1.34 (1.25 to 1.43) <0.001 1.44 (1.24 to 1.68) <0.001 1.36 (1.27 to 1.44) <0.001

Apixaban 1.13 (1.03 to 1.24) 0.009 1.35 (1.06 to 1.71) 0.02 1.16 (1.06 to 1.26) <0.001





Page 61 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



eTable 18 Complete case analysis and propensity score adjustment by database. Adjusted hazard ratios (95% CI) for

outcomes associated with exposure to study drugs compared with warfarin.

QResearch CPRD

Comparison with

warfarin

Complete case

analysis*

Propensity score

adjustment

Complete case

analysis*

Propensity score

adjustment

Major bleeding

Dabigatran 0.93 (0.78 to 1.11) 0.87 (0.73 to 1.04) 0.79 (0.51 to 1.21) 0.72 (0.46 to 1.11)

Rivaroxaban 1.03 (0.92 to 1.14) 0.97 (0.88 to 1.06) 1.14 (0.90 to 1.45) 1.01 (0.81 to 1.26)

Apixaban 0.60 (0.51 to 0.71) 0.60 (0.51 to 0.70) 0.73 (0.47 to 1.13) 0.71 (0.47 to 1.08)

Intracranial bleed




Haematuria




Haemoptysis




All GI bleed




Upper GI bleed




Rectal bleed




Ischaemic stroke




VTE



Apixaban 0.44 (0.35 to 0.56) 0. 59 (0.47 to 0.73) 0.43 (0.19 to 0.99) 0.38 (0.16 to 0.89)

Mortality








Page 62 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


STROBE 2007 Statement—Checklist of items that should be included in reports of cohort studies

Section/Topic Item

# Recommendation

Reported on

page #

Title and abstract 1 (a) Indicate the study’s design with a commonly used term in the title or the abstract 1,2

(b) Provide in the abstract an informative and balanced summary of what was done and what was

found

2-3

Introduction

Background/rationale 2 Explain the scientific background and rationale for the investigation being reported 5-6

Objectives 3 State specific objectives, including any prespecified hypotheses 6

Methods

Study design 4 Present key elements of study design early in the paper 7

Setting 5 Describe the setting, locations, and relevant dates, including periods of recruitment, exposure,

follow-up, and data collection

7-8

Participants 6 (a) Give the eligibility criteria, and the sources and methods of selection of participants. Describe

methods of follow-up

7-8

Variables 7 Clearly define all outcomes, exposures, predictors, potential confounders, and effect modifiers. Give

diagnostic criteria, if applicable

8-10

Data sources/

measurement

8* For each variable of interest, give sources of data and details of methods of assessment

(measurement). Describe comparability of assessment methods if there is more than one group

7-9

Bias 9 Describe any efforts to address potential sources of bias 11-12

Study size 10 Explain how the study size was arrived at In protocol

Quantitative

variables

11 Explain how quantitative variables were handled in the analyses. If applicable, describe which

groupings were chosen and why

8-9

Statistical methods 12 (a) Describe all statistical methods, including those used to control for confounding 10-12

(b) Describe any methods used to examine subgroups and interactions 9, 11

(c) Explain how missing data were addressed 10

(d) Cohort study—If applicable, explain how loss to follow-up was addressed 7

(e) Describe any sensitivity analyses 11-12

Page 63 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Results

Participants 13* (a) Report numbers of individuals at each stage of study—eg numbers potentially eligible,

examined for eligibility, confirmed eligible, included in the study, completing follow-up, and

analysed

Fig1

(b) Give reasons for non-participation at each stage 13

(c) Consider use of a flow diagram Fig1

Descriptive data 14* (a) Give characteristics of study participants (eg demographic, clinical, social) and information on

exposures and potential confounders

24-27

(b) Indicate number of participants with missing data for each variable of interest 24-27

(c) Cohort study—Summarise follow-up time (eg average and total amount) 28

Outcome data 15* Cohort study—Report numbers of outcome events or summary measures over time 28

Main results 16 (a) Report the numbers of individuals at each stage of the study—eg numbers potentially eligible,

examined for eligibility, confirmed eligible, included in the study, completing follow-up, and

analysed

Fig1

(b) Give reasons for non-participation at each stage Fig1

(c) Consider use of a flow diagram Fig1

Other analyses 17 Report other analyses done—eg analyses of subgroups and interactions, and sensitivity analyses 15-16

Discussion

Key results 18 Summarise key results with reference to study objectives 17

Limitations 19 Discuss limitations of the study, taking into account sources of potential bias or imprecision.

Discuss both direction and magnitude of any potential bias

17-19

Interpretation 20 Give a cautious overall interpretation of results considering objectives, limitations, multiplicity of

analyses, results from similar studies, and other relevant evidence

19-20

Generalisability 21 Discuss the generalisability (external validity) of the study results 17

Other information

Funding 22 Give the source of funding and the role of the funders for the present study and, if applicable, for

the original study on which the present article is based

31-32

Page 64 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


http://eprints.nottingham.ac.uk/35133/

1

Abstract—Anticoagulants are used to treat thrombosis and

to prevent thrombosis and stroke among those at high risk.

Older anticoagulants such as warfarin can be effective but

require regular monitoring and may lead to major bleeding

and even death. Newer anticoagulants may not need blood tests but they are quite expensive. There is a lack of information on how safe these new treatments are in the longer term since the original trials were done over relatively short periods of time. Also the trials were done in selected patients who may be different from patients in real world settings. So there is a need to establish how safe the new anticoagulants are compared with the older ones when used in real world settings over longer periods of time. Our aim is use a large primary care research database (QResearch) to investigate the unintended effects of novel anticoagulants (NOACS) in primary care populations compared with warfarin. Our specific objective is to evaluate bleeding, stroke and thrombosis outcomes associated with NOACs (both individual drugs & class effect) compared with warfarin in NHS patients

Index Terms—warfarin, novel anticoagulants, incident

user design, stroke prevention, atrial fibrillation, drug

safety, primary care, primary care databases

1 INTRODUCTION

Anticoagulants are used in the prevention & treatment of

venous thromboembolism1. They are also used to reduce risk of

ischaemic stroke2, especially among patients with atrial

fibrillation3. The use of anticoagulants is likely to increase in

future especially since international guidelines encourage more

systematic identification of patients at high risk of venous

thromboembolism or stroke who might benefit from

anticoagulation4-6. For example, in 2010, NICE issued new

guidance to improve prevention of venous thromboembolism

for patients using cost-effective interventions1. In 2014, NICE

issued guidance on the management of atrial fibrillation which

included assessment of the risks of stroke as well as the risks &

benefits associated with anticoagulation7.Although the risk of

Submission Date 13 July 2016. Julia Hippisley-Cox is Professor of Clinical

Epidemiology & General Practice at the University of Nottingham and Medical

Director of ClinRisk Ltd (email: [email protected]). Carol Coupland is Professor of Medical Statistics in Primary Care at the University

stroke in patients with atrial fibrillation can be reduced by

anticoagulation2, many patients with atrial fibrillation are not

currently prescribed anticoagulation even though it is

incentivized8 9. This may reflect concerns about monitoring of

warfarin since patients may be poorly maintained within the

therapeutic range & suffer adverse consequences. It could also

reflect uncertainties regarding potential balance of risks &

benefits for an individual patient including the potential adverse

haemorrhagic effects of traditional anticoagulants such as

warfarin.

1.1 Evidence from clinical trials

Novel oral anticoagulants (NOACs) which include dabigatran,

rivaroxaban and apixaban are a recent alternative to warfarin

and are now recommended in international guidelines or

position statements4-6. They have the stated advantage of not

requiring regular international normalisation ratio (INR) blood

test monitoring10. Clinical trials have established the non-

inferiority of these agents compared with warfarin in controlled

trial settings11-13 but have left residual concerns regarding

safety, particularly in ‘real world’ settings. This is important

since such bleeds can result in hospitalisations and may be fatal.

Although research is underway, there is currently no clinically

tested antidote to stop bleeding with the NOACs comparable to

that available for warfarin14. The Randomised Evaluation of

Long Term Anticoagulant therapy (RE-LY study) examined

dabigatran compared with warfarin over a median period of 2

years. Overall rates of major bleeding were similar in both

groups although more patients discontinued dabigatran due to a

serious adverse event (2.7% for dabigatran vs 1.7% for

warfarin)11. The trial reported an increased rate of major

bleeding with the higher dose of dabigatran (150mg) compared

with the 110mg dose11. In a second study, the rivaroxaban

verses warfarin in non-valvular atrial fibrillation study

(ROCKET-AF), there were lower rates of fatal bleeding in the

rivaroxaban arm compared with warfarin12.

Data on safety & efficacy of the novel anticoagulants is still

accumulating & the US Federal Drug Administration (FDA)

has highlighted15 the need to further investigate potential safety

issues with the new anticoagulants once more data is

of Nottingham and consultant statistician for ClinRisk Ltd (email:

[email protected]).

Julia Hippisley-Cox, Professor. Carol Coupland, Professor.

Protocol to assess the effectiveness and safety of novel

oral anticoagulants (NOAC) vs warfarin in real world

settings: cohort study in primary care

Page 65 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



2

available15. This was because of an unusually high number of

serious and fatal bleeding events reported to the FDA Adverse

Event Reporting System (FAERS) associated with the use of

dabigatran shortly after its approval in October 201016. A recent

meta-analysis of three randomised controlled trials17 reported

that the new anticoagulants are more effective at reducing all

cause stroke & systemic embolism (RR 0.78, 95% CI 0.67 to

0.92) in people with atrial fibrillation compared to warfarin.

However, data regarding risks of major bleeding (RR 0.88, 95%

CI 0.71 to 1.09) was inconclusive with new oral anticoagulants

associated with a lower risk of intracranial bleeding (RR 0.49,

95% CI 0.36 to 0.66) and a suggestion of an increased risk of

gastro-intestinal bleeding (RR 1.25, 95% CI 0.91 to 1.72)

compared with warfarin17. This increased risk of gastro-

intestinal bleeding was subsequently confirmed in second meta-

analysis by Holster et al14 which reported a statistically

significant increased risk for NOACs with an odds ratio of 1.45

(95% CI 1.07 to 1.97). The odds ratio for dabigatran was 1.58

(1.29 to 1.93) and that for rivaroxaban was 1.48 (1.21 to 1.82).

Another meta-analysis by Ruff et al reported an increased risk

of gastro-intestinal bleeding (RR 1.25, 1.01-1.55) compared

with warfarin13 although there was significant heterogeneity

between the individual trials.

Clinical trials, & their associated meta-analyses, provide

invaluable information on the effectiveness & efficacy of drugs.

However, they are limited in providing information on adverse

events since such data are not always recorded or reported in a

consistent fashion. Trials tend to be of relatively short duration

(e.g. median of 2 years), under-powered for the detection of

adverse events & are susceptible to selection bias with

participants tending to be predominantly white, younger &

healthier than the general population. Most trials & meta-

analyses are designed to investigate, & report on, effectiveness

of drugs compared with placebo18. Very few are specifically

designed to investigate adverse events. Also comparisons

between real world populations & those in clinical trials are

important since trial populations can be enriched or depleted

with patients with particular characteristics which then tend to

make the effect size seem artificially high or low19.

1.2 Warfarin vs dabigatran vs rivaroxaban

Several observational studies comparing the safety of

dabigatran and rivaroxaban with warfarin have recently been

reported in the BMJ20-22. Abraham et al20 studied 8,578 users of

dabigatran, 16,253 users of rivaroxaban and 67,985 users of

warfarin in commercially insured US adults. Outcomes were

limited only to gastrointestinal bleeding. Overall, dabigatran

and rivaroxaban had similar risks to warfarin in younger

patients but higher risks in older patients. The authors list a

range of limitations including a lack of data on apixaban or

different doses of dabigatran (US does not allow 110mg

dabigatran which is widely used in other countries including

Europe).

Chang et al undertook a new user cohort study of commercially

insured US adults21, in which outcomes were limited to

gastrointestinal bleeding with 18 months of follow up, ending

in March 2012. Whilst Chang et al found no significant

increased risk of gastrointestinal bleeding compared with

warfarin for dabigatran (adjusted hazard ratio 1.21, 95% CI

0.96 to 1.53) or rivaroxaban (0.98, 0.36 to 2.69), the confidence

intervals were wide due to relatively small numbers (122 events

among 4,907 users of dabigatran and 4 events among 1649

users of rivaroxaban). Elderly patients were under-represented;

follow up was relatively short, there was no information on

different doses of dabigatran; mortality data was unavailable

and the generalisability to European populations is unknown.

Chang et al concluded that they could not rule out as much as a

50% increase in GI bleeding with dabigatran compared with

warfarin (which would be similar to that found in the trials) or

more than a twofold increased risk for rivaroxaban compared

with warfarin.

Larsen et all studied the effectiveness and safety of NOAC

(dabigatran, rivaroxaban and apixaban) compared with warfarin

in an anticoagulant naive population of patients with atrial

fibrillation in Denmark22. The study was conducted between

2011 and 2015 and included 35,436 warfarin users, 12,701

dabigatran 150mg users, 7,192 rivaroxaban 20mg users and

6,349 apixaban 5mg users and had an average follow up of 1.9

years. There was no significant difference in risk of ischaemic

stroke between NOACS and warfarin; rivaroxaban was

associated with lower risks of ischaemic strokes or systemic

embolism than warfarin but with comparable major bleeding

rates. Dabigatran and apixaban had no difference in risk of

ischaemic stroke or systemic embolism compared with warfarin

but major bleeding rates were significantly lower with reference

to warfarin.

1.3 Why quantifying risk is important?

Quantifying absolute risk of bleeding for different types of

anticoagulation treatment using a real world population is

important because it can

provide more complete & valid safety information to

help inform policy & national guidelines since the

information on adverse events from clinical trials

alone is often incomplete or not reported23 24 or may

not generalise to ‘real world settings’. The relative

safety of NOACs versus warfarin has yet to be

established.

help clinicians in identifying patients for whom the

benefits of anticoagulation are likely to outweigh the

risk of a bleed.

allow clinicians to monitor patients at higher risk more

carefully to help reduce their risk, for example, by (a)

avoiding drugs which might interact with NOACs or

warfarin to increase bleeding risk, or (b) preferentially

using drugs which might lower bleeding risk such as

proton pump inhibitors (should there be evidence to

support it).

help patients make a more informed choice before

initiating anticoagulation. Whilst physicians tend to

emphasise the benefits of treatment using information

Page 66 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



3

from trials patients need information on benefits &

adverse effects in ‘real life’ to aid making decisions

about treatments.

2 METHODS

2.1 Study design & data source

We will conduct an inception cohort study in a large primary

care population using the latest version of the QResearch

database (http://www.qresearch.org). QResearch is a large

validated primary care electronic database containing the health

records of 22 million patients registered from approximately

1200 general practices using the EMIS system25. Practices &

patients in the database are nationally representative26. All

practices in the latest version of the QResearch database are

linked at individual patient level to both the hospital admissions

data & mortality records27 28. We will include all QResearch

practices in England once they have been using their current

EMIS system for at least a year. This database and similar ones

(CPRD) have been used to undertake direct comparisons of the

populations and results of statistical analyses29-32. Therefore

methods to ensure comparable populations, outcomes,

confounders and exposures have already been developed29-32.

2.2 Overall cohort selection

We will initially identify an open cohort of patients aged 21-99

years at the study entry date, drawn from patients registered

with eligible practices between 01/09/2008 & 01/10/2015.

Follow up will be until 31/01/2016. We will exclude patients

without a valid postcode related Townsend deprivation score.

We will then restrict the cohort for our main analyses to patients

who received an anticoagulant during the study period. We will

identify new users of warfarin and new users of novel

anticoagulants (NOACS) so that our analysis is a direct

comparison of new users (NOACs) vs warfarin (new user

cohort). We will exclude existing or recent users of

anticoagulants by excluding patients prescribed anticoagulants

in the 365 days before their study entry date.

We will use an incident (new) user design to help minimize the

impact of indication bias and the other biases which can arise

from non-randomised studies such as this if prevalent users are

included33. We will determine an initial entry date to the cohort

for each patient, which will be the latest of the following dates:

date of registration with the practice plus one year, date on

which the practice computer system was installed plus one year,

& the beginning of the study period.

We will assign an index entry date which will be the date of

their first prescription of the anticoagulant during the study

period. Patients will be censored if they stop anticoagulant

treatment (censored 30 days after the expected end date of their

last anticoagulant prescription to allow for a wash-out period),

if they have a gap of more than 30 days between the end of one

prescription and the start of the next (censored 30 days from the

start of the gap) or if they switch to another anticoagulant

treatment (censored the day before the switch to a different

treatment).

Patients will be censored at the earliest of these dates if

applicable or on the earliest date of the first major bleed

(primary outcome analysis) or secondary outcome (ischaemic

stroke or VTE), death, deregistration with the practice, last

upload of computerised data, or the study end date if this occurs

before censoring due to stopping or switching anticoagulant

treatment. We will censor patients if they switch NOAC

treatment since the reasons for switching might relate to

indications for a bleed (such as gastric problems) which may

not be fully recorded in the database, and could lead to

indication bias if exposed periods after a switch were included

in the analysis and attributed to the new treatment.

2.3 Outcomes

Our primary outcome will be a composite outcome of any major

bleed leading to admission or death based on hospital or

mortality linked records occurring after study entry. Major

bleeds will include incident upper or lower gastrointestinal

bleed, intracranial bleed, haematuria or haemoptysis. Patients

who have had a prior major bleed will be included in the

analysis and will be separately identified according to the type

of bleed they have had.

Where there are sufficient numbers of events, we will analyse

outcomes for upper and lower GI bleed, haematuria,

haemoptysis and intracranial bleed separately to allow for

comparison with other studies. We will include these events

since they are potentially life threatening or life changing &

may be preventable. As in a previous study34, we will use

ICD10 codes to identify events which were recorded either on

the linked hospital record or the mortality record.

Whilst our main outcomes are unintended effects of

anticoagulants, we will also identify incident ischaemic stroke

and venous thromboembolism using linked GP, hospital and

mortality electronic records to measure the effect on intended

outcomes (i.e. reduction of ischaemic stroke or VTE). This will

allow comparison of the effect size in this study with previous

trials. Patients will be classified as having an ischaemic stroke

if this is recorded on either the GP or hospital or mortality

records and the date of the event will be the earliest date

recorded on any of the three data sources. Similarly, patients

will be classified as having a venous thromboembolism if this

is recorded on either the GP or hospital or mortality records and

the date of the event will be the earliest date recorded on any of

the three data sources. We will also carry out an analysis of all-

cause mortality.

2.4 Drug Exposure

The primary exposure of interest is the use of anticoagulants

including warfarin compared with use of any of the currently

Page 67 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

http://www.qresearch.org/



4

available NOACs. NOACs will include those used during the

study period i.e. dabigatran, rivaroxaban and apixaban.

Edoxaban is the newest NOAC but will not be included since it

was only licensed for use in the UK at the end of 2015 and will

not have been prescribed during the study period.

We will identify new users of anticoagulants during the study

period by identifying those with at least one prescription issued

after the initial study entry date. We will categorise the

anticoagulant exposure variable as either novel anticoagulant or

warfarin. We will compare the individual NOAC drugs with

warfarin where the numbers exposed are sufficient. We will

categorise duration of use (for example, 1-7 days; 8-28 days;

29-90 days; >90 days). We will include patients who have only

had one prescription in the analysis since this could be

associated with one of the outcomes under investigation. We

will assess the numbers of patients prescribed anticoagulants

which are rarely used such as phenindione to determine whether

these patients should be excluded in the analysis or included as

a separate exposure.

We will include an analysis of different doses of NOACs where

sufficient information is available based on information in the

table below .We will not undertake an analysis of warfarin dose

since precise information on the dosage of warfarin which a

patient takes will vary according to INR measurements and is

not consistently recorded on the GP record.

2.5 Confounder variables

We will adjust for the following demographic variables which

may be associated with an increased risk of stroke, VTE or of

bleeding34 or an indication for prescribing a particular

anticoagulant(i.e. patients at higher risk of bleeding may be

preferentially prescribed NOACs rather than warfarin. The

selection of confounders will depend on which outcome is

being analysed but is likely to include variables from the

following list: age at study entry35; year of study entry; sex; self-

assigned ethnicity; deprivation35 36 smoking status, alcohol

use37, BMI, systolic blood pressure, falls or hip fracture, alcohol

dependence, atrial fibrillation\flutter; treated hypertension37;

chronic renal disease37; COPD; chronic liver disease or

pancreatitis37; diabetes; coronary heart disease; previous

stroke/TIA, congestive cardiac failure; venous

thromboembolism; cancer (most commonly occurring 12 types

of cancer); peptic ulcer; dyspepsia\heartburn, esophageal

varices; previous bleed (e.g. intracranial, upper gastrointestinal,

haematuria, or haemoptysis). We will also adjust for recent and

concurrent medications as these may increase or decrease

bleeding risk or interact with anticoagulants including proton

pump inhibitors, macrolide antibiotics, antacids, antiplatelets37

antidepressants38, anticonvulsants (phenytoin or

carbamazepine), NSAIDS, corticosteroids and statins. We will

also adjust for drugs which may increase VTE risk such as HRT

and the oral contraceptive. The covariates will be assessed at

the date when the anticoagulant is first prescribed, as only the

first treatment period is included in the main analyses, with

patients censored if they stop or switch treatment. We will

undertake a supplementary analysis which allows patients to

switch between different anticoagulants. The covariates will be

reassessed at the start of each treatment period and we will flag

patients as having switched from another treatment so that this

can be incorporated in the analysis.

2.6 Statistical Analysis

2.6.1 Descriptive statistics

We will describe the baseline characteristics of patients starting

each NOAC and for warfarin. As in other studies22, we will use

logistic regression to calculate the odds ratios for each treatment

compared with any of the alternatives to determine which

characteristics are associated with different treatment choices.

We will compare characteristics of patients in the new user

cohort with those excluded from the main analyses due to prior

use of warfarin.

We will determine the person years of exposure to each drug

and the number of events occurring in each group so that we

can calculate crude and age standardised incidence for the

outcomes (with 95% confidence intervals) for each drug.

2.6.2 Analysis of bleeding, stroke and VTE outcomes

We will use a Cox regression analysis to derive adjusted hazard

ratios with 95% confidence intervals for the primary outcome

(major bleed) and secondary outcomes (other bleeds, ischaemic

stroke, VTE, all-cause mortality) comparing novel

anticoagulant exposure with warfarin as the primary reference

group. We will adjust for potential confounders as listed above,

evaluating them at the date of the first prescription for an

anticoagulant during the study interval. We will use fractional

polynomials to model non-linear risk relationships with

continuous variables such as age, body mass index and systolic

blood pressure39.

In the supplementary analysis where switching between drugs

is incorporated in the analysis we will use a time varying Cox

regression analysis and re-evaluate confounders each time a

switch occurs. Patients in this analysis will be censored on the

earliest date of the first major bleed (primary outcome analysis)

or secondary outcome (minor bleeds, ischaemic stroke or VTE),

death, deregistration with the practice, last upload of

computerised data, or the study end date.

We will use multiple imputation to replace missing values for

body mass index, systolic blood pressure, smoking status &

alcohol status & use these values in our main analyses40-43. We

will carry out 5 imputations44 and combine results using

Rubin’s rules. We will examine interactions between different

anticoagulants and age and sex. We will also examine

interactions between anticoagulants and other prescribed

medications and the co-morbidities listed as confounders.

Page 68 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



5

2.6.3 Power calculation:

Based on previous analyses34, with 80% power and 1%

significance then a total of 5548 outcome events (major bleeds)

are needed to detect a hazard ratio of 1.5 for NOACs compared

with warfarin (5466 events in warfarin group and 82 in NOAC

group), assuming the ratio of exposed time in the NOAC group

compared with the warfarin group is 0.01 (QBleed paper34). For

an exposed time ratio of 0.03 (to account for increased use of

NOACs over time – e.g. Prescription Cost Analysis value for

2013) then a total of 1950 outcome events are needed (1866

events in warfarin group and 84 in NOAC group), For a hazard

ratio of 1.3, with an exposure ratio of 0.03 then 5092 outcome

events are needed (4901 events in warfarin group and 191 in

NOAC group). We anticipate that there will be at least this

many outcome events since the QResearch database includes

over 1200 practices covering a population of 22 million. We

will use all available data on QResearch to maximise the power

& generalisability of the study. We will consider a p value of

<= 0.01 to be statistically significant.

3 SUMMARY OF PATIENT ENGAGEMENT PLANS

We will use the PPI infrastructure in Nottingham which is

currently being established. We will hold meetings to get advice

on how best to engage with the public regarding this type of

drug equity & safety research – to ensure key questions can be

answered & presented in a way which is useful &

understandable to the general public. We will also engage with

the patient representatives from the QResearch advisory board

which includes the National Patient Participation Association.

We will use existing websites & social media to raise

awareness.

4 OTHER INFORMATION

4.1.1 Acknowledgements

We acknowledge the contribution of EMIS practices who

contribute to the QResearch® and EMIS for expertise in

establishing, developing and supporting the database. We also

acknowledge the contribution of the Health and Social Care

Information for supplying the hospital episodes data and the

Office of National Statistics for supplying the mortality data

and Public Health England for supplying the cancer registration

data.

4.1.2 Approvals:

The project has been reviewed in accordance with the

QResearch® agreement with NRES Committee East Midlands -

Derby [reference 03/4/021].

4.1.3 Competing Interests

JHC is professor of clinical epidemiology at the University of

Nottingham and co-director of QResearch® – a not-for-profit

organisation which is a joint partnership between the University

of Nottingham and Egton Medical Information Systems

(leading commercial supplier of IT for 60% of general practices

in the UK). JHC is also a paid director of ClinRisk Ltd which

produces open and closed source software to ensure the reliable

and updatable implementation of clinical risk equations within

clinical computer systems to help improve patient care. CC is

Professor of Medical Statistics at the University of Nottingham

and a paid consultant statistician for ClinRisk Ltd. This work

and any views expressed within it are solely those of the co-

authors and not of any affiliated bodies or organisations.

Page 69 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



6

5 REFERENCES

1. National Institute for Clinical Excellence. Venous

thromboembolism: reducing the risk Reducing the

risk of venous thromboembolism (deep vein

thrombosis and pulmonary embolism) in patients

admitted to hospital. NICE guideline 92. 1 ed.

London: National Institute for Clinical Excellence,

2010:50.

2. Hart RG, Pearce LA, Aguilar MI. Meta-analysis:

antithrombotic therapy to prevent stroke in patients

who have nonvalvular atrial fibrillation. Ann Intern

Med 2007;146(12):857-67.

3. Mant J, Hobbs FD, Fletcher K, et al. Warfarin versus

aspirin for stroke prevention in an elderly community

population with atrial fibrillation (the Birmingham

Atrial Fibrillation Treatment of the Aged Study,

BAFTA): a randomised controlled trial. Lancet

2007;370(9586):493-503.

4. You JJ, Singer DE, Howard PA, et al. Antithrombotic

therapy for atrial fibrillation: Antithrombotic Therapy

and Prevention of Thrombosis, 9th ed: American

College of Chest Physicians Evidence-Based Clinical

Practice Guidelines. Chest 2012;141(2 Suppl):e531S-

75S.

5. Skanes AC, Healey JS, Cairns JA, et al. Focused 2012

update of the Canadian Cardiovascular Society atrial

fibrillation guidelines: recommendations for stroke

prevention and rate/rhythm control. The Canadian

journal of cardiology 2012;28(2):125-36.

6. Camm AJ, Lip GY, De Caterina R, et al. 2012 focused

update of the ESC Guidelines for the management of

atrial fibrillation: an update of the 2010 ESC

Guidelines for the management of atrial fibrillation--

developed with the special contribution of the

European Heart Rhythm Association. Europace :

European pacing, arrhythmias, and cardiac

electrophysiology : journal of the working groups on

cardiac pacing, arrhythmias, and cardiac cellular

electrophysiology of the European Society of

Cardiology 2012;14(10):1385-413.

7. National Clinical Guideline Centre. Atrial fibrillation: the

management of atrial fibrillation. Clinical guideline,

methods, evidence and recommendations. In:

Excellence NIfC, ed. London, 2014:418.

8. Centre NI. The Quality and Outcomes Framework.

http://www.ic.nhs.uk/services/qof Accessed 25th

February 2015, 2015.

9. Scowcroft AC, Lee S, Mant J. Thromboprophylaxis of

elderly patients with AF in the UK: an analysis using

the General Practice Research Database (GPRD)

2000-2009. Heart 2013;99(2):127-32.

10. Lip GY. Atrial fibrillation in 2011: Stroke prevention in

AF. Nat Rev Cardiol 2012;9(2):71-3.

11. Connolly SJ, Ezekowitz MD, Yusuf S, et al. Dabigatran

versus Warfarin in Patients with Atrial Fibrillation.

New England Journal of Medicine

2009;361(12):1139-51.

12. Patel MR, Mahaffey KW, Garg J, et al. Rivaroxaban

versus warfarin in nonvalvular atrial fibrillation. N

Engl J Med 2011;365(10):883-91.

13. Ruff CT, Giugliano RP, Braunwald E, et al. Comparison

of the efficacy and safety of new oral anticoagulants

with warfarin in patients with atrial fibrillation: a

meta-analysis of randomised trials. Lancet

2014;383(9921):955-62.

14. Holster IL, Valkhoff VE, Kuipers EJ, et al. New oral

anticoagulants increase risk for gastrointestinal

bleeding: a systematic review and meta-analysis.

Gastroenterology 2013;145(1):105-12 e15.

15. Federal Drug Administration. A protocol for assessment of

dabigatran and selected safety outcomes 2013

[Available from: http://mini-

sentinel.org/assessments/medical_events/details.aspx

?ID=219.

16. Southworth MR, Reichman ME, Unger EF. Dabigatran

and postmarketing reports of bleeding. N Engl J Med

2013;368(14):1272-4.

17. Miller CS, Grandi SM, Shimony A, et al. Meta-analysis of

efficacy and safety of new oral anticoagulants

(dabigatran, rivaroxaban, apixaban) versus warfarin

in patients with atrial fibrillation. Am J Cardiol

2012;110(3):453-60.

18. Brugts JJ, Yetgin T, Hoeks SE, et al. The benefits of

statins in people without established cardiovascular

disease but with cardiovascular risk factors: meta-

analysis of randomised controlled trials. BMJ

2009;338(jun30_1):b2376-.

19. van Staa T-P, Leufkens HG, Zhang B, et al. A Comparison

of Cost Effectiveness Using Data from Randomized

Trials or Actual Clinical Practice: Selective Cox-2

Inhibitors as an Example. PLoS Med

2009;6(12):e1000194.

20. Abraham NS, Singh S, Alexander GC, et al. Comparative

risk of gastrointestinal bleeding with dabigatran,

rivaroxaban, and warfarin: population based cohort

study. 2015;350.

21. Chang H-Y, Zhou M, Tang W, et al. Risk of

gastrointestinal bleeding associated with oral

anticoagulants: population based retrospective cohort

study. BMJ 2015;350.

22. Larsen TB, Skjøth F, Nielsen PB, et al. Comparative

effectiveness and safety of non-vitamin K antagonist

oral anticoagulants and warfarin in patients with

atrial fibrillation: propensity weighted nationwide

cohort study. BMJ 2016;353.

23. Hughes S, Cohen D, Jaggi R. Differences in reporting

serious adverse events in industry sponsored clinical

trial registries and journal articles on antidepressant

and antipsychotic drugs: a cross-sectional study. BMJ

Open 2014;4(7).

24. Horn H, Nink K, McGauran N, et al. Early benefit

assessment of new drugs in Germany - Results from

2011 to 2012. Health Policy 2014;116(2-3):147-53.

25. Hippisley-Cox J, Coupland C, Vinogradova Y, et al.

Derivation and validation of QRISK, a new

Page 70 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

http://www.ic.nhs.uk/services/qof

http://mini-sentinel.org/assessments/medical_events/details.aspx?ID=219





7

cardiovascular disease risk score for the United

Kingdom: prospective open cohort study. BMJ

2007:bmj.39261.471806.55.

26. Hippisley-Cox J, Vinogradova Y, Coupland C, et al.

Comparison of key practice characteristics between

general practices in England and Wales and general

practices in the QRESEARCH data. Report to the

Health and Social Care Information Centre.:

University of Nottingham, 2005.

27. Hippisley-Cox J, Coupland C. Predicting risk of

emergency admission to hospital using primary care

data: derivation and validation of QAdmissions score.

BMJ Open 2013;3(8):e003482.

28. Hippisley-Cox J. Validity and completeness of the NHS

Number in primary and secondary care electronic

data in England 1991-2013. 2013; 1. Hippisley-Cox

J. Validity and completeness of the NHS number in

primary and secondary care: electronic data in

England 1991-2013

http://eprints.nottingham.ac.uk/3153/1/Validity%26C

ompletenessNHSNumber.pdf (accessed June 2013).

29. Hippisley-Cox J, Coupland C, Brindle P. The performance

of seven QPrediction risk scores in an independent

external sample of patients from general practice: a

validation study. BMJ Open 2014;4(8):e005809.

30. Vinogradova Y, Coupland C, Hippisley-Cox J. Exposure

to bisphosphonates and risk of gastrointestinal

cancers: series of nested case-control studies with

QResearch and CPRD data. BMJ 2013;346:f114.

31. Vinogradova Y, Coupland C, Hippisley-Cox J. Exposure

to combined oral contraceptives and risk of venous

thromboembolism: a protocol for nested case-control

studies using the QResearch and the CPRD

databases. BMJ Open 2014;4(4):e004499.

32. Vinogradava Y, Coupland C, Hippisley-Cox J. Combined

oral contraceptives and risk of venous

thromboembolism: nested case–control studies using

the QResearch and the CPRD databases. BMJ 2015.

33. Johnson ES, Bartman BA, Briesacher BA, et al. The

incident user design in comparative effectiveness

research. Pharmacoepidemiol Drug Saf 2013;22(1):1-

6.

34. Hippisley-Cox J, Coupland C. Predicting risk of upper

gastrointestinal bleed and intracranial bleed with

anticoagulants: cohort study to derive and validate

the QBleed scores. BMJ 2014;349:g4606.

35. Button L, Roberts S, Evans P. Hospitalized incidence and

case fatality for upper gastrointestinal bleeding from

1999 to 2007: a record linkage study. Aliment

Pharmacol Ther 2011;33(1):64 - 76.

36. Crooks C, West J, Card T. Upper gastrointestinal

haemorrhage and deprivation: a nationwide cohort

study of health inequality in hospital admissions. Gut

2011;61:514 - 20.

37. Pisters R, Lane DA, Nieuwlaat R, et al. A novel user-

friendly score (HAS-BLED) to assess 1-year risk of

major bleeding in patients with atrial fibrillation: the

Euro Heart Survey. Chest 2010;138(5):1093-100.

38. Tata LJ, Fortun PJ, Hubbard RB, et al. Does concurrent

prescription of selective serotonin reuptake inhibitors

and non-steroidal anti-inflammatory drugs

substantially increase the risk of upper

gastrointestinal bleeding? Aliment Pharmacol Ther

2005;22(3):175-81.

39. Royston P, Ambler G, Sauerbrei W. The use of fractional

polynomials to model continuous risk variables in

epidemiology. Int J Epidemiol 1999;28:964-74.

40. Schafer J, Graham J. Missing data: our view of the state of

the art. Psychological Methods 2002;7:147-77.

41. Group TAM. Academic Medicine: problems and solutions.

BMJ 1989;298:573-79.

42. Steyerberg EW, van Veen M. Imputation is beneficial for

handling missing data in predictive models. J

Epidemiol Community Health 2007;60:979.

43. Moons KGM, Donders RART, Stijnen T, et al. Using the

outcome for imputation of missing predictor values

was preferred. J Epidemiol Community Health

2006;59:1092.

44. White IR, Royston P, Wood AM. Multiple imputation

using chained equations: Issues and guidance for

practice. Stat Med 2011;30(4):377-99.

Page 71 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

http://eprints.nottingham.ac.uk/3153/1/Validity%26CompletenessNHSNumber.pdf

http://eprints.nottingham.ac.uk/3153/1/Validity%26CompletenessNHSNumber.pdf


08 August 2016 Version 1.0

ISAC APPLICATION FORM PROTOCOLS FOR RESEARCH USING THE CLINICAL PRACTICE RESEARCH DATALINK

(CPRD)

For ISAC use only

Protocol No. Submission date (DD/MM/YYYY)

........................... ...........................

IMPORTANT Please refer to the guidance for ‘Completing the ISAC application form’

found on the CPRD website (www.cprd.com/isac). If you have any queries, please contact the ISAC Secretariat at [email protected].

SECTION A: GENERAL INFORMATION ABOUT THE PROPOSED RESEARCH STUDY

1. Study Title§ (Please state the study title below)

Unintended effects of novel oral anticoagulants (NOAC) vs warfarin in real world settings §Please note: This information will be published on the CPRD’s website as part of its transparency policy. 2. Has any part of this research proposal or a related proposal been previously submitted to ISAC?

Yes * No

*If yes, please provide the previous protocol number/s below. Please also state in your current submission how this/these

are related or relevant to this study.

3. Has this protocol been peer reviewed by another Committee? (e.g. grant award or ethics committee) Yes

* No

*If Yes, please state the name of the reviewing Committee(s) below and provide an outline of the review process and

outcome as an Appendix to this protocol : School of Primary Care Research

4. Type of Study (please tick all the relevant boxes which apply)

Adverse Drug Reaction/Drug Safety Drug Effectiveness Drug Utilisation Pharmacoeconomics Disease Epidemiology Post-authorisation Safety Health care resource utilisation Methodological Research Health/Public Health Services Research Other

*

*If Other, please specify the type of study in the lay summary

5. Health Outcomes to be Measured§

§Please note:

This information will be published on CPRD’s website as part of its transparency policy.

Please summarise below the primary/secondary health outcomes to be measured in this research protocol:

• Incidence of major bleeding

• Ischaemic stroke

• Venous thromboembolism

• minor bleeding • All-cause mortality

•

• • • [Please add more bullet points as necessary]

Page 72 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



6. Publication: This study is intended for (please tick all the relevant boxes which apply):

Publication in peer-reviewed journals Presentation at scientific conference Presentation at company/institutional meetings Regulatory purposes Other

*

*If Other, please provide further information:

SECTION B: INFORMATION ON INVESTIGATORS AND COLLABORATORS

7. Chief Investigator§

Please state the full name, job title, organisation name & e-mail address for correspondence - see guidance notes for eligibility. Please note that there can only be one Chief Investigator per protocol.

Prof Julia Hippisley-Cox, Professor of Clinical Epidemiology & General Practice, University of Nottingham [email protected] §Please note:

The name and organisation of the Chief Investigator and will be published on CPRD’s website as part of its transparency

policy

CV has been previously submitted to ISAC CV number: 925_16 A new CV is being submitted with this protocol An updated CV is being submitted with this protocol

8. Affiliation of Chief Investigator (full address)

13th Floor Tower Building Department of Primary Care University Park Nottingham NG7 2RD

9. Corresponding Applicant§

Please state the full name, affiliation(s) and e-mail address below:

Yana Vinogradova, Research Statistician, University of Nottingham, [email protected] §Please note:

The name and organisation of the corresponding applicant and their organisation name will be published on CPRD’s

website as part of its transparency policy

Same as chief investigator CV has been previously submitted to ISAC CV number: 927_16 A new CV is being submitted with this protocol An updated CV is being submitted with this protocol

10. List of all investigators/collaborators§

Please list the full name, affiliation(s) and e-mail address* of all collaborators, other than the Chief Investigator below: §Please note:

The name of all investigators and their organisations/institutions will be published on CPRD’s website as part of its

transparency policy

Other investigator: Dr Carol Coupland, Professor of Medical Statistics, University of Nottingham [email protected] CV has been previously submitted to ISAC CV number: 926_16 A new CV is being submitted with this protocol An updated CV is being submitted with this protocol

Page 73 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



Other investigator: Yana Vinogradova, Research Statistician, University of Nottingham, [email protected] CV has been previously submitted to ISAC CV number: 927_16 A new CV is being submitted with this protocol An updated CV is being submitted with this protocol Other investigator: CV has been previously submitted to ISAC CV number: A new CV is being submitted with this protocol An updated CV is being submitted with this protocol Other investigator: CV has been previously submitted to ISAC CV number: A new CV is being submitted with this protocol An updated CV is being submitted with this protocol [Please add more investigators as necessary] *Please note that your ISAC application form and protocol must be copied to all e-mail addresses listed above at the time of submission of your application to the ISAC mailbox. Failure to do so will result in delays in the processing of your application.

11. Conflict of interest statement* Please provide a draft of the conflict (or competing) of interest (COI) statement that you intend to include in any publication which might result from this work

JHC is unpaid director of QResearch® and paid director of ClinRisk Limited, which produces open and closed source software to ensure the implementation of clinical risk algorithms within clinical computer systems. *Please refer to the International Committee of Medical Journal Editors (ICMJE) for guidance on what constitutes a COI.

12. Experience/expertise available Please complete the following questions to indicate the experience/ expertise available within the team of investigators/collaborators actively involved in the proposed research, including the analysis of data and interpretation of results.

Previous GPRD/CPRD Studies Publications using GPRD/CPRD data None 1-3 > 3

Experience/Expertise available Yes No

Is statistical expertise available within the research team? If yes, please indicate the name(s) of the relevant investigator(s)

Y Vinogradova, C Coupland

Is experience of handling large data sets (>1 million records) available within the research team? If yes, please indicate the name(s) of the relevant investigator(s)

J Hippisley-Cox, C Coupland, Y Vinogradova

Is experience of practising in UK primary care available to or within the research team? If yes, please indicate the name(s) of the relevant investigator(s)

J Hippisley-Cox

13. References relating to your study Please list up to 3 references (most relevant) relating to your proposed study: Ruff, C.T., et al., Comparison of the efficacy and safety of new oral anticoagulants with warfarin in patients with atrial

fibrillation: a meta-analysis of randomised trials. Lancet, 2014. 383(9921): p. 955-62. Miller, C.S., et al., Meta-analysis of efficacy and safety of new oral anticoagulants (dabigatran, rivaroxaban, apixaban)

versus warfarin in patients with atrial fibrillation. Am J Cardiol, 2012. 110(3): p. 453-60. Holster, I.L., et al., New oral anticoagulants increase risk for gastrointestinal bleeding: a systematic review and meta-

analysis. Gastroenterology, 2013. 145(1): p. 105-112 e15.

Page 74 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



SECTION C: ACCESS TO THE DATA

14. Financial Sponsor of study§

§Please note:

The name of the source of funding will be published on CPRD’s website as part of its transparency policy

Pharmaceutical Industry Please specify name and country: Academia Please specify name and country: University of Nottingham, UK Government / NHS Please specify name and country: Charity Please specify name and country: Other Please specify name and country: None

15. Type of Institution conducting the research

Pharmaceutical Industry Please specify name and country: Academia Please specify name and country: University of Nottingham, UK Government Department Please specify name and country: Research Service Provider Please specify name and country: NHS Please specify name and country: Other Please specify name and country:

16. Data access arrangements The financial sponsor/ collaborator* has a licence for CPRD GOLD and will extract the data The institution carrying out the analysis has a licence for CPRD GOLD and will extract the data** A data set will be provided by the CPRD

¥€

CPRD has been commissioned to extract the data and perform the analyses€

Other: If Other, please specify: *Collaborators supplying data for this study must be named on the protocol as co-applicants. **If data sources other than CPRD GOLD are required, these will be supplied by CPRD ¥Please note that datasets provided by CPRD are limited in size; applicants should contact CPRD ([email protected]) if a dataset of >300,000

patients is required. €Investigators must discuss their request with a member of the CPRD Research team before submitting an ISAC application. Please

contact the CPRD Research Team on +44 (20) 3080 6383 or email ([email protected]) to discuss your requirements. Please also state the name of CPRD Research team with whom you have discussed this request (provide the date of discussion and any relevant reference information):

Name of CPRD Researcher Reference number (where available) Date of contact

17. Primary care data Please specify which primary care data set(s) are required) Vision only (Default for CPRD studies Both Vision and EMIS

®*

EMIS® only*

Note: Vision and EMIS are different practice management systems. CPRD has traditionally collected data from Vision practice. Data collected from EMIS is currently under evaluation prior to wider release. *Investigators requiring the use of EMIS data must discuss the study with a member of the CPRD Research team before submitting an ISAC application

Please state the name of the CPRD Researcher with whom you have discussed your request for EMIS data: Name of CPRD Researcher Reference number (where available) Date of contact

SECTION D: INFORMATION ON DATA LINKAGES

Page 75 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



18. Does this protocol seek access to linked data

Yes* No If No, please move to section E.

*Research groups which have not previously accessed CPRD linked data resources must discuss access to these resources with a member of the CPRD Research team, before submitting an ISAC application. Investigators requiring access to HES Accident and Emergency data, HES Diagnostic Imaging Dataset and PROMS data must also discuss this with a member of the CPRD Research team before submitting an ISAC application. Please contact the CPRD Research Team on +44 (20) 3080 6383 or email [email protected] to discuss your requirements before submitting your application.

Please state the name of the CPRD Researcher with whom you have discussed your linkage request. Name of CPRD Researcher Reference number (where available) Date of contact Please note that as part of the ISAC review of linkages, your protocol may be shared - in confidence - with a representative of the requested linked data set(s) and summary details may be shared - in confidence - with the Confidentiality Advisory Group of the Health Research Authority.

19. Please select the source(s) of linked data being requested§ §Please note: This information will be published on the CPRD’s website as part of its transparency policy.

ONS Death Registration Data MINAP (Myocardial Ischaemia National Audit Project) HES Admitted Patient Care Cancer Registration Data* HES Outpatient PROMS (Patient Reported Outcomes Measure)** HES Accident and Emergency CPRD Mother Baby Link HES Diagnostic Imaging Dataset Practice Level Index of Multiple Deprivation (Standard) Practice Level Index of Multiple Deprivation (Bespoke) Patient Level Index of Multiple Deprivation*** Patient Level Townsend Score *** Other**** Please specify:

*Applicants seeking access to cancer registration data must complete a Cancer Dataset Agreement form (available from CPRD). This should be submitted to the ISAC as an appendix to your protocol. Please also note that applicants seeking access to cancer registry data must provide consent for publication of their study title and study institution on the UK Cancer Registry website. **Assessment of the quality of care delivered to NHS patients in England undergoing four procedures: hip replacement, knee replacement, groin hernia and varicose veins. Please note that patient level PROMS data are only accessible by academics *** ‘Patient level IMD and Townsend scores will not be supplied for the same study ****If “Other” is specified, please provide the name of the individual in the CPRD Research team with whom this linkage has been discussed. Name of CPRD Researcher Reference number (where available) Date of contact

20. Total number of linked datasets requested including CPRD GOLD

Number of linked datasets requested (practice/ ’patient’ level Index of Multiple Deprivation, Townsend Score or the CPRD

Mother Baby Link should not be included in this count) 3 Please note: Where ≥5 linked datasets are requested, approval may be required from the Confidentiality Advisory Group (CAG) to access these data

21. Is linkage to a local

¥ dataset with <1 million patients being requested?

Yes * No

*If yes, please provide further details: ¥

Data from defined geographical areas i.e. non-national datasets.

22. If you have requested one or more linked data sets, please indicate whether the Chief Investigator or any of the collaborators listed in question 5 above, have access to these data in a patient

Page 76 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



identifiable form (e.g. full date of birth, NHS number, patient post code), or associated with an identifiable patient index. Yes* No

* If yes, please provide further details:

23. Does this study involve linking to patient identifiable data (e.g. hold date of birth, NHS number, patient post code) from other sources? Yes No

SECTION E: VALIDATION/VERIFICATION

24. Does this protocol describe a purely observational study using CPRD data? Yes* No**

* Yes: If you will be using data obtained from the CPRD Group, this study does not require separate ethics approval from an NHS Research Ethics Committee. ** No: You may need to seek separate ethics approval from an NHS Research Ethics Committee for this study. The ISAC will provide advice on whether this may be needed.

25. Does this protocol involve requesting any additional information from GPs?

Yes* No * If yes, please indicate what will be required:

Completion of questionnaires by the GPψ Yes No Is the questionnaire a validated instrument? Yes No If yes, has permission been obtained to use the instrument? Yes No Please provide further information: Other (please describe) ψ Any questionnaire for completion by GPs or other health care professional must be approved by ISAC before circulation for completion.

26. Does this study require contact with patients in order for them to complete a questionnaire?

Yes* No *Please note that any questionnaire for completion by patients must be approved by ISAC before circulation for completion.

27. Does this study require contact with patients in order to collect a sample?

Yes* No * Please state what will be collected:

SECTION F: DECLARATION

28. Signature from the Chief Investigator

� I have read the guidance on ‘Completion of the ISAC application form’ and ‘Contents of CPRD ISAC Research Protocols’ and have understood these;

� I have read the submitted version of this research protocol, including all supporting documents, and confirm that these are accurate.

� I am suitably qualified and experienced to perform and/or supervise the research study proposed. � I agree to conduct or supervise the study described in accordance with the relevant, current protocol � I agree to abide by all ethical, legal and scientific guidelines that relate to access and use of CPRD data for research � I understand that the details provided in sections marked with (

§) in the application form and protocol will be published on

Page 77 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



the CPRD website in line with CPRD’s transparency policy. � I agree to inform the CPRD of the final outcome of the research study: publication, prolonged delay, completion or termination of the study.

Name: Julia Hippisley-Cox Date: 21 March 2017 e-Signature (type name)

Page 78 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960



PROTOCOL INFORMATION REQUIRED The following sections below must be included in the CPRD ISAC research protocol. Please refer to the guidance on ‘Contents of CPRD ISAC Research Protocols’ (www.cprd.com/isac) for more information on how to complete the sections below. Pages should be numbered. All abbreviations must be defined on first use.

Applicants must complete all sections listed below Sections which do not apply should be completed as ‘Not Applicable’

A. Study Title

§

§Please note:

This information will be published on CPRD’s website as part of its transparency policy

Unintended effects of novel oral anticoagulants (NOAC) vs warfarin in real world settings

B. Lay Summary (Max. 200 words)§

§Please note:


Anticoagulants (also called ‘blood thinning’ agents) are used both to treat blood clots and to help prevent further

blood clots and stroke in people with previous history of those , people with irregular heartbeats or after hip/knee

replacement operations. Warfarin is a medicine also used to treat blood clots and to prevent strokes, but it can

cause problems for some patients. It needs ongoing blood tests and can cause bleeding – such as vomiting or

coughing up blood – which may need hospital care or even be fatal.

Some new anticoagulants, which may not need blood tests, now exist, but they are quite expensive. It is also

unclear how safe these treatments will be in the longer term since the original trials were done over relatively short

periods of time (only up to 2 years) and because trials were done using selected patients, who may be different

from patients in real world settings. We need, therefore, to research the safety of the new anticoagulants when

compared with the older ones over longer periods of time and in normal clinical practice. Because of their novelty,

the drugs are not widely prescribed and the adverse effects are also rare. Their study will require a very large

number of patient records – available only from routinely-collected primary care data.

C. Technical Summary (Max. 200 words)§

§Please note:


Objective:

The study will compare in patients prescribed different types of novel anticoagulant drugs (NOAC) the risks of major bleeding requiring hospitalisation with those in patients prescribed warfarin.

Methods:

This will be a cohort study following patients who received anticoagulant prescription (NOAC or warfarin) between

1 September 2008 and 1 September 2016. We will use new-user design so patients with anticoagulant

prescriptions in the previous 12 months will be excluded. Outcomes will be hospitalisations and deaths identified

from HES and ONS mortality data. Primary outcomes will include gastrointestinal and intracranial bleed.

Secondary outcomes will include ischaemic stroke and venous thromboembolism (VTE).

Analysis:

Exposure to different NOAC (dabigatran, rivaroxaban and apixaban) will be compared with exposure to warfarin

using a Cox regression analysis. The results will be adjusted for potential confounders evaluated at the date of the

first prescription.

Page 79 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960




D. Objectives, Specific Aims and Rationale

i. Objective: to investigate the unintended effects of novel anticoagulants (NOACS) in primary care populations compared with those of warfarin

ii. Specific aims:

• to evaluate the following outcomes associated with NOACs compared with warfarin: i. all-cause mortality ii. major bleeding iii. ischaemic stroke or VTE

• to determine whether these risks vary by type, duration and dose iii. Rationale: to provide more complete and accurate safety information based on the UK general

population

E. Study Background

Clinical trials have established the non-inferiority of NOACs compared with warfarin in controlled trial settings[1-3]

but have left residual concerns regarding safety, particularly in ‘real world’ settings. A recent meta-analysis of

three randomised controlled trials[4] reported that the new anticoagulants are more effective at reducing all cause

stroke and systemic embolism (RR 0.78, 95% CI 0.67 to 0.92) in people with atrial fibrillation compared to warfarin.

Data regarding risks of major bleeding (RR 0.88, 95% CI 0.71 to 1.09) was, however, inconclusive, with new oral

anticoagulants associated with lower risk of intracranial bleeding (RR 0.49, 95% CI 0.36 to 0.66) but a suggestion

of increased risk of gastro-intestinal bleeding (RR 1.25, 95% CI 0.91 to 1.72) compared with warfarin[4]. This

increased risk of gastro-intestinal bleeding was subsequently confirmed in second meta-analysis by Holster et al[5]

which reported a statistically significant increased with of NOACs with a relative risk of 1.45 (95% CI 1.07 to 1.97).

The odds ratio for dabigatran was 1.58 (1.29 to 1.93) and that for rivaroxaban was 1.48 (1.21 to 1.82). Another

meta-analysis by Ruff et al[3] also reported an increased risk of gastro-intestinal bleeding (RR 1.25, 1.01-1.55)

compared with warfarin although there was significant heterogeneity between the individual trials.

Seven major population-based cohort studies have compared NOAC safeties with warfarin[6-12], each

investigating one or more outcomes. Five studies have investigated general bleeds[6-10], but have used several

different outcome definitions. One of these reported an increased risk (1.58, 95%CI 1.36 to 1.83)[6], while another

appeared to demonstrate that overall increased risk (2.29. 95%CI 1.77 to 2.98) applied only to the group

previously exposed to warfarin (3.30, 95%CI 2.40 to 4.53)[9]. The other three studies reported no difference in

major bleeding between dabigatran and warfarin[7, 8, 10].

Of the site-specific outcomes, gastro-intestinal has been more studied[6-8, 10-12], but the results reported for

dabigatran compared to warfarin have been similarly inconsistent, with three studies reporting increased risks[6-8]

and three studies reporting no difference[10-12]. Overall, the adjusted hazard ratios varied between 0.6 to 1.85.

Only two studies included rivaroxaban in their analysis, neither showing any difference in risk of gastro-intestinal

bleedings compared to warfarin[11, 12].

Four studies have investigated the difference in risk of intracranial haemorrhage for dabigatran compared to

warfarin, all reporting a decreased risk, with a range of hazard ratios between 0.24 to 0.51[6-8, 10]. Only one

study looked at bleeding in other anatomical sites, but none of those were associated with increased risks[6].

The clinical effectiveness of dabigatran compared to warfarin was investigated in four studies[7-10] including both

risks of thrombosis and ischaemic stroke. One study[9] showed an increase risk of venous thromboembolism

(adjusted hazard ratio 2.92, 95%CI 1.68 to 5.07) associated with dabigatran – most expressed in patients

previously exposed to warfarin (3.52, 95%CI 1.40 to 8.84) and with no difference for warfarin naïve patients – but

the others[7, 8, 10] reported no such difference. None of these studies reported any difference in risk of ischaemic

stroke between dabigatran and warfarin[7, 8, 10].

Three studies[7, 8, 10] also considered such outcomes as myocardial infarction and mortality in patients on

warfarin and NOACs, two finding decreased mortality risks[7, 10] associated with dabigatran compared with

Page 80 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960




warfarin. None, however, reported any difference in risk of myocardial infarction between dabigatran and

warfarin.[7, 8, 10].

Two limitations, which all the above seven studies have in common, were short follow-up time (between 4 and 18

months) and wide confidence intervals, probably due to relatively small numbers in the analyses. Most of these

observational studies were undertaken using data from US commercially insured populations, where older patients

were underrepresented and only one dosage (150mg) of dabigatran was analysed. The two Danish studies[9, 10]

did consider both dosages of dabigatran (110mg and 150mg), but the authors of one[9] noted significant

deviations from EMA-recommended prescribing for the higher dosage and suggested the need for additional real-

life follow-up studies.

Importantly, only one out of all these studies included patients who had switched from warfarin to dabigatran[9].

This is a situation that would not occur in clinical trials but is very likely to occur in ‘real world’ settings, so this also

raises safety questions clearly needing further research in a normal clinical context, particularly because patients

who had switched from warfarin to dabigatran were found to have had poorer outcomes. To date there have been

no such ‘real world’ observational studies in the UK.

F. Study Type

This will be a hypothesis-testing study using hazard ratios to investigate risks of a range of outcomes associated

with use of NOACs in comparison with warfarin.

G. Study Design

We will undertake an inception cohort study following patients on anti-coagulant treatment.

H. Feasibility counts

We have identified 60,620 patients from CPRD practices currently linked to HES and ONS data who started

anticoagulant treatment during the study period. Of those,11,669 were prescribed a NOAC (1,456 dabigatran,

7,463 rivaroxaban and 2,750 apixaban) and 48,951 were prescribed warfarin. For patients linked to HES up to

March 2014, we identified 35 primary outcome major bleeds in the NOAC group (2,192 patients) and 2,307 such

events in the warfarin group (41,369 patients). Considering the increased use of NOACs in the years following

(9,477 new users), we anticipate about 186 major bleeds in patients taking a NOAC during the study period.

I. Sample size considerations

Based on the available data and with an allocation ratio between NOAC and warfarin of 0.24, we will need 291 events in the warfarin group and 70 events in the NOAC group to detect a hazard ratio of 1.5 with 80% power and 1% significance level.

J. Data Linkage Required (if applicable):§

§Please note that the data linkage/s requested in research protocols will be published by the CPRD as part of its transparency policy

CPRD Gold will be used for identifying patients who started anticoagulants. Only patients with linkages will be

included. Inpatient Hospital Episode Statistics and ONS Mortality Data will be used for detecting knee/hip

replacement operations and the outcomes. Townsend deprivation score will be used as a confounder.

K. Study population

Page 81 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960




We will initially identify an open cohort of patients aged 21-99 years at the study entry date, drawn from patients registered with eligible Vision practices and linked to HES and ONS data during the study period – between 01/09/2008 and 01/09/2016. Start of follow-up will be defined as the latest of crd, uts, start of HES coverage, start of study period. End of follow-up will be defined as the earliest of tod, deathdate, lcd, end of HES coverage, end

of study period. We will restrict the cohort for our main analyses to patients who received an anticoagulant during the study period. We will then exclude patients prescribed anticoagulants in the 365 days before their study entry date to identify only new users of warfarin and novel anticoagulants (NOACS) so that our analysis directly compares new users of NOACs with new users of warfarin. We will use an incident (new) user design to help minimize the impact of indication and other biases which can arise if prevalent users are included in non-randomised studies[14]. We will determine an initial entry date to the cohort for each patient, which will be the latest of the following dates: date of registration with the practice plus one year; date on which the practice computer system was installed plus one year; the beginning of the study period. We will assign an index entry date which will be the date of the first anticoagulant prescription during the study period. Patients will be censored at the earliest of: stopping anticoagulant treatment (censored 30 days after the expected end date of their last anticoagulant prescription to allow for a wash-out period); a gap of more than 30 days between the end of one prescription and the start of the next (censored 30 days from the start of the gap); if they switch to another anticoagulant treatment (censored the day before the switch to a different treatment); date of the first major bleed (primary outcome analysis) or secondary outcome (ischaemic stroke or VTE); death; deregistration with the practice; last upload of computerised data; the study end date. We will censor patients if they switch NOAC treatment since the reasons for switching might relate to indications for a bleed (such as gastric problems), which may not be fully recorded in the database and which could lead to indication bias if exposed periods after a switch were included in the analysis and attributed to the new treatment.

L. Selection of comparison group(s) or controls

The comparison group will be patients prescribed warfarin.

M. Exposures, Health Outcomes§ and Covariates

§Please note:

Summary information on health outcomes (as included on the ISAC application form above )will be published on CPRD’s website

as part of its transparency policy

Outcomes Our primary outcome will be a composite outcome of any major bleed leading to admission or death based on hospital or mortality linked records occurring after study entry. Major bleeds will include incident upper or lower gastrointestinal bleed, intracranial bleed, haematuria or haemoptysis. Patients who have had a prior major bleed will be included in the analysis and will be separately identified by the type of bleed they had. We will look at primary and underlying causes of death and primary causes of hospital admissions. Where there are sufficient numbers of events, we will analyse separately outcomes for upper and lower GI bleed, haematuria, haemoptysis and intracranial bleed to allow for comparison with other studies. We will include these events since they are potentially life threatening or life changing and may be preventable. As in a previous study[13], we will use ICD10 codes to identify events which were recorded either on the linked hospital record or the mortality record. Whilst our main outcomes are unintended effects of anticoagulants, we will also identify incident ischaemic stroke and venous thromboembolism using linked GP, hospital and mortality electronic records to measure the effect on intended outcomes (reduction of ischaemic stroke or VTE). This will allow comparison of the effect size in this study with previous trials. Patients will be classified as having an ischaemic stroke if this is recorded on either the GP or hospital or mortality records, and the date of the event will be the earliest date recorded on any of the three data sources. Similarly, patients will be classified as having a venous thromboembolism if this is recorded on either the GP or hospital or mortality records, and the date of the event will be the earliest date recorded on any of

Page 82 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960




the three data sources. We will also carry out an analysis of all-cause mortality. Exposure The primary exposure of interest is the use of anticoagulants including warfarin compared with use of any of the currently available NOACs. NOACs will include those used during the study period i.e. dabigatran, rivaroxaban and apixaban. Edoxaban is the newest NOAC, but will not be included since it was licensed for use in the UK only at the end of 2015 so will not have been prescribed during the study period. We will identify new users of anticoagulants during the study period by identifying those with at least one prescription issued after the initial study entry date. We will categorise the anticoagulant exposure variable as either novel anticoagulant or warfarin. We will compare the individual NOAC drugs with warfarin where the numbers exposed are sufficient. We will categorise duration of use (for example, 1-7 days; 8-28 days; 29-90 days; >90 days). We will include patients who have only had one prescription in the analysis since this could be associated with one of the outcomes under investigation. We will assess the numbers of patients prescribed rarely-used anticoagulants, such as phenindione, to determine whether these patients should be excluded in the analysis or included as a separate exposure. Where sufficient information is available, we will include an analysis of different doses of NOACs based on information in the table below. We will not undertake an analysis of warfarin dose since precise information on the dosage of warfarin which a patient takes will vary according to INR measurements and is not consistently recorded on the GP record. Table 1 NOAC preparations and available doses.

Name of drug Date first used in CPRD

Tablets/ capsules available

Recommended dose

Dabigatran (pradaxa)

Sept 2008

75mg 110 mg 150 mg

110 mg BD (if >=80 years or at risk of bleeding or on verapamil or with moderate renal impairment) 150 mg BD – usual dose

Rivaroxaban (Xarelto)

Nov 2008 10 mg 15 mg 20 mg

10 mg OD 15 mg BD 20 mg OD

Apixaban (Eliquis)

Feb 2012 2.5 mg 5mg

2.5 mg BD 5mg BD 10 mg BD (loading Rx for DVT for 7 days then mg BD)

Covariates

We will adjust for demographic variables which may be associated with an increased risk of stroke, VTE or of

bleeding[13] or an indication for prescribing a particular anticoagulant (patients at higher risk of bleeding may be

preferentially prescribed NOACs rather than warfarin). The selection of confounders will depend on which

outcome is being analysed but is likely to include variables from the following list: age at study entry[15]; year of

study entry; sex; self-assigned ethnicity; deprivation[15, 16]; smoking status; alcohol use[17]; BMI; systolic blood

pressure; falls or hip fracture; hip/knee replacement operation; alcohol dependence; atrial fibrillation\flutter;

treated hypertension[17]; chronic renal disease[17]; COPD; chronic liver disease or pancreatitis[17]; diabetes;

coronary heart disease; previous stroke/TIA; congestive cardiac failure; venous thromboembolism; cancer (the

12 most commonly occurring types); peptic ulcer; dyspepsia\heartburn; esophageal varices; previous bleed

(intracranial, upper gastrointestinal, haematuria, or haemoptysis). We will adjust for recent and concurrent

medications, which may increase or decrease bleeding risk or interact with anticoagulants, including: proton

pump inhibitors; macrolide antibiotics; antacids; antiplatelets[17]; antidepressants[18]; anticonvulsants (phenytoin

or carbamazepine); NSAIDS; corticosteroids; statins. We will also adjust for drugs which may increase VTE risk,

such as hormone replacement therapy and the oral contraceptive. As only the first treatment period is included in

the main analyses, the covariates will be assessed at the date when the anticoagulant is first prescribed, with

Page 83 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960




patients censored if they stop or switch treatment. In the supplementary analysis incorporating switching between

different anticoagulants by patients, the covariates will be reassessed at the start of each treatment period, with

the switch flagged so that this can be included in the analysis.

N. Data/ Statistical Analysis

We will use a Cox regression analysis to derive adjusted hazard ratios with 95% confidence intervals for the

primary outcome (major bleed) and secondary outcomes (ischaemic stroke, VTE, all-cause mortality) comparing

novel anticoagulant exposure with warfarin as the primary reference group. We will adjust for potential

confounders as listed above, evaluating them at the date of the first prescription for an anticoagulant during the

study interval. We will use fractional polynomials to model non-linear risk relationships with continuous variables

such as age, body mass index and systolic blood pressure[19].

In a supplementary analysis, where switching between drugs is incorporated into the analysis, we will use a time-

varying Cox regression analysis and re-evaluate confounders each time a switch occurs. Patients in this analysis

will be censored on the earliest date of the first major bleed (primary outcome analysis) or secondary outcome

(minor bleeds, ischaemic stroke or VTE), death, deregistration with the practice, last upload of computerised

data, or the study end date.

A study with this protocol will also be carried out using data from another primary care database (QResearch)

which currently contains about 1300 English practices linked to HES and ONS data. The two studies will be

conducted identically, selecting the same confounders and running the same procedures. All observations will be

from general practices in the UK, from the same study period, with similar exposures and using similar methods

for recording outcomes. Adjusted hazard ratios from the Cox regression analyses of the two datasets will be

pooled using a fixed effect model with inverse variance weights. We will also run a sensitivity analysis using a

random effect model to allow for any heterogeneity.

A 1% level of statistical significance will be used to allow for multiple comparisons. Stata v 14 will be used for all

the analyses.

O. Plan for addressing confounding

We will consider all potential confounding factors listed above. Patient characteristics, selected medications and

chronic and acute conditions will be considered as a priori confounders because they are known risk factors and

affect doctor’s prescribing decision. We will also assess for interactions between the exposures and confounders.

P. Plans for addressing missing data

We will use multiple imputation to replace missing values for body mass index, systolic blood pressure, smoking

status and alcohol status, and use these values in our main analyses[20-23]. As ascertained through preliminary

data exploration we expect only a small proportion of missing data so will carry out up 5 imputations[24] and

combine results using Rubin’s rules. We will examine interactions between different anticoagulants and age and

sex. We will also examine interactions between anticoagulants and other prescribed medications and the co-

morbidities listed as confounders.

Q. Patient or user group involvement (if applicable)

We will use the PPI infrastructure in Nottingham, which is currently being established. We will hold meetings to

get advice on how best to engage with the public regarding this type of drug equity and safety research in order

to ensure that key questions are answered and will be presented in a way both useful and understandable to the

Page 84 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960




public. We will also engage with the patient representatives on the QResearch advisory Board which include the

National Patient Participation Association. We will use existing websites and social media to raise awareness of

our study and engage with the public.

R. Plans for disseminating and communicating study results, including the presence or absence of any restrictions on the extent and timing of publication

The findings will be published in a peer-review journal, and disseminated through scientific conferences and

general practices and in community settings.

S. Limitations of the study design, data sources, and analytic methods

The limitation of the study is the potential misclassification of exposure to anti-coagulants. The treatment may be

prescribed in hospitals or anti-coagulation clinics and such records are not available through primary care

databases. Such patients will appear as false stoppers. These potential misclassification is likely to be small but

might shift the hazard ratios towards unity.

T. References

1. Connolly, S.J., et al., Dabigatran versus Warfarin in Patients with Atrial Fibrillation. New England Journal of Medicine, 2009. 361(12): p. 1139-1151.

2. Patel, M.R., et al., Rivaroxaban versus warfarin in nonvalvular atrial fibrillation. N Engl J Med, 2011. 365(10): p. 883-91.

3. Ruff, C.T., et al., Comparison of the efficacy and safety of new oral anticoagulants with warfarin in patients with atrial fibrillation: a meta-analysis of randomised trials. Lancet, 2014. 383(9921): p. 955-62.

4. Miller, C.S., et al., Meta-analysis of efficacy and safety of new oral anticoagulants (dabigatran, rivaroxaban, apixaban) versus warfarin in patients with atrial fibrillation. Am J Cardiol, 2012. 110(3): p. 453-60.

5. Holster, I.L., et al., New oral anticoagulants increase risk for gastrointestinal bleeding: a systematic review and meta-analysis. Gastroenterology, 2013. 145(1): p. 105-112 e15.

6. Hernandez, I., et al., Risk of bleeding with dabigatran in atrial fibrillation. JAMA Internal Medicine, 2015. 175(1): p. 18-24.

7. Graham, D.J., et al., Cardiovascular, Bleeding, and Mortality Risks in Elderly Medicare Patients Treated With Dabigatran or Warfarin for Nonvalvular Atrial Fibrillation. Circulation, 2015. 131(2): p. 157-164.

8. Lauffenburger, J.C., et al., Effectiveness and Safety of Dabigatran and Warfarin in Real‐World US Patients

With Non‐Valvular Atrial Fibrillation: A Retrospective Cohort Study. Journal of the American Heart Association, 2015. 4(4).

9. Sørensen, R., et al., Dabigatran use in Danish atrial fibrillation patients in 2011: a nationwide study. BMJ Open, 2013. 3(5): p. e002758.

10. Larsen, T.B., et al., Efficacy and Safety of Dabigatran Etexilate and Warfarin in “Real-World” Patients With Atrial Fibrillation: A Prospective Nationwide Cohort Study. Journal of the American College of Cardiology, 2013. 61(22): p. 2264-2273.

11. Abraham, N.S., et al., Comparative risk of gastrointestinal bleeding with dabigatran, rivaroxaban, and warfarin: population based cohort study. BMJ, 2015. 350: p. h1857.

12. Chang, H.-Y., et al., Risk of gastrointestinal bleeding associated with oral anticoagulants: population based retrospective cohort study. BMJ, 2015. 350.

13. Hippisley-Cox, J. and C. Coupland, Predicting risk of upper gastrointestinal bleed and intracranial bleed with anticoagulants: cohort study to derive and validate the QBleed scores. BMJ, 2014. 349: p. g4606.

14. Johnson, E.S., et al., The incident user design in comparative effectiveness research. Pharmacoepidemiol Drug Saf, 2013. 22(1): p. 1-6.

15. Button, L., S. Roberts, and P. Evans, Hospitalized incidence and case fatality for upper gastrointestinal bleeding from 1999 to 2007: a record linkage study. Aliment Pharmacol Ther, 2011. 33(1): p. 64 - 76.

16. Crooks, C., J. West, and T. Card, Upper gastrointestinal haemorrhage and deprivation: a nationwide cohort study of health inequality in hospital admissions. Gut, 2011. 61: p. 514 - 520.

17. Pisters, R., et al., A novel user-friendly score (HAS-BLED) to assess 1-year risk of major bleeding in patients with atrial fibrillation: the Euro Heart Survey. Chest, 2010. 138(5): p. 1093-100.

Page 85 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960




18. Tata, L.J., et al., Does concurrent prescription of selective serotonin reuptake inhibitors and non-steroidal anti-inflammatory drugs substantially increase the risk of upper gastrointestinal bleeding? Aliment Pharmacol Ther, 2005. 22(3): p. 175-81.

19. Royston, P., G. Ambler, and W. Sauerbrei, The use of fractional polynomials to model continuous risk variables in epidemiology. Int J Epidemiol, 1999. 28: p. 964-974.

20. Schafer, J. and J. Graham, Missing data: our view of the state of the art. Psychological Methods, 2002. 7: p. 147-177.

21. Group, T.A.M., Academic Medicine: problems and solutions. British Medical Journal, 1989. 298: p. 573-579. 22. Steyerberg, E.W. and M. van Veen, Imputation is beneficial for handling missing data in predictive models.

J Epidemiol Community Health, 2007. 60: p. 979. 23. Moons, K.G.M., et al., Using the outcome for imputation of missing predictor values was preferred. J

Epidemiol Community Health, 2006. 59: p. 1092. 24. White, I.R., P. Royston, and A.M. Wood, Multiple imputation using chained equations: Issues and guidance

for practice. Stat Med, 2011. 30(4): p. 377-99.

List of Appendices (Submit all appendices as separate documents to this application) ICD10 codes for primary and secondary outcomes 2 peer reviews from School of Primary Care Research (National Institute for Health Research)

Page 86 of 86


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Documents

Confidential: For Review Only · 2010-10-17 · Confidential: For Review Only ABSTRACT Objectives: To investigate associations compared to warfarin between direct oral anticoagulants