Trends in Hypertension Hypertensive Emergency Case Decision and Classification

Layered Procedures inHypertensive Emergencies

Dr. Klaus Fiedler IHAMB, University of Basel, Basel, Switzerland

Trends in Hypertension

Hypertensive EmergencyCase Decision and ClassificationTherapy and Cardiovascular Complications

Primary Events

Secondary Events

Comparison of Primary and Secondary Events in this Study

Conclusions


Time period 80 – 09Values from WHO

Systolic blood pressure in mmHg (not age-standardized)

In 180countries

Blood Pressure Mean: Mencoefficient of regression ( )b

Average systolic values from 1980 to 2009 from WHO

Average systolic in mmHg (not age-standardized)

p<0.05

Male Blood Pressures

1980 – 2009 Systolic in mmHg

Blood Pressure Mean: Womencoefficient of regression ( )b

Average systolic values from 1980 to 2009 from WHO

Average systolic in mmHg (not age-standardized)

p<0.05

Female Blood Pressures

1980 – 2009 Systolic in mmHg

Simple Trends in Blood Pressure

betas of linear predictions (male and female) based on raw values from WHO (1980 – 2009)

Male

Fem

ale

Blood Pressure

coefficient of regression female against male

1.43 (CI 1.36-1.52); p=0.001

Simple Trends in Blood Pressure

• if the development is towards higher values of blood pressure (or negligible in male), female trends are ever so often more rapid

• if the development is towards lower values of blood pressure, it is usually more rapid in the female sex

female trends are highly significant

Þ countries with largest gender gaps are…

Top 2% Burkina Faso towards higher valuesMaliNigerNigeria

Top 2% Czech Republic towards lower valuesSpainEstoniaMalta

Physiology: Male and Female with 3 Parameters

Coefficient Male

Coeffi

cien

t Fem

ale

Countries

CI

=b 0.85 (CI 0.78-0.92)

PCA

Blood GlucoseBlood PressureBMI

• women’s values are non-uniformely distributed amongst different countries

• men’s values are uniformely distributed amongst different countries

Trends in Hypertension: 2016

Predictions based on 1980 – 2009Raw Values from WHO

Non-linear predictions; Average systolic blood pressureAdd to baseline of 2009(in mmHg)



Primary Events

Secondary Events


Conclusions

Hypertensive Emergencies

• available patient cohort study from May 2008 in Kanton Basel (BS, BL) and Luzern

• inclusion criteria: blood pressure >180 mmHg/>110 mmHg, age > 20 y

• discriminant analysis, logistic regression, categorical regression, multi layer perceptrons • primary care questionnaire

… in addition to Age and Gender, data on Blood pressure, Drugs, Emergency therapy, Follow-uptherapy, Cardiovascular risk factors were requested to be answered. Further questions referred to Accompanying conditions and the History of cardiovascular complications, Cardiovascular complications within 3 months

• 1 year follow-up… Cardiovascular events, Follow-up therapy changes and data on newly described Cardiovascular risk factors were requested to be answered

Hypertension Prevalence in the Elderly

Prevalence

Age group (approx.) 65-74

Wolf-Maier et al. (2003); Kearney et al. (2005); Pitsavos et al. (2006) ; eurostat.ec.europa.eu (2008); National Health and Nutrition Survey of Japan (2006) and WHO

Test: Hypertensive Emergencies and Case DecisionModelling of the „clinical decision“ path

E

E + H

H

- H

- E

-E -H

E: Emergency Therapy

H: Hospitalization

Hypertensive Emergencies: SymptomsClassificationEmergency Urgency Asymptomatic*

abdominal pain abdominal pain ataxia altered mental status chest pain Angst confusion atactic gait dizziness/vertigo cold intolerance drops dizziness/vertigo dyspnoea dyspnoea headache epiphora incontinence epistaxis limb weakness flush nausea gait disturbances pallor headache palpitations limb pain pruritus nausea speech disturbances neck pain sweating palpitations visual disturbances pruritus vomitus restlessness

shoulder painsweating tinnitus tremor tympanic pressure visual disturbances

* Does not lead to Emergency or Urgency classification

back paindizzinessepistaxisflushheartburnnauseaslightly altered mental status

included in the statistics,yet, do not count as overt symptoms

Differential Diagnosis

Medical conditions

cardiac acute pulmonary edema with systolic/diastolic dysfunction

malignant hypertension, stroke, aortic dissection

neurologic intracranial / subarachnoid hemorrhage, cerebral embolism / stroke, hypertensive encephalopathy

traumatic head trauma

renal chronic pyelonephritis, renal parenchymal diseases (glomerulonephritis), renovascular disease

drugs monoamine oxidase inhibitor interactions, cocaine overdose

endocrine pheochromocytoma (excess catecholamines), aldosteronism, renin-secreting tumors, glucocorticoid excess

metabolic hypercalcemia

pregnant preeclampsia, eclampsia

psychogenic hyperventilation, Angst, sympathetic hyper-tonus, panic attack

Brennan et al. 2010, Critical Care Study Guide

„Available Patient Cohort“ Study

164 patients included

blood pressure determination14 monthsn= 137

time course 18.4 months

8 patientsdied

lost to follow-upn=26

Cardiovascular Differential Diagnosis: Reevaluation

No Yes

Hypertension

8 Cases were selected

6: Other5: Congestive Heart Failure4: TIA3: Cerebrovascular Insult2: Occlusive PAD1: Coronary Artery Disease0: None

Hospitalization of Patients

Emergency hospitalization patient class diagnosis

1 Emergency acute myocardial infarction2 Emergency acute coronary syndrome3 Emergency suspected cerebrovascular insult4 Emergency suspected cerebrovascular insult5 Emergency suspected cerebrovascular insult6 Emergency hypertensive encephalopathy7 Emergency hypertensive encephalopathy8 Emergency hyponatremia, hypertensive encephalopathy9 Urgency mesenteric ischemia, bowel obstruction, hyperthyroidism10 Urgency suspected cerebrovascular insult

Summary of Patient Data164 patients

sex f 103 m 61

age Average 70,4 years (21-96)

Asymptomatic (n) 99

Urgency (n) 50

Emergency (n) 15

systolic hypertension (n) 153

diastolic hypertension (n) 40

sytol./diastol. hypertension (n) 29

systolic blood pressure Median 195 mmHg (145-255)

diastolic blood pressure Median 100 mmHg (60-130)

hypertension 78,7%first observed hypertensive emergency 62,2%

Anz

ahl

Emergency Urgency Asymptomatic Classification

num

ber

sex mf

Therapy by Emergency Classes

num

ber

Emergency Urgency AsymptomaticClass

no therapymonotherapy combination therapy

Time-Series

systolic

diastolic

hypertensive emergency

1 h

6 h

12 h

6 days

15 days

3.5 months

14 months

Cardiovascular Risk Factors

perc

enta

ge o

f pati

ents

none

hyperlipidemia

lack of exerciseobesity

smoking

Diabetes mellitus

family history

renal insufficiency

Sex-Specific DistributionFactorsDiabetes mellitusrenal dysfunctionphysical inactivityhyperlipidemiaobesitysmoking

cum

ulati

ve ri

sk fa

ctor

s

m f Sex

Cardiovascular Risk Factors After 1 Year

Accompanying Medical Conditions

OSAS

perc

enta

ge o

f pati

ents

Increase in body weight

Infection

Alcoholism

Asthma

NSAR

Pain

Non-compliance

Stress

White Coat hypertension

None

Patients and Frequency

• first hypertensive emergencies occur often with new patients

• White Coat hypertensive patients show fewer symptoms upon examination

• symptoms correlate with stress, NSAR medication and infects

correlative evidence shows…

Neural Network

The network identified 92% of the patients with acute myocardial infarction, 96% of the patients without infarction. When all patients with the ECG evidence of infarction were removed from the cohort, the network correctly identified 80%

(modified from Baxt 1990)

Key Question

“Does the evaluation contain previously inapparent informationthat can be used to improve on the diagnostic accuracy of predicting…”Baxt, 1990

Neural Network

• hypertension• gender• lack of exercise / obesity

Retained Parameters in Logistic Regression

n=164

0.020 0.035 0.042 0.017 0.000 0.021 0.004 0.046p=

Asymptomatic

E E E E E U E U

Pseudo R2=0.321, goodness of classification 64%

Class

Infection

New patient

StressW

hite Coat hypertension

History of cardiovascular events

NSARW

hite Coat hypertension

History of cardiovascular events

Dynamic Scaling of Emergency Classes

Separation of U and A requires use of Symptom variable?

E U A-0.5

00.5

11.5

22.5

33.5

44.5

Model: Predictors in logistic regression

Model: Predictors + Symptoms

Classification

Goodness of Classification

30 %

80 %

„Structural Equation“model without parameters

E U A-0.5

00.5

11.5

22.5

33.5

44.5

Model: Predictors in logistic regression

Model: Predictors + Symptoms

Model: Final

Classification

Dynamic Scaling of Emergency Classes

„The Diagnostic Gap“

Goodness of Classification

30 %

80 %

44 %

Classification in Hypertensive Emergencies

Importance

Model unifying 7 different „output functions“ counting 2400

E 79.8% +/-15.8%U 96.9% +/-2.5%A 96.3% +/-2.4%

Total 95.0% +/-2.4%

(+/- SD)

proposedcut-off

Top 10%

max. epochs 500

Optimized Top 500 Models of 2400

Hidden Layer AF

Hyperbolic tangent 321Sigmoid 182

Ouput Layer AF

Hyperbolic tangent 83Identity 93Sigmoid 37Softmax 290

AUCs ≥ 0.986

Optimized Top 500 Models of 2400

Importance

Model unifying 7 different „output functions“ - Top 500

E 100.0% +/-0.0%U 97.7% +/-1.1%A 99.6% +/-0.7%

Total 99.0% +/-0.6%

(+/- SD)

Classification in Discriminant Analysis

Function 1

Func

tion

2

Asymptomatic

Urgency

Emergency

EmergencyUrgencyAsymptomatic

Classification in Discriminant Analysis

Geschlecht

Erste Entgleisung

Hypertonie

Neuer Patient

Stress

Schmerz

Infekt

NSAR

Weisskittelhypertonie

Alter

Anam. Kardiovaskuläre Kompli.

Symptome

0 10 20 30 40 50 60 70 80 90 100

Varianz Modell 1Varianz Modell 2Varianz Modell 3

symptoms

history of cardiovascular events

age

White Coat hypertension

new patient

hypertension

first hypertensive emergency

sex

explained varianceof 200

NSAR

infection

pain

stressproposed cut-off

variancemodel 1model 2model 3

New data on hypertension Method % Classification Overall Error1 Neuronal network 1,0 (+/- 0,6%) - 5,0 (+/- 2,4%)2 Discriminant analysis 10,03 CatReg 16,04 Logistic regression £ 29,9

Hypertensive Emergency: Goodness of Classification

Symptoms were not introduced in logistic regression

Treatment

Patient Recall?

Seemed not absolutelynecessary…

Drug History and Emergency Therapyp <0.05 Cardiovascular treatment track

•ACE inhibitors in drug history lead to a significantly higher prescription rate of sedatives

• previous treatment with ARB results in frequent emergency treatment with ARB

• ARBs in drug history lead to infrequent therapy with ACE inhibitors in hypertensive emergencies

• patients receiving ARB in emergency treatment regularly receive ARB in follow-up treatment

Adverse Effects in Therapies

Drugs causing drug-induced hypertension and hypertension treatment

Therapy: According to Guideline

Blood pressure reduction

Sub-group Significance p<0.05

6 h systolic alltogether yes

1 h systolic E yes

Combination Therapy

Blood pressure reduction

Sub-group Significance p<0.05

all systolic alltogether or E, U, A none

all diastolic alltogether or E, U, A none

Mono-Therapy

Mann-Whitney U Test

Time-Series

blood pressurerecommendation1 to 2 h and 2 to 6 h

systolic

diastolic

hypertensive emergency

1 h

6 h

12 h

6 days

15 days

3.5 months

14 months

Acutely treated patients

Time-Series: Emergency

mmHg


Time-Series: Urgency

mmHg


Time-Series: Asymptomatic

mmHg


Time-Series: Emergency

mmHg


Time-Series: Urgency

mmHg

mmHg


Time-Series: Asymptomatic

Average of Blood Pressures: Treated Versus Non-Treated

Time Treatment None

Initial 200 / 102 194 / 98

1 h 186 / 97 186 / 92

6 h 160 / 87 163 / 80

in mmHg

All emergency classes

blood pressure* reduction achieved by 20-30% in 1-2 hours

• the acutely treated study population was not over-treated, one emergency and three urgency, and no other cases were likely loosing systolic blood pressure too fast (not all 6 hour blood pressure values were obtained)

• at least 16 cases showed systolic blood pressure values of more than 160 mmHg at 6 hours

Textbook Guideline

* Dieterle T, Zeller A, Martina B, Battegay E. Der hypertensive Notfall. Praxis. 2001

Average of Blood Pressures: Treated Versus Non-Treated

Initial 0 h 1 h 6 h

% above 160 99,0 95,2 55,2

% below 100 40,4 55,4 75,9

Treatment

Systolic

Diastolic

Initial 0 h 1 h 6 h

% above 160 96,7 92,7 44,4

% below 100 43,3 63,4 88,9

No Treatment

Systolic

Diastolic

All emergency classes

in mmHg

in mmHg

Cardiovascular Complications

Arrhythmias were not considered as primary events in the following canonical correlation.They were only included with secondary events if they led to or were followed by death.



Primary Events

Secondary Events


Conclusions

Recurrent Hypertensive Crises - Model and Correlation

Canonical correlation

modelled initial (day 6) blood pressure and risk factors

Recurrent Hypertensive Crises

Perceptrons: Primary Events and Predictors

Importance

Model unifying 6 different „output functions“ – 2400

Top 1%


1° Event 100.0% +/-0.0%

Total 98.4% +/-0.8%

(+/- SD)

1° Event 40.4% +/-26.7%

Total 91.2% +/-2.7%

Classification

Top 1% 2400 models

Other 2 selected models

1° Event 94.9% +/-2.5%

Total 98.6% +/-2.5%

(+/- SD)

Obesity Paradox

Flegal, K. M., Kit, B. K., Orpana, H., & Graubard, B. I. (2013). Association of all-cause mortality with overweight and obesity using standard body mass index categories: a systematic review and meta-analysis. JAMA, 309(1), 71–82.


Importance

Model unifying 6 different „output functions“ – 2400

2 100% Predictions

Models of Primary Events: Discriminant Analysis

Age p=0.011Multiple (recurrent) hypertensive crises p=0.004

in stepwise algorithm

(no prior probabilty) set classified 78.2%


Classification Method Average Overall % Primary Events %1 neuronal network 91,2 - 98,4 40,4 - 100,02 discriminant analysis 64,1 - 84,8 9,1 - 63,33 logistic regression 88,0 - 89,1 9,1 - 27,3

Primary Events: Goodness of Classification

n=92



Primary Events

Secondary Events


Conclusions

Diastolic Blood Pressure - Model and Correlation


Chillon and Baumbach (1997)


New Tentative Guideline for Secondary Prevention

Non-Linear Blood Pressure Model with CatReg

Parameter Importance P of b (regression)

Age 0.122 0.254

Sex 0.237 0.148

Systolic 0.291 0.034

Diastolic 0.349 0.005

R2 = 0.51

Day 6

Blood Pressure of day 6

Tentative New Guideline for Secondary Prevention

QuantificationResiduals



A

ge

S

ysto

lic B

lood

Pre

ssur

e

D

iast

olic

Blo

od P

ress

ure

Secondary Event

Secondary Event

Secondary EventCategories

Categories

Categories



Primary Events

Secondary Events


Conclusions

Secondary Versus Primary in 2198 Neuronal Networks: Classification of 100% versus <90%

Importance

< 90%

Secondary Versus Primary Event in Categorical Regression

Age Alcoholism

Asthma Diabetes mellitus

Family History of Cardiovascular Disease Hyperlipidemia

Hypertension Lack of Exercise

Non-Compliance Obesity

OSAS Renal Insufficiency

Second Hypertensive Emergency Sex

Smoking White Coat Hypertension

-2 -1 0 1 2 3 4

Coefficient Beta

Tendency Primary Event Secondary

Modalized Family History of Cardiovascular Disease

Ambiguity of Well Known and Other Risk Factors

-60 -40 -20 0 20 40 60 80

Second Hypertensive Emergency

Family History of Cardiovascular Disease

Diabetes mellitus

Lack of Exercise

OSAS

Obesity

Alcoholism

Asthma

Renal Insufficiency

Hyperlipidemia

White Coat Hypertension

Non-Compliance

Hypertension

Sex

Smoking

Age

From primary versus secondary event with perceptron and CatReg data

1

2

34

56

Stratification

Kardiovask. Komplikationen

Kardiovask. Komplikationen

12 3 4 5 6 1

2

3 46

emergencyscores lowest

5

controlhistory of cv event

1° cardiovascular event

2° cardiovascular event

death (cardiovascular)

death (other)

-1

0

1

2

3

1

23

45

6 Diabetes mellitusrenal dysfunctionphysical inactivityhyperlipidemiaobesitysmoking

Age Group

cum

ulati

ve ri

skcu

mul

ative

risk

year

s

Stratification: Risk Ratios

• in patients with hypertensive emergencies history of cardiovascular events is associated with hyperlipidemia with a RR 3.3 (CI 1.2-8.9); p=0.017 (Fisher Exact)

• a history of cardiovascular events is associated with antihypertensive pretreatment with a RR 4.4 (CI 1.2-15.8); p=0.024 (Fisher Exact)

• secondary cardiovascular events are associated with smoking with a RR 3.2 (CI 0.8-12.6); p=0.1 (Fisher Exact)

this is relative to healthy cohort patients

Anamnestic symptoms

Antihypertensives pretreatment

AsthmaWhite Coat hypertension

-2

0

2

4

b

regressioncoefficient

0.016 0.029 0.006 0.049p

Model prob. p = 0.005; df 21

Stratified Population: Further PredictionsGLM

Sympathetic Drive

Grassi (2009) Assessment of sympathetic cardiovascular drive in human hypertension. HypertensionLlewellyn et al. (2011) MnPO and SFO drive renal sympathetic nerve activity via a glutamatergic mechanismwithin the paraventricular nucleus.* Am J Physiol Regul Integr Comp Physiol

sub-category

pulmonary reflexes stemming from cardiopulmonary volume-sensitive receptors

vascular arterial chemoreceptors altered in obesity or overweight-dependent OSAS

metabolic leptin, insulin

renal on neuronal

angiotensin II

neuronal on renal

subfornical organ and median preoptic nucleus in part via hypothalamus

nutritional/behavioral

low sodium intake, alcohol excess

Sympathetic Drive

Koeppen & Stanton: Berne & Levy Physiology

2: Increased secretion of renin, which results in higher angiotensin II levels

*

* Neuronal on renal signalling

Cardiovascular Progression Model from Strata3200 models

Importance

/ selection of 1 best predictor

1 95.8% +/- 4.2%2 81.2% +/-16.4%3 26.0% +/-24.2%4 31.8% +/-31.6%5 17.5% +/-30.0%6 8.1% +/-23.8%

Total 78.4% +/-6.8%

Top 1

Total 94.9%



Primary Events

Secondary Events


Conclusions

White Coat Hypertension - Model



White Coat Hypertension

Health Ranking: The Physiology According to the Big 4Female Male

From WHO Data

higher physiologicalvalues

lower physiologicalvalues

lower values malelower values female

Blood GlucoseBlood PressureBMICholesterol

Rank 133

SwitzerlandRank 121

Female Risk Ratio forWhite Coat Hypertension2.3

Decision Path

E U S01234

Initial Final

2.36

4.62

2.08

2.38

4.62

2.36

2.01

2.03

3.82

Model: Green leads to higher and Red to lower ranking

• primary cardiovascular outcome (primary event) in this study cohort does not positively correlate with the number of risk factors

• secondary cardiovascular outcome and risk factor association is shown for renal insufficiency and smoking

• neural networks can produce predictions for primary events based on known cardiovascular risk factors

• causal relationships in these neural networks may end upside down - reasons are discovered in the obesity and low alcohol consumption

(Hyperplane Extraction Procedures may be found in e.g. Saad and Wunsch II, Neural Networks 20 (2007), 78)

Cardiovascular Risk Factors: Not Known?

• primary cardiovascular events after multiple hypertensive emergencies are elevated to a rate (per year) of 56%

• multiple hypertensive crises are overwhelmingly important in the neuronal network prediction

• in models with stratification, white coat hypertension has protective influence relative to hypertension

• in non-linear models, secondary or higher ranked events might be prevented by slightly elevated diastolic blood pressure of 90 mmHg and higher after hypertensive emergencies during 6 days, and systolic values from 130 – 165 mmHg

Study: Results

Acknowledgements

Prof. Dr. Benedikt Martina, IHAM BaselDr. Christoph Merlo, Luzern

and especially the primary care doctors

Members of my laboratory in Basel and in particular Elena Kouzmenko and Christian Chatenay-Rivauday

Documents

Trends in Hypertension Hypertensive Emergency Case Decision and Classification