Prolonged Prothrombotic Effectsof Antecedent Hypoglycemia inIndividuals With Type 2 DiabetesDiabetes Care 2018;41:2625–2633 | https://doi.org/10.2337/dc18-0050

OBJECTIVE

Hypoglycemia has been linked to persistent increases in cardiovascular (CV) mortalityin type2diabetesafter theevent.Ouraimwastoexamineacuteanddownstreameffectsof hypoglycemia on markers of thrombosis risk and inflammation in type 2 diabetes.

RESEARCH DESIGN AND METHODS

Twelve individuals with type 2 diabetes with no history of CV disease and 11 age- andBMI-matched volunteers without diabetes underwent paired hyperinsulinemic-euglycemic (glucose 6 mmol/L for two 60-min periods) and hypoglycemic (glucose2.5 mmol/L for two 60-min periods) clamps on separate occasions on day 0. Fibrinclot properties, platelet reactivity, and inflammatory markers were measured atbaseline, end of and after recovery from the initial clamp, day 1, and day 7 usingvalidated assays and electron microscopy.

RESULTS

Euglycemic hyperinsulinemia reduced platelet reactivity, decreased fibrin clotdensity, and improved fibrinolytic efficiency in both groups. Platelet reactivity andaggregation increased during acute hypoglycemia in both groups, resolving atrecovery. In type 2 diabetes, clot lysis times and clot maximum absorbanceincreased up to day 7 (P = 0.002 and 0.001 vs. euglycemia, respectively), but clotsfrom control subjects without diabetes showed limited changes. Fibrin networkdensity increasedD 1.156 0.28 fibers/mm2 at day 7 after the hypoglycemic clamp(P< 0.01 for glycemic arm), whereas fibrinogen and complement C3 increased afterhypoglycemia up to day 7 in type 2 diabetes only.

CONCLUSIONS

Antecedent hypoglycemia has acute and persistent prothrombotic effects, lastingat least 7 days, that were enhanced in individuals with type 2 diabetes. Thesefindings identify mechanisms by which hypoglycemia might increase short- andmedium-term risk of CV mortality.

Cardiovascular (CV) disease is the leading cause of death in type 2 diabetes. Anumber of large clinical trials have attempted to address the role of intensiveglucose control on vascular events and have shown either no reduction (1,2) or anincrease in mortality (3). Hypoglycemia is associated with increased CV mortality(2,4), but evidence establishing cause and effect is lacking. In interventional trials,the increased mortality associated with hypoglycemia extended well beyond theacute event, with elevated risk months later (2). Mechanisms by which hypoglycemiamight lead to increased CV mortality in the short to medium term are unclear.

1Department of Infection Immunity and Car-diovascular Disease, University of Sheffield,Sheffield, U.K.2Department of Oncology and Metabolism, Uni-versity of Sheffield, Sheffield, U.K.3Sheffield Teaching Hospitals NHS FoundationTrust, Sheffield, U.K.4Leeds Institute of Cardiovascular andMetabolicMedicine, University of Leeds, Leeds, U.K.5School of Life Sciences, University of Notting-ham, Nottingham, U.K.

Corresponding author: Simon R. Heller, s.heller@shef.ac.uk.

Received 8 January 2018 and accepted 17 Sep-tember 2018.

This article contains Supplementary Data onlineat http://care.diabetesjournals.org/lookup/suppl/doi:10.2337/dc18-0050/-/DC1.

© 2018 by the American Diabetes Association.Readers may use this article as long as the workis properly cited, the use is educational and notfor profit, and the work is not altered. More infor-mation is available at http://www.diabetesjournals.org/content/license.

Elaine Chow,1,2 Ahmed Iqbal,1,2

Emma Walkinshaw,2,3 Fladia Phoenix,4

Ian A. Macdonald,5 Robert F. Storey,1,3

Ramzi Ajjan,4 and Simon R. Heller2,3

Diabetes Care Volume 41, December 2018 2625

CARDIOVASCULA

RANDMETA

BOLIC

RISK

Platelet hyperreactivity, altered fibrinclot characteristics, and hypofibrinolysishave been linked to excess CV events,especially in type 2 diabetes (5). Exper-imental hypoglycemia in people withtype 1 diabetes has increased plate-let aggregation (6) and inflammatorymarkers, including CD40 expression andsoluble CD40 ligand, platelet-monocyteaggregates, plasma levels of vascularadhesion molecules, and P-selectin (7,8).Repeated episodes of hypoglycemiaimpair nitric oxide–mediated endothe-lial function and increase thrombin/antithrombin complex (9). In clinicalstudies, hypoglycemia increased factorVIII and vonWillebrand factor, which areprocoagulant (10).However, studies examining the

longer-term effects of hypoglycemia onthrombotic and inflammatory markersremain scarce. In particular, no studyto date has studied prothromboticchanges in individuals with type 2 di-abetes in response to hypoglycemia,where insulin resistance and clusteringof metabolic risk factors may differenti-ate them from those with type 1 dia-betes, predisposing them to a state ofdepressed fibrinolysis. Antecedent hy-poglycemia has been shown to alter au-tonomic responses and decrease vagalactivity downstream of the episode(11,12). Suppression of cholinergic anti-inflammatory pathways (13) may lead toprothrombotic and proinflammatory re-sponses after hypoglycemia.We hypothesized that hypoglycemia

might oppose the CV benefits of intensiveglycemic control by inducing prothrom-botic responses both during the episodeand downstream. Thus, the aim of thisstudy was to investigate the effects ofhypoglycemia in patients with type 2diabetes on key thrombotic mechanisms,including 1) fibrin clot structure andfibrinolysis, 2) platelet reactivity andactivation, and 3) plasma levels of inflam-matory markers. We studied prothrom-botic changes during acute hypoglycemiaand up to 7 days after experimentallyinduced moderate hypoglycemia inboth individuals with type 2 diabetesand matched healthy control subjects.

RESEARCH DESIGN AND METHODS

ParticipantsTwelve participants with type 2 diabe-tes aged 18–65 years and with an HbA1cof 6.5–10.5% (48–91 mmol/mol) were

recruited from Sheffield Teaching Hos-pitals outpatient clinics between 2011and 2014. Eleven age- and BMI-matchedindividuals without diabetes were re-cruited as control subjects from staffat Sheffield Teaching Hospitals and theUniversity of Sheffield. The same partic-ipants also had cardiac electrophysiolog-ical measurements recorded during themorning hypersinsulinemic clamps in astudy of the effect of hypoglycemia onautonomic function and repolarization(14). Participants with diabetes weretaking oral hypoglycemic agents and/or glucagon-like peptide 1 analogs and/or insulin for#2 years. Exclusion criteriawere previous myocardial infarction,ischemic heart disease, cardiac arrhyth-mias, stroke or peripheral vascular dis-ease, or other known CV disease; epilepsy;untreated hyperthyroidism; pregnancy;and serious intercurrent illness. Nonewere taking antiplatelet agents or anti-coagulants apart from two participantswith diabetes who were taking aspirin.All participants had normal full bloodcount, renal function (estimated glomer-ular filtration rate .60 mL/min/1.73 m2),electrocardiogram at baseline, and cardiacautonomic function on the basis of stan-dard autonomic function tests (15). Allparticipants gave written informed con-sent. The studywas approvedby the SouthYorkshire local research ethics committee.

Study DesignIndividuals participated in paired hyper-insulinemic-euglycemic and hypoglyce-mic clamp studies separated by atleast 4 but no more than 8 weeks tominimize carryover effects. We used acrossover design such that each partic-ipant served as his or her own controlsubject to reduce variability. Euglycemicstudies preceded hypoglycemic studiesbecause hypoglycemia may have persis-tent effects of the CV system. In previousstudies, two periods of hypoglycemiaresulted in altered autonomic and vagalfunction that could lead to proinflamma-tory and prothrombotic effects in thedays after hypoglycemia (11,12). In onestudy, these changes were observed upto 6 days later (11). Thus, in the hypo-glycemic arm, arterialized blood glucosewas maintained at 2.5 mmol/L for two60-min periods in the morning and af-ternoon with rapid-acting intravenousinsulin at 120 and 240 mU/m2/min, re-spectively, alongside a variable infusion

of 20% dextrose. Participants wereblinded to blood glucose levels. Bloodglucose was raised to euglycemic levelsbetween morning and afternoon hypo-glycemic clamps. In the euglycemic arm,arterialized blood glucose was main-tained at 6 mmol/L for 60 min duringboth morning and afternoon, using sim-ilar rates of insulin infusion as duringhypoglycemia. Insulin was administeredat the same rates in both participantgroups. Prothrombotic and inflamma-tory markers were measured at theendof themorning clampandat recovery(30 min after end of morning clamp) butnot during the afternoon clamp. To in-vestigate downstream effects, pro-thrombotic and inflammatory markerswere measured in the morning at 1and 7 days after both euglycemic andhypoglycemic studies. The studydesign isoutlined in Fig. 1, and details of thehyperinsulinemic clamp protocol are de-scribed in Supplementary Methods 1.

Biochemical MeasurementsFor epinephrine and norepinephrine,whole blood (6 mL) was collected intochilled lithium-heparin tubes containing50 mL EGTA/glutathione as a preserva-tive. These samples were assayed byhigh-performance liquid chromatogra-phy (interassay coefficient of variationfor norepinephrine 6.03% and epineph-rine 15.9%). For free insulin, whole blood(3 mL) was collected into a 6-mL lithium-heparin tube and immediately subjectedto centrifugation. The resulting plasma(0.5 mL) was added to a chilled plastictube containing 0.5 mL polyethyleneglycol for precipitation of immune com-plexes and mixed. The product was an-alyzed by an immunometric assay (InsulinELISA [interassay coefficient of variation7.1%]; Invitron,Monmouth, U.K.). Serumcortisol analyzed from 4 mL of venousblood collected in a serum separatortube was measured using a commercialimmunoassay (Roche e602 serum corti-sol assay [interassay coefficient of vari-ation 3.2%]; Roche Diagnostics, WestSussex, U.K.). Biochemical parameterswere measured at baseline and at endof euglycemic and hypoglycemic armsin both groups.

Turbidimetric and Lysis AssayVenous blood was collected into tubescontaining 3.2% sodium citrate (BDVacutainer Glass Citrate Tube) on ice. Exvivo fibrin polymerization characteristics

2626 Prothrombotic Effects of Hypoglycemia Diabetes Care Volume 41, December 2018

of plasma samples were investigatedby a validated turbidimetric clotting as-say (16) described in detail in Supple-mentary Methods 2.

Markers of Fibrin Dynamics andInflammationFibrinogen and plasminogen activatorinhibitor1 (PAI-1) assayswereperformedon venous blood (4.5 mL) collected intotubes containing 3.2% buffered sodiumcitrate solute (BD Vacutainer Glass Cit-rate Tube). hsCRP was analyzed usingan immunoturbidmetric assay (CardiacC-Reactive Protein [Latex] High Sensi-tive; Roche Diagnostics, Indianapolis, IN).Interleukin-6 (IL-6) was determined usingcytometric bead array (Human IL6 FlexSet; BD Biosciences, Oxford, U.K.). Plasmalevels of complement C3, a protein thatis incorporated into the fibrin clot andmodulates fibrinolysis (17), were deter-mined using ELISA (Biosources, San Diego,CA). Additional details are described inSupplementary Methods 2.

Scanning Electron MicroscopyPooled samples of plasma were analyzedfrom 10 participants with type 2 diabetesand 10 control subjects. Fibrin clots were

made as previously reported (18) anddescribed in detail in SupplementaryMethods 2.

Platelet Impedance AggregometryPlatelet aggregation was analyzed usingmultiple electrode impedance aggreg-ometry (Multiplate; Roche Diagnostics,Basel, Switzerland) and described in de-tail in Supplementary Methods 2.

Platelet P-Selectin ExpressionHirudin-anticoagulated blood was addedto tubes containing phycoerythrin Cy5mouse anti-CD62P antibody (BD Biosci-ences) and platelet marker CD41a (BDBiosciences) for measurement of plateletP-selectin expression and is described indetail in Supplementary Methods 2.

Statistical AnalysisBaseline demographic data and pro-thrombotic and inflammatory markersare summarized as mean6 SD for para-metric data, unless otherwise stated, ormedian (interquartile range) for non-parametric data. hsCRP and IL-6 werelogarithmically transformed because of askewed distribution. End-of-clamp bloodglucose and hormone concentrations

were compared between euglycemicand hypoglycemic time points usingtwo-way ANOVA.

We used a linear mixed model withrepeated measures to compare 1)whether there was significant changein prothrombotic markers over timewithin glycemic arms in each participantgroup and 2) the effect of the glyce-mic arm on changes in prothromboticmarkers in each participant group. Foranalysis 1, time point was specified as afixed effect with multiple comparisonsagainst baseline adjusted by the Sidakcorrection. In mixed-model analysis 2,glycemic arm, time, and interactionbetween glycemic arm and time werespecified as fixed effects and the partic-ipant as random effect. Glycemic arm andtime point were specified as repeatedmeasures within each participant. Re-peated measures were fitted with anunstructured, compound symmetry, orautoregressive 1 covariance structure,and the model with the best fit (low-est Akaike information criterion) wasselected. P values were obtained byrestricted maximum likelihood estima-tion. The effect of glycemic arm andinteraction between time and glyce-mic arm on fiber network density andfibrin thickness were analyzed usingtwo-way repeated-measures ANOVA. Wefurther calculated the within-individualcorrelations between clot maximumabsorbance (MA) or clot lysis time withfibrinogen, PAI-1, and C3, taking intoaccount repeated measures, withANCOVA (19). P, 0.05 was consideredsignificant. Results were analyzed us-ing SPSS version 20 software (IBMCorporation).

On the basis of our initial pilot data, asample size of 10 participants per grouphad a power of 81% to detect a clini-cally relevant 15% difference in clotlysis time, assuming an SD of 100 s. Weaimed to recruit 12–13 participantsfor each group, allowing for a 20–30%dropout rate.

RESULTS

Baseline CharacteristicsTwelve patients with type 2 diabetes(nine male) and 11 individuals withoutdiabetes (five male) participated in thestudy. Participants with diabetes weresimilar in age (54 years [50–58 years])compared to control subjects (52 years[47–59 years]; P = 0.90). Mean BMI was

Figure 1—Study design. Each subject with or without type 2 diabetes participated in euglycemicand hypoglycemic clamp visits separated by 4–8 weeks. Euglycemic clamp preceded hypogly-cemic clamps. In the euglycemic clamp visit, blood glucosewasmaintained at 6mmol/L for 60minin the morning and 60 min in the afternoon. In the hypoglycemic clamp visit, blood glucosewas maintained at 2.5 mmol/L for 60 min in the morning and 60 min in the afternoon. Bloodsamples for platelet function, clotting assays, coagulation proteins, and counterregulatoryhormones were collected at baseline, end of clamp at 60 min, recovery from morning clamp,and the morning of day 1 and day 7 after the euglycemic and hypoglycemic clamp visits,respectively. EU, euglycemic; HYPO, hypoglycemic.

care.diabetesjournals.org Chow and Associates 2627

comparable (346 5 vs. 316 8 kg/m2 inthe diabetes vs. nondiabetes group, re-spectively; P = 0.18). The median dura-tion of diabetes was 10 years (8–12years), and mean HbA1c was 7.8 61.3% (62 6 14 mmol/mol). Amongpatients with diabetes, five were takingoral hypoglycemic agents only, five weretaking oral hypoglycemic agents and aglucagon-like peptide 1 analog, and twowere taking oral hypoglycemic agentsand basal insulin. Baseline prothrom-botic and inflammatory markers areshown in Supplementary Table 1. Par-ticipants with diabetes had higher plate-let reactivity to ADP, but no significantdifferences were detected in fibrino-gen, PAI-1, C3, and hsCRP plasma levels(Supplementary Table 1).

Blood Glucose and CounterregulatoryHormonesBlood glucose levels were similar at end-of-morning euglycemic and hypoglycemicclamps in both groups (Supplemen-tary Table 2). Glucose targets were

reached during afternoon clamps withineach arm. During day 1, blood glucosewas higher after hypoglycemic clampscompared with euglycemia in the diabe-tes group only. Blood glucose levelswere similar between glycemic arms atday 7. Participants with diabetes did notreport symptomatic hypoglycemia or cap-illary glucose values ,3 mmol/L in theweek after each clamp.

Median (interquartile range) free in-sulin levels at 120 min were 576 pmol/L(468–627pmol/L) during euglycemia and689 pmol/L (477–1,076 pmol/L) duringhypoglycemia in the diabetes group. Inthe nondiabetes group, these valueswere 865 pmol/L (509–952 pmol/L) inthe euglycemic arm and 665 pmol/L(468–967 pmol/L) in the hypoglycemicarm. Insulin levels between groups dur-ing both euglycemia and hypoglycemiawere not different (P = 0.23).

Counterregulatory hormones wereunchanged during euglycemia (Sup-plementary Table 2). During acute hypo-glycemia, epinephrine, norepinephrine,

and cortisol increased significantly (allP , 0.05 vs. baseline) in both groups(Supplementary Table 2). Plasma epi-nephrine and cortisol returned to base-line at day 1 and day 7 after both arms inboth participant groups. However, inthe nondiabetes group, plasma norepi-nephrine levels increased significantlyat day 1 after both clamps.

Fibrin Clot Properties

Clot Lysis Time

In the nondiabetes group, clot lysis timesdecreased after euglycemia (from 729 6216 to 611 6 159 s, D 2146 6 110; P =0.001 vs. baseline) but returned to base-line at day 1 (Fig. 2A). No changes in clotlysis times were found in the hypogly-cemic arm in the nondiabetes group. Inthe diabetes group, clot lysis times de-creased at the end of the euglycemicclamp (D 281 6 86 s) but were pro-longed at the end of hypoglycemia, withfurther increases at day 1 (D 716 153 s)and day 7 (D 676 107 s). Changes in clotlysis times were significantly different

Figure 2—Effect of euglycemia and hypoglycemia on clot lysis time (A) and clot density (B) by turbidimetric and lysis assay. Differences between glycemicarm and the interaction between time and glycemic arm by a mixed model with repeated measures are shown. Data are mean 6 SE. ††P , 0.01euglycemia vs. hypoglycemia at equivalent time points; *P , 0.05 vs. baseline. DM, diabetes mellitus; EU, euglycemic; HYPO, hypoglycemic.

2628 Prothrombotic Effects of Hypoglycemia Diabetes Care Volume 41, December 2018

between the euglycemic and hypoglyce-mic arms in participants with diabetes(P = 0.001) (Fig. 2A).

Clot MA

In the nondiabetes group, clot MA de-creased at the end of both clamps butto a lesser extent during hypoglycemia(D 20.02 6 0.05 arbitrary units [AU])compared with euglycemia (D 20.05 60.05 AU; P = 0.02 for glycemic arm) (Fig.2B). In the diabetes group, clot MA de-creased during euglycemia and recovery.A nonsignificant increase was observedin clot MA at the end of hypoglycemia(D 0.02 6 0.05 AU), which resolved atrecovery, followed by gradual increasesat day 1 and 7 (Fig. 2B). Significant dif-ferences were found in clot MA be-tween glycemic arms in the diabetesgroup (P = 0.002) and interaction be-tween time and glycemic arm (P = 0.02).

Scanning Electron Microscopy

Fibrin Diameter

Fibrin diameter did not change signifi-cantly during euglycemia or hypoglyce-mia in the nondiabetes group (Fig. 3A).In the diabetes group, no changes wereobserved in fibrin fiber diameter duringeuglycemia, which increased at day 1and 7 after hypoglycemia (P , 0.01 vs.euglycemia at equivalent time points)(Fig. 3A). A significant difference wasfound between glycemic arms (P ,0.0001) and in the interaction betweentime and glycemic arm (P , 0.0001).

Fibrin Network Density

In the nondiabetes group, fiber net-work density decreased at day 1 aftereuglycemia compared with no changeafter hypoglycemia (P , 0.001 for gly-cemic arm). The fibrin network densityreturned to baseline at day 7 aftereuglycemia but increased at day 7 afterhypoglycemia (P , 0.01 for glycemicarm) (Fig. 3B).In the diabetes group, fiber network

density decreased at day 7 after eugly-cemia, whereas after hypoglycemia, anincrease was found in fiber network den-sity at day 1 and 7 (both P , 0.01 forglycemic arm) (Fig. 3B). Significant differ-ences were found between the glycemicarms and the interaction between timeand glycemic arm in both groups (bothP , 0.001). Representative examples ofscanning electron micrographs of pooledfibrin clots are shown in Fig. 3C.

Figure 3—Fibrin network properties and scanning electron micrographs of fibrin clots aftereuglycemia vs. hypoglycemia in participants with andwithout diabetes. Fiber network density (A)and fibrin fiber thickness (B) after euglycemia vs. hypoglycemia. Fiber thickness wasmeasured fora total number of 160 fibers (40 fibersmeasured from four different clot areas at each time point).Data are mean 6 SE. C: Visualization of ex vivo fibrin clots from pooled plasma samples inparticipants with (n = 10) and without (n = 10) diabetes. A decrease is seen in the fiber networkafter euglycemia in both groups as opposed to an increase in the network density after hy-poglycemia at day 7. *P , 0.01, **P , 0.01, ****P , 0.0001 euglycemia vs. hypoglycemia. EU,euglycemic; HYPO, hypoglycemic.

care.diabetesjournals.org Chow and Associates 2629

Coagulation ProteinsIn the nondiabetes group, fibrinogentended to decrease at the end of theclamp and recover similarly during botheuglycemia and hypoglycemia. However,in the diabetes group, fibrinogen levelsdid not change during euglycemia butincreased: D 0.20 6 0.10 mg/mL at day1 andD 0.856 0.69mg/mL at day 7 afterhypoglycemia (P = 0.05 for glycemic arm)(Supplementary Fig. 1A).PAI-1 falls in both groups were similar

during euglycemia and hypoglycemia(Supplementary Fig. 1B). PAI-1decreasedin the nondiabetes group during eugly-cemia and hypoglycemia (P = 0.15 and0.005 for time, respectively), with nosignificant differences between arms(P = 0.56). In the diabetes group, PAI-1 decreased from baseline to D28116204 pg/mL at the end of euglycemia (P =0.02 vs. baseline) and remained lowerat day 7. PAI-1 decreased to a similar ex-tent during hypoglycemia (D 2888 6256; P = 0.02 vs. baseline), with no dif-ferences detected between glycemicarms (P = 0.85).Change in clot MA correlated with

changes in fibrinogen (r = 0.98; P ,0.001), PAI-1 (r = 0.99; P , 0.001),and C3 (r = 0.99; P , 0.001) acrossboth groups. Change in clot lysis timecorrelated with changes in fibrinogen(r = 0.98; P , 0.001), PAI-1 (r = 0.99;P , 0.001), and C3 (r = 0.98; P , 0.00).No correlation was found among epi-nephrine, norepinephrine, and clot MAor clot lysis times.

Inflammatory MarkersIn the nondiabetes group, C3 levels weresimilar during euglycemia and hypogly-cemia and did not change significantly(Supplementary Fig. 1C). C3 levels didnotchange during euglycemia in the diabe-tes group but tended to rise at day 7after hypoglycemia (from 996 5 mg/mLat hypoglycemic baseline to 108 6 4mg/mL at day 7).In the nondiabetes group, hsCRP in-

creased similarly at day 1 after euglycemiaand hypoglycemia in both glycemicarms (Supplementary Fig. 1D). In thediabetes group, hsCRP decreased aftereuglycemia at day 7 (log hsCRP D20.116 0.01; P = 0.009 vs. baseline) comparedwith no change after hypoglycemia (loghsCRP D 0.06 6 0.01 at day 7). A sig-nificant interaction was observed be-tween glycemic arm and time (P = 0.04).

In the nondiabetes group, IL-6 washigher at day 1 after euglycemia but didnot change after hypoglycemia (data notshown). In the diabetes group, IL-6 didnot change during the euglycemic armand trended toward an increase in IL-6after hypoglycemia at day 7.

Platelet Reactivity and Activation

Collagen- and ADP-Induced Platelet

Aggregation

In the nondiabetes group, a nonsignif-icant change was found in collagen-induced platelet aggregation duringeuglycemia (Fig. 4A). During hypoglyce-mia, collagen-induced platelet aggre-gation increased significantly morethan during euglycemia, which resolvedat recovery. The overall difference inplatelet aggregation response to colla-gen between arms was significant (P =0.04).

In the diabetes group, collagen-induced platelet aggregation tended todecrease during acute euglycemia(D2156 23 units; P = 0.11 vs. baseline)in contrast to an increase after hypogly-cemia (P , 0.01 between glycemic arms),persisting to the recovery period but notbeyond. A significant interaction wasobserved between time and glycemicarm (P = 0.03).

ADP-induced platelet aggregationdecreased in the diabetes group dur-ing euglycemia compared with an in-crease at the end of hypoglycemia (P =0.03 between arms). In the nondiabetesgroup, secondary rises in platelet aggre-gation induced by both collagen andADP at day 1 occurred after both hy-poglycemia and euglycemia (Fig. 4Aand B).

No significant changes were found inplatelet reactivity to 5-hydroxytryptamineduring euglycemia in both groups (Fig.4C). Trends toward increased plateletreactivity were seen after hypogly-cemia in both groups, resolving at re-covery. In the nondiabetes group, nosignificant changes in platelet activationoccurred during euglycemia. Platelet ac-tivation tended to increase after hypo-glycemia and at day 1, with no overalldifference between arms (Fig. 4D). Inthe diabetes group, platelet activation,as measured by unstimulated P-selectinexpression, decreased after euglycemiaand maximally at day 1 but increasedimmediately afterhypoglycemia (P=0.01for glycemic arm) (Fig. 4D).

CONCLUSIONS

Toour knowledge, this study is thefirst toinvestigate the effects of hypoglycemiaon both cellular and protein arms ofthe thrombosis pathway in the periodbeyond a hypoglycemic challenge. Thenovel findings are that 1) hypoglyce-mia was associated with early and lateprothrombotic changes in the fibrin net-work in the participants with diabetesbut was less evident in healthy controlsubjects, 2) a rise in fibrinogen and C3levels may contribute to late prothrom-botic changes in fibrin network proper-ties, and 3) hypoglycemia enhancedplatelet reactivity in participants withand without diabetes, lasting less than24 h after the event.

The antiaggregatory, anti-inflamma-tory, and profibrinolytic responses toeuglycemic hyperinsulinemia are consis-tent with previous human studies(20,21). In the current work, a strikingobservation was the difference in clotlysis time between euglycemia and hy-poglycemia in the diabetes group. Ini-tially, the decrease in clot lysis timeduring euglycemia may be explained bya reduction in PAI-1. Insulin-mediatedsuppression of fibrinogen synthesis alsomay result in enhanced clot lysis but isunlikely to contribute to early changesgiven the long half-life of the protein,unlike PAI-1. On the other hand, earlyprolongation in clot lysis in the diabetesgroup did not appear to be PAI-1 medi-ated because protein levels were re-duced after hypoglycemic clamps. In aprevious study, PAI-1was decreased dur-ing both hyperinsulinemic hypoglyce-mia and euglycemia in individuals withtype 1 diabetes; however, the authorsfound increased PAI-1 after hypoglyce-mia in healthy control subjects (8).

A novel and intriguing observation isthe persistent effect of hypoglycemia onclot density and impaired fibrinolysis atleast 7 days after hypoglycemia in thediabetes group but not in the nondia-betes group. This finding may be impor-tant clinically, particularly in individualsat higher vascular risk. The mechanismsfor the late impairment in lysis arenot entirely clear, although the rise infibrinogen and complement C3 plasmalevels provide plausible explanations.Increased levels of fibrinogen are asso-ciated with denser clots, as we havepreviously shown (18), whichmay impair

2630 Prothrombotic Effects of Hypoglycemia Diabetes Care Volume 41, December 2018

Figure 4—A–D: Effect of euglycemia and hypoglycemia on platelet reactivity and platelet activation. Data represent the change in value from baseline.The euglycemic arm preceded the hypoglycemic arm. Differences between glycemic arm and interaction between time and glycemic arm by mixedmodelwith repeatedmeasuresareshown.Dataaremean6SE.†P,0.05,††P,0.01euglycemiavs.hypoglycemiaatequivalent timepoints;*P,0.05,**P , 0.01 vs. baseline. 5HT, 5-hydroxytryptamine; AUC, area under the curve; DM, diabetes mellitus; EU, euglycemic; HYPO, hypoglycemic; MFI,median fluorescence intensity.

care.diabetesjournals.org Chow and Associates 2631

fibrinolysis. This finding is supported byan observational study in type 2 diabeteswherein low fasting glucose was associ-ated with denser ex vivo fibrin clotscompared with fasting glucose in thenormal range (22). We have demon-strated that C3 modulates clot lysis in-dependently of PAI-1, and this effect isparticularly pronounced in individualswith diabetes (17,23). Taken together,repeated hypoglycemic episodes possi-bly lead to a state of chronic low-gradeinflammation, resulting in elevated fi-brinogen and C3 levels, which in turncompromise fibrinolysis. However, wecannot exclude the possibility thatother plasma components also mayhave an effect. Joy et al. (9) demon-strated that nitric oxide–mediated en-dothelial dysfunction, which creates aninflammatory environment, is exagger-ated by repeated experimental hypogly-cemia 24 h later, but effects beyond 48 hwere not reported.We also show that hypoglycemia

modulates platelet reactivity in type 2diabetes, probably mediated by cate-cholamine release, but the effects areshort-lived. Our results are consistentwith previous reports of individualswith type 1 diabetes demonstrating in-creased platelet reactivity during acutehypoglycemia (24). The hypoglycemia-mediated increase in platelet reactivitycan be abolished by a-receptor blockade,implicating a2-adrenoreceptors in medi-ating this effect (6). In contrast to fibrinclot studies, we did not observe consistentdifferences between platelet responses toacute hypoglycemia in the groups studied.The current data identify a mechanism

whereby hypoglycemia could contributeto increased CV mortality by opposingthe benefits of intensive glycemic control.In a post hoc analysis of the Normoglyce-mia in Intensive Care Evaluation-SurvivalUsing Glucose Algorithm Regulation(NICE-SUGAR) study, which reportedexcess mortality in patients treated totight glucose levels, the median timeto death as a result of a hypoglycemicevent was 7 days for moderate hypogly-cemia and 8 days for severe hypoglyce-mia (25). The time course matches ourfindings of a worsening atherothrom-botic risk profile 1 week after hypogly-cemia. In the Action in Diabetes andVascular Disease: Preterax and DiamicronMR Controlled Evaluation (ADVANCE)study and Veterans Affairs Diabetes Trial

(VADT), the increased CV risk extendedfor many months after a severe hypogly-cemic event, and it is unlikely that thechanges we observed would persistover this period. On the other hand, lesssevere episodes (equivalent to those in-duced in this study) were not measuredconsistently in these trials. Arguably, theyare occurring more frequently in thosewho experience severe events and, if re-current, might contribute to thromboticevents, particularly in individuals at in-creased CV risk.

Our study had the following relevantlimitations. First, the relatively smallnumber of participants, although ade-quate to demonstrate changes in fibrinclot properties, may have left us short ofstatistical power to identify subtle dif-ferences between the type 2 diabetesand nondiabetes groups. However, re-cruiting individuals into experimentalstudies is challenging, and this studyremains one of the largest to investigateboth participants with type 2 diabetesand control subjects with combinedplatelet and fibrin network analyses. Sec-ond, the experimental model requiressupraphysiological doses of insulin tomaintain stable hypoglycemia higherthan that observed during routine clin-ical care. This might artifactually dimin-ish prothrombotic and proinflammatorychanges during experimental hypoglyce-mia. Third, two participants were receiv-ing aspirin treatment, which can affectcollagen-induced platelet aggregationand clot lysis (26). However, these indi-viduals remained on aspirin throughoutthe study and displayed patterns ofcollagen-induced platelet aggregationsimilar to those of other participants(data not shown). In these studies, weinduced hypoglycemia on two occasionsduring day 0, and we cannot be certainthat similar persistent prothromboticchanges would have been observed aftera single hypoglycemic episode or shorterhypoglycemic episodes. Future studiesshould be designed to address this. Forobvious ethical reasons, we did not inducehypoglycemia in individuals with knownCV disease. It is unknown whether thedeleterious atherothrombotic effects ofhypoglycemia would be exaggerated inpatients with a high CV risk and theprotective effects of insulin attenuatedcompared with our study group.

In conclusion, we have shown thattwo episodes of moderate hypoglycemia

have acute prothrombotic effects in in-dividuals with and without type 2 di-abetes. We also have demonstrated thatthe impact of moderate, short-lived hy-poglycemia is maintained for at least7 days after the event, with adverseeffects on fibrin clot properties, fibrino-lysis, and subclinical inflammation. Theseeffects were more prominent amongindividuals with type 2 diabetes com-pared with control subjects without di-abetes. We have identified potentialmechanisms whereby hypoglycemiacould increase the risk of CV eventsduring and after an episode and so op-pose the benefits of intensive glycemiccontrol. The precise clinical relevanceof these findings remains to be estab-lished by additional studies. Neverthe-less, we believe that clinicians shouldconsider these results when addressingvascular health in patients with type 2diabetes and choose approaches thatminimize hypoglycemia when optimiz-ing glycemic management.

Acknowledgments. The authors thank the Di-abetes Department and Clinical Research Facil-ity in Sheffield Teaching Hospitals for assistanceand all the patients who gave time to this study.The authors also thank Jenny Freeman (Univer-sity of Leeds) and Eric Lau (Chinese Universityof Hong Kong) for statistical advice.Funding. This article is a summary of indepen-dent research funded in part by the NationalInstitute for Health Research (NIHR) Biomedi-cal Research Fellowship, U.K. (BRF-2011-004)and carried out at the NIHR Sheffield ClinicalResearch Facility.The views expressed are those of the authors

and not necessarily those of the National HealthService, NIHR, or the Department of Health.Duality of Interest. R.F.S. reports institutionalresearch grants/support from AstraZenecaand PlaqueTec; consultant fees from Actelion,AstraZeneca, Avacta, Bayer, Bristol-Myers Squibb/Pfizer, Idorsia, Novartis, PlaqueTec, and Throm-boserin; and honoraria from AstraZeneca andBayer. R.A. received honoraria and educationaland research support from Abbott DiabetesCare, AstraZeneca, Novo Nordisk, Eli Lilly, Bayer,Sanofi, MSD, and Boehringer Ingelheim. S.R.H.has served as an advisory board panel memberfor Eli Lilly, Novo Nordisk, Boehringer Ingelheim,Zealand Pharma, UNEEG medical, and TakedaPharmaceuticals and is a member of speakers’bureaus for AstraZeneca, Novo Nordisk, Eli Lilly,andMSD. No other potential conflicts of interestrelevant to this article were reported.AuthorContributions. E.C. designed the study,collected and analyzed the data, and wrote themanuscript. A.I., E.W., and F.P. helped to collectthe data, conducted clotting and platelet assays,and reviewed the manuscript. I.A.M. analyzedthe catecholamine data and reviewed the man-uscript. R.F.S., R.A., and S.R.H. designed the

2632 Prothrombotic Effects of Hypoglycemia Diabetes Care Volume 41, December 2018

study, reviewed the data, and edited and re-drafted the manuscript. E.C. is the guarantor ofthis work and, as such, had full access to all thedata in the study and takes responsibility for theintegrity of the data and the accuracy of the dataanalysis.Prior Presentation. Parts of this study werepresented in abstract form at the 73rd ScientificSessions of the American Diabetes Association,Chicago, IL, 21–25 June 2013.

References1. Patel A, McMahon S, Chalmers J, et al.;ADVANCE Collaborative Group. Intensive bloodglucose control and vascular outcomes in pa-tients with type 2 diabetes. N Engl J Med 2008;358:2560–25722. DuckworthW,AbrairaC,Moritz T, et al.; VADTInvestigators. Glucose control and vascular com-plications in veterans with type 2 diabetes. NEngl J Med 2009;360:129–1393. Gerstein HC, Miller ME, Byington RP, et al.;Action to Control Cardiovascular Risk in DiabetesStudy Group. Effects of intensive glucose low-ering in type 2 diabetes. N Engl J Med 2008;358:2545–25594. Zoungas S, Patel A, Chalmers J, et al.;ADVANCE Collaborative Group. Severe hypogly-cemia and risks of vascular events and death.N Engl J Med 2010;363:1410–14185. Bridge KI, Philippou H, Ariens R. Clot prop-erties and cardiovascular disease. Thromb Hae-most 2014;112:901–9086. Trovati M, Anfossi G, Cavalot F, et al. Studieson mechanisms involved in hypoglycemia-induced platelet activation. Diabetes 1986;35:818–8257. Wright RJ, Newby DE, Stirling D, Ludlam CA,Macdonald IA, Frier BM. Effects of acute insulin-induced hypoglycemia on indices of inflam-mation: putative mechanism for aggravatingvascular disease in diabetes. Diabetes Care 2010;33:1591–1597

8. Gogitidze Joy N, Hedrington MS, Briscoe VJ,Tate DB, Ertl AC, Davis SN. Effects of acute hypogly-cemia on inflammatory and pro-atherothromboticbiomarkers in individuals with type 1 diabetes andhealthy individuals. Diabetes Care 2010;33:1529–15359. Joy NG, Tate DB, Younk LM, Davis SN. Effectsof acute and antecedent hypoglycemia on en-dothelial function and markers of atherothrom-botic balance in healthy humans. Diabetes 2015;64:2571–258010. Grant PJ, Stickland MH, Wiles PG, Davies JA,Wales JK, Prentice CR. Hormonal control ofhaemostasis during hypoglycaemia in diabetesmellitus. Thromb Haemost 1987;57:341–34411. Robinson AM, Parkin HM, Macdonald IA,Tattersall RB. Antecedent hypoglycaemia innon-diabetic subjects reduces the adrenalineresponse for 6 days but does not affect thecatecholamine response to other stimuli. ClinSci (Lond) 1995;89:359–36612. Adler GK, Bonyhay I, Failing H, Waring E,Dotson S, Freeman R. Antecedent hypoglycemiaimpairs autonomic cardiovascular function: im-plications for rigorous glycemic control. Diabetes2009;58:360–36613. Tracey KJ. Physiology and immunology ofthe cholinergic antiinflammatory pathway. JClin Invest 2007;117:289–29614. Chow E, Bernjak A, Walkinshaw E, et al.Cardiac autonomic regulation and repolarizationduring acute experimental hypoglycemia intype 2 diabetes. Diabetes 2017;66:1322–133315. Tesfaye S, BoultonAJ, Dyck PJ, et al.; TorontoDiabetic Neuropathy Expert Group. Diabeticneuropathies: update on definitions, diagnosticcriteria, estimation of severity, and treatments.Diabetes Care 2010;33:2285–229316. Carter AM, Cymbalista CM, Spector TD,Grant PJ; EuroCLOT Investigators. Heritabilityof clot formation, morphology, and lysis: theEuroCLOT study. Arterioscler Thromb Vasc Biol2007;27:2783–2789

17. Hess K, Alzahrani SH,MathaiM, et al. A novelmechanism for hypofibrinolysis in diabetes: therole of complement C3. Diabetologia 2012;55:1103–111318. Hooper JM, Stuijver DJ, Orme SM, et al.Thyroid dysfunction and fibrin network struc-ture: a mechanism for increased thrombotic riskin hyperthyroid individuals. J Clin EndocrinolMetab 2012;97:1463–147319. Bland JM,AltmanDG.Calculatingcorrelationcoefficients with repeated observations: part 1--correlation within subjects. BMJ 1995;310:44620. Chaudhuri A, Janicke D, Wilson MF, et al.Anti-inflammatory and profibrinolytic effect ofinsulin inacuteST-segment-elevationmyocardialinfarction. Circulation 2004;109:849–85421. Dandona P, Ghanim H, Bandyopadhyay A,et al. Insulin suppresses endotoxin-induced ox-idative, nitrosative, and inflammatory stress inhumans. Diabetes Care 2010;33:2416–242322. Gajos G, Konieczynska M, Zalewski J, UndasA. Low fasting glucose is associated with en-hanced thrombin generation and unfavorablefibrin clot properties in type 2 diabetic patientswith high cardiovascular risk. Cardiovasc Diabe-tol 2015;14:4423. Hess K, Alzahrani SH, Price JF, et al. Hypo-fibrinolysis in type 2 diabetes: the role of theinflammatory pathway and complement C3.Diabetologia 2014;57:1737–174124. Dalsgaard-Nielsen J, Madsbad S, Hilsted J.Changes in platelet function, blood coagulationand fibrinolysis during insulin-induced hypogly-caemia in juvenile diabetics and normal subjects.Thromb Haemost 1982;47:254–25825. Finfer S, Liu B, Chittock DR, et al.; NICE-SUGAR Study Investigators. Hypoglycemia andrisk of death in critically ill patients. N Engl J Med2012;367:1108–111826. Ajjan RA, Standeven KF, Khanbhai M, et al.Effects of aspirin on clot structure and fibri-nolysis using a novel in vitro cellular system.Arterioscler Thromb Vasc Biol 2009;29:712–717

care.diabetesjournals.org Chow and Associates 2633