OUP user menu

★ fast track ★

Does hypoglycaemia increase the risk of cardiovascular events? A report from the ORIGIN trial

DOI: http://dx.doi.org/10.1093/eurheartj/eht332 3137-3144 First published online: 2 September 2013


Aims Hypoglycaemia caused by glucose-lowering therapy has been linked to cardiovascular (CV) events. The ORIGIN trial provides an opportunity to further assess this relationship.

Methods and results A total of 12 537 participants with dysglycaemia and high CV-risk were randomized to basal insulin glargine titrated to a fasting glucose of ≤5.3 mmol/L (95 mg/dL) or standard glycaemic care. Non-severe hypoglycaemia was defined as symptoms confirmed by glucose ≤54 mg/dL and severe hypoglycaemia as a requirement for assistance or glucose ≤36 mg/dL. Outcomes were: (i) the composite of CV death, non-fatal myocardial infarction or stroke; (ii) mortality; (iii) CV mortality; and (iv) arrhythmic death. Hazards were estimated before and after adjustment for a hypoglycaemia propensity score. During a median of 6.2 years (IQR: 5.8–6.7), non-severe hypoglycaemic episodes occurred in 41.7 and 14.4% glargine and standard group participants, respectively, while severe episodes occurred in 5.7 and 1.8%, respectively. Non-severe hypoglycaemia was not associated with any outcome following adjustment. Conversely, severe hypoglycaemia was associated with a greater risk for the primary outcome (HR: 1.58; 95% CI: 1.24–2.02, P < 0.001), mortality (HR: 1.74; 95% CI: 1.39–2.19, P < 0.001), CV death (HR: 1.71; 95% CI: 1.27–2.30, P < 0.001) and arrhythmic death (HR: 1.77; 95% CI: 1.17–2.67, P = 0.007). Similar findings were noted for severe nocturnal hypoglycaemia for the primary outcome and mortality. The severe hypoglycaemia hazard for all four outcomes was higher with standard care than with insulin glargine.

Conclusion Severe hypoglycaemia is associated with an increased risk for CV outcomes in people at high CV risk and dysglycaemia. Although allocation to insulin glargine vs. standard care was associated with an increased risk of severe and non-severe hypoglycaemia, the relative risk of CV outcomes with hypoglycaemia was lower with insulin glargine-based glucose-lowering therapy than with the standard glycaemic control. Trial Registration (ORIGIN ClinicalTrials.gov number NCT00069784).

  • Hypoglycaemia
  • Prognosis
  • Glucose-lowering treatment
  • Insulin
  • Glargine

See page 3102 for the editorial comment on this article (doi:10.1093/eurheartj/eht374)


Hypoglycaemia is a common adverse effect of glucose-lowering therapy with insulin, sulfonylureas, and meglitinides. In addition to its acute clinical effects, experimental studies suggest that hypoglycaemia may directly or indirectly induce abnormal myocardial repolarization, QT-prolongation, ventricular arrhythmias, and myocardial ischaemia probably mediated by adrenergic surges or increased adrenergic tone.13 Concerns that these and other effects may cause cardiovascular events are supported by epidemiological studies and exploratory analyses of trials linking hypoglycaemic events with cardiovascular and other long-term health outcomes.4,5 However, whether this relationship is related to hypoglycaemia itself or is influenced by factors linked to the propensity for hypoglycaemia remains unresolved.

The recently completed Outcomes Reduction with an Initial Glargine Intervention (ORIGIN) trial of insulin-mediated normoglycaemia vs. standard care with oral glucose-lowering agents in people with early type 2 diabetes and dysglycaemia at high risk for cardiovascular outcomes prospectively measured both non-severe and severe episodes of hypoglycaemia.6 Data from this trial therefore provide a unique opportunity to assess the relationship between hypoglycaemia and cardiovascular events in this well-defined population. They also permit analyses of whether any such relationship differs in people allocated to glucose lowering with basal insulin glargine vs. standard glycaemic care with oral agents.


The design and main results of the ORIGIN trial have already been published.68 Briefly, 12 537 people aged 50 years of age or older with impaired fasting glucose, impaired glucose tolerance, or early type 2 diabetes, who all had additional cardiovascular risk factors, were recruited between 2003 and 2005 in 40 participating countries. Eligible patients had either: (i) IFG, IGT, or newly detected diabetes [i.e. an FPG ≥6.1 mmol/L (110 mg/dL) or a 2 h plasma glucose ≥7.8 mmol/L (140 mg/dL) after a 75 g oral glucose load]; or (ii) established type 2 diabetes on stable therapy with 0 or 1 oral agent for ≥3 months. These criteria were used so that the locally measured glycated haemoglobin level of participants with established diabetes would be low enough to allow investigators to add or adjust oral agents to manage diabetes during the trial without the need for insulin in those participants allocated to the control group.

Participants were randomly allocated to either the addition of basal insulin glargine titrated to a fasting plasma glucose ≤5.3 mmol/L (95 mg/dL) or to standard glycaemic treatment according to local guidelines. At the same time, they were randomly allocated to 1 g of an n-3 fatty acid supplement or placebo according to a 2 × 2 factorial design. During a median follow-up period of 6.2 years (IQR: 5.8–6.7) both interventions had a neutral effect on the primary and secondary clinical cardiovascular and microvascular outcomes. The main endpoint for the insulin glargine trial was a composite of non-fatal myocardial infarction, non-fatal stroke, or death from cardiovascular causes.


The present report comprises all ORIGIN participants (mean age, 63.5 years; 35% female) including 6264 allocated to basal insulin glargine and 6273 to standard care.6 As previously published, baseline characteristics were similar for the two allocated groups. Approximately 60% had a prior cardiovascular event, 80% had a prior diagnosis of diabetes (of whom 72% were taking oral glucose-lowering agents), 6% had newly detected type 2 diabetes and 12% had impaired glucose tolerance or impaired fasting glucose. The median baseline HbA1c was 6.4% (IQR: 5.8–7.2) and fasting plasma glucose was 6.9 mmol/L (125 mg/dL) (IQR: 109–148).6

Ascertainment and definitions of hypoglycaemia

All the participants were provided with glucose meters and diaries and were instructed to record episodes of hypoglycaemia along with concomitant capillary glucose levels. Study personnel reviewed these diaries, explicitly asked about hypoglycaemic episodes, and recorded their occurrence on specific case report forms at every visit.

Non-severe hypoglycaemia was defined as an event with symptoms consistent with hypoglycaemia confirmed by a concomitant glucose reading ≤3.0 mmol/L (≤54 mg/dL). Severe hypoglycaemia was defined as a symptomatic hypoglycaemic event in which the participant required assistance of another person and there was: (i) prompt recovery after oral carbohydrate, i.v. glucose, or glucagon administration; and/or (ii) a documented self-measured or laboratory plasma glucose level ≤2.0 mmol/L (≤36 mg/dL). Nocturnal severe hypoglycaemia was defined as a severe hypoglycaemic episode occurring between midnight and 06:00 a.m.


Four outcomes were analysed for this report: (i) the primary composite outcome of cardiovascular death (defined as any death for which no non-cardiovascular cause could be identified), or non-fatal myocardial infarction (diagnosed on the basis of clinical presentation, elevated cardiac markers, and/or new electrocardiographic changes), or stroke (diagnosed on the basis of clinical presentation and imaging); (ii) mortality; (iii) cardiovascular mortality; and (iv) arrhythmic death (comprising sudden unexpected death, death from documented arrhythmia, unwitnessed death, and resuscitated cardiac arrest). An independent committee unaware of the allocated group of the affected participant adjudicated all events.

Statistical analysis

Continuous variables are summarized as means with standard deviations or as medians with inter-quartile ranges and compared using t-tests or Kruskall–Wallis tests, respectively. Categorical variables summarized as the number with percentages are compared using χ2 tests.

Propensity scores were used to account for confounding variables when estimating the hazard of an episode of non-severe or severe hypoglycaemia on the occurrence of clinical outcomes. Two propensity scores were calculated for each participant: one for non-severe hypoglycaemia and one for severe hypoglycaemia. Each score was calculated using a logistic regression model in which the dependent variable was at least one episode of non-severe hypoglycaemia or at least one episode of severe hypoglycaemia, respectively. Independent variables added to each model were age, female gender, ethnicity (white, black, south Asian, other Asian, Latin, other), education (≤8, 9–12, or >12 years), a prior cardiovascular event, hypertension, depression, current smoking, alcohol intake of >2 drinks/week, an albumin creatinine ratio ≥30 mg/g, treatment with metformin, sulfonylurea, statin, ACE/ARB, beta-blocker, thiazides and platelet stabilizing drugs, BMI, waist–hip ratio, HbA1c, fasting plasma glucose, total, HDL- and LDL-cholesterol, triglyceride, serum creatinine, mini-mental status examination, and prior diabetes mellitus.

The hazard ratios relating hypoglycaemic episodes to clinical outcomes occurring at any time after such episodes were estimated from the coefficient for either non-severe hypoglycaemia, severe hypoglycaemia, or nocturnal severe hypoglycaemia which were included as time-varying covariates in Cox regression models. These models also included allocation to glargine/standard care, allocation to omega-3-FA/placebo, diabetes status, and a history of a cardiovascular event before randomization as independent variables. Additional models included the appropriate propensity score (i.e. non-severe or severe) as independent variables. These additional models were also used to analyse the relationship between an episode of hypoglycaemia and a clinical outcome occurring within either 1 or 7 days of that episode. The relationship between number of episodes of hypoglycaemia and outcomes was assessed by calculating separate estimates of the hazard ratios for 1 or more, 2 or more, 3 or more, 4 or more, and 5 or more episodes of non-severe or severe hypoglycaemia. To determine whether the hazard increased with the number of episodes, a variable representing the number of episodes was added to the Cox models and the significance of the coefficient was tested.

Treatment group allocation

The effect of the randomly assigned treatment group on the relationship between hypoglycaemic episodes and outcome was assessed by including an interaction term (treatment allocation × hypoglycaemic episode) in the Cox regression models. A similar approach was used to assess the effect of prior diabetes. Any effect of sulfonylurea use prior to the cardiovascular outcome was adjusted for by including it in the regression. Data were analysed with the use of the SAS software (version 9.1 for Solaris). The nominal level of significance for all analyses was P < 0.05.


Baseline characteristics for participants with and without hypoglycaemia are presented in Table 1. During the median follow-up of 6.2 years (inter-quartile range, 5.8–6.7), 3518 participants had at least one episode of hypoglycaemia. Of these 2614 (74.3%) occurred in the glargine group and 904 (25.7%) in the standard group. Of the 472 participants with at least one episode of severe hypoglycaemia, 76.1% (359) occurred in the glargine group and 23.9% (113) in the standard group with an estimated annual incidence of 0.9 and 0.3%, respectively. Compared with participants who did not experience any hypoglycaemic episodes, those who experienced an episode of non-severe hypoglycaemia were younger and those who experienced a severe episode were older. For both types of episodes, affected participants were more likely to be men than women, had a higher baseline fasting plasma glucose and HbA1c, and were more often treated with a sulfonylurea (Table 1).

View this table:
Table 1

Clinical characteristics of the participants by the presence or not of non-severe and severe hypoglycaemic episodes

VariableNon-severe hypoglycaemiaP-valueSevere hypoglycaemiaP-value
Yes (n = 3518)No (n = 9019)Yes (n = 472)No (n = 12 065)
Demographic and clinical characteristics
 Age (years; mean ± SD)63.1 ± 7.563.7 ± 8.0<0.00165.9 ± 8.363.5 ± 7.8<0.001
 Female sex1176 (33)3210 (36)0.022144 (31)4242 (35)0.038
 Prior cardiovascular events/disease
  MI, stroke, and revascularization2079 (59.1)5299 (58.8)0.726301 (63.8)7077 (58.7)0.027
  Stroke457 (13.0)1199 (13.3)0.65263 (13.4)1593 (13.2)0.928
  Myocardial infarction2282 (64.9)5822 (64.6)0.741302 (64.0)7802 (64.7)0.761
  Hypertension2784 (79.1)7179 (79.6)0.564397 (84.1)9566 (79.3)0.011
 Current smoker463 (13.2)1089 (12.1)0.09754 (11.4)1498 (12.4)0.528
 Alcohol (>2/week)841 (23.9)2007 (22.3)0.047129 (27.3)2719 (22.5)0.015
 Albumin creatinine ratio (≥30 mg/g)3299 (93.8)8632 (95.7)<0.001436 (92.4)11 495 (95.3)0.004
Glycaemic characteristics
 Prior diabetes3296 (94.0)7785 (86.3)<0.001436 (92.4)10 645 (88.2)0.006
 Fasting plasma glucose (mg/dL; mean ± SD)138 ± 38130 ± 35<0.001135 ± 38131 ± 360.049
 HbA1c (%; mean ± SD)6.7 ± 1.06.5 ± 0.9<0.0016.6 ± 1.06.5 ± 1.00.004
Glycaemic drugs
 Insulin glargine allocation2614 (74.3)3650 (40.5)<0.001359 (76.1)5905 (48.9)<0.001
 Metformin941 (26.8)2494 (27.7)0.308125 (26.5)3310 (27.4)0.649
 Sulfonylurea1462 (41.6)2249 (24.9)<0.001181 (38.4)3530 (29.3)<0.001
 Other glucose-lowering drugs106 (3.0)244 (2.7)0.34720 (4.2)330 (2.7)0.052
Non-glycaemic cardiovascular risk factors
 Blood pressure (mm Hg; mean ± SD)
  Systolic147 ± 23145 ± 21<0.001150 ± 25146 ± 22<0.001
  Diastolic84 ± 1284 ± 120.16284 ± 1384 ± 120.809
 Body mass index [weight/height (m2); mean ± SD]30 ± 530 ± 5<0.00129 ± 530 ± 50.003
 Waist-to-hip ratio
  Men0.99 ± 0.090.98 ± 0.090.6540.98 ± 0.100.99 ± 0.090.359
  Women0.90 ± 0.090.90 ± 0.090.6210.91 ± 0.090.90 ± 0.090.260
 Cholesterol (mg/dL; mean ± SD)
  Total190 ± 48189 ± 460.143189 ± 48189 ± 460.995
  LDL113 ± 40112 ± 400.452113 ± 42112 ± 400.721
  HDL46 ± 1346 ± 120.76548 ± 1446 ± 120.009
 Triglycerides (mg/dL; median IQR)140 (97–196)142 (98–195)0.954133 (90–181)142 (98–196)<0.001
Creatinine (mg/dL; mean ± SD)1.0 ± 0.31.0 ± 0.20.0281.1 ± 0.31.0 ± 0.2<0.001
 Albumin–creatinine ratio, median (inter-quartile range)0.62 (0.28–2.50)0.57 (0.27–1.97)0.0040.72 (0.32–3.64)0.58 (0.28–2.09)<0.001
Cardiovascular drugs
 Statin1828 (52.0)4912 (54.5)0.012258 (54.7)6482 (53.7)0.689
 Thiazide610 (17.3)1761 (19.5)0.00588 (18.6)2283 (18.9)0.880
 Beta-blocker1806 (51.3)4792 (53.1)0.070255 (54.0)6343 (52.6)0.535
 ACE/ARB2389 (67.9)6292 (69.8)0.043334 (70.8)8347 (69.2)0.466
 Antiplatelet2411 (68.5)6255 (69.4)0.372346 (73.3)8320 (69.0)0.045
Median N Hypo (non-severe)0.34 (0.17–0.84)N/AN/AN/A
Median N Hypo (severe)N/AN/A0.17 (0.15–0.26)N/A
Propensity score (non-severe)a0.320 (0.109)0.264 (0.101)<0.0010.306 (0.108)0.279 (0.106)<0.001
Propensity score (severe)a0.041 (0.025)0.036 (0.022)<0.0010.051 (0.029)0.037 (0.022)<0.001
  • Data presented are n (%) if not stated otherwise.

  • aPropensity scores for non-severe and severe hypoglycaemia were calculated as described in the ‘Statistical analysis’ section.

Non-severe hypoglycaemia and outcomes

The crude incidence of outcomes in people who had 1, 2, 3, 4, or 5 or more non-severe hypoglycaemic episodes was lower than the incidence in people without severe hypoglycaemia (Supplementary material online, Table S1A). This is reflected in Supplementary material online, Table S1C which shows that people who had 1, 2, 3, 4, or 5 or more episodes of non-severe hypoglycaemia had a significantly lower hazard of outcomes than people who did not have this number of episodes after adjustment for the relevant propensity score. Moreover, when the number of hypoglycaemic episodes was modelled as a continuous variable, there was evidence of a slightly lower hazard with higher numbers of episodes (P < 0.001).

When all non-severe hypoglycaemic episodes were included within the model as a time-varying covariate (which accounts for both the numbers of episodes and the date during the study at which they occurred relative to the date of the outcome), they were associated with an increased risk of death (HR: 1.21; 95% CI: 1.08–1.35) and cardiovascular death (HR: 1.16; 95% CI: 1.00–1.34) but not the primary outcome or arrhythmic death in minimally adjusted models. These estimates became non-significant when the propensity score for non-severe hypoglycaemia was added to these models (Table 2). The pattern was similar for outcomes occurring within 1 or 7 days following an episode. The effect of non-severe hypoglycaemia on outcomes did not differ in participants with or without prior diabetes (interaction P > 0.05).

View this table:
Table 2

The risk for the different outcomes attributable to all non-severe hypoglycaemic episodes and to all severe hypoglycaemic episodes before and after applying the propensity scores adjusting for confounding factors (see the ‘Statistical analysis’ section)

OutcomeUnadjusted hazardP-valueAdjusted hazard with propensity scoreP-value
Non-severe hypoglycaemia
 CV death or non-fatal MI or stroke1.10 (0.98–1.23)0.1151.00 (0.88–1.12)0.947
 Mortality1.21 (1.08–1.35)<0.0011.12 (0.99–1.26)0.066
 Cardiovascular death1.16 (1.00–1.34)0.0491.03 (0.88–1.20)0.688
 Arrhythmic death1.19 (0.97–1.47)0.0911.10 (0.88–1.36)0.403
Severe hypoglycaemia
 CV death or non-fatal MI or stroke1.77 (1.39–2.25)<0.0011.58 (1.24–2.02)<0.001
 Total mortality2.05 (1.65–2.55)<0.0011.74 (1.39–2.19)<0.001
 Cardiovascular death2.02 (1.52–2.69)<0.0011.71 (1.27–2.30)<0.001
 Arrhythmic death2.14 (1.43–3.18)<0.0011.77 (1.17–2.67)0.007
Severe nocturnal hypoglycaemia
 CV death or non-fatal MI or stroke1.88 (1.18–3.00)0.0081.64 (1.01–2.65)0.044
 Total mortality1.95 (1.25–3.04)0.0031.64 (1.04–2.58)0.033
 Cardiovascular death1.99 (1.13–3.53)0.0191.61 (0.89–2.93)0.116
 Arrhythmic death2.04 (0.91–4.57)0.0831.79 (0.80–4.02)0.155
  • Data presented are hazard ratios (HR) and 95% confidence intervals (CI). (CV, cardiovascular; MI, myocardial infarction).

Severe hypoglycaemia and outcomes

The crude incidence of outcomes in people who had 1, 2, 3, 4, or 5 or more severe hypoglycaemic episodes was higher than the incidence in people without severe hypoglycaemia (Supplementary material online, Table S1B). After adjustment for the relevant propensity score the hazard of 1, 2, 3, 4, or 5 or more episodes of severe hypoglycaemia generally did not differ from 1 (Supplementary material online, Table S1C) and there was no evidence of a trend or dose–response (P > 0.1).

When all severe hypoglycaemic episodes were included within the model as a time-varying covariate, episodes of severe hypoglycaemia increased the risk of all four outcomes by >50% even after adjusting for the severe hypoglycaemia propensity score (Table 2). A similar pattern was seen for outcomes that occurred within 1 or 7 days following an episode and did not differ according to baseline diabetes status. Similar patterns were also noted for nocturnal severe hypoglycaemia.

Treatment group allocation and outcomes

Insulin glargine vs. standard care

The numbers and crude incidence of outcomes in people who had hypoglycaemic episodes (of any kind) in the glargine group were lower than in people who did not have hypoglycaemic episodes. Conversely for the standard care group the crude incidence was lower for non-severe hypoglycaemia but higher for severe and severe nocturnal hypoglycaemia (Table 3). Thus, although more people in the glargine vs. standard care group had severe hypoglycaemia, people who had such episodes in the glargine group had a lower incidence of outcomes than people who had such episodes in the standard care group.

View this table:
Table 3

Number and percentage of participants with and without hypoglycaemia who had an outcome by treatment group

Glargine groupStandard care
≥1 Episode of hypoglycaemiaNo hypoglycaemia≥1 Episode of hypoglycaemiaNo hypoglycaemia
Non-severe episodes
 No. of participants261436509045369
 CV death or non-fatal MI or stroke318 (12.2)723 (19.8)100 (11.1)913 (17.0)
 Mortality329 (12.6)622 (17.0)129 (14.3)836 (15.6)
 Cardiovascular death199 (7.6)381 (10.4)64 (7.1)512 (9.5)
 Arrhythmic death103 (3.9)197 (5.4)31 (3.4)253 (4.7)
Severe episodes
 No. of participants35959051136160
 CV death or non-fatal MI or stroke50 (13.9)991 (16.8)21 (18.6)992 (16.1)
 Total mortality.52 (14.5)899 (15.1)34 (30.1)931 (15.1)
 Cardiovascular death32 (8.9)548 (9.3)18 (15.9)558 (9.1)
 Arrhythmic death15 (4.2)285 (4.8)11 (9.7)273 (4.4)
Severe nocturnal episodes
 No. of participants1006164186225
 CV death or non-fatal MI or stroke14 (14.0)1027 (16.7)4 (22.2)1009 (16.1)
 Total mortality17 (17.0)934 (15.2)3 (16.7)962 (15.4)
 Cardiovascular death10 (10.0)570 (9.2)2 (11.1)574 (9.2)
 Arrhythmic death5 (5.0)295 (4.8)1 (5.6)283 (4.5)
  • All cells expressed as n (%); CV, cardiovascular; MI, myocardial infarction.

To determine whether treatment allocation influenced the relationship between hypoglycaemia and clinical outcomes and to account for confounders, hypoglycaemic episodes were included in the adjusted Cox models as a time-varying covariate and an interaction term representing allocation × hypoglycaemic episode was also included. As noted in Table 4, there was no evidence that treatment allocation affected the relationship between non-severe hypoglycaemia (P for interaction all P> 0.4) or nocturnal severe hypoglycaemia (P for interaction all P > 0.2), and the four outcomes that were analysed. Conversely, the hazard of all four outcomes following severe hypoglycaemic episodes did differ based on treatment allocation. Thus, for the primary composite outcome, the hazards associated with severe hypoglycaemia in people allocated to glargine and standard care were 1.38 (95% CI: 1.03, 1.86) and 2.39 (95% CI: 1.55, 3.70) respectively (P for interaction P = 0.047). The standard group participants who experienced severe hypoglycaemic episodes were 1.70 times (95% CI: 1.01, 2.87) more likely to have a subsequent primary composite outcome than glargine participants. Greater interaction effects were noted for the other three outcomes, which were two to three times more likely to be experienced by participants allocated to the standard group vs. the glargine group (Table 4). Adjustment for sulfonylurea use prior to the cardiovascular outcome did not substantively affect the estimates (Table 4).

View this table:
Table 4

The risk for the different outcomes by glucose-lowering strategy, insulin glargine or standard care, following adjustment for confounding factors according to the propensity scores (see the ‘Statistical analysis’ section)

OutcomeNon-severe hypoglycaemiaP-valueSevere hypoglycaemiaP-valueNocturnal severe hypoglycaemiaP-value
CV death or non-fatal MI or stroke
 Glargine1.01 (0.88–1.17)1.38 (1.03–1.85)1.45 (0.84–2.51)
 Standard care0.95 (0.76–1.18)2.39 (1.55–3.70)2.94 (1.90–7.89)
 Standard care vs. Glarginea0.93 (0.72–1.20)0.5891.70 (1.01–2.87)0.0471.91 (0.62–5.90)0.258
 Standard care vs. Glargineb1.13 (0.88–1.45)0.3441.69 (1.00–2.85)0.0491.89 (0.61–5.83)0.267
Total mortality
 Glargine1.09 (0.94–1.26)1.34 (1.00–1.79)1.68 (1.02–2.76)
 Standard care1.18 (0.97–1.45)3.13 (2.20–4.46)1.55 (0.49–4.79)
 Standard care vs. Glarginea1.10 (0.87–1.40)0.4102.31 (1.47–3.64)<0.0010.88 (0.25–3.02)0.834
 Standard care vs. Glargineb1.33 (1.06–1.68)0.0142.26 (1.44–3.56)<0.0010.87 (0.25–3.00)0.824
Cardiovascular death
 Glargine1.09 (0.90–1.31)1.37 (0.94–2.00)1.60 (0.82–3.09)
 Standard care0.95 (0.72–1.25)2.89 (1.80–4.65)1.81 (0.45–7.32)
 Standard care vs. Glarginea0.89 (0.65–1.22)0.4712.09 (1.15–3.82)0.0161.08 (0.23–5.06)0.918
 Standard care vs. Glargineb1.09 (0.80–1.49)0.5812.03 (1.11–3.71)0.0211.06 (0.23–4.95)0.939
Arrhythmic death
 Glargine1.18 (0.91–1.53)1.24 (0.71–2.16)1.81 (0.74–4.39)
 Standard care0.97 (0.65–1.43)3.66 (1.99–6.76)1.74 (0.24–12.5)
 Standard care vs. Glarginea0.86 (0.55–1.35)0.5182.94 (1.29–6.70)0.0100.96 (0.11–8.34)0.973
 Standard care vs. Glargineb1.08 (0.69–1.67)0.7452.79 (1.23–6.35)0.0140.95 (0.11–8.18)0.959
  • Data presented are hazard ratios (HR) and 95% confidence intervals (CI) (CV, cardiovascular; MI, myocardial infarction).

  • aThis row shows the relative increase in the hazard of the predefined outcomes following the indicated hypoglycaemic episodes in the standard care group vs. the glargine-mediated normoglycaemia group.

  • bThis row shows the estimate after also adjusting for use of sulfonylurea at the visit prior to the cardiovascular outcome.

Omega-3-fatty acid vs. placebo group

There was no significant interaction between allocation to omega n-3 fatty acid and outcome with respect to any of the measures of hypoglycaemia (not shown).


This analysis of data collected during the ORIGIN trial shows that both severe hypoglycaemia and nocturnal severe hypoglycaemia independently predicted cardiovascular events and mortality in people having cardiovascular risk factors and early type 2 diabetes or with dysglycaemia. As reported previously, participants allocated to insulin glargine vs. standard care experienced a higher incidence of severe and non-severe hypoglycaemia. Despite this difference, the present analysis also shows that the relationship between severe hypoglycaemia and these outcomes was substantially higher in individuals allocated to standard glycaemic control vs. those who targeted normal glucose levels with insulin glargine.

Hypoglycaemia induces transient cardiac stress due to an enhanced adrenergic tone that increases heart rate, blood pressure, myocardial contractility, and cardiac output.3 These effects plus increases in blood coagulability and viscosity suggest that hypoglycaemia may promote myocardial ischaemia and other serious cardiovascular outcomes, especially in the elderly or otherwise vulnerable people.911 The current finding that severe hypoglycaemia and nocturnal severe hypoglycaemia are both risk factors for cardiovascular outcomes after adjustment for the potential confounders that were included in the propensity score is consistent with similar findings from other trials of a glucose-lowering intervention.911 The fact that the adjusted hazard of nocturnal hypoglycaemia did not quite reach statistical significance for two of these outcomes is most likely due to the fewer numbers of these outcomes. This is of interest in light of reports attributing sudden death while asleep (so-called dead-in-bed syndrome) to hypoglycaemia in people with type 1 diabetes.3

Explanations for the observed relationship between severe hypoglycaemia and cardiovascular outcomes include the pathophysiological mechanisms noted above. However, it is also possible that people who are prone to hypoglycaemic episodes have some other attribute that puts them at risk for cardiovascular outcomes.4,5,12 Indeed the finding of a substantially stronger relationship between severe hypoglycaemia and outcomes in the standard care group, within which guideline suggested but not normal fasting glucose levels were being sought, than in the insulin glargine-mediated normoglycaemic group supports the possibility that severe hypoglycaemia may be a marker rather than an accelerator of future cardiovascular outcomes. If this assumption were true, a stronger relationship to cardiovascular outcomes would indeed be expected when severe hypoglycaemia occurs in response to standard approaches to glucose lowering than when it appears in response to targeting lower indeed even, as in the insulin glargine arm of ORIGIN, normal glucose levels. This is because people with a lower threshold for severe hypoglycaemia (i.e. those who develop it when targeting higher glucose levels) may also share other risk factors for cardiovascular outcomes. In support of this possibility, the following arguments can be made: first, although previous trials have linked severe hypoglycaemia with cardiovascular prognosis they have not shown that this is related to the intensity of glucose lowering;4 second meta-analyses of trials comparing more vs. less intensive glycaemic control suggested a decrease in major cardiovascular events in the intensive groups despite an increased number of severe hypoglycaemic episodes;13,14 and third, a similar finding was reported in the Action to Control Cardiovascular Disease in Diabetes (ACCORD) trial, in which people allocated to intensive glucose lowering who had a severe hypoglycaemic event had a lower risk of death than people in the standard arm with a severe hypoglycaemic event.15 The fact that we observed such an effect within the ORIGIN trial within which glucose lowering was mediated by insulin glargine (as opposed to the menu of drugs used in ACCORD) provides further support for this possibility.

Other alternative or supplementary explanations are possible. One is the much higher use of sulfonylureas in the standard arm (Table 1). This may be particularly relevant in light of the small between-group difference in HbA1c noted in the trial. In addition to promoting hypoglycaemia, sulfonylureas stimulate insulin secretion from the pancreatic beta cells by binding to ATP-dependent potassium channels.16 Such channels are located in many tissues, among them the heart17 and they are involved in myocardial adaptation to ischaemia. The use of sulfonylurea may therefore worsen the consequences of any ischaemic episode. Another is the possibility that the greater frequency of non-severe hypoglycaemia in the glargine arm might have induced a greater tolerance for subsequent severe events in the glargine group, and this may have accounted for lower absolute incidence of events (and hazard) in people who had severe hypoglycaemic episodes in the glargine vs. standard group.

Strengths of the study include the high contrast in insulin use between the two groups that was achieved and maintained throughout the trial, the independent adjudication of outcomes, the fact that both severe and non-severe hypoglycaemic episodes were collected prospectively from the beginning of the trial and the fact that criteria for non-severe and severe hypoglycaemia were chosen to maximize the specificity of the diagnosis of hypoglycaemia. Moreover, the randomized study design facilitates the comparison of hypoglycaemic events between the two treatment strategies. Limitations include the fact that hypoglycaemia was based on interrogation of participants and their logbooks and not on independent measures of glucose. Participants were carefully instructed to record the occurrence of hypoglycaemic symptoms and to measure glucose levels at the same time. However, the open design of the study and the self-reported nature of episodes cannot exclude the possibility of differential ascertainment of episodes by the treatment group. The fact that the study promoted self-glucose monitoring more in the insulin glargine than in the standard care group as a means of achieving glycaemic goals may have reduced ascertainment of hypoglycaemia in the standard care group. Alternatively, this may have led to an overestimate of the prevalence of hypoglycaemia in the insulin glargine arm. Another limitation relates to the number of severe hypoglycaemic episodes overall and within each treatment group which limits the power to explore ‘dose–response’ relationships. Finally, the ORIGIN participants had fairly low HbA1c levels at enrolment. Thus, these findings may not apply to patients with substantially higher HbA1c levels.

In summary, these analyses confirm the relationship between severe hypoglycaemia and cardiovascular outcomes in people with dysglycaemia at high cardiovascular risk. When viewed in the light of evidence from other outcome trials, they are consistent with the possibility that the observed association may be due to confounding by unmeasured risk factors for cardiovascular outcomes. They also provide further support for the conclusion that targeting normoglycaemia by means of insulin glargine had a neutral effect on cardiovascular outcomes despite a higher number of hypoglycaemic episodes.


The ORIGIN trial was funded by Sanofi while Pronova BioPharma, Norway, provided n-3 fatty acid supplements and placebo. In addition this particular part of ORIGIN received funding from the Swedish Heart-Lung Foundation. The international steering committee designed and conducted the trial, and all data were collected and analysed by the ORIGIN Project Office located at the Population Health Research Institute in Hamilton, Ontario Canada. The current analyses were developed, analysed, and interpreted by the writing committee who made all final decisions regarding content and who made the decision to submit the paper for publication.

Conflict of interest: L.G.M. reports research grants from The Swedish Heart-Lung Foundation, The Swedish Diabetes Foundation, MSD, Bayer AG, Sanofi-Aventis and lecture honoraria from MSD, Sanofi-Aventis, Novartis, Bayer-Schering, Astra Zeneca, Roche and Lilly. L.R. reports grants from Swedish Heart-Lung Foundation, AFA insurance, Swedish Diabetes Association, County Council Stockholm, Roche consulting honoraria from Sanofi-Aventis, Roche, Bristol-Myers-Squibb and AstraZeneca; lecture honoraria from Sanofi-Aventis, Roche, Bayer, Bristol-Myers-Squibb and AstraZeneca. M.R. reports honoraria for consulting from Sanofi-Aventis, Lilly, Amylin, Valeritas, and Elcelyx; and research grant support from Sanofi-Aventis, Lilly, and Amylin. These potential conflicts of interest have been reviewed and managed by Oregon Health & Science University. J.P. reports honoraria from Sanofi-Aventis for serving on advisory committees; and research funding directly sent to the university from Sanofi-Aventis. J.R. reports consulting honoraria from Roche, Sanofi-Aventis, Novo Nordisk, Eli Lilly, MannKind, Glaxo-Smith-Kline, Takeda, Merck, Daiichi Sankyo, Johnson & Johnson, Novartis, Boehringer Ingelheim, Halozyme, Intarcia and Lexicon; and received grants/research support from Merck, Pfizer, Sanofi-Aventis, Novo Nordisk, Roche, Bristol-Myers-Squibb, Eli Lilly, Glaxo-Smith-Kline, Takeda, Novartis, AstraZeneca, Amylin, Johnson & Johnson, Janssen, Daiichi Sankyo, MannKind, Boehringer Ingelheim, Intarcia and Lexicon. R.D. reports consulting honoraria from Merck Sharp & Dohme; lecture honoraria from Sanofi-Aventis, Bristol-Myers-Squibb; funding for research is given to his group from Sanofi-Aventis. S.Y. reports research grants to the institution and honoraria and related travel for lectures for Sanofi-Aventis, Glaxo-Smith-Klein, Novartis, Bristol-Myers-Squibb, AstraZeneca, Bayer and Boehringer Ingelheim. H.G. reports consulting honoraria from Sanofi-Aventis, Lilly, Roche, Novo Nordisk, Bayer, Glaxo-Smith-Kline, Novartis, Bristol-Myers-Squibb, and AstraZeneca; lecture honoraria from Sanofi-Aventis and Bayer; grants from Takeda, and funds given to his institution for research or educational initiatives from Sanofi-Aventis, Lilly, Novo Nordisk, Boehringer Ingelheim, Bristol-Myers-Squibb, and AstraZeneca.


The members of the writing committee are as follows: Linda G. Mellbin, MD and Lars Rydén, MD, The Department of Medicine Solna, Karolinska Institutet, Stockholm, Sweden; Matthew C. Riddle, Oregon Health and Science University, Portland; Jeffrey Probstfield, MD, University of Washington, Seattle, USA; Julio Rosenstock, MD, Dallas Diabetes and Endocrine Center at Medical City and University of Texas Southwestern Medical Center, Dallas, TX, USA; Rafael Díaz, Estudios Clinicos Latino America, Rosario, Argentina; Salim Yusuf, D.Phil., and Hertzel C. Gerstein, MD, the Department of Medicine and Population Health Research Institute, McMaster University and Hamilton Health Sciences, Hamilton, ON, Canada.


  • Writing committee for this paper is listed in the Appendix.


View Abstract