Severe low blood sugar and death in ACCORD

Overview

Bonds DE et al.. BMJ 2010 · PMID 20061358 · DOI 10.1136/bmj.b4909

People who had severe low-sugar episodes died more often, but this did not explain why the intensive group had more deaths.

Does the study support the specific claim? Partly · Methodological quality: moderate

Original publication

Plain-language summary

Scientific in-depth review

1. Full reference

Authors

Bonds DE, Miller ME, Bergenstal RM, Buse JB, Byington RP, Cutler JA, Dudl RJ, Ismail-Beigi F, Kimel AR, Hoogwerf B, Horowitz KR, Savage PJ, Seaquist ER, Simmons DL, Sivitz WI, Speril-Hillen JM, Sweeney ME

Journal

BMJ

Year

2010

Identifier

PMID 20061358 DOI 10.1136/bmj.b4909 · PMC2803744

Verification

verified – PubMed E-utilities efetch, 2026-10-25

Retracted

no

Corrections

none

Extent checked

full text

Conflicts of interest and funding

Funded by NHLBI and other NIH institutes, CDC and General Clinical Research Centers; study drugs/supplies donated by multiple companies (incl. Novo Nordisk, Sanofi-Aventis, Merck, GSK). Several authors report industry research support/consulting; one author (BH) is a full-time Lilly employee.

2. Research question

Link between severe hypoglycaemia and mortality in ACCORD

3. Study design and population

Study design

Cohort study – Retrospective (post hoc) epidemiological analysis of the ACCORD randomised trial (NCT00000620); hypoglycaemia as time-dependent exposure within and across randomised glycaemia arms; blinded adjudication of cause of death and hypoglycaemia involvement.

Population

ACCORD participants with type 2 diabetes, HbA1c 7.5-11%, aged 40-79 with CVD or 55-79 with subclinical atherosclerosis/risk factors; excluded recent/frequent serious hypoglycaemia, BMI >45, creatinine >133 umol/L.

Country

USA and Canada

Period

Data to 10 December 2007 (closure of intensive glycaemia arm; mean follow-up 3.5 years).

Sample size

10,194 of 10,251 ACCORD participants (intensive 5,106; standard 5,088); 451 deaths; hypoglycaemia requiring any assistance (HA) 816 (15.9%) intensive vs 256 (5.0%) standard; requiring medical assistance (HMA) 528 (10.3%) vs 175 (3.4%).

Inclusion and exclusion criteria

As described in the abstract; details not available in abstract.

4. Methods and endpoints

Exposure / intervention

Symptomatic severe hypoglycaemia requiring any assistance (HA) or medical assistance (HMA) (glucose <2.8 mmol/L or prompt recovery with treatment), as time-dependent covariate; also self-monitored values <3.9 mmol/L.

Comparator

Participants without such episodes; intensive vs standard arm among those with episodes.

Endpoints

All-cause and cause-specific mortality; adjudicated hypoglycaemia involvement in death (unlikely/possible/probable/definite).

Statistical methods

Cox models with time-dependent hypoglycaemia covariates, stepwise-selected baseline covariates, arm x hypoglycaemia interaction; Levin's attributable (aetiological) fraction.

5. Actual results

Results

Annual mortality, intensive arm: 2.8% with vs 1.2% without HA (aHR 1.41, 95% CI 1.03-1.93; unadjusted 1.79); standard arm 3.7% vs 1.0% (aHR 2.30, 1.46-3.65; unadjusted 2.93). HMA: intensive aHR 1.28 (0.88-1.85), standard 2.87 (1.73-4.76). Among those with HMA, intensive vs standard HR 0.55 (0.31-0.99); with HA HR 0.74 (0.45-1.23). Among those without HMA, intensive vs standard HR 1.25 (1.03-1.52). Attributable fraction of deaths to HA: 5.9% (intensive) and 5.1% (standard); only about 5 of 57 excess intensive-arm deaths attributable to excess HA. Hypoglycaemia not involved in 389/431 (90.3%) adjudicable deaths, possible in 38 (8.8%), probable in 3 (0.7%), definite in 1. Only 6 of 74 deaths after HA occurred within 30 days of the episode.

Absolute risks

Annual mortality 2.8% vs 1.2% (intensive) and 3.7% vs 1.0% (standard) with vs without HA; overall ACCORD 1.42% vs 1.14% per year (intensive vs standard).

Authors' conclusion

Symptomatic severe hypoglycaemia was associated with increased risk of death within each arm, but among people with hypoglycaemia the risk was lower in the intensive arm; no temporal relation; only one death definitely hypoglycaemia-related; severe hypoglycaemia does not account for the excess mortality of intensive therapy; susceptibility to hypoglycaemia may be a marker of underlying disorder.

6. Strengths

• Large randomised-trial cohort with prospectively, systematically collected hypoglycaemia data
• Blinded central adjudication of cause of death and hypoglycaemia involvement
• Time-dependent exposure modelling and attributable-fraction analysis

7. Weaknesses

• Post hoc observational analysis within an RCT: hypoglycaemia is not randomised; confounding by frailty/comorbidity likely (authors)
• No continuous glucose monitoring; unrecognised hypoglycaemia near death cannot be excluded; glucose missing for 15% of HMA episodes
• Protocol-driven differences between arms (visit frequency, glucose targets, relaxation of targets after 3 HMA) complicate interpretation
• Stepwise covariate selection
• Patients with recent severe hypoglycaemia excluded at entry

8. Risk of bias

Orienting tool

ROBINS-E

Type of assessment

structured assessment based on the full text by a single reviewer (no formal double review)

Full formal assessment

no

Overall risk of bias

some concerns

Rationale

Single-reviewer ROBINS-E-oriented judgement from the full text; formal assessment requires two independent reviewers. The association hypoglycaemia-death is likely confounded (high confounding risk), but the paper's key conclusion (hypoglycaemia does not explain the ACCORD excess mortality) rests on robust adjudication and between-arm comparisons; overall unchanged (some_concerns).

Full-text review (25 October 2026): Exposure definitions (HA/HMA), episode counts by arm, HMA-based HRs, attributable fractions (~5 of 57 excess deaths), adjudication breakdown (90.3% not involved, 1 definite) and temporal analysis added; funding/COI completed (NIH; one Lilly employee); ROBINS-E domains added, overall unchanged; relevance clarified as not supporting mechanistic claims.

9. Assessment of the statistical interpretation

Severe hypoglycaemia marks a 1.4-2.3-fold higher adjusted mortality, stronger in the standard arm where it is more likely a frailty marker; lack of temporal clustering and adjudication (1 definite death of 451) argue against a direct lethal effect explaining ACCORD's excess deaths.

10. Relevance to the specific claim

Supports the hypoglycaemia-mortality association part of C-INS-04; limits causal interpretation; does not support the mechanistic claims C-INS-02, C-INS-09, C-ODD-03.

11. Independent contrary findings

Same paper.

12. Transferability to Germany and Europe

Multinational/US/UK/European data; directly transferable to Germany/EU clinical practice.

13. Overall methodological judgement

Robust evidence that severe hypoglycaemia is a marker of high mortality risk in type 2 diabetes and should be avoided, but that it did not explain ACCORD's excess mortality; does not support mechanistic claims that insulin 'starves' cells or halts metabolism.

14. Open questions

None beyond those stated.

15. Linked fact checks

• Injected insulin halts the Krebs cycle, blocks protein and DNA synthesis, starves the brain of glucose and causes seizures and blood clots. – Studies with mixed or unclear results
• Large trials in type 2 diabetes showed that intensive glucose lowering towards normal HbA1c can be harmful: ACCORD was stopped after intensive therapy raised all-cause mortality by 22%, and severe hypoglycaemia was associated with roughly doubled mortality. – Studies supporting the claim
• Insulin is not needed for glucose to enter cells (the HbA1c inside red blood cells proves it); insulin therefore does not treat diabetes but 'starves' cells, especially the brain, of sugar. – Studies with mixed or unclear results
• Insulin and diabetes drugs pull glucose out of the blood and out of the cells, producing an internal 'famine'. – Studies with mixed or unclear results

Use by the Tayebat archive

Note: according to the archive, the studies were compiled by the archive operators. Whether Dr. Elawady himself cited this study is not established.

What the archive says about the study

Bonds (BMJ 2010): in ACCORD's 10,194 patients, symptomatic severe hypoglycemia was tied to higher mortality in both arms.

A. What was actually studied?

Link between severe hypoglycaemia and mortality in ACCORD

B. What did the authors find?

aHR 1.41 (intensive) and 2.30 (standard) for death after severe hypoglycaemia requiring assistance; among those with hypoglycaemia, mortality lower in intensive arm (HR 0.55 for HMA); only 1 of 451 deaths definitely hypoglycaemia-related; ~5 of 57 excess deaths attributable.

C. What conclusions do the authors themselves draw?

Severe hypoglycaemia was associated with increased death within each arm, but did not appear to account for the mortality difference between arms.

D. What does the archive derive from it?

Narrative: hypoglycaemia 'doubled the mortality' and supports insulin caution.

E. Is this derivation scientifically justified?

Numbers correct; the archive omits the authors' central conclusion that hypoglycaemia did not explain the excess deaths of intensive therapy, that only one death was definitely hypoglycaemia-related, and that hypoglycaemia is likely a marker of vulnerability. It provides no support for claims that insulin halts the Krebs cycle or causes clots.

Adequacy of the interpretation (by the archive)

essentially correct but incomplete