Sulfonylurea vs metformin as first drug

Overview

Roumie CL et al.. Ann Intern Med 2012 · PMID 23128859 · DOI 10.7326/0003-4819-157-9-201211060-00003

Starting a sulfonylurea instead of metformin was linked to about 1-4 extra heart events or deaths per 1,000 people per year.

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

Original publication

Plain-language summary

Scientific in-depth review

1. Full reference

Authors

Roumie CL, Hung AM, Greevy RA, Grijalva CG, Liu X, Murff HJ, Elasy TA, Griffin MR

Journal

Ann Intern Med

Year

2012

Identifier

PMID 23128859 DOI 10.7326/0003-4819-157-9-201211060-00003 · PMC4667563

Verification

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

Retracted

no

Corrections

none

Extent checked

full text

Conflicts of interest and funding

Funded by the US Agency for Healthcare Research and Quality (DEcIDE contract 290-05-0042); VA Career Development Awards; NIH/NCRR CTSA. Authors report no disclosures.

2. Research question

Sulfonylurea vs metformin monotherapy on CVD events/death

3. Study design and population

Study design

Cohort study – Retrospective new-user cohort (national VHA databases linked to Medicare/Medicaid), sulfonylurea vs metformin monotherapy initiators; on-treatment (persistence-required) primary analysis; multivariable Cox with multiple imputation; 1:1 propensity-score-matched (incl. facility) secondary analysis; quantitative bias analysis. Full text is the PMC author manuscript.

Population

Veterans >=18 years initiating metformin or sulfonylurea (glyburide or glipizide) after >=365 days without antidiabetic drugs; excluded heart failure, cancer, HIV, transplant, end-stage kidney/liver disease, respiratory failure, cocaine use, creatinine >=1.5 mg/dL. 97% male, 75% white, median age 62; median HbA1c 7.0-7.3%.

Country

USA

Period

Incident prescriptions 1 October 2001 - 30 September 2008; median follow-up 0.78 years (metformin) and 0.61 years (sulfonylurea).

Sample size

253,690 incident episodes (155,025 metformin, 98,665 sulfonylurea; 5% contributed >1 episode); propensity-matched 80,648 vs 80,648.

Inclusion and exclusion criteria

New users with regular VHA care; censoring at switch/addition of another drug, 90 days without drug, creatinine >=1.5 mg/dL, 181 days without VHA contact, or study end.

4. Methods and endpoints

Exposure / intervention

Sulfonylurea monotherapy (55% glyburide, 45% glipizide).

Comparator

Metformin

Endpoints

Hospitalisation for AMI or stroke, or death

Statistical methods

Cox regression with robust variance, adjusted for demographics, HbA1c, BMI, LDL, creatinine, BP, medications, utilisation, comorbidities (third-degree polynomials), secular trend; multiple imputation; PS-matched analysis; persistence-not-required (ITT-like) sensitivity; complete-case analysis; subgroups by CVD history, age, BMI, albuminuria; unmeasured confounder analysis.

5. Actual results

Results

Composite AMI/stroke/death: 1,844 vs 1,871 events; crude 18.2 vs 10.4 per 1,000 person-years; aHR 1.21 (95% CI 1.13-1.30); glyburide 1.26 (1.16-1.37), glipizide 1.15 (1.06-1.26). AMI/stroke only: 13.5 vs 8.2 per 1,000 PY, aHR 1.16 (1.06-1.25). PS-matched: 15.2 vs 13.0 per 1,000 PY, aHR 1.15 (1.07-1.25); CV events aHR 1.13 (1.03-1.23). Persistence not required: aHR 1.21 (1.16-1.27). No prior CVD: aHR 1.16 (1.06-1.29), ~1 excess event per 1,000 PY; prior CVD: aHR 1.25 (1.13-1.39 in text; abstract states 1.13-1.55), ~4 excess per 1,000 PY. Microalbuminuria subgroup aHR 1.13 (0.74-1.73). An unmeasured confounder with HR 1.25 would need to be >=53% more prevalent among sulfonylurea users to explain the result.

Absolute risks

Crude 18.2 vs 10.4 per 1,000 person-years; authors estimate excess of 1 (no CVD history) and 4 (CVD history) events per 1,000 person-years. (An adjusted rate difference of 2.2 per 1,000 PY cited from the PubMed abstract is not stated in this author manuscript.)

Authors' conclusion

Initiating sulfonylureas rather than metformin was associated with a modest but clinically important 21% increased hazard of AMI, stroke or death (excess 1-4 events per 1,000 person-years), supporting metformin as first-line therapy; residual confounding cannot be ruled out.

6. Strengths

• Very large active-comparator new-user cohort
• Adjustment for clinical measures (HbA1c, BMI, creatinine, BP, LDL) rarely available in claims studies
• Consistent results across PS matching, ITT-like and complete-case analyses; quantitative bias analysis
• Public funding; no COI

7. Weaknesses

• Observational; residual confounding by indication (sulfonylurea users older, more loop diuretics/digoxin, lower BMI) despite adjustment
• Very short median follow-up (~0.6-0.8 years) with heavy censoring (66-73% stopped therapy) - possibly informative
• Baseline-only covariates; no time-varying adjustment
• 97% male veterans; non-VHA events may be missed
• Glyburide (higher hypoglycaemia risk) over-represented; results not generalisable to gliclazide
• Minor abstract-text discrepancy in a subgroup CI

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 (author manuscript); formal assessment requires two independent reviewers. Overall unchanged (some_concerns).

Full-text review (25 October 2026): Exclusion criteria, follow-up (median 0.6-0.8 years), censoring patterns, event counts, CV-only, PS-matched (1.15) and ITT-like results, subgroup excess (1 and 4 per 1,000 PY) and bias analysis added; absolute risk corrected to the manuscript's 1-4 excess per 1,000 PY (2.2 figure not found in author manuscript); ROBINS-E domains added, overall unchanged.

9. Assessment of the statistical interpretation

Modest relative excess (aHR 1.15-1.21) robust across analyses, translating into a small absolute excess (1-4 per 1,000 person-years); short on-treatment follow-up; comparison is versus metformin, so the result may partly reflect a metformin benefit rather than sulfonylurea harm (authors).

10. Relevance to the specific claim

Supports C-ODD-11.

11. Independent contrary findings

I-INS-01, A-INS-16, I-INS-02.

12. Transferability to Germany and Europe

US veterans.

13. Overall methodological judgement

Credible, well-adjusted association of sulfonylurea (vs metformin) initiation with a small absolute excess of CV events/death; not proof of causation; supports metformin as first-line choice rather than abandoning glucose-lowering treatment.

14. Open questions

None beyond those stated.

15. Linked fact checks

• Every major class of glucose-lowering drug (metformin, sulfonylureas, thiazolidinediones, DPP-4 inhibitors, SGLT2 inhibitors, GLP-1 agonists) carries a documented signal of serious harm in the peer-reviewed literature. – Studies supporting the claim
• Sulfonylureas cause more myocardial infarctions, strokes and deaths than metformin (21% higher in 253,690 veterans). – Studies supporting the claim

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

Veterans cohort of 253,690 patients: sulfonylurea therapy carried a 21% higher hazard of MI, stroke, or death versus metformin.

A. What was actually studied?

Sulfonylurea vs metformin monotherapy on CVD events/death

B. What did the authors find?

aHR 1.21 (1.13-1.30) for AMI/stroke/death; PS-matched 1.15; excess 1-4 events per 1,000 person-years; median follow-up under 1 year.

C. What conclusions do the authors themselves draw?

Sulfonylurea vs metformin initiation associated with increased CVD events or death.

D. What does the archive derive from it?

Used as evidence of sulfonylurea harm.

E. Is this derivation scientifically justified?

Numbers correct; the archive presents the association as causation and omits the small absolute excess, the authors' acknowledgement of possible residual confounding and that metformin benefit could explain part of the difference.

Adequacy of the interpretation (by the archive)

essentially correct but incomplete