UK database: insulin users vs metformin users

Overview

Currie CJ et al.. J Clin Endocrinol Metab 2013 · PMID 23372169 · DOI 10.1210/jc.2012-3042

People treated with insulin alone had about twice the death rate of metformin users, but they were also sicker.

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

Original publication

Plain-language summary

Scientific in-depth review

1. Full reference

Authors

Currie CJ, Poole CD, Evans M, Peters JR, Morgan CL

Journal

J Clin Endocrinol Metab

Year

2013

Identifier

PMID 23372169 DOI 10.1210/jc.2012-3042 · PMC3612791

Verification

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

Retracted

no

Corrections

none

Extent checked

full text

Conflicts of interest and funding

Supported by a European Foundation for the Study of Diabetes grant for a related study; funder had no role. C.J. Currie is director of, and two co-authors consult for, Pharmatelligence (consultancy funded by pharmaceutical companies); authors report grants/consulting from many manufacturers incl. Sanofi-Aventis, Lilly, Novo Nordisk, BMS, Merck.

2. Research question

Risk of adverse events with five glucose-lowering regimens

3. Study design and population

Study design

Cohort study – Retrospective cohort study in the UK General Practice Research Database (GPRD) comparing five glucose-lowering regimens by exposure period (>=180 days on the same regimen), censoring at regimen switch.

Population

Incident type 2 diabetes patients in UK primary care (mean age 61.9); insulin-monotherapy users had longer diabetes duration (5.2 vs 1.5 years for metformin), higher HbA1c (9.5% vs 8.5%), more prior large-vessel disease (22.3% vs 11.1%), more creatinine >130 umol/L (21.1% vs 2.2%), higher Charlson index (2.4 vs 1.6) and more GP contacts (16.8 vs 10.9/year).

Country

UK

Period

January 2000 - December 2010; mean follow-up per exposure period 2.8 years (insulin monotherapy 3.3 years).

Sample size

84,622 patients; 105,123 exposure periods (metformin 58,532; sulfonylurea 16,218; metformin+sulfonylurea 23,049; insulin monotherapy 3,944; insulin+metformin 3,380); 298,530 person-years.

Inclusion and exclusion criteria

Incident T2DM (>=180-day wash-in after registration), T2DM by decision rules; T1DM excluded; >=180 days on same regimen; for each outcome, patients with prior history of that morbidity excluded (renal analysis: baseline creatinine above normal excluded). Patients could contribute to several cohorts after switching.

4. Methods and endpoints

Exposure / intervention

Insulin monotherapy, insulin+metformin, sulfonylurea monotherapy, metformin+sulfonylurea (no insulin dose analysis).

Comparator

Metformin monotherapy

Endpoints

Primary: first of all-cause mortality, incident cancer or MACE (non-fatal MI, non-fatal stroke, CV death); secondary: components, MI, stroke, eye complications, neuropathy, renal complications (Read codes).

Statistical methods

Cox proportional hazards with a priori covariates (age, sex, SBP, HbA1c, cholesterol, creatinine, BMI, smoking, antihypertensive/lipid-lowering/antiplatelet therapy, diabetes duration, prior cancer, LVD, microvascular disease, GP contacts, Charlson index); multiple imputation; subgroups by HbA1c tertile and Charlson index; direct comparisons with insulin monotherapy as referent.

5. Actual results

Results

Crude primary endpoint per 1,000 person-years: metformin 36.6, sulfonylurea 67.7, insulin monotherapy 58.9, insulin+metformin 30.8, metformin+sulfonylurea 31.4. Adjusted vs metformin: primary endpoint insulin monotherapy aHR 1.808 (95% CI 1.630-2.005), sulfonylurea 1.436 (1.354-1.523), insulin+metformin 1.309 (1.150-1.491). Insulin monotherapy: all-cause mortality 2.197 (1.983-2.434), MACE 1.736 (1.441-2.092), MI 1.954 (1.479-2.583), stroke 1.432 (1.159-1.771), cancer 1.437 (1.234-1.674), renal 3.504 (2.718-4.518), neuropathy 2.146 (1.832-2.514), eye 1.171 (1.057-1.298). Sulfonylurea monotherapy mortality 1.749 (1.643-1.863). Insulin vs metformin mortality across HbA1c/morbidity subgroups 1.897-2.776; primary endpoint 1.469 (0.978-2.206) to 2.644 (1.896-3.687).

Absolute risks

Crude rates per 1,000 person-years: all-cause mortality metformin 14.9, insulin monotherapy 46.0, sulfonylurea 50.7, insulin+metformin 15.7; cancer 15.6 vs 20.3 (insulin monotherapy); MACE 11.6 vs 16.7.

Authors' conclusion

In T2DM, exogenous insulin therapy was associated with increased risk of complications, cancer and all-cause mortality; baseline differences between groups should be considered; confounding by indication is the most relevant limitation; data should raise concerns and the risk-benefit profile of insulin should be reviewed.

6. Strengths

• Large real-world primary-care dataset with ~300,000 person-years
• Pre-specified covariate set incl. HbA1c, BMI, creatinine, comorbidity; multiple imputation
• Multiple comparisons with alternative referents and subgroup analyses

7. Weaknesses

• Severe confounding by indication: insulin is started late in progressive disease; insulin users had 3.5-fold longer diabetes duration, more renal impairment (metformin contraindicated), more vascular disease and comorbidity
• Short follow-up (~3 years) for cancer outcomes (authors)
• Censoring at regimen switch is likely informative (treatment failure leads to switching)
• No dose or insulin-type analysis; no adjustment for time-varying factors
• Outcomes from Read codes without adjudication; possible type 1/type 2 misclassification biased towards insulin monotherapy (authors)
• Large hazard ratios for outcomes such as renal complications (3.5) and neuropathy are typical of reverse causation/channelling
• Authors have extensive industry ties; analysis not pre-registered

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

critical

Rationale

Single-reviewer ROBINS-E-oriented judgement from the full text; formal assessment requires two independent reviewers. Raised from 'high' to 'critical' because the full text documents very large baseline imbalances (diabetes duration 5.2 vs 1.5 years, creatinine >130 umol/L 21.1% vs 2.2%) that make causal inference about insulin itself not credible; the authors themselves identify confounding by indication as the key flaw.

Full-text review (25 October 2026): Cohort sizes per regimen, baseline imbalances, crude event rates per 1,000 PY and subgroup ranges added; censoring-at-switch and Read-code outcome methods documented; funding (EFSD) and extensive author industry COI (Pharmatelligence) added; ROBINS-E domains added, overall high -> critical; methodological_quality low -> very_low; noted that no dose-response was analysed.

9. Assessment of the statistical interpretation

Precise but not interpretable as causal effects: hazard ratios of 2-3.5 for mortality and renal outcomes coincide with large baseline differences in disease stage and renal function. Randomised evidence (ORIGIN, UKPDS) does not show such harms. The paper does not report any insulin dose-response, so it cannot support a '+54% per IU/kg/day' statement.

10. Relevance to the specific claim

Main observational support for C-INS-05.

11. Independent contrary findings

A-INS-07, I-INS-05, I-INS-02.

12. Transferability to Germany and Europe

UK data.

13. Overall methodological judgement

Large observational association between insulin monotherapy and adverse outcomes that is almost certainly dominated by confounding by indication; authors explicitly caution about baseline differences. Not evidence that insulin itself doubles mortality.

14. Open questions

None beyond those stated.

15. Linked fact checks

• Insulin therapy in type 2 diabetes roughly doubles all-cause mortality and raises cancer, renal and cardiovascular events, with a dose-response (each extra IU/kg/day +54% mortality), compared with metformin or sulfonylureas. – 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

Currie (JCEM 2013): in 84,622 UK patients, insulin monotherapy roughly doubled all-cause mortality (HR 2.20) and raised MI, stroke, renal failure and cancer vs metformin.

A. What was actually studied?

Risk of adverse events with five glucose-lowering regimens

B. What did the authors find?

Insulin monotherapy vs metformin: mortality aHR 2.197 (1.983-2.434), primary endpoint 1.808, cancer 1.437, renal 3.504; insulin users were markedly sicker at baseline (longer diabetes, more renal and vascular disease).

C. What conclusions do the authors themselves draw?

Exogenous insulin was associated with increased complications, cancer and death; baseline differences between groups should be considered.

D. What does the archive derive from it?

Used as evidence that insulin itself raises mortality and complications in type 2 diabetes.

E. Is this derivation scientifically justified?

Numbers correct, but the causal wording ('doubled mortality') is not justified: the full text shows large baseline imbalances, short follow-up, censoring at switch and the authors' own statement that confounding by indication is the main limitation; no dose-response was analysed; randomised trials (ORIGIN) do not confirm harm.

Adequacy of the interpretation (by the archive)

partly correct but over-interpreted