Amount of insulin and cancer deaths

Overview

Bowker SL et al.. Diabetologia 2010 · PMID 20407744 · DOI 10.1007/s00125-010-1750-8

More insulin prescriptions went with more cancer deaths, which may reflect that very ill people need more insulin.

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

Original publication

Plain-language summary

Scientific in-depth review

1. Full reference

Authors

Bowker SL, Yasui Y, Veugelers P, Johnson JA

Journal

Diabetologia

Year

2010

Identifier

PMID 20407744 DOI 10.1007/s00125-010-1750-8

Verification

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

Retracted

no

Corrections

none

Extent checked

full text

Conflicts of interest and funding

Funded in part by Canadian Institutes of Health Research (MOP-82737; team grant OTG-88588); personal awards from AHFMR, Canadian Diabetes Association, Canada Research Chairs. Authors declare no duality of interest.

2. Research question

Glucose-lowering agents and cancer mortality

3. Study design and population

Study design

Cohort study – Population-based retrospective cohort (Saskatchewan Health administrative databases) of new users of oral glucose-lowering agents, with time-varying exposure Cox models; secondary new-insulin-user sub-cohort.

Population

New users of metformin or sulfonylureas aged >=30 years with >=1 year prior drug coverage and >=1 year of drug exposure after index; gestational diabetes excluded. Insulin used was human insulin (pre-analogue era).

Country

Canada

Period

Index 1991-1996; follow-up to 31 December 1999 (mean 5.4 years, max 9).

Sample size

10,309 new oral users (6,969 metformin alone or with sulfonylurea; 3,340 sulfonylurea monotherapy); 1,443 subsequent new insulin users; 407 cancer deaths (84 in insulin sub-cohort). 1,963 with <1 year exposure excluded.

Inclusion and exclusion criteria

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

4. Methods and endpoints

Exposure / intervention

Time-varying metformin and sulfonylurea use (1-year windows); insulin as 'average cumulative insulin dispensations per year' (<3, 3-11, >=12); insulin type and dose not recorded.

Comparator

Sulfonylurea monotherapy (for metformin); no insulin (primary) or <3 dispensations/year (insulin sub-cohort).

Endpoints

Cancer death (vital statistics); cancer incidence not studied.

Statistical methods

Cox regression with time-varying exposures, adjusted for age, sex and chronic disease score (CDS, computed over the full follow-up) and, in insulin sub-cohort, duration of prior oral therapy.

5. Actual results

Results

Cancer mortality 6.3 (metformin) vs 9.7 (sulfonylurea) per 1,000 person-years; metformin vs sulfonylurea aHR 0.80 (95% CI 0.65-0.98). Insulin vs none: <3 dispensations/year aHR 2.22 (0.99-5.00; 6 deaths), 3-11: 3.33 (2.26-4.89; 29 deaths), >=12: 6.40 (4.69-8.73; 49 deaths). New insulin user cohort (vs <3/year): 3-11 aHR 1.35 (0.56-3.28), >=12 aHR 2.47 (1.05-5.84); metformin vs sulfonylurea in this cohort 1.09 (0.66-1.80).

Absolute risks

Cancer mortality 6.3 vs 9.7 per 1,000 person-years (metformin vs sulfonylurea); rates by insulin category not estimable (authors).

Authors' conclusion

Supports reduced cancer mortality with metformin vs sulfonylurea and provides 'preliminary evidence' of a gradient of cumulative insulin dispensations and cancer mortality; insulin more likely to accelerate than cause cancer death; caution with early/aggressive insulin until more evidence; confirmation in larger cohorts with cancer incidence needed.

6. Strengths

• New-user (inception) cohort design
• Time-varying exposure reduces immortal-time bias of earlier time-fixed analysis
• Population-based linked databases; no declared COI

7. Weaknesses

• No data on BMI, smoking, HbA1c, alcohol (authors); adjustment only for age, sex and a drug-based comorbidity score
• Insulin 'dose' is number of dispensations, which rises with illness, hospital discharge and terminal care - strong reverse causation/protopathic bias for cancer death
• Exposure assessed in the same time window as the event (authors), so terminal-illness insulin use is counted
• Comorbidity score computed over the full follow-up (uses post-baseline information)
• Few events in insulin categories (6/29/49 deaths in primary; 84 in insulin cohort)
• Outcome is cancer mortality, not incidence; no non-diabetic control
• Implausibly large HR 6.40 for a drug effect within ~2.5 years of insulin use

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

high

Rationale

Single-reviewer ROBINS-E-oriented judgement from the full text; formal assessment requires two independent reviewers. Overall unchanged (high); full text confirms concurrent-window exposure and dispensation-based dose, making reverse causation the most likely explanation of the gradient.

Full-text review (25 October 2026): Cohort composition, death counts per category, secondary new-insulin-user cohort results (1.35, 2.47) and cancer mortality rates added; exposure definition (dispensation count, concurrent window) and covariate limits documented; funding (CIHR) and no-COI statement added; ROBINS-E domains added, overall unchanged (high).

9. Assessment of the statistical interpretation

The steep gradient (HR 2.2 to 6.4) is attenuated to 1.35-2.47 in the new-insulin-user cohort and is based on few deaths. Insulin dispensations measured in the same window as death are expected to increase near the end of life, so the association is not interpretable as a dose effect of insulin.

10. Relevance to the specific claim

Association for C-INS-05.

11. Independent contrary findings

I-INS-05.

12. Transferability to Germany and Europe

Canadian data.

13. Overall methodological judgement

Hypothesis-generating association between frequency of insulin dispensing and cancer death, highly prone to reverse causation and confounding; the authors call it preliminary. It does not address all-cause mortality or a per-IU/kg dose-response.

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

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

Listed in the archive bibliography for 'insulin-injections' without its own interpretation; used as general support for the claim that insulin intensification harms.

A. What was actually studied?

Glucose-lowering agents and cancer mortality

B. What did the authors find?

Metformin vs sulfonylurea cancer death aHR 0.80; insulin dispensation gradient aHR 2.22/3.33/6.40 vs no insulin; attenuated in new-insulin-user cohort (1.35, 2.47); 407 cancer deaths in total.

C. What conclusions do the authors themselves draw?

Supports decreased cancer risk with metformin and a gradient between insulin dispensations and cancer mortality.

D. What does the archive derive from it?

Support for insulin dose-cancer link.

E. Is this derivation scientifically justified?

Not justified as causal: the authors label the gradient preliminary, lacked BMI/smoking/HbA1c and measured insulin dispensations concurrently with the outcome; the study covers cancer mortality only, not the all-cause mortality dose-response cited in C-INS-05.

Adequacy of the interpretation (by the archive)

partly correct but over-interpreted