Skip to content
CholesterolResearch
← All studies
High tierMeta Analysis Of RctsCitation verified

Statins and risk of incident diabetes: a collaborative meta-analysis of randomised statin trials

Naveed Sattar, David Preiss, Heather M Murray, Paul Welsh, Ian Ford - The Lancet, 2010

The defining quantification of the statin-diabetes harm. Pooling 13 statin RCTs, statin therapy was associated with a 9% relative increase in incident diabetes (OR 1.09). In absolute terms the harm is small - about one extra case for every 255 patients treated for four years - which the authors judged low against the cardiovascular benefit.

Key findings

Effect measures

  • Odds Ratio: 1.09 (incident diabetes)95% CI 1.02-1.17
  • Number Needed To Harm: 255 treated for 4 years per extra diabetes case95% CI 150-852

Why this evidence tier (High)

Risk of bias:
Collaborative meta-analysis of randomized trials; low risk of bias for this harm estimate.
Precision:
Very large (91,140); precise estimate with little heterogeneity.
Directness:
Directly measures incident diabetes across statin RCTs.
Consistency:
Low heterogeneity; later work shows the effect is dose-dependent.
Funding / COI:
The meta-analysis itself received no external funding; several authors report ties to statin manufacturers, and the pooled trials were largely industry-sponsored.

High certainty that statins modestly raise incident diabetes; the absolute effect is small and concentrated in the already-predisposed.

Population:
91,140 participants across 13 randomised statin endpoint trials (each >1000 patients, follow-up >1 year), excluding transplant and dialysis patients; mean follow-up 4 years.
Conflicts of interest:
No funding for the meta-analysis itself. Several authors report honoraria, consultancy, or grants from statin manufacturers; the underlying trials were largely industry-sponsored.
Funding:
None (the meta-analysis received no external funding, per the abstract).

Limitations

  • The source says "incident diabetes" generically; type is not specified, though these populations are predominantly type 2.
  • Trial-level diabetes ascertainment was not the trials primary purpose.

How this study is used