Statin-related side effects: the recent Lancet publication is biased toward false-negatives

Dear DataMethods Users,

A recent Lancet meta-analysis assessed whether adverse events listed in statin product labels are causally attributable to statin therapy, using double-blind randomized trials and false discovery rate control across multiple outcomes. The authors concluded that the evidence does not support causal associations for most labeled adverse effects, a conclusion that was rapidly echoed in headlines such as “Statins do not cause the majority of side effects listed in package leaflets”.

However, in our new letter we argue that this conclusion arises from the combination of multiple, compounding sources of bias that systematically favor false-negative findings in safety assessment. These include stringent multiplicity control, threshold-based overinterpretation of statistically “non-significant” results, dichotomous readings of interval estimates, and reliance on randomized trial settings with limited real-world transportability together with intention-to-treat dilution under non-adherence.

Because “non-significant” does not mean “no causal effect”, we call for an interpretation framework centered on effect sizes and overall uncertainty, along with sensitivity analyses across alternative error trade-offs, risk strata, and dosages. Importantly, the consequences of inferential errors are not symmetric: false-positive signals primarily impose costs on researchers and regulators through additional scrutiny, whereas false-negative conclusions disproportionately affect the most vulnerable stakeholders, namely patients.

We look forward to your thoughts and welcome further discussion on this issue.

Rovetta et al., 2026, letter on ‘Assessment of adverse effects attributed to statin therapy in product labels’, v1.0.pdf (144.5 KB)