Blinded trials do not support most side effects listed on statin labels

The list of possible side effects on a medicine's label is not, as most readers assume, a list of effects the medicine has been shown to cause. It is a list of events reported during development and surveillance, assembled from sources that often include non-randomised and non-blinded studies — study designs in which a patient's expectation of harm and the reporting of harm are not separable. A meta-analysis published in The Lancet takes the statin labels and tests them against the only evidence that can distinguish the two.

How the list was tested

The researchers first built the list itself, searching an electronic medicines compendium for the undesirable-effect terms listed in the Summaries of Product Characteristics for five statins: atorvastatin, fluvastatin, pravastatin, rosuvastatin and simvastatin [s1].

They then assembled individual participant data from randomised trials meeting three conditions: at least 1,000 participants, a scheduled treatment period of at least two years, and a double-blind comparison of statin against placebo, or of a more intensive against a less intensive statin regimen [s1]. Nineteen trials compared statin with placebo, covering 123,940 participants with a median follow-up of 4.5 years (IQR 3.1–5.4) [s1]. Rate ratios were calculated with statistical significance assessed after controlling the false discovery rate at 5% [s1] — a necessary correction when testing dozens of outcomes at once, because at conventional thresholds a handful would come up positive by chance alone.

What survived

Beyond the previously reported effects on muscle outcomes and diabetes, 66 further undesirable outcomes had been attributed to statins. Four were significant after false-discovery-rate correction [s1]:

• Abnormal liver transaminases: 783 participants (0.30% per annum) on statin versus 556 (0.22% per annum) on placebo, rate ratio 1.41 (95% CI 1.26–1.57) [s1].
• Other liver function test abnormalities: 651 (0.25% per annum) versus 518 (0.20% per annum), RR 1.26 (1.12–1.41). Combined, the liver function test abnormalities represent an absolute annual excess of 0.13% [s1].
• Urinary composition alteration: 556 (0.21% per annum) versus 472 (0.18% per annum), RR 1.18 (1.04–1.33) [s1].
• Oedema: 3,495 (1.38% per annum) versus 3,299 (1.31% per annum), RR 1.07 (1.02–1.12) [s1].

The absolute numbers matter more than the ratios. An annual excess of 0.13% for liver enzyme abnormalities means roughly one additional case per 770 people per year, and the outcome is a laboratory value, not an illness.

The four trials comparing more intensive with less intensive regimens found significant excesses for abnormal transaminases and other liver function abnormalities — supporting a dose-dependent effect — but no significant excess for urinary composition alteration or oedema [s1]. That internal split is informative: a genuine drug effect would generally be expected to strengthen with dose.

What did not survive

The outcomes that did not show a causal signal in blinded data include the ones most often cited by patients who stop taking statins: cognitive impairment, depression, sleep disturbance and peripheral neuropathy [s1].

The authors' conclusion is that adverse event data from blinded randomised trials do not support causal relationships between statin therapy and most of the conditions listed in product labels as potential undesirable effects, and that such labelling and other official health information sources should be revised so patients and doctors can make appropriately informed decisions [s1].

The journal published two accompanying pieces the same fortnight — a linked comment arguing that product labels downplay the safety of statin therapy on the evidence from randomised trials [s2], and an editorial framed around statin safety warnings that outlive the evidence for them [s3]. The titles convey the position; the substance rests on the meta-analysis.

The limits

Blinded trials are the right instrument for separating drug effect from expectation, but they have known blind spots. Trial populations are healthier than general prescribing populations, and follow-up here had a median of 4.5 years [s1] — long by trial standards, short compared with the decades many people take a statin. Rare events are also hard to detect even in 123,940 people, which is one reason surveillance systems exist alongside trials in the first place.

The finding does not say that people who report symptoms on a statin are imagining them. It says that in blinded comparison, those symptoms occurred at similar rates whether or not the tablet contained a statin — which is a statement about attribution, not about whether the experience is real.

A separate window on the same question

A study published in Atherosclerosis the day before offers a different vantage: 15 years of clinical practice in children with heterozygous familial hypercholesterolaemia, for whom statins are the first pharmacological step [s4]. Among 696 children (329 girls, 47.3%) with a median age at statin initiation of 11.0 years (IQR 8.3–14.0), statin-associated symptoms were reported in about 1 in 8.5 children at the first assessment and about 1 in 8.2 at the second and third [s4].

Most symptoms were transient and mild, and the most frequent were muscle symptoms without clinically significant creatine kinase elevation, defined as no more than three times the upper limit of normal [s4]. No cases of rhabdomyolysis or other adverse events requiring hospital admission were reported [s4]. Thirteen children (1.9%) were statin intolerant, discontinuing permanently because of persistent symptoms [s4].

Reported symptoms are common; discontinuation because of them is not. That gap is the same one the Lancet meta-analysis measures with a control arm.

What to watch

The practical question is whether regulators revise the Summaries of Product Characteristics. Labels are not easily changed, and a warning removed is politically harder than a warning added. Whether the European and UK medicines regulators act on this analysis — and how quickly — is the concrete thing to follow.

The meta-analysis was funded by the British Heart Foundation, the UK Medical Research Council and the Australian National Health and Medical Research Council [s1].

This article describes research findings and is not guidance about taking or stopping any medicine.

Sources

• [s1] Assessment of adverse effects attributed to statin therapy in product labels: a meta-analysis of double-blind randomised controlled trials. The Lancet, 5 February 2026. https://doi.org/10.1016/S0140-6736(25)01578-8
• [s2] Product labels downplay the safety of statin therapy: evidence from randomised controlled trials. The Lancet, 5 February 2026. https://doi.org/10.1016/S0140-6736(25)02013-6
• [s3] Statin safety: when warnings outlive the evidence. The Lancet, 13 February 2026. https://doi.org/10.1016/S0140-6736(26)00303-X
• [s4] Statin-associated symptoms and statin intolerance in children with familial hypercholesterolemia: insights from 15 years of clinical practice. Atherosclerosis, 4 February 2026. https://doi.org/10.1016/j.atherosclerosis.2026.120659