· ST-SAF-01 · v2
Borderline: within 4 points of the Established/Likely boundary, so a re-grade could land either side.
Across different study designs and health systems, the evidence points the same way: people taking statins develop new type 2 diabetes somewhat more often than people who do not. A randomised comparison of rosuvastatin against placebo saw more new diabetes in the statin group [2], and four large population studies found the same elevated risk [11][12][14][19]. Short mechanistic trials show statins can nudge up insulin resistance and change how the pancreas releases insulin [9][16], which gives a plausible biological reason. The main weaknesses are that most of the supporting studies are observational, so sicker, higher-risk people may simply be the ones put on statins, and the single placebo-controlled trial in this corpus is only a brief conference abstract [2]. The effect is also not uniform: pravastatin and pitavastatin look neutral or lower-risk [5][10]. The panel treated the direction of the claim as well supported while noting these caveats.
A placebo-controlled randomised comparison reports increased incident T2D with statin (HR 1.25 [95% CI: 1.05-1.49]) [2], and four large independent cohorts show consistently elevated new-onset diabetes with statin exposure: HR 1.18 [14], HR 1.31 in normotensive patients [12], HR 1.46 [19], and raised risk versus pravastatin for atorvastatin (1.22), rosuvastatin (1.18) and simvastatin (1.10) [11]. Mechanistic trials demonstrate the proposed pathway, with atorvastatin raising insulin resistance by 8% and insulin secretion by 9% [9], and statin users showing lower insulin sensitivity and beta-cell function over nine years [16]; the METSIM cohort attributes the 46% excess risk to a 24% fall in insulin sensitivity and 12% fall in secretion [19]. Within-statin data reveal agent heterogeneity, with pitavastatin carrying lower NODM risk than atorvastatin or rosuvastatin (HR 0.72 [95% CI 0.59-0.87]) [10]. Two small trials found no insulin-sensitivity effect, on statin withdrawal [3] and for pravastatin [5]. The contested point: The division is whether the observational core ([11][12][14][16][19]) establishes causation or reflects confounding by indication, since statin recipients are selected for higher metabolic and cardiovascular risk [1]. A second fault line is agent heterogeneity: the unqualified class term masks that pravastatin and pitavastatin appear neutral or lower-risk [5][10][18].
All 21 papers behind this claim with their DOIs and what each does to it, the panel’s arbitration, the evidence gap map and the open research questions. Free, and sent to your inbox.
We’ll also send you other claims from Statins, one at a time, and tell you when a release or a pivotal trial changes a grade. Unsubscribe by replying.
One reviewer's only job is to attack the finding. The challenges below were weighed before the grade was set.
The included causal backbone beyond [2] is observational ([11][12][14][16][19]); confounding by indication is unaddressed because statin recipients are selected for higher metabolic/cardiovascular risk, and covariate adjustment cannot fully remove this.
The only placebo-controlled randomised comparison in the included set, [2], is a conference abstract with limited reported methods and short follow-up; the gold-standard individual-participant meta-analysis ([17]) and the MR/RCT synthesis ([21]) were removed by the study-type filter, leaving thin top-tier RCT evidence.
Where it remains uncertain. The causal backbone beyond [2] is observational and vulnerable to confounding by indication, and the effect is heterogeneous, with pravastatin and pitavastatin appearing neutral or lower-risk [5][10].
Aleatory
Hazard ratios cluster around 1.1-1.5 with wide confidence intervals ([2][11][12][14][19]), reflecting a low per-person probability and random variation in who develops diabetes.
Epistemic
The cohorts ([11][12][14][16][19]) cannot fully separate the drug from the higher metabolic and cardiovascular risk of the people selected to receive it [1].
Inferential
Five studies report insulin resistance, secretion or beta-cell markers ([3][4][5][9][16]) rather than diagnosed T2DM, requiring an inferential leap to the clinical endpoint.
The load-bearing papers from a corpus of 21, each classified by what it does to the claim. Every DOI is taken from the corpus the claim was graded on.
Only included placebo-controlled randomised comparison; direct evidence that statin allocation raises incident T2D.
“Overall, statin vs. placebo increased T2D (HR 1.25 [95% CI: 1.05-1.49])”
Panel auditInternal validity: high · scope of inference: Primary CVD prevention population on high-intensity rosuvastatin.Key limitations: secondary analysis focused on BCAAs · conference abstract · short follow-up (2y, max 5y)
Primary trial evidence for the physiological mechanism linking statins to diabetes risk.
“atorvastatin increased insulin resistance (steady-state plasma glucose) by a median of 8% (P=0.01) and insulin secretion (ISRAUC) by a median of 9% (P<0.001)”
Panel auditInternal validity: medium · scope of inference: Short-term insulin physiology in non-diabetic adults on high-intensity atorvastatin.Key limitations: no control arm · proxy not T2DM onset · 10-week duration
Large real-world cohort showing within-class heterogeneity, qualifying the uniformity of the class effect.
“pitavastatin resulted in a significantly reduced risk of NODM than atorvastatin + rosuvastatin (HR 0.72; 95% CI 0.59-0.87)”
Panel auditInternal validity: medium · scope of inference: Relative new-onset diabetes risk among individual statin agents.Key limitations: no non-statin comparator · retrospective EHR data
The abstract compares genotypes within an all-statin cohort and reports no statin-versus-non-user contrast, so it cannot speak to whether statins increase risk.
Panel auditInternal validity: medium · scope of inference: Genetic modifiers of new-onset T2DM among atorvastatin/rosuvastatin users over four years.Key limitations: all-statin cohort, no non-user comparator · small sample · association lost after correction
4 of 21 papers shown. The full review lists all of them, with the panel’s arbitration →
Claims are graded at each quarterly release, and re-graded between releases only when a pivotal trial lands. Every change is a new version of the claim.
Convergent primary evidence from one RCT and multiple large independent cohorts, supported by a coherent mechanism, places the directional claim at the bottom of the Established band despite observational confounding and a single thin RCT abstract.
Release 2026-q4 re-graded every claim from its corpus under protocol v0.19. An earlier release's grade is shown for the record; it came from an earlier protocol and is not a movement.
Assumptions that carry weight but are not themselves established.
The panel records the gap that would most move the claim, and the study that would close it.
Statin-versus-placebo RCT on diagnosed T2DM. The only randomised placebo comparison on the endpoint is a thin abstract [2], leaving the causal claim resting mainly on observational cohorts vulnerable to confounding by indication. The panel rates this the highest-importance open gap for the claim, and names the study design that would close it: A dedicated placebo-controlled trial, or full-paper individual-participant pooling of existing trials, powered on physician-diagnosed new-onset T2DM with prespecified glycaemic measurement.
Open question: Does statin therapy versus placebo increase physician-diagnosed new-onset T2DM when assessed in full-paper randomised evidence? Why it is still open: The only placebo-controlled randomised comparison in the corpus is a conference abstract with limited methods and short follow-up [2], and stronger syntheses were filtered out.
Priors is a sceptical first reader, not peer review, and not medical advice.
Release 2026-q4 · graded 26 Oct 2026 · protocol v0.19