Even if there were unknown factors influencing both statin use and diabetes risk, they would need to be more than five times stronger than any known risk factor to completely explain why statin users develop diabetes more often.
Claim Context
The E-value analysis in this study indicates that an unmeasured confounder would need to have a risk ratio greater than 5.93 to fully explain the observed association between statin use and new-onset diabetes, suggesting that the association is unlikely to be entirely due to unmeasured bias.
“The E-values for the point estimate and lower confidence bound for NODM were 5.93 and 1.52, respectively.”
Evidence from Studies
No evidence studies found yet.
What Would Prove This
Per GRADE and EBM methodology, here is what ideal scientific evidence would look like to definitively prove or disprove this claim, ordered from strongest to weakest.
A meta-analysis of E-values from multiple studies could determine whether this level of robustness is consistent across populations and adjustment models.
A systematic review pooling E-values from all published observational studies on statin use and new-onset diabetes, stratified by population, adjustment variables, and statin intensity, to assess consistency of confounding resistance.
An RCT would eliminate unmeasured confounding entirely, making E-value analysis unnecessary, as randomization balances all known and unknown confounders.
A double-blind RCT of 5000 adults aged 50–70 with cardiovascular risk factors, randomized to statin or placebo, with annual HbA1c and metabolic measurements, to determine whether statin use increases diabetes incidence.
A prospective cohort with comprehensive measurement of potential confounders could replicate the E-value analysis and test whether additional variables reduce the association.
A prospective cohort of 10,000 Emirati adults with cardiovascular risk factors, followed for 15 years with annual data on diet, physical activity, sleep, stress, and genetic markers, to recalculate E-values after full adjustment.
A case-control study with detailed confounder data could estimate the E-value for the association, but is less reliable due to recall bias.
A case-control study of 1000 incident diabetes cases and 1000 controls, with detailed historical data on statin use and potential confounders, to calculate E-values for the adjusted odds ratio.
A cross-sectional survey could estimate an E-value for prevalence, but cannot assess temporal sequence or causality.
A national survey measuring statin use, HbA1c, and potential confounders simultaneously in 20,000 adults to estimate an E-value for the association between statin use and diabetes prevalence.