High cholesterol is a common cardiovascular risk factor in the USA and internationally.1 Statins have been among the most prescribed class of drugs with an estimated 92 million patients on statin therapy in 2019 in the USA alone.2 Statins have been shown to have generally good safety with comparable efficacy between men and women in terms of lipid-lowering effect and reduced onset of atherosclerotic cardiovascular disease. Despite this, a number of sex differences in the utilization of statins have been observed ranging from differences in prescribing rates and statin dose to safety profile and long-term adverse effects.3–7 Optimization of lipid-lowering therapy must therefore consider efficacy and safety in choosing agent, dose, and perhaps even class of medication in the context of patient’s sex.

Sex differences in statin-induced diabetes mellitus (DM) were first suggested by Culver, et al in the Women’s Health Initiative, where the OR of developing DM across all statins was~1.7.8 In comparison, randomized clinical trials involving statins comprised predominately of men showed lower ORs ranging from 0.95 to 1.14.9 The mechanisms of statin-induced DM are unclear but may include decreased insulin production by pancreatic β-cells, decreased production of ubiquinone, decreased glucose transporter 4 (GLUT4) in adipocytes, or inhibition of glucose-mediated insulin release in cells.10

Awareness of DM as a potential consequence of statin use is growing and is now included in patient-facing educational resources from agencies like the Centers for Disease Control.11 Despite the growing acceptance of DM as an adverse side effect of statins, there is relatively little data directly comparing risk between men and women. Here, we summarize a pharmacovigilance analysis that identifies signals of disproportionate reporting (SDR) of sex differences of statin-associated DM in the Food and Drug Administration’s (FDA) Adverse Event Reporting System (FAERS). We hypothesized that statin-associated DM is reported disproportionately more in women than in men in post-marketing adverse drug event (ADE) surveillance representing an important sex difference in the safety profile of this commonly used class of medications.

All data are fully de-identified by the FDA and are available for unrestricted public download. After review with the Colorado Multi-Institutional Review Board, it was determined that our analysis therefore did not require ethics approval under the National Institutes of Health Human Subjects Research Exemption 4.

Patients and the public were involved in this study through voluntary, spontaneous submission of ADEs to the FDA, either directly or via the drug manufacturer. Patients and providers can currently report ADEs via the MedWatch website (https://www.accessdata.fda.gov/scripts/medwatch/). There was no direct patient/public contact as part of this study, and all data were fully de-identified prior to download.

We obtained publicly available data from FAERS (https://open.fda.gov/data/faers/). Sex is not available in the FAERS data prior to 1997.12 We therefore analyzed ADE reports received from January 1997 through December 2023. ADEs in FAERS have two separate identifiers. The PRIMARYID identifier reflects a specific report submitted, for example, through MedWatch. The CASEID identifier denotes a unique clinical event, which may be described by multiple individual reports from different sources. To avoid redundant counting, we generated results using unique case identifiers (CASEID).

FAERS ADE reports include a list of all drugs present in each ADE, which may include trade names, combination drugs, and obsolete or foreign names. Therefore, all FAERS cases were matched with distinct pharmacologic components using a multitiered algorithm of whole and partial string matching as well as limited manual matching by the authors. Drugs@FDA was used as the reference for drug brand names and active ingredients as described in FAERS documentation.13 Using this approach, more than 95% of all FAERS reports were matched to at least one active ingredient from Drugs@FDA. Unmatched names (eg, “HC”, “cranberry”, “BLU-U Blue light photodynamic therapy illuminator”) were excluded from this analysis. The algorithms used to match verbatim drug names are described in the online supplement, and the verbatim drug names matched to 3-hydroxy-3-methyl-glutaryl-coenzyme A (HMG-CoA) reductase inhibitors, or statins, (n=9,827) are provided in a spreadsheet in the online supplement. Access to this transformed data set can be provided via comma-separated value files or access to our Google BigQuery repository on reasonable request.

FAERS uses the Medical Dictionary for Regulatory Activities (MedDRA, V.25.1) Preferred Terms (PT) to classify ADEs. The specific PTs used to identify DM in this analysis were “diabetes mellitus”, “diabetes mellitus inadequate control”, “diabetes with hyperosmolarity”, “diabetes complicating pregnancy”, “gestational diabetes”, “increased insulin requirement”, “insulin resistant diabetes”, “pancreatogenous diabetes”, “insulin-requiring type 2 diabetes mellitus”, “type 1 diabetes mellitus”, “type 2 diabetes mellitus.”

The primary analysis used the aggregate of all approved statins combined (atorvastatin, fluvastatin, lovastatin, pitavastatin, pravastatin, rosuvastatin, simvastatin). We did not include cases involving cerivastatin primarily because it is not available for clinical use. In addition, it was on the market for only 4 years at the beginning of the study period, there was a maximum dose during that time which provoked severe side effects, and it was not clinically available during the period after statin-associated DM was recognized. Analyses were repeated for each of the four most reported statins (atorvastatin, pravastatin, rosuvastatin, simvastatin, figure 1). The primary outcome was the presence of a significant difference between men and women in the spontaneous reporting rate of DM associated with all statins combined. Secondary outcomes were sex differences in spontaneous reporting of DM associated with each individual statin.

Request permission

Total unique statin cases in Food and Drug Administration’s Adverse Event Reporting System, January 1, 1997, to December 31, 2023.

Pharmacovigilance analyses of spontaneously reported ADE data like those from FAERS generally involve identifying SDRs, where certain ADEs are reported more frequently with the drug(s) of interest compared with the reporting frequency of the same ADE in all other reports combined. When studying large numbers of events, the Proportional Reporting Ratio (PRR) has been used and recommended by drug safety monitoring agencies such as the FDA, EudraVigilance, and the UK’s Yellow Card Scheme.14 We therefore used the PRR as the primary indicator of an SDR for DM associated with statins compared with all other drugs. The PRR was considered significant if the PRR point estimate was>2, χ2≥4, and there were≥3 reports of DM associated with statins.15 The PRR is calculated as follows:

We then tested for differences between reporting in men and women using a modified reporting OR (ROR), which has been used in prior analyses to compare SDR between subgroups including in analyses of sex differences.16 Briefly, this ROR compares the ratio of statin-associated DM to all other statin-associated ADEs in women to that in men, where higher ROR indicates a greater difference between women and men. A difference in statin-associated DM reporting between women and men was considered significant if the lower bound of the ROR 95% CI was>1 and there were≥3 reports of DM associated with statins.14 The ROR was calculated as follows:

Categorical and continuous variables were otherwise compared using χ2 analysis, and analysis of variance, respectively. A p value<0.05 was considered significant for both. All analyses were performed using RStudio (V.2023.12.0, Posit Software, Boston, Massachusetts, USA) and the R statistical package (V.4.3.2, R Foundation for Statistical Computing, Vienna, Austria). SDR analyses were performed using the mdsstat package (V.0.3.2, ASM, Temecula, California, USA). All FAERS, Drugs@FDA, and MedDRA data are hosted for analysis in Google BigQuery (cloud.google.com, Mountain View, California, USA). Analysis code and access to our FAERS BigQuery repository are available on reasonable request.