Among commonly prescribed statins, rosuvastatin appears to be linked to the highest adverse event risks reported across most muscle-related side effect categories in post-marketing patient populations. Lovastatin and pravastatin appeared to have the lowest risk rates. These findings, based on a significant volume of case reports in the FDA AERS database linking statins with muscle adverse effects, corroborate and extend existing knowledge regarding the association of statin drugs with muscle-related adverse events. Our results parallel those of Sakaeda et al., 2011 
, Cham et al., 2010 
, and Alsheikh-Ali et al., 2005 
. Additionally, the findings generally corroborate those of Cham et al., 2010 in observing that fluvastatin, an agent that was not commonly prescribed, was an apparent potency outlier 
. Using a patient-targeted survey approach, they demonstrated that: (i) the highest potency statins (rosuvastatin and atorvastatin) showed higher muscle adverse event rates, (ii) simvastatin, with intermediate potency, showed intermediate rates, and finally (iii) pravastatin and lovastatin, with their lower potencies, showed the lowest rates 
. (See Table S3
for relative dose equivalence of statins.) We think a likely reason is that fluvastatin, which is far less frequently prescribed than other statins, may be primarily reserved by physicians for patients who have failed to tolerate other statins. Disproportionate use in statin non-tolerators may produce higher apparent adverse effect rates (indeed, use of fluvastatin 80 mg for those intolerant to other statins is advised by some 
). It might also be selectively used in settings in which drug interactions or other factors heighten toxicity. Alternatively, of course, fluvastatin might actually engender risk of muscle adverse effects beyond expectation for its potency. Head-to-head randomized assessments of fluvastatin versus other agents (or within-person crossover comparisons like the potency comparisons of Cham et al. 
) are desirable to resolve this.
Sakaeda et al., 2011 
analyzed the AERS database using methods similar to ours and found that muscle-related adverse events were more commonly observed with rosuvastatin treatment when compared with other statins such as atorvastatin and pravastatin. However, their inferences differ from our own regarding the foundation for such differences, as they do not ascribe a primary role to statin potency 
. Alsheikh-Ali et al., 2005 analyzed the first year of rosuvastatin AERS data against other major statins with a “first year of marketing analysis” and a “concurrent time period analysis” 
. Corresponding to our findings, their analysis showed that rosuvastatin had a higher risk rate for important muscle-related side effects.
Other studies that have analyzed post-marketing adverse events linked to rosuvastatin include Wolfe and Zipes et al. 
. Both used data taken from the small time window of approximately one year following rosuvastatin's introduction into the US market. In contrast to our findings and those cited above, the Zipes' et al., 2006 study appeared to show no difference in risk rates between rosuvastatin and other statins 
. However, that study used the ratio of a given adverse event report to all adverse event reports for the drug as the index. This approach may preclude detection of even large increases in adverse events if they are proportional (all adverse events increased together), as might arise from factors like greater potency. The Wolfe 2004 paper cited high renal and muscle-related side effect rates from both pre- and post-marketing data 
. Wolfe specifically focused on potentially high rhabdomyolysis risks in calling for rosuvastatin to be pulled from the market 
From a pharmacokinetic perspective (without consideration of potency), rosuvastatin has a profile that might be expected to yield fewer, not more, adverse events. It has high hepatoselectivity, high hydrophilicity, low rates of metabolism via cytochrome P450 enzymes, and moderate systemic bioavailability 
. Typical statements in the scientific literature appear to suggest rosuvastatin's safety equals or surpasses that of other statins, with assertions that rosuvastatin: is “safe and well tolerated,” 
has a “safety profile comparable to other statins” 
, “has a superior safety profile,” 
or even “an improved clinical safety profile” 
Our data suggest, however, that any benefits attending such factors may be overridden by other factors, such as potency considerations. Higher potency agents, and rosuvastatin in particular, were associated with elevated relative risk of adverse events. This finding has important implications for statin treatment decisions in general, and particularly with regard to patients who have already experienced muscle-related adverse events from statin therapy.
The parallels between our results and findings from prior adverse event surveys corroborate and validate the idea that valuable information can be obtained from within-class drug comparisons based solely upon AERS data. For key purposes, then, such an approach may lessen the need for, and costs associated with, separate adverse event surveys targeting individual drug classes.
This is important because of the central role of postmarketing information in the assessment of drug safety. Randomized clinical trials (RCTs) are considered the highest quality evidence for causal inference, but they have significant limitations for adverse effect detection and analysis. These include methods that exclude subjects who may have polypharmacy risks, are elderly, have comorbidities, and known risk factors for harm. Such patient-selection practices differentially exclude precisely those subjects who might be most likely to experience adverse events. Additional factors that can limit clinical trial utility for adverse effect understanding include: (i) exclusions based on statin compliance 
(lower compliance is linked to statin adverse effects 
), (ii) restrictive definitions for “statin myopathy” (e.g. requiring CK elevations >10× ULN), (iii) drug exposure times that are short relative to those experienced in many post-marketing consumer populations, (iv) treatment discontinuation at the first sign of a problem (yielding different findings than real-world usage), and (v) lack of comparator data across statin agents.
The relationship noted here between reported muscle-related adverse effects and statin potency (in terms of LDL reduction) does not necessarily imply that muscle adverse events are caused by lowering LDL cholesterol. The potency of LDL reduction relates to the magnitude of mevalonate inhibition, which, in turn, affects coenzyme Q10 levels, testosterone reduction in men, reduced antioxidant transport, and numerous other factors.
Postmarketing surveillance bears well-recognized limitations, such as lack of randomization, and is not intended to replace RCT approaches. Nonetheless, for the reasons noted above, RCTs are disadvantaged in adverse event detection, and case reports and postmarketing surveillance are commonly responsible for the first identification of important adverse events, including those that ultimately lead to regulatory actions such as “black box” warnings and product withdrawals 
. Each approach has an important role, and complements the other to extend the understanding of drug benefits versus risks. Limitations of our analysis include: (i) the FDA AERS database is only as accurate as the information inputted into it from various sources. (ii) AERS does not filter, correct, or make any analysis of the quality or potential bias of inputted data. (iii) Exogenous factors such as publicity and marketing can influence reporting. (iv) Physicians might disproportionately report effects associated with newer drugs, and rosuvastatin is the newest of the statins studied. Our analysis, however, did not find clear support for this limitation. (v) Dose data are not available. Physicians could prescribe higher doses of one statin within the recommended dosage range. However, maximum potencies of use bear an expected relation to average potencies of use. Moreover, in the Cham analysis of adverse effects, consistent results were obtained whether looking at results by expected potency equivalencies without consideration of dose and when evaluating individual rechallenge cases with known drug and dose. Both presumed potency for the statin, and known potency based on dose and dose equivalencies, were predictors of relative adverse event rates 
. (vi) Reports submitted to the FDA contain mistakes, including spelling errors leading to misclassifications, important data either missing or inadequately reported, and duplicate reports; however our analysis systems included multiple processing steps, safeguards, and manual oversight to lessen the impact of such factors. (vii) Only a minority of post-marketing adverse events are believed to be successfully logged into AERS 
. Therefore, any calculated rates are apt to substantially underestimate the actual incidence of these side effects in broad consumer populations. We address this by use of comparative rates.
Given that: (i) relative adverse event risks appear to be higher with higher potency statins, and (ii) other data sources (meta-analyses of head to head statin trials) have demonstrated that mortality outcomes are not more favorable with higher potency statin use (except in acute coronary syndrome) 
, we suggest these findings favor use of lower potency statins (if statin use is clearly indicated), particularly where a previous statin myopathy has occurred.
The data presented in this report may offer important reference points regarding the selection of statins for cholesterol management in general, and especially for the rechallenge of patients that have experienced muscle-related side effects. If statin reinitiation is considered following muscle-related adverse effects, agents of lower expected potency should be preferred. We believe that our results warrant the attention of healthcare providers, drug developers, patients, and regulatory professionals involved with statins and other cholesterol-related medications. Moreover, the data mining approach employed appears promising, and may have application in evaluation of adverse reactions from other drug classes.