Let's not be so quick to stop statins
Expert opinion on the importance of awareness of reporting bias and outstanding questions in studies.
Investigators at University of California, San Diego, recently published a report in the Archives of Internal Medicine, “Effects of Statins on Energy and Fatigue With Exertion.” Given the widespread use of statins, this topic is highly important and worthy of close scrutiny. In their brief research letter, the authors report an association between statin use and a subjective, survey measure of “fatigue.” This potentially important paper carried an important subtitle, “Results From a Randomized Controlled Trial,” implying the highest level of scientific evidence.
We decided to investigate the quality of this report according to the modern, evidence-based medicine standards expected of a randomized controlled trial (RCT). A high-quality RCT should not only be a randomized comparison of intervention and control groups, but should employ an appropriate randomization scheme, use proper blinding techniques, prespecify outcomes, choose outcomes of clinical importance, register the study design and hypothesis/aim (eg, on clinicaltrials.gov), achieve a high proportion of patient follow-up and data capture, and report on primary outcomes expeditiously in a full-disclosure manuscript. We have concerns that this study lacks several of these elements expected of a high-quality RCT and suffers from multiple potential sources of bias.
Examination of the report
Michael J. Blaha
Roger S. Blumenthal
Seth S. Martin
According to clinicaltrials.gov and the design manuscripts, the purpose of the UCSD Statin Study was to study the effect of statins on three specific primary outcome measures: (1) cognition; (2) irritability/aggression; and (3) serotonin levels. The study started in April 2000, and the final data were collected by March 2004. To date, there is no published manuscript on the primary results of the UCSD Statin Study. To put this situation in context, instructions to authors at Archives’ sister journal, the Journal of the American Medical Association, state: “Manuscripts based on data from randomized controlled trials should be reported … ideally within 1 year after follow-up has been completed.” The publication of this report in 2012 on the EnergyFatigEx score (an exploratory outcome that was neither prespecified nor previously validated), despite no published manuscripts on the results of the primary outcome measures of the study, raises concern about bias in reporting.
In addition, there are outstanding questions about the methods of this study. Regarding the EnergyFatigEx score, what exactly is the difference between energy vs. fatigue with exertion? What questions were posed to the patients and how was this done (eg, via written questionnaire or another method)? Did the researchers need to combine the two parameters to show an effect (ie, double counting)?
A larger question is the clinical relevance of the EnergyFatigEx score. The score has never previously appeared in the literature. It has not been validated. Therefore, it is difficult to know how to interpret results based on the score. The authors emphasize the link between their energy measure and actual exercise. However, the correlation was weak, and it is uncertain whether there is true clinical importance to the finding. We would like to know how many patients discontinued statins due to complaints of decreased energy or fatigue with exertion and whether such complaints improved after drug withdrawal.
Another important issue was not made transparent in the report and deserves emphasis — the issue of imputation. Unlike the standard RCT, which bases its analyses on directly measured data, some data were imputed by the authors in the study. To be clear, imputation means that the authors assigned values to patients with missing data by inference. However, the authors do not report how many data points were imputed and also do not include other standard information (eg, single vs. multiple imputation, variables in imputation model, etc).
So, we really need more information about this study to know what to make of its findings. In an attempt to learn such information and open a discussion, we submitted a Letter to the Editor at the Archives of Internal Medicine and have emailed the corresponding author.
Although many questions remain outstanding, the authors state that the effects they report “merit consideration when prescribing or contemplating use of statins.” A UCSD news headline reads, “Statins Shown to Cause Fatigue.” In the UCSD article, the lead investigator is quoted as saying, “Side effects of statins generally rise with increasing dose, and these doses were modest by current standards. … Yet occurrence of this problem was not rare — even at these doses and particularly in women.”
In an article about this study on theheart.org, the lead author had this advice for clinicians: “Statins are fine in patient populations where a mortality benefit has been shown — ie, men under 70 with heart disease or primary-prevention patients with raised [C-reactive protein; CRP] or who smoke. But I would think twice for other groups. Primary-prevention patients who don’t smoke or don’t have raised CRP are far more likely to experience an adverse effect than to have a cardiac event.”
As Mark Twain once said, “Many a small thing has been made large by the right kind of advertising.” Beyond advertisements throughout the news media, the Archives editors emphasized the clinical importance of the study in a letter to JAMA arguing against the use of statins for men at risk for CVD. We are concerned that kind of message instills undue fear and means that patients may be inappropriately taken off of statins — a generally safe and efficacious class of medications.
RCTs and efficacy of statins
Indeed, more than two dozen statin RCTs have consistently demonstrated the efficacy of statins in reducing MI, stroke, revascularization and death in patients at risk for CVD. In weighing the risk and benefits of statins, patients and physicians must consider an obvious fact — fatigue, whether it occurs or not with statins, is potentially reversible, but death, stroke and MI are not.
Moreover, statin trials showing efficacy were RCTs in the full sense of the term — they were not just randomized; they included the other elements of RCTs that scientists widely associate with this term. The outcome measures were definitely clinical important: MI, stroke, revascularization and death. These outcome measures were prespecified. Outcomes were measured, not imputed. Methods and results were thoroughly reported and in a timely fashion.
In the JUPITER trial, of 17,802 primary prevention patients with raised high-sensitivity CRP, there was no difference in the report of fatigue between the placebo group and treatment arm. There were 297 of 8,150 (3.6%) patients in the placebo group who reported fatigue and 309 of 8,154 (3.8%) patients receiving rosuvastatin 20 mg. There was also no significant difference in fatigue by level of LDL attained with an incidence rate per 100 person-years of 1.7% in the placebo group, 1.6% in the treatment arm patients who did not attain an LDL level <50 mg/dL and 1.8% in treatment arm patients who attained an LDL level <50 mg/dL.
Despite the publicity, this Archives report is, in our view, just hypothesis-generating. We found that closely scrutinizing the information that is available and debating this Archives report amongst ourselves and other colleagues led us to view the findings with more caution. We found that the study deviates from the scientific rigor of a high-quality RCT and remain concerned about multiple potential sources of bias, including exploratory analyses of an unvalidated outcome measure of unclear clinical significance, untimely reporting and outstanding questions about the methods of the study. As Benjamin Haydon said, “Fortunately for serious minds, a bias recognized is a bias sterilized.”
We encourage all clinicians to closely read the literature and closely examine the methods of studies making potentially game-changing headlines. Especially if one notices red flags — like a seemingly important “RCT” published as a brief research letter. Rather than passively accepting the results and conclusion, ask oneself important questions, discuss them with colleagues, write letters to the editor and email the corresponding author. Participating in this process leads to better science, and, ultimately, better patient care.
Seth S. Martin, MD; Roger S. Blumenthal, MD; and Michael J. Blaha, MD, MPH, are all from the Johns Hopkins Ciccarone Center for the Prevention of Heart Disease. Blumenthal is also the Cardiology Today CHD and Prevention Section Editor.
For more information:
- Baigent C. Lancet. 2010;376:1670-1681.
- Beckman JA. JAMA. 2012;308:666.
- Clinicaltrials.gov. Effects of statin medications on mental processes, behavior, and serotonin levels. Available at: clinicaltrials.gov/show/NCT00330980. Accessed Aug. 28, 2012.
- Golomb BA. Arch Intern Med. 2004;164:153-162.
- Golomb BA. Arch Intern Med. 2012;72:1180-1182.
- Golomb BA. Control Clin Trials. 2004;25:178-202.
- Hsia J. J Am Coll Cardiol. 2011;57:1666-1675.
- UC San Diego. Statins shown to cause fatigue. Available at: ucsdnews.ucsd.edu/pressreleases/statins_shown_to_cause_fatigue. Accessed Aug. 28, 2012.
- Martin, Blumenthal and Blaha report no relevant financial disclosures.
Seth S. Martin, MD, can be reached at the Johns Hopkins Division of Cardiology, 600 N. Wolfe St./Carnegie 568, Baltimore, MD 21287; his email is firstname.lastname@example.org.
Roger S. Blumenthal, MD, can be reached at the Johns Hopkins Division of Cardiology, 600 N. Wolfe St./Carnegie 568, Baltimore, MD 21287; his email is email@example.com.
Michael J. Blaha, MD, can be reached at the Johns Hopkins Division of Cardiology, 600 N. Wolfe St./Carnegie 568, Baltimore, MD 21287; his email is firstname.lastname@example.org.