Statins and the “Healthy User Bias” in Cardiac Surgery

The manipulation of plasma lipids with medication is a relatively new phenomenon. In the late 1960s and early 1970s, animal experimentation drew the association between hypercholesterolemia and increased inflammation, atherosclerosis, and activation of platelets. The Lipid Research Clinics Primary Prevention Trial1 was conducted in >3800 men with primary hypercholesterolemia. After an intensive screening program, these men were given multiple risk factor interventions including counseling on diet, exercise, and smoking cessation. After an initial run in period, if lipids were still elevated, these men were randomized to cholestyramine, a resin binding agent that was poorly tolerated, or a placebo. The reduction in low-density lipoprotein was found to reduce the incidence of a combined cardiac outcome in a dose-dependent manner. In this trial, mortality was not decreased. The correlation between reduced low-density lipoprotein and cardiac outcomes has been replicated in 2 other large-scale randomized controlled trials, the Helsinki2 and Coronary Drug Project.3 The resultant overwhelming positive effects of lipid lowering led to the synthesis, testing, and marketing of the 3-hydroxy-3-methylglutaryl CoA reductase inhibitors (“statins”), which are now the most widely sold pharmaceutical on the market. There can be no debate: statins reduce cardiovascular morbidity and mortality.4 The cardioprotective effects have been enumerated in large-scale randomized trials in both primary and secondary prevention. Over the long term, these cardiovascular effects are cost-effective.5 What is interesting and controversial is the growing list of observational trials that have associated positive effects of the statins in what can only be classed as biologically diverse pathologies. Statins are now associated with many positive outcomes, including better psychological well-being,6 a reduction in strokes,7 fewer venous thromboembolism events,8 fewer hospital admissions for chronic obstructive pulmonary disease,9 fewer admissions and decreased hospital lengths of stay for community-acquired pneumonia,10 less colorectal11 and prostate cancer,12 and fewer hip fractures.13 In cardiac surgery, statins have been associated with improved survival, fewer cardiac events, improved renal outcomes, and less delirium.14,15 Although these associations are potentially important, we would suggest tempering the enthusiastic approach to these findings. As in all observational trials, a clear understanding of the potential biases is paramount. In the case of the 3-hydroxy-3-methylglutaryl CoA reductase inhibitors, there are at least 4 important potential biases that need to be assessed in each individual report: A “healthy user effect” refers to the phenomena in which the patients who adhere to a medication schedule also engage in a healthy lifestyle, which may result in better overall health.* Patients taking statins are on average younger and see a primary care physician more often. Thus, it has been shown that a patient who takes his or her medication as prescribed also engages in practices that secondarily result in better outcomes. It has been shown that patients who adhere to statins more frequently use fecal occult blood testing, have regular mammography, receive influenza vaccination, pneumococcal inoculation, and prostate-specific antigen testing.16 Furthermore, patients who adhere to a statin dosing regimen are also likely to engage in smoking cessation, eat a healthy diet, and exercise regularly. This “healthy user effect” with statins has also been associated with a reduction in adverse events that are clearly unrelated to their use, including better dental hygiene; less drug dependency; and involvement in fewer workplace accidents, falls and/or fractures, and fewer automobile accidents.17 Thus, we submit that the “healthy user effect” is almost certainly a factor in the plethora of reports pertaining to the protective effects of statins. Simvastatin is more potent on a milligram-per-milligram basis than atorvastatin.18 However, few 3-hydroxy-3-methylglutaryl CoA reductase inhibitor observational studies list either the drugs or the specific dose. Remember, aggressive incremental dosing may be associated with better protection against stroke, myocardial infarction, and death.19 This argument also leads to a consideration of an unmeasured population selection bias as well. Certain patients are not tolerant of statins when an aggressive dosing regimen is initiated. In an evaluation of >25,000 patients receiving aggressive dosing of statins, patients were found to have a 28% increased likelihood of having to withdraw from the drug because of side effects. Listed side effects include increased liver enzymes, increased creatinine kinase levels, and malaise.20,21 This type of selection bias† is an issue in observational studies because patients who have a history of intolerance are usually labeled as nonusers as opposed to nonresponders or a dosing failure. Alternatively, in randomized controlled trials, using the appropriate intention-to-treat analysis, nonresponders (patients who cannot tolerate statins, or those who are withdrawn from therapy) show no positive effect, bringing the result closer to the null hypothesis. The bias is doubly strong in observational studies because these nonresponders are now compared with patients who are able to tolerate aggressive dosing and receive optimal care with the maximal result. Selection bias should be considered in all observational studies, especially where the demonstrated treatment effect is small. Finally, the most obvious factor is publication bias.‡ There are few negative trials published, and in a meta-analysis of the effect of statin use and reduction of hip fractures, a publication bias, with missing negative trials, was suggested.13 It should be stated that Anesthesia & Analgesia does not have either a policy or a bias against publishing negative results; it does receive fewer reports of negative trials, and therefore, statistically, will publish fewer negative reports. The process of writing a manuscript and enduring the review process is daunting, even for the most prolific investigators. The process is even more work for negative trials, thus many investigators may not even attempt to report the results of a negative observational investigation. In this respect, the present study is an important addition to the literature.22 In this edition of Anesthesia & Analgesia, Argalious et al.22 from the Cleveland Clinic are to be congratulated for persevering and publishing an evaluation of the effects of 3-hydroxy-3-methylglutaryl CoA reductase inhibitors on postoperative renal function in 10,648 patients undergoing elective cardiac surgery. Using the RIFLE (Risk of renal dysfunction, Injury to the kidney, Failure of kidney function, Loss of kidney function, and End-stage kidney disease) criteria to evaluate postoperative renal function, the authors found no significant association between statins and renal failure, a 20% reduction in postoperative dialysis which was not statistically significant, and a 17% reduction in death, which was, again, not significant. Clearly, as one of the very few negative reports on the effectiveness of statins in our surgical populations, this report will be instructive and yet controversial. The strengths of this report are the generally good adherence to the STROBE Statement (www.strobestatement.org), the well-described population of interest, and an unselected consecutive cohort with limited exclusions. The data source is likewise well described, collected prospectively, and monitored on a regular basis for accuracy and completeness. The outcome of interest is postoperative renal failure and uses the definitions in the RIFLE classification, which is widely accepted as the “gold standard” in this type of outcome.23 Another particular strength is the manner and attention to the adjustment for the important confounders. Argalious et al.22 included all of the recognized factors contributing to postcardiopulmonary bypass–mediated renal dysfunction, including bypass time, anemia, and blood transfusion. Finally, the adjustment for the confounders used both logistic regression and a propensity score-matched cohort analysis, which use different mathematical assumptions. Both methods resulted in the same conclusion. One of the weaknesses of the study is the fact that the authors were unable to control for the important biases listed above. To their credit, the discussion acknowledges “the healthy user effect”; however, this report, similar to all other perioperative studies, was unable to control for this effect. Second, there is neither a description of the types of statins nor any mention of the dosing. Such a study is likewise unable to assess the factors such as adherence, intolerance to the medication, or prior dose escalations. Third, this report failed to assess sample size a priori. An important limitation because, as with all negative studies, it is possible that the negative outcome is attributable to inadequate sample size and the power to demonstrate a significant difference. In a post hoc analysis, Argalious et al. revealed that although it is possible that there is a protective effect of the 3-hydroxy-3-methylglutaryl CoA reductase inhibitors, the cohort would have had to include >15,000 patients to demonstrate the effect. Although these biases remain unaddressed, both a healthy user effect and correction for dose and adherence would tend to minimize any protective effect. The authors correctly conclude on the basis of their evidence that: “Even if found to have a beneficial effect on renal outcomes and mortality after cardiac surgery with a larger patient population, the magnitude of renoprotection by statins is likely to be low.” Argalious et al.22 have delivered what is probably the most exhaustive and complete risk adjustment to date for the confounders of statin therapy in the cardiovascular surgery literature. This investigation finds that statins, presented in the most optimistic light, may confer a small beneficial effect. Readers and investigators should consider the following before calling for large-scale randomized trials: a large proportion of patients presenting for coronary surgery are currently receiving statins, but a significant number of patients cannot tolerate them. Second, given the large sample size that would be needed to assess any meaningful outcome, a large trial is probably unfeasible. However, what is abundantly clear is that statins confer protection to patients with cardiovascular disease and, as demonstrated in this study, continuation of statins into the perioperative period is safe.