Author + information
- Received May 2, 2011
- Revision received July 22, 2011
- Accepted September 20, 2011
- Published online February 7, 2012.
- William J. Kostis, PhD, MD⁎,⁎ (, )
- Jerry Q. Cheng, PhD†,‡,
- Jeanne M. Dobrzynski, BA‡,
- Javier Cabrera, PhD§,‡ and
- John B. Kostis, MD‡
- ↵⁎Reprint requests and correspondence:
Dr. William J. Kostis, Cardiology Division, Massachusetts General Hospital, Harvard Medical School, 55 Fruit Street, GRB-800, Boston, Massachusetts 02114
Objectives The aim of this study was to evaluate the effect of statins in decreasing cardiovascular events in women and men.
Background Published data reviews have suggested that statins might not be as effective in women as in men in decreasing cardiovascular events.
Methods Published data searches and contacts with investigators identified 18 randomized clinical trials of statins with sex-specific outcomes (N = 141,235, 40,275 women, 21,468 cardiovascular events). Odds ratios (ORs) and 95% confidence intervals (CIs) for cardiovascular events were calculated for women and men separately with random effects meta-analyses.
Results The cardiovascular event rate was lower among those randomized to statin intervention than in those randomized to control (low-dose statin in 4 studies, placebo in 11 studies, usual care in 3 studies) and similar in women and men (OR: 0.81, 95% CI: 0.75 to 0.89; p < 0.0001, and OR: 0.77, 95% CI: 0.71 to 0.83, p < 0.0001, respectively). The benefit of statins was statistically significant in both sexes, regardless of the type of control, baseline risk, or type of endpoint and in both primary and secondary prevention. All-cause mortality was also lower with statin therapy both in women and men without significant interaction by sex (p for interaction = 0.4457).
Conclusions Statin therapy is associated with significant decreases in cardiovascular events and in all-cause mortality in women and men. Statin therapy should be used in appropriate patients without regard to sex.
Randomized controlled clinical trials and meta-analyses have shown a benefit of statins in decreasing morbid and mortal cardiovascular events in apparently healthy individuals and in those with clinically evident cardiovascular disease (CVD) (1–6). However, there is insufficient information on the benefits of statins in women especially in primary prevention (7–9). Reviews and meta-analyses have shown improved outcomes with statins in both women and men without significant interaction by sex (10–11). However, they did not show statistically significant effects in women. This could be related to under-representation of women in trials and underscores the need to explore sex-related differences that would provide a basis for clinical strategies to improve outcomes for women (12–15).
The purpose of this report is to present a meta-analysis of sex-specific outcomes in controlled randomized clinical trials of statin therapy.
The meta-analysis was performed according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement (16). With Medline, the Cochrane Library, the Central Register of Controlled Trials, Web of Science, and ClinicalTrials.gov, we performed a systematic published data search of randomized clinical trials of statin therapy where sex-specific data were presented through June 30, 2010. Trials were eligible for inclusion in the meta-analysis if they were controlled, randomized, and investigator- and patient-blind and if they presented data by sex. We sought to obtain missing mortality data by contacting the investigators of all studies where sex-specific all-cause mortality data were not published. Studies with fewer than 100 patients or fewer than 5 deaths/randomized group were excluded. With the search strategy described in Online Table 1, we identified 2,332 potentially appropriate titles for possible inclusion in the analysis (Fig. 1, Table 1). Each of these titles was evaluated by 2 investigators (W.J.K., J.B.K.) for possible inclusion in the study. Eighteen trials fulfilled all criteria for inclusion in the meta-analysis. The studies were evaluated with regard to the similarity of baseline characteristics, having defined eligibility criteria, blinding, use of placebo or other control, intention-to-treat analysis, information on adherence, and the percentage lost to follow-up (Online Table 2) (17).
Data extraction and quality assessment
The 18 studies that fulfilled the inclusion criteria were categorized as primary prevention (5) (Online References e22–e24,e26,e31,e33,e34) or secondary prevention, (6) (Online References e21,e25,e27–e30,e32,e35,e36) as designed by the investigators. All studies except 1 were funded by the pharmaceutical industry (Online Reference e30). In the majority of the studies, an independent data center was responsible for the data and analysis. The approximate relative potency of the statin used (1 for lovastatin and pravastatin, 2 for simvastatin, 4 for atorvastatin, and 8 for rosuvastatin) and the relative dose used (calculated as the product of the dose and the relative potency) were tabulated. A measure of the use of active therapy in the intervention and control groups (active medication difference [AMD]) was defined as the percentage of patients who actually received active therapy among those randomized to receive it, minus the percentage of those who actually took active therapy among those who were randomized to control. The number of randomized patients, the number of primary endpoints, and the number of deaths in the intervention and control groups were recorded. The difference in the decrease of serum low-density lipoprotein cholesterol (LDL-c) concentration between the intervention and control groups was tabulated for each study. There were no assumptions or modifications made concerning the data included in the analyses. Sex-specific data on the number of randomized patients, the number of primary endpoints, and the number of deaths in the intervention and control groups were recorded.
Data synthesis and analysis
Pre-defined outcomes were all-cause mortality and the primary endpoint as defined by the investigators of each study. The rates of occurrence of the primary endpoint and all-cause mortality (where available) in the intervention and control groups were calculated for each study as a whole as well as by sex. Statistical analyses were performed with JMP (version 7.0, SAS Institute, Cary, North Carolina), Comprehensive Meta-Analysis (version 2.2, Biostat, Englewood, New Jersey), and R software (R Project for Statistical Computing). Odds ratios (ORs) and 95% confidence intervals (CIs) were calculated for each endpoint. Weighted pooled treatment effects were calculated for each study as a whole and by sex with random effects models. In addition, analyses were done separately for primary and secondary prevention trials, by level of baseline risk and by type of endpoint. Heterogeneity of the effects was evaluated with the Q statistic. Sensitivity analysis was performed by repeating the analysis 18 times, removing 1 study at a time. Also, analyses were performed by classifying the trials into 3 groups: primary prevention (5) (Online References e22,e33), secondary prevention (6) (Online References e21,e25,e27–e30,e32,e35,e36), and “mixed”, which included 5 studies where a significant proportion of patients might have had CVD (Online References e23,e24,e26,e31,e34). Publication bias was examined by performing cumulative meta-analysis, by the Duval and Tweedie's Trim and Fill method (18) and the fail-safe N models of Rosenthal (19) and Orwin (20). The types of these analyses were pre-specified. In addition, analyses were performed by classifying the studies according to annual risk for mortality derived by dividing the observed absolute risk by the duration of the study in years. Meta-analysis was performed by analyzing separately and comparing the 9 studies with the higher annual risk with the 9 studies with the lower annual risk as well as by classifying the studies into high annual risk (2 or higher  [Online References e23,e25,e26,e28,e31,e34]), medium (above 1 but lower than 2  [Online References e21,e24,e27,e29,e32,e35,e36]), and low risk (up to 1 [Online References e22,e30,e33]) (21). Also, we performed meta-analysis including the 3 studies (5,14) [Online Reference e33) that reported specific outcomes (stroke, coronary heart disease [CHD] event). We examined, in addition to meta-analysis, effect of sex on outcomes with meta-regression. A more detailed description of the methods is included in the online appendix.
Description of studies
Table 1 includes data on the 18 studies included in the meta-analysis. The statin used in the intervention group was atorvastatin in 4 trials (Online References e24,e30,e35,e36), lovastatin in 1 (Online Reference e22), pravastatin in 6 (Online References e23,e27,e29,e32–e34), rosuvastatin in 3 (5) (Online References e26,e28), and simvastatin in 4 (6) (Online References e21,e25,e31). Eight trials were designed as primary prevention trials (5) (Online References e22–e24,e26,e31,e33,e34), and 10 were designed as secondary prevention trials (6) (Online References e21,e25,e27–e30,e32,e35,e36). Five of the primary prevention studies included a proportion of patients with CVD (Online References e23,e24,e26,e31,e34). These studies were defined as “mixed” in the sensitivity analysis. Data on all-cause mortality were not available for 5 studies (6) (Online References e24–e26,e28). Overall, the meta-analysis included 141,235 patients, 21,468 primary events, and 13,710 deaths (3,898 deaths in studies with sex-specific mortality data). Different studies used study-specific definitions of primary events (Table 1) and had different durations of follow-up (47.9 ± 18.9 months). The mean relative dose in the intervention group was 98.2 ± 93.5 mg. The mean difference in relative statin dose between the intervention and the control groups was 89.4 ± 79.8 mg. Mean follow-up was 48 ± 19 months. On average, 85.5 ± 9.9% of those randomized to active therapy actually received it compared with 17.2 ± 9.9% of those randomized to the control group. On average, 72.9 ± 15.9% of the patients in the clinical trials were men. The mean age was 62.7 ± 5.1 years. Seventeen percent (17.1 ± 8.8%) of the patients had diabetes, 51.3 ± 21.3% had hypertension, 23.3 ± 13.9% were current smokers, 57.6 ± 29.5% were taking aspirin, and 66.9 ± 42.8% had prior CVD (ranging from 0% in 3 trials to 100% in 10 trials). The difference in LDL-c lowering (compared with baseline) between the intervention and control groups was 35.3 ± 16.3 mg/dl. When expressed as a percentage of the LDL-c at baseline, this difference was 26.7 ± 12% (Table 1).
Primary events in women and men
A statistically significant decrease in the primary endpoint was observed in women (OR: 0.81, 95% CI: 0.75 to 0.89, p < 0.0001) as well as in men (OR: 0.77, 95% CI: 0.71 to 0.83, p < 0.0001), with similar lowering in both sexes (p for interaction = 0.1837) (Table 2,Online Fig. 1). In women, the benefit with respect to the primary event seemed more pronounced in secondary prevention trials than in primary prevention trials (OR: 0.78, 95% CI: 0.70 to 0.88, p < 0.0001, and OR: 0.85, 95% CI: 0.75 to 0.98, p = 0.0209, respectively, p for interaction = 0.3397) (Online Fig. 2). Also, the benefit of the statin intervention was similar in studies where placebo/usual care or low-dose statin were used in the control group (OR: 0.81, 95% CI: 0.72 to 0.91, p = 0.0005 for both types of studies, p for interaction = 0.4545) (Online Fig. 3).
Sensitivity analysis of the primary event in women was performed 18 times with the leave-one-out method, resulting in ORs ranging from 0.80 (95% CI: 0.73 to 0.88, p < 0.0001) to 0.83 (95% CI: 0.77 to 0.90, p < 0.0001). Publication bias was assessed by performing cumulative meta-analysis with trials ordered by increasing weight and by funnel plot analysis. With cumulative meta-analysis, the associated funnel plot, and the Duval and Tweedie's Trim and Fill method, the effects of statins in lowering the primary endpoint remained statistically significant (OR: 0.84, 95% CI: 0.76 to 0.93, p = 0.035), by including 3 imputed trials (18). Rosenthal's fail-safe N indicated that 152 missing negative studies would be needed to bring the p value of the effect to >0.05 (19). Orwin's fail-safe N needed to bring the OR to 0.95 was 50 (20).
The primary endpoint was also lower in men when all studies were examined (OR: 0.77, 95% CI: 0.72 to 0.84, p < 0.0001) and when primary prevention trials were analyzed separately from secondary prevention trials (OR: 0.73, 95% CI: 0.63 to 0.84, p < 0.0001 for primary prevention, and OR: 0.79, 95% CI: 0.72 to 0.87, p < 0.0001 for secondary prevention, p for interaction by type of prevention 0.2122) (Online Fig. 4). Meta-analysis by level of risk indicated a statistically significant benefit of statin therapy at all levels of risk in both women (OR: 0.88, 95% CI: 0.81 to 0.95, p = 0.0014 for high risk, OR: 0.75, 95% CI: 0.64 to 0.89, p = 0.0011 for medium risk, and OR: 0.59, 95% CI: 0.41 to 0.87, p = 0.0066 for low risk) and men (OR: 0.87, 95% CI: 0.77 to 0.98, p = 0.0254 for high risk, OR: 0.73, 95% CI: 0.67 to 0.80, p < 0.0001 for medium risk, and OR: 0.61, 95% CI: 0.41 to 0.92, p = 0.0170 for low risk) for the primary event (Figs. 2 and 3).⇓⇓ This more-pronounced benefit in groups at low risk was also observed by meta-regression. Thus, meta-regression showed a statistically significant relationship of annual risk of mortality of each trial to the OR for the primary endpoint in both women (slope of log(OR): 0.01819, 95% CI: 0.00017 to 0.03620, p = 0.04783) and men (slope of log(OR): 0.01925, 95% CI: 0.00819 to 0.03032, p = 0.00065), indicating a greater benefit (lower OR) in low-risk groups. A statistically significant benefit with respect to stroke was observed in the meta-analysis of the 3 studies with sex-specific outcomes (OR: 0.74, 95% CI: 0.55 to 0.99, p = 0.0396 for women, and OR: 0.70, 95% CI: 0.57 to 0.84, p = 0.0002 for men). The benefit for CHD was also statistically significant (OR: 0.78, 95% CI: 0.67 to 0.94, p = 0.0090 for women, and OR: 0.73, 95% CI: 0.66 to 0.81, p < 0.0001) (Fig. 4). In the 2 studies with sex-specific reports on adverse effects, there was no significant difference between women and men (JUPITER [Justification for the Use of Statins in Prevention: An Intervention Trial Evaluating Rosuvastatin], TNT [Treating to New Targets]). In the TNT trial women were slightly more likely to report myalgia in both active and control treatment groups, and there was no difference between treatment groups for women and men.
Meta-regression showed a statistically significant relationship between the benefit of statin therapy and the difference in LDL-c lowering between intervention and control groups (slope log(OR) vs. LDL: −0.00416, 95% CI: −0.00642 to −0.00910, p = 0.0003) (Online Fig. 5).
Pairwise comparisons of the primary event for women versus men did not show a significant difference between men and women (log(OR)men − log(OR)women) = −0.026, SD = 0.062, p = 0.68) in multivariate meta-regression. This corresponds to an OR ratio (ORmen/ORwomen) of 0.77 ± 0.54, a nonsignificant trend for men to benefit more from statins. Also, a significant difference in all-cause mortality was not observed (log(OR): difference = −0.002, SD = 0.215, p = 0.993, indicating a nearly identical effect of statins in women and men, OR: 0.998).
To examine whether differences in primary events in women versus men were influenced by trial characteristics, we performed multivariate meta-regression analyses. They showed a trend implying that men might be more likely to benefit from active therapy than women (Online Appendix).
All-cause mortality in men and women
All-cause mortality was lower in women when all studies were examined (OR: 0.90, 95% CI: 0.82 to 0.99, p = 0.0344) (Table 3, Online Fig. 6) as well as when primary prevention trials were analyzed separately (OR: 0.87, 95% CI: 0.78 to 0.97, p = 0.0142, p for interaction = 0.5122). The effect on all-cause mortality in women was not statistically significant for secondary prevention trials (OR: 1.03, 95% CI: 0.84 to 1.25, p = 0.7926). All-cause mortality was also lower in men when all studies were examined (OR: 0.84, 95% CI: 0.77 to 0.92, p = 0.0003) (Online Fig. 6) as well as when the secondary prevention trials were analyzed (OR: 0.76, 95% CI: 0.66 to 0.87, p = 0.0001). In men, all-cause mortality was not significantly lower in the intervention group for primary prevention trials (OR: 0.92 95% CI: 0.84 to 1.01, p = 0.0664). However, there was no statistically significant interaction between the decrease in mortality and type of prevention in men (p for interaction = 0.2122).
The findings of this study are consonant with a wealth of information from randomized clinical trials and meta-analyses. The Cholesterol Treatment Trialists' Collaboration individual participant data meta-analyses showed decreases in cardiovascular events and mortality with statin use and that more intensive lowering of LDL-c produced further reductions in cardiovascular events (1,22). Our analysis, indicating a decrease in the primary endpoint and all-cause mortality and a positive association between LDL-c lowering and lower risk, is consistent with the Cholesterol Treatment Trialists' findings. In the present study, the benefit of statins in reducing primary endpoints was observed in women and men and in both primary and secondary prevention without significant difference between the 2 sexes. Lower all-cause mortality was also observed in both women and men. Meta-regression with stepwise variable selection suggested that the benefit was more pronounced in men in studies with higher percentages of smokers or patients with CVD and lower percentage on aspirin. The mortality benefit was statistically significant for primary prevention in women and for secondary prevention in men, although there was no significant interaction by sex in these analyses.
Studies classified as primary prevention frequently include patients with CVD, and the distinction between primary and secondary prevention is ambiguous (21,23). For this reason, we performed analyses by risk level. They indicate a benefit across risk levels in both women and men, and meta-regression shows a relationship between risk and the benefit of statins. The reasons for the more pronounced benefit of statins in studies that enrolled low-risk patients, both by meta-regression and by subsetting, are not known. Also, because the primary endpoint was different across studies, we analyzed the data according to whether the endpoint included or did not include stroke or angina, endpoints more likely to occur in women. The analyses indicated a benefit was shown for studies including either stroke or angina as a component of the combined endpoint. A recent Cochrane report on the use of statins for primary prevention agrees with our finding of reduction of cardiovascular events but does not include sex-specific results (21). The authors stated that caution should be taken in prescribing statins for persons at low cardiovascular risk (below 1% annual all-cause mortality risk). We found a benefit in both women and men in the 3 studies with annual risk 1% or below (Online References e22,e30,e33). However, the costs and safety of the medications should be considered in addition to the absolute benefit in decreasing events (24). The estimated cost of preventing 1 cardiovascular event with rosuvastatin in the JUPITER trial is $287,000 (21), although with the decrease in the price of generic statins the cost would be lower.
Significant differences between women and men were not observed in the 2 studies reporting sex-specific adverse events, and women are under-represented in clinical trials (5,12,14,25). This underscores the importance of collecting sex-specific data and of increasing the percentage of women in clinical trials.
Although LDL-c is a strong predictor of CHD in both women and men and similar treatment approaches are recommended for both sexes, CHD might have different manifestations in women, and medications might have different effects in women than in men (7–10,12–15). For example, aspirin seems to have differential effects in primary prevention of CVD, with men deriving benefit primarily in reduction of myocardial infarction and women in reduction of ischemic stroke (26). The present study, however, does not indicate any sex differences in the beneficial effects of statins in either primary or secondary prevention.
Previous meta-analyses, most conducted before the publication of the JUPITER trial, have implied that some of the benefits of statins do not pertain to women in primary prevention and that all-cause mortality is not decreased in women (7–10). Rosenberg and Allard (7) have stated that safety meta-analyses do not disaggregate for women and do not consider female vulnerability to statin-induced muscle problems and women-centered concerns such as breast cancer, miscarriage, and birth defects. Wenger et al. (13–15) has emphasized that, as the population ages, the incidence of CVD will increase among women and that the great majority of persons over 80 years are women and are more likely to suffer cardiovascular events. Thus, meta-analyses such as the present study provide important information from a large proportion of the population. Although statin therapy benefits both women and men, lipid-lowering therapy with statin drugs is currently underused. Barham et al. (27), examining a sample of 60 North Carolina primary care practices participating in a randomized practice-based trial, did not find large differences by sex in screening or appropriateness of management. Lifestyle interventions are preferable in primary prevention but they might be difficult to implement and to maintain (28).
Limitations of the study include possible publication bias, although it is unlikely that this has a significant sex-specific effect, because in all trials the hypothesis and the primary endpoints analyzed were not sex-specific. Exclusion of studies without sex-specific data diminishes the number of studies available for the meta-analysis and the power of the analyses. This study indicates that statin therapy is beneficial in both women and men. The trend suggesting higher benefit in men than in women must be interpreted with caution, because the power of the study in this respect was low (24%). Analysis of patient-level data that might be available will provide more precise estimates than those presented in this report (1,22). It is possible that such analyses, although showing benefits of statin therapy in women, will reveal a more pronounced benefit in men among certain subsets, because trends in that direction were observed in this study. However, we should not over-interpret the data of the 18 trials that represent, to our knowledge, all available information at the time of this report.
Statins decrease cardiovascular events and all-cause mortality in both women and men. The effect on cardiovascular events is present in both primary and secondary prevention trials. Therefore, statin therapy should be used in appropriate patients without regard to sex. It seems that, with respect to statin therapy, what is good for the gander is good for the goose (29).
For supplementary text, tables, figures, and references, please see the online version of this article.
This study was supported in part by the Robert Wood Johnson Foundation. All authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- coronary heart disease
- confidence interval
- cardiovascular disease
- low-density lipoprotein cholesterol
- odds ratio
- Received May 2, 2011.
- Revision received July 22, 2011.
- Accepted September 20, 2011.
- American College of Cardiology Foundation
- Amarenco P.,
- Labreuche J.,
- Lavallee P.,
- Touboul P.J.
- Study of the Effectiveness of Additional Reductions in Cholesterol and Homocysteine (SEARCH) Collaborative Group
- Wenger N.K.,
- Lewis J.S.,
- Welty F.K.,
- Herrington D.M.,
- Bittner V.,
- TNT Steering Committee and Investigators
- Wenger N.K.
- Taylor F.,
- Ward K.,
- Moore T.H.,
- et al.
- Bandyopadhyay S.,
- Bayer A.J.,
- O'Mahony M.S.