Statins are one the world’s most commonly prescribed drugs and by 2020, global sales were estimated to have reached approximately $1trillion. They were first approved for patients at high-risk of cardiovascular disease in 1987. Their approval and subsequent use was on the basis that the accumulation of low-density-lipoprotein-cholesterol (LDL-C) in blood vessels walls is a causative factor in the development of atherosclerosis and hence, cardiovascular disease (CVD). Since the 1980s, iterative clinical practice guidelines for the prevention of CVD have tended to expand the proportion of patients who are potentially eligible for statin treatment. Doctors are now encouraged to aggressively lower LDL-C both in patients both with and without a history of CVD and, in Ireland, about one in three adults over the age of 50 take statins.
The expansion in statin utilisation has been partly driven by assumption that the lower a patients LDL-C, the better. For example, the 2018 American Heart Association guidelines note that LDL-C is the ‘primary cause of atherosclerosis’ and promote ‘the general principle that the lower the better for LDL-C’. This loglinear relationship between LDL-C and cardiovascular outcomes has been reported in a series of meta-analyses by the Cholesterol Treatment Trialist’s (CTT) Collaboration and others.
Limitations of current analyses
While such analyses are highly influential, particularly in informing clinical guidelines, they are subject to a number of serious limitations. Firstly, the studies by the CTT are based on individual patient data which are inaccessible to independent researchers.
Secondly, the relative risk reduction (RRR) from taking statins is often described in terms of reductions in a ‘composite outcome’. A composite outcome is a combination of a number of cardiovascular outcomes which is given a title, for example, ‘major vascular events’. However, reported RRRs in composite outcomes may be associated with reductions in potentially subjective outcomes, such as revascularisation or hospitalisation, the frequency of which may depend on opinions or preferences of the attending doctor, rather than arguably more objective outcomes such as all-cause mortality, myocardial infarction [MI], or stroke. Reporting composite outcomes may give misleading impressions of the effect of treatment.
Thirdly, reporting RRRs in cardiovascular outcomes without reporting the associated absolute risk reductions (ARRs) can be misleading. RRRs can inflate the importance of a clinical intervention and may exaggerate trivial associations.
What we did
To facilitate shared-decisionmaking between patients and their clinicians, we undertook a reanalysis of the seminal statin trials. We decided to focus only on the ‘hard’ clinical outcomes of all-cause death, MI and stroke, (reporting absolute as well as relative risk reductions) and to explore the association between the degree of LDL-C lowering from statins and these clinical outcomes.
We undertook a systematic review and meta-analysis of statin trials and published our paper in JAMA Internal Medicine in March 2022.
What we found
‘We included 21 statin trials in our analysis. We appraised the quality of these studies and found them to be largely at low risk of bias.
We conducted a meta-analysis on 19 of the 21 trials that reported data on all-cause mortality and 18 trials reporting data on MI and stroke. The ARR was 0.8% for all-cause mortality, 1.3% for MI, and 0.4% for stroke. The RRR for all-cause mortality was 9%, 29% for MI, and 14% for stroke for the groups randomized to receive statin therapy compared with placebo or usual care.
We undertook meta-regressions to explore the potential mediating association of LDL-C reduction with relative and absolute treatment effects. Our findings were inconclusive, indicating very little, if any, association between the magnitude of LDL-C reduction and size of the treatment effect.
Our results suggested a benefit from taking statins in all clinical outcomes. However, this benefit was modest, particularly when considering ARR rather than RRR. Reporting only RRR has been described as ‘the first sin against transparent reporting’ as it may increase people’s willingness to receive a treatment, advise treatment, and pay to prevent the risk compared with ARR or other methods for communicating risk. Some patients who experience harms might choose to discontinue their use of the drug if they were counselled about the absolute degree of risk and benefit. However, some studies report that doctors tend to ‘succumb to the lure of relative risk reports’ and recommend treatment more on the basis of relative rather than absolute risk reductions.
In addition, the degree of benefit for an individual depends on that person’s baseline risk. The online algorithm QRisk, for example, calculates a person’s 10-year risk of heart attack or stroke based on factors such as blood pressure, sex, age, smoking status among others. For example, an overweight 65-year-old man who smokes, has high blood pressure and total cholesterol may be at high risk of cardiovascular disease, compared with a 45-year-old, non-smoking woman with slightly raised cholesterol and blood pressure and no other risk factors. If a doctor were to assess their risk of dying in the next ten years, the estimated risk for the man might be 38%, for example, whereas the woman’s risk might be only 1.4%. Now consider the impact of taking statins for both. According to our study, statins would reduce the relative risk of dying by 9%. In absolute terms, the man would reduce his risk from 38% to 34.6%, and the woman from 1.4% to 1.3%.
For all patients, but particularly those at low risk of CVD, the decision to take statins should include consideration of potential benefits and harms and there has been vigorous debate about the extent of harms from statins. For example, another CTT analysis reported that the risk treating 10000 patients with statins for 5 years could result in 5 cases of myopathy, 50 to 100 new cases of diabetes, and 5 to 10 cases of haemorrhagic stroke. However, observational studies have estimated much larger rates of muscle pain, (although these types of studies are considered lower certainty evidence compared with randomised controlled trials (RCTs)). For example, Buettner et. al reported that 22% of those taking statins in their study reported musculo-skeletal pain in at least one anatomical region in the previous 30 days compared with 16.7% of those who did not use a statin. In addition, some have argued that the CTT definition may be a high bar for diagnosing muscle symptoms among statin users who may simply define myopathy as any muscle symptom. More recently, researchers from the University of Sheffield undertook a study of people who had discontinued statins because of pain. This was a blinded N-of- 1 study design. An N-of-1 is a randomised trial of an individual patient designed to determine the best course of action for an individual patient. Participants were randomised to a sequence of six double-blinded treatment periods during which they received either statin or placebo daily. The study found that fewer muscle symptoms were reported during the period when people were on statins than on placebo. However, there are limitations to this study design including the potential carry-over of statin related adverse effects from the statin period into the placebo period and whether aggregation of a number of N-of-1 studies is methodologically valid. Thus, there is still considerable uncertainty regarding muscle symptoms in statin users. High-quality RCTs designed specifically to measured adverse effects would be required, the results of which should be open to independent scrutiny.
In our analyses we considered the association between LDL-C reduction and individual ‘hard’ outcomes of all-cause mortality,
MI, and stroke rather than focusing on the composite outcomes reported in other meta-analyses. We did so because of the inherent problems associated with composite outcomes including the fact that they can be inconsistently defined and inadequately reported. In addition, it can be difficult to quantify the components within a composite. Mora et al. for example, examined the outcome ‘total CVD events’ in a statin trial called JUPITER and found that women had a significant reduction in revascularisations and unstable angina, but not in other components of the composite outcome, including stroke.
Patients’ and prescribers’ decision making may be influenced by seemingly impressive reductions in a composite outcome even though larger treatment effects may be associated with the less clinically important components.
Current clinical guidelines for prevention of CVD assume a log-linear relationship as I noted already. In contrast, our metaregression analysis yielded inconsistent results, thus we could neither prove nor disprove an association between the magnitude of LDL-C reduction and the size of treatment effect.
In our analysis, we reported significant clinical heterogeneity among the included trials. The trials differed in terms of the type and dose of statins, compliance with treatment, base-line levels of LDL-C, population characteristics, study duration, control treatment and definition of outcomes.
We also identified statistical heterogeneity. As a result of these limitations our calculations of ARR and RRR must be interpreted with caution, as do those of all prior analyses based on these trials. These issues have received little of no attention in prior meta-analyses of statins, many of which form the basis for clinical guidelines.
In addition, there has been some recent debate as to whether LDL-C is the most appropriate metric for measuring atherogenic risk; instead the concentration of Apoprotein B (ApoB) may be more reliable. However, LDL-C was the metric used in the design of the 21 statin trials we analysed, whereas few reported ApoB levels. In addition, ApoB is not routinely measured by doctors or used to inform decisions to prescribe statins.
Conclusion and future research
Our study found modest reductions in all-cause death, MI and stroke associated with statin use in terms of ARR. We argued that patients and their clinicians need to be clear on the absolute reduction in clinically important outcomes in order to make informed decisions about statins. In addition, we failed to demonstrate a linear relationship between the degree of LDL-C reduction and these clinical outcomes. Thus, our findings have implications for the individual patient and their doctors, for the development of future clinical guidelines and for policy makers and payers considering the opportunity cost of statins.