Review 2532 www.thelancet.com Vol 388 November 19, 2016 Interpretation of the evidence for the efficacy and safety of statin therapy Rory Collins, Christina Reith, Jonathan Emberson, Jane Armitage, Colin Baigent, Lisa Blackwell, Roger Blumenthal, John Danesh, George Davey Smith, David DeMets, Stephen Evans, Malcolm Law, Stephen MacMahon, Seth Martin, Bruce Neal, Neil Poulter, David Preiss, Paul Ridker, Ian Roberts, Anthony Rodgers, Peter Sandercock, Kenneth Schulz, Peter Sever, John Simes, Liam Smeeth, Nicholas Wald, Salim Yusuf, Richard Peto Summary This Review is intended to help clinicians, patients, and the public make informed decisions about statin therapy for the prevention of heart attacks and strokes. It explains how the evidence that is available from randomised controlled trials yields reliable information about both the effi cacy and safety of statin therapy. In addition, it discusses how claims that statins commonly cause adverse effects reflect a failure to recognise the limitations of other sources of evidence about the effects of treatment. Large-scale evidence from randomised trials shows that statin therapy reduces the risk of major vascular events (ie, coronary deaths or myocardial infarctions, strokes, and coronary revascularisation procedures) by about one-quarter for each mmol/L reduction in LDL cholesterol during each year (after the first) that it continues to be taken. The absolute benefits of statin therapy depend on an individual’s absolute risk of occlusive vascular events and the absolute reduction in LDL cholesterol that is achieved. For example, lowering LDL cholesterol by 2 mmol/L (77 mg/dL) with an eff ective low-cost statin regimen (eg, atorvastatin 40 mg daily, costing about £2 per month) for 5 years in 10 000 patients would typically prevent major vascular events from occurring in about 1000 patients (ie, 10% absolute benefi t) with pre-existing occlusive vascular disease (secondary prevention) and in 500 patients (ie, 5% absolute benefi t) who are at increased risk but have not yet had a vascular event (primary prevention). Statin therapy has been shown to reduce vascular disease risk during each year it continues to be taken, so larger absolute benefi ts would accrue with more prolonged therapy, and these benefi ts persist long term. The only serious adverse events that have been shown to be caused by long-term statin therapy—ie, adverse eff ects of the statin—are myopathy (defined as muscle pain or weakness combined with large increases in blood concentrations of creatine kinase), new-onset diabetes mellitus, and, probably, haemorrhagic stroke. Typically, treatment of 10 000 patients for 5 years with an eff ective regimen (eg, atorvastatin 40 mg daily) would cause about 5 cases of myopathy (one of which might progress, if the statin therapy is not stopped, to the more severe condition of rhabdomyolysis), 50–100 new cases of diabetes, and 5–10 haemorrhagic strokes. However, any adverse impact of these side-eff ects on major vascular events has already been taken into account in the estimates of the absolute benefits. Statin therapy may cause symptomatic adverse events (eg, muscle pain or weakness) in up to about 50–100 patients (ie, 0·5–1·0% absolute harm) per 10 000 treated for 5 years. However, placebo-controlled randomised trials have shown defi nitively that almost all of the symptomatic adverse events that are attributed to statin therapy in routine practice are not actually caused by it (ie, they represent misattribution). The large-scale evidence available from randomised trials also indicates that it is unlikely that large absolute excesses in other serious adverse events still await discovery. Consequently, any further fi ndings that emerge about the eff ects of statin therapy would not be expected to alter materially the balance of benefi ts and harms. It is, therefore, of concern that exaggerated claims about side-eff ect rates with statin therapy may be responsible for its under-use among individuals at increased risk of cardiovascular events. For, whereas the rare cases of myopathy and any muscle-related symptoms that are attributed to statin therapy generally resolve rapidly when treatment is stopped, the heart attacks or strokes that may occur if statin therapy is stopped unnecessarily can be devastating. Introduction Used appropriately, modern medical therapies have the potential to prevent a large proportion of the burden of cardiovascular disease. However, their appropriate use relies on the availability of robust data on safety and efficacy, as well as on a sound understanding of the interpretation and application of such evidence. Randomised controlled trials of adequate size are needed to be confi dent that any moderate benefits and any moderate harms of a treatment have been assessed sufficiently reliably.1–4 In certain circumstances, available evidence from randomised trials about the eff ects of a treatment may be limited (perhaps because it is deemed not possible or too diffi cult to do adequate trials).2 However, the particular context that this Review addresses is the appropriate interpretation of evidence about the safety and effi cacy of a treatment when randomised trials of it have been conducted in large numbers of many diff erent types of patient (as is the case for statin therapy), as well as the additional value of information from observational studies based on cohorts, health-care databases, or other sources.3–5 Not only have the limitations of observational studies4,6–9 often been underestimated when attributing adverse effects to treatment (such as misleading claims that statins cause side-effects in one-fi fth of patients10–12), but also the strengths of randomised trials with masked treatment allocation and systematic ascertainment of many Lancet 2016; 388: 2532–61 Published Online September 8, 2016 http://dx.doi.org/10.1016/ S0140-6736(16)31357-5 Clinical Trial Service Unit & Epidemiological Studies Unit and MRC Population Health Research Unit, Nuffield Department of Population Health, University of Oxford, Oxford, UK (Prof R Collins FRS, C Reith FRCP (Glasg.), J Emberson PhD, Prof J Armitage FRCP, Prof C Baigent FRCP, L Blackwell BSc, D Preiss PhD, Prof R Peto FRS); Ciccarone Center for the Prevention of heart disease, Division of Cardiology, Johns Hopkins University School of Medicine, Baltimore, MD, USA (Prof R Blumenthal MD, S Martin MD); MRC/BHF Cardiovascular Epidemiology Unit, Department of Public Health and Primary Care, University of Cambridge, Cambridge, UK (Prof J Danesh FMedSci); MRC Integrative Epidemiology Unit, University of Bristol, Bristol, UK (Prof G Davey Smith DSc); Department of Biostatistics and Medical Informatics, University of Wisconsin, Madison, WI, USA (Prof D DeMets PhD); Department of Medical Statistics (Prof S Evans MSc), Clinical Trials Unit (Prof I Roberts PhD), and Department of NonCommunicable Disease Epidemiology (Prof L Smeeth FRCGP), London School of Hygiene & Tropical Medicine, University of London, London, UK; Wolfson Institute of Preventive Medicine, Barts and The London School of Medicine and Dentistry, Queen Mary University of London, London, UK (Prof M Law FRCP, Prof N Wald FRS); The George Institute for Global Health (Prof S MacMahon FMedSci, Review www.thelancet.com Vol 388 November 19, 2016 2533 Prof B Neal PhD, Prof A Rodgers MBChB), and the National Health and Medical Research Council Clinical Trial Centre (Prof J Simes MD), University of Sydney, Sydney, Australia; International Centre for Circulatory Health & Imperial Clinical Trials Unit (Prof N Poulter FMedSci), and the International Centre for Circulatory Health, National Heart and Lung Institute (Prof P Sever FRCP), Imperial College London, London, UK; Center for Cardiovascular Disease Prevention, Brigham and Women’s Hospital, Harvard Medical School, Boston, MA, USA (Prof P Ridker MD); Centre for Clinical Brain Sciences, University of Edinburgh, Edinburgh, UK (Prof P Sandercock DM); FHI 360, University of North Carolina School of Medicine, University of North Carolina, Chapel Hill, NC, USA (Prof K Schulz PhD); and Population Health Research Institute, Hamilton Health Sciences and McMaster University, Hamilton, ON, Canada (Prof S Yusuf DPhil) Correspondence to: Prof Rory Collins, Nuffield Department of Population Health, University of Oxford, Richard Doll Building, Old Road Campus, Oxford OX3 7LF, UK [email protected] different types of adverse event have been underestimated for the reliable assessment of the safety and efficacy of treatment.3,9,13–15 This Review first considers the generic strengths and limitations of randomised trials and observational studies for assessing the effects of treatment, and then considers the specifi c evidence that is available on the efficacy and safety of statin therapy. It concludes by considering the public health implications of the failure to recognise the full benefits of using statin therapy and of the exaggerated claims that have been made about the rates of side-effects. Randomised controlled trials: strengths and weaknesses for assessing the benefi ts and harms of treatment (panel 1) Like-with-like comparisons within randomised trials The key strength of randomised controlled trials is that the process of randomisation results in groups of patients who diff er from each other only by the play of chance with respect to their risks of having all types of health outcome (ie, the randomised treatment groups are balanced with respect to both known and unknown risk factors, irrespective of whether or not these have been assessed).3,9,13–16 In addition, masking assignment of study treatment with a placebo minimises the differential assessment of adverse events between the study treatment groups following randomisation.17,18 Continued follow-up of all randomised patients (even if some stop taking their assigned treatment) maintains the like-with-like comparison produced by the randomisation process (since, for example, the patients who stop may differ between the randomised groups).3 Consequently, subject to statistical tests of the likely impact of chance, the observed diff erences in the rates of health outcomes between the randomly assigned patient groups within a trial (ie, intention-to-treat comparisons) can be attributed causally to diff erences in the study treatment. Information about a health outcome does not need to be obtained in the same way in the different randomised trials of an intervention (eg, different statin trials recorded muscle-related outcomes differently; appendix) for the comparisons of the rates of the outcome between the randomly allocated groups within each separate trial to provide unbiased assessments of any real effects of the treatment. However, biases can be introduced by making non-randomised comparisons between rates of events across different trials, not only because the outcome definitions may differ but also because the types of patient studied and the duration of follow-up may differ. Such between-trial comparisons may be seriously misleading,19 which is the reason why meta-analysis of randomised trials involves statistical methods that are based on the within-trial diff erences in a particular outcome.20,21 Robustness for detecting real treatment effects It has been suggested that ascertainment of adverse events in randomised trials may not be sufficiently specific or sensitive to detect adverse eff ects of treatment reliably.11,12,22–24 However, comparisons within randomised trials with unbiased ascertainment of outcomes between the treatment groups are robust against both overascertainment and under-ascertainment.25 For example, if the study treatment produced a 20% proportional decrease (or increase) in the rate of an outcome that occurred in 10% of control patients, then (as shown in table 1) the ability to detect such an eff ect in a randomised trial of 20 000 patients would not be much altered by the random addition of 10–20% of reported events that were not actually the outcome of interest (ie, false positives). Likewise, similar amounts of under-ascertainment (ie, false negatives), would not materially aff ect the ability to detect such effects in a trial. Moreover, these false positives would have little or no impact on estimates of the absolute eff ects, and the false negatives would have limited impact. The robustness of these within-trial randomised comparisons applies not only to the detection of beneficial effects, but also to the detection of harms that a treatment might cause (such as any musclerelated symptoms with statin therapy). It has been suggested that, when data for some types of health outcome are not available from all of the relevant randomised trials of a treatment, this will bias the assessment of its effects.11,26,27 However, although some of these trials may have recorded all types of health outcome reported by the participating patients, See Online for appendix Panel 1: Contribution of randomised trials for assessing treatment effects Like-with-like patient comparisons Randomisation results in groups of patients that differ from each other only by the play of chance with respect to their risks of suffering all types of health outcome, so observed diff erences in rates of health outcomes can generally be attributed causally to diff erences in study treatment. Like-with-like outcome comparisons Non-diff erential outcome ascertainment between the randomised treatment groups within a trial helps to minimise bias in the assessment of treatment eff ects. It can be enhanced by masking, which is likely to be of most value for symptomatic adverse events that are subjective. Robustness for detecting effects Comparisons within randomised trials with unbiased ascertainment of outcomes between treatment groups are robust for the detection of both benefi cial and harmful effects of treatment. Generalisability of evidence Randomised trials with different eligibility criteria that involve large numbers of many diff erent types of patient (ideally combined in meta-analyses of individual patient data) can provide reliable information about treatment effects that can be widely generalised to different circumstances . Review 2534 www.thelancet.com Vol 388 November 19, 2016 others may have only recorded those outcomes that were considered serious (typically defi ned as resulting in admission to hospital or death), perhaps because previous trials had ruled out material diff erences in less serious outcomes. If information on a particular outcome is not available from a randomised trial because it was not recorded that would not bias assessment of the eff ects of the treatment based on trials that did record the outcome. Also, if randomised trials have already reported results based on large numbers of occurrences of a particular outcome (as with muscle-related outcomes in statin trials; appendix) then the inclusion of any unpublished data from other trials that did record such outcomes is not likely to materially alter the assessment of the eff ect of the treatment on that outcome. Intention-to-treat analyses based on comparisons between all randomised patients, irrespective of whether they were adherent to their assigned study treatment (ie, stopped taking the active drug or, if assigned to the control group, started taking it), will tend to underestimate the effects produced by actually taking the treatment. However, rather than using potentially biased on-treatment comparisons among only those patients who took their assigned study treatment, more appropriate allowance can be made by applying an approximate estimate of the level of adherence to estimates of the treatment effects provided by the intention-to-treat comparisons.28 For example, if the average adherence to treatment assignment is two-thirds and the observed relative risk reduction (or increase) is 20%, then the adjusted estimate of the eff ect of actual use of the treatment would be a 30% proportional reduction (or increase). Specificity versus sensitivity of composite outcomes When there is clear evidence that a treatment produces effects on the incidence of diff erent types of outcome that are in the same direction and of similar magnitude (eg, the reductions in coronary events, strokes, and revascularisations produced by statin therapy29–34), combination of these outcomes in a composite outcome (eg, major vascular events in the statin trials) may well provide more robust assessments of the eff ects of the treatment because they involve larger numbers of events than for any of the constituent outcomes. That does not necessarily mean that—when deciding whether the absolute benefi ts of the treatment outweigh the harms for any particular type of patient (eg, offering statin therapy to individuals at lower vs higher risk of cardiovascular events)—equal weight should be given to the diff erent components of such composite outcomes. Instead, such analyses of composite outcomes may allow more reliable evidence to emerge about the effects of the treatment in diff erent circumstances (eg, the similar proportional reductions in major vascular events that have been found with statin therapy among many diff erent types of patient;29–34 figure 1). However, when a treatment has eff ects on different outcomes that diff er in direction, then their combination in a composite outcome will reduce the ability to detect these outcome-specific effects and limit generalisability of the analyses.35–38 For example, if a treatment reduces the incidence of ischaemic strokes but increases the incidence of haemorrhagic strokes (as appears to be the case for statin therapy31,39) then the adverse effect on haemorrhagic strokes may be missed by an assessment based on the composite of all stroke types since ischaemic strokes occur more commonly in most circumstances. By contrast, the assessment of the eff ects of the treatment on ischaemic and haemorrhagic strokes considered separately would not only be more sensitive to any benefits and harms, but it would also yield fi ndings that are more readily generalised to different settings (as with the use of aspirin in primary and secondary prevention40). Likewise, if treatment produced similar proportional reductions in vascular mortality and increases in nonvascular mortality, then the effect on the composite outcome of all-cause mortality would depend on the ratio of vascular to non-vascular deaths in a particular setting: the treatment would appear to be beneficial when vascular deaths predominated, but harmful when non-vascular deaths predominated. Instead, the application of the proportional reductions and increases in the separate causes of death to the expected rates of these outcomes in the population of interest would yield estimates of the absolute effects of treatment on each type of death and, thus, of the net effect on survival for particular types of individual (as is described later in the context of statin therapy).4,41 The lack of sensitivity and generalisability of composite outcomes can be even more problematic when they Active (n=10 000) Control (n=10 000) Relative reduction Absolute reduction Z score* True events 800 (8·0%) 1000 (10·0%) 20% 2·0% 4·9 Extra false outcomes (evenly distributed†) +10% +20% 890 (8·9%) 1090 (10·9%) 18% 980 (9·8%) 1180 (11·8%) 17% 2·0% 2·0% 4·7 4·5 Missing real outcomes (unevenly distributed†) –10% –20% 720 (7·2%) 640 (6·4%) 900 (9·0%) 800 (8·0%) 20% 20% 1·8% 1·6% 4·6 4·3 *For context, a Z score of 4·0 is equivalent to a p value of <0·0001. †The numbers of participants in whom true events would not have occurred would be slightly diff erent between the treatment groups, but this produces little imbalance in the numbers of false events that can be recorded among such patients in the two treatment groups when true events are relatively uncommon (as in this example). Consequently, false events have been approximately evenly distributed because they would not be aff ected by treatment assignment. By contrast, there would be fewer real outcomes to be missed in the active treatment group (since the treatment reduces the rate of the outcome), so the numbers of missed real outcome events are unevenly distributed between the treatment groups. Table 1: Illustrative example of the robustness to misclassified outcomes (false positives) and missing outcomes (false negatives) of within-trial comparisons of the effects of treatment in randomised controlled trials Review www.thelancet.com Vol 388 November 19, 2016 2535 involve very disparate outcomes. It has been suggested that the assessment of statin therapy should be based on the composite outcome of all serious adverse events of any kind (eg, mixing vascular outcomes that are known to be prevented by statin therapy with outcomes in gastrointestinal, genitourinary, neuropsychiatric, and other systems that may not be affected).11 A key problem with such an approach is that it can prevent the identification of both specific benefits and specific hazards of treatment. For example, analyses of specific outcomes among the 25 673 randomised patients in the THRIVE trial were able to detect that niacin therapy (nicotinic acid) is associated with unexpected hazards (ie, increases in serious infections and bleeding)42 that would have been missed by analyses based on a composite of adverse events (as in the case of the AIM-HIGH trial43 of niacin). Consideration of the eff ects of treatment on specific outcomes allows any diff erences in its eff ects to be determined, and its use can then be appropriately targeted at those who are likely to get more benefi t than harm. Figure 1: Similar proportional reductions in risks of major vascular events per mmol/L LDL cholesterol reduction in randomised trials of statin therapy among people with different presenting characteristics Adapted from CTT Collaboration website. RRs are plotted for the combined comparisons of MVE rate in randomised trials of routine statin therapy versus no routine statin therapy and of more versus less intensive statin therapy, weighted per 1·0 mmol/L LDL cholesterol reduction at 1 year. The size of the squares is proportional to the numbers of events recorded (ie, statistical information) in the particular comparison. CHD=coronary heart disease. RR=rate ratio. MVE=major vascular event. 0·5 0·75 1 1·25 Total number of MVEs Presenting characteristics Annual event rate in control arm (% per year) RR (CI) per 1 mmol/L reduction in LDL cholesterol p value for heterogeneity or trend 99% CI 95% CI <2·5 5256 4·3 ≥2·5 to <3·0 4182 4·0 ≥3·0 to <3.5 4604 4·1 ≥3·5 10 563 3·9 ≤65 13 623 3·6 >65 to ≤75 9211 4·6 >75 2123 5·5 Male 19 922 4·4 Female 5035 3·0 CHD 19 097 5·6 Non-CHD vascular 1529 3·7 None 4331 1·8 Type 1 diabetes 337 6·0 Type 2 diabetes 5621 5·1 No diabetes 18 862 4·0 Yes 13 939 4·5 No 10 471 3·5 Current smokers 5225 4·7 Non-smokers 19 728 3·9 <5% 421 0·6 ≥5 to <10% 1453 1·6 ≥10 to <20% 7810 3·5 ≥20 to <30% 9028 5·8 ≥30% 6245 9·8 All patients 24 957 4·0 Pre-treatment LDL cholesterol (mmol/L) Age (years) Sex History of vascular disease Diabetes Treated hypertension Smoking status 5-year MVE risk 0·78 (0·69-0·89) 0·77 (0·70-0·85) 0·76 (0·70-0·82) 0·80 (0·77-0·84) 0·78 (0·75-0·82) 0·79 (0·74-0·83) 0·87 (0·76-0·99) 0·78 (0·75-0·81) 0·84 (0·78-0·91) 0·79 (0·76-0·82) 0·83 (0·73-0·94) 0·75 (0·69-0·82) 0·77 (0·58-1·01) 0·80 (0·74-0·86) 0·78 (0·76-0·82) 0·80 (0·77-0·84) 0·77 (0·73-0·81) 0·79 (0·73-0·85) 0·79 (0·76-0·82) 0·62 (0·47-0·81) 0·69 (0·60-0·79) 0·79 (0·74-0·85) 0·81 (0·77-0·86) 0·79 (0·74-0·84) 0·79 (0·77-0·81) p=0·22 p=0·14 p=0·02 p=0·18 p=0·78 p=0·11 p=0·88 p=0·04 LDL cholesterol lowering worse LDL cholesterol lowering better For the CTT Collaboration website see www. cttcollaboration.org Review 2536 www.thelancet.com Vol 388 November 19, 2016 Value of meta-analyses of randomised trials Meta-analyses of randomised trials may be required when the eff ects of a treatment on some particular outcome are likely to be moderate and too few cases of it have occurred in any individual trial to assess the effects sufficiently reliably.3,20,44–47 For example, table 2 shows that a metaanalysis of 100 000 randomised patients (as is available for statin therapy33) would have 90% statistical power at p=0·01 to detect an absolute excess of 0·5% in the incidence of events that occur in 5% of patients in the control group (ie, a 10% proportional increase) and an absolute excess of 1% for events that occur in 20% of patients in the control
Science News from research organizations Study explaining side effects of statins finds drug can have unexpected benefits Statins can be beneficial during heart attacks and cancer cell metastasis Date: March 19, 2019 Source: University of Toledo Summary: By suppressing the activity of key cellular receptors called G protein-coupled receptors (GPCRs) and their interacting partners called G proteins, statins have the potential to alter various bodily functions controlled by this important pathway, according to researchers. Share: University of Toledo. Note: Content may be edited for style and length. Journal Reference: Mithila Tennakoon, Dinesh Kankanamge, Kanishka Senarath, Zehra Fasih, Ajith Karunarathne. Statins Perturb Gβγ Signaling and Cell Behavior in a Gγ Subtype Dependent Manner. Molecular Pharmacology, 2019; 95 (4): 361 DOI: 10.1124/mol.118.114710
Science News from research organizations Statins have low risk of side effects Date: December 10, 2018 Source: American Heart Association Summary: Cholesterol-lowering statin drugs are associated with a low risk of side effects. The benefits of statin therapy for most people outweigh the risks. Share: Materials provided by American Heart Association. Note: Content may be edited for style and length. Journal Reference: Connie B. Newman et al. Statin Safety and Associated Adverse Events: A Scientific Statement From the American Heart Association. Circulation: Arteriosclerosis, Thrombosis and Vascular Biology, 2018 DOI: 10.1161/ATV.0000000000000073
Statins are more effective for those who follow the Mediterranean diet Date: December 21, 2018 Source: Istituto Neurologico Mediterraneo Neuromed I.R.C.C.S. Summary: For those who have already had a heart attack or a stroke, the combination of statins and Mediterranean Diet appears to be the most effective choice to reduce the risk of mortality, especially from cardiovascular causes. Share: Istituto Neurologico Mediterraneo Neuromed I.R.C.C.S.. Note: Content may be edited for style and length. Journal Reference: Marialaura Bonaccio, Augusto Di Castelnuovo, Simona Costanzo, Mariarosaria Persichillo, Amalia De Curtis, Chiara Cerletti, Maria Benedetta Donati, Giovanni de Gaetano, Licia Iacoviello. Interaction between Mediterranean diet and statins on mortality risk in patients with cardiovascular disease: Findings from the Moli-sani Study. International Journal of Cardiology, 2018; DOI: 10.1016/j.ijcard.2018.11.117
