WHAT IS ALREADY KNOWN ON THIS TOPIC
WHAT THIS STUDY ADDS
In the long-term 20-year follow-up of participants formerly assigned atorvastatin in the Anglo-Scandinavian Cardiac Outcomes Trial, there remained a significant reduction in non-fatal myocardial infarction and fatal coronary heart disease, total coronary events and cardiovascular mortality compared with those assigned placebo, despite evidence that post-trial lipid levels were similar in the two groups.
HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY
Introduction
Cardiovascular (CV) disease remains a leading global health concern, requiring continuous exploration of interventions that moderate its impact. Data from the Anglo-Scandinavian Cardiac Outcomes Trial (ASCOT) have contributed to the evidence base supporting the role of statins in the primary prevention of CV disease.
ASCOT was a prospective randomised controlled trial of two different blood pressure-lowering strategies in 19 342 hypertensive subjects with three additional risk factors for CV disease—ASCOT Blood Pressure-Lowering Arm (BPLA). In a factorial design, a subgroup of participants (n=10 305) with total cholesterol <6.5 mmol/L were re-randomised in a double-blind manner to receive either atorvastatin 10 mg or placebo—referred to as the lipid-lowering arm (LLA). In ASCOT LLA, atorvastatin compared with placebo significantly reduced the incidence of several major CV end points, including myocardial infarction (MI) and stroke, over an average follow-up period of 3.3 years.1
Two years after the termination of the trial, and despite the fact that most participants from both original trial arms (atorvastatin or placebo) were receiving atorvastatin, the risk reductions for CV end points in favour of atorvastatin remained unchanged.2 These observations prompted further investigation into the long-term legacy effects of atorvastatin. Twelve years after the trial, we reported only on mortality outcomes, which suggested a legacy benefit in participants previously assigned atorvastatin. Deaths from all CV causes were significantly reduced by 15%, and although there was no significant reduction in coronary deaths, they were reduced by 22%. However, mortality alone does not account for the majority of the health burden of CV disease, making non-fatal events of great interest
Twelve years after the trial, we reported on only mortality outcomes, which suggested a legacy benefit in participants formerly assigned atorvastatin. Deaths from all CV causes were significantly reduced by 15%, and although there was no significant reduction in coronary deaths, they were reduced by 22%.3 However, mortality alone does not account for the majority of the health burden of CV disease, making non-fatal events of great interest.
In these analyses, conducted approximately 20 years from the initial randomisation and 17 years postclosure of the LLA, we have incorporated data on non-fatal events and mortality from NHS England and Public Health Scotland. This provides a comprehensive assessment of the potential long-term legacy effects of atorvastatin treatment and evaluates the association between achieved cholesterol levels during the trial and long-term CV outcomes.
Methods
ASCOT: lipid-lowering arm and participant profile
Briefly, ASCOT was designed to compare two antihypertensive treatment strategies: amlodipine, with perindopril added as necessary (amlodipine-based) and atenolol, with bendroflumethiazide added as necessary (atenolol-based) in the BPLA. By factorial design, those patients with a total cholesterol level <6.5 mmol/L were further randomised to atorvastatin 10 mg or placebo in the lipid-lowering arm (LLA).4
This population consisted of hypertensive men and women, aged between 40 and 79 years at randomisation, with at least three additional risk factors for CV disease. Participants had no history of coronary heart disease (CHD) events, treated angina, stroke or transient ischaemic attack (TIA) within the 3 months prior to randomisation. The primary outcome was non-fatal MI and fatal CHD. Patients were initially recruited between February 1998 and May 2000, with most participants coming from family practices. In the Nordic countries, 686 family practices were randomised, while in the UK and Ireland, the majority of participants were referred from family practices to regional centres for the study. Total cholesterol and other lipid markers were measured throughout the trial, initially 6 weeks after randomisation, followed by measurements at 3 months, 6 months and then 6 monthly thereafter until the end of the trial. A total of 10 305 participants were randomised to the LLA, which was prematurely stopped in September 2002 at the recommendation of the Data Safety and Monitoring Board due to a highly significant reduction in the primary end point, along with reductions in several other CV end points in favour of atorvastatin.1
At the end of the trial, all participants were offered atorvastatin 10 mg. Approximately two-thirds of participants from each arm of the trial took atorvastatin for the following 2 years, at which point lipid profiles were identical in both groups.2
ASCOT-Legacy cohort
Long-term follow-up of participants originally recruited in the Nordic countries was not possible using electronic records.
Out of 10 305 ASCOT LLA participants, 4605 participants from the UK were originally randomised to either atorvastatin (2317) or placebo (2288). All of these participants were followed up until the end of the LLA, during which 372 died. Of the remaining participants, 3920 who consented were flagged for death and hospitalisation with NHS England and Public Health Scotland (figure 1). In this report, we have used all reported deaths or hospitalisations up until 31 January 2019 (online supplemental file 2).
Figure 1Diagram of LLA of the ASCOT-Legacy cohort. ASCOT, Anglo-Scandinavian Cardiac Outcomes Trial; CHD, coronary heart disease; CV, cardiovascular; HF, heart failure; LLA, lipid-lowering arm; MI, myocardial infarction.
A team of two physicians independently adjudicated all deaths using prespecified criteria that were consistent with the definitions used during the in-trial period. Outcomes for non-fatal events were obtained from electronic hospital records, using previously validated code lists and classified according to the original trial outcomes.
The primary objective was to compare the time to the first occurrence of the following CV outcomes between those assigned atorvastatin or placebo: non-fatal MI and fatal CHD, total coronary events (defined as non-fatal MI, fatal CHD, non-fatal and fatal heart failure (HF) and the development of angina), fatal and non-fatal stroke, fatal and non-fatal HF, total CV events, CV and all-cause mortality. A secondary objective was to evaluate in-trial cholesterol levels to identify predictors of the legacy effect, should it be apparent.
In 2011–12, 9 years after the closure of the LLA, a subgroup of living participants (n=2156) from 18 regional centres in the UK that had participated in the original trial was identified as part of the ASCOT-10 substudy. These participants were invited to complete a questionnaire requesting information on their health and current medications. Of these, 1142 had participated in the LLA. An additional subgroup (n=239 of those randomised to the LLA) was recalled to the clinic for lipid measurements.
This report investigates whether assignment to atorvastatin in the ASCOT-Legacy study conferred any long-term benefits on CV or other outcomes, and examines whether the cholesterol levels achieved during the trial were associated with CV outcomes over the 20-year follow-up period.
Statistical methods
We analysed all UK participants in ASCOT. Where linkage was not possible, primarily due to participants not consenting to long-term mortality follow-up, we censored participants at the end of the trial follow-up period. For all analyses, censoring occurred at the time of death, withdrawal of consent or the end of the linkage period.
To compare the incidence of end points between treatment groups, survival analyses were conducted using Cox proportional hazards models, with both crude and corrected models reported. Analyses were adjusted for potential confounders, including baseline age, sex, self-reported ethnicity, socioeconomic status, systolic blood pressure, smoking habit, baseline cholesterol, body mass index, history of diabetes, allocation to blood pressure-lowering medication and a history of vascular diseases (coronary, cerebral and peripheral).
Sensitivity analyses regarding lipid-lowering effects were performed from baseline to the end of the LLA period (3.3 years) and were repeated for the post-trial period, from the end of the ASCOT trial to the end of the legacy period, in patients who survived the ASCOT trial.
In an additional sensitivity analysis, we repeated the analysis using competing risks regression models, which accounted for mortality and other causes as competing risks to the primary outcomes.5
The impact of the 3-year in-trial low-density lipoprotein (LDL)-cholesterol (defined as the mean LDL-cholesterol calculated from samples obtained 6 months postrandomisation until trial closure) on each long-term outcome per mmol/L decrease in LDL in patients allocated to atorvastatin (n=2151) was assessed through both unadjusted and adjusted analyses.
Kaplan-Meier survival curves were generated, and log-rank tests were employed to compare atorvastatin and placebo groups. Stata V.18 was used for all analyses, and GraphPad Prism V.10 was employed for data visualisation.
We prepared this report in accordance with the guidelines outlined in the Reporting of Studies Conducted Using Observational Routinely Collected Health Data Statement.6
Results
A total of 8580 participants from England, Wales and Scotland were included. Of these, 4605 participants were randomised to either atorvastatin or placebo. The participants were followed for a median duration of 17 years (IQR 9–19), with a maximum follow-up of up to 21 years (figure 1).
The UK participants were well-matched across the allocated groups. On average, participants were 64 years of age (SD 8) at trial entry, with the majority being male (88%) and having left education before 16 years of age (79%). At baseline, the mean total cholesterol was 5.5 mmol/L (SD 0.8). Participants had a history of diabetes (27%), stroke or TIA (10%) or other vascular diseases (22%) (table 1).
Table 1Baseline characteristics of those in the ASCOT-Legacy cohort randomised by atorvastatin or placebo
Figure 2 illustrates the Kaplan-Meier curves and log-rank test results, demonstrating the cumulative incidence of CV and coronary events among those allocated to atorvastatin versus placebo in the ASCOT-Legacy population.
Figure 2Kaplan-Meier curves showing the cumulative incidence of cardiovascular (CV) events and mortality among those allocated to atorvastatin versus placebo in the Anglo-Scandinavian Cardiac Outcomes Trial (ASCOT)-Legacy population. Kaplan-Meier curves show the association of atorvastatin versus placebo allocation with (A) non-fatal myocardial infarction and fatal coronary heart disease (CHD), (B) total coronary events, (C) fatal and non-fatal heart failure (HF), (D) fatal and non-fatal stroke, (E) total CV events and procedures, (F) CV mortality and (G) all-cause mortality in the ASCOT-Legacy population.
Figure 3 shows the adjusted HRs associated with assignment to atorvastatin compared with placebo. Participants allocated to atorvastatin exhibited a significant reduction in non-fatal MI and fatal CHD events (HR 0.81, 95% CI 0.69 to 0.94, p=0.006), total coronary events (HR 0.88, 95% CI 0.80 to 0.98, p=0.017) and CV mortality (HR 0.86, 95% CI 0.74 to 0.99, p=0.048). However, no significant reduction was observed in HF, strokes, total CV events and all-cause mortality.
Figure 3Morbidity and mortality outcomes related to lipid-lowering treatment from the baseline until the end of the legacy period (20 years). The figure shows the adjusted HRs associated with assignment to atorvastatin compared with placebo. CHD, coronary heart disease; CV, cardiovascular; HF, heart failure; MI, myocardial infarction. *Per 1000 person-years. Adjusted for baseline age, sex, ethnicity, socioeconomic status (education), systolic blood pressure, body mass index, diabetes status, smoking habit, baseline cholesterol, allocation to blood pressure-lowering medications and history of vascular diseases (coronary, cerebral and peripheral).
In sensitivity analyses, we initially examined the association of atorvastatin allocation versus placebo from baseline to the end of the LLA period (3.3 years) within the UK population (n=4605). The results revealed evidence of a significant decrease in non-fatal MI and fatal CHD events (HR 0.62, 95% CI 0.41 to 0.93, p=0.021), total coronary events (HR 0.73, 95% CI 0.56 to 0.97, p=0.029) and total CV events and procedures (HR 0.73, 95% CI 0.60 to 0.90, p=0.003) (online supplemental table S1). Furthermore, we assessed the effect of atorvastatin compared with placebo during the trial on post-trial outcomes (15 years) among the UK population who survived the ASCOT trial. There were no significant reductions in total coronary events (HR 0.89, 95% CI 0.80 to 1.00, p=0.053) and CV mortality (HR 0.86, 95% CI 0.73 to 1.02, p=0.075) despite modest reductions (online supplemental table S2).
In those participants assigned atorvastatin in the trial, mean LDL-cholesterol from 6 months from the time of allocation until the end of ASCOT LLA was strongly associated with long-term CV outcomes (table 2). Each unit decrease in LDL was associated with adjusted HRs of non-fatal MI and fatal CHD events (0.69, 95% CI 0.57 to 0.85, p<0.001), total coronary events (0.70, 95% CI 0.61 to 0.79, p<0.001), non-fatal and fatal HF (0.68, 95% CI 0.57 to 0.81, p<0.001), non-fatal and fatal stroke (0.74, 95% CI 0.59 to 0.92, p=0.006), total CV events and procedures (0.74, 95% CI 0.66 to 0.81, p<0.001), CV mortality (0.66, 95% CI 0.55 to 0.81, p<0.001) and all-cause mortality (0.81, 95% CI 0.71 to 0.90, p<0.001).
Table 2Morbidity and mortality outcomes from baseline until the end of the legacy period (20 years) in those originally assigned atorvastatin, expressed per 1 mmol/L decrease in mean LDL from 6 months after baseline until the end of LLA (n=2317)
To investigate the potential interaction between age and the outcomes of interest, we performed a sensitivity analysis using an interaction term within a Cox regression model. Participants were stratified into two age groups: those younger than 65 years (n=2337) and those aged 65 years and older (n=2268). This approach allowed us to evaluate whether the effects observed in the primary analysis were consistent across different age categories. On detailed examination, no statistically significant differences in outcomes were observed between the two groups.
The subgroups of participants recalled during the follow-up period, approximately 9 years after the trial’s conclusion, were found to be generally representative of the overall ASCOT-Legacy population (online supplemental table S3). Cholesterol levels, assessed 9 years post-trial closure, were similar between individuals formerly assigned atorvastatin and placebo (online supplemental table S4). At this time point, 72.6% of those formerly assigned atorvastatin were taking a statin compared with 70.2% of those formerly assigned placebo. Further information on lipid-lowering medications in participants previously assigned to the two treatment groups is available in online supplemental table S5.
In sensitivity analyses, we repeated the analysis using competing risks regression models, which considered mortality and other causes as competing risks to our primary outcomes. The results were consistent with our previous findings.
Discussion
These results confirm and extend our previous findings on the long-term benefits of atorvastatin.3 We previously reported that all CV deaths were significantly reduced in those formerly assigned atorvastatin compared with placebo after 16 years of follow-up and 11 years after the closure of ASCOT-LLA. In the current report, we have incorporated data from digital health records over a prolonged period of follow-up to provide a more comprehensive assessment of non-fatal and fatal CV outcomes. The observations have confirmed the long-term benefits of atorvastatin, particularly on MI and CV death.
Several lipid-lowering trials have now confirmed that following an intervention period of only a few years, despite both arms of the original trial subsequently being treated with statins, the risk reductions in favour of statins reported from the trial are maintained, although slightly diminished, for up to 20 years of follow-up.7–11
The so-called legacy effects of former treatment have most commonly been reported in trials of lipid-lowering with statins.7 11 However, the phenomenon has also been observed in trials of antihypertensive drugs and oral hypoglycaemic agents,12–14 suggesting that it may be a more generalised effect, not limited to the specific actions of individual drugs. In most of these long-term follow-up studies, differences in cholesterol, blood sugar and blood pressure observed during the trials—which accounted for the differences in trial outcomes—were minimised in the post-trial period, and therefore could not explain the long-term benefits.
Several hypotheses have been suggested to explain these findings, including plaque stabilisation by statins during the trial, off-target effects of statins and epigenetic effects. However, the fact that a similar phenomenon of long-term legacy has been reported following trials of blood pressure and glucose-lowering treatments suggests that the explanation may be more generalised, although it remains unknown at present. A possible explanation is that those initially randomised to placebo, despite being exposed to statins post-trial, never fully catch up with those originally assigned to statin treatment, which is why the Kaplan-Meier curves never converge.
We have also demonstrated that the achieved cholesterol levels during the trial in participants receiving statins are strongly associated with long-term CV outcomes, including coronary events, HF and total CV events.
In the original trial, atorvastatin was shown to reduce stroke incidence by 27%,1 but we found no evidence of a long-term benefit on this end point. The pathophysiology of stroke is complex and involves multiple causes, several of which are unrelated to atherosclerotic vascular disease. Therefore, the lack of a demonstrated long-term benefit from statin treatment is perhaps unsurprising.
There are limitations to our analyses. Follow-up of participants from the Nordic countries using electronic records was not possible, so data are only available for approximately half of the patients originally randomised in the LLA. However, we have no reason to believe that the current analyses are not representative of the original trial population. In the original trial, all non-fatal events were adjudicated by an independent panel. However, data from electronic hospital records of post-trial events provided by NHS England and Public Health Scotland were not available for adjudication. We acknowledge that non-fatal outcomes identified using the International Classification of Diseases, Tenth Revision codes have their limitations; however, the adjudicated mortality data support the results of the combined assessment of fatal and non-fatal outcomes.
Two years after the trial closure, during which time approximately two-thirds of participants from each arm of the trial were taking atorvastatin, lipid levels were identical in both arms (LDL-cholesterol 2.4 mmol/L). In addition, we have limited cholesterol measurements and lipid treatment data from the substudy samples, collected 9 years post-trial. However, these are likely to be representative of the trial population. Surprisingly, despite evidence from questionnaires showing that most participants were taking lipid-lowering drugs by this time, LDL-cholesterol levels had risen to 3.4 mmol/L but remained similar between those previously assigned atorvastatin and those assigned placebo. In this study, we examined a number and variety of outcomes. We did not adjust for multiplicity and so we acknowledge that some inflation of overall type I error may have occurred.
The findings from this and other long-term follow-up trials provide robust evidence for the benefits of early intervention in preventing CV outcomes. Delays in initiating statin treatment in individuals at risk of atherosclerotic CV disease result in a continued excess risk of CV events compared with those whose treatment is started earlier.
Data availability statement
Data may be obtained from a third party and are not publicly available. A data sharing statement was included in the original submission. However, since the trial recruited participants in 1999–2000, this predates the requirement for data sharing.
Ethics statements
Patient consent for publication
Not applicable.
Ethics approval
We obtained approval from the South East Scotland Research Ethics Committee (18/SS/0016), the Health Research Authority Confidentiality Advisory Group (18/CAG/0044), the Independent Group Advising on the Release of Data of NHS England and the Public Benefit and Privacy Panel for Health and Social Care of Public Health Scotland. The study adhered to good clinical practice guidelines and the Declaration of Helsinki. The protocol, along with subsequent amendments, underwent review and approval by central and regional ethics review boards in the UK and national ethics and statutory bodies in Ireland and the Nordic countries (Sweden, Denmark, Iceland, Norway and Finland).
Participants gave informed consent to participate in the study before taking part.
Acknowledgments
We are grateful to NHS England, Public Health Scotland, the ASCOT-10 investigators and to participants in the Anglo-Scandinavian Cardiac Outcomes Trial.
