Reappraisal of statin primary prevention trials: implications for identification of the statin-eligible primary prevention patient

Abstract

Aims

Identification of patients eligible for primary prevention statin therapy is complex, often relying upon risk algorithms that diverge internationally. Our goal was to develop a simpler global definition of statin-eligible primary prevention patients.

Methods and results

Randomized clinical trials (RCTs) cited in North American and European dyslipidaemia guidelines justifying primary prevention statins for cardiovascular risk reduction were critically reappraised according to eligibility criteria and characteristics of actual enrollees. Statin-eligibility based on meeting minimal enrolment criteria vs. risks calculated using the Framingham risk score, the pooled cohort equation, and the systematic coronary risk estimate two were contrasted. Patient scenarios meeting minimal RCT eligibility criteria seldom attained high enough 10 year risk of events according to the algorithms tested and thus would not be eligible for statin therapy. Overall, enrollees were 63.9 ± 8.9 years (mean ± SD) with low density lipoprotein-cholesterol (LDL-C) 3.53 ± 0.91 mmol/L. Enrollees in trials studying the lowest LDL-C levels were generally older and had additional risk factors.

Conclusion

Results of primary prevention RCTs justify treatment of more subjects and lower risk subjects than current risk algorithm-based guidelines. Based on a synthesis of RCT inclusion/exclusion criteria and the characteristics of enrollees, we propose that a statin-indicated primary prevention subject is one who is 40 to 70 years with a low density lipoprotein-cholesterol (LDL-C) ≥ 3.0 mmol/L or is 55 to 80 years with LDL-C ≥ 1.8 mmol/L and additional risk factors.

Structured Graphical Abstract

Lay Summary

Based on available primary prevention clinical trials, their inclusion criteria and the characteristics of enrollees, a simplified definition of a statin-eligible primary prevention patient is one who is:

• 40–70 years old with LDL-C ≥ 3.0 mmol/L or

• 55–80 years old with LDL-C ≥ 1.8 mmol/L and additional risk factors.

Statin, Primary prevention, Framingham risk score, Pooled cohort equation, Systematic coronary risk estimate, Randomized clinical trials

Translational Perspective

Although most guidelines endorse the use of risk algorithms to identify subjects who might benefit from statin therapy, this method tends to focus on subjects with risk higher than subjects who actually meet eligibility criteria for successful primary prevention statin trials. We propose that subjects eligible for these trials should be offered statins and, in general, such subjects are 40 to 70 years with LDL-C ≥ 3.0 mmol/L or 55 to 80 years with both LDL-C ≥ 1.8 mmol/L and additional risk factors.

Introduction

Once randomized clinical trials (RCTs) affirm the benefit of an intervention, effective knowledge translation ensures that the general population benefits. Implementing therapy to patients similar to those enrolled in RCTs is a basic principle of medical practice. But for pragmatic reasons, guideline recommendations are sometimes modified to simplify identification of eligible patients in the real world. For secondary prevention of low density lipoprotein-cholesterol (LDL-C)-related cardiovascular (CV) risk, multiple RCTs clearly indicate that established atherosclerotic cardiovascular disease (ASCVD) is a statin-indicated condition.^1–4 Patient characteristics such as history of an acute coronary syndrome or other clinical features may influence whether a patient should be offered more intensive therapy than a statin alone, but the fundamental core principle still holds that ASCVD is a statin-indicated condition.

In contrast, there is no analogous simple way to identify a primary prevention patient who may benefit from statin treatment. Current guidelines recommend primary prevention in diverse ways. For example, treatment is recommended for patients fulfilling criteria for heterozygous familial hypercholesterolemia or those with an LDL-C ≥ 5.0 mmol/L despite lack of dedicated RCTs of ASCVD outcomes for these indications. In contrast, other recommendations directly follow from RCTs (e.g. for patients with diabetes or chronic kidney disease [CKD]). But frequently, recommendations in primary prevention patients are based upon a predicted risk using diverse algorithms.^1–6 The diversity of algorithms leads to variable risk stratification in different international jurisdictions.⁷ Moreover, many practitioners bypass this risk stratification step entirely, thereby introducing further heterogeneity in identification of statin-eligible patients and initiation of treatment. Finally, these approaches focus treatment of higher risk patients in order to optimize numbers needed to treat (NNTs), but fewer events in absolute terms may be prevented at the population level.^8,9 For instance, in a cohort of patients with pre-mature coronary artery disease (CAD), only 17% had received statin therapy prior to presentation with CAD, representing a major missed opportunity to reduce ASCVD burden.¹⁰

The purpose of this study was to develop, through reappraisal of primary prevention RCTs, an alternative method to reduce heterogeneous identification of patients who may benefit from statin therapy while prioritizing the importance of the causal exposure of LDL-C in the development of atherosclerosis.¹¹

Methods

Inclusion and exclusion criteria

The RCTs evaluating the effects of statin vs. placebo and cited in support of the primary prevention recommendations for general populations in Canada, the United States, and in Europe were identified from the respective guidelines publications.^12–18 Trials were excluded if they provided the basis for non-algorithm-dependent recommendations for primary prevention with statins (i.e. most patients with diabetes [CARDS-Collaborative Atorvastatin Diabetes Study]⁶ and non-dialysis/non-transplant subjects with CKD [SHARP-Study of Heart and Renal Protection]⁵). Similarly, a publication published after the official guidelines and pertaining exclusively to persons living with human immunodeficiency virus (HIV) was excluded.¹⁹ Finally, a recent study undertaken in both the primary and secondary prevention setting but restricted to statin intolerant patients and using a non-statin (bempedoic acid) was also excluded.²⁰

One trial enrolled only male patients (West of SCOtland Primary Prevention Group, WOSCOPS).¹² Another enrolled only elderly patients (age 70–82 y) (PROspective Study of Pravastatin in the Elderly at Risk, PROSPER).¹⁴ A third trial enrolled only patients with elevated high sensitivity C-reactive protein (hs-CRP, Justification for the Use of statins in Prevention: an Intervention Trial Evaluating Rosuvastatin, JUPITER).¹⁷ These three trials were retained for our analysis because there are no non-algorithm-based recommendations per se for either males, those with advanced age or those with elevated hs-CRP. Accordingly, our focus was on the following set of trials, listed chronologically: WOSCOPS,¹² AFCAPS/TEXCAPS (Air Force/Texas Coronary Atherosclerosis Prevention Study),¹³ PROSPER,¹⁴ ASCOT-LLA (Anglo-Scandinavian Cardiac Outcomes Trial—Lipid Lowering Arm),¹⁵ MEGA (Management of Elevated Cholesterol in the Primary Prevention Group of Adult Japanese),¹⁶ JUPITER,¹⁷ and HOPE-3 (Heart Outcomes Prevention Evaluation-3).¹⁸

Statistical analyses

Eligibility inclusion criteria were extracted as well as demographics of participants enrolled in the trials. These were reported either as percentages or as mean ± standard deviation. Medians and interquartile ranges were converted to means and standard deviations using the Luo and Wan methods respectively.^21,22 The Cochrane method was used to combine means and standard deviations.²³ Percentiles were calculated in standard fashion for the thresholds of 2.5%, 25%, 75%, and 97.5% for selected variables. The 10 year risk was calculated using the Framingham risk score (FRS) used in Canada²⁴ (without incorporating a fixed multiplier for family history of pre-mature cardiovascular disease), the pooled cohort equation (PCE) used in the USA,²⁵ and the systematic coronary risk estimate 2 (SCORE2) equations used in different regions of Europe.^26,27 The SCORE2 equations are tailored and calibrated for low, moderate, high, and very high risk regions as described.² The definitions of low, moderate, and high risk of events per 10 years were those proposed in the respective guidelines employing the specific algorithms.

To contrast, the risk-based vs. the trial eligibility-based identification of statin-eligible patients, we created cases that just met the inclusion criteria of trials (i.e. lowest age, lowest total cholesterol [TC], highest high-density lipoprotein-cholesterol [HDL-C], etc.), which would translate into the lowest calculated risk using the diverse algorithms. If a study’s eligibility criteria did not stipulate the lowest TC or highest HDL-C compatible with inclusion, the lowest and highest values reported in another primary prevention RCT were used, respectively. Although HDL-C levels were not considered inclusion criteria for HOPE-3, the trial stated that low HDL-C levels for both men and women could be counted as a potential, lipid-related risk factor that would allow enrolment. If a study did not stipulate the lowest LDL-C value compatible with trial enrolment, we imputed it using the Friedewald equation assuming a triglyceride (TG) of 1.5 mmol/L. The scenarios excluded family history of pre-mature ASCVD, smoking, CKD, or diabetes. If hypertension was an inclusion requirement, we modelled treated hypertension assuming a systolic blood pressure (SBP) of 120 mmHg.

Results

Eligibility criteria of primary prevention trials

Seven primary prevention RCTs were included in our analysis based on the above criteria (n = 67 648 subjects). Table 1 summarizes the main eligibility criteria of these RCTs. All trials were open to males and females with the exception of WOSCOPS, which enrolled only males. ASCOT-LLA and MEGA did not stipulate a requirement for females to be older than males to be eligible, nor to have an enhanced risk burden as in HOPE-3. The youngest age eligible for at least one of the RCTs was 40 years for both males and females (MEGA) while the oldest was 82 years (PROSPER). Where stipulated, the lowest eligible TC was 4.0 mmol/L (PROSPER); the highest HDL-C was 1.16 mmol/L for males and 1.22 mmol/L for females, both from the AFCAPS/TEXCAPS trial. Notably, HOPE-3 considered an HDL-C normal if ≥0.9 or ≥1.3 mmol/L for men and women, respectively. The presence or absence of hypertension diagnosis and treatment were diverse across trials but the general goal was exclusion of subjects with poorly controlled hypertension. Some of the trials were not exclusive to subjects without established ASCVD. For example, PROSPER excluded ASCVD features occurring within 6 months of enrolment and ASCOT-LLA allowed enrolment of subjects with peripheral arterial disease, previous stroke or transient ischaemic attack. These ASCVD features were excluded from any case modelling.

Table 1

Inclusion parameters for the main, primary prevention statin trials

Variable	WOSCOPS12	AFCAPS/TexCAPS13	PROSPER14a	ASCOT-LLA15b	MEGA16	JUPITER17	HOPE-318c
N	6595	6605	5804	10 305	7832	17 802	12 705
Men (%)	100%	85%	48%	81%	32%	62%	54%
Women (%)	0%	15%	52%	19%	68%	38%	46%
Age (years) Men Women	45–64	45–73 55–73 (post-menopausal)	70–82	40–79	40–70	≥50 ≥60	≥55 ≥65 (≥60 for women if had at least 2 RFs)
History of vascular disease	Excluded history of MI (stable angina not hospitalized within previous 12 months eligible)	Excluded prior history, signs, or symptoms of definite MI, angina, claudication, cerebrovascular accident, or TIA	Excluded recent vascular disease ≤6 months before study entry	See footnote.b	Excluded history of CHD or stroke	Excluded history of CVD	Excluded participants with CVD
Hypertension		Excluded uncontrolled hypertension
Diabetes						Excluded diabetics
BMI		Excluded if weight >50% than desirable for height
Systolic BP (mmHg)				≥160 (for untreated) ≥140 (for treated)		Excluded >190
Diastolic BP (mmHg)				≥100 (for untreated) ≥90 (for treated)		Excluded >100
Total cholesterol (mmol/L)		4.65–6.82	4.0–9.0	≤6.5	4.69–6.98
LDL (mmol/L)	≥4.5–6.0	3.36–4.91 (3.23–3.34 included when TC to HDL ratio >6.0)				<3.4
HDL (mmol/L) Men Women		≤1.16 ≤1.22
Triglyceride (mmol/L)		≤4.52	<6.0	<4.5		<5.6
Fasting glucose (mmol/L)						Exclude >7.0
Glycated haemoglobin (%)		>10% (20% ULN) excluded
Serum creatinine (mmol/L)						Excluded >176.8 µmol/L
hs-CRP (mg/L)						≥2.0

^aPROSPER—participants had to have either pre-existing vascular disease (coronary, cerebral, or peripheral) > 6months prior to study entry or at least 1 risk factor for vascular disease: Current smoker, hypertension (currently receiving drug treatment); known diabetes mellitus or fasting glucose > 7 mmol/L.

^bASCOT-LLA—participants had to have at least 3 risk factors for CVD: LV hypertrophy, other specified abnormalities on electrocardiogram, type 2 diabetes, peripheral arterial disease, previous stroke or transient ischaemic attack, male sex, age 55 years or older, microalbuminuria or proteinuria, smoking, ratio of plasma total cholesterol to HDL-cholesterol of 6 or higher, or pre-mature family history of CHD.

^cHOPE-3—no mandated BP or lipid levels. Participants had to have at least one CV risk factor: Waist/hip ratio ≥ 0.85 in women and ≥ 0.90 in men, history of current or recent smoking (regular tobacco use within 5 years), low HDL-C (HDL-C < 1.0 mmol/L in men and <1.3 mmol/L in women), dysglycaemia (impaired fasting glucose, impaired glucose tolerance, or uncomplicated diabetes treated with diet only), early renal dysfunction, family history of pre-mature coronary heart disease in first degree relatives (men <55 years or women <65 years).

Eligibility for RCTs compared with calculated risk

Figure 1 shows cases meeting the minimal eligibility criteria for each trial based on age, SBP, and cholesterol measurements. These are juxtaposed with the guideline recommendation for statin use according to the risk algorithms used in North American and European guidelines. This analysis reveals that patients meeting eligibility criteria for primary prevention RCTs may not be identified by risk algorithms as being eligible for statin therapy, generally because their risk is considered too low.

Figure 1

Trial-indicated vs. risk algorithm-indicated statin allocation. The left hand panel shows the acronym of the primary prevention randomized clinical trial (RCT) and the minimum age, systolic blood pressure (SBP), and total cholesterol and maximal HDL-C that would render a subject with these characteristics eligible for the individual RCT. The right hand panel shows the risk %/10 years as calculated by the FRS, pooled cohort equation (PCE, for both white and black subjects) and the systematic coronary risk estimate 2 (SCORE2) calibrated for low, moderate, high, and very high risk regions of Europe. Using the appropriate thresholds applicable to each algorithm, calculations yielding a low risk are designated in green and imply that statin would generally not be offered. Calculations yielding moderate risk are coloured orange and imply that statins would be considered. Calculations yielding high risk, or scenarios meeting RCT inclusion criteria, are coloured red and imply that statin use is indicated. *The lipoprotein-cholesterol (LDL-C) column represents either the lower limit defined in the specific trial or a calculated LDL-C (Friedewald equation) assuming a TG of 1.5 mmol/L and use of the lowest total cholesterol (4.0 mmol/L) stipulated in PROSPER and highest high-density lipoprotein-cholesterol (HDL-C) (1.16 and 1.22 mmol/L for males and females, respectively) stipulated in AFCAPS/TEXCAPS. ** No history of hypertension and TC:HDL-C ≤ 6.0. † History of hypertension. ‡ On blood pressure treatment. § Female with more than one risk factor other than smoking, or low HDL-C (i.e. dysglycaemia, abnormal waist:hip ratio, and family history of pre-mature coronary disease). ¶ CRP ≥ 2 mg/L.

Characteristics of patient enrolled in primary prevention trials

We then analysed the characteristics of the patients enrolled in the selected trials. Table 2 summarizes the main demographics. The percentage of women ranged from 0% to 68% with the highest percentage noted in the MEGA trial. Smoking rates ranged from 12.5% to 44% with the lowest rate in the AFCAPS/TEXCAPS trial. Hypertension rates ranged from 15.5% to 80.4% with the lowest rate in the WOSCOPS trial. The mean LDL-C ranged from 2.77 to 4.97 with the lowest mean value from the JUPITER trial. Mean TG values ranged between 1.41 and 1.84 mmol/L with the lowest value recorded in the JUPITER trial. In the two trials reporting hs-CRP the mean value was 4.76 mg/L in JUPITER and 2.33 mg/L in HOPE-3. When the trials were aggregated, the percentage of males and females was 65% and 35%, respectively. 95% of the enrollees were 47 to 81 years old (mean 63.9 ± 8.9 years), and 95% had LDL-C between 1.75 and 5.31 mmol/L (mean 3.53 ± 0.91 mmol/L). Systolic pressure was 142 ± 20 with 95% of enrollees representing a range from 102 to 181 mmHg. It should be noted that a minority of subjects had established ASCVD in any of the trials except PROSPER. Exclusion of this trial did not substantially alter the overall age or lipid ranges represented by these trials (see Supplementary Materials, Supplementary material online, Figure S1). Figure 2 shows the age vs. LDL-C ranges from each trial with error bars depicting 95% of the enrollees. Notably, while both JUPITER and HOPE-3 had the lowest ranges of LDL-C, the mean age was over 65 years old. On the other hand, PROSPER enrolled patients with the most advanced ages and with LDL-C values comparable to MEGA and AFCAPS/TEXCAPS.

Figure 2

Relationship between age and lipoprotein-cholesterol in primary prevention randomized clinical trials of statin therapy. Bars represent the 2.5th to 97.5th percentiles that encompass 95% of subjects enrolled in the trials. The average and 1.96 standard deviation of the systolic blood pressure is also provided for each trial.

Table 2

Characteristics of participants enrolled in the main primary prevention statin trials

	Individual randomized clinical trials							Aggregation of all trials
Variable	WOSCOPS12	AFCAPS/TexCAPS13	PROSPER14	ASCOT-LLA15	MEGA16	JUPITER17	HOPE-318	# studies included	N	N (%) or Mean (SD)	Mean ± 0.67 SD (25th to 75th percentile, 50% of population)	Mean ± 1.96 SD (2.5th to 97.5th percentile, 95% of population)
N	6595	6605	5804	10 305	7832	17 802	12 705	7	67 648	—	—	—
Men (%)	100%	85%	48%	81%	32%	62%	54%	7	67 648	43 678 (64.6%)	—	—
Women (%)	0%	15%	52%	19%	68%	38%	46%	7	67 648	23 970 (35.4%)	—	—
Age	55.2 (5.5)	58.3 (7.0)	73.4 (13.4)	63.1 (8.5)	58.3 (7.2)	65.7 (8.2)b	65.8 (6.4)	7	67 648	63.9 (8.9)	57.9, 69.9	46.5, 81.3
History of vascular disease		—	44.2%	—	—	—	—	1	5804	2565 (44.2%)
Angina	5.1%		26.9%					2	12 399	1897 (15.3%)
Claudication	2.9%		6.7%					2	12 399	583 (4.7%)
MI			13.4%					1	5804	776 (13.4%)
Stroke/TIA			11.2%	9.7%				2	16 109	1650 (10.2%)
Revascularization			4.1%					1	5804	237 (4.1%)
PAD			2.1%	5.0%				2	16 109	637 (4.0%)
Other CVD				3.8%				1	10 305	395 (3.8%)
Family history of pre-mature CHD (%)	—	15.7%	—	—	—	11.5%	26.2%	3	37 112	6415 (17.3%)	—	—
Smoking (%)a	44%	12.4%	26.8%	32.7%	20.6%	15.8%	27.7%	7	67 648	16 613 (24.6%)	—	—
Hypertension or on hypertensive medications (%)	15.7%	20.5%	61.9%	80.4%	41.8%	—	37.9%	6	49 846	22 364 (44.9%)	—	—
Diabetes (%)	1.2%	6.0%	10.7%	24.6%	20.8%	0%	—	6	54 943	5257 (9.6%)	—	—
Pre-diabetes or early diabetes	—	—	—	—	—	—	18.5%	1	12 705	2347 (18.5%)	—	—
Metabolic syndrome (%)	—	—	—	—	—	41.4%	—	1	17 802	7375 (41.4%)	—	—
BMI	26.0 (3.2)	26.95 (3.16)	26.8 (4.2)	28.65 (4.65)	23.8 (3.1)	28.56 (4.97)b	27.11 (4.78)	7	67 648	27.19 (4.57)	24.1, 30.3	18.2, 36.1
Systolic BP (mmHg)	135.5 (17.5)	138 (17)	154.7 (21.8)	164.2 (17.8)	132.2 (16.8)	134.4 (15.6)	138.1 (14.8)	7	67 648	141.6 (20.2)	128.1, 155.1	102.0, 181.2
Diastolic BP (mmHg)	84.0 (150.5)	78 (10)	83.8 (11.5)	95.0 (10.3)	78.6 (10.3)	80.7 (8.9)	81.9 (9.3)	7	67 648	83.2 (11.2)	75.7, 90.7	61.2, 105.2
Total cholesterol (mmol/L)	7.03 (0.59)c	5.71 (0.54)	5.7 (0.9)	5.5 (0.8)	6.27 (0.31)	4.77 (0.6)b,c	5.21 (1.09)c	7	67 648	5.53 (1.01)	4.85, 6.21	3.55, 7.51
LDL-C (mmol/L)	4.97 (0.44)c	3.89 (0.43)	3.8 (0.8)	3.4 (0.7)	4.05 (0.46)	2.77 (0.48)b,c	3.31 (0.93)c	7	67 648	3.53 (0.91)	2.92, 4.14	1.75, 5.31
HDL-C (mmol/L)	1.14 (0.23)c	0.95 (0.14)	1.3 (0.4)	1.3 (0.4)	1.49 (0.39)	1.28 (0.39)b,c	1.16 (0.36)c	7	67 648	1.24 (0.38)	0.99, 1.49	0.50, 1.98f
Triglyceride (mmol/L)	1.84 (0.78)c	1.78 (0.86)d	1.5 (0.7)	1.65 (0.90)	1.51 (0.69)b	1.41 (0.70)b,c	1.50 (0.71)b,c	7	67 648	1.56 (0.77)	1.04, 2.08	0.05, 3.07
ApoB (g/L)	—	—	—	—	—	—	1.03 (0.26)	1	12 705	1.03 (0.26)	0.86, 1.20	0.52, 1.54
Lp(a) (mmol/L)	—	—	—	—	0.88 (0.91)	—	—	1	7832	0.88 (0.91)	0.27, 1.49	0g, 2.66
Fasting glucose (mmol/L)	—	—	—	6.2 (2.1)	—	5.25 (0.63)b,c	5.34 (0.82)b,c	3	40 812	5.52 (1.29)	4.66, 6.38	2.99, 8.05
Glycated haemoglobin (%)e	—	—	—	—	—	5.68 (0.34)b	—	1	17 802	5.68 (0.34)	5.47, 5.93	5.03, 6.37
Serum creatinine (mmol/L)	—	—	—	99.0 (16.7)	—	—	79.1 (19.5)c	2	23 010	88.01 (20.79)	74.08, 101.94	47.26, 128.8
eGFRe	—	—	—	—	—	74.0 (14.3)b	—	1	17 802	74.0 (14.3)	64.4, 83.6	45.97, 102.03
hs-CRP (mg/L)	—	—	—	—	—	4.76 (3.23)b	2.33 (2.19)b	2	30 507	3.75 (3.09)	1.68, 5.82	0h, 9.81

^aCurrent smoking reported for all studies except for MEGA, which reported one number for current/past smoking.

^bOriginally reported as Median (IQR) and converted to Mean (SD).

^cOriginally reported in mg/dL and converted to mmol/L.

^dOriginally reported as Median (SD) and converted to Mean (SD).

^eRenal criteria from SHARP trial and T2D from CARDS trial are already used to consider statin-eligibility.

^fHDL-C, only one study reported an upper HDL-C eligibility criterion of 1.16 mmol/L for males and 1.22 mmol/L for females (AFCAPS/TEXCAPS).

^gLp(a) reported in only one study.

^hhs-CRP, mean—1.96SD is negative.

Table 3 summarizes the event rates, relative and absolute risk reductions associated with statin therapy, percent reduction of LDL-C, and NNT. The table also summarizes the study duration which allowed extrapolation of the NNT to 10 years [NNT (10 year)] so that the trials could be compared with each other. Baseline risk in the placebo groups ranged from 2.5% in MEGA to 16.2% in PROSPER. Relative risk reduction ranged from 15% in PROSPER to 44% in JUPITER. Absolute risk reduction ranged from 1.1% in ASCOT and HOPE-3 to 2.1% in WOSCOPS and PROSPER. Using these values and the mean study duration, NNT (10 year) ranged from 15 (PROSPER) to 51 (HOPE 3). Supplementary material online, Table S1, excluding PROSPER, shows that these ranges did not change materially.

Table 3

Summary of event rates, risk reduction and number needed to treat in primary prevention randomized trials

Study	Primary endpoint	Placebo event rate	Treatment event rate	Study duration	Percent reduction of LDL-C	Percent reduction in primary endpoint	Absolute risk reduction	Number needed to treat	Number needed to treat (10 y)
WOSCOPS	Occurrence of non-fatal MI or death from CHD as a first event	248 in 3293 = 7.5%	174 in 3302 = 5.3%	Mean 4.9 years	26%	31%	2.1%	48	24
AFCAPS	First acute major coronary event defined as fatal or non-fatal myocardial infarction, unstable angina, or sudden cardiac death.	183 in 3301 = 5.5% Rate per 1000 patient-years—10.9	116 in 3304 = 3.5% Rate per 1000 patient-years = 6.8	Mean 5.2 years	25%	37%	2.0%	50	26
PROSPER	Composite of coronary death, non-fatal MI, and fatal or non-fatal stroke	473 in 2913 = 16.2%	408 in 2891 = 14.1%	Mean 3.2 years	34%	15%	2.1%	48	15
ASCOT	Non-fatal MI and fatal CHD	154 in 5137 = 3.0%	100 in 5168 = 1.9%	Median 3.3 years	35%	36%	1.1%	91	30
MEGA	First occurrence of coronary heart disease	101 in 3966 = 2.5% 5.0 cases per 1000 patient-years	66 in 3866 = 1.7% 3.3 cases per 1000 patient-years	Mean 5.3 years	18.0%	33%	1.8%	56	29
JUPITER	MI, stroke, arterial revascularization, hospitalization for unstable angina, or death from cardiovascular causes.	251 in 8901 = 2.8% Rate per 100 person-year = 1.36	142 in 8901 = 1.6% Rate per 100 person-year = 0.77	Median 1.9 years	50%	44%	1.2%	83	16
HOPE-3	First coprimary outcome was composite of death from cardiovascular causes, non-fatal myocardial infarction, or non-fatal stroke, and the second coprimary outcome additionally included resuscitated cardiac arrest, heart failure, and revascularization	304 in 6344 = 4.8% (1st coprimary outcome) 363 in 6344 = 5.7% (2nd coprimary outcome)	235 in 6361 = 3.7% (1st coprimary outcome) 277 in 6361 = 4.4% (2nd coprimary outcome)	Median 5.6 years	27%	24% (1st coprimary outcome) 25% (2nd coprimary outcome)	1.1–1.3%	77–91	43–51

Relative risk reduction was reported in WOSCOPS and similar metrics were derived from reported relative risk (AFCAPS/TexCAPS) and an unadjusted hazard ratio (ASCOT-LLA). Other relative risk reductions were derived from the reported hazard ratios. These relative risks pertain to the diverse primary endpoints broadly considered to represent major adverse cardiac events as follows: WOSCOPS-non-fatal MI or death from coronary heart disease; AFCAPS/TexCAPS-acute coronary events defined as fatal or non-fatal MI, unstable angina, or sudden cardiac death; PROSPER-Death from coronary heart disease or non-fatal MI or fatal or non-fatal stroke; ASCOT-LLA-non-fatal MI (including silent MI) plus fatal coronary heart disease; MEGA-coronary heart disease (includes fatal and non-fatal MI, angina, cardiac and sudden death, and coronary revascularization); JUPITER-first major cardiovascular event (includes non-fatal MI, non-fatal stroke, hospitalization for unstable angina, arterial revascularization procedure, or confirmed death from cardiovascular causes); HOPE-3-death from cardiovascular causes, non-fatal MI, or non-fatal stroke.

Simplified flow diagram

By considering both the simplest inclusion criteria of the RCTs as well as the actual characteristics of the enrollees, we developed a simplified flow diagram that reflects both the relatively high LDL-C ranges studied in patients over a broad, younger age range and the relatively low LDL-C ranges studied in JUPITER and HOPE-3, but in somewhat older individuals (Figure 3A and B). Based on this approach, a statin-eligible primary prevention patient is 40 to 70 years old with an LDL-C ≥ 3.0 mmol/L. When the LDL-C is in a lower range (≥1.8 mmol/L) but the subject is older (55 to 80 years old), at least one RCT will provide the rationale for statin therapy if additional risk factors are present, as driven largely by the HOPE-3 and JUPITER trials. In all of these trials, concomitant lifestyle and behavioural counselling was also provided. The flow diagram includes follow-up to ensure adherence, safety and attainment of whatever goals or targets are relevant to the relevant jurisdiction of practice. Supplementary files provide a link to an app that literally incorporates all the diverse inclusion/exclusion criteria of the RCTs and several additional case examples.

Figure 3

(A, B) Simplified flow diagrams for consideration of statin therapy in primary prevention. Panel (A): This algorithm is an alternative to the app-based method described in the text. It factors age, lipoprotein-cholesterol (LDL-C) and presence or absence of other risk features, derived from JUPITER and HOPE-3. The age range and lipid ranges are approximations from the trial criteria Table 1, and the characteristics of patients enrolled in the trials and 95% confidence intervals shown in Table 2 and Figure 2. *Dysglycaemia includes pre-diabetes, impaired fasting glucose, glucose intolerance, and uncomplicated, diet-controlled diabetes. + WHR = waist to hip ratio. Panel (B): This panel highlights the overlap of age ranges for statin-indicated patients and when secondary criteria should be used to assist in determining statin-eligibility. In subjects >70 years with LDL-C ≥ 3.0 mmol/L, the likelihood of having no additional risk factors is low and would be eligible for statins as per the bottom row. Conversely, patients <55 with LDL-C ≥ 1.8 mmol/L may or may not have additional risk factors. It is anticipated that risk algorithms, novel risk markers or imaging studies may be most impactful in this setting.

Discussion

Based on the fundamental, primary prevention RCTs using statins, we propose an alternate approach to identify statin-eligible primary prevention patients that is grounded in the RCT evidence and that is independent of non-evidence-based risk calculation (Graphical Abstract). This synthesis proposes that a statin-eligible primary prevention patient is a person between 40 and 70 years with an LDL-C ≥ 3.0 mmol/L or between 55 and 80 years with an LDL-C ≥ 1.8 mmol/L who has additional risk factors. This alternate approach is intended to serve as a starting point for a discussion in the health community regarding the advisability of considering statin in addition to lifestyle and behavioural counselling for primary prevention within a patient-centred decision-making process (Figure 3A and B).

Counselling a patient that they have characteristics similar to patients enrolled in successful primary prevention RCTs may be compelling. The suggested approach elevates patient education and patient choice regarding their personal priority for primary prevention. Such choices will be affected by many factors including whether the goal is to prevent events or to prolong life and whether or not there are other countervailing issues at play such as frailty or other co-morbidities. Barring such impediments to therapy, if the patient remains reluctant, other approaches such as measurement of novel risk factors, performance of imaging studies or calculation of algorithm-based risk would be secondary tools to use in the context of motivational interviewing. These additional tools may also be very useful to determine differential aggressiveness of therapy or consideration of non-generic or expensive statin add-on therapies.

A key finding from this analysis is that patients who meet minimal eligibility criteria of these RCTs rarely meet recommended guideline thresholds of risk that would endorse statin therapy (Figure 1). This implies that primary prevention recommendations are substantially narrower than would be supported by RCT evidence, and that despite the massive expenditures on executing these landmark RCTs, appropriate translation into clinical practice has been sub-optimal. While this was perhaps acceptable in the era of expensive, non-generic statins, our findings signal that this may no longer be justified. Moreover, rapidly emerging concepts about the early effects of LDL-C on development of atherosclerosis and the greater potential for regression when LDL-C lowering is applied during early phases of the disease cannot be ignored.²⁸

It has been demonstrated that while allocation of statins in primary prevention using risk algorithms lowers the theoretical NNT, the counterbalancing result ironically is that fewer events are prevented at the population level in absolute terms.⁹ One reason for this is that the risk imparted by dyslipidaemia is intertwined with the risks imparted by non-lipid risk factors such as age, sex, smoking, or hypertension, which are all factors that markedly influence the calculation of risk. This approach does not adequately reflect that the trials have also shown substantial risk reductions and with reasonable NNT (Table 3). Moreover, it is well accepted that a 20% reduction in CV events can be achieved per 1 mmol/L reduction in LDL-C, a benefit that may be even greater in younger patients.^29,30 A 1 mmol/L reduction in LDL-C can be achieved with high efficacy statin dosing at low levels of baseline LDL-C (e.g. 1.8 mmol/L) and even with low efficacy statins at high levels of baseline LDL-C (e.g. at 5 mmol/L). This range of LDL-C harmonizes with the range of LDL-C studied in the RCTs (Figure 2; Supplementary material online, Figure S1).

To emphasize the causal role of LDL-C while also trying to minimize NNT, the causal-benefit model has been proposed.^31,32 A commensurately greater benefit of lipid lowering is predicted with either higher baseline levels of LDL-C or lower targets of therapy or both. However, the expected absolute risk reduction required to calculate NNT must rely on one of several diverse algorithms. Applying this approach to the National Health and Nutrition Examination Survey (NHANES), the authors demonstrated that the PCE identified only 7.9 million candidates for statins out of 24.7 million that met eligibility criteria of primary prevention RCTs.³² Furthermore, the PCE identified 7.1 million candidates for statins who had features not studied in RCTs. Similarly, the causal-benefit model passed over 11.1 million subjects meeting eligibility criteria of RCTs while endorsing treatment in 10.9 million subjects not studied in RCTs. While it is conceptually appealing in clinical practice to balance NNT and benefit of magnitude of LDL-C lowering, we should recognize that the population health benefit, and the evidence base supporting this strategy have not been proven. Moreover, while adherence to inclusion/exclusion criteria of RCTs is often criticized as not reflective of the general population, the prioritization of risk-based approaches paradoxically identifies even fewer patients than proven to benefit in the RCTs.

Faridi et al. have recently reported findings in asymptomatic adults age 40 to 60 years from the Miami Heart Study without traditional, non-lipid risk factors and with a <5% 10 year risk as assessed by PCE.³³ Coronary atherosclerosis was present in over 20% of this group and was increasingly prevalent at higher levels of atherogenic lipoproteins including LDL-C and Lipoprotein (a) (Lp(a)). This is a striking example of the discordance between risk calculation and the ongoing vascular toxicity imparted by lipids, even when non-lipid risk factors are absent.

Our proposal also attenuates sex bias currently integrated into the existing algorithms. Both men and women of the same age were enrolled into PROSPER, ASCOT-LLA and MEGA and a woman of 40 years with an LDL-C ≥ 3.0 mmol/L would be eligible for inclusion in at least one RCT (MEGA). We believe that eliminating sex when considering LDL-C-related CV risk is another benefit of this simplified approach. A recent analysis by Delabays et al. compared European and US guidelines that make treatment recommendations dependent upon risk algorithm calculations.³⁴ Nearly half of women facing an incident ASCVD event were not eligible for lipid-lowering therapy. For the same degree of lipid-associated risk and even with the same co-morbidities, women are less likely to be advised on or offered risk reduction.⁷ While this can be partially justified by the protective influence of the pre-menopausal years, lifetime CV burden is substantial and CV disease remains the leading cause of death in women. Substantial evidence further suggests that women are prone to the legacy CV risks related to complications of childbearing, such as hypertensive disorders of pregnancy, or withholding of lipid-lowering therapy during child bearing years. These and perhaps other influences are not fully accounted for by any of the commonly used algorithms.^35–37 Allowing women more choice in managing their personal lipid-related risk by this new approach may be advantageous over historical sex-based inequities in lipid management.

Earlier intervention in young and middle-aged adults would provide even greater reductions in clinical events over the longer term.^38–45 Such an approach would likely be highly cost-effective in young adult men and moderately cost-effective in young adult women.⁴⁶ Although most guidelines endorse screening for lipids and for treatment at 40 years old, the current algorithms, particularly for women and in the absence of co-morbidities such as smoking and hypertension, seldom generate sufficiently high risk to warrant therapy even when LDL-C is elevated. In the RCTs to date, we demonstrate that 95% of the subjects were 47 to 81 years old. While awaiting RCTs specifically performed in the young,^47,48 we should not ignore the biological insights from studies such as PESA^28,49 applicable to subjects 40 to 55 years old. In addition, a Presidential Advisory from the American Heart Association describes ‘Life’s Essential 8’ and promotes the concept that, irrespective of co-morbidities, non-HDL-C in adults ≥20 years old is ideal if <130 mg/dL, which equates to an LDL-C of ∼100 mg/dL or ∼2.6 mmol/L.⁵⁰ Similarly, even the European guidelines include a general recommendation for patients <70 years old to maintain LDL-C <2.6 mmol.² Thus, our RCT-derived proposal is already substantially concordant with more simplified, but less emphasized recommendations in both the USA and Europe. However, our definition is somewhat more conservative with respect to age (≥40 years old) and LDL-C (≥3 mmol/L) and with the requirement to be even older and with additional risk factors if LDL-C is ≥1.8 mmol/L. This new but more conservative paradigm is justified by the existing RCTs and because the implication of our proposal is that statin, not just lifestyle and behavioural counselling, would be indicated.

Based on this study, and with the recent publication of a landmark RCT of statins in patients living with HIV, we propose a short list of primary prevention scenarios that do not require calculation of risk before consideration of statin in addition to lifestyle and behavioural recommendations (Table 4).¹⁹

Table 4

Current and proposed statin-indicated primary prevention patients not identified through calculation of risk

1. Familial hypercholesterolemia

2. LDL-C ≥ 5.0 mmol/L

3. Type 2 diabetes (most patients)

4. CKD (excluding dialysis or renal transplantation)

5. Persons living with HIV

6. Persons meeting eligibility criteria for primary prevention randomized clinical trials or 40–70 years old with LDL-C ≥ 3.0 mmol/L or 55–80 years old with LDL-C ≥ 1.8 mmol/L and additional risk factors.

Limitations

This proposal has limitations. Not all of the trials entirely excluded subjects with some forms of ASCVD. This pertained particularly to PROSPER but even after exclusion of that trial, the overall age and LDL-C ranges represented by this set of predominantly primary prevention trials and the anticipated relative risk reductions were not altered materially (See Supplementary material online, Figure S1 and Table S1). But foremost, at this time we consider this study to be a thought experiment intended to spark discussion about rationally enhancing primary prevention beyond the current approach, remaining concordant with available RCTs and reducing heterogeneity in patient identification world-wide. Such dialogue needs to occur if the full impact of these RCTs is to be actualized. It is especially important to engage primary care practitioners who often find national guidelines challenging to implement, especially when new iterations recommend different risk-based algorithms. Some practitioners may still prefer to use a theoretical risk prediction despite their heterogeneity. Such algorithms constantly need recalibration, which can both improve statistical performance while also adversely affecting practice patterns and possibly overall public health.^51,52 We also appreciate that literal implementation of the proposal might be construed as leading to overtreatment with statins in primary prevention. However, this paradigm would emphasize informed patient choice about accepting results of the scientific trials showing the value of statins in patients with a similar profile as their own. Whether this approach would augment uptake of statins, or improve practice patterns and overall public health warrants consideration and comparison to the current, risk calculation approach. In statin-eligible primary prevention patients, the practitioner should use national recommendations to determine how to follow-up adherence, to determine achievement of suitable lipid levels, safety etc., all of which are beyond the scope of this analysis.

In conclusion, using reasonable and conservative assumptions to encourage translation of available primary prevention RCTs into routine practice, we propose that adults meeting eligibility criteria of primary prevention RCTs, i.e. adults 40–70 years old with LDL-C ≥ 3.0 mmol/L or those 55–80 years old with LDL-C ≥ 1.8 mmol/L and additional risk factors, deserve to have a well-informed conversation regarding the advisability of incorporating statin therapy as part of a patient-centred primary prevention risk reduction strategy.

Supplementary material

Supplementary material is available at European Journal of Preventive Cardiology.

Acknowledgements

We wish to thank Craig Kamimura for artistic support.

Author contribution

G.B.J.M. conceived of the idea explored in this study and wrote the original drafts, including revisions. A.R. performed all case analyses and development of software algorithms. G.B.J.M., A.R., and E.Y. undertook critical appraisal through extraction of key information from the trials, and all formal analyses of aggregated data. I.I., L.B., and R.H. contributed to the interpretation of data for the work and critically reviewed and edited the original and revised manuscripts. All gave final approval and agree to be accountable for all aspects of work ensuring integrity and accuracy.

Funding

No external funding was obtained for this research.

Data availability

Upon reasonable request, data are available.

Author notes

Conflict of interest: G.B.J.M. has received a research grant from Amgen, Novartis, HLS Therapeutics, Sanofi, Esperion, Ultragenyx. A.R. and E.Y. no conflicts of interest. I.I. has received a research grant from Novartis, Amgen, HLS Therapeutics, Ultragenyx and Sanofi. R.A.H. has received a research grant from Akcea-Ionis, Amgen, Arrowhead, Boston Heart, HLS, Novartis, Pfizer, Regeneron, Sanofi, Ultragenyx. L.R.B. has received a research grant from Amgen, Novartis, HLS Therapeutics.