Atherosclerotic Coronary Plaque Regression and Risk of Adverse Cardiovascular Events

Key Points

Question

What is the association between plaque regression and major adverse cardiovascular events (MACEs)?

Findings

This systematic review and meta-regression analysis of 23 lipid treatment trials included 7407 patients. In adjusted analyses, a 1% decrease in percent atheroma volume (PAV) showed a 19% reduction in odds of a MACE, and a 1% difference in PAV between intervention and reference arms was associated with a 25% reduction in odds of experiencing a MACE.

Meaning

These findings indicate that change in PAV has potential as a surrogate marker for MACEs, but given the heterogeneity in findings, additional data are needed.

This systematic review and meta-regression evaluates the association between coronary plaque regression induced by lipid-lowering therapies and major adverse cardiovascular events.

Abstract

Importance

The association between changes in atherosclerotic plaque induced by lipid-lowering therapies (LLTs) and reduction in major adverse cardiovascular events (MACEs) remains controversial.

Objective

To evaluate the association between coronary plaque regression assessed by intravascular ultrasound (IVUS) and MACEs.

Data Sources

A comprehensive, systematic search of publications in PubMed, Embase, Cochrane Central Register of Controlled Trials, and Web of Science was performed.

Study Selection

Clinical prospective studies of LLTs reporting change in percent atheroma volume (PAV) assessed by IVUS and describing MACE components were selected.

Data Extraction and Synthesis

Reporting was performed in compliance with Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines. The association between mean change in PAV and MACEs was analyzed by meta-regression using mixed-effects, 2-level binomial logistic regression models, unadjusted and adjusted for clinical covariates, including mean age, baseline PAV, baseline low-density lipoprotein cholesterol level, and study duration.

Main Outcome and Measures

Mean PAV change and MACE in intervention and comparator arms were assessed in an updated systematic review and meta-regression analysis of IVUS trials of LLTs that also reported MACEs.

Results

This meta-analysis included 23 studies published between July 2001 and July 2022, including 7407 patients and trial durations ranging from 11 to 104 weeks. Mean (SD) patient age ranged from 55.8 (9.8) to 70.2 (7.6) years, and the number of male patients from 245 of 507 (48.3%) to 24 of 26 (92.3%). Change in PAV across 46 study arms ranged from −5.6% to 3.1%. The number of MACEs ranged from 0 to 72 per study arm (17 groups [37%] reported no events, 9 [20%] reported 1-2 events, and 20 [43%] reported ≥3 events). In unadjusted analysis, a 1% decrease in mean PAV was associated with 17% reduced odds of MACEs (unadjusted OR, 0.83; 95% CI, 0.71-0.98; P = .03), and with a 14% reduction in MACEs in adjusted analysis (adjusted OR, 0.86; 95% CI, 0.75-1.00; P = .050). Further adjustment for cardiovascular risk factors showed a 19% reduced risk (adjusted OR, 0.81; 95% CI, 0.68-0.96; P = .01) per 1% decrease in PAV. A 1% reduction of PAV change between intervention and comparator arms within studies was also associated with a significant 25% reduction in MACEs (OR, 0.75; 95% CI, 0.56-1.00; P = .046).

Conclusions and Relevance

In this meta-analysis, regression of atherosclerotic plaque by 1% was associated with a 25% reduction in the odds of MACEs. These findings suggest that change in PAV could be a surrogate marker for MACEs, but given the heterogeneity in the outcomes, additional data are needed.

Introduction

Randomized clinical trials have unequivocally established that lowering serum cholesterol prevents major adverse cardiovascular events (MACEs), with a magnitude of clinical benefit proportional to the reduction in low-density lipoprotein cholesterol (LDL-C) levels.¹ As the mainstay of coronary artery disease (CAD) therapy, lipid-lowering medications represent a cornerstone in cardiovascular risk reduction through beneficial effects on coronary plaque composition. However, the effect of plaque volume reduction on clinical outcomes remains to be established.² The use of intravascular ultrasound (IVUS) has become a globally established method to evaluate the effect of lipid-lowering therapies (LLTs) on atherosclerotic plaque through accurate serial measurements of atheroma burden. Studies by Bhindi et al³ and others⁴ have shown that changes in percent atheroma volume (PAV) defined by IVUS represent a potential surrogate measure of cardiovascular events. With the advent of new clinical trials demonstrating the benefit of early initiation of intensive LLT on plaque burden in high-risk patients,^5,6 we evaluated the association between plaque regression and MACEs by providing an updated systematic review and meta-regression analysis of lipid treatment trials reporting MACEs and IVUS measures of change in coronary atheroma.

Methods

Study Identification and Selection

In this meta-regression analysis, we systematically reviewed prospective studies of LLTs published from July 2001 up to July 2022 that reported change in PAV or plaque volume at baseline and follow-up, as assessed by IVUS and explicitly describing presence or absence of any component of MACE (myocardial infarction, stroke, transient ischemic attack, unstable angina, or all-cause mortality). Results were limited to studies involving participants older than 18 years and articles published in English. The literature search and selection strategy is shown in Figure 1.

Figure 1. Flowchart of Literature Search and Selection.

IVUS indicates intravascular ultrasound; MACE, major adverse cardiovascular event.

Study selection and data extraction were performed by 3 independent reviewers (I.I., E.Y., and G.B.J.M.), including a member of the statistical team (M.G.). Disagreements were resolved through discussion to consensus. Studies selected in a previous meta-regression³ were included in our analysis (Figure 1). Information on baseline characteristics and treatment from each study was extracted as detailed in eTable 1 in Supplement 1. Data on cardiovascular risk factors can be found in eTable 2 in Supplement 1. The study followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) reporting guideline.

Data Preparation

The PAV was calculated as done previously.³ Mean change in PAV was obtained as follow-up minus baseline PAV. Major adverse cardiovascular events were reported as the total number of events. If an element of MACEs was not described in studies specifically mentioning other MACE components, the number of that event was assumed to be 0. The ratio of the MACE count to total number of individuals in each study group was calculated.

Meta-Regression of Change in PAV and MACE

We undertook a meta-regression analysis to determine the association between reported mean change in PAV and proportion of MACE at the study arm level, which allowed a direct comparison with the previous analysis by Bhindi et al.³ Due to heterogeneity between and within studies and the presence of several 0 counts of MACEs, mixed-effects binomial logistic regression models were used to examine the association between mean change in PAV and MACEs as described previously.³ Briefly, this association was first examined by using an unadjusted model, then by models with adjustment for individual clinical factors, including age, sex, baseline PAV, baseline LDL-C level, and study duration (Table). Race and ethnicity were reported in the minority of studies and with any degree of detail in only 1 study and were not considered further. All clinical factors associated with MACEs with P < .10 in the aforementioned models were selected as covariates in an adjusted model. In addition to this adjusted model, the association was further examined by incorporating cardiovascular risk factors (hypertension, diabetes, and smoking) and type of patient (ie, those with stable vs acute coronary syndrome). To include all identified studies, our analysis considered each arm of the studies because not all trials had comparator arms.

Table. Major Adverse Cardiovascular Events per 1% Reduction in Mean Percent Atheroma Volume (PAV) by 2-Level Binomial Logistic Mixed-Effects Modeling.

Model	OR per 1% reduction in mean PAV (95% CI)	P value
Unadjusted	0.83 (0.71-0.98)	.03
Adjusted for single covariate
Change in PAV + baseline PAV	0.90 (0.77-1.07)	.24
Change in PAV + baseline LDL-C	0.72 (0.83-0.97)	.02
Change in PAV + study duration	0.74 (0.62-0.89)	.001
Change in PAV + age	0.82 (0.66-0.99)	.046
Change in PAV + sex	0.77 (0.62-0.96)	.03
Change in PAV + change in LDL-C	0.73 (0.88-1.04)	.14
Change in PAV + follow-up LDL-C	0.69 (0.84-1.01)	.07
Multivariable adjusteda	0.86 (0.75-1.00)	.050
Fully adjustedb	0.81 (0.68-0.96)	.01

Abbreviation: LDL-C, low-density lipoprotein cholesterol.

^a In the multivariable analysis, only covariates independently associated with major adverse cardiovascular events (P < .10) after accounting for the effect of PAV were included and consisted of baseline PAV, baseline LDL-C level, study duration, and age.

^b This model also included hypertension, diabetes, smoking, and type of patient (ie, those with stable vs acute coronary syndrome).

To extend these analyses, a mixed-effects, 2-level binomial logistic regression was used to examine the association of adjusted change in PAV between treatment and comparator arms in individual studies with MACE outcomes. Specifically, we assessed the association between the difference in PAV change between the intervention arms (high-intensity treatment) and the comparator arms (reference group of either background LLT or lesser-intensity lipid treatment) and MACEs. Two single-arm studies were excluded: COSMOS (Coronary Atherosclerosis Study Measuring Effects of Rosuvastatin Using Intravascular Ultrasound in Japanese Subjects)⁷ (mean [SD] intervention rosuvastatin dose, 16.9 [5.3] mg per day; n = 126) and ASTEROID (A Study to Evaluate the Effect of Rosuvastatin on Intravascular Ultrasound-Derived Coronary Atheroma Burden)⁸ (intervention rosuvastatin dose, 40 mg per day; n = 507). For the YOKOHAMA-ACS (Yokohama Assessment of Fluvastatin, Pravastatin, Pitavastatin and Atorvastatin in Acute Coronary Syndrome) study,⁹ which had 4 arms (2 intervention groups of 26 participants each and 2 comparator groups of 25 participants each), an average of both groups was taken given the nearly identical sample sizes per group.

Sensitivity and Subgroup Analyses

To assess robustness of the findings, the association between mean change in PAV and MACEs was examined by removing 1 study at a time and by excluding (1) nonrandomized trials, (2) stable CAD cohorts, (3) acute coronary syndrome (ACS) cohorts, and (4) mixed cohorts separately. Subgroup analyses were also subsequently performed for stable CAD and ACS groups independently. All analyses were conducted using RStudio, version 1.0.136 software (R Project for Statistical Computing). In the mixed-effects logistic regression models, the t test was applied to examine the statistical significance of the association between change in PAV and MACEs. A 2-sided P < .05 was considered statistically significant.

Results

Characteristics of Included Studies

Among 1338 screened publications, 23 studies with 46 study arms published between 2001 and 2022 met eligibility criteria (eTable 1 in Supplement 1). Three studies were nonrandomized trials,^7,8,10 8 were ACS cohorts,^{5,6,9,11,12,13,14,15} 10 were stable CAD cohorts,^{16,17,18,19,20,21,22,23,24,25} and 2 included patients with either stable CAD or ACS.^26,27 Among the 23 studies, 5 were described as placebo-controlled,^5,6,10,14,28 and the rest were LLT studies with a wide range of types and doses administered.^{7,8,9,11,12,13,14,15,18,19,20,21,22,23,24,25,26,27}

A total of 7407 patients were included in the analysis, with numbers ranging from 39 to 1380 according to individual studies (eTable 1 in Supplement 1). Among these, 4023 patients were in the intervention arms (n = 24), namely high-intensity treatment, and 3384 were in the comparator arms (n = 22), with the latter including patients receiving either background LLT or lesser-intensity lipid treatment. Mean (SD) age ranged from 55.8 (9.8) to 70.2 (7.6) years. The number of women ranged from 2 of 26 (7.7%) to 262 of 507 (51.7%) and men from 245 of 507 (48.3%) to 24 of 26 (92.3%).

Mean (SD) baseline PAV ranged from 36.0% (8.3%) to 55.2% (6.1%), and higher baseline PAV was associated with older patients. Study duration ranged from 11 to 104 weeks, with a median of 44 weeks; 4 studies had a more than 1-year follow-up. Mean (SD) change in PAV ranged from −5.6% (5.5%) to 3.1% (6.2%). The MACE counts ranged from 0 to 72, with 17 of 46 (37%) study groups having no events, 9 (20%) having 1 to 2 events, and 20 (43%) having 3 or more events. Higher numbers of events were typically associated with larger sample sizes and longer follow-up times, as expected. The number of patients with MACEs ranged from 0 to 72 (12.2%).

Meta-Regression Analysis of Mean Change in PAV and MACE

The odds ratios (ORs) and 95% CIs for MACEs per 1% reduction in mean PAV from different models are shown in the Table. In the unadjusted model, an association was observed between mean change in PAV and MACE, with PAV accounting for an estimated 33% of the variation in MACE. Per 1% reduction in mean PAV, the odds of experiencing a MACE outcome decreased 17% (OR, 0.83; 95% CI, 0.71-0.98; P = .03). Inclusion of either study duration, mean age, sex, or baseline LDL-C level in the regression model did not attenuate the association between mean change in PAV and MACE. In addition, with the exception of the proportion of men, these covariates were independently associated with the occurrence of MACEs, using the conservative P < .10. Baseline PAV was also independently associated with MACE (P = .002), and after the addition of baseline PAV to the model of change in PAV on MACE, the association was no longer significant. With multivariable adjustment for baseline PAV, study duration, mean age, and baseline LDL-C levels, the association between PAV and MACE was slightly attenuated, showing a 14% reduction in the odds of MACE per 1% reduction in mean PAV and a reduction in mean PAV (adjusted OR, 0.86; 95% CI, 0.75-1.00; P = .050) (Table). Further adjustment for hypertension, diabetes, smoking, and type of patients showed a significant 19% reduction in odds of incurring a MACE (adjusted OR, 0.81; 95% CI, 0.68-0.96; P = .01).

The association between the observed mean change in PAV and corresponding odds of a MACE (log scale) is depicted in Figure 2A. The regression line is based on the adjusted model, and the slope reflects the estimated correlation between average change in PAV and the proportion of patients with a MACE at 52 weeks, with age fixed at a mean of 62 years, baseline LDL-C fixed at a mean of 123 mg/dL, and baseline PAV fixed at a mean of 45%. To further expand these analyses, the association between the reference-adjusted change in PAV (ie, difference between intervention and comparator arms in individual studies having both) and corresponding OR of MACE (log scale) was examined (Figure 2B). The slope of the regression line reflects the estimated correlation between the difference in PAV change between arms in each trial and the reduction in MACE between arms after exclusion of single-arm studies. Thus, a 1% reduction of change in PAV between study arms was associated with a significant 25% reduction in the odds of experiencing a MACE (OR, 0.75; 95% CI, 0.56-1.00; P = .046).

Figure 2. Association of Percent Atheroma Volume (PAV) With Odds of Major Adverse Cardiovascular Events (MACEs).

A, Each circle represents a single study group. The size of the circle is proportional to the sample size of that study group. The MACE proportion was converted to log odds, and a constant 0.5 was added to 0 counts to allow the conversion of log odds. The regression line is based on the adjusted mixed-effects logistic regression model with follow-up at 52 weeks, age fixed at a mean of 62 years, baseline low-density lipoprotein cholesterol fixed at a mean of 123 mg/dL (to convert to mmol/L, multiply by 0.0259), and baseline PAV fixed at a mean of 45%. B, The black line indicates the regression line; shaded area, 95% CI; light blue circles, observed log(odds ratio [OR]) between the intervention and high-intensity statin arm and the comparator or reference arm for the change in PAV; size of circles, size of study (ie, total number of patients in each study); and brown circles, studies for which the log(OR) was calculated with addition of a non-0 constant (0.1) to 0 events to allow conversion of log odds.

Evaluation of Study Heterogeneity in MACE

We subsequently evaluated the heterogeneity of the association between change in PAV and observed MACEs. The observed χ²₄₁ of 193.5 (P < .001) indicates substantial heterogeneity across studies. However, the estimated intraclass correlation coefficient of 2.9% implies that the contribution of interstudy variability to the overall variability was not large (Figure 3).

Figure 3. Evaluation of Heterogeneity of Observed Major Adverse Cardiovascular Events (MACEs) by Study Group.

Included are only study groups with active lipid-lowering therapy. The MACEs are presented for each study group as total counts and as the proportion of patients with a MACE. For 0 MACE counts, the lower CI was set as 0 and the upper CI as [1 – 0.05^(1/n)], where n is the sample size in the study group. For non-0 MACE counts, the CI was obtained from the log odds and then converted back to proportions. EPA indicates eicosapentaenoic acid.

Sensitivity and Subgroup Analyses

Figure 4 illustrates the association between the mean change in PAV and MACE outcome after removal of 1 study at a time. The association of mean change in PAV with MACE was found to be consistent irrespective of the removal of any study. We further examined this association by excluding studies grouped by nonrandomized trials, stable CAD, ACS, and mixed cohorts. The association remained when excluding the 3 nonrandomized trials. Removal of studies with stable CAD resulted in an OR of 0.96 (95% CI, 0.85-1.09; P = .53), exclusion of the ACS cohort in an OR of 0.85 (95% CI, 0.68-1.08; P = .18), and removal of mixed studies in an OR of 0.72 (95% CI, 0.59-0.87; P < .001). In a separate subgroup analysis, a significant association between mean change in PAV and MACE was observed in the stable CAD cohort (OR, 0.68; 95% CI, 0.52-0.90; P = .002) but not in the ACS cohort (OR, 0.84; 95% CI, 0.70-1.01; P = .057).

Figure 4. Sensitivity Analysis of Major Adverse Cardiovascular Events (MACEs) per 1% Reduction in Mean Percent Atheroma Volume (PAV).

The odds ratios (ORs) and 95% CIs are shown after exclusion of 1 trial at a time, nonrandomized studies, studies performed in patients with acute coronary syndrome (ACS), studies of mixed ACS and stable coronary artery disease (CAD), and studies of stable CAD exclusively.

Discussion

This systematic review and meta-regression analysis found through different approaches that pharmacologically induced regression of atherosclerotic plaque assessed by IVUS was associated with a clinically important and statistically significant reduction in MACEs. Specifically, every 1% decline in mean PAV was associated with a significant 17% reduction in odds of incurring a MACE. After adjusting for baseline PAV, age, study duration, and baseline LDL-C level, this reduction was attenuated slightly to a nonsignificant 14%, whereas additional adjustment for cardiovascular risk factors led to a significant reduction of 19%. These results included all of the available data, including studies without a comparator arm. In examining the reduction in MACEs as a function of PAV difference among the trials that had both treatment and comparator arms, for each 1% reduction of PAV change between treatment groups, the odds of experiencing cardiovascular events was significant at 25%. Whether through analysis methods of previous work by Bhindi et al³ examining the association from each individual treatment arm or through new approaches including only studies with both a treatment and comparator arm, we provide evidence of an association between PAV change and MACE outcomes where for each 1% change in PAV, there is a 14% to 25% range in reduction of MACEs. As such, these results confirm the findings of Bhindi et al³ and others,⁴ supporting the change in PAV as a potential surrogate end point for MACEs, even in the face of new clinical trials in patients with ACS, earlier initiation of LLT, or more intensive combinations of lipid-lowering agents, including proprotein convertase subtilisin/kexin type 9 (PCSK9) inhibitors. Accordingly, to our knowledge, these analyses represent the most comprehensive and up-to-date meta-regression studies of this nature.

Despite a mild attenuation in association strength between PAV change and MACEs when adjusting for different covariates, the fully adjusted association was significant. As expected, the change in PAV was associated with duration of LLT, older age, and baseline LDL-C concentrations. Furthermore, in extensive sensitivity analyses, the association of mean change in PAV with occurrence of MACEs was found to be consistent irrespective of study removal, in both stable and unstable forms of atherosclerotic disease. Figure 4 illustrates the robustness of this association when individual studies are removed one at a time. In an exploratory analysis, we further examined this association when excluding studies grouped by stable CAD, ACS, and mixed cohorts. Removal of the 3 nonrandomized studies,^7,8,10 as well as the ACS population,^{5,6,9,11,12,13,14,15} maintained a similar OR between PAV and the occurrence of MACEs. The latter was, however, attenuated when eliminating stable CAD studies^{16,17,18,19,20,21,22,23,24,25} and became significantly pronounced when mixed ACS and stable trials were removed.^26,27 This analysis shows a heterogeneity between studies based on several underlying factors, such as population characteristics, number of individuals, and event rate or MACE components, among others. The findings are also in line with the slight difference in odds of MACE observed between the current study and the previous meta-regression by Bhindi et al,³ where a reduction in mean PAV of 1% was associated with an adjusted MACE reduction of 18% (95% CI, 5%-30%; P = .01) and an unadjusted MACE reduction of 22% (95% CI, 4%-37%; P = .02) for baseline PAV, age, study duration, and baseline LDL-C levels.

Although it has been noted by Räber et al⁵ and Nicholls et al⁶ that atherosclerotic coronary disease in patients with ACS may be more amenable to regression by LLT, our analysis does not substantiate that this regression translates into a greater MACE reduction compared with stable CAD. As such, previous IVUS analyses of clinical trials with emphasis on LDL-C lowering through PCSK9 inhibitors and high-intensity statin treatment of patients with ACS have shown a larger absolute regression of PAV compared with more stable disease.^5,6,29 Our study explored the association between plaque regression and MACE. While patients with ACS may be more sensitive to LLT given underlying plaque vulnerability, on the basis of the limited data available, one cannot conclude that in this particular group, plaque regression leads to a more effective MACE risk reduction, possibly because of an interplay between absolute risk of different cohorts and duration of studies.

However, there is now evidence to suggest that such changes in overall atheroma burden are associated with remodeling of lesions in more complex ways, such as diminution of lipoid components, inflammatory cells, and thickening of fibrous caps.¹ These mechanisms of lesion remodeling are compatible with stabilization of atherosclerotic plaques, which has important implications for understanding how intensive LLT reduces cardiovascular risk. Numerous studies support the role of statins beyond cholesterol lowering, with anti-inflammatory, immunomodulatory, and antioxidative effects, as well as increased plaque stabilization.³⁰ Puri et al³⁰ reported that high-intensity statin therapy was associated with PAV regression from baseline, whereas both low-intensity statins and no statin therapy were associated with PAV progression. Independent of lipid lowering, and as a means to provide plaque stabilization, statins also promote coronary atheroma calcification, which has been reported by several investigators, including from SATURN (Study of Coronary Atheroma by Intravascular Ultrasound: The Effect of Rosuvastatin vs Atorvastatin).³¹ Interestingly, these microcalcifications are commonly found within a fibrous cap, with a very low proportion actually rupturing, and with statins rendering these microcalcifications more confluent and denser, which contributes to plaque stability.^29,32 Supportive of this stabilizing of plaque with statin therapy are also the PACMAN-AMI (PCSK9 Antibody Alirocumab on Coronary Atherosclerosis in Patients With Acute Myocardial Infarction)⁵ clinical trial and the HUYGENS (High-Resolution Assessment of Coronary Plaque in a Global Evolocumab Randomized Study) study,^6,33 which assessed the effect of early administration of PCSK9 inhibitors to high-intensity statin therapy on plaque characteristics in patients with ACS. The premise of these studies was that patients remain at increased risk of recurrent atherothrombotic events due to high-risk plaque characteristics, particularly in noninfarct-related vessels. In PACMAN-AMI, coronary plaque regression assessed by IVUS was associated with a reduction in lipid core burden index, increased fibrous cap thickening, and decreased mean angular extension of macrophages, indicating benefits for coronary plaque composition and stabilization. Similar results were shown in HUYGENS. The emerging data from these studies show that plaque regression may be associated with atheroma remodeling and a favorable outcome of intensive cholesterol lowering on plaque lipids.

Limitations

A few factors limit our analysis. First, this systematic review and meta-analysis is based on studies that generally had low event rates and small changes in PAV. Second, the individual studies did not provide, in some cases, information on timing of events. As such, these findings are hypothesis generating and should not be taken as implying direct causality between plaque regression and MACE outcomes. On the other hand, these factors also warrant and justify consideration of approaches such as meta-regression methods for detection of the associations reported herein. Third, a proportion of our patient samples came from studies performed by a small number of highly experienced investigative groups, which may limit generalizability of the findings.^{4,6,8,17,26,27,28,29,30,31,33,34} Fourth, the proportion of patients with a MACE may be underestimated, as not all MACE components were reported in all studies, and losses to follow-up could not be taken into account. Thus, although our results suggest that PAV has potential as a surrogate end point for MACEs, long-term studies of antiatherosclerotic agents powered for clinical events and associated with concomitant imaging would be ideal to firmly establish the value of PAV as a predictor of clinical benefit. However, such studies to date have not been practical to undertake.

Conclusions

With data from more than 7000 patients, this systematic review and meta-analysis is, to our knowledge, the largest performed to date, providing evidence that atherosclerotic plaque reduction has the potential to be a surrogate measure of cardiovascular risk reduction. Nevertheless, given the heterogeneity in findings and borderline significant association when adjusting for other factors, additional data are needed. Utility of novel lipid-lowering medications using measures of atheroma regression may be considered as hypothesis generating prior to completion of large-scale clinical end point studies, but definitive trials will always be required to confirm clinical efficacy and inform potentially important adverse effects or unexpected, off-target toxic outcomes.

Supplement 1.

eTable 1. Baseline Characteristics and Treatment of Included Studies

eTable 2. Baseline Cardiovascular Risk Factors of Included Studies

Supplement 2.

Data Sharing Statement