Prediction of Cardiometabolic Health Through Changes in Plasma Proteins With Intentional Weight Loss in the DiRECT and DIADEM-I Randomized Clinical Trials of Type 2 Diabetes Remission

The Diabetes Remission Clinical Trial (DiRECT) in the U.K. primary care population demonstrated, for people with type 2 diabetes of <6 years’ duration, remission occurred in 46% at 1 year, with a mean weight loss of 10.2 kg (1). A maintained weight loss >10 kg led to remission for ∼70% of intervention group participants at both 12 and 24 months (2). The Diabetes Intervention Accentuating Diet and Enhancing Metabolism (DIADEM-I) remission trial included younger participants from 13 different countries in the Middle East and North Africa (MENA) region and shorter diabetes duration (≤3 years). It showed that a mean weight loss of 11.98 kg at 12 months resulted in 61% remission (3). These trials have placed remission as an option for people with early type 2 diabetes, now translated into routine patient care across some countries, e.g., England where a successful pilot has led to wider adoption by the National Health Service. Furthermore, DiRECT showed weight loss was accompanied by substantial reductions in liver fat, and hepatic triglyceride export, as well as improvements in pancreatic function and morphology (4,5).

Observational evidence in people with type 2 diabetes has suggested improved survival with intentional weight loss in a clinic-based study of newly diagnosed patients (6) and in insurance data (7). In the Look AHEAD (Action for Health in Diabetes) trial, there was no overall cardiovascular benefit from an intervention designed to improve cardiovascular fitness with intensive lifestyle intervention over 4 years (8); however, in post hoc analysis, 10 kg weight loss by year one was associated with 21% lower risk (95% CI 2–36) of the primary composite cardiovascular outcome (9). Ample observational data also suggest that significant weight loss with bariatric surgery is associated with lower cardiovascular risk, including in those with type 2 diabetes (10).

In this post hoc analysis of proteomic data from DiRECT and DIADEM-I, we examined a broader set of clinically relevant outcomes beyond those available from routine clinical data. Our group recently established circulating protein-based algorithms that predict several cardiometabolic outcomes in combined analyses of seven prospective cohorts (the U.K. Fenland Study, the Norwegian Trøndelag Health Study (HUNT3), the Swiss Biochemical and Electrocardiographic Signatures in the Detection of Exercise-induced Myocardial Ischemia (BASEL VIII) study, the international EXenatide Study of Cardiovascular Event Lowering (EXSCEL), and the U.S. Covance, Atherosclerosis Risk in Communities (ARIC), and HEalth, RIsk factors, exercise Training And GEnetics (HERITAGE) studies (11,12). In that proof-of-concept study, we demonstrated that these protein-based tests can provide individualized health assessment across multiple conditions simultaneously from a single blood sample. Our hypothesis was that plasma protein changes linked to ≥10 kg weight loss in DiRECT and DIADEM-I would predict cardiovascular benefits and improvements in cardiovascular fitness. We also wished to demonstrate the dynamic properties of these tests, further supporting their potential utility as tools for precision medicine. These are important questions to inform patients embarking on intentional weight loss as well as health care services.

DiRECT was designed to target type 2 diabetes and cardiovascular risk factors through weight loss (13). DiRECT is a large-scale randomized trial that demonstrated dietary and lifestyle intervention in a primary care setting could achieve and maintain remission of type 2 diabetes (defined as HbA_1c <6.5% after at least 2 months off all diabetes medications) and improve blood pressure control (1,13). The intervention group was taken off all glucose-lowering medications, and diuretic and antihypertensive medications, and put on a total diet replacement with use of formula-based meal replacement products (825–853 kcal/day formula) for 3–5 months, followed by a stepped food reintroduction (2–8 weeks) and structured support for long-term weight loss maintenance (1,13). Protocols based on national guidelines were used to reintroduce medications if blood glucose or blood pressure exceeded treatment thresholds. The DiRECT data set used for these analyses included all participants (n = 262) with analyzable baseline and 1-year follow-up biological samples (n = 524 samples). More than half (50.9%) of the patients in the intervention arm analyzed in this data subset showed remission of type 2 diabetes. EDTA plasma samples were collected from fasting individuals and spun at 2,000g at 4°C for 15 min within 4 h of collection. Samples were decanted within 30 min of centrifugation and stored at −80°C.

DIADEM-I was a randomized controlled clinical trial designed with a dietary intervention similar to that of DiRECT, delivered in a primary care and community setting (14). Participants recruited originated from the MENA region and were younger and had a shorter diabetes duration in comparison with DiRECT participants. The dietary intervention included total diet replacement using formula-based meal replacement products (800–820 kcal/day) for 3 months, followed by a stepped food introduction over the subsequent 3 months followed by regular food intake. In addition to the dietary intervention, DIADEM-I participants received regular unsupervised support for physical activity. Diabetes medications were stopped at the outset in those undergoing the intervention. Blood pressure medication management was through clinical evaluation following national and international guidelines. The measured outcomes were weight loss and diabetes remission (defined as HbA_1c <6.5% after the individual was off of all diabetes medications for at least 3 months). The DIADEM-I data set used for these analyses included all participants (n = 142) with analyzable baseline and 1-year follow-up samples (n = 284 samples) with longitudinal analysis on baseline plus 3-monthly follow-up biological samples up to 1 year (n = 597 samples). EDTA plasma samples were collected from fasting individuals and spun at 1,300g at 4°C for 10 min at ∼2–3 h after collection. Samples were decanted within 30 min of centrifugation and stored at −80°C. Both trials received ethics approval by their respective institutional review boards.

Sensitivity to change in response to the DiRECT and DIADEM-I interventions was assessed for 10 SomaSignal tests; results for 8 of these SomaSignal tests (described below) were included in the analyses. Two SomaSignal tests, Alcohol Impact (11) and Primary Cardiovascular Risk – 4 years (11), were run for exploratory purposes (Alcohol Impact test was included as a negative control and Primary Cardiovascular Risk test is not appropriate for participants with elevated cardiovascular disease [CVD] risk factors, such as type 2 diabetes) and were accounted for in Bonferroni adjustments for multiple testing but are not included in the results.

The SomaSignal tests applied to the DiRECT and DIADEM data sets were previously developed and validated (11,12). The Liver Fat test was developed and validated in >10,000 participants from the Fenland Study cohort with liver ultrasound scores and reports results as follows: “no excess fat” or “some excess fat” based on standard ultrasound scoring (score ≤4, normal, and score ≥5, mild/moderate/severe grades of fat) (11). The body composition tests, Lean Body Mass, Body Fat Percentage, and Visceral Fat, were developed and validated in >11,000 participants from the Fenland Study cohort with DEXA measurements and report results as continuous measurements of body fat (percentage), lean body mass (kilograms), and visceral fat (kilograms) (11). The Cardiorespiratory Fitness – VO2 Max test predicts maximal oxygen uptake on a cycle ergometer as a continuous measurement (millimeters per kilograms per minute) and was developed and validated on >700 participants from the HERITAGE exercise intervention study (11).

The Cardiovascular Risk – 4 years test predicts the likelihood of a cardiovascular event (myocardial infarction, stroke, trans-ischemic attack, hospitalization for heart failure, or all-cause death) in individuals ≥40 years with one or more known causes of elevated cardiovascular risk, including individuals presenting with stable CVD (defined as a history of myocardial infarction or stroke [>6 months prior], history of heart failure, revascularization, abnormal stress test or imaging suggesting coronary heart disease including coronary artery stenosis >50% by angiogram or elevated coronary calcium score), type 2 diabetes, symptoms consistent with chronic obstructive coronary artery disease, or age >65 years old without known elevated cardiovascular risk (12). The test was developed in 813 participants from the HUNT3 and ARIC cohorts and independently validated in 11,609 participants from HUNT3, ARIC, BASEL VIII, and EXSCEL cohorts, with results reported as a risk category and percentage range (12).

The resting energy rate test was developed and validated in >9,000 participants from the Fenland Study cohort with resting energy expenditure measured by indirect calorimetry; results are reported as a continuous measure (integer) in kilocalories per day. The glucose tolerance test was developed and validated in >11,700 participants from the Fenland Study cohort with 2-h oral glucose tolerance test results and reports results as “normal tolerance” (<7.8 mmol/L) or “impaired tolerance” (≥7.8 mmol/L) based on standard oral glucose tolerance test cutoffs. For all SomaSignal tests statistical analysis plans and minimum performance criteria were predefined and documented in a regulatory document vault (Arena Solutions).

The modified aptamer binding reagents (15) and SomaScan Assay (16) and its performance characteristics (17) have previously been described. The SomaScan Assay uses ∼5,000 individual DNA-based slow off-rate modified aptamers (SOMAmer reagents), which complex with proteins in the diluted plasma. After various steps to eliminate nonspecific binding, the DNA-based reagents are quantified by hybridization on an Agilent array as described in Supplementary Material.

SomaSignal tests were performed on both baseline and 1-year follow-up samples to look for favorable changes in the model outputs in the intervention and control arms, as well as to compare differences in changes between the two arms. Within the intervention arm these tests were also assessed on participants stratified by weight loss with a 10 kg cutoff (i.e., participants who lost ≥10 kg vs. those who did not). Six of the SomaSignal tests associated with cardiometabolic health (Cardiovascular Risk – 4 years, Liver Fat, Body Fat Percentage, Visceral Fat, Resting Energy Rate, and Glucose Tolerance) were hypothesized a priori to improve in the intervention group in both DiRECT and DIADEM-I. Additionally, the Cardiorespiratory Fitness – VO2 Max and Lean Body Mass tests results were hypothesized to improve in the intervention group in DIADEM-I, which included specific physical activity support as part of the structured weight loss protocol. For tests with a continuous output (the Cardiovascular Risk – 4 years, Body Fat Percentage, Visceral Fat, Cardiorespiratory Fitness – VO2 Max, Lean Body Mass, and Resting Energy Rate tests) paired t tests were used to determine whether the means of the predictions are significantly lower (at α = 0.05, Bonferroni adjusted for multiple testing) after the intervention. Tests with binary outputs (Liver Fat and Glucose Tolerance) consisted of a one-tailed binomial test of proportions to determine whether the proportion of individuals predicted as impaired (probability >0.5 for predictions of some increased liver fat or impaired glucose tolerance, respectively) decreases significantly after the intervention. One-tailed tests were used to maximize power to detect cardiometabolic changes hypothesized a priori to be associated with weight loss. Two-tailed paired t tests were used to determine whether there was a significant overall change (α = 0.05) due to the intervention for SomaSignal tests that were not hypothesized to change in response to treatment. This same approach was used for the SomaSignal tests in the control arms in both studies. Model outputs were not expected to change in the control arms, and two-tailed paired t tests were used for all SomaSignal tests to determine whether predictions change significantly (α = 0.05) in the control group. Finally, group-wise changes (mean of continuous predictions or proportion for binary outputs) were compared with t tests to determine whether the changes in the control and intervention arms were significantly different (at α = 0.05).

Longitudinal analyses of the DIADEM-I samples were assessed for change in each SomaSignal test across time points: baseline, 3 months, 6 months, 9 months, and 1 year. Repeated-measures ANOVA was used to determine whether the means of the predictions are significantly different between each time point (α = 0.05), with time and treatment arm used as fixed effects and subject identifier as the random effect.

In supplementary analyses, we combined data from both trials and tested the effect of intervention with the cardiometabolic SomaSignal tests (in the entire trial) using a mixed-effects linear regression model accounting for study, sex, age, and subject identifier as a random effect and including an interaction term of time point × treatment arm.

Demographics for the DiRECT and DIADEM-I participants included in the analyses are described in Table 1. Participants in DiRECT were approximately a decade older than DIADEM-I participants, and more had existing CVD and hypertension. The majority of DiRECT participants were White European, while in DIADEM-I the majority of participants were from the MENA region. DIADEM-I included a larger number of males. Weight loss was comparable between the two trials.

We used SomaSignal tests related to cardiometabolic health to assess change in response to the DiRECT and DIADEM-I weight loss interventions. These tests include measures of future cardiometabolic risk (Cardiovascular Risk – 4 years) and current health state (Liver Fat, Cardiorespiratory Fitness – VO2 Max, Lean Body Mass, Body Fat Percentage, Visceral Fat, Glucose Tolerance, and Resting Energy Rate). Six of the SomaSignal tests were hypothesized to change in concordance with the DiRECT intervention and reversal of type 2 diabetes, and eight of the SomaSignal tests were hypothesized to change in concordance with the DIADEM-I intervention (a priori hypotheses included in RESEARCH DESIGN AND METHODS). In assessments of the change in proteins within treatment groups, six of the SomaSignal tests showed improvement for the DiRECT intervention group and seven for the DIADEM-I intervention group. In the control arms, SomaSignal test results remained unchanged or improved to a lesser extent. Key performance indicators for paired testing of intervention and control groups for the SomaSignal tests are shown in Supplementary Tables 1 and 2.

In addition, the magnitude of the mean change between the control and intervention groups (control vs. intervention) for the SomaSignal tests was assessed. These changes were assessed using the same hypotheses for directionality (described above). In comparisons of changes in proteins between the DiRECT intervention and control arms (control vs. intervention), all six of the SomaSignal tests that showed significant improvement in the intervention arm also exhibited significantly higher improvements in the intervention versus control group, with results adjusted for multiple testing; in addition, the differences in CVD risk were significant (see Table 2). Similarly, in DIADEM-I, six of the seven SomaSignal tests with improvement in the intervention arm alone had significantly stronger improvement in the intervention group versus control group when we controlled for multiple testing. However, while directionally consistent with DiRECT, the effect on cardiovascular risk in DIADEM-I was not statistically significant. The mean change between the control and intervention groups for the SomaSignal tests is illustrated in Supplementary Fig. 1.

Since remission of type 2 diabetes and its maintenance at 2 years were closely related to sustaining >10 kg weight loss in DiRECT (1,2), the SomaSignal tests were also assessed on participants within the intervention arms with stratification by weight loss using a 10 kg cutoff (i.e., participants with ≥10 vs. <10 kg weight loss). The DiRECT and DIADEM-I results mirrored the results of paired testing done in the intervention group without stratification (Table 2), with the same SomaSignal tests showing improvement with intervention. However, the effect size was generally greater in those who lost ≥10 kg of weight in comparison with those who lost <10 kg of weight (see Supplementary Table 3 and Fig. 1). Figure 1 also illustrates that relative predicted risks were reduced most in participants in the intervention group with ≥10 kg loss, with particularly strong (>50%) predicted reductions for glycemia and ectopic fat depot SomaSignal tests, and more modest, ∼20–30%, relative risk reductions for the CVD SomaSignal test in DIRECT and DIADEM-I. In nearly all cases, a dose response effect was evident: those in the intervention arm who lost >10 kg had the greatest magnitude of change, followed by those with <10 kg loss (typically intermediate changes) and then the control arm with the smallest or no change from baseline.

Supplementary analyses combining data from both trials mirrored the results seen in the separate DiRECT and DIADEM-I analyses (Supplementary Table 4). Change in cardiovascular risk was the only proteomic test where results changed significantly in the combined analyses but did not in both trials independently. This combined result was driven mainly by the results of the larger study, DiRECT, which showed a significant effect on the mixed-effects linear regression model and in which there were higher baseline cardiovascular risk scores in comparison with DIADEM-I (Supplementary Fig. 1) and a greater increase in cardiovascular risk in the control arm from baseline to 1 year.

Finally, with the availability of serial samples in DIADEM-I, temporal effects were identified whereby rapid (within 3 months) changes were seen in protein-predicted metabolic/adiposity parameters as well as in cardiorespiratory fitness and resting energy rate, whereas changes in cardiovascular risk appeared to take longer (Fig. 2 and Supplementary Fig. 2).

Our results show that weight loss, which led 46% of people with recently diagnosed type 2 diabetes to be in remission after 1 year in DiRECT, alters protein levels in a validated algorithm commensurate with future reductions in a range of cardiometabolic outcomes, including cardiovascular outcomes. Although from a post hoc analysis, the data stem from a well-characterized randomized trial of typical people living in the U.K. with early type 2 diabetes and add to the evidence base on the effects of large-scale weight loss on cardiovascular outcomes (18). The SomaSignal protein tests indicate meaningful reductions in cardiovascular risk and also suggest that the 1-year weight loss of ≥10 kg leads to ∼10% reduction in resting energy rate and a comparable ∼13% rise in cardiorespiratory fitness. Finally, greater predicted improvements in glycemia, liver fat, and body composition measures, all of which improved rapidly as measured by gold standard measurements in trial, lend strong internal validity (4). Such findings concur with our recent concept article on the timing of differing cardiometabolic benefits post–weight loss (19). We saw broadly similar results in DIADEM-I, a trial conducted in the MENA region, which also showed fast improvements in metabolic parameters, whereas any cardiovascular benefit would take longer to emerge. Indeed, body composition and cardiometabolic fitness significantly improved within the first 6 months before plateauing during the weight loss maintenance period. Lean body mass and resting energy rate decreased significantly within the first 3 months but then slowly increased from 3 months to 1 year, potentially consistent with an introduction of exercise regimen plus a modest regain in weight. Collectively, the work adds further evidence for rapid multifactorial metabolic and predicted longer-term cardiovascular benefits of intentional weight loss in the context of type 2 diabetes.

Diabetes is a major public health problem across the world, and CVD remains the leading cause of death and disability (20). The lack of overall benefit in hard cardiovascular outcomes in the Look AHEAD trial (8) fueled a perception in some quarters that there is little vascular benefit to be gained from lifestyle improvements in people with type 2 diabetes. Our results, albeit using a surrogate protein-based index of cardiovascular risk, but one recently shown to correctly predict the results of multiple intervention trials (12), challenge this concept. In doing so, they add support to the post hoc results of Look AHEAD in which people who had lost at least 10% of their body weight in the first year of the study had a 21% lower risk (95% CI 2–36) of cardiovascular outcomes (9). This risk reduction level is close to the ∼20–30% reductions in CVD SomaSignal test predictions we report for those in the intervention arm who lost 10 kg of weight (∼10%) by 1 year in DiRECT and DIADEM-I (Supplementary Table 3 and Fig. 1), even though the cardiovascular outcome included in our analyses was somewhat different to those in the Look AHEAD analysis. It will be of interest to see whether results reported from the Semaglutide Effects on Heart Disease and Stroke in Patients with Overweight or Obesity (SELECT) study (21) (semaglutide vs. placebo, subjects without diabetes and with prevalent atherosclerotic CVD) and SURPASS-CVOT (22) (tirzepatide vs. dulaglutide, subjects with prevalent diabetes and CVD), both of which anticipate substantial weight loss, show major adverse cardiac events reduction. If so—and a preliminary press release from SELECT is encouraging in this respect—this may add further credence to the notion that institutions with intervention development pipelines may look to proteomics-based approaches to gain insights into potential outcomes, to help inform on decisions about the best weight loss candidates for phase 3 trials.

The SomaSignal tests were developed and validated with large cohorts (11), but it is worth noting that liver fat algorithms were developed with use of less sensitive ultrasound, not MRI, and the glucose tolerance SomaSignal test was developed with use of oral glucose tolerance test data. Even so, the significant >50% relative reductions in these measures in both trials in those who lost >10 kg of weight over the first year are strongly consistent with findings with liver fat (via MRI) and glycemia changes (via HbA_1c) that we reported previously in DiRECT (4,13).

The predicted reduction in the resting energy rate in both trials is in accord with expectations and with other controlled weight loss studies that saw a 5.6–14.6% reduction in resting energy rate with similar weight loss (23–25). Lighter people must burn fewer calories unless more active and accelerometer data from DiRECT showed minimal change in activity (26). Notably, the improvement in cardiorespiratory fitness appeared to be rapid in DIADEM-I.

Together these new data provide further evidence that patterns of plasma proteins can meaningfully approximate a range of important metabolic outcomes or risk parameters (some of which require imaging or dynamic testing) as well as capture, the often rapid, changes in such risk parameters following dietary-driven intentional weight loss. These findings thereby extend the notion that protein-based tests not only could simultaneously provide information for individualized health guidance across multiple conditions from a single blood sample but also could be used to track treatment-induced improvements. Most importantly, they highlight a potential cardiovascular benefit of intentional weight loss, especially when such weight loss is >10 kg in people with type 2 diabetes, a topic of extreme interest given availability of newer therapies that can now help people with type 2 diabetes lose an average of >10 kg in weight, e.g., tirzepatide (27).

We accept several limitations. First, we had blood samples only for people in the intervention arm who remained in the trial up to 1 year in DiRECT. Lifestyle trials tend to have larger dropout rates than drug trials, but even so, near 80% of participants in the active arm provided biomarker samples, with 97% of the control participants providing such samples. We did not control for drug changes, but, notably, patients in the intervention group were on less oral diabetes agents (mostly metformin) and blood pressure medications at year 1, a pattern that may be predicted to minimize group differences in cardiovascular risk rather than to exaggerate them. We also accept that predicted cardiovascular benefits were not nominally significant in DIADEM-I, but notably, this cohort was of lower baseline cardiovascular risk, being more than a decade younger and with far fewer having hypertension and only one patient having established CVD. This means that there may have been less scope and statistical power to identify a potentially beneficial effect on the proteins corresponding to changes in CVD risk. Finally, we accept that SomaSignal protein-predicted outcomes are not substitutes for hard outcomes, but, even so, it is notable that a large amount of weight loss due to surgery in a 10-year follow-up to a trial in patients with diabetes suggested reductions in vascular outcomes (both macro- and microvascular events, though numbers were small) in comparison with much less weight loss in a medical arm (28). While other surgery-based information also suggests benefits of substantial weight loss on hard outcomes, such information comes from observational data rather than randomized clinical trials. The totality of evidence from DiRECT and DIADEM-I biomarkers reported here, together with post hoc Look AHEAD data (9), and surgical data in people with type 2 diabetes (10), suggests that intentional (as opposed to unintentional) weight loss is likely to lessen a range of cardiometabolic outcomes, including, over time, atherosclerotic CVD. Such findings, in turn, support intentional weight loss as a likely contributory factor for any cardiovascular benefits in future incretin-based or similar outcome trials.

In summary, analyses of changes in plasma proteins from DiRECT and DIADEM-I predicted significant improvements in many domains beyond glycemia and adiposity to include future cardiovascular risk, resting energy rate, and cardiorespiratory fitness with use of intention-to-treat intervention data, especially for those achieving >10% weight loss. These findings lend confidence to the notions that intentional weight loss early during type 2 diabetes is advantageous for multiple metabolic traits and that if weight loss is sustained future cardiovascular end points could be reduced, and that such benefits could be captured or monitored by proteomic analyses of plasma.

Acknowledgments. The authors thank Liz Coyle, University of Glasgow, for excellent assistance in the preparation of this manuscript. The authors thank Elaine Butler and Philip Stewart from the University of Glasgow for providing technical assistance in sample collection and curation. The authors also thank the SomaLogic sample management, assay services, and production bioinformatics teams for assistance in running the samples and quality control.

Funding. DIADEM-I was funded by the Qatar National Research Fund through the National Priorities Research Program grant (NPRP 8-912-3-192) awarded to S.T. S.T., O.C., and H.Z. were supported by funding from the Biomedical Research Program by the Qatar Foundation to Weill Cornell Medicine Qatar. N.S. is supported by the British Heart Foundation Research Excellence Award (RE/18/6/34217).

The statements made in this publication are solely the responsibility of the authors.

Duality of Interest. N.S. has consulted for and/or received speaker fees from Abbott Laboratories, Amgen, AstraZeneca, Boehringer Ingelheim, Eli Lilly, Hanmi Pharmaceutical, Janssen, Merck Sharp & Dohme, Novartis, Novo Nordisk, Pfizer, Roche Diagnostics, and Sanofi and received grant support paid to his university from AstraZeneca, Boehringer Ingelheim, Novartis, and Roche Diagnostics outside the submitted work. SomaLogic Operating Co., Inc., funded the costs of the proteomic assays and the biomarker measurements, and S.W., M.A.H., J.C., D.P.A., and E.V.T. are employees of the company. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. N.S. conceived the idea and wrote the first draft of the manuscript. S.T. led the DIADEM-I trial and conceived idea and contributed to writing of the manuscript. D.P.A. planned and carried out the analyses, contributed to the interpretation of the results, and aided in the writing of the manuscript. J.C. conceived the presented idea and planned the analyses, contributed to the interpretation of the results, and aided in the writing of the manuscript. M.A.H. planned and carried out the analyses, contributed to the interpretation of the results, and aided in the writing of the manuscript. M.V.H., P.W., H.Z., and O.C. all contributed to the writing of the manuscript. E.V.T. contributed to the interpretation of the results and aided in the writing of the manuscript. S.W. conceived the presented idea, planned the analyses, and contributed to the interpretation of the results. M.L. and R.T. co-led DiRECT and contributed to writing of the manuscript. N.S. and J.C. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented in abstract form at the American Heart Association Scientific Sessions 2022, Chicago, IL, 5–7 November 2022.