Comparative Discrimination of Life’s Simple 7 and Life’s Essential 8 to Stratify Cardiovascular Risk: Is the Added Complexity Worth It?

Abstract

Background:

Life’s Simple 7 (LS7) is an easily calculated and interpreted metric of cardiovascular health (CVH) based on 7 domains: smoking, diet, physical activity, body mass index, blood pressure, cholesterol and fasting glucose. The Life’s Essential 8 (LE8) metric was subsequently introduced, adding sleep metrics and revisions of the prior 7 domains. While calculating LE8 requires additional information, we hypothesized that it would be a more reliable index of CVH.

Methods:

Both the LS7 and LE8 metrics yield scores with higher values indicating lower risk. These were calculated among 11,609 Black and White participants free of baseline cardiovascular disease (CVD) in the REasons for Geographic And Racial Differences in Stroke study, enrolled in 2003–7, and followed a median of 13 years. Differences in 10-year risk of incident CVD (coronary heart disease or stroke) were calculated as a function LS7 and LE8 scores using Kaplan-Meier and proportional hazards analyses. Differences in incident CVD discrimination was quantified by difference in the c-statistic.

Results:

For both LS7 and LE8, the 10-year risk was approximately a 5% for participants around the 99^th percentile of scores, and 4-times higher 20% risk for participants around the 1^st percentile. Comparing LS7 to LE8, 10-year risk was nearly identical for individuals at the same relative position in score distribution. For example, the “cluster” of 2,013 participants with LS7 score of 7 was at the 35.8^th percentile in distribution of LS7 scores, and had an estimated 10-year CVD risk of 8.4% (95% CI: 7.2% to 9.8%). In a similar location in the LE8 distribution, the 1,457 participants with LE8 score of 60±2.5 at the 39.4^th percentile of LE8 scores, with a 10-year risk of CVD of 8.5% (95% CI: 7.1% to 10.1%), similar to the cluster defined by LS7. The age-race-sex adjusted c-statistic of the LS7 model was 0.691 (95% CI: 0.667, 0.705), and 0.695 for LE8 (95% CI: 0.681, 0.709) (p for difference 0.12).

Conclusions:

Both LS7 and LE8 were associated with incident CVD, with discrimination of the two indices practically indistinguishable. As a simpler metric, LS7 may be favored for use by the general population and clinicians.

Introduction

In 2010, the Goals and Metrics Committee of the Strategic Planning Task Force of the American Heart Association (AHA) introduced Life’s Simple 7 (LS7) as a metric to assess cardiovascular health (CVH), and monitor temporal changes.¹ Among the guiding principles used for the development were: 1) simplicity and accessibility for use by both the general population and clinicians, and 2) actionable items which the population, clinicians, and policy makers could use to measure cardiovascular health. The metric assesses health on 7 domains: diet, physical activity, smoking, body mass index (BMI), total cholesterol, fasting glucose, and blood pressure. For each of these domains, individuals were scored as having ideal, intermediate, or poor levels of the domain (see Table S1 for the criteria for the levels of each domain). For tracking and assessment, those with ideal levels are commonly scored with 2 points, those intermediate with 1 point, and those at poor levels 0 points; a summary score is calculated by summing the 7 individual scores, resulting in an overall score from 0 to 14 (higher is better).² This summary score is useful as an index of CVH for both individuals and clinicians, and it predicts not only incident cardiovascular events,^3–6 but a wide range of other health outcomes including cardiovascular risk factors like incident hypertension,⁷ diabetes,⁸ and atrial fibrillation,⁹ the development of conditions including nonalcoholic fatty liver disease,¹⁰ cognitive impairment,¹¹ peripheral artery disease,¹² cerebral small vessel disease,¹³ heart failure,¹⁴ end-stage kidney disease,¹⁵ stroke,² adverse outcomes after stroke,¹⁶ and overall health care expenditures.¹⁷

In 2022, the AHA introduced an updated and enhanced LS7 health metric with Life’s Essential 8 (LE8).¹⁸ This new metric added a measure of sleep,⁶ and levels of the LS7 previous health domains were expanded (see Table S1 for details of calculation). For example, in contrast to having only 3 levels of health for each domain, each of the 8 domains of LE8 is expressed as a semicontinuous variable. The diet, nicotine avoidance (only smoking for the LS7 metric), BMI, cholesterol, and blood pressure domains now have 5 levels, the new sleep domain has 6 levels, blood glucose has 7 levels, and physical activity 8 levels. For example, for BMI the LS7 scored 2 points for values <25.0 kg/m², 1 point for values from 25.0 to 29.9 kg/m², and 0 points for values >30.0 kg/m². In contrast, LE8 scores 100 points for values <25.0 kg/m², 70 points for values from 25.0 to 29.9 kg/m², 30 points for values from 30.0 to 34.9 kg/m², 15 points for values from 35.0 to 39.9 kg/m², and 0 points for values ≥40.0 kg/m². The levels of each of these 8 domains are scored on a scale of 0 to 100, and the final summary score is calculated as the average across the domains. The classification of each LS8 domain at a finer resolution holds the possibility of capturing important additional information to better quantify incident cardiovascular disease (CVD) risk (see Table S1 for scoring of the other 7 of the LE8 domains).

The LE8 metric has other changes compared to LS7, all with the goal to better quantify CVH. Some of these changes substantially complicate the information needed from individuals. For example, diet is quantified by percentiles of a DASH (Dietary Approaches to Stop Hypertension) diet score, which requires more information from individuals that might be harder to obtain accurately than the 5 food groups included in LS7. For the glucose domain, both metrics use fasting blood glucose, but LE8 added Hemoglobin A1c among individuals with diabetes. The LE8 tobacco metric was expanded beyond only cigarette smoking to include use of other tobacco products (cigars, cigarillos, pipes and snuff), inhalation from other nicotine delivery systems (a.k.a. vaping), the number of years since quitting for past smokers, and exposure to passive cigarette smoke. A less burdensome change in LE8 was use of non-HDL cholesterol rather than total cholesterol.

We hypothesize that the additional data and greater granularity in LE8 should improve the quantification of CVH, outweighing the burden of more data collection on both the individuals being evaluated and the clinical staff collecting the data. The goal of this report is to assess the magnitude of improvement in degree to which LE8, relative to LS7, can identify individuals at higher 10-year risk of having a cardiovascular event (coronary heart disease [CHD] or stroke).

Methods

Study Overview

Data for this study are from REasons for Geographic And Racial Differences in Stroke (REGARDS), a longitudinal cohort study of 30,239 Black and White community-dwelling individuals aged 45+ years who were recruited from the 48 contiguous US states and the District of Columbia between 2003 and 2007. These individuals were randomly selected from commercially available lists, and initially recruited by mail and telephone. For those agreeing to participate, demographics and cardiovascular risk factor data were collected by trained staff using a computer assisted telephone interview (CATI). Following the interview, at an in-home assessment, technicians measured anthropometrics, blood pressure, and obtained an ECG; collected blood and urine specimens; conducted a medication inventory; and provided self-administered questionnaires to be completed and returned by mail. A second assessment similar to the first was conducted between 2013 and 2016. Participants are contacted by telephone every 6 months for information about potential heart and stroke events, and assessment of cognitive performance. The study was approved by the Institutional Review Board, and written informed consent was obtained from all participants. REGARDS welcomes opportunities to share the data with appropriate investigators under a data use agreement that can be obtained at https://www.uab.edu/soph/regardsstudy/. Details of the study protocol have been previously described.¹⁹

Assessment of Risk Factors

At the in-home examination, systolic and diastolic blood pressure levels were calculated as the mean of 2 measurements performed after a 5-minute seated rest. Both LS7 and LE8 include fasting glucose, with classification thresholds at 100 mg/dL and 126 mg/dL. Participants were asked to fast for collection of blood specimens; however, 14% of the participants did not. These non-fasting participants were included in the analysis using accepted criteria of non-fasting levels of 140 mg/dL and 200 mg/dL, respectively.²⁰ REGARDS did not collect HbA1c on participants with diabetes as included in the LE8 criteria. Therefore, we conservatively assigned participants with diabetes a score of 20 points, near the middle of the scores assigned for HbA1c in the LE8 criteria. Treatment for high blood pressure, hyperlipidemia, and diabetes were assessed by self-report. For diet, both the consumption of the 5 foods used in the LS7 and the DASH diet score used in the LE8 were calculated from the self-administered validated Block 98 food frequency questionnaire.^{21, 22} Physical activity was assessed using the question “How many times per week do you engage in intense physical activity, enough to work up a sweat?”, a validated index of physical activity level.^23–25 Cigarette use, smoking cessation, smoking of non-cigarette tobacco products (cigars, cigarillos and pipes) and living with individuals who regularly smoke cigarettes were assessed by self-report. Height and weight were measured during the in-home assessment, and BMI calculated as kg/m². The average number of hours of sleep was assessed by self-report at a mean follow up of 3 years after baseline, a number of participants were excluded because these data were not available (we note that since participants were recruited in random order and followed up in that same general order, these missing data are quite likely “missing completely at random.”)²⁶

Cardiovascular Outcomes

The time to CHD death, non-fatal acute myocardial infarction, or fatal or non-fatal stroke event over 10 years was the outcome; hereafter referred to as “CVD events.” Suspected events were documented during the 6-month telephone contact, with medical records retrieved and adjudicated by a physician panel using published guidelines.^27–30 Medical records during the last year of life, autopsy reports, and death certificates were retrieved and next of kin interviewed to assess if the death was from CVD. Similarly, medical records for suspected stroke events associated with self-reported stroke hospitalization, or hospitalization associated with a report of stroke symptoms on the Questionnaire for Verifying Stroke-Free Status,³¹ were retrieved and adjudicated by a physician panel. Stroke was defined using the World Health Organization (WHO) definition,³⁰ or by an event with symptoms lasting less than 24 hours with neuroimaging evidence consistent with acute stroke. CHD and stroke events occurring before December 31,2019 were included in these analyses.

Statistical Analyses

LS7 was measured on a scale from 0 to 14 and LE8 on a scale from 0 to 100. Our focus was on risk difference between the LS7 and LE8 for an individual with the same relative ranking for the scores, e.g., how does the risk compare for a person at the 75^th percentile of each score. The scores for each person were considered as their percentile rank of the LS7 and LE8 distributions respectively, contrasting the risk difference for an individual at the same relative position in the distributions of the two scales.

Three analyses were performed: 1) a “univariate” analysis using Kaplan-Meier estimates of 10-year risk, 2) a proportional hazards analysis where the exposure was considered in categorical levels, and 3) a proportional hazard analysis where the LS7 and LE8 were considered as continuous predictors. Individuals who died before an incident event were censored at the time of their death, and individuals who reached the current end of follow-up without an incident event were censored at the end of follow-up. Since the LS7 scale naturally “pools” participants along the 0–14 scale and LE8 can take on many more unique values, to provide stable estimates for comparison, the LE8 score was rounded to the closest value of 5 for both the Kaplan-Meier and categorial proportional hazards analysis. In addition, for both analyses there were relatively small sample sizes in the “tails” of the distribution that would also contribute to unstable estimates. As such, strata with less than 50 participants were pooled, specifically: 1) LS7 values of 0 (n=1), 1 (n=17), and 2 (n=44) were pooled with the LS7 stratum of 3, 2) LE8 values of 15 (n=5), 20 (n=12), and 25 (n=39) were pooled with the LE8 stratum of 30, and 3) LE8 values of 100 (n=28) were pooled with LE8 stratum of 95. We contrasted differences in the 10-year risk, estimated directly from the Kaplan-Meier curves for the univariate analysis, and at the mean age, race and sex the proportional hazards analysis.

Finally, a formal test of whether the LE8 provides more “information” than LS7 was performed by considering the difference in the c-statistic between models using each metric. Briefly, the c-statistic estimates for two randomly selected participants with one participant having an earlier CVD event, the likelihood that the risk score would be higher for the participant having the earlier event. As such, the c-statistic can serve as a clinical indicator of the “information” in the model, ranging from 0.5 (no information) to 1.0 (complete information). The difference in the c-statistics for the LS7 and LE8 models was assessed by a Wald test considering the ratio of the difference in the c-statistics divided by the standard error of the difference in the c-statistics. The standard error of the difference in the c-statistics was calculated using bootstrapping techniques with 500 replications.

Results

Of the 30,239 REGARDS participants, 56 (<1%) were excluded for data anomalies, 1,930 (6%) for prevalent stroke at baseline, and 4,694 (16%) for prevalent CHD at baseline. All variables for LS7 were available for 20,305 (67%), but only 14,136 (47%) had all LE8 variables available. The main reason for different sample sizes was missing data on the sleep domain as the sleep questionnaire was added after baseline, so 10,303 (34%) participants had missing data (see Figure S1 for details). All data were available on 11,609 (38%) of the participants, and a description of the study population is provided in Table 1. The distribution of LS7 and LE8 is shown in Figures 1A and 1B respectively, where the greater spread of LE8 (relative to LS7) is illustrated. The distribution of the individual components of LS7, and the mean and standard deviation of the individual components of LE8, are provided in Table S2. A total of 1,305 outcome events (713 CHD events and 592 strokes) occurred over a median follow-up of 13 years (although estimation of risk was assessed at the 10-years point).

Table 1:

Description of the study population for demographic characteristics and components of the Life’s Simple 7 and Life’s Essential 8 Indices of cardiovascular health (n = 11609)

Age (mean ± SD)		63.1 ± 8.8
Black (%)		3,694 (31.8%)
Female (%)		6,934 (59.7%
Dash diet score (mean ± SD)		24.2 ± 4.4
Times per week engaged in physical activity enough to work up sweat (%)	≥4	3533 (30.4%)
	1–3	4613 (39.8%)
	0	3463 (29.8%)
Smoking (%)	Current Smoker	1694 (14.6%)
	Former, quit <1 y or using inhaled NDS	68 (0.6%)
	Former, quit 1-<5y	250 (2.2%)
	Former, quit ≥5 y	3974 (34.2%)
	Never	5623 (48.4%)
Body Mass Index (mean ± SD)		29.0 ± 6.0
Cholesterol (mg/dL, mean ± SD)	Total	195 ± 38
	Non-HDL	142 ± 37
Glucose (mg/dL, mean ± SD)	Fasting (n = 10415)	98 ± 25
	Non-fasting (n = 1194)	112 ± 46
Diabetes (%)		1656 (14.3%)
Systolic blood pressure (mm Hg, mean ± SD)		125 ± 15
Diastolic Blood Pressure (mm Hg, mean ± SD)		76 ± 9
Number of hours of sleep (mean ± SD)		7.1 ± 1.2

Figure 1:

Distribution of Life’s Simple 7 (LS7) and Life’s Essential 8 (LE8). LE8 was rounded to closest value of 5.

As shown in Figure 2, there was a tight relationship between the LS7 and LE8 scores (correlation of 0.87), with the box-and-whisker plot of the distribution of the LE8 score compressed at each level of LS7. For example, for those with LS7 = 7, the interquartile range (containing 50 percent of participants) is 10 LE8 points (75^th – 25^th percentile: 63.8 – 53.8 = 10.0). The difference between the 95^th and 5^th percentile (containing 90% of participants) is 23.1 LE8 points (70.6 – 47.5 = 23.1).

Figure 2:

Box-and-whisker plot showing the distribution of Life’s Essential 8 at each level of Life’s Simple 7. The top of the box shows the 75^th percentile, line in box the 50^th percentile, and bottom of box the 25^th percentile. Top whisker shows the 95^th percentile, while bottom of box the 5^th percentile. The correlation between LS7 and LE8 is 0.87.

Figure 1A (with numeric estimates provided in Table S3A) shows the Kaplan-Meier estimate of the “crude” 10-year risk of incident CVD at each level of LS7 and LE8 (with LE8 rounded to the closest 5). Both LS7 and LE8 provided a substantial discrimination of the risk of incident cardiovascular disease, with those below the 10^th percentile of scores having a 15% to 20% 10-year risk, while those above the 90^th percentile had an approximate 5-times lower (3% to 4%) risk. The pattern of decreasing risk with increasing scores was similar for LS7 and LE8, where, for example, the 2,013 participants with LS7 = 7 were at the 35.8^th percentile of LS7 scores, and had an 8.4% (95% CI: 7.2% to 9.8%) 10-year risk of incident cardiovascular disease. This compares to the 1,457 participants with an LE8 score within 2.5 points of 60, who are at the 39.4^th percentile of LE8 scores, and who have an 8.5% (95% CI: 7.1%, 10.1%) 10-year risk of incident cardiovascular disease. For both the LS7 and LE8, the 10-year risk of incident cardiovascular disease was strikingly nonlinear, with a rapid increase below approximately the 10^th percentile of scores and rapid decrease above the 90% percentile of scores.

Figure 1B (with numeric estimates provided in Table S3B) provides the 10-year risk by percentile of LS7 and LE8 after adjustment for age, race, and sex, demonstrating a very similar association as the crude analysis. There were two notable differences: 1) a more consistently increasing pattern of risk below the 10^th percentile (i.e., a “smoother” increase in the estimated risk), and 2) a marginally higher estimated risk for those at a low percentile of risk, with 10-year risks reaching 20% or higher.

Both LS7 and LE8 discriminated the cardiovascular risk of individuals, with participants with low cardiovascular health scores at substantially higher risk than those with higher cardiovascular health scores. Using the LE8 scores to demonstrate this point, over the entire range of scores there was a 20.8-time risk difference between the 146 participants with LE8 scores of 30 or less (10-year risk of 27.0%) compared to the 166 participants with scores rounded to 95 (10-year risk of 1.3%). Likewise, for LS7, there was a 21.5-times greater risk for the 233 participants with a LS7 score of 3 or below (21.3% 10-year risk) compared to the 64 participants with a LS7 score of 13 (1.0% 10-year risk). For both metrics, the pattern of the risk difference in estimated 10-year risk over the percentiles of scores showed: 1) a rapid decline in the first decile of risk scores (i.e., below the 10^th percentile), 2) a steady but slower decline over the 2^nd to 9^th deciles of scores (between the 10^th and 90^th percentiles), and 3) then a rapid decline in the highest decile (above the 90^th percentile). Using LE8 to demonstrate this pattern, the 10-year risk decreased from 27.0% for scores near 30 (1.0^st percentile), to 16.5% for scores near 35 (2.1^st percentile), to 14.2% for scores near 40 (5.0^th percentile), and to 9.9% for scores near 45 (9.8^th percentile). So, within the first decile of the distribution of LE8, the risk monotonically decreased by approximately two thirds. There was a similar two-thirds decline in risk within the 10^th decile, a 4.7% 10-year risk for scores near 85, a 3.8% 10-year risk for scores near 90, and a 1.3% 10-year risk for scores near 95. These large decreases in the 1^st and 10^th deciles contrast with a steady but much less dramatic 50% decline in risk over the 2^nd to 9^th decile of risk. Specifically, individuals at the 9.8^th percentile of LE8 (score near 45) had a 10-year risk of 9.9%, and individuals at the 93.3^rd percentile of LE8 (score near 85) has a 4.7% risk, only approximately a 50% lower risk between percentiles 83.5 points apart.

Figure S2 (with numeric estimates provided in Table S3C) shows the 10-year risk of percentiles of LS7 and LE8 (after adjustment for age-race-sex) where the scores were considered continuously. The linear modeling of LS7 and LE8 both fail to capture the substantial increases/decreases in for the 1^st and 10^th decile of scores that was shown in the univariate Kaplan-Meier analysis (Figure 3a) and the proportional hazards modeling with categorical assessment of scores (Figure 3b). That is, the linear modeling did not have sufficient “flexibility” to capture these important and substantial abrupt changes in risk in the tails of the distribution of scores.

Figure 3:

Estimated 10-year risk for coronary or stroke events as a function of the percentile for the Life’s Simple 7 (LS7) score and the Life’s Essential 8 (LE8) score. Panel A provides the “crude” Kaplan-Meier estimated 10-year risk. The LS7 estimates, shown in red, are provided for the calculated LS7 score, shown as the first number in the annotation. The sample size in each stratum is provided in the parentheses. For LE8, strata were created by rounding to the closest score of 5, and the rounded score and sample size is similarly provided. The LS7 scores below 10 or below have been pooled and shown at the LS7 = 10 point, as were LE8 scores below 30 or below (shown at LE8 = 30), and LE8 above 95 (shown at LE8 = 95). Likewise, LS7 scores of 18 and above have been pooled, as were LE8 scores of 90 or higher. Panel B provides the age-race-sex adjusted 10-year event rates from a proportional hazards model, with similar pooling for low and high LS7 and LE8 scores employed in Panel A. Note that the numeric values plotted in this figure are provided in Tables S3A and S3B.

For this reason, the assessment of the statistical significance of potential gain in information with LE8 compared to LS7 focused on the age-race-sex adjusted categorical model shown in Figure 3B. The c-statistic for the LS7 categorical proportional hazards model was 0.691 (95% CI: 0.667, 0.705), and for LE8 it was 0.695 (95% CI: 0.681, 0.709), for an increase of 0.004 (95% CI: −0.001, 0.009; p = 0.12). In contrast, the c-statistic for the LS7 continuous proportional hazards model was 0.687 (95% CI:0.673, 0.701), while for LE8 it was 0.693 (95% CI: 0.679, 0.707), for an increase of 0.006 (95% CI: 0.001, 0.011; p = 0.013).

Discussion

Counter to our hypothesis, the risk stratifications provided by LS7 and LE8 were nearly identical, and there was no evidence of increased information provided by LE8 to predict incident cardiovascular events. This lack of improvement was surprising as: 1) LE8 included sleep duration, a proven risk factor for cardiovascular disease independent of other risk factors,³² 2) a finer gradation in the quantification of the other 7 domains, where increasing the number of levels should have captured additional information that would improve risk prediction as these are causal risk factors, and 3) the “enrichment” of assessment of some domains, including a more detailed assessment of diet, passive cigarette smoke and vaping, and moving from total cholesterol to non-HDL cholesterol. However, LE8 did not appear to improve assessment of cardiovascular risk. While LE8 may be useful for population health tracking, with the added complexity and burden of measuring LE8, individuals and clinicians might prefer using LS7 in practice.

These findings are discordant from a recent report by Gao and colleagues contrasting the LS7-versus-LE8 prediction of coronary events among 339 patients undergoing percutaneous coronary intervention (PCI), where LE8 was shown to have a greater area under the receiver operator characteristic curve relative to LS7 (0.662 vs. 0.615; p < 0.05).⁴ While the reasons for the discordant findings are not clear, there are many differences between these studies, where the study by Gao and colleagues: 1) used a select population of patients with more advanced cardiac disease profile, and 2) a focused on a more broad definition of coronary outcomes (i.e., major adverse cardiac events, or MACE) with the exclusion of stroke events. Among the 339 PCI patients, 105 had a MACE event, with the most common being revascularization (n = 45), angina pectoris recurrence (n = 32), and congestive heart failure (n = 12); with only 8 recurrent myocardial infarctions. These substantial differences in the population and outcome makes the comparison of the two studies a challenge, and raises many additional questions. Among these questions is the possibility of a tighter association of the LE8 scale with coronary events than stroke events. While there were 713 coronary events in our REGARDS cohort, the complexity of adequately addressing this question will require a separate paper. Additionally, the possibility of a stronger association for LE8 among those with more advanced cardiovascular disease could be addressed by examining recurrent events in the REGARDS cohort, also a question beyond the scope of the current report.

The pattern of the estimated risk over the range of LS7 and LE8 scores has implications for interpretation of scores. Relatively small differences among very low (within the 1^st decile) or very high (within the 10^th decile) scores were associated with dramatic differences in 10-year risk of cardiovascular disease. For example, a 10-point difference between LE8 scores of 30 and 40 was associated with an in estimated risk of 27.0% versus 14.2%, approximately a halving of risk. Similarly, the risk differences between LE8 score of 85 and 95 was a reduction of approximately two-thirds, going from an estimated risk of 4.7% versus 1.3%. For those with scores in the middle of the distribution, to get the same 50% or 66% reduction requires larger differences in the percentile of scores. For example, a person with an LE8 score of 50 has an estimated 9.2% 10-year risk. To get a similar approximately 50% difference in risk, a score of 80 (30-point increase) is required (which has a risk of 4.7%). So, small differences in very high, or in very low risk scores (for both LS7 and LE8) had a larger impact on estimated risk than did larger differences in the middle of the score distribution. Therefore, the presentation of the LE8 and LS7 scores on a linear scale may be misleading to users of these scores, who may erroneously assume that any 10-point improvement in their LE8 CVH score (or, for LS7, any 1-point improvement) has similar benefits in their 10-year risk of CVD events, regardless of their starting score.

The more rapid increases and decreases in the first and tenth decile are clearly shown in the crude Kaplan-Meier analysis and the categorical proportional hazards model; however, the continuous proportional hazards model did not have sufficient inflection to capture these substantial changes in CVD risk. That the continuous risk model fails to adequately fit the data in the tails makes it substantially less attractive, and we place greater confidence in the categorical model.

The increase in the c-statistic between the LS7 and LE8 models was used to assess whether the LE8 model contained significantly more information than the LS7. The interpretation of this question is complicated by a discordant finding between the model considering LS7 and LE8 as a categorical versus continuous measure. Specifically, the c-statistic was only significantly increased for LE8 compared to LS7 for the continuous measure; however, characterizing LS7 and LE8 fails to capture the substantial risk changes in the first and tenth decile of scores. We can then conclude that the LE8 fails to add significant new information above that available in LS7.

There are several strengths and weaknesses in this report. Perhaps the greatest strength is the large number of participants from a national biracial sample and the substantial number of incident cardiovascular events, allowing a reliable description of the differences in estimated risk over the range of LS7 and LE8 scores. Additional strengths are the robust measurement of health factors and physician-adjudicated stroke and coronary outcomes. The LS7 and LE8 were proposed as indices of cardiovascular health, including both coronary and stroke outcomes. However, the performance of the indices may differ for coronary and stroke outcomes, but this was not assessed based on the intent of use of the metrics for the composite outcome. We attempted to calculate the LS7 and LE8 indices as closely as possible to the proposed scales; however, we did not have all the needed data (see Table S1 for details). Perhaps the greatest concern was for the diabetes domain, where the LE8 required access to both fasting glucose and HbA1c, specifically to better classify risk among participants with diabetes. Unfortunately, HbA1c was not available, so we assigned a single “penalty” in the score for participants with diabetes. These only comprised 1,656 (14.3%) of participants. We also classified sleep status after baseline and it was therefore missing on a large number who were excluded from the analysis. The characteristics of those excluded did/did not differ from those included. The comparative discrimination of the metrics will naturally be affected by the available data, which has some limits in REGARDS. Performance of the scales might be improved with more precise quantification of their components, for example better measures of exercise or sleep than are available in REGARDS. However, the data available in REGARDS are representative of the data that are available in many longitudinal studies, and similar to what would be ascertained in routine clinical practice, so the findings are likely generalizable.

In conclusion, we hypothesized that the refinements introduced with LE8 compared to LS7 would better differentiate risk of cardiovascular events. However, there was virtually no difference between the scales in their association with cardiovascular risk, suggesting that LS7 may be a more practical index of CVH owing to ease of its data collection and calculation.

Clinical Perspective.

What is new?

• In 2010, the Life’s Simple 7 was proposed by the American Heart Association to be a wide-used metric of cardiovascular health that included the assessment of 7 health domains: smoking, diet, physical activity, body mass index, blood pressure, cholesterol and fasting glucose.

• This metric was updated by in 2022 with the Life’s Essential 8, adding a new domain of sleep and refining/enriching the previous 7 domains.

• Both metrics are strongly predictive of incident cardiovascular events, but despite the broadening and refinements of domains in Life’s Essential 8, the predictive performance of the two scales are practically indistinguishable.

What are the clinical implications?

• Both metrics are useful to discuss the risk of incident cardiovascular events with patients, and allow the general public to understand opportunities to improve their cardiovascular health; however, the simplicity of Life’s Simple 7 may allow a more understandable communication of that risk.

Non-standard Abbreviations and Acronyms:

AHA

American Heart Association

CVD

Cardiovascular disease

CVH

Cardiovascular health

DASH

Dietary Approaches to Stop Hypertension

LE8

Life’s essential 8

LS7

Life’s simple 7

MACE

Major adverse cardiovascular events

REGARDS

REasons for Geographic And Racial Differences in Stroke