Optimization of Time Interval for the Measurement of Plasma Lipids for Cardiovascular Disease Risk Assessment

Abstract

Background:

Lipid testing for atherosclerotic cardiovascular disease (ASCVD) risk is often performed every 4–6 years, but we hypothesized that the optimum time interval may vary depending on baseline risk.

Research Design and Methods:

Using lipid values and other risk factors from the National Health and Nutrition Examination Survey (NHANES) (n=9,704), we calculated a 10-year risk score with the pooled-cohort equations. Future risk scores were predicted by increasing age and projecting systolic blood pressure (SBP) and lipid changes, using the mean-percentile age group change in NHANES for SBP (n=17,329) and the Lifelines Cohort study for lipids (n=133,540). The crossing of high and intermediate-risk thresholds were calculated by time to determine optimum intervals for lipid testing.

Results:

Time to crossing risk thresholds depends on baseline risk, but the mean increase in the risk score plateaus at 1% per year for those with a baseline 10-year risk greater than 15%. Based on these findings, we recommend the following maximum time intervals for lipid testing: baseline risk <15%: 5-years, 16%: 4-years, 17%: 3-years, 18%: 2-years, and 19%: ≤1-year.

Conclusions:

Testing patients for lipids who have a higher baseline risk more often could identify high-risk patients sooner, allowing for earlier and more effective therapeutic intervention.

Introduction

Lipid testing is a key step in atherosclerotic cardiovascular disease (ASCVD) risk assessment for the prevention of cardiovascular disease (CVD) [1–3], a leading cause of death in the US and elswhere [4]. The US 2018-Multisociety Guideline on the Management of Blood Cholesterol [5] recommends that for adults between 40 and 75 years of age with a low-density lipoprotein cholesterol (LDL-C) between 70 and 189 mg/dL and without diabetes mellitus (Type 1and 2), therapeutic decisions should be based on the individual’s 10-year risk score, calculated using the pooled cohort risk equations (PCE) [6,7]. This ASCVD risk score is based on well-established CVD risk factors [6–9], namely, age, sex, race, total cholesterol (TC), high-density lipoprotein cholesterol (HDL-C), systolic blood pressure (SBP), diastolic blood pressure (DBP), diabetes status, use of blood pressure (BP) medication, and smoking status. Given that the risk for atherosclerosis is a function of the duration of exposure to the various risk factors, age is quantitatively the most important risk factor [10]. It has been estimated that age in combination with sex and race accounts for up to 80% of the predictive power of most CVD risk models [11].

According to the US guidelines, a 10-year risk score below 7.5% is considered low or borderline risk, and statin therapy is unlikely to be required. Statin treatment may be considered when an individual’s 10-year risk score is 7.5–20% (intermediate risk). All patients with a 10-year risk score ≥20% or with an LDL-C ≥190 mg/dL or with diabetes mellitus should be offered high-intensity statin treatment because of the evidence that statins can decrease both morbidity and mortality from ASCVD in these high risk groups [5]. After age 40, and until 75 years, the 10-year risk should be re-calculated at least every 4–6 years [5].

The European and Canadian recommendations for CVD differ from US guidelines in several ways. The major differences include using different risk score calculations, such as the systemic coronary risk estimation (SCORE) in Europe [12], and the Framingham risk score (FRS) or the cardiovascular life expectancy model (CLEM) in Canada [13]. Similar to US guidelines, however, these other primary prevention guidelines also recommend estimating CVD risk and measuring lipids at least every 5 years in adult males >40 and females >50 years of age [12,13].

Recently, it was reported that nearly 93 million U.S. adults had high or borderline-high plasma TC levels (≥200 mg/dL) between 2015 and 2019 [4]. Although severe hypercholesterolemia can manifest as xanthomas and can result in other physical findings, hypercholesterolemia is normally asymptomatic. Thus, many people with hypercholesterolemia may be unaware of their high ASCVD risk and would only discover their condition if they undergo lipid testing, which typically includes TC, HDL-C, triglycerides (TG) and measured or calculated LDL-C in the standard lipid panel [14]

In alignment with the current age of personalized medicine, we developed a simple rule for optimizing the frequency of lipid testing based on persons baseline risk. For those individuals with a baseline 10-year risk of 15% or greater, we recommend that the maximum annual time interval for repeat lipid testing be calculated by subtracting the integer value of baseline 10-year risk score from 20. This would result in the more frequent monitoring of lipids in patients who are at higher baseline risk and are possible candidates for statin therapy, which could reduce delays in initiating treatment and save lives because earlier lipid-lowering intervention is more effective in reducing ASCVD events [15,16].

Patients and Methods

A retrospective observational type study was performed to determine the optimal interval for lipid testing, using data from the National Health and Nutrition Examination Survey (NHANES), which was designed to demographically match the US general population. Deidentified data was obtained from NHANES for the years 1999–2020 for patients 18 years of age and older (n=28,155) at the following website: https://www.cdc.gov/nchs/nhanes/index.htm. Details on the survey design and how the lipid tests were performed can also be obtained at this website. Patients on lipid lowering medications (n=4,721), with diabetes mellitus (n=3,361) or with LDL-C >190 (n=733) were excluded. Of the remaining individuals, only those between age 40–75 were included to create a primary prevention cohort (n=9,704). Patients between age 18–75, excluding those taking anti-hypertensive medications (n=7,104), were used for the longitudinal modeling of SBP (n=17,329) but not for the optimization of the lipid testing interval for the primary prevention part of the study. Lipid data from the Dutch Lifelines cohort study [17] of patients with no CVD or use of lipid medications was used for longitudinal modeling of TC and HDL-C (n=133,450). Research for this study was considered non-human subject research and exempted from IRB review by the National Institutes of Health.

ASCVD risk was calculated for patients aged 40–75 in NHANES, using the PCE risk equations [18]. For Black men (BM) and Black women (BW), the specific PCE equation for these groups were used, whereas for all other racial and ethnic groups PCE equations for either White men (WM) or White women (WW) were used. For stratification purposes, the 10-year risk score was truncated to the calculated integer except for patients with scores between 7–8. For these individuals, separate categories for both 7.0–7.49 and for 7.5–7.99 were created.

To predict future PCE risk score, each patient was assigned a population percentile rank by age for TC, HDL-C and SBP. Each patient was assumed to stay within their population percentile rank as they aged. A second or third order polynomial was used to develop smooth population percentile lines for each parameter for both men and women. Results between the percentile lines shown were calculated by interpolation. Other risk factors used in the calculation (smoking or use of BP medications) were assumed to remain unchanged over the subsequent years. Results were then projected for the forthcoming 5 years, using the PCE equations, by chronologically adding 1 to 5 years to the patient’s age and using their extrapolated lipid test and SBP values.

The effect of immediate repeat lipid testing on PCE risk score was determined by considering the effect of both analytical and biological variation on the risk score as previously described [19]. We used combined coefficients of variation of 7% for HDL-C, and 6% for TC. For determining the effect of repeat SBP testing on the risk score, we used a coefficient of variation of 10% based on previous studies [20].

Cost-benefit analysis was done first by calculating the number of lipid tests performed in our primary prevention cohort with the current schedule of once every 5 years versus the number of tests that would be performed with the newly developed proposed rule. Next, we determined the number of expected ASCVD events over the 5-year time period by performing lipid testing once every 5 years versus the new proposed rule for those individuals with a PCE risk score ≥16%. We used each individuals PCE risk score to calculate their annual ASVD event rate. For those individuals that were predicted to cross the high-risk 20% threshold, we assumed that they would be placed on a statin and lowered their annual ASCVD event rate by 33% based on previous on studies on statin therapy [21].

Data analysis was done with JMP software (JMP,Cary, NC) or in Excel (Microsoft, Redmond, WA). Data for key findings and software and spreadsheets developed for this study can be downloaded at Fig Share website: (https://figshare.com/articles/software/Sampson_PCE_Risk_Projection_Calculator/24156333).

Results

A cohort (n=9,704) of individuals aged 40 to 75 was assembled from NHANES to represent a primary ASCVD prevention cohort from within the US general population. Individuals already on lipid-lowering medication or those with diabetes or with a LDL-C>190 mg/dL were excluded. The mean age was 54.6 and 48.4% were men. 79.1% were White and 20.9% were Black. Other demographic features of the cohort can be seen in Supplemental Table 1. Another cohort of all patients from age 18–85 was also assembled for examining the longitudinal risk factor changes over time (Supplemental Table 2).

As depicted in Figure 1, we first assessed the impact of immediately repeating lipid testing and SBP measurement on the 10-year risk score to determine the likelihood of crossing a risk threshold due to the biological and analytical variability of lipid testing and SBP measurement. To simulate repeat testing, we used a combined biological variation and analytical imprecision of 7% for HDL-C, 6% for TC and 10% for SBP based on previous studies on the variation of these parameters [19,22]. We tested the scenario of repeating only lipids (TC and HDL-C), repeating only SBP measurement, or both. Figure 1 shows that up to 40% of patients with an initial 10-year risk score just below the intermediate risk (7.5%) or high-risk (20%) thresholds are likely to cross into the next higher risk category after repeat testing. Based on this analysis, it may be useful to do immediate repeat testing of lipids and SBP on all patients with PCE scores that are within 1% of a risk threshold, particularly if it could possibly change clinical management, as we will discuss later.

Figure 1. Effect of biological and analytical variation on PCE risk score after repeat testing.

Individuals are grouped by their baseline PCE risk scores (x-axis) for those either below the intermediate risk (7.5%/10 years) threshold (Panel A) or below the high risk (20%/10 years) threshold (Panel B). The y-axis is represents the percent of each baseline risk group that crosses the intermediate (Panel A) or high risk threshold (Panel B). Three different scenarios are investigated: repeating lipids only (red), repeating SBP only (green) or repeating both lipids and SBP (blue).

Next, we did a cross-sectional analysis of our whole primary prevention cohort to examine their ASCVD risk categories by age (Fig. 2). Those aged 45 years or less mostly had a 10-year risk score of less than 5% (low risk) and relatively few were at intermediate or high risk. As age increases, first the borderline-risk group (5–7.4% 10-year risk) increases followed by the intermediate-risk group. The highest-risk group (>20% 10-year risk) does not begin to appear until about age 55 to 60. Thereafter, the frequency of this high risk group rapidly increases, and it becomes the majority risk category after about age 72.

Figure 2. Association of age with different PCE risk categories.

The number of individuals in the primary prevention cohort with PCE risk scores in the low risk (<5%/10 years, blue), borderline risk (5–7.5%/10 years, red), intermediate risk (7.5–20%/10 years, green) or high risk (≥20%/10 years, yellow) groups are shown by age (Panel A) or as a percentage of the total population (Panel B).

To determine the main drivers of the increased ASCVD risk with age, we examined the individual risk factors used in the PCE score after age 18 (Fig. 3). We did this for the four main race/gender groups for which the PCE scores were specifically developed [6]. Similar to what has been previously described [23], TC increased with age. It did so more quickly in men and peaked at about age 50 and then slowly decreased with age. The peak in TC for women does not occur until about age 60 and then decreases more slowly compared to men. In contrast, HDL-C steadily increases with age for both sexes and, as previously described, is higher for women versus men and for Blacks versus Whites [24]. SBP also steadily increases with age in an almost linear fashion. It increases more quickly in women than in men, but women start at a lower SBP baseline and therefore their mean SBP does not surpass that of men until after age 70. Both the Black men and Black woman groups have a higher mean SBP than their White counterparts. Blood pressure medication use also steadily increases with age and was more common among Black men and Black women. Diabetes steadily increases with age but decreases after 70 years, likely due to the decreased longevity associated with diabetes [25]. Finally, use of cigarettes rapidly increases in the early 20s and then plateaus until it starts to decrease in the late 50’s to early 60s. Although not assessed in this study, among all the risk factors tested, age is known to have the largest quantitative impact on the PCE risk score followed by SBP, HDL-C, TC, diabetes, BP medications and cigarette smoking when considered at the population level [10].

Figure 3. Changes in ASCVD risk factors with age.

The trends by age are shown for black women (BW, blue), black men (BM, red), white women (WW, green), and white men (WM, yellow) for mean total cholesterol (Panel A, TC), mean HDL-cholesterol (Panel B, HDLC), mean systolic blood pressure (Panel C, SBP), percentage taking blood pressure medications (Panel D, BP meds), percentage with diabetes (Panel E), and percentage who currently smoke (Panel F).

To project future risk scores, we developed population percentile based algorithms to forecast TC, HDL-C and SBP based on large population studies. Supplemental Figure 1 shows the mean percentile trend data for change over time for TC, HDL-C and SBP from large population studies [26]. An example of the smoothed curves is shown in Figure 4 for TC. Similar graphs for HDL-C and SBP are shown in Supplemental Figure 2. The population percentile level for each patient is calculated based on their age, sex and baseline test results. We assumed that in the next five years the patient will remain at their same percentile for each risk factor. The PCE score for each individual was then recalculated for each successive year with the predicted changes in TC, HDL-C and SBP over time and by chronologically increasing age. Other risk factors, such as diabetes, smoking or use of BP medications were assumed not to change from baseline.

Figure 4. Percentile population curves for changes in plasma total cholesterol by age.

Mean data from the Copenhagen Lifelines cohort for males (Panel A) and females (Panel B) are shown for total cholesterol (TC) for the various percentile levels seen in the study. Curves were fitted by second or third order polynomials.

Patients were grouped based on their baseline 10-year risk score, and the percentage of individuals per group who were projected to cross the intermediate or high-risk thresholds are shown in Figure 5. Not unexpectedly, the closer a patient’s baseline 10-year risk score is to either the intermediate or high-risk threshold, the sooner they cross it in subsequent years. By analyzing the rate of change in the PCE risk score (Fig. 6A), we observed that those patients with a higher baseline risk increased the fastest. For those, however, with a baseline PCE score of 12% or more, the mean annual increase in their PCE score plateaued at about 1% per year (Fig. 6B). When we further divided the primary prevention cohort into different age groups (40–49, 50–59, >60), the youngest age group showed a slower annual rate of change in their PCE score, but for those with a baseline risk of 15% or more it also started to approach 1% per year (Supplemental Figure 3). Based on this analysis, one can easily project the time that any individual will likely take to cross the high-risk threshold by subtracting from 20 the integer value of their baseline PCE score, which results in the values shown in Table 1. For those with a PCE score of 19%, it may be reasonable, however, to repeat lipid testing immediately, given the impact of analytical and biological variation of lipid testing on the PCE risk score (Fig. 1).

Figure 5. Predicted percent of population that cross intermediate and high risk thresholds by age.

The percentage of the population in the primary prevention cohort (y-axis) that are predicted to cross either the intermediate risk (Panel A, 7.5%/10 years) or high risk (Panel B, 20%/10 years) thresholds in the future years (x-axis). Results are separated into groups by baseline risk score (see legend) expressed as a whole number rounded down.

Figure 6. Annual predicted rate of change of PCE scores.

The change in the mean PCE score is shown between the baseline risk score and subsequent ones projected for 5 years for the primary prevention cohort. Results are separated into groups by the baseline risk score (see legend) expressed as a whole number rounded down, with the number in each group shown in the figure legend. Panel A shows results for each year. Panel B is the annual rate of change (slope of lines in Panel A) by each baseline risk score group.

Table 1:

Optimized lipid test frequency for patients with different baseline PCE scores.

PCE score (%)	<16	16.0–16.9	17.0–17.9	18.0–18.9	19.0–19.9
# years until repeat testing	5	4	3	2	≤1

Finally, we did a simple cost-benefit analysis for implementing the new rule for lipid test frequency on our cohort of primary prevention patients over a five-year period (Fig. 7). Compared to lipid testing once every 5 years, only a relatively small number of additional lipid tests (2.13 tests/patient versus 2 tests/patient) would be performed under the new optimized test algorithm rule (Table 1), because the percent of the total primary prevention population near the high-risk threshold is relatively low (Fig. 7A). Assuming that statin therapy lowers ASCVD events by approximately 33% [27], the projected number of total events decreases from 10.7% to 9.2% for an overall decrease of 14.4% (Fig. 7B). Although this may seem a relativley small improvement, the benefit is calculated for only 5 years. In addition, as shown in Fig. 7C, some subgroups, based on their baseline risk, showed an even greater improvement with the new lipid testing schedule. For those with a 19% baseline risk their ASCVD events decreased from 11.6% to 8.8% with the new lipid testing schedule. In contrast, comparatively less benefit was found for the group with a baseline risk of 16%. This is because the new lipid testing schedule only reduces their retesting interval by one year and, therefore, they could only potentially benefit from one additional year of statin therapy.

Figure 7. Cost benefit analysis for new lipid test frequency intervals.

The number of lipid tests performed for a 5-year period are shown for the primary prevention cohort for the current 5-year test schedule (blue) or for the new lipid testing schedule, which allows for more frequent testing in patients with risk scores close to the high-risk threshold (red) (Panel A). The percent of predicted ASCVD events are shown for the current 5-year testing schedule (blue) or the new lipid testing schedule (red) for the whole cohort (Panel B) or for the cohort subdivided (x-axis) into their baseline risk groups (Panel C).

Discussion

Cardiovascular risk assessment, particularly lipid testing, is a crucial part of preventive medicine. It aids in the identification of individuals who are at risk for ASCVD and could, therefore, benefit from lipid-lowering therapy [28]. While periodic lipid testing is widely acknowledged to be important in monitoring primary prevention patients [5,13,29], the optimum time interval for lipid testing for primary prevention has not been carefully evaluated. Our main finding, which is perhaps not unexpected but has important practical implications, is that those patients with a high baseline risk should have their lipids monitored more frequently.

Over an individual’s lifespan, lipid profiles can follow distinct trajectories, which significantly influence the development of CVD risk [26]. For example, as we observed for TC, between the ages of 20 and 40, it exhibits a rapid upward slope, which becomes more gradual as individuals approach 50 to 60 years of age and then frequently decreases. It is important to recognize, however, that it is the cumulative exposure to cholesterol in pro-atherogenic lipoproteins over time that contributes to an individual’s overall burden of CVD risk [15]. It is not only lipids, however, but also other risk factors, which frequently change with age, that are used for determining a patient’s 10-year risk score and their eligibility for statins. We, therefore, hypothesized that the 10-year risk score could be a major determinant of how often lipid testing should be done. In particular, those patients that are close to the high-risk threshold of 20% may need more frequent monitoring, if crossing this threshold would change clinical management. Evidence supporting the use of statin therapy in the high-risk group is much stronger than for other groups [30], so crossing this threshold adds compelling new information in favor of initiating statin therapy.

For our analysis, we developed an algorithm that allowed us to project a person’s 10-year risk score over the next 1–5 years. Given that age has such a dominant effect on increasing risk [10], simply recalculating risk after incrementally increasing the age of each person provides a reasonable estimate of future risk, particularly if one also considers the sex and race of the patient [11]. To further improve our risk forecasting, we also considered age-related changes in lipids and SBP. We did not, however, consider changes in other risk factors, such as diabetes, so we likely underestimated the true future risk. We also only used conventional lipid markers like HDL-C, which has not been shown to be a therapeutic target but nevertheless predicts ASCVD risk. We did not assess other alternative biomarkers like nonHDL-C and apoB, which may be superior risk markers. Despite these limitations, we uncovered a relatively simple relationship between conventional ASCVD risk markers and age that may have great practical value in the management of patients. We observed that risk increases by a minimum of approximately 1% per year for those with a baseline 10-year risk score of greater than 15%. Thus, the time after which a patient will likely cross the high-risk threshold (≥20% 10-year risk) can easily be calculated by subtracting from 20 the integer value of the baseline 10-year risk score when it is above 15% and does not require the much more complicated approach, we used in our initial analysis for projecting future risk. For those with a baseline 10-year risk of 15 or less, our results indicates that the current recommended practice of testing lipids every 4–6 years is reasonable. The newly proposed intervals for repeat lipid testing, however, should be just considered as a guide. More frequent lipid testing could be desirable, depending on the circumstances of a particular patient. If a patient, for example, starts smoking or develops diabetes or has some other risk factor that we did not consider in our analysis, they may benefit from even more frequent testing. Our analysis does indicate that because of the major effect of age on ASCVD risk those patients without any other change in risk factors should be more closely monitored for lipids than every 4–6 years, if they are close to the 20% 10-year risk score.

A possible exception to the above rule exists for patients who are within one unit of the high-risk score threshold. It might be best for these patients to immediately undergo immediate repeat testing rather than waiting a year, because the high biological and analytical variability of lipid and SBP estimations make it likely that the risk scores of these patients may exceed this threshold on immediate repeat testing. A similar immediate repeat testing strategy may also lead to earlier detection of patients who are close to crossing the 7.5%, intermediate-risk threshold. There is less clinical justification for doing so in this group, however, unless it is expected that it would change clinical management. If not, based on their slower mean annual increase in PCE score, testing this lower risk population for lipids every 4–6 years is reasonable. Another possible exception to our new repeat testing interval rule may exist for patients that have a LDL-C value close to 190 mg/dL. Although many of these patients will likely have a high 10-year risk score, some younger patients with a genetic predisposition to hypercholesterolemia may not, and could be missed. It would be reasonable, therefore, for patients with a known genetic predisposition to also undergo more frequent lipid testing.

Given the observed faster rate of change observed in the PCE score of the higher risk group, the use of the commonly used 4–6-year lipid testing interval in these patients may lead to unnecessary delays in starting lipid-lowering therapy. Based on our analysis, even within a short 5-year period, more frequent lipid-testing could reduce ASCVD events by earlier initiation of statin therapy in high-risk patients. Over the long-term, the positive impact of early statin intervention may be even greater owing to reduced the lifetime exposure to atherogenic lipoproteins [15]. This phenomenon was clearly observed in the follow-up of the West of Scotland Coronary Prevention Study (WOSCOPS), a landmark trial in primary prevention of CVD [16]. Despite eventually receiving statin therapy at comparable rates (38 vs. 35%) after the conclusion of the 5-year study, the original control group never achieved the same benefits observed in the original statin treatment group in terms of cardiovascular outcomes. In a 20-year follow-up of this study, individuals who were in the initial statin arm continued to demonstrate a significant 21% reduction in cardiovascular mortality compared to the control group who were put on statins later (hazard ratio: 0.79; 95% confidence interval: 0.69–0.90; p<0.01). Additionally, there was a notable decrease in cumulative hospitalization rates: a decrease of 18% for any coronary events, 24% for myocardial infarctions, and 35% for heart failure in the initial statin treatment group versus the control group. This observation highlights the fact that, when indicated, initiating statin therapy as early as possible is desirable because of its improved effectiveness in reducing ASCVD events.

Knowledge of the annual mean change of the PCE score at different baseline risks may also facilitate the shared decision-making process that is recommended between patients and their healthcare providers [30]. For example, a patient at intermediate risk, not already on a statin, but told that they will likely cross the high-risk threshold in a few years, may decide to start statins right away rather than wait. Such a clinical scenario may then obviate the need to do ancillary testing, such as other risk enhancer tests or obtaining a coronary artery calcium score, if a decision to start statins has already been made based on the projection of future risk.

It is important to note that our study has several strengths and weaknesses. First, our analysis was largely based on a cohort from the NHANES, a general US adult population study, making our findings possibly less relevant to other demographic groups and to non-US populations. Furthermore, other countries often use different types of ASCVD risk scores than the PCE risk scores used in our study. It is likely, however, that our main finding on how higher risk individuals would benefit from more frequent lipid testing will still hold in other populations and our approach, with some modifications, could possibly be used elsewhere but will require external validation. Another limitation of our study is that we did not directly assess the change in lipids, SBP and PCE risk scores over time, but rather predicted it, which required several assumptions that could decrease the certainty of our findings. We also only used two groups, males and females, in our modelling of the change in PCE scores over time. More precise predictions of future PCE scores can probably be obtained by more careful analysis of subgroups and by considering other factors. This would be at the expense, however, of increasing the complexity of the algorithm and making it harder to implement in routine clinical practice. It is also not clear that more precise estimates of future risk would significantly alter the simple lipid test frequency rules that we developed, which were also partly based on pragmatic considerations.

Conclusions

Individualizing the frequency of lipid testing based on baseline risk would allow a more prompt identification of candidates for statin therapy and would reduce lifetime exposure to atherogenic lipoproteins. The current one-size-fits-all approach of testing lipids once every 4 to 6 years fails to account for the substantial interindividual variability in baseline ASCVD risk and how this could affect the clinical management of patients. Adopting our new rules for the frequency of lipid testing would be relatively easy to implement and would likely also be cost-effective based on our modelling showing less ASCVD events in high-risk patients who have more frequent lipid testing.

Supplementary Material

Figure S1

Supplementary Figure 1. Percentile population curves for changes in plasma lipids and SBP by age. Mean lipid results from the Copenhagen Lifelines lipid cohort is summarized into 5 year age increments and indicated percentile levels for total cholesterol for men (Panel A, TC M), total cholesterol for women (Panel B, TC W), HDL-cholesterol for men (Panel C, HDL-C M), and HDL-cholesterol for women (Panel D, HDL-C W). Mean SBP from the NHANES population is summarized into 5 year age increments and indicated percentile levels for systolic blood pressure for men (Panel E, SBP M) and systolic blood pressure for women (Panel F, SBP W).

Figure S2

Supplementary Figure 2. Smoothed percentile population curves for changes in plasma lipids and SBP by age. Results from Supplementary Figure 1 were smoothed using second or third order polynomials. Results between main percentile curves were calculated by interpolation.

Figure S3

Supplementary Figure 3. Annual predicted rate of change of PCE scores. The change in the mean PCE score is shown between the baseline risk score and subsequent ones projected for 5 years for primary prevention cohort for those between age 40–49 (Panel A), between age 50–59 (Panel B) and for those ≥ 60 years (Panel C). Results are separated into groups by the baseline risk score (see legend) expressed as a whole number rounded down with the number in each group shown in the figure legend. Panels A-C shows results for each year. Panel D is the annual rate of change (slope of lines in Panel A-C) by each baseline risk score group and age group (see legend).