Age-stratified comparison of heart age and predicted cardiovascular risk in 370 000 primary care patients

Abstract

Background

Cardiovascular disease (CVD) preventive medications are recommended for patients at high short-term CVD risk. As most younger people with multiple raised CVD risk factors levels have low short-term risk, they could be falsely reassured to take no action. Heart age—the chronological age of a hypothetical person with the same short-term absolute CVD risk as the patient being assessed, but with an ‘ideal’ risk profile—is a complementary relative CVD risk metric developed to encourage these younger patients to make long-term lifestyle changes. However, clinicians sometimes use heart age to inform medication decisions. We assessed the appropriateness of this practice by comparing heart age and short-term CVD risk.

Methods

New Zealand primary care patients are recruited to the PREDICT cohort when their CVD risk is assessed. PREDICT is an ongoing prospective study in one-third of New Zealand general practices, designed to derive CVD risk prediction algorithms. Five-year CVD risk was calculated for 35–74-year-old PREDICT participants using published equations. Heart age was calculated using non-smoking, systolic blood pressure of 120 mm Hg and total cholesterol/high-density lipoprotein ratio of 3.5, as the ‘ideal’ risk profile. CVD risk and heart age gaps (difference between chronological age and heart age) were compared.

Results

Among 371 676 PREDICT participants, 5-year CVD risk increased with age, approximately doubling every 10 years, whereas heart age gaps decreased with increasing age, approximately halving between 35 and 44-year olds and 65–74-year olds. There were 5–40-year heart age gap differences between groups with similar 5-year CVD risks, but different ages.

Conclusion

Short-term CVD risk and heart age are not interchangeable risk metrics. Short-term risk increases with increasing age whereas heart age gaps generally decline, with major differences between younger and older people with similar short-term risk. If heart age is used to inform medication decisions rather than encourage long-term lifestyle changes, older people at high short-term risk could be undertreated and younger people at low short-term risk could be unnecessarily medicated.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Heart age is a commonly used relative risk metric designed to communicate and aid understanding of the future potential impact of modifiable risk factors on arterial health and to encourage lifestyle changes. In contrast, predicted short-term cardiovascular risk is an absolute measure of risk that can be used to estimate the short-term absolute benefits of drug treatment. However, these two measures of cardiovascular risk are sometimes used interchangeably to inform drug treatment decisions.

WHAT THIS STUDY ADDS

• While relative measures of risk, like heart age, are easier to communicate to patients than absolute measures, like predicted short-term cardiovascular risk, this study demonstrates that they are very different measures of risk and that informing treatment decisions using the calculated heart age could result in both under and over-treatment.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• These findings should help clinicians make more informative decisions about the management of cardiovascular disease risk, based on two distinctly different measures of risk.

Introduction

Most cardiovascular disease (CVD) risk management guidelines recommend that preventive medication decisions should be informed by patients’ predicted short-term (5 or 10 years) absolute CVD risk,¹,³ as the short-term benefits of preventive medications (eg, lipid-lowering and blood pressure-lowering drugs) have been shown to be directly proportional to patients’ predicted short-term CVD risk.^{4 5} However, younger people with a high burden of modifiable risk factors are generally at low short-term CVD risk because exposure to risk factors can take decades to impact significantly on absolute CVD risk. As a result, these younger patients may be falsely reassured that they do not need to be concerned about their CVD risk. To address this issue, other CVD risk metrics have been developed to help communicate longer term risk and encourage younger people to make lifestyle changes. A longer term risk metric now included in many CVD risk calculators is the estimated ‘heart age’.⁶,⁸

Heart age is the chronological age of a hypothetical person with ‘ideal’ modifiable CVD risk factor levels who has the same predicted short-term CVD risk as the patient being assessed. For example, if a 40-year-old men’s current predicted 5-year CVD risk is 5%, but he would not reach this level of risk until he was 57 years old if he had an ideal modifiable risk profile, then his heart age would be 57 years and the gap between his heart age and chronological age would be 17 years.

Heart age has considerable face validity as a CVD risk metric because risk is described in absolute units of years, which is easy to communicate. However, heart age is a relative measure of risk that was not intended to be used interchangeably with short-term predicted absolute CVD risk to inform short-term medication decisions.^{9 10} Yet clinicians sometimes use heart age to inform medication decisions as well as to encourage long-term lifestyle changes.

The objective of this study was to assess the appropriateness of this practice by empirically comparing heart age gaps and short-term absolute CVD risk in a primary care population.

Methods

Study population

The study population has been described in detail elsewhere.¹¹ In brief, PREDICT is an ongoing, prospectively designed, open cohort study that automatically recruits participants when primary healthcare practitioners in New Zealand complete CVD risk assessments using PREDICT decision support software. When opened, the software auto-populates the PREDICT CVD risk template from patient records. Clinicians then fill in any missing fields before a patient’s 5-year CVD risk is calculated and recruitment completed. Patient risk factor profiles are regularly linked to national databases documenting ICD-coded CVD hospitalisations and deaths. Approximately 90% of all eligible patients in New Zealand had completed standardised CVD risk assessments in the previous 5 years by the end of 2016¹² and approximately one-third of all New Zealanders eligible for CVD risk assessments are enrolled in primary care practices using PREDICT software. The study population (the PREDICT Heart Age cohort) used in these analyses includes patients recruited into PREDICT between 2004 and 2018, including patients receiving blood pressure and lipid-lowering medications.

We excluded people with diabetes from this study because New Zealand guidelines recommend risk assessing this patient group with equations derived specifically among people with diabetes¹ and this would have involved replicating all the analyses presented here. Recent European Society of Cardiology guidelines also recommend using diabetes-specific equations in people with diabetes.²

We also excluded people aged 30–34 years as they were not representative of primary care patients in this age group. The enrolment numbers and exclusions in the PREDICT Heart Age cohort used in the analyses presented in this paper are shown in online supplemental figure S1. Fewer than 1.5% of participants had missing variables, most due to out-of-date lipid measurements and were excluded from the analyses.

Calculating 5-year CVD risk, heart age and heart age gaps in the study population

Each person in the PREDICT Heart Age cohort had their heart age calculated by the three-step process illustrated in figure 1, using the methods described in our previous heart age paper.⁶ First, each participant had their baseline 5-year absolute CVD risk calculated using an updated version of published PREDICT equations,¹³ (online supplemental appendix table 1), based on their measured CVD risk factor profile at recruitment (the red dot shown in the left section of figure 1). Second, each participant’s ‘ideal’ 5-year CVD risk was calculated for each year from their current age, using their measured baseline CVD risk factor levels, except for smoking, systolic blood pressure (SBP) and the total cholesterol to high-density lipoprotein cholesterol ratio (TC/HDL), which were reset to the ‘ideal’ risk profile (the blue line shown in the right section of figure 1). In this study, we used non-smoking, an SBP of 120 mm Hg and a TC/HDL ratio of 3.5 as the ‘ideal’ modifiable risk factor profile. Third, heart age was calculated by determining how old each participant would have to be to reach their current calculated 5-year CVD risk, if they had the ‘ideal’ modifiable risk profile (the intersection of the horizontal line drawn from the patient’s current CVD risk to the blue line, shown in the right section of figure 1). Finally, the heart age gap was calculated by subtracting the current chronological age of participants from their calculated heart age.

Figure 1. The three-step process of calculating heart age. CVD, cardiovascular disease.

Statistical analyses

Baseline characteristics were reported as frequencies and/or percentages for categorical variables and means for continuous variables, stratified by age group and sex. Variation is represented by 95% CI. Heart age gaps were similarly reported by means and 95% CI, with stratification by age group, sex and 5-year CVD risk. 5-year CVD risk and heart age gaps, stratified by age group and sex, are also illustrated in Box and Whisker plot figures. The boxes show quartile and median levels while the Whiskers show maximum and minimum values. The non-parametric test for trend across ordered groups developed by Cuzick¹⁴ was used to assess the statistical significance of the trends for predicted median 5-year CVD risk and heart age gaps, by age group and sex. The non-parametric test is an extension of the Wilcoxon rank-sum test, with incorporated correction for ties. All analyses were done using STATA edition V.14.2.

Patient and public involvement

There was no patient or public involvement in this study.

Results

Study cohort

After exclusions, the PREDICT Heart Age cohort included 210 890 men and 160 786 women aged 35–74 years without prior CVD or diabetes (table 1, online supplemental appendix table S2 and figure S1).

Table 1. Baseline characteristics of the PREDICT Heart Age cohort by age group and sex.

	35–44 years		45–54 years		55–64 years		65–74 years
	Men	Women	Men	Women	Men	Women	Men	Women
Total: n (%)	46 011 (85%)	8087(15%)	88 224 (63%)	51 164 (37%)	51 332 (42%)	71 251(58%)	25 323 (46%)	30 284(54%)
Current smoking(%)95% CI	2322.9 to 23.7	1917.8 to 19.6	1716.5 to 17.0	1918.3 to 19.0	1313.1 to 13.7	99.0 to 9.4	88.0 to 8.7	66.0 to 6.6
Mean SBP (mm Hg)95% CI	125124.7 to 124.9	123123.0 to 123.7	127127.1 to 127.3	125125.0 to 125.3	131131.3 to 131.6	128128.3 to 128.5	134134.1 to 134.5	135134.7 to 135.0
Mean TC/HDL(ratio)95% CI	4.74.7 to 4.7	3.94.0 to 4.0	4.44.4 to 4.4	3.73.8 to 3.8	4.24.2 to 4.2	3.53.6 to 3.6	4.04.0 to 4.0	3.43.6 to 3.6

Point estimates of smoking prevalence, mean SBP and TC/HDL are rounded to nearest whole number, 95% CIs rounded to one decimal point. A complete table of participant characteristics is in online supplemental appendix table S2.

SBP, systolic blood pressure; TC/HDL, total cholesterol to high-density lipoprotein cholesterol ratio

As risk assessment is recommended 10 years earlier for men than women in New Zealand, there were more men than women in the younger age groups. The youngest age group (35–44 years) had the highest proportion of current smokers and smoking prevalence decreased with increasing age. Clear separation of the 95% CIs for these comparisons indicates these differences were all statistically significant. Mean SBP increased with age and was higher in men than women except in the oldest age group. The mean TC/HDL ratio was higher in men than women in all age groups and decreased significantly with increasing age in both sexes.

Five-year CVD risk and heart age gaps

The distributions of predicted 5-year CVD risk by age group and sex in the study population are shown in figure 2a and online supplemental appendix table S3. Predicted 5-year CVD risk increased with age, approximately doubling with each consecutive 10-year age group in both men and women. The test for trend was significant at p<0.01 in both men and women. The median 5-year CVD risk in men increased from approximately 1.4% in 35–44-year olds to 8.8% in 65–74-year olds. The equivalent median 5-year CVD risk by age group in women was about half those in men, increasing from 0.7% (35–44 years) to 5.1% (65–74 years). The ‘ideal’ median 5-year CVD risk by age group and sex, shown in figure 2 by the continuous solid line (also online supplemental appendix table S3), approximately aligns with the 25th centile of predicted CVD risk in both men and women.

Figure 2. (a) 5-year cardiovascular disease risk by age group and sex, (b) heart age gap by age group and sex. CVD, cardiovascular disease.

Figure 2, online supplemental appendix table S3, shows the heart age gap by age group and sex. The median heart age gap and its variance decreased with increasing age and the test for trend was significant at p<0.01 in both men and women. The median heart age gap approximately halved between the youngest (35–44 years) and the oldest (65–74 years) age groups and was about two-fold higher in men than women (online supplemental appendix table S3). Table 2 also shows a negative heart age gap for some members of the cohort in all age groups, particularly for women, indicating that a proportion of the population had a ‘better than ideal’ risk factor profile.

Table 2. Mean heart age gap by predicted 5-year CVD risk category, age group and sex.

	Heart age gap (years), mean (95% CI)
	Men				Women
	35–44 years	45–54 years	55–64 years	65–74 years	35–44 years	45–54 years	55–64 years	65–74 years
5-year CVD risk
<2.5%	3.93.9 to 4.0	2.22.1 to 2.2	−0.4−0.5 to −0.3	−3.6−4.3 to −2.9	2.72.5 to 2.8	1.71.7 to 1.8	0.60.5 to 0.6	−1.0−1.0 to −0.9
2.5-<5%	12.612.5 to 12.7	7.17.1 to 7.2	2.82.8 to 2.8	0.10.1 to 0.2	14.914.4 to 15.4	9.59.4 to 9.6	3.93.9 to 4.0	1.01.0 to 1.1
5-<10%	19.319.0 to 19.7	12.312.2 to 12.4	6.06.0 to 6.1	2.52.4 to 2.5	21.619.7 to 23.4	14.213.9 to 14.5	7.67.5 to 7.7	2.42.3 to 2.4
10-<15%	27.325.6 to 28.9	17.617.2 to 18.1	10.09.8 to 10.2	4.24.1 to 4.3	24.824.8	18.617.2 to 20.1	10.910.4 to 11.3	4.84.6 to 5.0
>15%	47.328.9 to 65.7	21.119.6 to 22.7	13.112.6 to 13.5	6.26.1 to 6.4	0*	18.414.7 to 22.1	12.711.2 to 14.2	5.75.3 to 6.0

^*nNo women aged 35–44 years had a predicted risk of CVD>15%.

CVDcardiovascular disease

Table 2 shows the mean heart age gap stratified by age group, sex and predicted 5-year CVD risk. The heart age gap increased as predicted 5-year CVD risk increased in all age groups and in both men and women, but more so in younger than older people. For example, in men aged 35–44 years, the mean heart age gap increased from 3.9 (95% CI 3.9 to 4.0) years in those with a predicted 5-year CVD risk <2.5%, to 47.3 (95% CI 28.9 to 65.7) years in those with a CVD risk >15%, whereas in men aged 65–74 years, the equivalent mean heart ages were −3.6 (95% CI −4.3 to −2.9) years and 6.2 (95% CI 6.1 to 6.4) years.

There were also clear differences in the heart age gap in different age groups among people in the same 5-year CVD risk category, with older people having smaller heart age gaps than younger people. For example, among women with a predicted 5-year CVD risk of 5 to <10%, the mean heart age gap decreased from 14.9 (95% CI 14.4 to 15.4) years in those aged 35–44 years to 1.0 (95% CI 1.0 to 1.1) years in those aged 65–74 years. All differences in mean heart age gaps by increasing age and increasing 5-year CVD risk were statistically significant as demonstrated by distinct, non-overlapping 95% CIs.

Discussion

To our knowledge, this is the first study to have systematically compared predicted short-term CVD risk and heart age gaps in a contemporary primary care population. The study demonstrates that these two metrics trend in opposite directions with increasing age and therefore reflect quite different dimensions of CVD risk. This observation has important implications for clinical practice, as anecdotally, clinicians and patients sometimes consider predicted short-term CVD risk and heart age to be almost interchangeable metrics and that either could be used to inform medication decisions. However, a patient’s predicted short-term CVD risk is an absolute measure that is directly proportional to the absolute short-term benefits of preventive treatment,^{4 5} whereas the heart age gap is a relative measure for communicating longer term CVD risk¹⁰ and is only weakly associated with short-term benefits of drug treatment. As a relative measure, it will also be sensitive to the choice of an ideal risk profile.

The main strength of this study is that predicted CVD risk and calculated heart age are compared in a large real-world population in whom regular CVD risk assessments is recommended by national guidelines.¹¹ Missing data were minimal (<1.5% of participants) as all data fields must be completed prior to submitting for CVD risk calculation, and the exclusions were largely due to our decision to exclude lipid measurements done more than 5 years before recruitment. Given the small proportion of participants with missing data, they were excluded from the analyses. We calculated heart age using our previously published methodology,⁶ which has been widely applied in clinical practice in New Zealand for over a decade. A potential limitation of the study is that men comprise the majority of participants under age 45 years, and to a lesser extent under 55 years. This reflects national guidelines that recommend CVD risk assessments at a younger age for men than women and for high-risk ethnic groups.¹ Moreover, the aim of these population-specific recommendations was to target higher risk patients, so while the study population was not representative of all New Zealanders aged 35–54 years, it was representative of younger people at high CVD risk who meet national guideline recommendations for CVD risk assessments. As discussed in the Methods section, people with diabetes were excluded from the analyses, because a different CVD risk assessment equation is used for people with diabetes.

A number of heart age calculators have been developed over the past two decades and they are frequently accessible on public-facing websites.⁶,^{815 16} They all include an ‘ideal’ modifiable risk profile as the reference point, which is typically based on smoking status (non-smoker) and ideal levels of blood pressure and blood lipids. However, the ‘ideal’ blood pressure and lipid levels vary between heart age calculators⁶,^{816 17} as there are no definitive ‘ideal’ levels. SBP levels of either 120 mm Hg or 125 mm Hg and TC/HDL ratios of either 3.5 or 4.0 are the most common values used in heart age calculators. We used an SBP of 120 mm Hg and a TC/HDL ratio of 3.5 because they gave 5-year CVD risk estimates that were close to the 25th centile values in each of the age/sex categories in the PREDICT cohort participants (continuous lines across age groups in figure 2), which we considered should be achievable for most people. In preliminary analyses, the higher SBP (125 mm Hg) and TC/HDL ratio (4.0) levels used in some heart age calculators, gave 5-year risk estimates that were close to or above the median values in approximately half the PREDICT cohort participants (not shown).

Bonner and colleagues have written extensively about the difference between heart age—a relative measure—and predicted short-term CVD risk—an absolute measure and why heart age should not be used to inform drug treatment decisions.^{9 10} Clinical guidelines worldwide consistently recommend that drug treatment decisions should be informed primarily by a person’s short-term absolute CVD risk,¹,^{318 19} and we have demonstrated empirically that a low short-term CVD risk can be associated with a large heart age gap in younger people, while a high short-term risk can be associated with a small heart age gap in older people. Therefore, if heart age was used to inform medication decisions, it could lead to both overtreatment in younger people and undertreatment in high-risk older people. An alternative approach to addressing long-term risk in young people at low short-term risk has been to predict longer term absolute CVD risk, for example, the American Heart Association new 30-year CVD risk prediction equations.¹⁷ However, as with heart age, basing current drug treatment decisions on the predicted risk of events that might occur decades into the future could also lead to overtreatment of young people at low predicted short-term risk.

In conclusion, this study demonstrates major differences between predicted short-term absolute CVD risk and heart age gaps. We believe both metrics are important complementary measures for communicating CVD risk, but it is important that clinicians and patients recognise that short-term predicted CVD risk should be primarily used to inform decisions to initiate medications and heart age to primarily communicate the need to make longer term lifestyle changes.