Relationship of Arterial Stiffness Index and Pulse Pressure With Cardiovascular Disease and Mortality

1Analyses of the largest arterial stiffness index data set (n=169 613) to date indicated that it was an independent predictor of incident cardiovascular disease and all‐cause mortality.Arterial stiffness index improved the 5.9‐year risk prediction model of incident cardiovascular events by 2.3% when added to the Framingham Risk Score.Pulse pressure improved the 5.9‐year risk prediction model by 5.4% when added to the Framingham Risk Score. Pulse pressure appears to have more added value than arterial stiffness index to improve the risk classification of incident cardiovascular disease.

2Analyses of the largest arterial stiffness index data set (n=169 613) to date indicated that it was an independent predictor of incident cardiovascular disease and all‐cause mortality.Arterial stiffness index improved the 5.9‐year risk prediction model of incident cardiovascular events by 2.3% when added to the Framingham Risk Score.Pulse pressure improved the 5.9‐year risk prediction model by 5.4% when added to the Framingham Risk Score.

3Analyses of the largest arterial stiffness index data set (n=169 613) to date indicated that it was an independent predictor of incident cardiovascular disease and all‐cause mortality.

4Arterial stiffness index improved the 5.9‐year risk prediction model of incident cardiovascular events by 2.3% when added to the Framingham Risk Score.

5Pulse pressure improved the 5.9‐year risk prediction model by 5.4% when added to the Framingham Risk Score.

6Pulse pressure appears to have more added value than arterial stiffness index to improve the risk classification of incident cardiovascular disease.

7Pulse pressure appears to have more added value than arterial stiffness index to improve the risk classification of incident cardiovascular disease.

8As the arterial system ages, the large elastic arteries undergo progressive luminal dilatation, thickening of the arterial wall, increased deposition of collagen, and combined fragmentation and degeneration of elastin.1 The result of these changes is stiffening of the arteries and consequent increase in pulse‐wave velocity (PWV), which is used to assess arterial stiffness. Increased arterial stiffness can cause isolated systolic hypertension, which increases pulse pressure (PP). Arterial stiffness and PP are independent measures of vascular aging.2 PP is strongly related to adverse outcomes such as coronary heart disease (CHD), and cardiovascular events in hypertensive patients,3 elderly,4 and the general population.5 Several studies have observed carotid‐femoral (aortic) PWV, which is considered the criterion standard of arterial stiffness, to be strongly related to risk factors such as atherosclerosis,6 hypertension,7 metabolic syndrome,7 diabetes mellitus,7 and future cardiovascular disease (CVD) events,8 including CHD, stroke,8 and all‐cause mortality.8 Arterial stiffness index (ASI) is a convenient and noninvasive method to measure arterial stiffness by using infrared light (photoplethysmography) to record the volume waveform of the blood in the finger. The shape of the waveform is directly related to the time it takes for the pulse wave to travel through the arterial tree. These tools might be of interest to quickly estimate CVD risk.9, 10

9In this study we investigated the association of vascular aging as indicated by ASI and PP with CVD risk factors, CVD events, and mortality in 169 613 participants from UK Biobank.

10The data from the UK Biobank resource are available for other researchers following an approved research proposal.11 The UK Biobank study design and population have been described in detail elsewhere.12 In brief, UK Biobank is a large community‐based prospective study in the United Kingdom that recruited >500 000 participants aged 40 to 69 years old with the aim of improving prevention, diagnosis, and treatment of a plethora of illnesses including cancer, diabetes mellitus, stroke, and heart diseases. A total of 190 077 participants had an ASI measurement during their first or 1 of the follow‐up visits. We analyzed data from 169 829 participants who had an ASI assessment at their first visit to the assessment centers in England and Wales. No ASI assessments were performed for participants from Scotland. All participants gave informed consent for the study via a touch‐screen interface that required agreement for all individual statements on the consent form as well as the participant's signature on an electronic pad.13 In this process all participants gave informed consent for data linkage as 1 statement requested consent for access to medical and other health‐related records, the long‐term storage and use of this and other information about the participants, also after incapacity or death, for health‐related research. The UK Biobank consent form is available at: http://www.ukbiobank.ac.uk/wp-content/uploads/2011/06/Consent_form.pdf. UK Biobank has approval from the institutional review boards, namely, the North West Multi‐centre Research Ethics Committee for the UK, from the National Information Governance Board for Health & Social Care for England and Wales, and from the Community Health Index Advisory Group for Scotland.14

11PWV for ASI assessment was measured during the first visit to the assessment center using the PulseTrace PCA2 (CareFusion, San Diego, CA) (Field‐ID 21021) in 169 829 participants from 2009 until 2010. The PulseTrace PCA2 uses finger photoplethysmography to obtain the pulse waveform during a 10‐ to 15‐s measurement using an infrared sensor clipped to the end of the index finger. The measurement was repeated on a larger finger or on the thumb if less than two thirds of the waveform was visible on the display of the PulseTrace PCA2 device, or if the waveform did not stabilize within 1 minute after clipping the infrared sensor to the end of the index finger.15 The shape of the waveform is directly related to the time it takes for the pulse wave to travel through the arterial tree.10, 15 The standing height (shoeless) was measured using a Seca 202 height measure and was manually entered into the assessment center software by a UK Biobank staff member. The software immediately alerted the staff member when he/she entered impossible or implausible values and was asked to correct it.13 Height (meters) was divided by the time between the peaks of the pulse waveform to obtain the ASI in m/s.15 This method has been validated by comparing it with carotid‐femoral PWV in 3 independent studies that concluded both measurements are highly correlated and that ASI is a simple, inexpensive, rapid technique that requires no training and is operator independent.9, 10, 16

12The prevalence and incidence of cardiovascular risk factors, conditions and events were captured through self‐reported data collected at the assessment center using a questionnaire and a verbal interview. Diagnoses were additionally captured using the “Spell and Episode” category from the Hospital Episode Statistics records. This category contains main and secondary diagnoses, coded according to the International Classification of Diseases Ninth Revision (ICD‐9) and 10th Revision (ICD‐10),17 made during hospital inpatient stay. The main diagnosis is taken to be the main reason for the hospital admission, while secondary diagnoses are more often contributory or underlying conditions. Furthermore, we used surgical procedures that were recorded according to the Office of Population, Censuses and Surveys: Classification of interventions and Procedures, version 4 coding.18 We used both the main and secondary diagnoses for recording prevalent and incident risk factors, conditions, and events. Incidence cases based on self‐reported diagnoses during follow‐up visits were included only if there were no events recorded according to ICD‐9/ICD‐10/Office of Population, Censuses and Surveys: Classification of interventions and Procedures and only if the participant did not report this in the previous visit. Date of event was then defined as reported age of diagnosis (when available) or the median date between the visit of the first self‐reported diagnosis and the previous visit. Follow‐up for new‐onset CVD, myocardial infarction (MI), CHD, heart failure (HF), stroke, and death because of these new‐onset cardiovascular conditions was from inclusion until March 31, 2015 for participants from England and until February 28, 2015 for participants from Wales. Please see Table S1 for the definitions used.

13Participant follow‐up for mortality started at inclusion in the UK Biobank study and was censored on January 31, 2016 for all participants from England and Wales. The information about cause of death was obtained from the National Health Service Information Centre. Detailed information about the linkage procedure is available online at http://biobank.ctsu.ox.ac.uk/crystal/refer.cgi?id=115559.

14Systolic and diastolic blood pressure (Field ID 4079, 4080) were measured twice at the assessment center using an automated blood pressure device (Omron 705 IT electronic blood pressure monitor; OMRON Healthcare Europe B.V. Kruisweg 577 2132 NA Hoofddorp), or manually (Field ID 93, 94) using a sphygmomanometer with an inflatable cuff in combination with a stethoscope if the blood pressure device failed to measure the blood pressure or if the largest inflatable cuff of the device did not fit around the participant's arm.19 All measurements were performed while the participant was sitting in a chair and were carried out by nurses trained in performing blood pressure measurements.19 During the first measurement nearly all participants (169 529) had blood pressure measurements using the automated blood pressure device. All had a second automated blood pressure measurement, except 153 individuals who did not have a second measurement. Manual sphygmomanometer blood pressure measurements during the first measurement were performed on 84 participants (0.05% of the total population). Of these, 11 participants had no second measurement and all others had a second manual blood pressure measurement. Multiple available measurements for 1 individual were averaged. PP was calculated by subtracting the (average) diastolic from the (average) systolic blood pressure value. Mean arterial pressure (MAP) was calculated by dividing the PP by 3 and adding this value to the diastolic blood pressure. The Omron 705 IT blood pressure monitor has satisfied the Association for the Advancement of Medical Instrumentation SP10 standard and has been validated according to the British Hypertension Society protocol, with an overall “A” grade for both systolic and diastolic blood pressure measurements.20 However, since automated devices tend to measure higher systolic blood pressures than manual sphygmomanometers, we adjusted both systolic and diastolic blood pressures that were measured using the automated device using algorithms by Stang et al.21 For systolic blood pressure we used the following algorithm: 3.3171+0.9201×level (systolic blood pressure in mm Hg)+6.0246×sex (male=1; female=0). For diastolic blood pressure we used: 14.5647+0.8092×level (diastolic blood pressure in mm Hg)+2.0108×sex (male=1; female=0). These adjusted blood pressure values were used for all calculations, including the PP and MAP calculations.

15Data are expressed as median (interquartile range) or as mean±SD for quantitative variables and as counts with percentages for discrete and categorical variables. Participants with ASI values below or above 4 SDs of the mean were excluded, as well as participants with incorrect inclusion dates and those younger than 40 years old, as this last group was considered less likely to undergo an event during follow‐up.

16To evaluate whether ASI and PP increased with age, we performed regression spline models with 95% confidence intervals per 5 years of age for females and males separately.

17We examined the effect of traditional cardiovascular risk factors on ASI and PP using unadjusted linear regression analyses and found all risk factors (sex, age, body mass index [BMI], MAP, diabetes mellitus, and smoking) were associated with both measures. We considered these risk factors in multivariable Cox regression models. We used up to 4 summative models. Model 1: Unadjusted; Model 2: Adjusted for age and sex; Model 3: Model 2+MAP, diabetes mellitus, smoking, and BMI; Model 4: Model 3+history of CVD, MI, CHD, HF, and stroke. To examine the predictive value of ASI and PP for new‐onset CVD, MI, CHD, HF, and stroke, we performed Cox regression analyses using Model 1 to 3. Participants with a history of CVD, MI, CHD, HF, or stroke, were excluded from the respective analyses. To examine the relationship between ASI and PP with all‐cause, CVD, and non‐CVD mortality, we applied Model 1 to 4. For the Cox regression analyses, we reported hazard ratios and 95% confidence intervals.

18Kaplan–Meier failure curves were plotted for outcomes associated with ASI or PP. Because older individuals tend to have more CVD events and die sooner, the use of single ASI and PP distributions for the entire study sample would result in a greater proportion of older participants to undergo events, in comparison to the smaller proportion of younger participants with fewer events. To improve the balance of these proportions, we stratified for deciles of age at ASI and PP measurement. Age decile cut points for men were at 44, 49, 52, 56, 59, 61, 63, 65, and 67 years. For women the cut points were at 44, 48, 52, 55, 58, 60, 62, 64, and 67 years. We then determined ASI and PP distributions for each decile, separately for ASI and PP and by sex. Participants with ASI and PP measurements below and above the median of their respective sex‐ and age‐specific deciles were pooled together and compared. Log‐rank testing was performed to estimate the statistical difference between the medians.

19Harrell's C‐indices, a generalization of the area under the receiver operating characteristics curve for data from Cox regression analyses, were computed for Model 3 for disease and Model 4 for mortality outcomes, both with and without additional adjustment for ASI or PP. Postestimation analysis was used to determine whether ASI or PP independently predicted outcomes compared with the traditional risk factors. We further investigated the possible clinical impact of ASI and PP measurements in UK Biobank by performing reclassification analyses using the net reclassification improvement (NRI) and the integrated discrimination improvement.22 For the reclassification analyses, we used a non–laboratory‐based Framingham Risk Score (FRS) in which BMI is used instead of cholesterol.23 Because the FRS calculates 10‐year risk and we had a maximum of 5.92 years of follow‐up, we divided the FRS risk estimates by 1.69 to represent the 5.92‐year risk. Individuals were classified into <5%, 5% to 15%, and >15% risk categories. P<0.05 was considered statistically significant. All analyses were performed using Stata version 14 (StataCorp. 2015; Stata Statistical Software: Release 14. College Station, TX: StataCorp LP).

20We studied 169 613 individuals (45.8% males; average age 56.8 years old) participating in UK Biobank and of whom PP and ASI had been measured. The sample selection strategy is presented in Figure S1 and baseline characteristics are presented in Table 1. In total, 333 042 UK Biobank participants were excluded from the analyses. Baseline characteristics of the excluded participants are shown in Table S2. Ethnic backgrounds of included and excluded participants (both >90% white) are provided in Table S2. The most common risk factor was past or current smoking, followed by hypertension. CHD was diagnosed most often (n=4326), followed by MI (n=1587), stroke (n=1319), and HF (n=1192). The overall mean ASI was 9.30±3.1 m/s and mean PP was 50.98±13.2 mm Hg. ASI and PP were weakly correlated with each other (R 2=0.01; P<0.001). Both increased with advancing age (R 2=0.04 for ASI; R 2=0.17 for PP; both P<0.001, Figure S2). Multiple cardiovascular risk factors (increased BMI, hypertension, MAP, diabetes mellitus, and smoking) were associated with ASI and PP (Table 1). Heart rate was positively associated with ASI and negatively associated with PP (Table 1).

21The median follow‐up duration for new‐onset disease events was 2.8 years (interquartile range 1.4–4.0). Kaplan–Meier failure curves for CVD divided by the median ASI and PP are shown in Figure. Curves for MI, CHD, and HF divided by the median ASI are shown in Figure S3 and curves for MI, CHD, and HF divided by the median PP in Figure S4. Log‐Rank testing showed a difference between below and above the median ASI and PP for most diseases. HF showed no difference between below and above the median ASI. In total, during a maximum of 5.9 years of follow‐up, CVD was diagnosed in 18 190 individuals. Cox regression analyses showed that both increased ASI and PP were associated with increased risk for CVD, MI, CHD, HF, and stroke (Table 2). Adjustment for age and sex did not alter the associations (Table 2). Additional adjustment for CVD risk factors (MAP, diabetes mellitus, smoking, and BMI) resulted in loss of the associations between ASI and PP with stroke and lead to small effect sizes for most outcomes (Table 2). Additional adjustment for PP did not affect the associations between ASI with the disease outcomes, nor did additional adjustment for ASI affect the association between PP with disease outcomes (Table S3). We also plotted the risk of quintiles of ASI and PP for overall CVD adjusted for the third model and observed semilinear increases in risk (Figure S5).

22Addition of ASI to traditional risk factors increased C‐indices for MI and CHD, independent of additional adjustment for PP (Table S4). Addition of PP to traditional risk factors increased C‐indices for CVD and CHD independently of adjustment for ASI, but it did not increase C‐indices for MI or stroke. The NRI showed improvement (2.3%; 95% confidence interval, 2.0–2.6; P<0.001) in reclassification when ASI was added to the FRS (Table 3). Also the integrated discrimination improvement was improved with an estimate of 0.0008 (P<0.001). When PP was added to the FRS, the NRI showed an improvement of 5.4% (95% confidence interval, 4.9–5.8; P<0.001; Table 4) and the integrated discrimination improvement was estimated at 0.006 (P<0.001). We calculated the cutoff values using Youden's index (indicating the value at which the measure has highest sensitivity and specificity) of ASI and PP for overall CVD at 9.21 m/s and 51.23 mm Hg, respectively.

23Median follow‐up for mortality was 6.1 years (interquartile range 5.8–6.3). In total, 3678 participants died, of which 1180 participants of CVD. Kaplan–Meier failure curves for all‐cause, CVD, and non‐CVD mortality divided by the age‐ and sex‐specific median ASI are shown in Figure S6 and a curve for CVD mortality divided by the median PP in Figure S7. Log‐rank testing showed a difference for all mortality outcomes, except CVD mortality, which showed no difference between the ASI values. In univariate Cox regressions we observed that higher ASI and PP were associated with increased risk for all mortality outcomes (Table 5). The association between ASI with increased risk of all mortality outcomes persisted when adjusted for traditional risk factors and history of CVD, MI, CHD, HF, or stroke (Table 5). To investigate whether the association between ASI and non‐CVD mortality was driven by cancer mortality or other causes, we divided non‐CVD mortality into cancer mortality and death by other causes (non‐CVD/noncancer). ASI was associated with both cancer mortality and non‐CVD/noncancer mortality, also after adjustment for all factors mentioned above (Table S5). The associations of PP with all‐cause and non‐CVD mortality were less strong (Table 5).

24To determine whether ASI is a predictor for all‐cause, CVD, and non‐CVD mortality independently of PP, we adjusted the largest significant ASI Cox regression models for PP, which did not affect any of the associations (Table S6). Similarly, adjusting the PP Cox regression models for ASI did not affect the PP associations either. Postestimation C‐indices for all‐cause mortality were improved after ASI was added to the traditional risk factors and history of disease, independent of adjustment for PP (Table S4). C‐indices for CVD mortality did not improve when PP was added to the traditional risk factors and history of disease (Table S4).

25In the literature there is ongoing debate whether PP24 and especially aortic PWV25, 26 measurements should be adjusted for heart rate. For this reason we performed sensitivity analyses adjusting the largest significant PP and ASI models for heart rate, but this did not affect our conclusions (Table S7).

26We studied 2 independent measures of vascular aging,2 ASI and PP, in relation to CVD events and mortality in 169 613 participants of UK Biobank. In the largest sample size available to date, we demonstrated that ASI was an independent predictor of new‐onset CVD outcomes and mortality. Furthermore, we have shown that PP was an independent predictor of CVD, MI, CHD, HF, and CVD mortality. PP improved C‐indices of CVD and CHD, but not of MI, stroke, and CVD mortality, compared with traditional risk factors. PP added to the nonlaboratory FRS improved the NRI twice as much compared with ASI.

27We found that both increased ASI and PP were associated with an increased risk of developing CVD. The predictive value of ASI and PP can be considered a reflection marking the (patho)physiological function of the cardiovascular system. Stiffer arteries allow the pressure wave in the arterial tree to travel faster, causing systolic rather than diastolic augmentation of the reflected pressure wave. Systolic augmentation increases PP and left ventricular load, which in turn can lead to cardiac hypertrophy.27 In addition, ischemia of the left ventricle could also occur through decreased aortic pressure during diastole causing reduced coronary filling and thereby reduced myocardial perfusion.27, 28

28In our study we found a positive association between overall CVD events and ASI whereby the risk increased with 6% per SD change. This association is in line with previous, smaller observations.8, 29 Both reported a higher risk (6.15% versus 1.6% risk per year) of developing CVD per SD increase in aortic PWV. This difference in effect size might be attributable to differences in baseline characteristics. For example, participants of the study by Mitchell et al29 were recruited from 1998 to 2001 and were on average 6.2 years older. The participants of the studies in the meta‐analysis of Vlachopoulos et al8 were recruited from 1980 to 2005 and were on average 2.4 years older than our participants.

29The association between PP and CVD incidence was previously reported by Blacher et al30 in older (average 67–72 years) hypertensive patients who had a 17% increased risk of CVD per 10 mm Hg higher PP. The risk of developing CVD in this cohort of hypertensive subjects is higher than the risk observed in our presumably healthier community‐based population, which was 3.8% when estimated per 10 mm Hg higher PP (1 SD in our population equals 13.2 mm Hg difference in PP). As noted before, ASI and PP are independent measures of vascular aging; also in our study the correlation between ASI and PP was weak. ASI and PP likely do not reflect the same properties or characteristics of the vasculature. However, improvement in NRI was observed when ASI and PP were added to the non–laboratory‐based FRS, indicating both ASI and PP may aid the risk prediction of overall CVD. The improvement of the NRI by addition of ASI was low (2.3%), whereas the improvement of the NRI by addition of PP was much larger (5.4%), indicating a superior possible clinical applicability of PP compared with ASI. The NRI possibly better reflects the added clinical value of ASI and PP to the risk prediction than the C‐indices, which for each end point were similar until the third decimal. We also performed sensitivity analyses in subgroups and found PP was modestly better than ASI at predicting mainly overall CVD in, for example, women, participants with diabetes mellitus, and both young and old participants (Tables S8 through S12). ASI was better at predicting overall CVD in men and all‐cause mortality in women, nondiabetics, and young participants (≤56.8 years). ASI and PP were not different in subgroups of hypertension and FRS (Tables S9 and S11). However, also in these additional analyses the C‐indices were very similar. The significant differences in C‐indices between different models of the main‐ and subgroup analyses may be attributable to the large sample size and does not represent a difference that is clinically relevant.

30Besides finding an association between ASI and PP with overall CVD, we observed associations between ASI and PP with several specific disease outcomes. Both ASI and PP were associated with MI and CHD in our community‐based population. In aortic PWV studies, MI was often included in the definition of outcome variables such as overall CVD events or CHD, but not investigated as an individual outcome as has been done in this study. These previous studies show an association between aortic PWV and CHD,8, 31, 32, 33 as is shown here as well. Our results are consistent with previous PP studies that reported associations with both incident MI3 and CHD34 after adjustment for traditional CVD risk factors in a hypertensive and community‐based cohort, respectively.

31The association between aortic PWV with HF is inconsistent across studies.32, 35 We found an association between ASI and HF after adjustment for traditional CVD risk factor but it had little discriminating power. We found a strong association between PP and HF, similar to 2 previous studies that found PP independently predicted chronic HF in the Framingham Heart Study population36 and an elderly population.37

32Finally, neither ASI nor PP was associated with stroke in our population in survival models after adjustment for MAP. The association between arterial stiffness measured by carotid‐femoral PWV index and stroke, which was also adjusted for MAP, has, however, been described previously.31 Also the associations of PP with stroke are inconsistent with earlier work in older people with isolated systolic hypertension, where they found an 11% increased risk per 10 mm Hg higher PP, but also showed that MAP increased risk of stroke.38 However, a later study in uncontrolled hypertensive subjects found, similar to this study, that the relationship between PP and stroke was dependent on MAP,39 indicating that MAP may be a confounder in the association between PP and stroke.

33In the present work we found that ASI was associated with all‐cause, CVD, and non‐CVD mortality, although no statistical difference between CVD mortality incidences was found between individuals with high and low ASI values. The association between ASI and non‐CVD mortality appears to be driven by both cancer mortality and mortality by other causes. To our knowledge, increased ASI has not previously been found to predict cancer mortality. The origin of this association should be subject to future research. Unlike ASI, and in contrast to previous reports of studies in 9431 (65–102 years old)40 and 2725 (20–80 years old)41 individuals from the general population, PP was not convincingly associated with all‐cause or non‐CVD mortality.

34The addition of heart rate to the ASI models had little effect, suggesting that heart rate had no confounding effect on ASI, unlike previously suggested for aortic PWV.25, 26 The addition of heart rate to the PP models had similar little effect, arguing against a need for heart rate adjustment for PP.24 It is interesting that heart rate was inversely associated with PP, whereas it was positively associated with ASI. However, it should be noted that the effect of heart rate is small for both measures, also when compared with the effects of the other characteristics in Table 1.

35The associations between ASI and PP with disease and mortality outcomes have been studied previously but our contemporary study is unique. Not only is the very large sample size unprecedented, but also we provide a community‐based population with both ASI and PP measurements in combination with detailed health‐ and mortality‐related data. Although the effect sizes were small to moderate for most outcomes in the unadjusted Cox regression analyses and even smaller in the adjusted analyses, the large sample size allowed us to detect these at the statistically significant level. One important limitation is the limited duration of disease follow‐up. A second limitation is that, although we could adjust for a number of important classical risk factors, we did not have data available on serum lipid levels to take into account in our multivariable models. A third limitation is that the accuracy of Hospital Episode Statistics data used for our analyses is not known for most data fields. Furthermore, the methodological differences of arterial stiffness measurement (aortic PWV versus ASI) are likely to play a role in the discrepancies found between our and previous studies. ASI derived by finger photoplethysmography is influenced by the elasticity of the large central arteries and the properties of the reflection sites, both central and peripheral.16, 25 In addition, the PWV measured in the peripheral arteries are usually higher than the aortic PWV because of the nearer reflection sites.25 These differences make results from our and previous studies (aortic PWV) more difficult to directly compare. Finally, brachial PP was used as a measure for systemic arterial stiffness instead of central PP, which is measured at the carotid artery. Brachial PP is generally higher than central PP because of the larger number of reflection sites in the peripheral arteries compared with the central arteries. The central arteries of younger individuals are more elastic than the peripheral arteries, which can further increase difference between brachial PP and central PP, and might have resulted in an overestimation of the stiffness of their arterial tree.7

36Higher ASI is associated with cardiovascular risk factors and is an independent predictor of new‐onset CVD outcomes as well as all‐cause, CVD, and non‐CVD mortality. Although finger photoplethysmography is a simple and fast method, ASI measurement added relatively little to the risk prediction in this community‐based population, limiting its potential clinical value. Similar to ASI, PP was an independent predictor of new‐onset CVD outcomes and CVD mortality. PP improved the CVD risk prediction classification by 5.4%, suggesting that PP could be used as a clinical tool to improve risk prediction for disease outcomes in patients. Also similar to ASI, PP is obtained with minimal efforts, further favoring the use of PP over ASI in clinical settings. ASI may, however, remain an interesting measure for studying vascular aging and stiffness.42

37Said was supported by the Royal Netherlands Academy of Arts and Sciences’ Van Walree grant. Verweij is supported by Marie Sklodowska‐Curie GF (call: H2020‐MSCA‐IF‐2014, Project ID: 661395) and an NWO VENI grant (016.186.125).

38None.