Evaluation of the Age Dependence of Conventional and Novel Photoplethysmography Parameters

Abstract

Background

Cardiovascular (CV) mortality increases with age partly due to physiological ageing of the CV system. Early vascular ageing raises CV risks. Personalizing CV risk assessment by defining CV age could reduce CV events. Photoplethysmography (PPG), which analyses the peripheral arterial pulse wave, may be an effective method for estimating CV age. Ageing index (AGEi) and some other PPG parameters were proven to have age correlation; however, the age dependence of many other pulse wave parameters remains unclear. We aimed to identify age correlations of PPG indices and pulse rate variability (PRV) parameters including a few novel parameters which were calculated to further investigate the various aspects of ageing.

Our study included 118 healthy (M/F: 53/65, mean age: 31.8 ± 11.8 SD) volunteers for PPG parameter calculation and 106 (M/F: 44/62, mean age: 32.6 ± 12.2 SD) for PRV parameters (age: 19–74). 2-min pulse wave recording was obtained using a pulse oximeter. An automated, proprietary software evaluated PPG and PRV parameter values, which were compared with chronological age (Pearson correlation and non-linear analysis).

Results

PPG parameters describing various time-dependent aspects of cardiac ejection positively correlated with age, while those indicating arterial elasticity showed negative correlation. Composite PPG parameters proposed as indicators of CV health and fitness had negative, non-linear correlation. Most PRV parameters exhibited negative correlation, indicating reduced adaptive capacity due to ageing (p < 0.05, IrI > 0.3).

Conclusions

PPG-based pulse waveform analysis provides a wide range of age-related parameters which display different patterns of age correlation, making it a promising method for estimating cardiovascular age. Future studies will include subjects with vascular ageing conditions beyond physiological values (e.g., hypertension, heart failure, coronary artery disease).

1 Introduction

Cardiovascular (CV) diseases, including atherosclerosis and stroke are major public health challenges, consistently ranking among the leading causes of death worldwide in recent decades, especially in the elderly population [1, 2]. Age-related phenotypic alterations in the CV system, and more importantly their accelerated development brought about by CV risk factors, are among the most relevant (patho)physiological changes that drive these diseases [3, 4]. Therefore, identifying new, affordable biomarkers that reflect (CV) aging is critical for improving treatments and preventive strategies.

Peripheral pulse wave analysis may offer a valuable method for monitoring CV health and predicting disease progression [5, 6]. Calculating heart rate from continuous pulse wave recordings may have relevance in diagnostics, as pulse rate variability (PRV) is an important indicator of various diseases [7,8,9]. Beyond PRV, the morphological characteristics of pulse waves have yielded considerable attention, with numerous studies suggesting that these parameters may be associated with CV disease states such as atherosclerosis and heart failure [5, 10, 11].

Photoplethysmography (PPG) is a simple, easily accessible, and highly repeatable method for real-time monitoring of pulse waves [12]. This non-invasive technique involves illuminating the skin and tissues below, typically the finger, with an LED and measuring the intensity of the reflected or transmitted light, which corresponds to pressure changes in the vascular system. Importantly, PPG has no known adverse effects [13].

The promising results from previous studies suggest that PPG-based pulse wave analysis could gain traction in CV diagnostics and home monitoring in the near future [14]. While it holds potential as a tool for assessing CV aging, its broader use is constrained by the limited investigation of age-related correlations in most PPG-derived parameters. Although some parameters have been linked to age-related changes, most studies have focused on the age dependence of individual or a few selected parameters, leaving the majority unexplored [6, 15,16,17].

However, a combination of parameters or composite measures derived from multiple parameters might better capture age-related changes than single parameters alone. PPG-based monitoring devices, equipped with advanced algorithms, enable the simultaneous assessment and complex analysis of numerous parameters [5, 18]. Consequently, research aimed at identifying a set of simultaneously recorded PPG features with the strongest correlation to CV age could significantly enhance the potential of PPG-based pulse wave analysis. Additionally, most published studies have assumed linear age dependence of parameters [15,16,17], which may not accurately reflect reality. Many parameters could exhibit non-linear relationships with age, particularly in women, where CV changes accelerate after menopause.

The primary goal of our research was to identify age-dependent changes in a large set of simultaneously recorded pulse wave parameters, including PRV parameters, pulse morphology parameters and newly developed composite score parameters, aiming to establish the utility of PPG-based pulse wave analysis as a tool to assess CV aging. For this purpose, we utilized an efficient, automated software that enables accurate, rapid, and reproducible evaluation of large datasets; and a comprehensive database of pulse wave data from a healthy adult population was established. To better characterize age-dependent parameter changes, we used both linear and non-linear analyses to describe age-related trends.

2 Methods

Participants were required to meet specific inclusion criteria, including self-reported good physical and mental health, absence of CV disease, no use of CV medications, non-pregnancy, a BMI between 18 and 26 kg/m², non-smoker status, negligible alcohol consumption, and no reported history of chronic or cancerous diseases.

Subjects were primarily recruited from among the healthy employees, relatives of employees, and students of Semmelweis University. Recruitment was facilitated by the University's Occupational Health Service and social networking platforms. All tests were conducted in the laboratory facilities of Semmelweis University. The study protocol was designed in accordance with the Declaration of Helsinki and approved by the Semmelweis University Regional and Institutional Committee of Science and Research Ethics (approval number: 120/2018).

Participants provided informed consent and completed a health questionnaire, which collected personal and health-related data, including medical history, lifestyle, and medication use. Blood pressure (BP) was measured three times using an automatic sphygmomanometer. Subjects with systolic BP higher than 140, and/or diastolic BP exceeding 90 mmHg were excluded from the study. All data was recorded anonymously.

Pulse wave recordings were obtained using a Berry BM 1000B pulse oximeter placed on the right index finger. This non-invasive device, certified by the manufacturer, recorded pulse waves for 140 s while the participant remained seated and still. The pulse oximeter transmitted data via Bluetooth to a mobile application (SCN4ALL/HeartReader), developed by E-Med4All Europe Ltd. (Budapest, Hungary), which uploaded the recordings to a secure online database. The studies for the repeatability and reliability of the measurements, along with the detailed description of signal processing methods of the system have already been published [19, 20]. Briefly, the measurement takes 140 s to be completed. Due to filtering and preprocessing reasons discussed in detail by Kulin et al. [19], 120 s of the recording is used for further analysis. Parameters were defined for each individual cycle that met certain predefined signal quality criteria, and the average of these values was reported.

The proprietary software used for analysis identified fiducial points on the pulse wave, allowing for the calculation of both classical and novel pulse wave parameters (PPG parameters), including pulse rate variability (PRV parameters) metrics. The primary criterion for selecting parameters was to choose those that, according to the literature, describe various aspects of CV function—such as temporal relationships, arterial elasticity, and autonomic function—and have previously been reported to correlate with CV age, mortality, and (severity) of various CV diseases. Table 1. shows the parameters and their descriptions.

Table 1 PPG and PRV parameter names and descriptions

The parameter values obtained from the pulse waveform analysis were compared with the age (in years) of the volunteers (JASP 0.19.1 software, JASP Team (2024)) using Pearson correlation and Generalised Additive Models (GAM) analysis (Google Colaboratory. Retrieved December 14, 2024, from https://colab.research.google.com/). GAM is an advanced statistical modelling method designed to capture both linear and non-linear relationships between variables. (see the ‘Additional file1.docx’ for a more detailed description of the model). A p value of <  = 0.05 was accepted as significant throughout.

During the preparation of this work the author(s) used ChatGPT and Grammarly to improve the readability and find shorter expressions to fit word limit. After using these tools/services, the authors reviewed and edited the content as needed and took full responsibility for the content of the publication.

3 Results

Our study included 118 healthy (M/F: 53/65, mean age: 31.8 ± 11.8 SD) volunteers for PPG parameter calculation and 106 (M/F: 44/62, mean age: 32.6 ± 12.2 SD) for PRV parameters. Participants were aged between 19 and 74 years.

The relationship between age and CV function may encompass both linear and non-linear factors. To comprehensively evaluate this, we performed two distinct analyses: a Pearson correlation to assess linear associations and a GAM analysis to capture potential non-linear trends.

Tables 2. and 3. summarize the results of Pearson correlation and GAM analysis between PPG and PRV parameters and age.

Table 2 Results of correlation analysis of PPG parameters and age

3.1 Pearson Correlation Analysis

Among the conventional PPG morphology parameters that significantly correlated with age, AGEi (r = 0.485), SysAlpha (r = − 0.418), and d/a (r = − 0.376) (Fig. 1A) demonstrated the strongest age dependence. Additionally, time-related parameters of the PPG curve that characterize ejection-related ventricular activity, such as ET(PPG) (r = 0.589), Crest Time (r = 0.570), LVETi (r = 0.539), and the proprietary parameters eLVET1* (r = 0.548) and eLVET2* (r = 0.450), also exhibited strong correlations with age (Table 2., Fig. 2.).

Fig. 1

Scatter plots of correlation results between age and PPG parameters (AGEi and DNi*). A The scatter plot of the correlation analysis between age and AGEi, and B the scatter plot of the correlation analysis between age and DNi* parameters

Fig. 2

Scatter plots of correlation results between age and the ejection-related PPG parameters. A The scatter plots of the correlation analysis between LVETi, B eLVET1, C Crest Time and D ET(PPG) and age

Furthermore, age correlation was observed in other novel parameters, including DNi* (r = -0.517) and c-d incidence* (r = 0.419) (Fig. 1B). Finally, all proprietary score parameters demonstrated significant correlations with age: Heart Fitness Score (r = − 0.493), CV Health Score (r =−  0.450), and Total Score (r = − 0.301) (p < 0.001 for all cases).

Several of the PRV parameters exhibited a moderate, but significant negative correlation with age (Table 3. and Fig. 3.). The cTotalPower (r = − 0.325) (Fig. 3A) and cSDRR (r = − 0.401) parameters (Fig. 3B) exhibited the strongest age dependence (p < 0.001) among frequency-domain and time-domain measures, respectively. The age correlation of non-linear PRV parameters proved to be weaker, except for cSD2 (r = − 0.428).

Table 3 Results of correlation analysis of PRV parameters and age

Fig. 3

Scatter plots of correlation results between age and PRV parameters A The scatter plot of the correlation analysis between age and cTotalPower, and B the scatter plot of the correlation analysis between age and cSDRR parameters

3.2 GAM Analysis

The GAM analysis allowed the identification of non-linear trends. Similar to the Pearson correlation analysis, this analysis also found significant correlations (p < 0.05) between age and PPG parameters, except for Si and b/a parameters. Among the PRV parameters, cSDRR, cTotalPower, cHFpow and cSD2 were significantly correlated with age based on GAM analysis. The GAM analysis confirmed linear (for AGEi, LVETi, eLVET2*, DNi*, SysAlpha) or near-linear (for b/a, d/a, Ri, Si) relationship between most PPG parameters and age. However, for some parameters, a non-linear trend with age was observed.

All Score parameters demonstrated a clear non-linear decline, especially after the age of 40. (Fig. 4A and B). The eLVET1* and c-d incidence* parameters showed a moderate non-linear upward trend followed by a plateau.

Fig. 4

Plots of GAM analysis. A, B The plot of the GAM analysis between age and Heart Fitness and Total Score parameters, C the plot of the GAM analysis between age and Crest Time PPG parameter and D the plot of the GAM analysis between age and cpNN50 PRV parameter

Crest Time exhibited extreme non-linearity with multiple inflection points (Fig. 4C).

For PRV parameters, cTotalPower and cSDRR showed a clear linear decrease with age. The cMHR did not have a significant correlation with age (nor did it when Pearson correlation analysis was performed) (Fig. 4D).

4 Discussion

From a public health perspective, addressing the assessment and monitoring of CV ageing is crucial, as CV diseases continue to be the leading cause of mortality, particularly in older populations. As the global population ages, the demand for reliable, non-invasive methods to meet this need is increasing. Photoplethysmography (PPG) appears to be a promising tool in this regard, as it offers a simple but effective way to monitor CV function by pulse wave analysis. Although the age correlation of some PPG parameters has been investigated, the full scope of age-related changes in pulse wave characteristics is not yet fully evaluated. [21,22,23] This study, by examining both conventional and novel PPG parameters, as well as PRV characteristics, provides a more comprehensive understanding of the effects of chronological ageing on the pulse waveform morphology. The importance of our research is emphasized by the fact that age is arguably the most significant risk factor for CV morbidity and mortality. This is supported by the results of Pencina et al. who found that age, sex, and race capture 63% to 80% of the prognostic performance of CV risk models [24]. This is further emphasized in the Framingham risk score, where age contributes more to the total risk score than any other variable. [25] Our study identified a diverse set of simultaneously recorded PPG parameters including ones that are related to cardiac ejection time, arterial elasticity and loss of PRV. These findings highlight the correlation of PPG parameters with chronological age, suggesting their potential use for monitoring age-related CV changes and evaluating CV health across different age groups. In addition, a major strength of this study lies in the use of a proprietary, automated software system capable of analyzing large datasets with high efficiency that enhance reliability, ensures the reproducibility of study's results [19].

Among the 16 PPG morphology parameters, ET, including its subcomponent eLVET1, as described by our research group, and LVETi, as described by Weber et al., demonstrated the strongest correlations with age, indicating a gradual decline in CV efficiency as individuals age [21, 26, 27]. Using GAM analysis, an extreme non-linear relationship with multiple inflection points was observed between crest time and age. This is probably due to sparse sampling in older age groups. This highlights the sensitivity of nonlinear models to small sample sizes and outliers.

Arterial stiffening due to loss of arterial elasticity and structural changes in the vascular wall, such as increased collagen deposition and reduced elastin, is a hallmark of CV aging and contributes to elevated CV risk [28]. Therefore, reliable characterization of arterial distensibility by easily accessible biomarkers is an important step toward early detection and prevention of CV diseases, as well as the assessment of vascular aging [5, 29].

Our results also confirmed the findings of previous studies describing age-dependent changes in the AGEI. AGEi is a parameter derived from the second derivative of the pulse contour wave, and its correlation with age and arterial stiffness is widely recognized (as noted by Takazawa and colleagues) [15].

While pulse wave velocity (PWV) is often considered a better measure for assessing CV aging because of its broader predictive power at the population level, AGEi shows considerable potential as a complementary tool, particularly in individual risk assessment. The strong correlation between the second-derivative PPG signal parameters, particularly AGEi and PWV, has been published in several publications [16, 30]. These results highlight the potential of AGEi as a practical, non-invasive measure of individual risk stratification, especially when measurement of PWV is less accessible. The sensitivity of AGEi to age-related vascular changes is a valuable addition to CV diagnostics, complementing PWV’s population-level insights. Our study has also shown that DNi has stronger age dependence than AGEi suggesting that it may have relevant potential in monitoring a progressive decline in arterial distensibility (DNi, a proposed marker of aortic distensibility and coronary flow pressure gradient). Si is another PPG parameter proposed by several authors to characterize arterial stiffening. Based on the previous publication of Millasseau (Determination of age-related increases in large artery stiffness by digital pulse contour analysis), PWV and Si are significantly correlated with each other, and both are correlated with age. Interestingly, the Si showed a weak correlation with age in our study [31]. One possible explanation for this may be the different age and sex distribution of the two studies. In the study of Millaseau et al., 29 of the 87 participants were women; the mean age was 47 years, with a range of 21–68 years. Whereas our study age distribution for females found to contain a higher proportion of women mostly in premenopausal age. These observations emphasize that precise characterization of age correlation may require accounting for sex-specific differences and other confounding factors in the analysis; however, this necessitates analysis performed on large datasets.

In addition to the individual parameters, "composite scores" of multiple PPG parameters, such as the Total Score, Heart Fitness Score and CV Health Score, also showed significant correlations with age, both using Pearson correlation and GAM analysis. This supports the unpublished observations of the manufacturer that suggested strong age dependence of these parameters in a large inhomogeneous patient population coming from real-world data of more than 98 000 processed measurements from more than 5 800 individuals in various age, sex and health status [32]. The composite scores were developed to simplify the interpretation of CV health indicators by aggregating multiple PPG-derived parameters into a single, more user-friendly metric. This approach can make it easier for end-users to track and understand their metrics, especially for non-specialists for whom interpretation of multiple individual parameters (e.g. 15–20) can be challenging. While the exact calculation methods for these scores are proprietary, they are based on established PPG signal features associated with vascular and cardiac health. These include parameters related to arterial stiffness, pulse wave characteristics, and temporal signal dynamics, all of which are linked to age-dependent CV changes. The validation of these composite scores as independent predictors of CV health. requires further studies. However, preliminary findings suggest that they could support CV risk evaluations. All score parameters in this study showed a clear non-linear, decreasing relationship with age, especially after age 40. This sharp decline is consistent with published data showing accelerated ageing during middle age [33].

Some PRV parameters, such as total power (cTotalPower) and SDNN (cSDRR), showed a significant correlation with age. Both parameters showed a decrease with increasing age; this could indicate a less sensitive autonomic nervous system, which may contribute to the reduced cardiovascular adaptive capacity observed in the elderly. This finding is consistent with the existing literature, which suggests that decreased heart rate variability reflects reduced autonomic control of the CV system, and highlights the importance of monitoring autonomic function through PRV parameters as part of a comprehensive CV health assessment. [7,8,9, 34].

In summary, our results reveal a set of PPG and PRV parameters associated with age-related changes with distinct differences between parameters in the aspect of linearity, emphasizing the potential of simultaneous recording and analysis of multiple PPG parameters in CV prevention, though further research is required. Additionally, combining different PPG parameters has yielded composite scores with unique age-dependent patterns which might reflect the non-linear trends of ageing, which may prove useful in identifying age-related CV events or conditions. We believe that our study may serve as a foundational step in developing personalized PPG-based CV age assessment tools. However, future research should explore whether individuals positioned above or below the correlation trend line represent distinct CV aging phenotypes, such as early vascular aging or supernormal vascular aging. [35, 36].

5 Limitations of the study

A limitation of our study is that the age distribution of the sample population is not fully uniform and may not be fully representative of the general population. Future research should, therefore, be extended to a wider, more diverse cohort to further verify these results.

Clinical validation of the proprietary PPG parameters introduced could be a critical next step towards their wider use and clinical utility. Although the aim of this study was primarily to explore the age dependence of these parameters, it is important to outline possible avenues for future validation. Future studies are planned to focus on the correlation of the new PPG composite scores with widely accepted CV risk scores such as the Framingham Risk Score or the HeartScore (European Society of Cardiology), as well as with established measures such as lipid profiles, hs-CRP, plasma creatinine, carotid Doppler and echocardiography results, and pulse wave velocity (PWV). Further validation efforts include analysing how composite scores interact with clinical and lifestyle factors, including patient history and modifiable risk behaviours, to increase their predictive accuracy. In addition, to ensure wider applicability, we plan to evaluate the performance of these scores in different patient subgroups, including individuals with different CV risk profiles and comorbidities. These studies may be beneficial to further refine the interpretation of the identified age-related indicators, as different PPG parameters may be more relevant in certain pathological contexts, such as hypertension or heart failure.

6 Conclusion

This study has successfully identified age-related linear and non-linear correlations across both conventional and novel PPG parameters, highlighting their potential as valuable indicators of CV ageing. The findings demonstrate that parameters related to cardiac ejection time, arterial elasticity, and PRV, among others, consistently correlate with age, offering a comprehensive view of how the CV system evolves over time. The introduction of novel composite PPG score parameters, which showed notable age correlations, may complement traditional metrics, although further validation is needed to confirm their specific contributions. The clinical relevance of these findings is that they draw attention to the potential of pulse wave analysis to monitor CV ageing non-invasively and position PPG as a promising tool in both clinical and preventive cardiology. However, translation of this method to clinical settings requires further research in patients with various CV conditions and comorbidities.