A novel non-invasive estimate of biological age: can an echocardiogram measure the patient’s age? Free

This editorial refers to ‘Five-year changes in weight and risk of atrial fibrillation in the Danish Diet, Cancer, and Health Cohort’, by T.C. Frederiksen et al., https://doi.org/10.1093/eurjpc/zwad300.

The impact of aging balances the cumulative burden of chronic stress and life events, the allostatic load, with the capacity for physiologic adaptation to these stressors. This load results in many structural and functional changes, particularly involving the heart and the vascular system. The associated oxidative stress, chronic low-grade inflammation, and decreased cellular regenerative capacity result in lower myocyte number, myocyte hypertrophy, and increased collagen deposition in the extracellular matrix of the heart.¹ The resulting myocardial fibrosis and stiffness can involve the atria, leading to left atrial myopathy, atrial enlargement, and atrial fibrillation, or the ventricles leading to concentric remodelling, hypertrophy, and diastolic dysfunction.² Concurrent changes in the vascular system include endothelial dysfunction and arterial intimal–medial thickening and can result in wall stiffness and arterial dilation.¹ To date, age is considered the strongest determinant of not only aging but also cardiovascular health. For example, age was shown to be stronger than magnetic resonance imaging measures in determining the risk of stroke.³

Despite the well-established association between aging and cardiovascular disease, it is widely accepted that individuals age at different rates. This inter-individual variability in aging results from a complex interaction between genetic predisposition, lifestyle, the environment, accompanying diseases, chronic load, and adaptation. As a result, different measures have been developed to better reflect health status while accounting for these variations, more than chronological age, leading to the concept of ‘biological age’. Frailty index is one commonly used measure that incorporates symptoms, signs, functional impairments, and laboratory abnormalities and can be used to reflect biological age in elderly patients, being more strongly associated with mortality than ‘chronological’ age in different settings.⁴ Vascular age is another measure based on cardiovascular risk factors, treatments, or pulse wave velocity as a marker of arterial stiffening⁵; the difference between chronological age and vascular age has been found to be a significant predictor of cardiovascular events. More recently, artificial intelligence has been used to estimate cardiovascular or biological age using a different paradigm not relying on cardiovascular risk factors or direct measurement of structural, functional, or physiological changes. Instead, it uses a relatively basic principle: predicting chronological age and using the gap between predicted age minus chronological age as a measure of age rate. First tested analysing the electrocardiogram through deep neural networks,⁶ the gap between predicted and chronological age has been associated to total and cardiovascular mortality,⁷ to genetic conditions causing accelerated aging, and also to cardiovascular risk factors.⁸ Other developed measures of biological age included DNA methylation age, telomere length, protein glycosylation, metabolites, and composite biomarker predictors. While many of these measures have proven successful in predicting health outcomes, some of them are invasive, and others are not validated enough.

The concept of biological age allows for the opportunity to identify patients at a higher risk of adverse future cardiovascular events regardless of absolute age who may be candidates for targeted intervention. An ideal measure of biological age would be feasible and non-invasive and would have a stronger association with outcomes than chronological age. In this issue of the journal, Ganau et al.⁹ developed a novel measure of biological age using non-invasive echocardiographic variables of the left ventricle (LV) and left atrium (LA).

The study population was that of the SardiNIA Project, a longitudinal population-based observational study in Ogliastra region of the Mediterranean island of Sardinia, known for its centenarians, which aimed towards identifying genetic bases for age-associated changes such as cardiovascular risk factors. The study authors performed offline measurements for LV mass index, relative wall thickness, LA volume index, E/A ratio, E/e′ ratio, and aortic root diameter index. They then used sex-specific linear regressions of age with each of these echocardiographic parameters to classify patients into slow, normal, and accelerated aging patterns based on where patients fell relative to the 95% two-sided tolerance intervals. The primary endpoint included non-fatal and fatal cardiovascular events while the secondary endpoint included cardiovascular and non-cardiovascular age-related events and all-cause mortality.

In this relatively healthy sample of 3817 participants, 76% of patients had normal aging pattern, 9% had slow aging, 14% had accelerated aging, while the rest (∼1%) could not be classified. Interestingly, the normal aging group was younger and had lower body mass index, hypercholesterolaemia, and diabetes compared with both the slow and the accelerated aging groups. However, both the primary and secondary endpoints occurred more frequently in the accelerated (vs. normal or slow) and normal (vs. slow) aging patterns over median 4 years of follow-up.

Notably, the separation in survival curves between the three aging patterns was noted at around 30-month follow-up for the primary endpoint and as early as 9 months for the secondary endpoint, suggesting that although the aging patterns were developed using echocardiographic parameters, they may be more reflective of the general health status of the individual than specifically cardiovascular health. This is further supported by the absence of significant differences between normal and either slow or accelerated aging patterns in the multivariable model assessing cardiovascular outcomes while these differences existed in the multivariable model assessing combined cardiovascular and non-cardiovascular outcomes. Although it is well recognized that ascertaining the cause of death is challenging, and inaccuracies frequently exist,¹⁰ the results suggest that the measurement of biological age using echocardiography may be a measure of overall health, not just cardiovascular. Another notable observation is the non-linear association of aging patterns with the secondary endpoint; while having a ‘slow’ vs. ‘normal’ aging pattern is associated with >70% reduction in the combined outcome (hazard ratio of 0.29 vs. normal aging), ‘accelerated’ vs. ‘normal’ aging is associated with only 40% increase in risk. Indeed, it may be overly simplistic to list some events as non-cardiovascular, such as diabetes, hypertension, and dementia. Emerging data identify a bi-directional causative interplay of traditional cardiovascular risk factors and cardiovascular disease. Given the increasing understanding that many cardiovascular diseases in turn predispose to a spectrum of pathologic processes, e.g. a perturbed metabolism, increased oxidative stress, and sympathetic overactivation, it may not be surprising that markers of cardiovascular health are in turn markers of various non-cardiovascular conditions.

The authors also developed and internally tested a computational model to estimate the phenotypic age of the heart (HeartPhAge) based on echocardiographic and clinical parameters. The HeartPhAge was strongly correlated with the chronological age (r = 0.82) and was 9 years younger than chronological age in the slow aging group, comparable in the normal aging group, and 4 years older in the accelerated aging group.

This is a carefully conceived and well-executed analysis developing a novel classification of biological age. This measure is relatively easy to calculate, non-invasive, and feasible, with only 1–2% of patients who could not be classified. Importantly, the developed classification of aging patterns was associated with outcomes independent of chronological age, an essential requirement for any measure of biological age. One challenge that many of the previously developed biological age estimates have faced is that the effects of many chronic diseases cannot be separated from normal aging. The authors did focus on a healthier subset of the population although diseases known to cause structural myocardial changes like diabetes still existed. Reassuringly, patients with the slow aging pattern more frequently had such diseases and yet showed a younger biological age and better outcomes than the normal aging pattern group.

As the authors correctly point out, these results have the potential for innumerable clinical implications; however, future studies validating this novel classification system in other populations are needed before this is considered ready for clinical use. Furthermore, studies evaluating the association of the developed HeartPhAge with outcomes, especially in comparison to the chronological age, will provide important insights to better understand mechanistic links between echocardiographic abnormalities and aging. Identifying patients who are at the highest risk of adverse outcomes will open the doors not only for implementing known preventive strategies and more aggressive treatment of associated comorbidities but also for identifying a target population where novel preventive strategies can be developed and tested and could potentially have a significant public health impact. This study by Ganau et al. represents an important step in the right direction, and the authors are to be commended for their exciting work.

References

Author notes