Biogerontology (2023) 24:845–866

https://doi.org/10.1007/s10522-023-10041-2

REVIEW ARTICLE

Measuring healthy ageing: current and future tools

Nádia Silva · Ana Teresa Rajado · Filipa Esteves · David Brito · Joana Apolónio ·
Vânia Palma Roberto · Alexandra Binnie · Inês Araújo · Clévio Nóbrega · José Bragança ·
Pedro Castelo‑Branco · ALFAScore Consortium

Received: 5 April 2023 / Accepted: 23 May 2023 / Published online: 13 July 2023
© The Author(s) 2023

Abstract Human ageing is a complex, multifacto- biological age, molecular biomarkers, ageing trajec-
rial process characterised by physiological damage, tories, and multi-omics ageing scores are reviewed.
increased risk of age-related diseases and inevita- We conclude that the ideal healthy ageing score is
ble functional deterioration. As the population of the multisystemic and able to encompass all of the poten-
world grows older, placing significant strain on social tial alterations associated with ageing. It should also
and healthcare resources, there is a growing need be longitudinal and able to accurately predict ageing
to identify reliable and easy-to-employ markers of complications at an early stage in order to maximize
healthy ageing for early detection of ageing trajecto- the chances of successful early intervention.
ries and disease risk. Such markers would allow for
the targeted implementation of strategies or treatments Keywords Healthy ageing · Ageing scores · Ageing
that can lessen suffering, disability, and dependence in biomarkers · Biological age
old age. In this review, we summarise the healthy age-
ing scores reported in the literature, with a focus on
the past 5 years, and compare and contrast the vari-
ables employed. The use of approaches to determine

The members of ALFAScore Consortium are listed in the

acknowledgement section.

N. Silva · A. T. Rajado · F. Esteves · D. Brito · P. Castelo‑Branco

J. Apolónio · V. P. Roberto · A. Binnie · I. Araújo · Faculty of Medicine and Biomedical Sciences (FMCB),
C. Nóbrega · J. Bragança · P. Castelo‑Branco (*) University of Algarve, Gambelas Campus, Bld. 2,
Algarve Biomedical Center Research Institute (ABC-RI), 8005‑139 Faro, Portugal
Campus Gambelas, Bld.2, 8005‑139 Faro, Portugal
e-mail: pjbranco@ualg.pt A. Binnie
Department of Critical Care, William Osler Health
V. P. Roberto · I. Araújo · C. Nóbrega · J. Bragança · System, Etobicoke, ON, Canada
P. Castelo‑Branco
ABC Collaborative Laboratory, Association for Integrated I. Araújo · C. Nóbrega · J. Bragança · P. Castelo‑Branco
Aging and Rejuvenation Solutions (ABC CoLAB), Champalimaud Research Program, Champalimaud Centre
8100‑735 Loulé, Portugal for the Unknown, Lisbon, Portugal

A. Binnie · I. Araújo · C. Nóbrega · J. Bragança ·

Vol.: (0123456789)
13
 846 Biogerontology (2023) 24:845–866

Introduction a certain severity, they eventually cause clinically-

measurable changes in anatomic and physiological
Ageing and healthy ageing parameters, limiting physical and cognitive function
(Ferrucci et al. 2018).
The term “healthy ageing” has been widely used to This buffering capacity is the homeodynamic
describe high-functioning older adults based on their space of a biological system, determining an individ-
physical and mental attributes. Initially, healthy age- ual’s health, and the ability to survive and maintain
ing was felt to preclude chronic disease (Rowe and a healthy state. The extent of homeodynamic space
Kahn 1997). However, recently, there has been a achieved by an individual depends both on genetic
shift from a disease-centred model of healthy ageing factors and on pre-natal and early-life epigenetic fac-
towards a function-centred paradigm (Cesari et al. tors, including nutrition, infections, mental stimula-
2018; Cosco et al. 2014). The World Health Organi- tion and physical activity (Rattan 2013, 2020).
zation (WHO) characterises healthy ageing as the Currently, the best strategies to increase health-
“process of developing and maintaining a functional span are physical exercise, healthy nutrition, and life
ability that enables well-being in old age” (World in a socially supportive environment (World Health
Report on Ageing and Health 2015). Functional abil- Organization 2012). Health-oriented and preven-
ity depends on the intrinsic capacity (IC), which is tive strategies, such as hormesis: heat/cold exposure,
the sum of the individual’s physical and mental com- dietary restriction, exercise, and cognitive stimula-
petencies, as well as the individual’s environment tion, have proven to be approaches that potentiate the
and risk factors (World Report on Ageing and Health homeodynamic space and delay ageing (Rattan 2012).
2015). However, these strategies are not always sufficient to
The remarkable increase in human longevity ensure healthy ageing and are difficult for many indi-
observed during the last century has led to a substan- viduals to sustain. Thus, there is significant interest in
tial increase in the number of elderly individuals alive developing new therapies to promote healthy ageing,
today (Vaupel 2010). This has been accompanied by and simultaneously implement tools to monitor and
an increase in the prevalence of numerous chronic, evaluate their efficacy.
non-communicable diseases that arise in old age,
such as cardiovascular disease, cancer, osteoarthri-
tis, and diabetes mellitus type II as well as neurode- Ageing scores
generative diseases such as Alzheimer’s Disease and
Parkinson’s Disease (Franceschi et al. 2018; Li et al. Chronological age only partially reflects an indi-
2021). In fact, the main risk factor for the develop- vidual’s functional and health characteristics. Age-
ment of these diseases is age itself (Hayflick 2021). ing scores are used in epidemiological and sociode-
Understanding the fundamental biology of ageing is mographic settings to characterise the health status
necessary but difficult to achieve, since the progres- of a population (Rodriguez–Laso et al. 2018) Scores
sion, rate and phenotype of ageing differs among typically include chronological age, sex, race, life-
organism, organ, cell types, and molecules within a style, body composition, and the presence of chronic
cell (Rattan 2018). At the cellular level, the molecu- diseases as well as other quantifiable phenotypical or
lar hallmarks of ageing include compromised cell and clinical inputs (Newman et al. 2008). The data used to
tissue function. These cellular effects lead to systemic derive ageing scores are taken from epidemiological
age-related pathologies that are accompanied by loss studies that may be longitudinal, following the same
of function and, ultimately, death (López–Otín et al. individuals over time, or cross-sectional, evaluating
2013; Schmauck-Medina et al. 2022; Singh et al. individuals at a single time point. Scores are typically
2019a, b). created by weighting the factors/variables according
Most ageing-related diseases have long latent peri- to their impact on the intended outcome, whether this
ods that precede their disease manifestations. In the is physical or cognitive performance, disease risk, or
early stages of the disease, buffering at the molecular mortality. Scores are calculated by factor analysis or
level, delays their influence on phenotype and func- by obtaining sub-scores for specific domains based
tional status. However, when perturbations reach on the distribution of the sample, such as z-scores,

Vol:. (1234567890)
13
Biogerontology (2023) 24:845–866 847

quartiles and categories. The combination of the sub- of the PIC that replaces the brain MRI with a cogni-
scores is then achieved by arithmetic sum or aver- tive performance test - the Mini-Mental Status Exam
age. In combination with socio-economic indicators, (MMSE) - and the carotid ultrasound with systolic
scores can ascertain the long-term impact of socio- blood pressure. The HAI is used extensively in epi-
economic and educational factors, lifestyle behav- demiological studies to characterise and compare the
iours, and occupational risks on the ageing quality ageing of different populations worldwide (Nie et al.
of a population (Dieteren et al. 2020; Liu et al. 2019; 2021; O’Connell et al. 2019; Wu et al. 2018; Zhang
O’Connell et al. 2019). In the clinical context, age- et al. 2021).
ing scores can help determine an individual’s disease Modified versions of the HAI have also been cre-
risk, providing a nuanced view of their ageing status ated by adding other variables, particularly functional
and potentially guiding early interventions for at-risk dimensions, to quantify the impact of lifestyle and
individuals. Numerous ageing scores have been pro- life experience on ageing quality (Table 1). The Suc-
posed, which can be loosely divided into three sub- cessful Ageing - Health domains score was derived
types: (a) phenotypic, (b) functional and (c) biologi- from an exploratory factor analysis that identified the
cal (Ferrucci et al. 2018). domains of healthy ageing and their predictive fac-
The choice of derivation cohort affects the appli- tors (Mount et al. 2019). Authors found that Insulin
cability of a given score to individuals. Thus, it is Growth Factor 1 (IGF-1) levels and arterial pulse
crucial that ageing scores are validated in multiple pressure (the difference between systolic and diastolic
populations to ensure generalized application (World blood pressure) were predictors of neuro-sensory
Report on Ageing and Health 2015). functional decline. Similarly, the Biological Health
Despite the change in paradigm of ageing, still in Score [(Karimi et al. 2019), Table 1] is designed to
present days epidemiologically viable metrics of age- measure the “wear and tear” of ageing and includes
ing biology are mostly based on factors which reflect biological markers from 4 physiological domains
an individual’s organismal deterioration. The frailty (endocrine, inflammatory, cardiovascular and meta-
index (FI) is one of the main methods of clinical bolic) and two organs (liver and kidney). The Bio-
evaluation to assess the quality of ageing (Searle et al. logical Health Score was used to examine the impact
2008). It represents the proportion of accumulated of socio-economic position (SEP) on biological age-
deficits of an individual using 40 variables (symp- ing, showing that education-related differences could
toms, signs, functional impairments and laboratory be detected even in young adults (20–40 years old),
abnormalities), reflecting the severity of illness and making a case for the early application of a HAS.
proximity to death (Mitnitski et al. 2001). Following an appeal from the WHO to improve
harmonisation between ageing scores, the Universal
Healthy Ageing Scale, was derived from a harmo-
Physiological and phenotypic healthy ageing scores nized dataset created from 16 worldwide longitudinal
cohorts, called the “Ageing Trajectories of Health:
The first ageing scores were based on physiological Longitudinal Opportunities and Synergies” (ATH-
and phenotypic parameters. The Physiological Index LOS) dataset [(Sanchez–Niubo et al. 2021), Table 1].
of Comorbidity (PIC) (Newman et al. 2008) score was It is hoped that the application of the Universal
designed to identify subjects who were at medium to Health Ageing Scale will help to harmonise future
low risk of disease for enrolment into clinical trials ageing studies globally.
(Charlson et al. 1986). It was based on a combina- Recently, the Intrinsic Capacity Construct (ICC)
tion of clinical measures to identify underlying dis- was proposed, which takes a slightly different view of
ease risk. These included: carotid ultrasound, pul- ageing based on the concept that, although an indi-
monary function testing, brain magnetic resonance vidual’s functional capacity may have fallen below its
(MRI) scan, serum cystatin-C, and fasting glucose peak, they may still be able to maintain key functions
levels. The PIC has been validated as a predictor of if they live in a supportive environment (Cesari et al.
mobility limitation, difficulties with activities of 2018). The ICC comprises 5 domains: cognition, psy-
daily life (ADL), and mortality. The Healthy Ageing chological, locomotion, sensory and vitality. The ICC
Index (HAI) (Sanders et al. 2014) is a simplification has been validated in populations around the world

Vol.: (0123456789)
13
 848 Biogerontology (2023) 24:845–866

Table 1  Scores of healthy ageing and intrinsic capacity

Name of score Objective Variables Outcome

Healthy ageing index (Dieteren Identify ageing trajectories and Systolic blood pressure, non- Classification in ’early’ and
et al. 2020) evaluate the role of baseline fasting plasma glucose levels, “gradual’ ageing population.
sociodemographic charac- global cognitive functioning, Lifestyle factors (e.g. nutrition
teristics and lifestyle factors. plasma creatinine levels and and physical activity) appear
Longitudinal study lung functioning to play an important role in
optimal ageing
Chinese healthy ageing index Creation of a composite Blood pressure, peak expira- Index range (0–12), from
(CHAI) (Nie et al. 2021) measure of healthy ageing tory flow, cognitive status healthiest to unhealthiest
in the Chinese population. score, fasting glucose, kidney
Investigate changes in the function and C-reactive
index over time. Longitudinal protein
study
Successful ageing—Health Use exploratory factor analysis Physical function, cognitive Prediction of objective but
domains (Mount et al. 2019) to identify domains of status, social interactions, not subjective measures of
healthy ageing. Longitudinal psychological status, blood successful ageing. IGF-1
study biomarkers, disease history, and pulse pressure levels are
and socioeconomic status related to neuro-sensory func-
allowed the identification of tion decline
4 domains of ageing: neuro-
sensory function, muscle
function, cardio-metabolic
function, and adiposity
Biological Health score Create a score capturing Endocrine: DHEAS, testos- Contribution of the inflamma-
(Karimi et al. 2019) the wear-and-tear of four terone (men); Inflammatory: tory and metabolic systems to
physiological systems and CRP, fibrinogen, and IGF-1; the overall score. Physiologi-
determine the impact of SEP Metabolic: A1C, HDL, total cal differences can already be
on biological ageing. Cross- cholesterol, and triglycerides; observed in the early-adult
sectional study Cardiovascular: systolic and group (20–40 years)
diastolic blood pressure,
pulses. Liver: ALT, AST and
GGT; Kidney: creatinine*
Universal Healthy ageing scale Creation of a universally 16 worldwide cohorts (343.915 Association with various
(Sanchez-Niubo et al. 2021) applicable scale to evaluate individuals). 41 items sociodemographic, life and
healthy ageing and ageing encompassing activities of health factors and healthy life
trajectories classification. daily living and cognitive and expectancy. Classification in 3
Cross-sectional study physical functioning. Scores ageing trajectories
were rescaled according to
the cohort
Intrinsic capacity (Yu et al. Examine the structure and pre- ICC domains: Locomotor, Prediction of incident IADL
2021) dictive capacity of the ICC. vitality, sensory, cognitive, limitations at the 7-year
Longitudinal study psychological follow-up
Multidimensional model of Applying the ICC, identify a) ICC domains: physiological Gender and economic situation
healthy Ageing (Rivadeneira indicators that discriminate and metabolic health, geriat- seem to play an important role
et al. 2021) healthy ageing from less ric syndromes, risk factors, in healthy ageing
healthy ageing. Cross-sec- physical capacity, cognitive
tional study capacity, and psychological
well-being. b) social and
political environment. c) the
interaction of the older adult
with the environment

Vol:. (1234567890)
13
Biogerontology (2023) 24:845–866 849

Table 1  (continued)
Name of score Objective Variables Outcome

Intrinsic capacity (Cheong Create ICC index. Explore the ICC domains using differ- Validity of 3-domain ICC using
et al. 2022) performance of combining ent variables of: locomotor Time Up-and GO + LogMAR
domain-specific measures. vitality, sensory, cognitive, (visual) + ENIGMA (nutri-
Cross-sectional study psychological tional). Showed excellent
correlations with known health
determinants
Intrinsic capacity (Gutiérrez- Describe the levels of intrinsic ICC domains: cognition, Decreased levels of intrinsic
Robledo et al. 2019) capacity and factors related depression, hearing, vision, capacity were associated with
to its decline. Cross-sectional anorexia, weight loss, and less schooling, self-rated
study mobility health, chronic diseases, visits
to a physician, and ADL

*Dehydroepiandrosterone sulfate (DHEAS), C-Reative protein (CRP), Insulin growth factor - 1 (IGF-1), Hemoglobin A1C (A1C),
high-density lipoprotein cholesterol (HDL), Alanine transaminases (Alt), Aspartate aminotransferase (Ast), Gamma-glutamyl trans-
peptidase (Ggt)

[(Cheong et al. 2022; Gutiérrez-Robledo et al. 2019; is significant interest in identifying early predictors of
Rivadeneira et al. 2021; Yu et al. 2021), Table 1], healthy ageing that can be measured and compared
showing that it can predict Instrumental Activities of at any age (Hartmann et al. 2021; Justice et al. 2018;
Daily Living limitations at 7-year follow-up (Yu et al. Lohman et al. 2021).
2021) and has an excellent correlation with other “Biological age” (BA), is conceptualized as a
known health determinants (Cheong et al. 2022). surrogate measure of a healthy lifespan at any age
However, the components of the ICC differ amongst (Kwon and Belsky 2021). The heritable contribution
studies (Table 1), making cross-study comparisons to lifespan is estimated to be only 25–30% (Brooks-
difficult. Standardisation and further validation are Wilson 2013; van den Berg et al. 2017), as shown
necessary to make the ICC a relevant and useful tool by studies of monozygotic twins (Zenin et al. 2019),
in the clinical and community setting (George et al. as well as populations living in the “blue zones” of
2021; Rivero-Segura et al. 2020). Despite the change healthy ageing, which include Okinawa in Japan,
in paradigm of ageing, still in present days epidemio- Sardinia in Italy, and Nicoya in Costa Rica (Buettner
logically viable metrics of ageing biology are mostly and Skemp 2016). In fact, genome-wide association
based on factors which reflect an individual’s organ- studies (GWAS) have identified only a few loci that
ismal deterioration. The frailty index (FI) is one of are consistently linked with longevity and health-
the main methods of clinical evaluation to assess the span, such as apolipoprotein E (ApoE), Forkhead
quality of ageing (Searle et al. 2008). It represents Box O3 (FOXO3), LDL Receptor Related Protein
the proportion of accumulated deficits of an individ- 1B (LRP1B) and Cyclin-Dependent Kinase Inhibitor
ual using 40 variables (symptoms, signs, functional 2A/B (CDKN2A/B) (Deelen et al. 2019; Melzer et al.
impairments and laboratory abnormalities), reflecting 2020). Thus, researchers have turned to physiological
the severity of illness and proximity to death (Mitnit- variables and multi-omics markers to help explain the
ski et al. 2001). observed variation in healthspan.
Biological ageing scores are derived from ageing
Biological ageing scores datasets that typically include demographic data, out-
come data - functional and physiological, and multi-
Although physiologic and phenotypic ageing scores omics data – epigenetics, transcriptomics, proteom-
are useful for assessing the health and functional sta- ics, metabolomics and microbiome data. Machine
tus of elderly individuals at a specific point in time, learning (ML) approaches allow hypothesis-free data
the window for disease prevention or behaviour cor- mining of these large datasets and can model many
rection may already have closed. Consequently, there different dimensions of the ageing process (Farrell

Vol.: (0123456789)
13
 850 Biogerontology (2023) 24:845–866

et al. 2022; Kwon and Belsky 2021). ML network stable’ group, and a ‘rapid decline’ group. Abstinence
analysis enables the connection between different of physical activity and specific multimorbidity pat-
types of information, and the relationships between terns were associated unfavourable ageing trajectories
different dimensions may represent effects which can- (Moreno-Agostino et al. 2020; Nguyen et al. 2021).
not be described just by statistical correlations (Dato
et al. 2021). These approaches have led to the devel- Epigenetic biological age
opment of biological ageing scores based on a variety
of data types as well as the concepts of ageing phe- Epigenetic biological ageing scores, also known as
notype, ageing trajectory, and ageotype, discussed epigenetic clocks, are collections of DNA methyla-
below. tion sites whose aggregate methylation status meas-
The difference between BA and chronological age ures age (Hannum et al. 2013; Horvath 2013). The
(CA) may be positive, indicating accelerated bio- most commonly applied clocks are the blood-based
logical ageing, or negative, indicating decelerated algorithm by Hannum (Hannum et al. 2013) and
(or healthy) biological ageing. The ideal marker of the multi-tissue algorithm by Horvath (Horvath
biological age should provide reliable prognostic 2013). Both produce a DNA methylation (DNAm)
information about future ageing-associated outcomes age that correlates very closely with CA (r = 0.94).
including comorbidities, functional status or mortal- Researchers have hypothesized that deviations from
ity. It should be able to predict disease onset in pre- CA observed in epigenetic clocks may reflect BA
symptomatic individuals and identify causal lifestyle and health status. Second-generation or “composite”
behaviours, aiding in the development of disease pre- epigenetic clocks include a larger number of DNA
vention strategies (Belsky et al. 2018). methylation sites and also incorporate DNAm sur-
rogates of ageing biomarkers previously described
Physiological ageing scores (Bergsma and Rogaeva 2020; Simpson and Chandra
2021). The DNAmPhenoAge epigenetic clock (Lev-
Several biological ageing scores have been derived ine et al. 2018) was created by regressing a physi-
using physiological variables. PhenoAge is an ML ological measure of mortality risk – PhenoAge – on
derived biological ageing score that captures mor- DNA methylation markers [(Levine et al. 2018; Liu
bidity and mortality risk across diverse populations, et al. 2018), Table 3]. Increased DNAmPhenoAge was
independent of chronological age (Levine et al. 2018; associated with increased activation of pro‐inflam-
Liu et al. 2018). It comprises 10 physiological vari- matory and interferon pathways as well as decreased
ables and is strongly associated with future disease activation of transcriptional/translational machinery,
count (Table 2), enabling researchers to evaluate the DNA damage response, and mitochondrial signatures,
benefits of early interventions. Biological Age is the suggesting that these pathways are important in age-
product of an ML approach in which investigators ing (Levine et al. 2018). DNAmGrimAge is based on
used a deep neural network (DNN) to identify blood surrogate DNAm markers of seven plasma proteins
biomarkers of healthy ageing [(Gialluisi et al. 2022), that increase with age as well as DNAm markers of
Table 2]. The strongest markers of mortality and hos- smoking [(Lu et al. 2019a), Table 3]. DNAmGrimAge
pitalisation risk were Cystatin-C, N-terminal-pro hor- was shown to predict longevity and was also sensi-
mone B-type natriuretic peptide (NT-proBNP), and tive to age-related pathologies, including cognitive
gender. The Physiological Ageing score (PA) [(Sun decline (Hillary et al. 2021), depression (Protsenko
et al. 2021), Table 2] was derived from two inde- et al. 2021), hypertension (Robinson et al. 2020), and
pendent cohorts of individuals in long-lived commu- long-term cardiovascular health (Joyce et al. 2021).
nities (SardiNIA and InCHIANTI). The ratio of PA In The Irish Longitudinal Study on Ageing (TILDA,
to chronological age (PAR) was found to be a sig- N = 590), DNAmGrimAge outperformed Horvath,
nificant predictor of survival as well as a proxy for Hannum, and DNAmPhenoAg epigenetic clocks in
whole-body ageing. The ATHLOS harmonised data- predicting all‐cause mortality and age‐related clinical
set modelled individual healthy ageing trajectories phenotypes.
over 10 years [(Nguyen et al. 2021), Table 2], defin- The ideal epigenetic clock should also detect the
ing 3 ageing trajectories: a ’high stable’ group, a ’low beneficial effects of an improved lifestyle. In a 2-year

Vol:. (1234567890)
13
Biogerontology (2023) 24:845–866 851

Table 2  Scores measuring ageing rates and biological age

Name Objective Variables Outcome

PhenoAge (Levine et al. 2018) Determine the applicability for Chronological age, albumin, A biological age measure;
differentiating risk for vari- creatinine, glucose, CRP, % highly predictive of mortality
ous health outcomes within lymphocyte, mean Red blood and independent of chrono-
diverse subpopulations that cell volume and distribution, logical age. Strong association
include healthy and unhealthy weight, alkaline phosphatase, with disease count. Used as a
groups and distinct age and White blood cell count base for the DNAmPhenoAge
groups. Cross-sectional study clock
Biological age (Gialluisi et al. Biological age algorithm using 36 clinical biomarkers and Δage (chronological age—bio-
2022) DNN. Longitudinal study gender logical age) significantly
predicted mortality and hospi-
talisation risk. Major contribu-
tors to BA were cystatin-C,
NT-proBNP and gender. A
decelerated BA was associated
with higher physical and men-
tal well-being, healthy lifestyle
and higher socioeconomic sta-
tus, while accelerated ageing
was associated with smoking
and obesity
Physiological ageing rate-PAR Predict physiological ageing ML analysis of 148 variables Predictor of physiological age.
(Sun et al. 2021) rate from quantitative traits. in the InCHIANTI and sar- Major contributors are pulse
Identify genetic loci by diNIA ageing cohorts. GWA​ mean velocity, CCA intima-
GWAS. Longitudinal study media thickness, peak systolic
velocity, diastolic CCA
diameter, waist circumfer-
ence and BMI. If PAR > 1, the
individual’s physiological age
is greater than their chrono-
logical age. GWAS 2 loci
associated with PAR: CFI/
GAR1, LINC00202
Universal Healthy ageing tra- Describes healthy ageing 41 items related to health Definition of 3 ageing patterns:
jectories (Moreno-Agostino trajectory patterns and asso- and functioning, such as "high stable", "low stable",
et al. 2020; Nguyen et al. ciation with multimorbidity. ADL cognitive and physical and "rapid decline" groups.
2021) Determine the impact of functioning, using data from The cardiorespiratory/arthritis/
groups of diseases over age- 7 harmonised cohorts cataracts population group
ing trajectories. Longitudinal was associated with the "rapid
study decline" and the "low stable"
groups

follow-up, DNAmGrimAge detected alterations in the An alternative epigenetic strategy for measuring
quality of dietary consumption (Fiorito et al. 2021). BA is to analyse DNAm associated with telomere
Similarly, the Horvath epigenetic clock showed evi- shortening, one of the hallmarks of cellular ageing.
dence of deceleration with improved dietary, lifestyle Leucocyte telomere length (TL) has been widely
behaviours, and medication (Fahy et al. 2019; Fitzger- studied as an ageing biomarker (Vaiserman and Kras-
ald et al. 2021; Gensous et al. 2020). Although these nienkov 2021). However, discrepancies in measure-
are small studies with only short-term follow-up, ment methodologies and issues with replicability
these results suggest that epigenetic clocks could be have undermined its utility (Lulkiewicz et al. 2020).
useful in assessing the efficacy of preventative strate- The DNAmTL is an epigenetic clock that indirectly
gies or treatments to decrease ageing rate or modify measures telomere length (Lu et al. 2019b). This
ageing trajectories. method is easier to use and more robust than standard

Vol.: (0123456789)
13
 852 Biogerontology (2023) 24:845–866

Table 3  Composite Next-Generation blood epigenetic clocks used in healthspan ageing research
Epigenetic clock of ageing Method Additional variables Outcome
CpG sites

DNAm PhenoAge (Levine et al. Illumina 450 K 513 CPGs DNAm surrogate of PhenoAge: DNAm PhenoAge is moderately
2018) chronological age, albumin, heritable and is associated with
creatinine, glucose, C-reactive activation of pro-inflammatory,
protein, lymphocyte %, mean interferon, DNAm damage repair,
red blood cell volume, red blood transcriptional/ translational
cell distribution weight, alkaline signalling, and various markers
phosphatase, White blood cell of immuno-senescence: a decline
count of naïve T cells and shortened
leukocyte telomere length
DNAmGrimAge DNAmGrim- Illumina 450 K&Epic 1030 CPGs DNAm based surrogates: ADM, Lifespan predictor. Results are given
AgeAA (Lu et al. 2019a) B2M, Cystatin-C,GDF-15, Lep- in years. High predictive ability
tin, PAI-1, TIMP-1, DNAm based for time‐to‐death. DNAm-based
estimator of smoking pack-years* surrogate biomarker for smoking
DNAmGrimAgeAA: DNAmGrim- pack-years is a better predictor of
Age and chronological age mortality than the self-reported
biomarker. Associated with age-
related changes in blood cell com-
position and leukocyte telomere
length. Correlated with lifestyle
factors and a host of age-related
conditions
DNAmTL DNAmTLadjAge (Lu Illumina 450 K&Epic 140 CPGs Developed by regressing measured Leukocyte DNAmTL has a strong
et al. 2019b) Leucocyte TL on blood methyla- association with several ageing-
tion related diseases, physical fitness/
DNAmTLadjAge: DNAmTL and functioning, dietary variables,
Chronological age educational attainment, and
income. DNAmTLadjAge is herit-
able and significantly associated
time-to-death, all-cause mortality,
time-to-CV disease, later age at
menopause and positive associa-
tion with physical activity
DunedinPace, Pace of Ageing Illumina Epic 173 CPGs Longitu- DNAm surrogates of 19 indicators Added incremental prediction of
Calculated from the Epigenome dinal study of organ-system integrity: BMI, morbidity, disability, and mortality
(Belsky et al. 2022) Waist-hip ratio, A1C, Leptin, BP, beyond DNAmGrimAge. Can be
­VO2Max, FEV1/FVC, FEV1, used to complement previously
Total cholesterol, Triglycerides, generated epigenetic clocks
HDL, Lipoprotein(a), ApoB100/
A1 ratio, eGFR, BUN, hs-CRP,
White blood cell count, mean
periodontal attachment loss, tooth
decay**
mDNAage (Vetter et al., 2022a) Illumina Epic MS-SNuPE 7 CPGs Chronological age and leukocyte Adaptation and development of a
cell distribution cost-effective epigenetic clock
base on 7 CpGs. Applicable to 2
sequencing techniques

*Adrenomedullin (ADM), beta-2-microglobulin (B2-M), growth differentiation factor 15 (GDF-15), Plasminogen activator inhibitor
1 (PAI-1), and tissue inhibitor metalloproteinases 1 (TIMP-1). **Body mass index (BMI), Hemoglobin A1C (A1C), Blood pressure
(BP), Maximal oxygen consumption (VO2Max), Forced expiratory volume (FEV1), Forced vital capacity (FVC), Apolipoprotein B
(ApoB100), Apolipoprotein A1 (ApoA1), estimated glomerular filtration rate (eGFR), blood urea nitrogen (BUN), high-sensitivity
c-reactive protein (hs-CRP)

TL measurements (Table 3) and is more sensitive to DNA methylation is dynamic, so longitudinal stud-
age-related conditions such as disease and physical ies are necessary to understand how DNAm changes
fitness, making it a potentially useful biomarker in during the life of an individual. A promising next-
ageing interventional studies. generation DNA-methylation biomarker was recently

Vol:. (1234567890)
13
Biogerontology (2023) 24:845–866 853

developed using data from the Dunedin Study of an organism. It is a dynamic entity, which varies
1972–1973 birth cohort (Belsky et al. 2020), which between cell types and changes rapidly in response to
includes 4 longitudinal timepoints. The Dunedin-Pace developmental and environmental cues. Using whole-
of Aging Calculated from the Epigenome Score (Dun- blood gene expression data, Peters et al. identified
edinPACE) [(Belsky et al. 2022), Table 3]. is based 1497 genes whose expression was associated with
on DNAm surrogates for 19 physiological markers of chronological age in leukocytes. The authors used
organ-system integrity (Dieteren et al. 2020; Sanders gene expression profiles to calculate a Transcriptomic
et al. 2014; Wu et al. 2017) combined with DNAm Age and showed that differences between CA and
markers of periodontal attachment loss and tooth Transcriptomic Age were associated with important
decay. The last two variables were incorporated to biological features of ageing including blood pres-
reflect lifestyle and income, socioeconomic factors sure, serum cholesterol, fasting glucose, and body
that have been linked to ageing quality. The use of mass index (Peters et al. 2015). More recently, a Self-
longitudinal cohort data to build the DunedinPACE Organizing Maps ML (SOM-ML) analysis of whole
score helped eliminate many potential confounding blood transcriptome data [(Schmidt et al. 2020),
factors, including survival bias (Vrijheid 2014), and Table 4] revealed two major blood transcriptome
has been validated in 5 epidemiological studies show- types. Type 1 was characterized by increased inflam-
ing improved prediction of morbidity, disability, and mation and increased heme metabolism and was more
mortality when compared to DNAmGrimAge (Belsky commonly found in men, older individuals, and obese
et al. 2022). The cost-effectiveness of the epigenetic individuals. Type 2 was characterised by transcrip-
clock are also a requirement for future clinical appli- tional activation and immune activation and was more
cation, the mDNAage clock [(Vetter et al. 2022a), commonly found in women, younger individuals, and
Table 3] was developed based on 7CpGs only and normal weight individuals.
is sensitive to cardiovascular health scores (Lemke Tools to determine age in vitro are also required
et al., 2022). to facilitate the study of cellular mechanisms of age-
Although studies suggest that DNAm clocks can ing and in vitro testing of anti-ageing therapies. The
predict future disability and mortality, the benefit of Binarized Transcriptomic Aging Clock (BiT age)
epigenetic clocks over more traditional phenotypi- [(Meyer and Schumacher 2021), Table 4] is a tran-
cal and physiological ageing scores is still uncertain. scriptional clock that was developed in C. elegans
A validation of 5 DNAm clocks using data from the and validated in human fibroblasts, where it showed a
Berlin Aging Study II failed to show an association high degree of accuracy in predicting BA. The genes
between DNAm results and health deterioration or included in BiT age support roles for transcription
loss of function after 7 years (Vetter et al., 2022b). factors, the innate immune response and neuronal sig-
In a separate study, markers of epigenetic age accel- nalling as key pathways in cellular ageing (Gill et al.
eration were unable to predict a change in frailty at 2022; Meyer and Schumacher 2021).
1.5 years of follow-up (Seligman et al. 2022). Further
validation is necessary to determine whether DNAm Proteomic biological age
clocks can accurately predict functional outcomes,
during long-term follow-up (Föhr et al. 2022; Mad- Proteins are appealing as biomarkers of ageing
dock et al. 2020). Intriguingly, healthy individuals because their role as direct biological effectors
display methylation changes that are associated with makes it likely that they will reflect the physiologi-
both accelerated and decelerated epigenetic ageing. cal changes of ageing (Tanaka et al. 2018). Recently,
Thus, an individual’s epigenetic age is a function of a plasma proteomic signature of age, PROAge, was
the relative contribution of each site to their overall designed to identify individuals who were ageing
DNA methylation profile (Shahal et al. 2022). faster than their CA [(Tanaka et al. 2020), Table 4].
PROAge includes 76 ageing-associated proteins and
Transcriptomic biological age predicts the development of both ageing-associated
diseases and mortality. A second, ultra-predictive
The transcriptome is the collection of all messenger ageing clock was generated that included 491 plasma
RNA (mRNA) transcripts expressed from the genes proteins [(Lehallier et al. 2020), Table 4]. This clock

Vol.: (0123456789)
13
 854 Biogerontology (2023) 24:845–866

predicted a younger BA for individuals who did regu- to faecal samples to measure the frequency of ribo-
lar exercise relative to those who were sedentary. Pro- somal RNA markers that are specific to certain
teins associated with the immune system were par- microbes or groups of microbes. Machine learning
ticularly useful in predicting CA and BA. techniques are then applied to the sequencing data to
identify features associated with ageing. The top pre-
Metabolomic biological age dictor of longevity in older age groups is alpha-diver-
sity, a measure of within-sample diversity (Biagi
The metabolome, defined as the collection of small et al. 2016; Kong et al. 2016). However, clustering
molecules, and their interactions, within a biologi- of individuals based on alpha-diversity is difficult
cal system, is altered during ageing and may reflect because the microbiome becomes increasingly diver-
underlying physiological function (Johnson et al. gent and unique with age. This individual uniqueness
2019). Plasma metabolome analysis by ultra-high is associated with the enrichment of health-associ-
performance liquid chromatography–mass spectrom- ated bacteria and may be a favourable adaptation to
etry (UHPLC-MS) was used to identify metabolites ageing (Biagi et al. 2016; Kong et al. 2016; Wilman-
that predict faster biological ageing [(Johnson et al. ski et al. 2021).
2019), Table 4]. The metabolites most strongly asso- Studies have shown associations between the
ciated with ageing included amino acid, fatty acid, makeup of the gut microbiome and diet, physical fit-
acylcarnitine, sphingolipid and nucleotide metabo- ness, and frailty, all of which affect health span (Jack-
lites. In a separate study, metabolomic predictors of son et al. 2016). Differences in the prevalence of
age were identified in urine and blood samples from specific microbial species have also been associated
a longitudinal UK cohort and validated in a longitudi- with key markers of health, including inflammation,
nal Finnish cohort [(Robinson et al. 2020), Table 4]. diastolic blood pressure, and weight, suggesting that
Accelerated metabolomic age, defined as metabo- the microbiome plays a role in healthy ageing (Claes-
lomic age greater than CA, was associated with obe- son et al. 2012). A recent study comparing the gut
sity, diabetes, alcohol use, and depression (Robinson microbiome of healthy and unhealthy older adults
et al. 2020). reported an abundance of Akkermansia and Erysip-
A new approach to metabolomics is the analysis elotrichaceae taxa in the healthy cohort. The authors
of volatile organic compounds (VOCs), low-weight hypothesised that these fermentative, complex carbo-
carbon-based molecules that can be detected in sweat, hydrate-digesting bacteria promote healthy intestinal
exhaled breath, blood, urine, and faeces. Urinary and barrier function and thereby contribute to healthy
faecal VOCs can distinguish different age groups ageing (Singh et al. 2019a, b). In contrast, the Entero-
and can also discriminate the offspring of centenar- bacteriaceae family have been associated with mor-
ians from age-matched controls (Conte et al. 2021, tality risk in the general population over an extended
2020).“Breathomics” is the quantification of VOCs in follow-up (Salosensaari et al. 2021).
breath samples. In one study, it was shown to detect The association between microbe prevalence and
age-related differences amongst females (Sukul et al. ageing recently led to the development of a micro-
2022). The use of VOCs as biomarkers of ageing biome-based ageing clock [(Galkin et al. 2020),
requires further validation, however, the development Table 4]. The taxa that were most predictive of CA
of a non-invasive tool to monitor ageing and/or age- were Bifidobacterium spp., Akkermansia muciniph-
related conditions would be invaluable. ila, and Bacteroides spp, which were associated with
good ageing quality, and Escherichia coli and Campy-
Microbiome measurements in ageing lobacter jejuni, which were associated with poor age-
ing quality. Notably, most microbes only impacted
The gut microbiome is responsible for diverse bio- age prediction when their relative abundance reached
logical and metabolic functions, including vitamin a minimum threshold, suggesting that low threshold
synthesis, digestion of dietary fibre, and regula- microbes play a limited role.
tion of the host immune response (Adak and Khan In addition to reflecting the health of the host,
2019; Knight et al. 2017). To assess an individual’s microbiome composition determines microbial meta-
microbiome, next generation sequencing is applied bolic outputs that are subsequently absorbed by the

Vol:. (1234567890)
13
Biogerontology (2023) 24:845–866 855

Table 4  Biological Age determination using omics and multi-omics tools

Name Reference Methodology Objectives Variables Outcome

Blood Transcriptome Whole blood transcrip-Characterising the diver- Lifestyle, obesity, dis- Identified 2 main blood
(Schmidt et al. 2020) tome, microarray sity of transcriptional ease history, medica- transcriptomes, whose
analysis states and their impact tion status and age signatures were shaped
on cellular functions by immune response and
and association with inflammatory processes
ageing phenotypes
BitAge (Meyer and RNA-seq, human dermal Develop a transcrip- Age and a progeria Longitudinal in C. elegans.
Schumacher 2021) fibroblast tomic ageing clock syndrome group Validated in human
base on c. elegans but fibroblasts showing
applicable to human contribution of the
fibroblast transcrip- innate immune response,
tome data neuronal signalling, and
single transcription fac-
tors for biological age
ProAge, PROAge Accel Plasma proteome 76 Develop a method for Age, disease and Identification of a 76-pro-
(Tanaka et al. 2020, proteins Longitudinal in-depth diagnostic mortality tein proteomic age signa-
2018) procedures and early ture PROAge, predictive
interventions in age- accumulation of chronic
ing. Used only healthy diseases and all-cause
adults mortality. Development
of PROAgeAccel for age-
ing rate quantification
Proteome Ageing clock Plasma proteome 491 Datamining of protein Age and lifestyle Proteins associated with
(Lehallier et al. 2020) proteins patterns. Reactome signal transduction, or
pathway analysis the immune system are
capable of predicting
human age. Aerobic-
exercised trained indi-
viduals have a younger
predicted age than
sedentary subjects
Metabolomic age (Rob- Urine and serum Determine metabo- Lifestyle and psycho- Correlated with chrono-
inson et al. 2020) metabolome lomic age. Relate the logical risk factors for logical age. metabolic
metabolome with premature mortality Age Acceleration (mAA)
determinants of accel- was related to over-
erated ageing weight/obesity, diabetes,
heavy alcohol use and
depression
The plasma metabolome Plasma metabolome by Identify plasma metabo- Klemera and Doubal Plasma metabolites are
(Johnson et al. 2019) LC–MS lomic signatures asso- biological age. 360 predictive of faster vs.
ciated with biological plasma metabolites slower ageing trajectory.
ageing in healthy Metabolites most associ-
adults ated with the rate of
biological ageing include
amino acid, fatty acid,
acylcarnitine, sphin-
golipid, and nucleotide
metabolites
Microbiome clock Stool 13 Illumina data- Gut Microbiome Ageing Age and disease Prediction of host age from
(Galkin et al. 2020) sets ENABrowser Clock Based on Taxo- gut microflora profiles.
nomic Profiling and The clock is sensitive to
Deep Learning disease presence. Could
be used as a starting
point for anti-ageing
intervention design

Vol.: (0123456789)
13
 856 Biogerontology (2023) 24:845–866

Table 4  (continued)
Name Reference Methodology Objectives Variables Outcome

Biological age (Earls Multi-omics Longitudi- Biological age estima- Genetic, clinical lval- Measures of metabolic
et al. 2019) nal study tion, applying the ues, metabolome, and health, inflammation,
Klemera-Doubal proteome and toxin bioaccumula-
algorithm using deep tion were strong predic-
phenotyping variables tors of increased BA over
time
Ageotype (Ahadi et al. Multi-omics Longitudi- Use deep phenotyping Transcriptomics, Individuals were grouped
2020) nal study to find a measure cor- proteomics, metabo- in ‘ageotypes’, based on
related with age lomics, cytokines, the types of molecular
microbiome, and pathways that changed
clinical laboratory over time in a given
values individual
Archetype (Zimmer Multi-omics Longitudi- Generate individual Lifestyle,Fitbit records, The model can be used
et al. 2021) nal study archetypes. Find genomics, microbi- for early detection of
enriched traits for each ome, metabolomics, transitions from health
archetype by deep and proteomics to disease state, identify
phenotyping aberrant health condi-
tions and ageing

host. These can induce physiological responses and Multi‑omics biological age
also impact the host metabolome (Lozupone et al.
2012). Indeed, proteomic analysis of the gut microbi- Biological ageing is a complex and multivariate pro-
ome has identified a protein biomarker that is associ- cess, and it is unlikely that a single biological data
ated with ageing: a decrease in tryptophan and indole type can quantify every facet of the ageing process.
synthesis as a consequence of a decline in the phylum Furthermore, there is a notable lack of agreement
Firmicutes in older individuals (> 54 years) (Ruiz- amongst different approaches to quantifying BA, sug-
Ruiz et al. 2020). gesting that different biological clocks may be meas-
Ageing patterns within the gut microbiome could uring different aspects of ageing (Belsky et al. 2018;
have significant clinical implications if beneficial Robinson et al. 2020; Vetter et al., 2022b). This has
interventions can be identified (Wilmanski et al. given rise to the hypothesis that clocks compiled from
2021). However, whether microbiome diversity and multiple data types may better evaluate BA and more
enriched beneficial bacteria are the cause or effect of accurately define ageing trajectories than individual
healthy ageing is still an open question. The utility data types (Table 4). In a recent longitudinal study,
of including microbiome data in ageing scores also authors performed deep phenotyping of 3558 individ-
needs to be established. In a recent longitudinal study, uals that included metabolomics, proteomics, genom-
microbiome features were informative for mortality ics, and clinical variables (Earls et al. 2019). The var-
risk but did not improve prediction relative to other iables most strongly associated with BA were plasma
covariates such as age, sex, BMI, smoking, diabetes, protein levels related to metabolic health, inflamma-
cardiovascular health, and medications (Salosensaari tion, and bioaccumulation of toxins. Interestingly, the
et al. 2021). Also, geography and ethnicity play an association of these biomarkers with BA was gender-
important role in microbiome composition (Kong specific. Notably, this multi-omics approach was sen-
et al. 2016), which may limit the applicability of sitive to changes in lifestyle, with a decrease in BA
microbiome biological ageing scores across different detected amongst participants who were taking part
countries and cultures. in a wellness program that comprised lifestyle coach-
ing on exercise, nutrition, stress management, and
sleep (Earls et al. 2019; Zubair et al. 2019).
Another recent multi-omics study from Ahadi et al.
tracked 106 healthy individuals over 4 years. Deep

Vol:. (1234567890)
13
Biogerontology (2023) 24:845–866 857

phenotyping - including proteomics, metabolomics, intrinsic capacity, it is first necessary to understand

transcriptomics and microbiomics - revealed that the variables and underlying mechanisms that con-
each individual had a specific molecular ageing pat- tribute to healthy ageing.
tern, which the authors termed an “ageotype” [(Ahadi Biological ageing is a multidimensional process,
et al. 2020), Table 4]. Ageotypes could be broadly and probably no single measurement is capable of
grouped into four categories: liver dysfunction, kid- quantify all of its aspects. Ageing scores typically
ney dysfunction, metabolism and inflammation, and measure loss of functionality, i.e. physiological age-
immunity pathways. Inter-individual variability was ing, and are used in different settings but mostly to
detected from a relatively young age, which suggests predict morbidity, disability and mortality. Until
that it may be difficult to create a global ageing score. recently, methodologies and variables used in ageing
However, categorization by ageotype could provide a studies relied mainly on the functional and societal
molecular assessment of an individual’s ageing qual- aspects of ageing with scarce application of molecu-
ity, that might prove useful in monitoring and inter- lar measurements (Dato et al. 2021; Stanziano et al.
vening in the ageing process. Longer-term follow-up 2010). Using new tools of biological age measure-
will be required to determine whether ageotypes can ments simultaneously with classical ageing scores,
predict changes in organ function over time. scientists could more quickly determine the utility
Finally, Zimmer et al. recently combined health of these molecular biomarkers (Levine 2020; Oblak
questionnaires with longitudinal multiomics data et al. 2021).
[(Zimmer et al. 2021), Table 4] to create a multi-
dimensional health model. Based on clinical data, Equivalence between ageing measurements
the authors identified four archetypes/wellness states
within the study population. These archetypes were The low equivalence among approaches used for
subsequently enriched with omics data to character- measurement of biological ageing observed in some
ise each archetype further. Using an individual’s lon- studies, reveals that each might be quantifying differ-
gitudinal data, the authors found that movement over ent aspects of the ageing process (Fiorito et al. 2021;
time within the multidimensional model space could McCrory et al. 2020; Vetter et al. 2019; 2022b). An
(1) detect transitions of ageing, (2) detect transi- example is the assessment of BA by metabolomics,
tions from health to disease, and (3) identify aberrant which was revealed to be complementary, but not
health conditions. associated with established epigenetic clocks, show-
These multidimensional multi-omics models are ing an association with distinct lifestyle risk factors
complex and unlikely to find practical application in instead (Robinson et al. 2020). A comparison of 9
the clinical setting. However, the results of deep phe- BA methodologies (telomere length, 4 DNA meth-
notyping in exploratory studies will help refine the ylation clocks, physiological age, cognitive function,
discovery of new and improved biomarkers of ageing functional ageing index (FAI), and frailty index (FI)
and health in ageing. was performed in a single longitudinal cohort (Li
et al. 2020). All BAs were correlated with each other
to some degree, in large part due to their correlations
Discussion and future perspectives with CA. However, except for telomere length, they
were also independently associated with mortality
Measurements of healthy ageing are a valuable risk, showing that BA can be better than CA at pre-
tool for understanding ageing dynamics within and dicting mortality. Of the BA methodologies that were
amongst populations. By characterizing populations compared, the best independent predictors of mor-
and their health/longevity outcomes it is possible tality were DNAmGrimAge and FI. In a joint model,
to expand the knowledge of lifestyle habits or envi- DNAmHorvath, DNAmGrimAge, and FI showed com-
ronmental conditions that contribute to the health- plementarity in predicting mortality risk.
span. Future improvements in ageing quality will The differences in outcome prediction when using
require both individual and policy-level changes. To different methodologies, may arise due to differences
understand which interventions mostly improve age- in sample size, study-specific age cut-offs to define
ing quality by increasing homeodynamic space and the affectation status, sex- specificity, and population

Vol.: (0123456789)
13
 858 Biogerontology (2023) 24:845–866

specificity, i.e., genetic and/or lifestyle heterogene- Use of MLin the measurements of health in ageing
ity among cohorts (Dato et al. 2021). The inclusion
of different population backgrounds is particularly Critics state that the use of ageing scores, especially
crucial in ageing, since it is heavily influenced by a biological ageing, reduces the comparison of complex
strong geographical component and environmen- biological states, such as the heterogeneity observed
tal exposure. The lack of homogeneity in the data in ageing, to the comparison of single numbers,
obtained from epidemiological, demographic, and which destroys information because it assumes that
even clinical markers is problematic. Also there is age-dependent differences between individuals can
limitation to the range of markers obtained from be depicted by a single dimension (Freund 2019).
each study (Kwon and Belsky 2021) since they are a Another challenge is that the combination of molec-
combination of multiple assays and sometimes differ- ular and phenotypic data is not able to distinguish
ent laboratory methodologies. The ATHLOS project between the effects and the causes of ageing (New-
(https://​athlos.​pssjd.​org/) and Maelstrom research man, 2015), with sometimes the presence of biomark-
catalogue (https://​www.​maels​trom-​resea​rch.​org/) are ers of chronic diseases associated with ageing, being
examples of resources intended to harmonize data the main drivers of the scores.
across studies in order to obtain universal scientific An attempt to overcome these limitations is using
data, applicable worldwide. While the universality modern analytic techniques to perform high-dimen-
of ageing scores is not established, researchers must sional analysis, more representative of biological real-
be familiar with the advantages and drawbacks of the ity (Cohen et al. 2019). The use of machine learning
different measurements for answering their research algorithms for assessing ageing quality allows for the
questions (Nelson et al. 2020). inclusion of more ageing manifestations as outcomes,
Epigenetic clocks are promising measures of age- which may improve the predictive value of the models
ing quality, that have demonstrated potential to serve (Sun et al. 2021). ML allows an hypothesis-free data-
as a reliable ageing biomarkers. They are generated mining, instead of an hypothesis-driven data testing
from a single multiplex array and include the same (Hägg et al. 2019). Given these advantages the appli-
measurements across studies making comparisons cation of these models to ongoing ageing cohorts is
and validation easier (Kwon and Belsky 2021). How- being implemented more routinely (Gomez-Cabrero
ever, these DNAm-based biomarkers tools are still not et al. 2021; Speiser et al. 2021; Varzaneh et al. 2022).
considered a replacement for validated measures of Recent reviews have approached the challenges
physical and cognitive performance in old age (Mad- associated with integrating omics measurements and
dock et al. 2020). Also, to understand the molecular ML data analysis in ageing research, calling out to
origins underlying the observable epigenetic differ- data integration, interpretation and sharing of high-
ences further investigation is needed. throughput data as the main issues to be resolved
However, although a lot of progress has been made (Dato et al. 2021; Zhavoronkov et al. 2019). Despite
in identifying biological markers of ageing, the use ML offering an alternative to traditional approaches
of molecular biomarkers in ageing scores remains for modelling outcomes in ageing, scepticism over
fundamentally challenging. First, the contribution of these methods persists due to lack of reproducibility
each molecular biomarker to BA is small, with high and interpretability of the complex algorithms that
variability and frequent replicability issues. Second, underlie these models (Speiser et al. 2021). Although
molecular biomarkers can be modified in response promising, ML algorithms warrant further characteri-
to multiple factors including genetics, lifetime expo- zation and validation, since their biological, clinical
some, and the presence of age-related diseases. Thus, and environmental correlates remain largely unex-
interpreting their significance with respect to ageing plored (Gialluisi et al. 2022).
can be complex. Third, validating molecular biomark-
ers as surrogates for health span will require evidence
that these scores are modifiable through interven- Application of ageing scores in younger populations
tion and that the resulting phenotypes have improved
long-term outcomes. While phenotypic and physiological ageing scores
are excellent tools for assessing ageing quality in the

Vol:. (1234567890)
13
Biogerontology (2023) 24:845–866 859

Fig. 1  Evaluation of the ageing quality throughout the lifes- and frailty index. In fact, ageing begins earlier in life with the
pan. Currently, measurement and evaluation of ageing begin molecular imbalance; application of new biomarkers of ageing
when ageing-related diseases arise, ending the health span quality (ex., epigenetic clocks, transcriptome or metabolome)
period of life. This usually occurs after 60 years of age when can be used from early adulthood to determine biological age
physiological imbalance gives rise to functional impairment. and ageing rate. In addition, ageing trajectories and ageotype
Current evaluation of ageing uses several approaches, among could be used to monitor ageing progression and allow imple-
them the healthy ageing index, intrinsic capacity construct, mentation of healthy ageing policies from a young adult age

elderly, they have lower utility for predicting ageing needed to understand the cellular and molecular pro-
quality in younger populations (Nelson et al. 2020). cesses that underlie the epigenetic changes of ageing
This gap can potentially be addressed by biological and the redout of the clocks to evaluate ageing qual-
ageing scores, thereby enabling the study of early ity (Bell et al. 2019; Oblak et al. 2021; Raj and Hor-
interventions to favour healthy ageing trajectories in a vath 2020).
precision medicine scenario (Fig. 1).
The observation that younger adults show vari-
able ageing rates and ageotypes (Ahadi et al. 2020;
Belsky et al. 2022; Dieteren et al. 2020; Karimi et al. Conclusion
2019) makes a strong case for longitudinal studies
of biological ageing scores in younger populations. In the last 5 years, the measurement of healthy age-
This would allow for the identification of key molec- ing has taken a significant leap forward. On the one
ular mechanisms of ageing before the emergence of hand, there has been the development of the concept
age-related diseases. The DunedinPace clock (Bel- of intrinsic capacity, recognizing the importance of
sky et al. 2022), which was derived from a cohort lifestyle, well-being, and societal participation in
of young adults followed until the age of 45, showed achieving healthy ageing. On the other hand, life sci-
sensitivity to changes in individual ageing trajecto- entists are plunging ever deeper into molecular meas-
ries. However, the ageing outcomes of these individ- urements of ageing, trying to establish new biomarker
uals, for the next 30 years, still need to be established panels to identify ageing trajectories and phenotypes.
to understand the relationship between early ageing To tackle the current and future challenges of an age-
trajectories and healthspan. Further investigation is ing population, robust ageing scores that encompass

Vol.: (0123456789)
13
 860 Biogerontology (2023) 24:845–866

all of the alterations suffered by an individual dur- input and manuscript writing and revision. PCB - Conceptual-
ing ageing are required. The ideal HAS should be ization, scientific guidance and critical revision of the manu-
script. All authors read and approved the final manuscript.
multisystemic, predictive of future health status, and
responsive to change, thereby capturing an individu- Funding Open access funding provided by FCT|FCCN
al’s current and future ageing trajectories. It is likely (b-on). This work was supported by CRESC Algarve 2020
that HAS will differ between the research environ- (Operation Code: ALG-01–0145-FEDER-072586).
ment, where in-depth phenotyping is possible and Declarations
desirable, and the clinical environment, where a more
pragmatic approach is required. However, the ideal Competing interests Authors declare no competing interests.
healthy ageing score for both research and clinical
purposes will probably adopt a multi-omics approach Open Access This article is licensed under a Creative Com-
to optimize reliability and ensure that the complexity mons Attribution 4.0 International License, which permits
of the ageing process is adequately captured. use, sharing, adaptation, distribution and reproduction in any
medium or format, as long as you give appropriate credit to the
original author(s) and the source, provide a link to the Crea-
Acknowledgements The authors acknowledge the Algarve
tive Commons licence, and indicate if changes were made. The
Biomedical Center for structural support.
images or other third party material in this article are included
The members of ALFAScore Consortium are Raquel P.
in the article’s Creative Commons licence, unless indicated
Andrade (Algarve Biomedical Center - Research Institute (ABC-
otherwise in a credit line to the material. If material is not
RI), University of Algarve, Faro, Portugal; Algarve Biomedical
included in the article’s Creative Commons licence and your
Center (ABC), University of Algarve, Faro, Portugal; Faculty
intended use is not permitted by statutory regulation or exceeds
of Medicine and Biomedical Sciences (FMCB), University of
the permitted use, you will need to obtain permission directly
Algrave, Faro, Portugal; Champalimaud Research Program,
from the copyright holder. To view a copy of this licence, visit
Champalimaud Centre for the Unknown, Lisbon, Portugal),
http://​creat​iveco​mmons.​org/​licen​ses/​by/4.​0/.
Sofia Calado (Algarve Biomedical Center - Research Institute
(ABC-RI), University of Algarve, Faro, Portugal; Algarve Bio-
medical Center (ABC), University of Algarve, Faro, Portugal;
Faculty of Medicine and Biomedical Sciences (FMCB), Univer- Glossary
sity of Algrave, Faro, Portugal), Maria Leonor Faleiro (Algarve
Biomedical Center - Research Institute (ABC-RI), University
of Algarve, Faro, Portugal; Algarve Biomedical Center (ABC), Ageing Multifactorial process
University of Algarve, Faro, Portugal; Faculty of Medicine and characterised by func-
Biomedical Sciences (FMCB), University of Algrave, Faro, tional deterioration, physi-
Portugal; Faculty of Science and Technology (FCT), University
ological damage, and mul-
of Algarve, Faro Portugal), Carlos Matos, Nuno Marques, Ana
Marreiros, Hipólito Nzwalo, Sandra Pais (Algarve Biomedical tiple age-related diseases.
Center - Research Institute (ABC-RI), University of Algarve, Ageing phenotype  Set of measurable age-
Faro, Portugal; Algarve Biomedical Center (ABC), University ing traits shared by a
of Algarve, Faro, Portugal; Faculty of Medicine and Biomedi-
population.
cal Sciences (FMCB), University of Algrave, Faro, Portugal),
Isabel Palmeirim (Algarve Biomedical Center - Research Insti- Ageing Scores  Set of criteria used for
tute (ABC-RI), University of Algarve, Faro, Portugal; Algarve the evaluation of ageing
Biomedical Center (ABC), University of Algarve, Faro, Por- quality.
tugal; University of Algrave, Faro, Portugal; Champalimaud
Ageing trajectories The behaviour of ageing
Research Program, Champalimaud Centre for the Unknown,
Lisbon, Portugal), Sónia Simão (Algarve Biomedical Center phenotypes over time.
- Research Institute (ABC-RI), University of Algarve, Faro, Ageotype  Ageing patterns that
Portugal; Algarve Biomedical Center (ABC), University of are classified based on
Algarve, Faro, Portugal; Faculty of Medicine and Biomedical
molecular pathways that
Sciences (FMCB), University of Algrave, Faro, Portugal), Naté-
rcia Joaquim (USF Balsa, Tavira, Portugal), Rui Miranda (USF change over time within
Balsa, Tavira, Portugal), António Pêgas (USF Ossónoba, Faro, an individual.
Portugal), Ana Sardo (USF Mirante, Olhão, Portugal). Biological Age (BA) Age measured according
to biomarkers/physiologi-
Author contributions NS - Conceptualization and writing of
cal parameters. The dif-
the original manuscript. ATR, FE, DB, JA, VPR, the ALFAS-
core Consortium, AB, IA, CN and JB -Provided intellectual ference between chrono-
logical age (CA) and

Vol:. (1234567890)
13
Biogerontology (2023) 24:845–866 861

biological age is consid- References

ered a measure of ageing
quality. Adak A, Khan MR (2019) An insight into gut microbiota and
its functionalities. Cell Mol Life Sci 76:473–493
Exposome  Sum of environmental
Ahadi S, Zhou W, Schüssler-Fiorenza Rose SM, Sailani MR,
exposures during an indi- Contrepois K, Avina M, Ashland M, Brunet A, Snyder M
vidual’s lifetime (such as (2020) Personal aging markers and ageotypes revealed by
lifestyle behaviours, pollu- deep longitudinal profiling. Nat Med 26:83–90
Bell CG, Lowe R, Adams PD, Baccarelli AA, Beck S, Bell JT,
tion, or stress).
Christensen BC, Gladyshev VN, Heijmans BT, Horvath
Frailty index (FI) 
An age-related condi- S, Ideker T, Issa JPJ, Kelsey KT, Marioni RE, Reik W,
tion of increased risk for Relton CL, Schalkwyk LC, Teschendorff AE, Wagner W,
adverse health outcomes Zhang K, Rakyan VK (2019) DNA methylation aging
clocks: challenges and recommendations. Genome Biol
caused by a decrease in
20:1–24
homeodynamic space Belsky DW, Moffitt TE, Cohen AA, Corcoran DL, Levine ME,
exposing the individual to Prinz JA, Schaefer J, Sugden K, Williams B, Poulton
a higher risk of adverse R, Caspi A (2018) Eleven telomere, epigenetic clock,
and biomarker-composite quantifications of biological
outcomes, such as multi-
aging: do they measure the same thing? Am J Epidemiol
morbidity, falls, disability, 187:1220–1230
nursing home placement, Belsky DW, Caspi A, Arseneault L, Baccarelli A, Corcoran
and death. D, Gao X, Hannon E, Harrington HL, Rasmussen LJH,
Houts R, Huffman K, Kraus WE, Kwon D, Mill J, Pieper
Healthy ageing  The process of develop-
CF, Prinz J, Poulton R, Schwartz J, Sugden K, Vokonas
ing and maintaining func- P, Williams BS, Moffitt TE (2020) Quantification of the
tional abilities that enable pace of biological aging in humans through a blood test,
well-being in older age. the DunedinPoAm DNA methylation algorithm. Elife
9:1–56
Healthspan  Period of life spent in
Belsky DW, Caspi A, Corcoran DL, Sugden K, Poulton R,
good health, free from Arseneault L, Baccarelli A, Chamarti K, Gao X, Hannon
chronic disease and E, Harrington HL, Houts R, Kothari M, Kwon D, Mill
disability. J, Schwartz J, Vokonas P, Wang C, Williams BS, Moffitt
TE (2022) DunedinPACE, a DNA methylation biomarker
Hormesis  Mild stress induced-acti-
of the pace of aging. Elife 11:1–26
vation of adaptive and pro- Bergsma T, Rogaeva E (2020) DNA methylation clocks and
tective pathways in cells their predictive capacity for aging phenotypes and health-
and organisms, presenting span. Neurosci Insights 15:1–11
Biagi E, Franceschi C, Rampelli S, Severgnini M, Ostan R,
numerous health-promot- Turroni S, Consolandi C, Quercia S, Scurti M, Monti D,
ing, ageing-modulatory Capri M, Brigidi P, Candela M (2016) Gut Microbiota
and lifespan-extending and extreme longevity. Curr Biol 26:1480–1485
effects. Brooks-Wilson AR (2013) Genetics of healthy aging and lon-
gevity. Hum Genet 132:1323–1338
Intrinsic Buettner D, Skemp S (2016) Blue zones: lessons from the
capacity (IC)  An individual’s biologi- world’s longest lived. Am J Lifestyle Med 10:318–321
cal capacity based on Cesari M, Araujo de Carvalho I, Amuthavalli Thiyagarajan J,
five functional domains: Cooper C, Martin FC, Reginster J-Y, Vellas B, Beard JR
(2018) Evidence for the domains supporting the construct
locomotion, cognition, of intrinsic capacity. J Gerontol Ser A 73:1653–1660
psychology, vitality, and Charlson ME, Sax FL, MacKenzie CR, Fields SD, Braham
sensory. IC goes beyond RL, Douglas RGJ (1986) Assessing illness severity: does
genetics and health sta- clinical judgment work? J Chronic Dis 39:439–452
Cheong CY, Yap P, Nyunt MSZ, Qi G, Gwee X, Wee SL, Yap
tus to encompass how the KB, Ng TP (2022) Functional health index of intrinsic
person functions in their capacity: multi-domain operationalisation and validation
environment. in the Singapore Longitudinal Ageing Study (SLAS2).
Lifespan  The time between birth Age Ageing 51:afac011
Claesson MJ, Jeffery IB, Conde S, Power SE, O’Connor EM,
and death of an organism. Cusack S, Harris HMB, Coakley M, Lakshminarayanan

Vol.: (0123456789)
13
 862 Biogerontology (2023) 24:845–866

B, O’Sullivan O, Fitzgerald GF, Deane J, O’Connor M, immunosenescent trends in humans. Aging Cell
Harnedy N, O’Connor K, O’Mahony D, van Sinderen D, 18:e13028
Wallace M, Brennan L, Stanton C, Marchesi JR, Fitzger- Farrell S, Mitnitski A, Rockwood K, Rutenberg AD (2022)
ald AP, Shanahan F, Hill C, Ross RP, O’Toole PW Interpretable machine learning for highdimensional tra-
(2012) Gut microbiota composition correlates with diet jectories of aging health. PLoS Comput Biol 18:1–30
and health in the elderly. Nature 488:178–184 Ferrucci L, Levine ME, Kuo PL, Simonsick EM (2018) Time
Cohen AA, Luyten W, Gogol M, Simm A, Saul N, Cirulli and the metrics of aging. Circ Res 123:740–744
F, Berry A, Antal P, Köhling R, Wouters B, Möller S, Fiorito G, Caini S, Palli D, Bendinelli B, Saieva C, Ermini I,
Fuellen G, Jansen L (2019) Health and aging: Unifying Valentini V, Assedi M, Rizzolo P, Ambrogetti D, Ottini
concepts, scores, biomarkers and pathways. Aging Dis L, Masala G (2021) DNA methylation-based biomark-
10:883–900 ers of aging were slowed down in a two-year diet and
Conte M, Conte G, Martucci M, Monti D, Casarosa L, Serra physical activity intervention trial: the DAMA study.
A, Mele M, Franceschi C, Salvioli S (2020) The smell of Aging Cell 20:1–13
longevity: a combination of Volatile Organic Compounds Fitzgerald KN, Hodges R, Hanes D, Stack E, Cheishvili D,
(VOCs) can discriminate centenarians and their offspring Szyf M, Henkel J, Twedt MW, Giannopoulou D, Herd-
from age-matched subjects and young controls. GeroSci- ell J, Logan S, Bradley R (2021) Potential reversal of
ence 42:201–216 epigenetic age using a diet and lifestyle intervention:
Conte M, Conte G, Salvioli S (2021) VOCs profile can discrim- a pilot randomized clinical trial. Aging (albany. NY)
inate biological age. Aging (albany. NY) 13:9156–9157 13:9419–9432
Cosco TD, Prina AM, Perales J, Stephan BCM, Brayne C Föhr T, Törmäkangas T, Lankila H, Viljanen A, Rantanen T,
(2014) Operational definitions of successful aging: a sys- Ollikainen M, Kaprio J, Sillanpää E (2022) The associ-
tematic review. Int Psychogeriatrics 26:373–381 ation between epigenetic clocks and physical function-
Dato S, Crocco P, Rambaldi Migliore N, Lescai F (2021) ing in older women: a 3-year follow-up. J Gerontol Ser
Omics in a digital world: the role of bioinformatics in A 77:1569–1576
providing new insights into human aging. Front Genet Franceschi C, Garagnani P, Morsiani C, Conte M, Santoro A,
12:1–17 Grignolio A, Monti D, Capri M, Salvioli S (2018) The
Deelen J, Evans DS, Arking DE, Tesi N, Nygaard M, Liu continuum of aging and age-related diseases: common
X, Wojczynski MK, Biggs ML, van der Spek A, Atz- mechanisms but different rates. Front Med 5:61
mon G, Ware EB, Sarnowski C, Smith AV, Seppälä I, Freund A (2019) Untangling aging using dynamic Organism-
Cordell HJ, Dose J, Amin N, Arnold AM, Ayers KL, Level Phenotypic Networks. Cell Syst 8:172–181
Barzilai N, Becker EJ, Beekman M, Blanché H, Chris- Galkin F, Mamoshina P, Aliper A, Putin E, Moskalev V, Gla-
tensen K, Christiansen L, Collerton JC, Cubaynes S, dyshev VN, Zhavoronkov A (2020) Human gut micro-
Cummings SR, Davies K, Debrabant B, Deleuze JF, biome aging clock based on taxonomic profiling and
Duncan R, Faul JD, Franceschi C, Galan P, Gudna- deep learning. iScience 23:101199
son V, Harris TB, Huisman M, Hurme MA, Jagger C, Gensous N, Garagnani P, Santoro A, Giuliani C, Ostan R,
Jansen I, Jylhä M, Kähönen M, Karasik D, Kardia SLR, Fabbri C, Milazzo M, Gentilini D, di Blasio AM, Pie-
Kingston A, Kirkwood TBL, Launer LJ, Lehtimäki T, truszka B, Madej D, Bialecka-Debek A, Brzozowska
Lieb W, Lyytikäinen LP, Martin-Ruiz C, Min J, Nebel A, Franceschi C, Bacalini MG (2020) One-year Medi-
A, Newman AB, Nie C, Nohr EA, Orwoll ES, Perls TT, terranean diet promotes epigenetic rejuvenation with
Province MA, Psaty BM, Raitakari OT, Reinders MJT, country- and sex-specific effects: a pilot study from the
Robine JM, Rotter JI, Sebastiani P, Smith J, Sørensen NU-AGE project. GeroScience 42:687–701
TIA, Taylor KD, Uitterlinden AG, van der Flier W, George PP, Lun P, Ong SP, Lim WS (2021) A rapid review of
van der Lee SJ, van Duijn CM, van Heemst D, Vaupel the measurement of intrinsic capacity in older adults. J
JW, Weir D, Ye K, Zeng Y, Zheng W, Holstege H, Kiel Nutr Heal Aging 25:774–782
DP, Lunetta KL, Slagboom PE, Murabito JM (2019) A Gialluisi A, Di Castelnuovo A, Costanzo S, Bonaccio M, Per-
meta-analysis of genome-wide association studies iden- sichillo M, Magnacca S, De Curtis A, Cerletti C, Donati
tifies multiple longevity genes. Nat Commun 10:3369 MB, de Gaetano G, Capobianco E, Iacoviello L (2022)
Dieteren CM, Samson LD, Schipper M, van Exel J, Brou- Exploring domains, clinical implications and environ-
wer WBF, Verschuren WMM, Picavet HSJ (2020) The mental associations of a deep learning marker of biologi-
healthy aging index analyzed over 15 years in the gen- cal ageing. Eur J Epidemiol 37:35–48
eral population: the doetinchem cohort study. Prev Med Gill D, Parry A, Santos F, Okkenhaug H, Todd CD, Hernando-
139:106193 Herraez I, Stubbs TM, Milagre I, Reik W (2022) Multi-
Earls JC, Rappaport N, Heath L, Wilmanski T, Magis AT, omic rejuvenation of human cells by maturation phase
Schork NJ, Omenn GS, Lovejoy J, Hood L, Price ND transient reprogramming. Elife 11:e71624
(2019) Multi-omic biological age estimation and its Gomez-Cabrero D, Walter S, Abugessaisa I, Miñambres-
correlation with wellness and disease phenotypes: a Herraiz R, Palomares LB, Butcher L, Erusalimsky JD,
longitudinal study of 3,558 individuals. J Gerontol— Garcia-Garcia FJ, Carnicero J, Hardman TC, Mischak H,
Ser A Biol Sci Med Sci 74:S52–S60 Zürbig P, Hackl M, Grillari J, Fiorillo E, Cucca F, Cesari
Fahy GM, Brooke RT, Watson JP, Good Z, Vasanawala M, Carrie I, Colpo M, Bandinelli S, Feart C, Peres K,
SS, Maecker H, Leipold MD, Lin DTS, Kobor MS, Dartigues JF, Helmer C, Viña J, Olaso G, García-Palm-
Horvath S (2019) Reversal of epigenetic aging and ero I, Martínez JG, Jansen-Dürr P, Grune T, Weber D,

Vol:. (1234567890)
13
Biogerontology (2023) 24:845–866 863

Lippi G, Bonaguri C, Sinclair AJ, Tegner J, Rodriguez- Kong F, Hua Y, Zeng B, Ning R, Li Y, Zhao J (2016)
Mañas L (2021) A robust machine learning framework Gut microbiota signatures of longevity. Curr Biol
to identify signatures for frailty: a nested case-control 26:R832–R833
study in four aging European cohorts. GeroScience Kwon D, Belsky DW (2021) A toolkit for quantification of
43:1317–1329 biological age from blood chemistry and organ func-
Gutiérrez-Robledo LM, García-Chanes RE, Pérez-Zepeda MU tion test data: BioAge. GeroScience 43:2795–2808
(2019) Allostatic load as a biological substrate to intrin- Lehallier B, Shokhirev MN, Wyss-Coray T, Johnson AA
sic capacity: a secondary analysis of CRELES. J Nutr (2020) Data mining of human plasma proteins gener-
Heal Aging 23:788–795 ates a multitude of highly predictive aging clocks that
Hägg S, Belsky DW, Cohen AA (2019) Developments in reflect different aspects of aging. Aging Cell 19:1–19
molecular epidemiology of aging. Emerg Top Life Sci Lemke E, Vetter VM, Berger N, Banszerus VL, König M,
3:411–421 Demuth I (2022) Cardiovascular health is associated
Hannum G, Guinney J, Zhao L, Zhang L, Hughes G, Sadda with the epigenetic clock in the Berlin Aging Study II
S, Klotzle B, Bibikova M, Fan J-B, Gao Y, Deconde (BASE-II). Mech Ageing Dev 201:111616
R, Chen M, Rajapakse I, Friend S, Ideker T, Zhang K Levine ME (2020) Assessment of epigenetic clocks as bio-
(2013) Genome-wide methylation profiles reveal quanti- markers of aging in basic and population research. J
tative views of human aging rates. Mol Cell 49:359–367 Gerontol Ser A 75:463–465
Hartmann A, Hartmann C, Secci R, Hermann A, Fuellen G, Levine ME, Lu AT, Quach A, Chen BH, Assimes TL,
Walter M (2021) Ranking biomarkers of aging by cita- Bandinelli S, Hou L, Baccarelli AA, Stewart JD, Li Y,
tion profiling and effort scoring. Front Genet 12:1–15 Whitsel EA, Wilson JG, Reiner AP, Aviv A, Lohman
Hayflick L (2021) The greatest risk factor for the leading cause K, Liu Y, Ferrucci L, Horvath S (2018) An epigenetic
of death is ignored. Biogerontology 22:133–141 biomarker of aging for lifespan and healthspan. Aging
Hillary RF, Stevenson AJ, Cox SR, McCartney DL, Harris SE, (albany. NY) 10:573–591
Seeboth A, Higham J, Sproul D, Taylor AM, Redmond P, Li X, Ploner A, Wang Y, Magnusson PKE, Reynolds C, Fin-
Corley J (2021) An epigenetic predictor of death captures kel D, Pedersen NL, Jylhävä J, Hägg S (2020) Longi-
multi-modal measures of brain health. Mol Psychiatry tudinal trajectories, correlations and mortality associa-
26(8):3806–3816 tions of nine biological ages across 20-years follow-up.
Horvath S (2013) DNA methylation age of human tissues and Elife 9:1–20
cell types. Genome Biol 14:R115 Li Z, Zhang Z, Ren Y, Wang Y, Fang J, Yue H, Ma S,
Jackson MA, Jeffery IB, Beaumont M, Bell JT, Clark AG, Ley Guan F (2021) Aging and age-related diseases: from
RE, O’Toole PW, Spector TD, Steves CJ (2016) Signa- mechanisms to therapeutic strategies. Biogerontology
tures of early frailty in the gut microbiota. Genome Med 22:165–187
8:8 Liu Z, Kuo P-L, Horvath S, Crimmins E, Ferrucci L, Lev-
Johnson LC, Parker K, Aguirre BF, Nemkov TG, D’Alessandro ine M (2018) A new aging measure captures morbidity
A, Johnson SA, Seals DR, Martens CR (2019) The and mortality risk across diverse subpopulations from
plasma metabolome as a predictor of biological aging in NHANES IV: A cohort study. PLOS Med 15:e1002718
humans. GeroScience 41:895–906 Liu Z, Chen X, Gill TM, Ma C, Crimmins EM, Levine ME
Joyce BT, Gao T, Zheng Y, Ma J, Hwang SJ, Liu L, Nannini (2019) Associations of genetics, behaviors, and life
D, Horvath S, Lu AT, Bai Allen N, Jacobs DR, Gross course circumstances with a novel aging and healthspan
M, Krefman A, Ning H, Liu K, Lewis CE, Schreiner measure: evidence from the health and retirement study.
PJ, Sidney S, Shikany JM, Levy D, Greenland P, Hou PLoS Med 16:e1002827
L, Lloyd-Jones D (2021) Epigenetic age accelera- Lohman T, Bains G, Berk L, Lohman E (2021) Predictors of
tion reflects long-term cardiovascular health. Circ Res biological age: the implications for wellness and aging
129:770–781 research. Gerontol Geriatr Med 7:1–13
Justice JN, Ferrucci L, Newman AB, Aroda VR, Bahn- López-Otín C, Blasco MA, Partridge L, Serrano M, Kroemer
son JL, Divers J, Espeland MA, Marcovina S, Pollak G (2013) The hallmarks of Aging. Cell 153:1194–1217
MN, Kritchevsky SB, Barzilai N, Kuchel GA (2018) Lozupone CA, Stombaugh JI, Gordon JI, Jansson JK, Knight
A framework for selection of blood-based biomark- R (2012) Diversity, stability and resilience of the human
ers for geroscience-guided clinical trials: report from gut microbiota. Nature 489:220–230
the TAME Biomarkers Workgroup. GeroScience Lu AT, Quach A, Wilson JG, Reiner AP, Aviv A, Raj K,
40:419–436 Hou L, Baccarelli AA, Li Y, Stewart JD, Whitsel EA,
Karimi M, Castagné R, Delpierre C, Albertus G, Berger E, Assimes TL, Ferrucci L, Horvath S (2019a) DNA meth-
Vineis P, Kumari M, Kelly-Irving M, Chadeau-Hyam ylation GrimAge strongly predicts lifespan and health-
M (2019) Early-life inequalities and biological ageing: span. Aging (albany. NY) 11:303–327
a multisystem Biological Health Score approach in U Lu AT, Seeboth A, Tsai PC, Sun D, Quach A, Reiner AP, Koo-
nderstanding S ociety. J Epidemiol Community Health perberg C, Ferrucci L, Hou L, Baccarelli AA, Li Y, Har-
73:693–702 ris SE, Corley J, Taylor A, Deary IJ, Stewart JD, Whit-
Knight R, Callewaert C, Marotz C, Hyde ER, Debelius JW, sel EA, Assimes TL, Chen W, Li S, Mangino M, Bell
McDonald D, Sogin ML (2017) The microbiome and JT, Wilson JG, Aviv A, Marioni RE, Raj K, Horvath S
human biology. Annu Rev Genomics Hum Genet (2019b) DNA methylation-based estimator of telomere
18:65–86 length. Aging (albany. NY) 11:5895–5923

Vol.: (0123456789)
13
 864 Biogerontology (2023) 24:845–866

Lulkiewicz M, Bajsert J, Kopczynski P, Barczak W, Rubis B Schramm K, Wilson YA, Kobes S, Tukiainen T, Ramos
(2020) Telomere length: how the length makes a differ- YF, Göring HHH, Fornage M, Liu Y, Gharib SA,
ence. Mol Biol Rep 47:7181–7188 Stranger BE, De Jager PL, Aviv A, Levy D, Murabito
Maddock J, Castillo-Fernandez J, Wong A, Cooper R, Richards JM, Munson PJ, Huan T, Hofman A, Uitterlinden AG,
M, Ong KK, Ploubidis GB, Goodman A, Kuh D, Bell Rivadeneira F, Van Rooij J, Stolk L, Broer L, Verbiest
JT, Hardy R (2020) DNA methylation age and physical MMPJ, Jhamai M, Arp P, Metspalu A, Tserel L, Milani
and cognitive aging. J Gerontol—Ser A Biol Sci Med Sci L, Samani NJ, Peterson P, Kasela S, Codd V, Peters A,
75:504–511 Ward-Caviness CK, Herder C, Waldenberger M, Roden
McCrory C, Fiorito G, McLoughlin S, Polidoro S, Cheallaigh M, Singmann P, Zeilinger S, Illig T, Homuth G, Grabe
CN, Bourke N, Karisola P, Alenius H, Vineis P, Layte R, HJ, Völzke H, Steil L, Kocher T, Murray A, Melzer D,
Kenny RA (2020) Epigenetic clocks and allostatic load Yaghootkar H, Bandinelli S, Moses EK, Kent JW, Cur-
reveal potential sex-specific drivers of biological aging. ran JE, Johnson MP, Williams-Blangero S, Westra HJ,
Journals Gerontol. - Ser. A Biol Sci Med Sci 75:495–503 McRae AF, Smith JA, Kardia SLR, Hovatta I, Perola M,
Melzer D, Pilling LC, Ferrucci L (2020) The genetics of Ripatti S, Salomaa V, Henders AK, Martin NG, Smith
human ageing. Nat Rev Genet 21:88–101 AK, Mehta D, Binder EB, Nylocks KM, Kennedy EM,
Meyer DH, Schumacher B (2021) BiT age: a transcriptome- Klengel T, Ding J, Suchy-Dicey AM, Enquobahrie DA,
based aging clock near the theoretical limit of accuracy. Brody J, Rotter JI, Chen YDI, Houwing-Duistermaat
Aging Cell 20:e13320 J, Kloppenburg M, Slagboom PE, Helmer Q, Den Hol-
Mitnitski AB, Mogilner AJ, Rockwood K (2001) Accumulation lander W, Bean S, Raj T, Bakhshi N, Wang QP, Oyston
of deficits as a proxy measure of aging. ScientificWorld- LJ, Psaty BM, Tracy RP, Montgomery GW, Turner ST,
Journal 1:321027 Blangero J, Meulenbelt I, Ressler KJ, Yang J, Franke L,
Moreno-Agostino D, Daskalopoulou C, Wu YT, Koukounari Kettunen J, Visscher PM, Neely GG, Korstanje R, Han-
A, Haro JM, Tyrovolas S, Panagiotakos DB, Prince M, son RL, Prokisch H, Ferrucci L, Esko T, Teumer A, Van
Prina AM (2020) The impact of physical activity on Meurs JBJ, Johnson AD, Nalls MA, Hernandez DG,
healthy ageing trajectories: evidence from eight cohort Cookson MR, Gibbs RJ, Hardy J, Ramasamy A, Zonder-
studies. Int J Behav Nutr Phys Act 17:1–12 man AB, Dillman A, Traynor B, Smith C, Longo DL,
Mount S, Ferrucci L, Wesselius A, Zeegers MP, Schols AMWJ Trabzuni D, Troncoso J, Van Der Brug M, Weale ME,
(2019) Measuring successful aging an exploratory factor O’Brien R, Johnson R, Walker R, Zielke RH, Arepalli S,
analysis of the InCHIANTI Study into different health Ryten M, Singleton AB (2015) The transcriptional land-
domains. Aging (albany. NY) 11:3023–3040 scape of age in human peripheral blood. Nat Commun
Nelson PG, Promislow DEL, Masel J (2020) Biomarkers 6:8570
for aging identified in cross-sectional studies tend to Protsenko E, Yang R, Nier B, Reus V, Hammamieh R, Ramp-
be non-causative. J Gerontol—Ser A Biol Sci Med Sci ersaud R, Wu GWY, Hough CM, Epel E, Prather AA,
75:466–472 Jett M, Gautam A, Mellon SH, Wolkowitz OM (2021)
Newman AB, Boudreau RM, Naydeck BL, Fried LF, Harris “GrimAge”, an epigenetic predictor of mortality, is
TB (2008) A physiologic index of comorbidity: relation- accelerated in major depressive disorder. Transl Psychia-
ship to mortality and disability. J Gerontol A Biol Sci try 11:193
Med Sci 63:603–609 Raj K, Horvath S (2020) Current perspectives on the cellular
Newman AB (2015) Is the onset of obesity the same as aging? and molecular features of epigenetic ageing. Exp Biol
Proc Natl Acad Sci USA 112:E7163 Med 245:1532–1542
Nguyen H, Moreno-Agostino D, Chua KC, Vitoratou S, Rattan SIS (2012) Biogerontology: From here to where?
Matthew Prin A (2021) Trajectories of healthy age- The Lord Cohen Medal Lecture-2011. Biogerontology
ing among older adults with multimorbidity: A growth 13:83–91
mixture model using harmonised data from eight ATH- Rattan SIS (2013) Healthy ageing, but what is health?
LOS cohorts. PLoS ONE 16:1–15 Biogerontology 14:673–677
Nie P, Li Y, Zhang N, Sun X, Xin B, Wang Y (2021) The Rattan SIS (2018) Biogerontology: research status, challenges
change and correlates of healthy ageing among Chinese and opportunities. Acta Biomed 89:291–301
older adults: findings from the China health and retire- Rattan SIS (2020) Homeostasis, Homeodynamics and Aging.
ment longitudinal study. BMC Geriatr 21:1–13 In: Rattan, S.I.S.-E. of B.G. (Ed.), . Academic Press,
O’Connell MDL, Marron MM, Boudreau RM, Canney M, Oxford, pp. 238–241.
Sanders JL, Kenny RA, Kritchevsky SB, Harris TB, Rivadeneira MF, Mendieta MJ, Villavicencio J, Caicedo-Gal-
Newman AB (2019) Mortality in relation to changes lardo J, Buendía P (2021) A multidimensional model of
in a healthy aging index: the health, aging, and body healthy ageing: proposal and evaluation of determinants
composition study. J Gerontol A Biol Sci Med Sci based on a population survey in Ecuador. BMC Geriatr
74:726–732 21:1–11
Oblak L, van der Zaag J, Higgins-Chen AT, Levine ME, Boks Rivero-Segura NA, Bello-Chavolla OY, Barrera-Vázquez OS,
MP (2021) A systematic review of biological, social and Gutierrez-Robledo LM, Gomez-Verjan JC (2020) Prom-
environmental factors associated with epigenetic clock ising biomarkers of human aging: In search of a multi-
acceleration. Ageing Res Rev 69:101348 omics panel to understand the aging process from a mul-
Peters MJ, Joehanes R, Pilling LC, Schurmann C, Conneely tidimensional perspective. Ageing Res Rev 64:101164
KN, Powell J, Reinmaa E, Sutphin GL, Zhernakova A,

Vol:. (1234567890)
13
Biogerontology (2023) 24:845–866 865

Robinson O, Chadeau Hyam M, Karaman I, Climaco Pinto R, discloses distinct inter-personal variability in epigenetic
Ala-Korpela M, Handakas E, Fiorito G, Gao H, Heard A, aging patterns. Epigenetics Chromatin 15:9
Jarvelin M-R, Lewis M, Pazoki R, Polidoro S, Tzoulaki Simpson DJ, Chandra T (2021) Epigenetic age prediction.
I, Wielscher M, Elliott P, Vineis P (2020) Determinants Aging Cell 20:e13452–e13452
of accelerated metabolomic and epigenetic aging in a UK Singh H, Torralba MG, Moncera KJ, DiLello L, Petrini J,
cohort. Aging Cell 19:e13149 Nelson KE, Pieper R (2019a) Gastro-intestinal and oral
Rodriguez-Laso A, McLaughlin SJ, Urdaneta E, Yanguas J microbiome signatures associated with healthy aging.
(2018) Defining and estimating healthy aging in Spain: a GeroScience 41:907–921
cross-sectional Study. Gerontologist 58:388–398 Singh PP, Demmitt BA, Nath RD, Brunet A (2019b) The genet-
Rowe JW, Kahn RL (1997) Successful Aging. Gerontologist ics of aging: a vertebrate perspective. Cell 177:200–220
37:433–440 Speiser JL, Callahan KE, Houston DK, Fanning J, Gill TM,
Ruiz-Ruiz S, Sanchez-Carrillo S, Ciordia S, Mena MC, Mén- Guralnik JM, Newman AB, Pahor M, Rejeski WJ, Miller
dez-García C, Rojo D, Bargiela R, Zubeldia-Varela E, ME (2021) Machine learning in aging: an example of
Martínez-Martínez M, Barbas C, Ferrer M, Moya A developing prediction models for serious fall injury in
(2020) Functional microbiome deficits associated with older adults. J Gerontol Ser A 76:647–654
ageing: Chronological age threshold. Aging Cell 19:1–11 Stanziano DC, Whitehurst M, Graham P, Roos BA (2010)
Salosensaari A, Laitinen V, Havulinna AS, Meric G, Cheng S, A review of selected longitudinal studies on aging:
Perola M, Valsta L, Alfthan G, Inouye M, Watrous JD, past findings and future directions. J Am Geriatr Soc
Long T, Salido RA, Sanders K, Brennan C, Humphrey 58:292–297
GC, Sanders JG, Jain M, Jousilahti P, Salomaa V, Knight Sukul P, Grzegorzewski S, Broderius C, Trefz P, Mittlmeier
R, Lahti L, Niiranen T (2021) Taxonomic signatures of T, Fischer DC, Miekisch W, Schubert JK (2022) Physi-
cause-specific mortality risk in human gut microbiome. ological and metabolic effects of healthy female aging on
Nat Commun 12:1–8 exhaled breath biomarkers. iScience 25:103739
Sanchez-Niubo A, Forero CG, Wu YT, Giné-Vázquez I, Prina Sun ED, Qian Y, Oppong R, Butler TJ, Zhao J, Chen BH, Tan-
M, De La Fuente J, Daskalopoulou C, Critselis E, De La aka T, Kang J, Sidore C, Cucca F, Bandinelli S, Abecasis
Torre-Luque A, Panagiotakos D, Arndt H, Ayuso-Mateos GR, Gorospe M, Ferrucci L, Schlessinger D, Goldberg I,
JL, Bayes-Marin I, Bickenbach J, Bobak M, Cabal- Ding J (2021) Predicting physiological aging rates from
lero FF, Chatterji S, Egea-Cortés L, García-Esquinas E, a range of quantitative traits using machine learning.
Leonardi M, Haro JM (2021) Development of a com- Aging (albany. NY) 13:23471–23516
mon scale for measuring healthy ageing across the world: Tanaka T, Biancotto A, Moaddel R, Moore AZ, Gonzalez-
results from the ATHLOS consortium. Int J Epidemiol Freire M, Aon MA, Candia J, Zhang P, Cheung F, Fan-
50:880–892 toni G, Semba RD, Ferrucci L (2018) Plasma proteomic
Sanders JL, Minster RL, Barmada MM, Matteini AM, signature of age in healthy humans. Aging Cell 17:1–13
Boudreau RM, Christensen K, Mayeux R, Borecki IB, Tanaka T, Basisty N, Fantoni G, Candia J, Moore AZ, Bian-
Zhang Q, Perls T, Newman AB (2014) Heritability of cotto A, Schilling B, Bandinelli S, Ferrucci L (2020)
and mortality prediction with a longevity phenotype: Plasma proteomic biomarker signature of age predicts
the healthy aging index. J Gerontol A Biol Sci Med Sci health and life span. Elife 9:e61073
69:479–485 Vaiserman A, Krasnienkov D (2021) Telomere length as a
Schmauck-Medina T, Molière A, Lautrup S, Zhang J, Chlop- marker of biological age: state-of-the-art, open issues,
icki S, Madsen HB, Cao S, Soendenbroe C, Mansell E, and future perspectives. Front Genet 11:630186
Vestergaard MB, Li Z, Shiloh Y, Opresko PL, Egly J-M, van den Berg N, Beekman M, Smith KR, Janssens A, Slag-
Kirkwood T, Verdin E, Bohr VA, Cox LS, Stevnsner T, boom PE (2017) Historical demography and longevity
Rasmussen LJ, Fang EF (2022) New hallmarks of age- genetics: back to the future. Ageing Res Rev 38:28–39
ing: a 2022 Copenhagen ageing meeting summary. Aging Varzaneh ZA, Shanbehzadeh M, Kazemi-Arpanahi H (2022)
(albany. NY) 14:6829–6839 Prediction of successful aging using ensemble machine
Schmidt M, Hopp L, Arakelyan A, Kirsten H, Engel C, Wirk- learning algorithms. BMC Med Inform Decis Mak
ner K, Krohn K, Burkhardt R, Thiery J, Loeffler M, 22:258
Loeffler-Wirth H, Binder H (2020) The human blood Vaupel JW (2010) Biodemography of human ageing. Nature
transcriptome in a large population cohort and its relation 464:536–542
to aging and health. Front Big Data 3:1–22 Vetter VM, Meyer A, Karbasiyan M, Steinhagen-Thiessen E,
Searle SD, Mitnitski A, Gahbauer EA, Gill TM, Rockwood K Hopfenmüller W, Demuth I (2019) Epigenetic clock
(2008) A standard procedure for creating a frailty index. and relative telomere length represent largely different
BMC Geriatr 8:24 aspects of aging in the Berlin aging study II (BASE-II).
Seligman BJ, Berry SD, Lipsitz LA, Travison TG, Kiel Journals Gerontol. - Ser A Biol Sci Med Sci 74:27–32
DP (2022) Epigenetic age acceleration and change in Vetter VM, Kalies CH, Sommerer Y, Bertram L, Demuth I
frailty in MOBILIZE Boston. J Gerontol: Series A (2022a) Seven-CpG DNA methylation age determined by
77(9):1760–1765 single nucleotide primer extension and illumina’s infin-
Shahal T, Segev E, Konstantinovsky T, Marcus Y, Shefer G, ium methylationEPIC array provide highly comparable
Pasmanik-Chor M, Buch A, Ebenstein Y, Zimmet P, results. Front Genet 12:1–7
Stern N (2022) Deconvolution of the epigenetic age Vetter VM, Kalies CH, Sommerer Y, Spira D, Drewelies J,
Regitz-Zagrosek V, Bertram L, Gerstorf D, Demuth I

Vol.: (0123456789)
13
 866 Biogerontology (2023) 24:845–866

(2022b) Relationship between 5 epigenetic clocks, tel- Zenin A, Tsepilov Y, Sharapov S, Getmantsev E, Menshikov
omere length, and functional capacity assessed in older LI, Fedichev PO, Aulchenko Y (2019) Identification of
adults: cross-sectional and longitudinal analyses. J Ger- 12 genetic loci associated with human healthspan. Com-
ontol: Series A 77(9):1724–1733 mun Biol 2:41
Vrijheid M (2014) The exposome: a new paradigm to study the Zhang H, Zhu Y, Hao M, Wang J, Wang Z, Chu X, Bao Z,
impact of environment on health. Thorax 69:876–878 Jiang X, Shi G, Wang X (2021) The modified healthy
Wilmanski T, Diener C, Rappaport N, Patwardhan S, Wiedrick ageing index is associated with mortality and disabil-
J, Lapidus J, Earls JC, Zimmer A, Glusman G, Robin- ity: the rugao longevity and ageing study. Gerontology
son M, Yurkovich JT, Kado DM, Cauley JA, Zmuda J, 67:572–580
Lane NE, Magis AT, Lovejoy JC, Hood L, Gibbons SM, Zhavoronkov A, Mamoshina P, Vanhaelen Q, Scheibye-Knud-
Orwoll ES, Price ND (2021) Gut microbiome pattern sen M, Moskalev A, Aliper A (2019) Artificial intelli-
reflects healthy ageing and predicts survival in humans. gence for aging and longevity research: recent advances
Nat Metab 3:274–286 and perspectives. Ageing Res Rev 49:49–66
World Health Organization (2012) Good health adds life to Zimmer A, Korem Y, Rappaport N, Wilmanski T, Baloni P,
years. Policies and priority interventions for healthy age- Jade K, Robinson M, Magis AT, Lovejoy J, Gibbons SM,
ing, Copenhagen Hood L, Price ND (2021) The geometry of clinical labs
World Report on Ageing and Health, (2015). Luxembourg. and wellness states from deeply phenotyped humans. Nat
Wu C, Smit E, Sanders JL, Newman AB, Odden MC (2017) Commun 12:1–13
A modified healthy aging index and its association with Zubair N, Conomos MP, Hood L, Omenn GS, Price ND,
mortality: the national health and nutrition examina- Spring BJ, Magis AT, Lovejoy JC (2019) Genetic predis-
tion survey, 1999–2002. J Gerontol A Biol Sci Med Sci position impacts clinical changes in a lifestyle coaching
72:1437–1444 program. Sci Rep 9:6805
Wu C, Newman AB, Dong B-R, Odden MC (2018) Index
of healthy aging in chinese older adults: China health Publisher’s Note Springer Nature remains neutral with regard
and retirement longitudinal study. J Am Geriatr Soc to jurisdictional claims in published maps and institutional
66:1303–1310 affiliations.
Yu R, Thiyagarajan JA, Leung J, Lu Z, Kwok T, Woo J (2021)
Validation of the construct of intrinsic capacity in a lon-
gitudinal chinese cohort. J Nutr Heal Aging 25:808–815

Vol:. (1234567890)
13