How human and animal frailty index studies can inform each other: a brief narrative review Authors: Andrew Rutenberg, Kenneth Rockwood Published: 2026 ABSTRACT AbstractFrailty is common in older people, who are much more often studied than older animals. This makes it less surprising that research in frailty largely began with insights from humans before being translated into frailty in animals. Back translation from animal models to humans followed. Here, we review this interplay. We highlight how properties of the deficit accumulation approach to frailty (our focus) have been reproduced in animals. These include the distribution of frailty scores, the regularity of their change, and the presence of a submaximal limit that appears to reflect the mortality risk associated with significant health deficit accumulation. We offer examples of how animal models have helped us to understand age-related health changes. These changes reflect an increase in health deficits but also a reduction in the ability to withstand lesser potentially damaging stress (robustness) or to repair such damage when it occurs (resilience). Investigations into how age and sex affect the dynamics of deficit accumulation, in both human and animal models, are underway. Studying frailty in preclinical models will benefit our understanding of health dynamics and health, both for preclinical models and for us. Introduction In this brief narrative review, we consider how studies using the deficit accumulation approach to frailty can inform our understanding of how aging can be quantified in ways that can benefit health, especially human health. We begin by considering how a frailty index based on deficit accumulation arose and then survey some potential properties and uses. Next, we consider how some features of human aging were translated into animal models. We make special note of the FI-Lab (a frailty index based on laboratory test data) as an instance of translation from animals back to humans. Finally, we move to consideration of how human frailty studies can influence what is studied in animal databases. The frailty index (FI) in humans The motivation to consider how health deficits accumulate and give rise to frailty was to provide a tractable summary measure of individual health. It viewed frailty as a graded, age-associated, multiply determined, and dynamic state that contributed to risk through loss of a (then) poorly defined “physiological reserve.”1 Having tried other approaches,2,3 it was clear that any new measure needed to be: easy to calculate, robust to missing data, relevant to patient experience, and informative when tracked over time. A 2001 paper introduced the notion of a frailty index (FI) based on the accumulation of deficits as a proxy measure of aging.4 This approach to frailty thus began with human studies, where the best data lay. The FI was defined as the proportion of health deficits present in a given individual in relation to the number of health deficits considered in epidemiological or clinical databases.4 Such databases included population-based studies4–8 or clinical samples.6,9,10 Our essential hypothesis when combining multiple types of health challenges into a single measure is that functional measures of health are strongly but not completely correlated. This means that including as many items as possible in a health measure should be more sensitive in capturing underlying latent health changes—while also more broadly capturing changes in patients’ experience and quality of life. Weighted schemes were considered, but they did not improve the FI construct and impaired generalizability. Indeed, PCA analysis11 shows that weights are not needed due to the strong correlations between deficits. What makes an FI that combines these items useful is that doing so efficiently quantifies the risk of accumulating further deficits or of dying12 and offers insights into how resilience might be conceptualized.13,14 The initial FI featured more than 90 items.4 It was drawn from the variables in the Canadian Study of Health and Aging, over 3 waves from 1991 to 2001.15 In the days of handwritten notes in paper charts, this was too onerous for clinical practice. Even so, health surveys often considered large numbers of health features, as did the main tool in the geriatricians’ toolkit: the Comprehensive Geriatric Assessment (CGA).9 The CGA was meant to supplant the notion of single diagnoses, which are not common in older people who live with frailty. Such patients’ needs extend to impaired function, cognition, mobility and balance, and social interactions in ways that, though reflecting often treatable impairments, fall outside traditional medical assessments. Because the FI economically summarized such complicated records, variants of the FI emerged, including one based largely on self-report data16 and another on CGA data.9 The FI efficiently quantifies the risk of accumulating further deficits or death.12 There is an empirical maximal FI score that is well below the theoretical limit of 1.17 This appears to be an effect of censoring due to mortality, since high FI drives high hazard.18 Indeed, in clinical practice many individuals display a clinical FI close to 0.85 (nearly total functional impairment) as they approach death—but this is rarely caught in large cross-sectional or longitudinal studies. For most population-based FIs, the accumulation of deficits is approximately exponential with age. This exponential increase was readily incorporated into the “damage-accumulation” paradigm of aging. Dynamical modeling of the FI, including both damage and repair mechanisms, suggests that deficits enhance damage rates and suppress repair rates—supporting damage accumulation13 but also offering insights into how resilience might be conceptualized.13,14 Both damage and repair also depend on age—that is, on subclinical health effects.18 The simplicity and flexibility of the FI as a tool for health assessment and comparison support its utility in clinical, public health, and health policy settings.12,14,19,20 The FI can be applied to a range of datasets21 to better understand how aging unfolds in different settings.22 By using a common measure, such studies have highlighted the relevance of different groups within one country, different care settings (community, care homes, and hospitals), and how all this might vary over time. We see similar utility in animal applications of the FI. Animal FI The FI was developed for use in humans in 2001. A decade later, it was successfully translated from human data to animal models23 and used data from terminal experiments. Soon thereafter, an animal FI with non-invasive health measures24 was developed, based on the (human) Comprehensive Geriatric Assessment.9 Several features of the 2 early animal FIs23,24 were found to mimic what was seen in human work, as did the earliest reports from a dog FI.25,26 A higher proportion of deficits indicated a higher risk of further adverse outcomes—such as cardiac myocyte hypertrophy and reduced contractility. Accordingly, the proportion of deficits on average increases with age.27,28 These findings are seen even when only a few measures are used to construct a frailty index, but with an important degree of variability between measures.29 Although important sex differences have been noted,30,31 many reports, including pharmacological studies on interventions to extend lifespan and health span, still use only male animals or do not report on sex differences.32 Papers reporting an animal frailty index made up of more than 2 dozen deficits typically report a submaximal limit. This was seen in human FIs that corresponded to an FI-CGA, with similar rates of deficit accumulation24,33 and a higher mortality risk.33 This was reproduced for deficit accumulation FI in rats.34 The approach is reliable given appropriate training and review,35–37 and is even susceptible to automation.38 Usefully, it offers insights that appear to be translatable to humans.32,39,40 What does this tell us about the construction and use of animal FI? Large discrete deficit accumulation is likely to be strongly correlated, similar to those captured in self-report and clinical FIs in humans. These functional deficits will drive mortality risk. Obvious differences between human and animal frailty indices are in time scale/lifespan and in ethical constraints. Further, we less often proactively prevent, manage, or repair health deficits in animals. As a result, we expect that the balance between damage and repair will be different in animal models. To use the terminology proposed by Ukraintseva and colleagues,41 robustness (the ability to resist a stressor deficit) and resilience (the ability to repair it) should also differ.13 Still, both animal and human FIs share common features. Variables chosen as potential deficits should increase with age and be associated with adverse outcomes, although the latter can be harder to establish. For example, in mice, grey fur is treated as a deficit.24 Hair greying can arise as a “natural” feature of aging and can also be induced by environmental toxicants,42 and several features consistent with oxidative stress in grey versus pigmented hair bulbs have been reported, sometimes within the same individual.43 Even so, it remains uncertain how well grey hair reflects more widespread phenotypic aging in humans. A 2024 review noted that although grey hair is characteristic of the rare Hutchinson–Gilford progeria syndrome, mechanistic work on premature greying has largely focused on the follicle pigmentation unit and that “salt and pepper” greying complicates efforts to disentangle genetic and environmental contributions.44 For these reasons, grey hair is not included as a deficit in human FIs. The FI-lab The FI-lab is an FI-like tool to capture aging health separately from functional impairments. It is assembled from clinical lab tests such as blood tests. It thresholds these tests into healthy/not binary deficits and assembles the FI-lab from these deficits. The first animal FI used several lab tests, but these (at least in 2012) were typically terminal experiments.23 They motivated a lab-based FI in humans that could be easily repeated in longitudinal studies. The first FI-lab report in humans came in 201445 and has been used with increasing frequency.46,47 In human studies, we have strong hints that FI-Lab deficits appear to precede clinically detectable/macroscopic deficits but do not increase as non-linearly as the FI. In a 2017 report, data from the National Health and Nutrition Examination Survey (NHANES), for example, mean FI-Lab scores were consistently higher than mean FI-self-report scores until about age 58, after which they were typically lower.48 The FI-lab is correlated with the FI but is not the same. In age- and sex-adjusted Cox proportional hazards regression analysis of mortality risk, both the FI-Lab and FI-Self-report remained individually significant in a combined model.48 A subsequent study, using data from the first wave of the Canadian Longitudinal Study of Aging, employed a 23-item blood-based laboratory FI-lab, a 47-item clinical examination FI (FI-Clin), and a 48-item self-report index (FI-self-report) that included information on co-morbidities and disabilities.49 All 3 FIs increased nonlinearly with age, and the FI-Lab was correlated with each of the other two at r = 0.33. All had similar Areas Under the Receiver Operating Characteristic Curve (AUROC) in relation to mortality of about 0.79; the 95% confidence intervals overall ranged from 0.767 to 0.802. For the 3 combined, the AUC was 0.805 (0.789–0.822) in an age- and sex-adjusted model. A detailed dynamical analysis of lab blood tests shows much weaker correlations between individual variables than between functional deficits and stable linear drifts of values with age rather than nonlinear accumulation.50 Observed nonlinear accumulation of FI-lab arises because of the thresholding (binarization) of blood-test values and approximately corresponds to a thresholding of risk from lab values.50 Accordingly, our current understanding is that biological age and biological aging rate drive preclinical changes in physiology. As these changes accumulate and contribute significant risk to functional deficit accumulation, they can be captured by the FI-lab. Increases in the FI-lab capture risks and drive subsequent changes to functional FIs (such as FI-clin or FI-self-report). In turn, we expect that functional deficits, or chronic disease, would modulate the rate of aging and, so, the accumulation of FI-lab. The details of these connections between FI-lab and FI-clinical scores remain to be elucidated. Nevertheless, we believe that to understand mortality and aging, we ideally need both clinical FIs and FI-lab. Without FI-Clin data, we are further away from understanding mortality risk. Without FI-lab, we are left with only chronological age (or various biological clocks) as a proxy for the effects of cellular health on individual deficit accumulation. For now, it seems reasonable to understand the FI-Lab as a marker of physiological reserve. Discussion The clinical literature has seen some push away from “frailty” and instead toward “resilience.” The case often appears to be made chiefly on aesthetic grounds: many older adults do not like “frailty,” and emphasizing deficits appears pessimistic.51 Nevertheless, many geriatricians might find it odd to dismiss frailty, since frailty matters most to older adults who come to the Emergency Department. They are more inclined to prioritize functional abilities (captured by frailty), whereas their fitter peers give more priority to more straightforward medical issues.52 Further, in an Emergency Department study that screened for frailty and prioritized people who live with it, wait times and length of stay were both shortened. Despite the many definitions of “resilience,” there is much to be learned too about it from quantitative studies.53 As above, we see merit in distinguishing between resisting potentially damaging stressors (“robustness”) and repairing them after damage occurs (“resilience”).41 Recent animal research suggests that measures of frailty and resilience are not well correlated with each other and show important sex differences.54 Research should be carried out on both resilience (together with robustness) and frailty, since one is not simply the opposite of the other. Resilience touches on items in the FI (eg, exercise) that have been considered (when absent) as a potential functional deficit rather than a protective behavior. Examining all such items, from both the FI and from, for example, intrinsic capacity deficits,55,56 seems to be a productive way forward that could complement the FI. Specific issues to consider include noting degrees or grades of an attribute (eg, using quantiles versus binary deficits57). Many social factors (eg, income and rural-dwelling) and physical factors (eg, sleep, diet, exercise, smoking, and the like) have been evaluated individually as “protective factors.”58 Overall, a range of influencing factors contribute to the trajectories of frailty.59,60 While some items associated with protection have been combined in a protection index, the information gained thus far has been variable.61,62 More work should be done. In this light, recent emphasis on social factors in animal studies of frailty is welcome.63,64 The FI-Lab can also be useful to bridge questions in translational geroscience. For example, in a randomized phase-2 clinical trial of autologous mesenchymal stem cell (MSC) transplantation for “aging frailty,”65 intravenous laromestrocel improved the degree of age-related clinical frailty (measured using the Clinical Frailty Scale (CFS))66 after 9 months compared with placebo. The trial, in 143 patients aged 70–85 years, with initial CFS scores of 5 (“mildly frail”) or 6 (“moderately frail”) with single laromestrocel infusions. By 6 months, improvement to being either fit or managing well was seen in 14.5% of placebo-treated subjects (95% CI: 4, 19%-33.73%) compared with 30.8% of treated subjects (22.3%-40.5%). In people still meeting frailty criteria, significant increases were also seen in walking distance and in peripheral circulation.65 One biomarker, soluble TIE2 (a tyrosine kinase receptor), indicated that the MSCs impacted small vessel vasculature: a positive treatment effect was indicated by lower blood levels of sTIE2 by the sixth month in a dose-response fashion. Exploring the effects of laromestrocel and sTIE2 in greater detail in preclinical animal models is warranted, and the FI-lab is a convenient tool in such studies. That the clinical FI as a useful summary measure of functional health in humans and animal models is clear. The convenience of the FI-lab is also apparent, as is its association with health outcomes. The relationship between FI-lab and FI-clin—particularly how they influence each other—remains to be established. It likely depends on how the FIs are assembled. Exploring these connections in animal models is natural, since controlled intervention studies are simpler and faster in smaller, short-lived organisms within a lab environment.