Frailty Index versus Physical Phenotype as complementary tools

There is a dearth of mice studies where frailty was determined by either one of the two major frailty assessment tools: the Frailty Index (FI) and the Physical Phenotype (PP). There have been multiple head-to-head comparisons of the Frailty Phenotype (FP) and Frailty Index (FI) in both clinical and preclinical settings. Clinical studies have highlighted key differences between these two tools in predicting outcomes like mortality, hospitalization, and functional decline [15,16,17,18]. Preclinical comparisons using murine models have similarly demonstrated differences in frailty outcomes, particularly depending on the assessment methodology, such as physical or lab-based measures [19, 20]. Seldeen and colleagues further emphasized the biological variability influencing these measures in female mice [21], reinforcing the importance of addressing these methodological differences when comparing the FP and FI in both clinical and preclinical research.

To date, it is not clear whether each tool would classify animals similarly. Determining which are the most relevant markers of frailty to be targeted by an assessment tool is a challenge, especially considering the multisystem nature of frailty-driven physiological decline. Still, the standardization of frailty classification across studies would greatly contribute to advances in the field of aging. Hence, one of the main objectives of this study was to determine the level of frailty of female C57BL/6 by comparing for the first time how both the mouse FI and the PP would classify female mice under the three main frailty categories: Robust, Prefrail, and Frail.

We observed that the two assessment tools did not identify the female cohort similarly: the FI identified nearly 3 times more female mice as frail compared to the PP (Fig. 1a). Previous findings comparing both tools in male C57BL/6 mice showed that the Frailty Index identified 44.4% of the cohort as frail, whereas the Frailty Phenotype found no mice to be frail and a modified version of the same tool identified only 16.6% of mice as frail [19]. In clinical practice, there is substantial evidence indicating that, at all ages, females tend to exhibit higher levels of frailty compared to males [22,23,24]. Sex differences in frailty in clinics are influenced by a complex interplay of behavioral, social, and biological factors, although the precise mechanisms remain unclear [25, 26]. Preclinical studies in murine models demonstrate sex-specific variations in frailty assessments depending on the methodologies employed [21, 27]. Inflammation and immune responses, hormonal dysregulation, and genetic factors have all been proposed as mechanisms contributing to sex differences in frailty [25, 28,29,30]. Further research, particularly in preclinical models, is essential for elucidating these complex sex differences in frailty.

Fig. 1

Frailty Index versus Physical Phenotype as complementary tools. Comparison of frailty classification using Frailty Index (FI), Physical Phenotype (PP), and Vitality Phenotype (VP) tools with the entire old cohort (21–28 months) represented as parts of a whole (A). B The cohort is distinctively classified according to each assessment tool, FI, PP, and VP. Each tool is represented by a node. The initial node on the left of the diagram represents the proportion of animals classified as Non-frail or Frail when utilizing the FI tool. The central node represents the proportion of animals from the same cohort classified as Robust, Prefrail, or Frail using the PP tool, while the rightmost node represents the proportion of animals from the same cohort classified as Robust, Prefrail, or Frail when employing the VP (*). Links connecting nodes illustrate how the cohort classification is influenced by each tool. Numbers within nodes represent the sample size classified in each category, whereas numbers within connecting links indicate how many mice changed or maintained their classificatory status between tools. Age distribution is described in Table 1

Our finding suggests that these two assessment tools are rather complementary than interchangeable. This suggests that each tool may target distinct frailty markers or pathways and could potentially be more effective if integrated into a new, more comprehensive classification system.

Using a cohort of 190 female mice, the FI classified ~ 77% as Non-frail and ~ 23% as Frail, while the PP identified ~ 78% of animals as Robust (Non-frail), ~ 15% as Prefrail (Non-frail), and ~ 7% as Frail (Fig. 1a). The most primordial difference between both tools is given by the evaluation method, and the Frailty Index is mostly based on the visual inspection of the animal, whereas the PP focuses on performance more than appearance. The second most relevant difference relies on the categorization method. The Frailty Index uses a binary classification, Frail versus Non-frail (positive or negative for frailty), while the PP adds an intermediary frailty state, the Prefrail. The observed reduction of nearly 16% of animals first classified as Frail when using the FI tool and then classified as Robust or Prefrail using the PP suggests that the FI overestimates considerably the frailty status across the cohort. Furthermore, when examining the Sankey diagram, we noted that 25 out of 146 animals (17%) classified as Robust when using the FI were then classified as Prefrail (21) and Frail (4, Fig. 1b) when categorized by the PP. Together, these differences indicate that the absence of an intermediary frailty state in the classification method not only implies the overestimation of Frailty, but also Robustness. Still, it is important to emphasize that not all changes in frailty classification were destined to Prefrail when using the PP. Indeed, 20 out of 44 animals (~ 45%) classified as Frail by the FI were then classified as Robust by the PP, while only 13 were reclassified as Prefrail (Fig. 1b). Interestingly, 45% of animals that accumulated a lot of health deficits scoring poorly with FI performed well physically when examined by the PP, once again suggesting that each tool targets distinct age-related impairments and that not all physiological declines occur simultaneously. Thus, it seems that combining tools to investigate aging decline within an organism might increase the chances of identifying a risk of frailty development.

Merging Frailty Index and Physical Phenotype, the Vitality Phenotype

Considering that both PP and FI are complementary and not interchangeable tools, we decided to merge both into a more comprehensive one. Since the PP combined multiple physical evaluations, we decided to incorporate the FI (health deficits) as one of the criteria. With a new criterion added to the evaluation, the updated frailty tool will be now referred to as the Vitality Phenotype (VP).

The VP increased the Prefrail classification by nearly 10% as compared to PP due to a reduction of nearly 2% of animals classified under the Frail category and 8% from Robust (Fig. 1a, b).

Defining cutoff values for Vitality Phenotype criteria

Frailty is a state of vulnerability defined by the loss of reserves due to the accumulative decline of multiple physiological systems. In this context, the VP aims to identify physical signs of frailty by means of an observational and physical performance evaluation. Due to the heterogeneity of the frailty syndrome, the VP encompasses several criteria, each of which accounts as positive or negative according to a given cutoff value. Defining cutoff values is a great challenge not only in clinical practice, but also in animal research. Indeed, in vivo models exhibit significant variability across studies; factors such as genetic background, age, sex, type of bedding, housing conditions, dietary composition, and microbiota can all influence research outcomes [31,32,33]. This variability often complicates standardization efforts and hinders the reproducibility of findings. We thus suggest that cutoff values should ideally be defined within each research protocol until the field establishes standardization of cutoff values for mice that can be reproduced within each laboratory [9]. In longitudinal studies, tracking individual aging mice is ideal for observing the gradual transition from robustness to frailty [9]. This approach recognizes that frailty is not solely based on age but also on the decline in physiological performance. In contrast, cross-sectional studies only capture a snapshot of differences between groups at a single point in time, rather than tracking changes within the same subjects over a prolonged period. Therefore, when employing a cross-sectional or short-term intervention design, we propose that frailty cutoff values be determined using a reference group. Since the field has not yet established reproducible standard values for frailty-related declines in physical performance, the major selection factor for the reference group is chronological age. For instance, if a cutoff value is set using a very old cohort in a cross-sectional study, a disabled mouse could be incorrectly classified as robust, and conversely, a healthy mouse could be misclassified as frail. Therefore, using an appropriate age as a reference for frailty remains the best approach when working with cross-sectional or short-term studies.

Defining a reference group for identifying cutoff values for frailty

Despite the enormous advantage regarding shortened lifespans, limited genetic variability, and regulated environmental conditions, defining animal models of frailty is rather challenging, especially concerning its true potential of reverse translation from clinical practice. This is also true when choosing the research design and frailty assessment tools. Defining the physiological status/profile that best represents a threshold for a given decline is crucial to determining positive criteria in frailty. Based on a longitudinal approach, our group has previously identified that the reference group for identifying cutoff values should have (1) 100% survival, (2) correspondence to human lifespan: 23 months for a mouse falls within the ~ 65–75 range of human years [34] that corresponds to the initial age brackets assessed by Fried et al. [2], and (3) reversibility potential that would provide adequate time to implement possible life-changing interventions [9]. Herein, we propose that in a cross-sectional study format, the age-driven reference group should be set at an age in which frailty onset is more likely to occur which would still meet 100% survival as well as provide the window for therapeutic interventions, favoring Prefrail classification.

Selecting an appropriate age range where frailty is expected to emerge is crucial for accurately identifying frailty thresholds and defining cutoffs, helping to avoid over- or underestimation in animal classification. For any criterion to be marked as positive, a well-established baseline is essential. In this regard, control groups provide a vital standard for assessing experimental outcomes. While controls are widely recognized in animal models, they are often overlooked in frailty research. Including a specific control for frailty (rather than just aging) helps differentiate frailty-specific findings from those driven by aging alone.

In a longitudinal study following the prevalence of frailty across the lifespan, our team identified frailty onset in both males and females at 17 months [35, 36]. We thus selected this age to establish an age-matched control group for frailty, which would run parallel to the frail/older cohort in cross-sectional studies. Cutoffs would then be defined by the 17-month-old control group and applied to the groups of interest which would increase the number of animals classified as “Prefrail” or “Frail” consistently (Fig. 2). This age selection would also enable earlier frailty detection, facilitating the identification of prefrail animals—a critical stage for implementing potentially life-changing interventions [35]. Without setting a prefrail cutoff, frailty status may be underestimated, with both Prefrail and Frail animals potentially being misclassified as Robust. Importantly, although we posit that chronological age represents the optimal selection criterion for the reference group in a cross-sectional study, this principle does not extend to the experimental group (old mice). A wide age range was employed for the old group to facilitate a more comprehensive classification of subjects across all three frailty categories. Restricting the selection to a specific age would have significantly reduced the number of animals classified across all frailty categories, thereby compromising the validity of our frailty analysis. Notably, this strategy prioritizes the establishment of consistent Robust, Prefrail, and Frail groups over stringent age selection, as the primary focus is on frailty rather than aging per se. Thus, the experimental groups were defined based on physical performance outcomes, not chronological age, ensuring a more accurate and meaningful assessment of frailty.

Fig. 2

Impact of adding a reference group to define cutoff for frailty classification. Evaluation of the impact of using a reference group (middle age (17–19 months)) to define cutoffs on frailty status within the old cohort. Two nodes represent how the old cohort (21–28 months) is classified when cutoffs are based on two age selections: old age (21–28 months) and middle age (17–19 months*). Within each node, the proportion along with the accurate number of animals classified as Robust, Prefrail, and Frail are illustrated. Links connecting the nodes trace how the cohort classification is influenced by each cutoff. Age distribution is described in Table 2

Correlating biological with frailty markers

Numerous pathophysiological mechanisms contribute to frailty. The current clinical conceptualization of frailty focuses on three main biological systems: metabolic, stress response, and neuromuscular, which involve endocrine dysregulation, cellular senescence, and chronic inflammation [3]. The Vitality Phenotype includes various criteria that investigate physical performance and assess neuromuscular function, such as the rotarod, treadmill, and grip force test. Moreover, the health deficits criteria include observations such as alopecia, distended abdomen, and rectal prolapse, all deficits that may reflect stress (e.g., chronic inflammation) and/or metabolic dysfunction. Yet, it is not clear whether physical measurements correlate with biological markers of frailty that reflect the decline of the previously mentioned systems. We thus investigated correlations between specific frailty criteria from the Vitality Phenotype and biological markers of glucose homeostasis (e.g., area under the curve from oral glucose tolerance test, AUC oGTT, Fig. 3 and Table 3), bone mineral density (BMD, Fig. 3 and Table 3), inflammation (plasma levels of MCP-1, Fig. 3 and Table 3), and heart rate (HR, shown in Table 3). We observed that the only biological marker that correlated with two frailty criteria was BMD (Fig. 3 and Table 3). BMD inversely correlated with both grip strength and voluntary wheel activity (Fig. 3C and E, respectively, and Table 3). These data suggest that among the analyzed biological markers, BMD is likely the only marker influenced by performance. Both grip force and wheel activity involve increased weight-bearing stress on the animals’ limbs, providing a reasonable explanation as to why animals with worse performance in these criteria had lower whole BMD.

Fig. 3

Correlating biological markers with frailty criteria. Correlation between various biological markers and physical frailty criteria. In all panels, Y axes represent AUC oGTT (a.u.), color grading BMD, and circle size MCP-1 levels. Panels are differentiated by X axes that represent each physical frailty criteria: A Health Deficits, B Walking Speed, C Grip Strength, D Endurance, and E Voluntary Wheel Activity. AUC, area under the curve; BMD, bone mineral density; MCP-1, monocyte chemoattractant protein-1; oGTT, oral glucose tolerance test. Frailty criteria units are described in Table 3

Table 3 Correlation values of biological markers and frailty criteria: This table provides the correlation values (R2) and their statistical significance (P values) between various biological markers and frailty criteria. It highlights the significant relationships that can help inform frailty assessments and potential intervention strategies. The table includes correlations for AUC oGTT, BMD, MCP-1, and HR with multiple frailty indicators, providing a comprehensive view of how these biomarkers relate to frailty status. AUC, area under the curve; a.u., arbitrary units; BMD, bone mineral density; HR, heart rate; MCP-1, monocyte chemoattractant protein-1; oGTT, oral glucose tolerance testLimitations to the study

Given the multi-project nature of this data analysis, our study is subject to certain limitations. One such limitation is the broad mice aging. Although our focus is on biological rather than chronological aging, the reproducibility of our findings may be challenging, which also applies to the research design. Our considerations did not define specific information such as the number of animals from each age range and the age range itself. Another limitation of the present study is the absence of interrater reliability. Due to working with multiple projects over an extended period, the assessments were not performed by the same rater across projects, but only within each project. As we deal with numerous subjective assessments, variability in the rating range may have occurred as a result of this inconsistency.