Vigor to Frailty As a Continuum—A New Approach in the Study of Muscle, Mobility, and Aging Cohort

Abstract

Background

Frailty can occur in older adults without disability or multimorbidity. Current methods focus on the most frail, but poorly discriminate among those “not frail.”

Methods

The Study of Muscle, Mobility, and Aging (SOMMA) included 879 adults aged 70 years and older without mobility disability. We operationalized frailty domains using: peak oxygen consumption (endurance), digit symbol substitution test (speed), leg power (strength), perceived fatigability, D3 creatine dilution (sarcopenia), and accelerometry (sedentary behavior) to construct a frailty score of 0–12 summing tertiles (0–2) of each component. We used linear or logistic regression with and without adjustment for confounders to examine associations with age, reported, and performance function.

Results

The SOMMA frailty score distribution was broad and strongly associated with age (r = 0.33, p < .0001). Each point was associated with a 30%–50% higher odds of having reported difficulty with activities of daily living or mobility. After grouping the total score (0–3, 4–7, and 8–12) those in the highest group were 9–31 times more likely to have functional limitation, and at least 8 times more likely to have poorer function after full adjustment. Higher scores identified those less likely to report ease of walking or higher physical activity. Peak oxygen consumption, leg power, fatigability, and digit symbol score contributed most to these associations.

Conclusions

The SOMMA frailty score characterizes frailty as a continuum from frail to vigorous with assessments that are amenable to change. Associations with age and function suggest utility for distinguishing a wide range of vigor and vulnerability in relatively well-functioning older adults.

Graphical Abstract

Graphical Abstract.

The literature on frailty in old age is extensive, and multiple approaches have been used for its assessment that variably include diagnoses, physical function, and signs and symptoms of diseases of aging. Conceptually, frailty is thought to represent an age-related deterioration in the functioning of the person that manifests as vulnerability to stressors and loss of resilience (1). Approaches to its assessment can generally be grouped into 2 categories, one defining frailty phenotypically using 3 or more of 5 clinical signs, or as a risk algorithm assessing the accumulation of deficits (2). Recent work in progress is evaluating tests of physiologic function that might better capture vulnerability and predict resilience to a future stressor such as surgery or a fall (3,4). This approach builds on the idea that multiple stress response systems may be impaired, and their sum and their interactions may best describe the clinical phenomenon of frailty (5).

Many previous frailty assessments discriminate among the more impaired but not among more functional older adults. Frailty is related to disability and to multimorbidity, though it can be present in older adults without disability or multimorbidity (6). It is in these individuals that frailty is most challenging to identify. Furthermore, the physiology underlying frailty falls on a continuum and failure to recognize differences among better-functioning people misses important opportunities for better science and intervention development and testing.

In the Study of Muscle, Mobility, and Aging (SOMMA), we conducted several tests in the domains of the frailty phenotype that are more demanding or specific than the components used to operationalize the original frailty phenotype and developed a new phenotype with these measures. Our approach raises the ceiling and distinguishes a continuum of frailty from vigorous to frail. Additionally, we wanted to distinguish speed from endurance by including a measure of processing speed (digit symbol substitution test) and peak oxygen consumption during exercise, a direct measure of endurance. Although longer-term validation will be needed, we report here our approach and associations of this novel score with age and with reported and performance measures of physical functioning, hypothesizing that cross-sectional associations would be apparent. Since such assessments are labor intensive, we also evaluated whether a more parsimonious score would have similar associations as well as the full score.

Methods

SOMMA is a longitudinal cohort study of 879 men and women aged 70 and older recruited between April 2019 and December 2021 at 2 clinical sites at the University of Pittsburgh and Wake Forest University. Eligibility criteria also included: no major mobility disability, defined as being able to walk 400 m without or without a straight cane but no assistance of a quad cane, walker, or person at baseline, free of any life-threatening disease, no reported neurodegenerative disease or dementia, ability to provide informed consent, and with no contraindications to muscle tissue sampling (no chronic anticoagulation), magnetic resonance (MR) scanning, or cardiopulmonary exercise testing (CPET) and no plans to leave the area during the study. The detailed baseline examination required 3–4 clinic visits for a period of 1–2 months to complete all tests. Of the 879 enrolled, all had at least 1 of the major components (400-m walk plus at least MR, CPET, or muscle tissue collection) (7). Sample size was chosen to have 80% power to detect differences in the rate of major mobility disability after 5 years of follow-up. Annual examination and semiannual phone follow-up are underway. The protocol was reviewed and approved by the WIRB-Copernicus Group (WCG Institutional Review Board) and all participants gave informed consent.

Frailty Score Components

Physical endurance was quantified by peak oxygen consumption. Cardiopulmonary exercise testing was performed on a treadmill and included 3 phases of testing (phase 1, preferred walking speed; phase 2, symptom-limited peak; and phase 3, fixed walking speed). Blood pressure and heart rate were continually assessed during each phase to monitor safety and oxygen consumption gathered throughout the test. Oxygen consumption at peak (VO₂ peak) was defined as the highest volume of oxygen consumption averaged over 30-second intervals during phase 2 (peak) testing using a modified Balke protocol (8). VO₂ peak was calculated using BreezeSuite metabolic cart software (MGC Diagnostics, St. Paul, MN) and adjusted for participant weight (9). Sarcopenia was defined as skeletal muscle mass assessed using the 24-hour D₃ creatine dilution method, and participants were classified using the ratio of D₃ creatine muscle mass in kg/weight in kg (10). Leg muscle power was assessed using a Keiser AIR 300 pneumatic resistance device (A420 model: Keiser Sports Health Equipment, Fresno, CA) (11). Power was calculated as the standardized peak power (watts/kg) from 40%–70% of one repetition maximum. The Pittsburgh Fatigability Scale assesses one’s perceived propensity for physical and mental fatigue for a set of tasks of increasing demand. We used the physical subscale scored 0–50 with higher levels indicating greater fatigability (12). Sedentary time was assessed using a wrist-worn ActiGraph GT9X (ActiGraph, Pensacola, FL) (13) for a total of 7 days to record the total time spent in sedentary activity, excluding time in bed (14) and summarized as minutes sedentary per 24 hours. The digit symbol substitution test was included here as a measure of processing speed (licensed from NCS Pearson, Inc., Bloomington, MN). This is a pencil and paper test that requires the participants to draw a symbol that corresponds to a digit (1–9), completing as many as possible in 120 seconds. The score is the total number correct with higher scores indicating better performance (15,16).

SOMMA Vigor to Frailty Score

Each of the 6 measures above was grouped into sex-specific tertiles of the continuous measure. Each component was weighted equally. The best tertile was assigned a score of zero and the worst a score of 2. The individual components were summed 0–12 with 0 indicating the most vigorous and 12 the most frail. Each component was weighted equally. If an individual was missing a component due to a medical exclusion or inability to perform the test (VO₂ peak, n = 39; leg power, n = 29), their performance metric was set to 0 so that they were included in the worst tertile (2) for that component. If one item was missing for logistical rather than health reasons, that item was assigned a score equal to the average of the remaining items (n = 135, 15.4%) of scores imputed. All had at least 4 components of the score.

Frailty Phenotype

The frailty phenotype was defined (17) as three or more of the following: the lowest sex-specific quintile of D3Cr muscle mass/weight; grip strength in the lowest quintile, stratified by sex and body mass index (BMI) quartile; the slowest quintile of walking speed, stratified by sex and median height; low physical activity—the lowest sex-specific quintile of energy expenditure from the Community Healthy Activities Model Program for Seniors (CHAMPS) (18); and exhaustion—answering “at least 3–4 days per week” to either of 2 questions from the Center for Epidemiological Studies Depression scale (CES-D)-10 (“I felt that everything I did was an effort” and “I could not get going”) (19). Participants with none of the components were considered not frail, 1–2 components were intermediate, and 3 or more components were frail. For those missing data on 1–2 of the components, the frailty index was extrapolated based on the mean of the nonmissing components (20).

Study Outcomes

In SOMMA, major mobility disability is the primary outcome, defined as inability to walk 400 m at the annual examination or adjudicated from medical records and self-report, with functional limitations assessed by self-report and performance as secondary outcomes. All participants were able to walk 400 m at the baseline exam, but may have had some limitations by self-report or performance testing. Some were well-functioning and physically active, and we used the upper quartile of walking performance and reported physical activity to assess the ability of the SOMMA scale to identify them. For this cross-sectional report, we examined all of these measures of function.

Functional limitations were assessed with self-report of any difficulty with one or more activities of daily living (ADLs) including transferring, bathing, eating, dressing, and toileting. Self-report of mobility limitation was defined as difficulty walking 2–3 blocks or climbing 10 steps (21). The mobility assessment tool, short form (MAT-sf) uses video vignettes to illustrate degrees of environmental walking challenges with the participant scoring each of 12 activities on a computer (22).The short physical performance battery (SPPB) was assessed at the baseline clinical exam and scored 0–12 (23). Because the SOMMA cohort was free of mobility disability at baseline, the proportions with low scores were small and the distributions of some continuous outcomes were skewed so that we dichotomized ADLs at 1 or more with difficulty, mobility limitations as any difficulty with either steps or distance, SPPB at a score ≤8 (24) and MAT-sf at a score of <60 (range 30–80) (25,26). A 400-m walk test at usual gait speed was used to assess mobility disability with performance expressed in seconds to complete the test (26). Participants were permitted to use a straight cane and to rest without sitting for up to 60 seconds; a quad cane or walker were not permitted. The upper quartile of time to complete the 400-m walk (≥429.76 seconds) was used to define those with poorer function.

High functioning was assessed by self-report. Participants were asked to rate the ease of walking 1 mile (8–12 blocks), grouped as very easy versus other responses (somewhat or not that easy, or any difficulty). Self-reported physical activity was assessed with the CHAMPS questionnaire (18). CHAMPS data were used to calculate the hours/week spent in moderate-intensity exercise-related activities (metabolic equivalents, METs ≥ 3) and the hours/week spent in vigorous-intensity exercise-related activities (METs ≥ 6). Data were categorized as the upper quartile of time spent in the activity level versus the lower 3 quartiles (hours/week ≥ 10.25 for moderate-intensity and ≥ 1.75 for vigorous intensity).

Potential Confounders

We examined age in years based on reported date of birth, self-reported sex, and race based on self-selection of current census categories and classified as non-Hispanic White versus Black, Hispanic, or other as potential confounders. Multimorbidity was classified using a modification to the Rochester Epidemiology Multimorbidity Scale (0–13) (27). The number of current prescription medications was defined using a review and inventory of prescription bottles (28). Weight was measured with a balance beam or digital scale. Height was measured using a wall-mounted stadiometer. Body mass index was calculated as weight in kg/height in m². Smoking was assessed by self-report and categorized as current or past versus never. Cognitive function was measured using the Montreal Cognitive Assessment (MoCA) for global cognition, and the Trail Making Test Part B for executive function (29,30).

Statistical Analysis

Cohort characteristics were described using means, proportions, and histograms. The associations of frailty with physical performance, functional limitations, and high functioning were examined with logistic regression models. The primary predictor examined, the SOMMA frailty score (0–12) was expressed as a continuous variable per 1 unit higher score and categories in 3 groups approximating tertiles of the score (0–3, 4–7, 8–12). Models were unadjusted and adjusted for age, sex, race, multimorbidity score, medication use, BMI, and smoking status. Results were also stratified by sex after testing for significant interactions of the frailty score and sex. To assess the contributions of the components, the component scores (0–2) were examined together with each outcome. Forward variable selection with an entrance criterion of 0.10 was used to determine if a more parsimonious model could determine outcomes as well as the 6-component score. Components that were not selected for most models were dropped to create a more parsimonious model; this parsimonious model was then reexamined for the associations of the total score with functioning.

A sensitivity analysis was performed repeating analyses in the subsets with one or no chronic health conditions and those with ≥2 chronic health conditions, to examine whether associations were primarily driven by disease prevalence. The interaction of number of chronic health conditions and frailty score was examined. Another sensitivity analysis was performed among those with data for all the frailty score components (ie, complete case analysis).

Age attenuation was assessed by examining the percent of the attenuation of the beta coefficient for age after adjustment for the frailty score for each of the independent functional status variables.

All significance levels reported were 2-sided and all analyses were conducted using SAS version 9.4 (SAS Institute Inc., Cary, NC).

Results

The SOMMA cohort (Table 1) included 879 men and women (59.2%) of mean age 76.4 years and included 13.2% Black participants with the remainder being White or other race. The majority (85%) had at least some college education. They had on average 1.5 health conditions and took about 4 prescription medications.

Table 1.

Characteristics of Study of Muscle, Mobility and Aging (SOMMA) Participants at the Baseline Exam

	Men (n = 359)	Women (n = 520)	Total (n = 879)
Demographics
Age in years, mean ± SD	76.3 ± 5.1	76.4 ± 5.0	76.4 ± 5.0
Sex, women, n (%)			520 (59.2)
Race, Black, n (%)	33 (9.2)	83 (16.0)	116 (13.2)
Education—some college or more, n (%)	312 (87.6)	429 (83.5)	741 (85.2)
Health status
Rochester Multimorbidity Index (0–13), mean ± SD	1.4 ± 1.1	1.7 ± 1.0	1.5 ± 1.1
Number of prescription medications, mean ± SD	4.2 ± 3.3	4.5 ± 3.5	4.4 ± 3.4
Body mass index, kg/m2, mean ± SD	27.8 ± 4.2	27.4 ± 4.8	27.6 ± 4.6
Past or current smoker, n (%)	160 (44.9)	221 (42.7)	381 (43.6)
Montreal Cognitive Assessment score, mean ± SD	24.56 ± 2.81	25.02 ± 2.90	24.83 ± 2.87
Trail Making Test Part B, s, mean ± SD	118.77 ± 62.80	116.03 ± 62.88	117.16 ± 62.83
Frailty score and components, mean ± SD
SOMMA frailty score (0–12), mean ± SD	6.05 ± 3.02	6.04 ± 2.96	6.05 ± 2.98
Cardiopulmonary exercise testing: VO2 peak (mL/kg/min), mean ± SD	22.27 ± 5.00	18.75 ± 4.11	20.22 ± 4.82
Sarcopenia: kg muscle D3 Creatine/kg total weight, mean ± SD	0.32 ± 0.07	0.27 ± 0.06	0.29 ± 0.07
Muscle power: leg press standardized peak power (Watts/kg total body weight), mean ± SD	5.70 ± 1.79	3.93 ± 1.19	4.66 ± 1.71
Pittsburgh Fatigability Scale, physical score (0–50), mean ± SD	14.13 ± 8.61	17.00 ± 8.53	15.83 ± 8.67
Sedentary time: accelerometry, min/24 h excluding sleep, mean ± SD	709.89 ± 97.07	652.95 ± 107.73	676.68 ± 107.11
Cognitive speed: digit symbol substitution test, # of correct boxes, mean ± SD	53.30 ± 13.79	56.74 ± 13.49	55.34 ± 13.71
Frailty phenotype
Frail n (%)	26 (7.24)	41 (7.88)	67 (7.62)
Intermediate n (%)	167 (46.52)	258 (49.62)	425 (48.35)
Not frail n (%)	166 (46.24)	221 (42.50)	387 (44.03)

The distributions of each component of the SOMMA frailty score by tertile (Table 2) show a range of performance on each measure and substantial differences between men and women, especially for VO₂ peak and muscle power.

Table 2.

Tertile Cutpoints for Components of Study of Muscle, Mobility, and Aging (SOMMA) Frailty Measured at Baseline, Where Lower or Worse Values Were Given Higher Scores

Component	Score	Men	Women
Cardiopulmonary exercise testing: VO2 peak (mL/kg/min)	2	<19.20	<16.40
	1	19.20 to <23.80	16.40 to <19.50
	0	≥ 23.80	≥ 19.50
Sarcopenia: kg muscle D3 creatine/kg total weight	2	<0.29	<0.24
	1	0.29 to <0.34	0.24 to <0.29
	0	≥0.34	≥0.29
Muscle power: leg press standardized peak power (Watts/kg total body weight)	2	<4.72	<3.36
	1	4.72 to <6.39	3.36 to <4.38
	0	≥6.39	≥4.38
Pittsburgh Fatigability Scale, physical score (0–50)	0	<9	<13
	1	9 to <17	13 to <21
	2	≥17	≥21
Sedentary time: accelerometry, min/24 h excluding sleep	0	<672.86	<603.71
	1	672.9 to <744.7	603.7 to <688.6
	2	≥744.7	≥688.6
Cognitive speed: digit symbol substitution test, # of correct boxes	2	<48	<51
	1	48 to <59	51 to <63
	0	≥59	≥63

The distribution of the SOMMA frailty score was broad and substantially associated with age (Figure 1). The mean score ranged from about 5 in the youngest group to 8 in the oldest group. The correlation between the score and age was 0.33, p < .001. For each 1 unit higher score, there was a 0.56 year (0.46, 0.67) higher age (p < .001). The distribution of the score was substantially skewed to higher scores in the age group older than 85 indicating higher frailty, whereas the 70–74-year-old group was much less likely to have higher scores.

Figure 1.

Distribution of baseline frailty scores by age group, Study of Muscle, Mobility, and Aging, Chi-square p trend < .0001.

The prevalence of each functional limitation was as follows: any ADL difficulty was 10.2%, mobility limitation was 8.5%, MAT-sf <60 was 15.9%. For performance, the SPPB ≤8 was 14.6%. The prevalence of those who found walking a mile very easy was 42.2%. The cross-sectional associations of the SOMMA frailty score with each measure of functional limitation are shown in Figure 2A. Each point in the score was associated with a 30%–50% higher likelihood of low performance. For example, the OR for difficulty with 1+ or more ADL per each 1-unit increment in the score was 1.3 (95% CI, 1.2 to 1.5). When grouping the total score into tertiles of 0–3, 4–7, and 8–12, those in the highest group were 9–31 times more likely to have lower self-reported function or physical performance. In fully adjusted models, the highest frailty group was still at least 8 times more likely to have poorer function. This was only modestly attenuated with adjustment for potential confounders. The SOMMA frailty score was also associated with all measures of high function (Figure 2B). Each point in the score was associated with a 20%–30% lower likelihood of high functioning. Those in the highest tertile group were 80%–90% less likely to have higher function. We assessed whether these associations were similar in men and women and found a few statistically significant interactions between the score and sex among the outcomes examined (Supplementary Table 1). There were significant interactions with the frailty score and sex for the high functioning outcomes of finding it very easy to walk a mile [p-interaction = .06, OR (95% CI): men, 0.75 (0.69, 0.81); women, 0.66 (0.61, 0.73)] and having more time spent in vigorous activity [p-interaction = .02, OR (95% CI): men, 0.87 (0.81, 0.93); women, 0.76 (0.71, 0.82)], with SOMMA frailty score having a stronger inverse association with high functioning in women than men.

Figure 2.

(A) SOMMA frailty score (0–12) and self-reported function and performance: odds ratios and 95% confidence intervals. Unadjusted (left) and adjusted (right) for age, sex, Rochester Multimorbidity Index, number of Rx medications, BMI, smoking, and race. (B) SOMMA frailty score (0–12) and self-reported high functioning: odds ratios and 95% confidence intervals for self-reported high functioning. Unadjusted (left) and adjusted (right) for age, sex, Rochester Multimorbidity Index, number of Rx medications, BMI, smoking, and race. SOMMA = Study of Muscle, Mobility, and Aging; BMI = body mass index.

We next examined how each component of the total score contributed to the association of each of the measures of functioning, by considering them as variables in a model for each measure of function (Table 3) with forward selection. Regardless of the functional outcome considered, measures of VO₂ peak, leg power, and perceived physical fatigability were selected for all outcomes examined, with VO₂ peak selected first for all models. The DSST was selected for 3 of the 5 functional limitation outcomes. The D₃ creatine was less consistently associated with functioning; sedentary time score was unrelated. In sex-specific analyses, D₃ creatine/weight was more important in men than in women for several outcomes. For the outcomes defining high functioning, VO₂ peak and perceived physical fatigability were selected for all models, with leg power also selected for the ease of walking a mile and sedentary time also selected for time spent in moderate activity. We subsequently constructed a frailty score excluding sedentary time and D₃ creatine/mass tertiles for the score. Associations of the 0–8 score with function were similar to the full score (Figure 3A and B).

Table 3.

Component Analysis: Odds Ratios and Order of Selection of Each SOMMA Frailty Component Subscore (0–2) With Self-Reported Function and Performance, Overall and Sex-Specific Models, Expressed Per 1 Unit Higher Score (0–2).

	Endurance: Peak VO2 Subscore (0–2)	Sarcopenia: D3Cr/weight Subscore (0–2)	Leg Power Subscore (0–2)	Pittsburgh Fatigability Scale Subscore (0–2)	Sedentary Time Subscore (0–2)	Cognitive Speed: DSST Subscore (0–2)
Outcome	OR (95% CI)	OR (95% CI)	OR (95% CI)	OR (95% CI)	OR (95% CI)	OR (95% CI)
Self-reported function
Difficulty with 1 + ADL	1.81 (1.22, 2.70) 1	0.82 (0.60, 1.14)x	1.51 (1.06, 2.15) 3	1.83 (1.26, 2.65) 2	0.99 (0.72, 1.36)x	1.33 (0.96, 1.84) 4
Mobility limitation	1.98 (1.25, 3.12) 1	1.06 (0.74, 1.51)x	1.37 (0.93, 2.03) 3	2.26 (1.46, 3.51) 2	1.01 (0.71, 1.44)x	1.32 (0.92, 1.89)x
MAT-SF < 60	2.34 (1.61, 3.39)1	1.43 (1.06, 1.92)4	2.13 (1.51, 2.99) 2	2.20 (1.57, 3.07) 3	0.80 (0.60, 1.06)x	1.10 (0.82, 1.47)x
Performance
SPPB ≤ 8/12	2.22 (1.52, 3.23) 1	1.06 (0.79, 1.42)x	1.83 (1.31, 2.54) 3	1.60 (1.16, 2.20) 4	1.01 (0.76, 1.36)x	2.00 (1.46, 2.73) 2
Q4 of 400 MW time	2.59 (1.90, 3.53) 1	1.19 (0.92, 1.53)x	1.66 (1.27, 2.18) 3	1.77 (1.35, 2.31) 2	0.88 (0.68, 1.13)x	1.53 (1.19, 1.97) 4
High-functioning
Walk a mile easily	0.50 (0.38, 0.64) 2	1.00 (0.79, 1.26)x	0.78 (0.62, 0.99) 3	0.34 (0.26, 0.43) 1	1.14 (0.91, 1.43)x	0.90 (0.72, 1.14)x
Q4 of METs ≥ 3	0.49 (0.38, 0.65) 1	0.81 (0.64, 1.03)x	1.08 (0.85, 1.37)x	0.54 (0.42, 0.68) 2	0.78 (0.62, 0.97) 3	0.88 (0.70, 1.11)x
Q4 of METs ≥ 6	0.60 (0.47, 0.76) 2	1.03 (0.83, 1.28)x	1.06 (0.85, 1.33)x	0.52 (0.42, 0.64) 1	0.85 (0.69, 1.05)x	0.98 (0.79, 1.21)x

Notes: The Odds Ratio (95% CI) Reported Are From Models With All 6 Components Allowed to Enter the Same Model Using Forward Selection, Entrance Criteria p < .10. Entry Order Given as Superscript. Bold font indicates OR (95% CI) that do not include 1. ADL = activities of daily living; CI = confidence interval; DSST = digit symbol score; MAT-SF = mobility assessment tool, short form; METs = metabolic equivalents; MW = meter walk; OR = odds ratio; Q4 = quartile 4; SPPB = short physical performance battery; SOMMA = Study of Muscle, Mobility, and Aging.

1 = First selected variable.

2 = Second selected variable.

3 = Third selected variable.

4 = Fourth selected variable.

X = Variable not selected.

Figure 3.

(A) SOMMA frailty parsimonious score (0–8) and self-reported function and performance: odds ratios and 95% confidence intervals. Unadjusted (blue) and adjusted (red) for age, sex, Rochester Multimorbidity Index, number of Rx medications, BMI, smoking, and race. (B) SOMMA frailty parsimonious score (0–8) and self-reported high functioning: odds ratios and 95% confidence intervals for self-reported high functioning. Unadjusted (blue) and adjusted (red) for age, sex, Rochester Multimorbidity Index, number of Rx medications, BMI, smoking, and race. SOMMA = Study of Muscle, Mobility, and Aging.

All associations of frailty and outcomes examined were significant when performed in the subsets who reported 0–1 chronic health conditions and ≥2 chronic health conditions, with little to no attenuation of effect size and no statistically significant interaction. There were 135 (15.4%) participants who did not have data on all 6 of the frailty score components. The component with the most missing data was sedentary time (n = 56), followed by muscle mass (D₃ creatine, n = 52), endurance (VO₂ peak, n = 20), leg power (n = 7), processing speed (Digit Symbol Substitution Test (DSST), n = 7), and physical fatigability (Pittsburgh Fatigability Scale, n = 6). Results were very similar after subsetting analyses to those with data for all components (Supplementary Table 2). Results were very similar after subsetting analyses to those with data for all components (Supplementary Table 2).

We examined the relationship of the SOMMA Vigor to Frailty score with a frailty phenotype in Supplementary Figure 1. Although scores were correlated, the range of SOMMA scores in those not considered “not frail” (no frailty components) was substantial.

Given its strong association with age, we were interested in how the score might attenuate the association of age with function. Age alone was significantly associated with most of our measures of function. For example, each 5 years of age was associated with a 70% higher likelihood of SPPB ≤8. The age association was attenuated to an OR of 1.3 with adjustment for our frailty score (49% attenuation) and was similarly attenuated in models for other outcomes, although age was not as strongly associated with self-reported function in SOMMA as with performance measures (Supplementary Table 3).

Discussion

The SOMMA cohort was selected to be free of mobility disability at baseline, yet showed a wide range of performance on tests designed to capture elements of frailty, including peak oxygen consumption, leg power, fatigability, muscle mass, objective sedentary behavior, and cognitive speed. A simple summary score of these tests was very strongly associated with age and functional limitations and inversely associated with high functioning. The components were chosen to more precisely measure many of the same domains included in the frailty phenotype of weakness, slowness, shrinking, low activity, and fatigue. We also attempted to disentangle speed from endurance using a timed cognitive test and maximal oxygen consumption during treadmill testing. The very strong associations with age and functional limitations support its potential precision for classifying a range from frailty to vigor in nondisabled older adults.

There are many frailty scales available and while all have been useful for identifying highest-risk groups, they have poor agreement (31). Potentially, the more challenging tests administered to the SOMMA cohort could be more directly assessing physiologic reserve and with greater precision for assessing the vulnerability and resilience of frailty. The components of the treadmill exercise test, the muscle power test, fatigability, which is anchored from low to higher levels of activity, and a timed cognitive test can be thought of as stress tests, thus may be more directly reflective of resilience than tests such as usual gait speed or questions about fatigue or exhaustion, which have ceiling effects.

The inclusion of a cognitive test in our score addresses the growing recognition of a central nervous system component to physical frailty (32). Some formulations of cognitive frailty are designed to include both physical and cognitive function to better capture the risk of cognitive impairment (33,34). The digit symbol substitution test used here assesses several aspects of cognition and is sensitive to executive cognitive impairment but is also very sensitive to psychomotor speed (35), which we felt was most relevant to physical frailty.

Most frailty assessments have focused on the identification of the most vulnerable, with the nonfrail group including a wide range of physiology from not frail to vigorous. Such scales have little room for improvement with intervention (36,37). The scoring used here raises the ceiling on frailty using measures that may not only provide greater precision but also provide for better detection of improvement with intervention and potential identification of those with the best function. The precision and sensitivity to change would make this assessment battery suitable as an outcome for clinical trials of geroscience-guided therapies.

The SOMMA frailty score was strongly associated with age. The score substantially attenuated the associations of chronological age with functional limitations and high functioning. The associations of the score with functional limitations and high functioning were independent of age as well as multimorbidity and number of medications. The associations remained significant with little or no attenuation of effect size in the subset with 0–1 chronic conditions or those with complete data for the frailty components. Associations were also significant in analyses subset by sex. In aggregate, these findings suggest that the SOMMA frailty score is reflective of physiologic aging and potentially the adverse effects of biologic aging on the score and its components. Further work in SOMMA will assess the association of the score with mitochondrial energetics and other markers of aging biology.

Clinicians may be concerned that these tests are not easy to do in primary care settings. This is a moot issue for older adults who are already disabled or too ill to tolerate testing. It is in those who are nondisabled that such assessments would be most useful. All SOMMA participants were able to walk 400 m without assistance other than a cane at the baseline examination, yet the range of SOMMA frailty was substantial. As with cardiovascular risk assessment (38), the benefits of testing most likely would accrue to those who appear to be at intermediate risk. This remains to be tested with additional follow-up.

We assessed which elements contributed most to the associations of the score with functional limitation and high functioning. Perhaps not surprisingly, it was the more challenging tests of endurance, power, fatigability, and cognitive speed that contributed most. Each test component is currently individually available in most health systems and could be colocated for testing older adults, such as in a geriatrics clinic.

Our cross-sectional analysis does not adequately validate our approach; we believe that longitudinal follow-up is needed to assess the association of SOMMA frailty with mortality, ADL disability, mobility disability, falls, hospitalization, and ability to recover from major health events. This follow-up is ongoing in SOMMA. It is also important to keep in mind that the SOMMA cohort was not a population-based sample; we excluded individuals on anticoagulation and who had contraindications to MR spectroscopy and imaging. Thus, the values used to define cutpoints by tertiles are specific to this population. Cutpoints used were sex-specific, so that it is not possible to directly compare the scores of men and women. However, we note that the original criteria in the Cardiovascular Health Study used sex-specific cutpoints as well (17). We found that the score was similarly associated with cross-sectional function in men and women, though somewhat stronger for some measures in women. Additional work will be conducted in SOMMA on prediction of future outcomes in men and women. External validation is needed in other populations as well, including in less selected, more representative groups.

In summary, we extend work in the field of frailty by proposing a new frailty score that addresses frailty as a continuum from vigor to frailty and includes tests that are challenging for older adults. These tests and the overall score also have the potential to change with intervention. The SOMMA frailty score was strongly associated with age and functional limitations and may reveal vulnerability and resilience to a greater extent than previous assessments of frailty.

Funding

The Study of Muscle, Mobility, and Aging is supported by funding from the National Institute on Aging (grant number AG059416). Study infrastructure support was funded in part by NIA Claude D. Pepper Older American Independence Centers at University of Pittsburgh (P30AG024827) and Wake Forest University (P30AG021332), and the Clinical and Translational Science Institutes, funded by the National Center for Advancing Translational Science, at Wake Forest University (UL1 0TR001420).

Conflict of Interest

None.