Introduction

The concept of frailty was introduced to account for variability in the aging process. This syndrome reflects a decrease in the physiological reserve, and reduces the ability to respond to stress1. In older adults, frailty is known to be associated with an increased risk of adverse outcomes, such as falls, fractures, unplanned hospitalizations, and death2. Several frailty domains may only be slightly altered, so that early-stage frailty is not necessarily clinically visible3.

The most commonly used operational definitions of frailty are based on two different conceptual frameworks. Fried’s rules-based criteria correspond to a physical phenotype4, whereas the “Rockwood accumulative model” defines frailty as the accumulation of multiple deficits5. These instruments have been studied and validated in populations of individuals aged 65 and over. However, many other measures of frailty have been suggested. According to two consensus papers, an operational definition of frailty should include components from the nutrition, mobility, physical activity, strength, endurance, balance, cognition, senses, mood and social domains6,7. Furthermore, frailty can also be found in younger adults8,9, and we have reported on frailty profiles in community-dwelling individuals aged 50–75 (median: 61.7)10. Moreover, a recent study reported that frailty was associated with an elevated mortality rate among younger adults8. We therefore hypothesized that the factors determining the main ageing-related adverse events are already present in middle age.

Although the early detection of frailty is potentially important (since the condition might be reversible in its early stages)11,12, the prevalence of various components of frailty assessments (hereafter referred to as “frailty parameters”) among middle-aged populations and the parameters’ relationships with further adverse events have not been extensively documented in the literature. Most of the literature studies have focused on older adults or on a small number of frailty parameters, and none investigated the parameters’ prognostic value in multivariate models1,8,9,13,14,15,16,17,18,19,20,21.

The present study’s objectives were therefore to investigate the prevalence of various frailty parameters in middle-aged community-dwelling individuals (aged 50–65) and assess the parameters’ prognostic value for future adverse health events.

Methods

Design and participants

We performed cross-sectional and longitudinal analyses of community-dwelling volunteers aged 50 or over having been prospectively included in the ongoing SUCCEED cohort at an outpatient clinic in a university medical center in the greater Paris area of France22. The participants underwent a comprehensive multi-domain geriatric assessment after attending the SUCCessful ageing outpatient Department (SUCCEED) for an initial “prevention and healthy ageing” consultation; they had not been referred by a physician or for a particular health problem. In the present analysis, we selected individuals aged 50–65 and having been recruited between 2010 and 2015. In 2016/2017, the participants were contacted by phone by a geriatrician and asked to provide information on any adverse health events that had occurred since the geriatric assessment.

Measures

The data were collected prospectively. Performance tests were conducted by trained nurses, and a comprehensive geriatric assessment was performed by a geriatrician. In the present study, we selected components from the comprehensive geriatric assessment’s nutrition, mobility, physical activity, strength, endurance, balance, cognition, senses, mood, and social domains. Self-reported unintentional weight loss over the previous year (regardless of the amount lost) was recorded. Mobility was assessed via a daily pedometer count over a week (low level of physical activity: < 7500 steps per day)23, and gait speed over 10 m (slowness: < 1 m/s)24. Muscle strength was estimated by the completion time in a five-time sit-to-stand test (≥ 11.19 s)25, and maximum dominant-hand grip strength (kg) measured with a dynamometer (JAMAR, Sammons Preston, Bolingbrook, IL, USA). Weakness was defined as grip strength stratified by sex and body mass index4. The appendicular lean mass index was estimated using dual-energy X-ray absorptiometry (low muscle mass: < 7.23 kg/m2 in men, < 5.67 in women)26; sarcopenia was defined as a combination of low muscle mass with low muscle strength (< 30 kg in men and < 20 kg in women) or slowness26. The sternal push test and ankle dorsiflexion (< 20°, taken as the cut-off for clinically relevant ankylosis) were used to assess balance. The times in a 10-m walking test with concurrent motor and cognitive tasks were recorded. Overall cognitive performance was evaluated using the Mini-Mental State Examination score adjusted for age and educational level27. Episodic memory was assessed using the five-word screening test (< 10/10)28. Executive and visuospatial functions were evaluated using the seven-point clock-drawing test (CDT) (< 7/7)29, and frontal lobe functions were evaluated using the Frontal Assessment Battery (FAB) (< 16/18)30. Mood was explored using the Geriatric Depression Scale (risk of depression: ≥ 11/30, or ≥ 5/15 for the short form)31. Hearing impairment was defined as hearing aid use or a poor result in the finger rub test. Living alone was considered to be a proxy for the social domain. Comorbidities were also recorded.

To replicate the Fried phenotype (according to the Cardiovascular Health Study (CHS) criteria)4, we considered shrinking (> 4.5 kg in the last year), weakness, exhaustion (a self-reported feeling of general fatigue over the previous year), slowness, and a low level of physical activity (< 7500 steps per day, and no regular physical activity). We classified individuals according to these modified CHS criteria, the number of positive items defined the individuals as frail (≥ 3), pre-frail (1–2) or robust (none).

Outcome

Due to the small number of adverse events observed during follow-up, we considered a composite outcome variable comprising non-accidental falls (i.e. those not related to sports, DIY or domestic accidents), fractures (hip, spine, and wrist), unplanned hospital admissions, and death. This information was gathered recorded during the 2016/2017 phone interviews with a geriatrician.

Statistical analysis

Quantitative variables were quoted as the median [interquartile range (IQR)], and qualitative variables as the number (%).

The prevalence of frailty parameters was estimated for the whole population and then (if the overall prevalence was ≥ 5%) in [50–55], [56–60] and [61–65] age groups selected a priori. Potential differences between age classes were tested using a chi-squared test for trend or Cuzick’s trend test, as appropriate. Pairwise comparisons were performed using a chi-squared or Kruskal–Wallis tests; P values from multiple pairwise comparisons were corrected using the false discovery rate method32.

To investigate frailty parameters that were potentially predictive of adverse events, we used logistic regression models adjusted for the length of follow-up. Given that the number of frail individuals was small, we pooled the pre-frail and frail categories into a "non-robust" phenotype. Odds ratios (ORs) with their 95% confidence intervals (CIs) were estimated for variables with P values < 0.15. These variables were then considered in a multivariable logistic regression model. After checking compliance with the missing-at-random hypothesis by exploring the pattern of missingness, we used multiple imputations to maximize the sample size33. We assessed the model’s discrimination [area under the receiver-operating characteristic curve (AUC)] and calibration (Hosmer–Lemeshow test). A sensitivity analysis was carried out on complete cases.

We used the same method to assess the relationship between frailty parameters and non-accidental falls.

The threshold for statistical significance was set to P < 0.05. Statistical analyses were performed using STATA software (version 15.0, StataCorp, College Station, TX).

The present observational study was reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology statement34.

Ethical approval and consent to participate

All participants gave their verbal informed consent prior to inclusion in the study. In line with the French legislation on observational studies, written informed consent was not required. All procedures were performed in accordance with the relevant guidelines and regulations. This study with all procedures was approved by an independent ethics committee (Comité de Protection des Personnes Ile-de-France X, Paris, France, reference: 96-2019).

Results

Of the 625 individuals included in the SUCCEED survey, the 411 aged ≤ 65 years were analyzed. After a median follow-up of 3 years [2–5], outcomes were available for 340 individuals (82.7%) (Fig. 1).

Figure 1
figure 1

Flow chart for the SUCCEED-04 survey.

Full size image

Table 1 reports on the participants’ baseline characteristics; the median age was 59.0 [55.9–62.1] years, 71.1% were female, and 34.6% were retired. According to the modified physical frailty phenotype, 233 (66.8%) of the participants were pre-frail and 18 (5.2%) were frail.

Table 1 Baseline characteristics of the 411 community-dwelling participants.
Full size table

Prevalence of frailty parameters

Six frailty parameters were highly prevalent (> 20%) in this middle-aged population: living alone, a low level of physical activity, weakness, exhaustion, limited ankle dorsiflexion (a proxy of altered balance), and executive dysfunction (assessed by the CDT).

The number of comorbidities, the time needed to walk ten meters during a concurrent task, and the prevalence of impaired CDT and FAB performance increased significantly with age (Table 2). Gait speed and the prevalence of depressive symptoms decreased significantly with age; a trend towards a decrease was observed for the prevalence of exhaustion (P = 0.07). Employment status was associated with both exhaustion and age; hence, after stratification by employment status, exhaustion was no longer associated with age (P > 0.47). Lastly, we observed a non-significant trend towards an increase in the prevalence of non-accidental falls in the past year with age (P = 0.09) (Table 2).

Table 2 Frailty parameters within a population of community-dwellers aged 50–65, by age class (N = 411).
Full size table

In pairwise comparisons, the number of comorbidities was significantly higher in the 56–60 group (relative to the 50–55 group), and significantly lower gait speed (during normal walking and during a cognitive dual task) was observed from the age of 61 onwards (Table 2).

Association between frailty parameters and adverse events

The following adverse outcomes were reported by 68 of the 340 middle-aged subjects (20%) with follow-up data: deaths (N = 2, 0.6%), non-accidental falls (N = 47, 14.0%), fractures (N = 3, 0.9%), and unplanned hospitalizations (N = 21, 6.3%).

We found that female sex, living alone, weakness, altered balance, abnormal FAB score, and exhaustion were significant predictors of adverse events in the univariable analysis (Table 3).

Table 3 Frailty parameters associated with the occurrence of adverse events among community-dwelling individuals aged 50–65 years.
Full size table

Non-significant trends (P ≤ 0.10) were observed for a lower educational level and a non-robust modified Fried phenotype.

In a multivariable analysis with multiple imputations, female sex, living alone, exhaustion, altered balance and a FAB score < 16 were independent predictors of subsequent adverse events (Table 4).

Table 4 Multivariable logistic regression analyses of frailty parameters predicting the occurrence of adverse health events (N = 402).
Full size table

Similar results were found in the sensitivity analyses of complete cases (Supplementary Table 1). The model had good discriminative power (AUC = 0.76 [0.69–0.82]) and was well calibrated (P = 0.39). Similar results were observed when considering non-accidental falls as an outcome (Table 4 and Supplementary Table 2).

Discussion

In a population of middle-aged community-dwellers (aged 50–65 years), three of the modified CHS criteria (exhaustion, weakness, and a low level of physical activity) and three other frailty parameters (living alone, altered balance, and executive dysfunction) were already highly prevalent (> 20%). Four of these parameters (namely living alone, exhaustion, altered balance, and executive dysfunction) and female sex were independently associated with subsequent adverse events—suggesting the need to screen for frailty among middle-aged adults, with a view to including them in targeted interventional programs. Lastly, gait speed and executive function worsened significantly with age.

To the best of our knowledge, the present study is the first to have (i) estimated the prevalence of a large number of frailty parameters among middle-aged community-dwellers (aged 50–65) and (ii) evidenced associations with adverse events. Most previous studies focused on older (over-65) participants or considered a small number of criteria (either the five CHS criteria or a few specific parameters), and none investigated the parameters’ prognostic value in multivariate models8,9,13,14,15,16,17,18,19,20,21. Although the frailty parameters were associated with adverse health events, the parameters’ clinical significance among younger adults may differ from that among older adults. Indeed, the concept of frailty should be applied to younger adults with caution. Moreover, we cannot say that a change in these parameters reflects a decrease in functional reserves. However, our findings indicate that these frailty components should be assessed in middle age.

Although the observed prevalences of exhaustion, a low level of physical activity and living alone, were in line with previous reports9,20,35,36,37, the prevalence of weakness was higher than expected8,9,20. Our finding of similar prevalence of weakness and altered balance (26.7% and 26.8%, respectively) is consistent with the literature data on the link between these two frailty parameters in older women38; however, the underlying mechanisms have yet to be elucidated. To the best of our knowledge, the prevalence of balance impairment among the middle-aged adults has not previously been reported. However, a deterioration in balance among young and middle-aged individuals and then further deterioration after the age of 60 has been reported39. Although the community-dwellers included in the SUCCEED cohort were not expected to have dementia, a substantial proportion (23.2%) displayed impairments in executive function. In line with a previous report, we also observed an age-related decline in executive and visuospatial functions40 but not in global cognitive performance. Indeed, normal ageing appears to be associated with selective changes in certain specific cognitive domains coordinated by the prefrontal area (e.g. executive function), rather than with overall cognitive decline41.

In line with the literature data, we observed a gradual increase with age in reduced speed and a decrease in depressive symptoms13,35,42. Although exhaustion is one of the CHS criteria, we observed a trend towards a decrease in its prevalence with age—in line with some previous reports8,20. However, our results suggest that employment status (i.e. being retired or not) is a confounding factor in this association.

Interestingly, the five independent predictors of adverse events in middle-aged community-dwellers reported here (female sex, living alone, exhaustion, altered balance, and executive function impairment) are well-known risk factors for adverse events in persons 65 years or older14,15,16,43. In line with our results, a few studies have documented an increased risk of falls, fractures or unplanned hospitalizations associated with female sex, living alone, and altered balance in under-65 community-dwellers13,17,18. However, none of these studies reported on all these parameters in a multivariable model. The present study is the first to have observed an association between the risk of adverse events and executive dysfunction in healthy middle-aged adults. Although an earlier study showed that executive dysfunction is a likely precursor of frailty and disability in older adults, our results suggest that executive function impairment is also important in apparently healthy middle-aged individuals44.

Loneliness and social frailty are known predictors of physical frailty among individuals 65 years or older45,46. In our middle-aged population, participants living alone (a proxy of social frailty) were more likely to experience an adverse event than participants not living alone. Thus, these individuals may be potential target for frailty prevention.

Very few studies have investigated the relationships between the Fried phenotype and adverse events in healthy middle-aged populations8. The small number of participants with a frail physical phenotype (according to the modified CHS criteria) prevented us from assessing whether the modified phenotype predicted adverse events. However, exhaustion was the only modified CHS criterion that predicted adverse events in our study. Our results therefore suggest that although it might be useful to detect frailty parameters among middle-aged individuals, focusing solely on the physical frailty criteria may not be advisable. Hence, we recommend screening for alterations in balance, executive functions and exhaustion in middle age—especially among people living alone and women. Once our present findings have been validated in another setting, a screening tool that takes account of the impaired domains could be developed.

Clinical implications

Although the effects were unclear and inconsistent, several studies have shown that interventions based on cognitive training and nutritional interventions or physical exercise are effective in preventing pre-frailty or slowing frailty progression in older adults11,12. Given that the middle-aged individuals studied here already had a high prevalence of frailty parameters and that some of the latter were associated with adverse events, our results emphasize the need for targeted actions in a younger-than-usual age group.

Strengths and weaknesses

We prospectively recorded frailty parameters from almost all domains of geriatric assessment using validated scales.

The recruitment of participants wishing to benefit from a comprehensive geriatric assessment at a single center may limit the external validity of our results. However, since the prevalences of many of the frailty parameters observed here were similar to those reported previously, major selection bias is unlikely to have been present9,13,20,35,36,37. Furthermore, the use of a multiple imputation procedure enabled us to consider participants who were lost to follow-up and thus to limit potential selection bias. The fact that a sensitivity analysis of complete cases produced similar results suggests that our findings are robust.

We compared prevalence rates across age groups, rather than changes over time in frailty parameters among the same population of individuals; hence, we cannot rule out a potential generational effect. The recording of adverse events through phone interviews constitutes a further study limitation.

Conclusion

In a population of middle-aged community-dwellers aged 50–65, we observed a high prevalence of frailty parameters known to be associated with adverse events in older people. Furthermore, we found that parameters from various frailty domains (living alone, exhaustion, balance impairment, and executive dysfunction) and female sex were independent predictors of adverse events. Our findings highlight a possible need for frailty screening and preventive programs that target middle-aged adults.