The authors have declared that no competing interests exist.
Frailty is associated with increased risk of various health conditions, disability, and death. Health behaviors are thought to be a potential target for frailty prevention, but the evidence from previous studies is based on older populations with short follow-ups, making results susceptible to reverse causation bias. We examined the associations of healthy behaviors at age 50, singly and in combination, as well as 10-year change in the number of healthy behaviors over midlife with future risk of frailty.
In this prospective cohort study of 6,357 (29.2% women; 91.7% white) participants from the British Whitehall II cohort, healthy behaviors—nonsmoking, moderate alcohol consumption, ≥2.5 hours per week of moderate to vigorous physical activity, and consumption of fruits or vegetables at least twice a day—were measured at age 50, and change in behaviors was measured between 1985 (mean age = 44.4) and 1997 (mean age = 54.8). Fried’s frailty phenotype was assessed in clinical examinations in 2002, 2007, 2012, and 2015. Participants were classified as frail if they had ≥3 of the following criteria: slow walking speed, low grip strength, weight loss, exhaustion, and low physical activity. An illness–death model accounting for both competing risk of death and interval censoring was used to examine the association between healthy behaviors and risk of frailty. Over an average follow-up of 20.4 years (standard deviation, 5.9), 445 participants developed frailty. Each healthy behavior at age 50 was associated with lower risk of incident frailty: hazard ratio (HR) after adjustment for other health behaviors and baseline characteristics 0.56 (95% confidence interval [CI] 0.44–0.71;
Our findings suggest that healthy behaviors at age 50, as well as improvements in behaviors over midlife, are associated with a lower risk of frailty later in life. Their benefit accumulates so that risk of frailty decreases with greater number of healthy behaviors. These results suggest that healthy behaviors in midlife are a good target for frailty prevention.
Severine Sabia and colleagues investigate whether healthy behaviors at midlife are associated with reduced risk of frailty at older ages.
Frailty is a clinical syndrome associated with increased risk of several adverse health outcomes, including fracture, disability, and mortality.
Health behaviors at older ages have been found to be associated with risk of frailty, but short follow-up in these studies raises the concern that the findings may reflect changes in health behaviors consequent to health-related conditions occurring in the years preceding frailty onset rather than a causal association between health behaviors and incident frailty.
Data on smoking, alcohol consumption, physical activity, and fruits and vegetables consumption at age 50 were assessed among 6,357 participants of the Whitehall II study who were followed for incident frailty over 20.4 years.
Frailty assessed at clinical examinations in 2002, 2007, 2012, and 2015 was defined as having 3 or more of the following criteria: slow walking speed, low grip strength, weight loss, exhaustion, and low physical activity.
Each healthy behavior at age 50—nonsmoking, moderate alcohol consumption, practice of physical activity at least 2.5 hours per week, and consumption of fruits and vegetables at least twice a day—was associated with lower risk of frailty onset at older ages. In addition, participants with a greater number of healthy behaviors at age 50 had lower risk of frailty later in life, with those presenting all 4 healthy behaviors being at around 70% lower risk of developing frailty than those with none of these behaviors.
Change in healthy behaviors between mean ages 44.4 and 54.8 years suggests that among individuals with no or only 1 healthy behavior, engagement in a greater number of healthy behaviors was associated with a reduced subsequent risk of frailty.
Our findings suggest that health behaviors at age 50 are important determinants of frailty at older ages. We also found a dose–response association, such that the benefits for frailty are higher among those with a greater number of healthy behaviors.
In our cohort, improvement in health behaviors over midlife was associated with reduced risk of developing frailty, suggesting that lifestyle interventions in midlife may be beneficial in frailty prevention.
The expected doubling of people aged 60 or older by 2050 [
Unhealthy behaviors, comprising smoking, excessive alcohol consumption, physical inactivity, and poor diet, are recognized risk factors for several chronic diseases [
To address these limitations, this analysis of the Whitehall II cohort study aimed to examine the association between health behaviors at age 50 and subsequent risk of frailty over a mean follow-up of 20 years. Each healthy behavior and the number of healthy behaviors were investigated to assess independent and cumulative associations with risk of frailty. A further (post hoc) objective was to examine the association of change in the number of healthy behaviors over midlife and subsequent risk of frailty.
This study is reported following the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guideline (
Participants were drawn from the Whitehall II study, an ongoing cohort study established in 1985–1988 among 10,308 British civil servants aged 35–55 years at recruitment [
Data from the first 4 data collection waves (in 1985, 1991, 1997, and 2002) were used to extract information on health behaviors at age 50 for each participant, allowing a 5-year margin. Health behaviors were assessed by questionnaire and categorized into 3 groups. Smoking status was categorized as “Never smoked,” “Ex-smoking,” and “Current smoking.” The number of units of alcohol consumed in the last week was categorized as “None,” “Moderate alcohol consumption” (1–14 units/week), and “High alcohol consumption” (>14 units per week) [
Exploratory analysis was undertaken to examine the shape of the association of continuous health behavior variables, alcohol consumption, and physical activity with onset of frailty (
Frailty was measured at the clinical examination waves in 2002, 2007, 2012, and 2015 using the Fried’s frailty phenotype [
Slow walking speed was defined as when the time spent walking 8 feet was ≥3.73 seconds for men (women) with height ≤173 (≤159) cm and ≥3.20 seconds for men (women) with height >173 (>159) cm.
Low grip strength, assessed using a Smedley hand grip dynamometer, was defined for men as ≤29 kg for body mass index (BMI) ≤24 kg/m2, ≤30 kg for BMI 24.1–28 kg/m2, and ≤32 kg for BMI >28 kg/m2. For women, low grip strength was defined as ≤17 kg for BMI ≤23 kg/m2, ≤17.3 kg for BMI 23.1–26 kg/m2, ≤18 kg for BMI 26.1–29 kg/m2, and ≤21 kg for BMI >29 kg/m2.
Weight loss was defined as unintentional weight loss of 5% or more over the previous year according to Fried’s criterion. Because weight was measured every 5 years, we used a cutoff of 10% of loss on body weight to define weight loss as used in the Women’s Health Aging Study-I [
Low physical activity was denoted by an energy expenditure of <383 kcal/week for men and <270 kcal/week for women, assessed based on responses to a questionnaire on frequency and duration of participation in 20 physical activities (e.g., cycling, housework, gardening activities). A metabolic equivalent value was assigned to each activity to calculate the energy expenditure of each participant.
Exhaustion was defined based on responses to 2 items extracted from the Center for Epidemiology Studies Depression (CES-D) scale: “I felt that everything I did was an effort in the last week” and “I could not get going in the last week.” If participants answered “occasionally or moderate amount of the time (3–4 days)” or “most or all of the time (5–7 days)” to either of these items, they were categorized as exhausted.
A frailty score was calculated as the number of the above criteria met, resulting in a score ranging from 0 to 5. Participants were considered as “frail” if their score was at 3 or higher [
Mortality data until August 2017 were drawn from the British national mortality register (National Health Services Central Registry). The tracing exercise was carried out using the National Health Service identification number of each participant.
Apart from sex, ethnicity, and education, which were assessed at baseline, covariates were drawn from the same wave as the measure of health behaviors at age 50 for each participant. Demographic factors included exact age, sex, ethnicity, and marital status (“married or cohabiting” versus “single, divorced or widowed”). Socioeconomic factors consisted of education (nonacademic qualification, high school, higher secondary, university, higher university degree) and occupational position (high, intermediate, and low; representing income and status at work). The number of morbidities at age 50 was calculated based on history of diabetes, coronary heart disease, stroke, chronic obstructive pulmonary disease, depression, arthritis, cancer, hypertension, and obesity using data from clinical examinations and electronic health records (the cancer registry, the National Hospital Episode statistics database, and the Mental Health Services Data Set, which in addition to in- and out-patient data, also has data on care in the community).
Characteristics of participants were described by the number of healthy behaviors at age 50 and the frailty status at the end of the follow-up. Pearson’s chi-squared test, Fisher’s exact test, or chi-squared trend test was used to assess differences across categorical variables, and
Frailty was assessed at clinical evaluations in 2002, 2007, 2012, and 2015, but the exact date of frailty onset was unknown (interval-censored data). In addition, some participants may have developed frailty and died between 2 data collection waves without being identified as frail, making death a competing event. We therefore used an interval-censored illness–death model with a Weibull distribution to assess hazard ratio (HR) of frailty with respect to health behavior categories. This analysis accounts for both the interval-censored nature of the data and competing risk of death [
The follow-up for all participants started at age 50 (±5), when health behaviors were assessed. The analysis used age as the timescale, and separate models for each health behavior, in 3 and then 2 categories, were serially adjusted for demographic factors and study wave in which follow-up began (model 1), socioeconomic factors (model 2), and the number of morbidities (model 3). The final model included all 4 health behaviors to assess their independent association with frailty onset (model 4). In further analyses, the number of healthy behaviors was entered into the models (models 1–3) as a categorical variable and then as a continuous variable to assess the HR of frailty associated with 1-point increment in the number of healthy behaviors.
Several sensitivity analyses were undertaken to assess robustness of the findings. First, we used an alternative definition of frailty based on meeting at least 2 of the Fried’s criteria instead of 3. Second, as physical activity is included in both one of the exposures and the outcome, albeit coded differently, we redefined frailty without the “low physical activity” criterion and examined the impact of this possible bias on our results. Third, we used Cox regression instead of an illness–death model for comparison of results. Fourth, to take missing data into account, we used inverse probability weighting with information on the target population to calculate the probability of being included in the analytical sample using a logistic model that included demographic, socioeconomic, and behavioral factors at recruitment, morbidities and mortality over follow-up, and stepwise-selected interactions between covariates and health conditions. The inverse of these probabilities was used to weight results in the Cox regression.
We conducted several post hoc analyses in response to comments from reviewers. (1) Data on health behaviors assessed in 1985 and 1997 at mean ages 44.4 and 54.8 years were used to examine the association of change in health behaviors and subsequent frailty. The number of healthy behaviors was categorized into 3 groups (0–1, 2, and 3–4) at each time point, and change in these categories was examined in the analysis. Participants were followed for incident frailty from 1997 to the end of follow-up. Covariates in the analysis were assessed in 1997. For participants with missing data in 1997, data on health behaviors and covariates from 1991 or 2002 were used. (2) We examined the association between alternative definitions of healthy alcohol consumption and onset of frailty. (3) We undertook exploratory analysis on sedentary time as an additional health behavior. Sedentary time was assessed only in 1997 and could not be extracted at age 50; thus, we examined the association between sedentary time in 1997 and incident frailty. Sedentary time was measured as time spent sitting, calculated based on self-reported number of hours per week spent “sitting at work, driving, commuting or other” and “sitting at home e.g. watching TV, sewing, at desk,” and was categorized into tertiles given the lack of definition of unhealthy sedentary time. A score of healthy behavior was calculated with all 5 behaviors, with healthy sedentary time defined as not being in the highest tertile of time spent sitting.
Of the 10,308 persons recruited to the Whitehall II study, 24 died before the age of 50 years, 337 were over 50 years at recruitment, and 6 were frail at age 50 (
Characteristics at age 50 | Total study sample | Nonfrail | Frail | |
---|---|---|---|---|
( |
( |
( |
||
Sex | <0.001 | |||
Men | 4,501 (70.8) | 4,261 (72.1) | 240 (53.9) | |
Women | 1,856 (29.2) | 1,651 (27.9) | 205 (46.1) | |
Ethnicity | <0.001 | |||
White | 5,830 (91.7) | 5,464 (92.4) | 366 (82.3) | |
Nonwhite | 527 (8.3) | 448 (7.6) | 79 (17.7) | |
Marital status | <0.001 | |||
Married/cohabiting | 4,942 (77.7) | 4,658 (78.8) | 284 (63.8) | |
Single, divorced, or widowed | 1,415 (22.3) | 1,254 (21.2) | 161 (36.2) | |
Education | <0.001 | |||
No academic qualification | 663 (10.4) | 590 (9.9) | 73 (16.4) | |
High school | 2,062 (32.4) | 1,903 (32.2) | 159 (35.7) | |
Higher secondary | 1,737 (27.3) | 1,638 (27.7) | 99 (22.2) | |
University | 1,411 (22.2) | 1,319 (22.3) | 92 (20.7) | |
Higher university degree | 484 (7.6) | 462 (7.8) | 22 (4.9) | |
Occupational position | <0.001 | |||
Low | 845 (13.3) | 717 (12.1) | 128 (28.8) | |
Intermediate | 2,824 (44.4) | 2,624 (44.4) | 200 (44.9) | |
High | 2,688 (42.3) | 2,571 (43.5) | 117 (26.3) | |
Number of morbidities | <0.001 | |||
0 | 4,370 (68.7) | 4,113 (69.6) | 257 (57.7) | |
1 | 1,549 (24.4) | 1,408 (23.8) | 141 (31.7) | |
2 or more | 438 (6.9) | 391 (6.6) | 47 (10.6) | |
Smoking status | <0.001 | |||
Never smoked | 3,202 (50.4) | 2,967 (50.2) | 235 (52.8) | |
Ex-smoking | 2,378 (37.4) | 2,251 (38.1) | 127 (28.5) | |
Current smoking | 777 (12.2) | 694 (11.7) | 83 (18.6) | |
Alcohol consumption | <0.001 | |||
None | 989 (15.6) | 873 (14.8) | 116 (26.1) | |
Moderate | 3,543 (56.2) | 3,318 (56.1) | 225 (50.6) | |
High | 1,825 (28.7) | 1,721 (29.1) | 104 (23.4) | |
Physical activity | <0.001 | |||
Inactive | 703 (11.1) | 599 (10.1) | 104 (23.4) | |
Moderately active | 1,500 (23.6) | 1,358 (23.0) | 142 (31.9) | |
Active | 4,154 (65.3) | 3,955 (66.9) | 199 (44.7) | |
Fruits and vegetables consumption | <0.001 | |||
Less than once a day | 2,186 (34.4) | 2,007 (33.9) | 179 (40.2) | |
Once a day | 2,445 (38.5) | 2,262 (38.3) | 183 (41.1) | |
At least twice a day | 1,726 (27.2) | 1,643 (27.8) | 83 (18.6) | |
Slow walking speed | 503 (7.9) | 230 (3.9) | 273 (61.3) | <0.001 |
Low grip strength | 1,399 (22.0) | 1,075 (18.2) | 324 (72.8) | <0.001 |
Exhaustion | 794 (12.5) | 511 (8.6) | 283 (63.6) | <0.001 |
Low physical activity | 2,210 (34.8) | 1,799 (30.4) | 411 (92.4) | <0.001 |
Weight loss | 302 (4.7) | 181 (3.1) | 121 (27.2) | <0.001 |
*Values are numbers (percentages). Percentages are reported in column.
†End of follow-up corresponds to date of frailty diagnosis or last wave of clinical examination for nonfrail participants.
Because there was no evidence of an interaction of the number of healthy behaviors with sex (
Health behaviors | Frail % | Model 1 |
Model 2 |
Model 3 |
Model 4 |
|||||
---|---|---|---|---|---|---|---|---|---|---|
HR (95% CI) | HR (95% CI) | HR (95% CI) | HR (95% CI) | |||||||
Smoking status | ||||||||||
Never smoked | 235/3,202 | 7.34 | 0.59 (0.46–0.77) | <0.001 | 0.62 (0.48–0.81) | <0.001 | 0.62 (0.47–0.80) | <0.001 | 0.68 (0.52–0.89) | 0.01 |
Ex-smoking | 127/2,378 | 5.34 | 0.50 (0.38–0.67) | <0.001 | 0.52 (0.39–0.69) | <0.001 | 0.50 (0.37–0.66) | <0.001 | 0.53 (0.40–0.71) | <0.001 |
Current smoking | 83/777 | 10.68 | 1 (ref) | 1 (ref) | 1 (ref) | 1 (ref) | ||||
Alcohol consumption | ||||||||||
None | 116/989 | 11.73 | 1.22 (0.91–1.65) | 0.18 | 1.08 (0.79–1.46) | 0.64 | 1.11 (0.81–1.51) | 0.51 | 1.15 (0.85–1.57) | 0.37 |
Moderate | 225/3,543 | 6.35 | 0.76 (0.59–0.97) | 0.03 | 0.72 (0.56–0.92) | 0.01 | 0.75 (0.58–0.97) | 0.03 | 0.76 (0.59–0.98) | 0.03 |
High | 104/1,825 | 5.70 | 1 (ref) | 1 (ref) | 1 (ref) | 1 (ref) | ||||
Physical activity | ||||||||||
Inactive | 104/703 | 14.79 | 1 (ref) | 1 (ref) | 1 (ref) | 1 (ref) | ||||
Moderately active | 142/1,500 | 9.47 | 0.81 (0.62–1.05) | 0.11 | 0.87 (0.65–1.17) | 0.37 | 0.90 (0.68–1.19) | 0.47 | 0.94 (0.71–1.25) | 0.67 |
Active | 199/4,154 | 4.79 | 0.55 (0.42–0.72) | <0.001 | 0.59 (0.45–0.78) | <0.001 | 0.61 (0.47–0.81) | <0.001 | 0.66 (0.48–0.88) | 0.001 |
Fruits and vegetables consumption | ||||||||||
Less than once a day | 179/2,186 | 8.19 | 1 (ref) | 1 (ref) | 1 (ref) | 1 (ref) | ||||
Once a day | 183/2,445 | 7.48 | 0.83 (0.67–1.02) | 0.08 | 0.87 (0.70–1.08) | 0.19 | 0.85 (0.68–1.06) | 0.14 | 0.91 (0.73–1.13) | 0.40 |
At least twice a day | 83/1,726 | 4.81 | 0.64 (0.49–0.85) | <0.001 | 0.68 (0.52–0.90) | 0.01 | 0.65 (0.50–0.87) | <0.001 | 0.70 (0.53–0.92) | 0.01 |
*Models 1, 2, and 3 were estimated for each health behavior separately.
‡Model 2: model 1 additionally adjusted for education and occupational position.
§Model 3: model 2 additionally adjusted for the number of morbidities at age 50.
‖Model 4: model 3 additionally adjusted for all other health behaviors.
Abbreviations: CI, confidence interval; HR: hazard ratio; ref, reference group
Subsequent analyses of each healthy behavior as binary variables (
Healthy behaviors | Frail % | Model 1 |
Model 2 |
Model 3 |
Model 4 |
|||||
---|---|---|---|---|---|---|---|---|---|---|
HR (95% CI) | HR (95% CI) | HR (95% CI) | HR (95% CI) | |||||||
Noncurrent smoking | ||||||||||
No | 83/777 | 10.68 | 1 (ref) | 1 (ref) | 1 (ref) | 1 (ref) | ||||
Yes | 362/5,580 | 6.49 | 0.56 (0.44–0.72) | <0.001 | 0.58 (0.45–0.75) | <0.001 | 0.57 (0.44–0.73) | <0.001 | 0.56 (0.44–0.71) | <0.001 |
Moderate alcohol consumption | ||||||||||
No | 220/2,814 | 7.82 | 1 (ref) | 1 (ref) | 1 (ref) | 1 (ref) | ||||
Yes | 225/3,543 | 6.35 | 0.68 (0.56–0.82) | <0.001 | 0.69 (0.57–0.84) | <0.001 | 0.71 (0.59–0.86) | 0.001 | 0.73 (0.61–0.88) | <0.001 |
Physical activity | ||||||||||
No | 246/2,003 | 12.28 | 1 (ref) | 1 (ref) | 1 (ref) | 1 (ref) | ||||
Yes | 199/4,154 | 4.79 | 0.63 (0.51–0.78) | <0.001 | 0.64 (0.52–0.79) | <0.001 | 0.66 (0.53–0.81) | <0.001 | 0.66 (0.54–0.81) | <0.001 |
Fruits and vegetables consumption at least twice a day | ||||||||||
No | 362/4,631 | 7.82 | 1 (ref) | 1 (ref) | 1 (ref) | 1 (ref) | ||||
Yes | 83/1,726 | 4.81 | 0.71 (0.56–0.92) | 0.01 | 0.74 (0.57–0.95) | 0.02 | 0.72 (0.55–0.94) | 0.02 | 0.76 (0.59–0.98) | 0.03 |
*Models 1, 2, and 3 were estimated for each healthy behavior separately.
‡Model 2: model 1 additionally adjusted for education and occupational position.
§Model 3: model 2 additionally adjusted for the number of morbidities at age 50.
‖Model 4: model 3 additionally adjusted for all other healthy behaviors.
Abbreviations: CI, confidence interval; HR, hazard ratio; ref, reference group
Lower risk of incident frailty was observed in participants who had 2 or more healthy behaviors at age 50 compared with those who had none (
Model 1: age as a timescale, adjusted for sex, ethnicity, marital status, and wave at inclusion. Model 2: model 1 additionally adjusted for education and occupational position. Model 3: model 2 additionally adjusted for the number of morbidities at age 50. Associated estimations are in
All sensitivity analyses yielded results that were similar to those in the main analyses so that the risk of frailty decreased as the number of healthy behaviors at age 50 increased (
Change in the number of healthy behaviors was calculated among 6,435 participants with data on health behaviors available in 1985 and 1997 (
Number of healthy behaviors | Frail % | Model 1 |
Model 2 |
Model 3 |
|||||
---|---|---|---|---|---|---|---|---|---|
In 1985, mean age = 44.4 years | In 1997 |
HR (95% CI) | HR (95% CI) | HR (95% CI) | |||||
0–1 | 0–1 | 66/422 | 15.64 | 1 (ref) | 1 (ref) | 1 (ref) | |||
0–1 | 2 | 51/555 | 9.19 | 0.61 (0.42–0.90) | 0.01 | 0.63 (0.43–0.92) | 0.02 | 0.64 (0.44–0.94) | 0.02 |
0–1 | 3–4 | 38/426 | 8.92 | 0.56 (0.37–0.85) | 0.007 | 0.57 (0.37–0.87) | 0.009 | 0.57 (0.38–0.87) | 0.009 |
2 | 0–1 | 25/248 | 10.08 | 0.65 (0.41–1.04) | 0.71 | 0.66 (0.41–1.05) | 0.08 | 0.69 (0.43–1.10) | 0.12 |
2 | 2 | 84/999 | 8.41 | 0.54 (0.38–0.75) | <0.001 | 0.54 (0.39–0.76) | <0.001 | 0.57 (0.41–0.81) | 0.001 |
2 | 3–4 | 79/1,329 | 5.94 | 0.35 (0.24–0.50) | <0.001 | 0.36 (0.25–0.51) | <0.001 | 0.37 (0.26–0.53) | <0.001 |
3–4 | 0–1 | 7/72 | 9.72 | 0.66 (0.29–1.48) | 0.31 | 0.67 (0.29–1.52) | 0.34 | 0.56 (0.25–1.27) | 0.17 |
3–4 | 2 | 29/545 | 5.32 | 0.39 (0.25–0.62) | <0.001 | 0.39 (0.25–0.63) | <0.001 | 0.41 (0.26–0.65) | <0.001 |
3–4 | 3–4 | 85/1,839 | 4.62 | 0.29 (0.21–0.42) | <0.001 | 0.30 (0.21–0.43) | <0.001 | 0.32 (0.23–0.46) | <0.001 |
*Or in 1991 (
‡ Model 2: model 1 additionally adjusted for education and occupational position.
§ Model 3: model 2 additionally adjusted for the number of morbidities at the second measurement of health behaviors.
Abbreviations: CI, confidence interval; HR, hazard ratio; ref, reference
Associations between alternative definitions of healthy alcohol consumption and frailty onset are presented in
Post hoc analysis on the association between sedentary time in 1997 and incident frailty based on 5,296 participants (
In this longitudinal study based on over 6,000 participants, healthy behaviors—defined as not smoking, moderate alcohol consumption, at least 2.5 hours per week of moderate to vigorous physical activity, and consumption of fruits and vegetables at least twice a day—assessed at age 50 were independently associated with a reduced risk of frailty onset over a mean 20-year follow-up, with risk reduction ranging from 24% for fruits and vegetables consumption at least twice a day to 44% for nonsmoking. The benefits of healthy behaviors were cumulative, as the risk of frailty decreased progressively with greater number of healthy behaviors to reach a risk reduction of 72% among those with all 4 healthy behaviors at age 50 compared with those without any healthy behaviors at that age. Our results also show that favorable change in the number of healthy behaviors in midlife was associated with a reduced risk of subsequent frailty compared with persistently low number of healthy behaviors.
Previous studies have suggested that current nonsmoking (including never and ex-smokers) [
In addition to the independent association of each health behavior with the risk of frailty, the present study suggests that their benefits have a cumulative effect so that for each additional healthy behavior at age 50, there was a 31% risk reduction in frailty. These results are in accordance with a recent study based on 1,309 participants from the general population (mean age at health behaviors measure = 70 years) showing the number of favorable health behaviors among nonsmoking, physically active, healthy diet, adequate sleeping duration, not being sedentary, and daily social interaction to be associated with lower incidence of self-reported frailty symptoms assessed twice over an 8-year period [
Multiple mechanisms may explain the association between healthy behaviors and frailty. Physical activity, moderate alcohol intake, and healthy diet are associated with increase in adiponectin and high-density lipoprotein (HDL) levels [
The present study has several strengths, including the measurement of health behaviors for all participants at age 50 (SD, 2.1 years) and a mean follow-up duration of 20 years. In contrast, most previous studies were undertaken among older populations, with a short follow-up for the onset of frailty. As the frailty phenotype develops over several years and includes an intermediate state of prefrailty [
Our findings need to be considered in light of the study limitations. First, the population composed of British civil servants is not representative of the general population. However, this is an unlikely source of bias because previous analyses show that although participants from the cohort are healthier in terms of risk factor levels and prevalence of cardiovascular disease, the associations between cardiovascular risk factors, including health behaviors, and cardiovascular disease are similar to those in general population studies [
In the context of aging of populations worldwide, effective prevention is key in order to allow older adults to remain healthy as long as possible and reduce the societal burden of aging. This includes prevention of frailty, a geriatric syndrome associated with higher risk of several health conditions and increased healthcare needs. Frailty is more prevalent among women and participants from the lower socioeconomic group; in our analyses, the number of healthy behaviors at age 50 and the change in this number over midlife were similarly associated with risk of incident frailty in both sexes and in different socioeconomic groups. Because health behaviors are modifiable, they are a good target for frailty prevention at the population level. The present findings along with those from previous studies support the development of interventions in clinical settings to encourage a healthy lifestyle as a whole rather than focusing on specific healthy behaviors to promote healthy aging. Among the multiple tools to assess frailty, we chose to use Fried’s frailty phenotype given its robust association with health outcomes; its use in several studies, making our results amenable to replication; and the fact that its association with health outcomes such as mortality has been shown to be similar to that of more extensive measures such as the frailty index based on the accumulation of deficits model [
Our analysis based on an observational cohort study showed that a greater number of healthy behaviors at age 50 as well as improvement in health behaviors over midlife were associated with lower risk of frailty over a 20-year follow-up period. These findings highlight the importance of healthy behavior in midlife for prevention of frailty at older ages.
(DOCX)
(DOCX)
(DOCX)
CI, confidence interval; HR, hazard ratio.
(DOCX)
CI, confidence interval; HR, hazard ratio.
(DOCX)
CI, confidence interval; HR, hazard ratio.
(DOCX)
CI, confidence interval; HR, hazard ratio.
(DOCX)
Blue line represents the hazard ratio for frailty compared with “14 units of alcohol per week” (reference), and black dashed line represents the corresponding 95% confidence interval estimated based on a Cox regression model with age at timescale adjusted for sex, ethnicity, marital status, and wave at inclusion.
(TIF)
Blue line represents the hazard ratio for frailty compared with 2.5 hours of moderate and vigorous physical activity (reference), and black dashed line represents the corresponding 95% confidence interval estimated based on a Cox regression model with age at timescale adjusted for sex, ethnicity, marital status, and wave at inclusion.
(TIF)
Age was used as timescale, and models are adjusted for sex, ethnicity, marital status, wave of inclusion, education, occupational position, and number of morbidities at age 50. HR, hazard ratio.
(TIF)
(DOCX)
STROBE, Strengthening the Reporting of Observational Studies in Epidemiology.
(DOCX)
We are grateful to the participants from civil service departments and their welfare, personnel, and establishment officers and all members of the Whitehall II study teams.
The lead author affirms that this manuscript is an honest, accurate, and transparent account of the study being reported; that no important aspects of the study have been omitted; and that any discrepancies from the study as planned have been explained. The lead author in this statement is the study guarantor.
Center for Epidemiology Studies Depression
confidence interval
high-density lipoprotein
hazard ratio
standard deviation
Strengthening the Reporting of Observational Studies in Epidemiology
Dear Dr. Sabia,
Thank you very much for submitting your manuscript "Association between healthy behaviors at age 50 and frailty at older ages: 20-year follow-up of the Whitehall II cohort study" (PMEDICINE-D-19-04222) for consideration at PLOS Medicine.
Your paper was discussed among the editorial team, evaluated by an academic editor with relevant expertise, and sent to independent reviewers, including a statistical reviewer. The reviews are appended at the bottom of this email and any accompanying reviewer attachments can be seen via the link below:
[LINK]
In light of these reviews, we will not be able to accept the manuscript for publication in the journal in its current form, but we would like to invite you to submit a revised version that fully addresses the reviewers' and editors' comments. You will appreciate that we cannot make a decision about publication until we have seen the revised manuscript and your response, and we expect to seek re-review by one or more of the reviewers.
In revising the manuscript for further consideration, your revisions should address the specific points made by each reviewer and the editors. Please also check the guidelines for revised papers at
In addition, we request that you upload any figures associated with your paper as individual TIF or EPS files with 300dpi resolution at resubmission; please read our figure guidelines for more information on our requirements:
We hope to receive your revised manuscript by Feb 27 2020 11:59PM. Please email us (
***Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.***
We ask every co-author listed on the manuscript to fill in a contributing author statement, making sure to declare all competing interests. If any of the co-authors have not filled in the statement, we will remind them to do so when the paper is revised. If all statements are not completed in a timely fashion this could hold up the re-review process. If new competing interests are declared later in the revision process, this may also hold up the submission. Should there be a problem getting one of your co-authors to fill in a statement we will be in contact. YOU MUST NOT ADD OR REMOVE AUTHORS UNLESS YOU HAVE ALERTED THE EDITOR HANDLING THE MANUSCRIPT TO THE CHANGE AND THEY SPECIFICALLY HAVE AGREED TO IT. You can see our competing interests policy here:
Please use the following link to submit the revised manuscript:
Your article can be found in the "Submissions Needing Revision" folder.
To enhance the reproducibility of your results, we recommend that you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. For instructions see
Please ensure that the paper adheres to the PLOS Data Availability Policy (see
Please let me know if you have any questions. Otherwise, we look forward to receiving your revised manuscript in due course.
Sincerely,
Richard Turner PhD, for Thomas McBride, PhD
Senior Editor, PLOS Medicine
-----------------------------------------------------------
Requests from the editors:
Please adapt the data statement to state "... because of constraints dictated by the study's ethics approval ..." or similar.
Please adapt the title to include a study descriptor and accord with journal style. We suggest "Healthy behaviors at age 50 years and frailty at older ages in a 20-year follow-up of the Whitehall II cohort: a longitudinal study".
Please add summary demographic information on participants to the abstract.
In the abstract and elsewhere, please add p values alongside 95% CI where available.
Please add a new final sentence to the "methods and findings" subsection of your abstract to summarize the study's main limitations.
At line 45, please begin the sentence with "Our findings suggest that ..." or similar.
After the abstract, we will need to ask you to add a new and accessible "author summary" section in non-identical prose. You may find it helpful to consult one or two recent research papers published in PLOS Medicine to get a sense of the preferred style.
Early in the methods section of your main text, please state whether the study had a protocol or prespecified analysis plan and if so attach the document(s) as a supplementary file (referred to in the text). Please highlight analyses that were not prespecified.
In the discussion of study limitations at line 281, please add some additional detail. For example, in noting the limitations of observational evidence, you could mention the possible existence of unmeasured confounders.
Throughout the text, please style reference call-outs as follows: "... involved in frailty development [49,50]." (i.e., preceding punctuation).
In the reference list, please ensure that journal names are abbreviated correctly and consistently (e.g., "Lancet" for references 2, 4 and 8).
Please additional access details to reference 10.
Please add a completed checklist for the most appropriate reporting guideline, which we suspect will be STROBE, as a supplementary document (referred to in the methods section). In the checklist, individual items should be referred to by section (e.g., "Methods") and paragraph number rather than by page or line numbers, as the latter generally change in the event of publication.
Comments from the reviewers:
*** Reviewer #1:
"Association between healthy behaviors at age 50 and frailty at older ages: 20-year follow-up of the Whitehall II cohort study" analyzes the well-known Whitehall II cohort of British civil servants, in particular as to how "healthy behaviors" at age 50 affect "frailty" approximately 20 years later (with a range of plus or minus five years from the starting age, and also with a range for the follow-up depending on which of the four later clinical evaluations (2002, 2007, 2012, 2015) was most appropriate. The general findings were unsurprising, in that exhibiting more examples of healthy behaviours at age 50 correlated with reduced frailty later on. The hazard ratio for each additional healthy behaviour was estimated to be 0.69, and for all vs. no healthy behaviours as 0.28.
The authors used an appropriate interval-censored illness-death model for the main statistical analysis. Furthermore, the expected additional demographic and socioeconomic covariates were considered. On the whole, the analysis and conclusions seem broadly valid. However, there remain a few points that might be considered.
On the variability of the cohort participants in terms of initial age and follow-up age, it might be helpful to include more detailed data, such as a table with rows as initial age and columns as years to follow-up, with each cell containing the number of participants with that initial age/years to follow-up combination. Further on the data, while the demographic and healthy behaviour characteristics are provided in Table 1, there is no such information for frailty characteristics. This could be added.
On "Moderate alcohol consumption" being defined as a healthy behavior (line 95), the referenced source [32] states that "no robust association with alcohol consumption was found". In any case, it is unclear why "None" is not also defined as healthy behavior. As such, more involved analysis might be performed on alcohol consumption (e.g. comparing "None" only, and "None+Moderate" as criteria)
A general comment on the categorizations used for frailty and healthy behavior measures: while various justifications are cited (e.g. most prominently Fried's frailty phenotype for frailty), the availability of raw values for at least some of these measures suggests that more nuanced analyses might be possible for those measures at least. For example, the "slow walking speed" measure is suddenly discontinuous at a height of 173/159cm for males and females respectively. It could be interesting to examine effects with greater precision (e.g. for "Moderate alcohol consumption", the range of 1 to 14 units is also rather large and could conceivably be split further). Of course, it remains up to the authors as to whether these might constitute reasonable secondary analyses.
*** Reviewer #2:
This original article evaluates the association between health behaviors at age 50 in Whitehall II cohort study participants and the incidence of the frailty syndrome, using Fried's phenotype, at 4 follow-up visits, in 2002, 2007, 2012 and 2015. The findings show that each healthy behavior was independently associated with lower risk of incident frailty, and that the impact of aggregate health behaviors was cumulative in further lowering the risk for frailty.
Questions to authors:
Methods: p. 4: Baseline recruitment occurred in 1985-88, when those recruited were 35-55; it would be useful to know the range of length of follow-up after age 50 for those who were included in this analysis. Mean follow up length is mentioned on p. 16 as being 20 years.
Results: Given that p. 9 indicates confounding by higher SES and fewer comorbidities among those with greater number of healthy behaviors at age 50, it would be useful to understand if those with lower socio economic background who also had positive health behaviors had comparable outcomes to those of higher socioeconomic background. Further, is it possible to offer an analysis of length of exposure to adverse behaviors prior to age 50, by confounders?
p. 11: could the comparable incidence of frailty for moderate and high alcohol drinkers be associated with loss to mortality among the high drinkers?
p. 11: associations attenuated after adjustment for socioeconomic factors, comorbidities. Is the association of occupational position with frailty attenuated by healthy behaviors, and if so which?
Discussion:
It would be useful to hear analysis of the finding on p. 11 that there was no evidence of interaction between sex and number of healthy behaviors in association with frailty.
p. 18. Lines 309-312. The aggregate health outcomes affected by these health behaviors warrants further discussion. Since those with positive health behaviors had fewer comorbidities at age 50, please discuss the findings re: independence of association of health behaviors and frailty from presence of chronic diseases, comorbidity.
*** Reviewer #3:
Thank you for the opportunity to review the manuscript by Gil-Salcedo who investigated the association of health behaviors with developing physical frailty in adults who were 50 years old at baseline. Midlife represents a critical time in which healthy lifestyle behaviors could play a major role in helping to prevent the health problems of old age (e.g., dementia PMID: 28735855) and understanding health behaviors may be pivotal in frailty prevention (PMID: 26805753). In this study, participants from the Whitehall II cohort study (n= 6357) had their smoking status, alcohol consumption, physical activity, and fruit/vegetable consumption measured at their first study visit (at age 50). Frailty was assessed at four time points, over a period of 20 years. All health behaviors were independently associated with developing frailty (445/6357 developed frailty) even when all health behaviors were included in their model which considered death as a competing risk. Non-smokers had the lowest risk of developing frailty, followed by moderate alcohol consumption, physical activity, and then fruit/vegetable consumption. A greater number of health behaviors resulted in a dose-associated lower risk of frailty.
Major comments:
-Have the authors considered examining sedentary time in their models? Sedentary behaviors (low energy expenditure activities while seated/lying/reclining during awake hours) are associated with frailty (PMID: 30355522) and may provide insights into their data. It might be possible that even though someone is physically active, they could still report high levels of sitting time.
- A novelty of this study is the examination of the impact of multiple health behaviors on developing frailty. However, it is reasonable to assume that health behaviors changed over the follow up period in the present study, and such changes are likely to impact whether an individual develops frailty or not. Based on the author's description of the Whitehall II study, it is feasible to assess changes in health behaviors. It is strongly recommended that changes in health behaviors be assessed in relation to developing frailty, as these results would be quite novel.
-The authors adjust for several chronic conditions at age 50 in their statistical models, some of which are more severe than others. Can the authors determine which baseline conditions were more likely to result in developing frailty?
Minor comments:
-It appears that disability was not considered in their models. How might they affect the findings here? Were participants free of disability at baseline?
-Can the authors please comment on the multiple frailty tools which exist? Although the frailty phenotype is among the most widely used in research, how might the findings differ if another frailty tool, say the frailty index based on the accumulation of deficits model, were used?
-It is recommended that the authors should comment on the implications of their findings relation to public health and in the clinical context.
***
Any attachments provided with reviews can be seen via the following link:
[LINK]
Submitted filename:
Dear Dr. Sabia,
Thank you very much for re-submitting your manuscript "Healthy behaviors at age 50 years and frailty at older ages in a 20-year follow-up of the Whitehall II cohort: a longitudinal study" (PMEDICINE-D-19-04222R1) for review by PLOS Medicine.
I have discussed the paper with my colleagues and the academic editor and it was also seen again by two reviewers. I am pleased to say that provided the remaining editorial and production issues are dealt with we are planning to accept the paper for publication in the journal.
The remaining issues that need to be addressed are listed at the end of this email. Any accompanying reviewer attachments can be seen via the link below. Please take these into account before resubmitting your manuscript:
[LINK]
Our publications team (
***Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.***
In revising the manuscript for further consideration here, please ensure you address the specific points made by each reviewer and the editors. In your rebuttal letter you should indicate your response to the reviewers' and editors' comments and the changes you have made in the manuscript. Please submit a clean version of the paper as the main article file. A version with changes marked must also be uploaded as a marked up manuscript file.
Please also check the guidelines for revised papers at
We expect to receive your revised manuscript within 1 week. Please email us (
We ask every co-author listed on the manuscript to fill in a contributing author statement. If any of the co-authors have not filled in the statement, we will remind them to do so when the paper is revised. If all statements are not completed in a timely fashion this could hold up the re-review process. Should there be a problem getting one of your co-authors to fill in a statement we will be in contact. YOU MUST NOT ADD OR REMOVE AUTHORS UNLESS YOU HAVE ALERTED THE EDITOR HANDLING THE MANUSCRIPT TO THE CHANGE AND THEY SPECIFICALLY HAVE AGREED TO IT.
Please ensure that the paper adheres to the PLOS Data Availability Policy (see
If you have any questions in the meantime, please contact me or the journal staff on
We look forward to receiving the revised manuscript by May 08 2020 11:59PM.
Sincerely,
Thomas McBride, PhD
Senior Editor
PLOS Medicine
------------------------------------------------------------
Requests from Editors:
1- Please consider addding "UK" to the title: "Healthy behaviors at age 50 years and frailty at older ages in a 20-year follow-up of the UK Whitehall II cohort: a longitudinal study"
2- In the "Study population" section of the Methods, please specify whether consent is written.
3- In the Discussion, please switch the order of the “Strengths and limitations” and “Comparison with previous studies” sections.
4- Regarding your response to reviewer 2, point 5, it may be relevant to mention this point and reference the Brunner study in the Discussion.
Comments from Reviewers:
Reviewer #1: We thank the authors for carefully considering our previous comments, and performing appropriate post-hoc analyses. The remaining minor suggestions pertain mostly to the supplementary material:
1. There appears to be an issue with the image for S3 Figure (green background, blurred words)
2. For Table 2/3 & S4 to S7 Tables, it might be helpful for the N frail/N total ratio to also be expressed as a percentage for easy comparison, if formatting permits.
Reviewer #3: The reviewers have done a great job addressing reviewer comments. I accept the article in its revised form.
Any attachments provided with reviews can be seen via the following link:
[LINK]
Submitted filename:
Dear Dr. Sabia,
On behalf of my colleagues and the academic editor, Dr. Sanjay Basu, I am delighted to inform you that your manuscript entitled "Healthy behaviors at age 50 years and frailty at older ages in a 20-year follow-up of the UK Whitehall II cohort: a longitudinal study" (PMEDICINE-D-19-04222R2) has been accepted for publication in PLOS Medicine.
PRODUCTION PROCESS
Before publication you will see the copyedited word document (in around 1-2 weeks from now) and a PDF galley proof shortly after that. The copyeditor will be in touch shortly before sending you the copyedited Word document. We will make some revisions at the copyediting stage to conform to our general style, and for clarification. When you receive this version you should check and revise it very carefully, including figures, tables, references, and supporting information, because corrections at the next stage (proofs) will be strictly limited to (1) errors in author names or affiliations, (2) errors of scientific fact that would cause misunderstandings to readers, and (3) printer's (introduced) errors.
If you are likely to be away when either this document or the proof is sent, please ensure we have contact information of a second person, as we will need you to respond quickly at each point.
PRESS
A selection of our articles each week are press released by the journal. You will be contacted nearer the time if we are press releasing your article in order to approve the content and check the contact information for journalists is correct. If your institution or institutions have a press office, please notify them about your upcoming paper at this point, to enable them to help maximize its impact.
PROFILE INFORMATION
Now that your manuscript has been accepted, please log into EM and update your profile. Go to
Thank you again for submitting the manuscript to PLOS Medicine. We look forward to publishing it.
Best wishes,
Thomas McBride, PhD
Senior Editor
PLOS Medicine