| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
|
|
|
|||||||
|
|||||||||
SPECIAL ARTICLE |
1 Department of Medicine, School of Medicine, Johns Hopkins University, Baltimore, Maryland.
2 Center on Aging and Health, Johns Hopkins Medical Institutions, Baltimore, Maryland.
3 Department of Biostatistics, Bloomberg School of Public Health, Johns Hopkins University, Baltimore, Maryland.
4 Mailman School of Public Health, Columbia University, New York, New York.
Address correspondence to Qian-Li Xue, PhD, 2024 E. Monument Street, Suite 2-700, Baltimore, MD 21205. E-mail: qxue{at}jhsph.edu
A
Background. Understanding points of onset of the frailty syndrome is vital to early identification of at-risk individuals and to targeting intervention efforts to those components that are first affected, when reversal may be most possible. This study aims to characterize natural history by which commonly used frailty criteria manifest and to assess whether the rate of progression to frailty depends on initial manifestations.
Methods. The investigation was based on a 7.5-year observational study of 420 community-dwelling women aged 70–79 years who were not frail at baseline, with frailty defined as meeting 
3 of 5 criteria: weight loss, slow walking speed, weakness, exhaustion, and low physical activity level.
Results. The 7.5-year incidence of frailty was 9% among women who were nonfrail at baseline. Despite significant heterogeneity, weakness was the most common first manifestation, and occurrence of weakness, slowness, and low physical activity preceded exhaustion and weight loss in 76% of the women who were nonfrail at baseline. Women with exhaustion or weight loss as initial presenting symptoms were 3–5 times more likely to become frail than were women without any criterion (p <.05).
Conclusions. Our findings suggest that weakness may serve as a warning sign of increasing vulnerability in early frailty development, and weight loss and exhaustion may help to identify women most at risk for rapid adverse progression.
Key Words: Exhaustion • Muscle weakness • Physical activity • Physiologic reserve • Sarcopenia • Walking speed • Weight loss
|
To test these hypotheses, we characterized the natural history of frailty in a longitudinal cohort study of older women who were not frail at baseline. Specifically, we described patterns of accumulation of the phenotypic components of frailty, and evaluated whether there was a predictable order of decline of these components and whether the rate of progression to frailty depended on initial manifestations.
METHODS
Study Population
The Women's Health and Aging Study II (WHAS II) is a prospective cohort study of 436 community-dwelling women who were 70–79 years of age, cognitively intact, and had no or mild physical disability at baseline (10,11). Interviews and physical examinations were conducted at baseline, beginning in 1994, and at four follow-up examinations 18 months apart (except for the interval between the third and the fourth examination, which was, on average, 3 years). An average follow-up time of 7.5 years resulted. The study was approved by the Johns Hopkins University Institutional Review Board. This article concerns 420 participants who were not frail (i.e., nonfrail or prefrail) at baseline.
Frailty Phenotype
We used the frailty phenotype that was cross-validated in the WHAS (8), closely akin to that proposed in the Cardiovascular Health Study (CHS) (1). It consists of five binary criteria: (i) weakness (grip strength 
17, 17.3, 18, 21 kg by a JAMAR dynamometer for body mass index [BMI] 23, 23.1–26, 26.1–29, and >29, respectively); (ii) slowness (usual-pace 4-m walking speed 0.65 m/s if height 159 cm or 0.76 if height 159 cm); (iii) low physical activity (total energy expenditure <90 Kcal/wk on six activities [walking, doing strenuous household chores, doing strenuous outdoor chores, dancing, bowling, exercise]); (iv) weight loss (unintentional weight loss of at least 7.5% between examinations spanning 18-months intervals or 15% between examinations spanning a 36-month interval [examinations 3 and 4], or having a BMI < 18.5); and (v) exhaustion (low energy level [<3 on a Likert scale of 0–10] or feeling unusually tired or weak most or all the time by self-report). We classified women meeting 3 or more criteria as frail, 1 or 2 as prefrail, and 0 as nonfrail. Frailty status was treated as missing if >2 criteria were missing. At baseline, 14 participants had one frailty criterion missing, which was conservatively treated as an absence of that criterion in the analyses.
Statistical Analyses
We used three approaches to study patterns of accumulation of frailty criteria over time and degree of heterogeneity in these. First, we compared the prevalence of the five frailty criteria at the examinations in which prefrailty and frailty were first observed, among women who were nonfrail at baseline. We also characterized prevalence of individual frailty criteria at frailty onset among women who were prefrail at baseline.
Second, we compared population-averaged criteria incidences among women who were nonfrail at baseline. Because criteria could only be observed at clinic visits, we used a multivariate extension of the univariate "discrete-time" proportional hazard model (12,13) (see Appendix) to estimate the relative hazards of developing each frailty criterion.
Third, we analyzed the individual-level order of emergence of frailty criteria. The observed orders were summarized into groups based on hypotheses derived from the prior discrete-time proportional hazard analysis. To test the null hypothesis of within-person independence in the order of initial manifestations of frailty, and separately of last manifestations upon frailty onset, we used Fisher's exact or chi-square tests for comparisons between the expected (under the null) and observed frequency distributions of the order groups. Failure to reject the null would indicate lack of within-person temporal ordering among the frailty manifestations.
Finally, to evaluate whether the rate of progression to frailty depended on initial manifestations, we modeled the relative hazard of incident frailty comparing by baseline status (i) those with each specific criterion to those without any criterion (i.e., nonfrail), (ii) those with only one specific criterion to those without any criterion, and (iii) the number of criteria, using discrete-time proportional hazard models, adjusting for age, race, education, and number of chronic diseases. MIXOR (version 2.0; available at http://tigger.uic.edu/
hedeker/mix.html) was used to fit discrete-time proportional hazards models, and the rest of the analyses used S-PLUS (version 2000; Insightful Inc., Seattle, WA).
RESULTS
Of the 420 women in this study, the mean age was 74 years; 19% were African American; 30% reported having mobility difficulty; and 60% were either overweight or obese. On average, they had 13 years of education and 1.5 diseases (Table 1). Two hundred sixty-eight (62%) and 152 (35%) were nonfrail and prefrail, respectively. Compared to the nonfrail women, the prefrail women were less educated (mean [standard deviation {SD}]: 11.9 years [3.3] vs 12.9 [3.2]), more likely to report fair or poor health (17.1% vs 6.7%) and difficulty with mobility tasks (46.1% vs 20.2%), and had higher average disease count (mean [SD]: 1.8 [1.1] vs 1.4 [1.0]; of 14), prevalence of obesity (30.9% vs 16.9%), coronary artery disease (21.7% vs 11.2%), and congestive heart failure (4.0% vs 0.4%).
|
Next, we sought to infer patterns of frailty criterion accumulation from frequency distributions of frailty criterion presence at the time of prefrailty or frailty onset. At the time of prefrailty onset, the frequencies of meeting each of weak strength (44%), slow walking speed (23%), and low physical activity (29%) were noticeably higher than for weight loss (14%) and exhaustion (10%). At the time of frailty onset, five times as many women reported exhaustion with onset of frailty than with prefrailty, and there was a two- to threefold increase for the other criteria. Weight loss was the least prevalent condition among women who became frail (Table 2).
|
|
|
Finally, given the heterogeneity in the order by which frailty manifestations initially presented, we evaluated whether the rate of progression to frailty depended on initial manifestations. Compared to the nonfrail women, women who were prefrail at baseline had a threefold higher risk of developing frailty over the 7.5 years of follow-up (hazard ratio [HR] = 3.00, p <.01). However, among the women who were prefrail at baseline, frailty incidence was not differentiated by whether one or two criteria were present (Table 5, Model III), but it did vary depending on initial manifestations. Specifically, women reporting weight loss (exhaustion) at baseline, with or without another criterion present, were 4.4 (3.7) times more likely to become frail than were women without any criterion, after adjusting for baseline age, race, education, and comorbidity (p <.01). Weakness also significantly predicted frailty onset (HR = 2.6, p =.02; Table 5, Model I). Notably, slow walking speed did not. The results were qualitatively similar when comparing women experiencing only one specific criterion to women without any criterion, although the weakness association lost significance (Table 5, Model II).
|
This study presented evidence of a partially hierarchical order in the onset of frailty manifestations over time. Weakness tended to develop first, and was moderately strongly predictive of frailty onset. However, there was heterogeneity in the initial manifestations of frailty, with early development of weight loss or exhaustion predicting more rapid onset of the frailty syndrome. Weight loss and exhaustion, representing mismatches between energy production, utilization, and intake, were frequent tipping points for frailty onset.
That weakness should presage frailty onset is consistent with earlier reports that loss of muscle strength begins in midlife (14–16). Decline in strength has been attributed to the loss of muscle mass and muscle quality referred to as sarcopenia, resulting from anatomic and biochemical changes in the aging muscle (17). The causal mechanisms underlying sarcopenia are many, including oxidative stress, dysregulation of inflammatory cytokines and hormones, malnutrition, physical inactivity, and muscle apoptosis (18,19), all of which have been hypothesized to contribute to frailty through interactive pathways at multiple temporal and spatial scales (20).
The finding of heterogeneity in initial criteria is consistent with the hypothesis that the cycle of frailty may be initiated by insults at many points in a hypothesized cycle of dysregulated energetics (1,9). Notably, it was not the number of early manifestations (i.e., 1 or 2) but the specific manifestations initially present that distinguished the risk and rate of onset of frailty. Women who experienced weight loss and/or exhaustion at baseline had a significantly higher risk of developing frailty compared to women without any criterion, whereas neither slow walking speed nor low activity at baseline was significantly associated with incident frailty. It remains to be determined whether the different patterns of initial accumulation of frailty criteria represent different etiologic pathways with different rates of progression to frailty, either organ-specific or representing systemic physiologic dysregulations of aging. Alternatively, certain of our criterion measures may be more sensitive than others to changes associated with "normal aging," for instance performance-based as opposed to self-reported criteria.
Despite heterogeneous entry points into the cycle of frailty, 80% of transitions to frailty involved adding exhaustion and/or weight loss. This finding raises the possibility that decreased energy production or increased utilization, as in wasting conditions, may be involved in the threshold transition in a final common pathway toward frailty. That weight loss and exhaustion rarely developed alone, but rather co-occurred with other manifestations, is consistent with the reliability theory (21) whereby an emergent aggregation of multiple frailty manifestations would result from depletion of system redundancy or compensatory mechanisms, such that any new deficit leads to failure of the whole organism (22–28). Then, early detection of subclinical changes or deficits at the molecular, cellular, and/or physiologic level would be key to preventing or delaying the development of frailty.
We initially hypothesized that pairwise associations of the five frailty criteria could be unified into a cycle of frailty as in Figure 1 (2,4,5,9,29). However, the observed patterns of aggregation of initial manifestations provided insufficient evidence to link weakness, slowness, and low physical activity in a progressive process within persons. Possible explanations include tied event times resulting from discretely spaced data collection, measures with varying predictive value, or potentially disparate etiologic causes. Longitudinal analyses with specified causal sequence will be needed to test the hypothesis of frailty as a syndrome in which the manifestations accrue in a vicious cycle. The findings reported here are more consistent with a tornado model whereby progression may be initiated by a variety of criteria but tends to culminate in a threshold transition adding exhaustion or weight loss, consistent with a complex nonlinear system.
The clinical utility of these findings lies in the fact that weakness was the most common initial manifestation of the frailty phenotype. It evidenced only moderate predictive validity for incident frailty; however, by our conceptualization the development of frailty is progressive and multisystemic, and any one specific criterion alone, especially at an early stage in the process as in the case of weakness, may be neither sufficient nor specific for frailty prediction. Given that the criterion defining thresholds for grip strength are known to be associated with meaningfully greater risk of adverse outcomes including disability and mortality (30), weakness may nevertheless be a clinically meaningful indicator of increasing vulnerability at a relatively early stage of the frailty process, when preventive intervention could be easiest to implement and theoretically most effective. Although the subsequent or "concurrent" onset of weight loss or exhaustion with the other criteria may better predict frailty onset, by the time someone experiences weight loss or exhaustion, it may be too late to implement frailty interventions. Therefore, consideration should be given to the possible tradeoff between risk prediction and potential for benefits in deciding the proper timing and targets of interventions.
Strengths of our study include its prospective design, long-term follow-up, and its population of initially quite high functioning women. However, there are several limitations. First, despite the 7.5-year follow-up, incidence of frailty was low; hence, the precision for specifying hierarchy among women who developed frailty was limited. Second, the 18-month follow-up interval is too long to observe all individuals' full temporal sequences of events. Third, death or loss to follow-up may also contribute to incomplete observation of incident events in ways leading to underestimation of frailty incidence. Fourth, the reported partial hierarchy may be sensitive to the chosen percentile cutoffs for the frailty criteria. Notably, however, each cutoff has clinical significance based on other studies (30–33). Finally, definitive pathophysiological conclusions cannot be derived from the findings of this study, as this would require analyses relating pathophysiological changes to the frailty criteria. However, characterizing the natural history by which clinical frailty criteria accrue is a crucial first step to the understanding of possible physiologic mechanisms that may contribute to frailty.
Summary
This study, to our knowledge, has presented the first data on the natural history by which commonly used frailty criteria manifest. Taken together, our findings suggest that weakness may serve as a warning sign of increasing vulnerability in early frailty development, and weight loss and exhaustion may help to identify women most at risk for rapid adverse progression. However, it is important to note that it may be premature to start mass screening for weakness before the results of this study can be independently validated by other cohorts, the pathophysiology of frailty is better understood, and the aims and the targets for screening are well-specified.
APPENDIX
The comparison of population-averaged risks of incident frailty criteria faces two analytic challenges. First, frailty status was ascertained at each of the five visits. The resulting incidence data are therefore grouped into four intervals defined by the visits. Because of the discrete nature of these time-to-event data, the discrete-time proportional hazards model (DTPH) (12) was deemed to be more appropriate. Second, to conduct hypothesis testing of the differences in risks of incident criteria, the incident events of different criteria need to be analyzed jointly in a model that takes into account the correlation among the incident events within a person. To address these challenges, we applied the multivariate discrete-time proportional hazards model proposed by Hedeker and colleagues (13) as a multivariate extension of the univariate DTPH model. The model was specified as:
|
|
j are the parameters characterizing the population mean difference in common baseline hazard between interval j (j = 2, 3, 4) and the first interval (as reference), therefore allowing the baseline hazard to vary over intervals of different length; and 
0(t) is the probability of experiencing weakness in the first interval given event-free at the beginning of the interval (i.e., discrete hazard) when 
ij = 0. Terms ij are random, subject-specific effects with 0 mean that allow for within-subject correlations among incidences of the multiple frailty criteria. The model assumes constant population average difference in baseline hazard j across frailty criteria.
A
For data acquisition, management, and analysis, this research was supported, in part, by the Johns Hopkins Older Americans Independence Center under contracts P30-AG02133 and R37-AG19905 from the National Institute on Aging, National Institutes of Health.
F
Decision Editor: Luigi Ferrucci, MD, PhD
Received August 9, 2007
Accepted January 6, 2008
References
| ||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| All GSA journals | The Gerontologist |
| Journals of Gerontology Series B: Psychological Sciences and Social Sciences | |
