ORIGINAL CONTRIBUTION

Gait Speed and Survival in Older Adults

Stephanie Studenski, MD, MPH Context Survival estimates help individualize goals of care for geriatric patients, but
Subashan Perera, PhD life tables fail to account for the great variability in survival. Physical performance mea-
sures, such as gait speed, might help account for variability, allowing clinicians to make
Kushang Patel, PhD
more individualized estimates.
Caterina Rosano, MD, PhD
Objective To evaluate the relationship between gait speed and survival.
Kimberly Faulkner, PhD
Design, Setting, and Participants Pooled analysis of 9 cohort studies (collected
Marco Inzitari, MD, PhD between 1986 and 2000), using individual data from 34 485 community-dwelling older
Jennifer Brach, PhD adults aged 65 years or older with baseline gait speed data, followed up for 6 to 21
years. Participants were a mean (SD) age of 73.5 (5.9) years; 59.6%, women; and
Julie Chandler, PhD 79.8%, white; and had a mean (SD) gait speed of 0.92 (0.27) m/s.
Peggy Cawthon, PhD Main Outcome Measures Survival rates and life expectancy.
Elizabeth Barrett Connor, MD Results There were 17 528 deaths; the overall 5-year survival rate was 84.8% (confi-
Michael Nevitt, PhD dence interval [CI], 79.6%-88.8%) and 10-year survival rate was 59.7% (95% CI, 46.5%-
70.6%). Gait speed was associated with survival in all studies (pooled hazard ratio per 0.1
Marjolein Visser, PhD m/s, 0.88; 95% CI, 0.87-0.90; P!.001). Survival increased across the full range of gait
Stephen Kritchevsky, PhD speeds, with significant increments per 0.1 m/s. At age 75, predicted 10-year survival across
Stefania Badinelli, MD the range of gait speeds ranged from 19% to 87% in men and from 35% to 91% in wom-
en. Predicted survival based on age, sex, and gait speed was as accurate as predicted based
Tamara Harris, MD on age, sex, use of mobility aids, and self-reported function or as age, sex, chronic con-
Anne B. Newman, MD ditions, smoking history, blood pressure, body mass index, and hospitalization.
Jane Cauley, PhD Conclusion In this pooled analysis of individual data from 9 selected cohorts, gait
speed was associated with survival in older adults.
Luigi Ferrucci, MD, PhD
JAMA. 2011;305(1):50-58 www.jama.com
Jack Guralnik, MD, PhD

R
EMAINING YEARS OF LIFE VARY among the older adults.14 The purpose of adults with gait speed data at baseline,
widely in older adults, and phy- this study is to evaluate the association of monitored survival for at least 5 years.
sicians should consider life ex- gaitspeedwithsurvivalinolderadultsand Analyses performed herein were con-
pectancy when assessing goals to determine the degree to which gait ducted in 2009 and 2010. All studies
ofcareandtreatmentplans.1 However,life speed explains variability in survival af- required written informed consent and
expectancy based on age and sex alone ter accounting for age and sex. institutional review board approval.
provideslimitedinformationbecausesur- METHODS Populations
vivalisalsoinfluencedbyhealthandfunc-
Overview All studies recruited community-
tional abilities.2 There are currently no
well-established approaches to predict- We used individual participant data dwelling older adults. Although some
inglifeexpectancythatincorporatehealth from 9 cohort studies, baseline data for sought representative samples,8,15,20,23
and function, although several models which were collected between 1986 and others focused on healthier partici-
havebeendevelopedfromindividualdata 2000 (TABLE 1).8,15,16,18-23 Each study, pants,16,17 single sex,19,22 or older adults
sources.3-5 Gait speed, also often termed which included more than 400 older from primary care practices.21 Only
walking speed, has been shown to be as-
Author Affiliations: Department of Medicine, Divi- Autonomous University of Barcelona, Barcelona, Spain
sociated with survival among older adults sion of Geratric Medicine, School of Medicine (Drs (Dr Inzitari); Merck Research Laboratories, North Wales,
in individual epidemiological cohort Studenski and Perea), Department of Epidemiology, Pennsylvania (Dr Chandler); California Pacific Medi-
School of Public Health (Drs Rosano, Newman, and cal Center Research Institute (Dr Cawthon), and Uni-
studies6-12 and has been shown to reflect Cauley), Department of Physical Therapy, School of versity of California at San Francisco (Dr Nevitt), Uni-
health and functional status.13 Gait speed Health and Rehabilitation (Dr Brach), University of Pitts- versity of California at San Diego (Dr Barrett Connor);
burgh, and National Personal Protective Technology VU University, Amsterdam, the Netherlands (Dr Vis-
has been recommended as a potentially Laboratory (Dr Faulkner), Pittsburgh, Pennsylvania; ser); Wake Forest University, Winston-Salem, North
useful clinical indicator of well-being Laboratory of Epidemiology, Demography, and Bi- Carolina (Dr Kritchevsky); and Geriatric Unit, Azienda
ometry (Drs Patel, Harris, and Guralnik), and Clinical sanitaria firenze, Florence, Italy (Dr Badinelli).
Research Branch, Intramural Research Program (Dr Fer- Corresponding Author: Stephanie Studenski, MD,
For editorial comment see p 93. rucci), National Institute on Aging, Bethesda, Mary- MPH, Kaufmann Bldg, Ste 500, 3471 Fifth Ave, Pitts-
land; Pere Virgili Hospital and Institute on Aging, burgh, PA 15143 (sas33@pitt.edu).

50 JAMA, January 5, 2011—Vol 305, No. 1 (Reprinted) ©2011 American Medical Association. All rights reserved.

Downloaded From: http://jama.jamanetwork.com/ on 06/07/2014

 GAIT SPEED AND SURVIVAL IN OLDER ADULTS

Table 1. Characteristics of Participants in the 9 Cohort Studies

No. (%) of Participants by Study

Osteoporotic Study of
Health, Hispanic Invecciare Fractures NHANES Osteoporotic
Study CHS 8
EPESE 15
ABC16,17 EPESE8 in Chianti18 in Men19 III20 PEP 21
Fractures22
Sample size, No. 5801 2128 3048 1905 972 5833 3958 491 10 349
Women 3336 (57.51) 1404 (65.98) 1575 (51.67) 1098 (57.64) 541 (55.66) 0 2044 (51.64) 216 (43.99) 10 349 (100)
Race/ethnicity
White 4854 (83.68) 2126 (99.91) 1783 (58.50) 0 972 (100) 5223 (89.54) 2535 (64.05) 394 (80.24) 9662 (93.36)
Black 909 (15.67) 2 (0.09) 1265 (41.50) 0 0 235 (4.03) 699 (17.66) 89 (18.13) 654 (6.32)
Hispanic 0 0 0 1905 (100) 0 122 (2.09) 623 (15.74) 0 0
Other 38 (0.66) 0 0 0 0 253 (4.34) 101 (2.55) 8 (1.63) 33 (0.32)
Age mean (SD), y 72.81 (5.58) 78.85 (5.52) 73.62 (2.87) 74.74 (5.96) 74.58 (7.08) 73.61 (5.84) 75.17 (6.93) 74.08 (5.74) 71.81 (5.21)
Age group, y
65-74 3852 (66.40) 559 (26.27) 1912 (62.73) 1083 (56.85) 555 (57.10) 3401 (58.31) 2033 (51.36) 279 (56.82) 7486 (72.34)
75-84 1732 (29.86) 1204 (56.58) 1136 (37.27) 668 (35.07) 302 (31.07) 2183 (37.42) 1484 (37.49) 188 (38.39) 2596 (25.08)
"85 217 (3.74) 365 (17.15) 0 154 (8.08) 115 (11.83) 249 (4.27) 441 (11.14) 24 (4.89) 200 (1.93)
Missing 0 0 0 0 0 0 0 0 67 (0.65)
Gait speed, mean 0.86 (0.22) 0.83 (0.13) 1.12 (0.23) 0.56 (0.23) 1.00 (0.28) 1.19 (0.23) 0.68 (0.23) 0.88 (0.24) 0.95 (0.22)
(SD), m/s
Gait speed class, m/s
!0.4 149 (2.57) 0 4 (0.13) 515 (27.03) 35 (3.60) 11 (0.19) 480 (12.13) 20 (4.07) 33 (1.19)
"0.4 to !0.6 526 (9.07) 78 (3.67) 20 (0.66) 621 (32.60) 59 (6.07) 54 (0.93) 897 (22.66) 40 (8.15) 466 (4.50)
"0.6 to !0.8 1887 (32.53) 791 (37.17) 189 (6.20) 467 (24.51) 110 (11.32) 206 (3.53) 1368 (34.56) 110 (22.40) 1752 (16.93)
"0.8 to !1.0 2076 (35.79) 1105 (51.93) 705 (23.13) 220 (11.55) 246 (25.31) 875 (15.00) 887 (22.41) 166 (33.81) 3768 (36.41)
"1.0 to !1.2 1077 (18.57) 135 (6.34) 1093 (35.86) 77 (4.04) 305 (31.38) 1774 (30.41) 294 (7.43) 116 (22.63) 3054 (29.51)
"1.2 to !1.4 0 17 (0.80) 684 (22.44) 4 (0.21) 170 (17.49) 1911 (32.76) 32 (0.81) 36 (7.33) 970 (9.37)
"1.4 86 (1.48) 2 (0.09) 353 (11.58) 1 (0.05) 47 (4.84) 1002 (17.18) 0 3 (0.61) 217 (2.10)
Mobility aid use
None NA 1962 (92.20) 3048 (100) 1817 (95.38) 881 (90.64) 5792 (99.30) 3664 (92.57) 463 (94.11) 10 165 (98.22)
Cane NA 87 (4.09) 0 49 (2.57) 8 (0.82) 38 (0.65) 201 (5.08) 21 (4.27) All aids
Walker NA 67 (3.15) 0 23 (1.21) 3 (0.31) 0 74 (1.87) 5 (1.01) Combined
Other/missing NA 12 (0.56) 0 16 (0.84) 80 (8.23) 3 (0.05) 19 (0.48) 3 (0.61) 184 (1.78)
BMI, mean (SD) 26.68 (4.71) 26.63 (4.64) 27.40 (4.82) 27.91 (5.13) 27.51 (4.11) 27.39 (3.83) 26.66 (5.11) 27.53 (5.12) 26.61 (4.57)
BMI category
!25 2237 (38.65) 803 (38.24) 983 (32.25) 555 (29.13) 276 (28.40) 1593 (27.31) 1544 (39.01) 156 (31.77) 4352 (42.05)
25-30 2407 (41.49) 886 (42.64) 1288 (42.26) 758 (39.79) 437 (46.96) 2991 (51.28) 1559 (39.39) 211 (42.97) 3842 (37.12)
#30 1144 (19.72) 411 (19.31) 777 (25.49) 577 (30.29) 243 (25.00) 1247 (21.38) 852 (21.53) 123 (25.05) 2155 (20.82)
Missing 13 (0.22) 28 (1.32) 0 15 (0.79) 16 (1.65) 2 (0.03) 3 (0.08) 1 (0.20) 0
Hospitalized NA 395 (18.57) 456 (14.98) 304 (15.96) 129 (13.27) NA 775 (19.58) 97 (19.76) 1116 (11.51)
past year
Diseases
Cancer 830 (14.33) 486 (22.84) 575 (18.91) 115 (6.04) 95 (9.77) 1697 (29.09) 387 (9.78) 113 (23.01) NA
Arthritis 2977 (51.94) 2055 (96.57) 1706 (56.72) 812 (42.62) 304 (31.31) 2764 (47.39) 1827 (46.16) 286 (58.25) 6003 (63.10)
Diabetes 690 (11.90) 335 (15.74) 453 (14.88) 455 (23.88) 106 (10.91) 624 (10.70) 607 (15.34) 84 (17.11) 681 (7.04)
Heart disease 1230 (21.20) 312 (14.66) 652 (22.03) 155 (8.14) 49 (5.05) 1379 (23.64) 484 (12.23) 89 (18.13) NA
Self-reported health 2177 (37.61) 542 (74.48) 1343 (44.12) 870 (45.67) 591 (62.61) 5012 (85.95) 1204 (30.47) 229 (46.64) 8537 (82.49)
excellent/
very good
Total deaths during 3851 (66.39) 1955 (91.87) 848 (27.82) 972 (51.02) 187 (19.24) 1073 (18.40) 2837 (71.68) 293 (59.55) 5512 (53.26)
follow-up
Median survival years 13.25 9.57 NE 11.70 NE NE 9.86 11.15 17.23
(95% CI) (13.00-13.56) (9.17-9.92) (11.11-NE) (9.53-10.19) (9.82-11.92) (16.97-17.47)
Follow-up period, 13.25 9.57 9.00 11.54 6.00 6.84 9.86 11.15 15.03
median (range), y (0.01-18.06) (0.10-20.65) (0.02-9.00) (0.07-12.29) (0.18-6.00) (0.04-8.26) (0.08-17.75) (0.12-13.76) (0.02-21.00)
Length of walk 15 feet 8 feet 6m 8 feet 4m 6m 4m 4m 6m
Year of baseline data 1989-90, 1987-1989 1997-1998 1995-1996 1998-2000 2000-2002 1988-1994 1996 1986-1988,
collection 1992-93 1997
Year of most recent 2007 2008 2007 2007 2006 2008 2006 2010 2008
mortality follow-up
Abbreviations: ABC, Aging and Body Composition; BMI, body mass index, calculated as weight in kilograms divided by height in meters squared; CHS, Cardiovascular Health Study; CI,
confidence interval; EPESE, Established Populations for the Epidemiological Study of the Elderly; NA, not applicable; NE, not estimable due to insufficiently long follow-up and resulting
in low mortality rate less than or close to 50%; NHANES III, Third National Health and Nutrition Examination Survey; PEP, Predicting Elderly Performance.

©2011 American Medical Association. All rights reserved. (Reprinted) JAMA, January 5, 2011—Vol 305, No. 1 51

Downloaded From: http://jama.jamanetwork.com/ on 06/07/2014

 GAIT SPEED AND SURVIVAL IN OLDER ADULTS

participants 65 years and older with mous variable reflecting difficulty in in- ordered 0.2-m/s gait speed categories.
baseline gait speed data were included strumental ADLs based on report of To examine the influence of early
in this study. Individual study goals, re- difficulty or dependence with shopping, deaths, we repeated analyses exclud-
cruitment methods, and target popu- meal preparation, or heavy housework ing deaths within 1 year of gait speed
lations have been published.8,15-23 due to a health or physical problem. Par- measurement and moved up the 0 time
ticipants were then classified into 1 of 3 for survival assessment (results were
Measures groups; dependent in ADLs, difficulty similar; eTable 1 available at http://www
Gait speed was calculated for each par- with instrumental ADLs, or independent. .jama.com). Subgroup analyses were re-
ticipant using distance in meters and Physical activity data were collected in 6 peated in strata by age (65-74, 75-84,
time in seconds. All studies used instruc- studies, but time frames and items varied or "85 years), sex, race, self-reported
tions to walk at usual pace and from a widely. Two studies used the Physical Ac- health status, smoking history, BMI,
standing start. The walk distance varied tivity Scale for the Elderly (PASE).27 We functional status, use of mobility aids,
from 8 ft to 6 m. For 8 ft, we converted dichotomizedthePASEscoreat100.28 We and hospitalization and by report of
to 4-m gait speed by formula.24 For 6 m, created operational definitions of other cancer, arthritis, diabetes, and heart dis-
we created a conversion formula (4-m covariatesthatwerereasonablyconsistent ease.29 Results were pooled across sex
speed=−0.0341$(6-m speed)%0.9816 across studies. Covariates were identical because no substantial sex differences
with R2 =0.93, based on a cohort of 61 in- forheight,weight,BMI,andsystolicblood existed in HRs within subgroup strata.
dividuals with concurrent 4- and 6-m pressure. Hospitalization within the prior To obtain simple and clinically us-
walks). For 15 feet (4.57 m),23 speed was yearwasdeterminedlargelybyself-report, able estimates of survival probability
simply meters divided by time. Where andchronicconditionswerebyself-report based on sex, age, and gait speed, we fit
available, data on fast gait speed (walk as of physician diagnosis, with heart disease logistic regression models separately for
fast as comfortably able25) and the Short encompassingangina,coronaryarterydis- each sex with dichotomized 5- and 10-
Physical Performance Battery were ob- ease, heart attack, and heart failure. year survival as the response variable and
tained.26 Survival for each individual used age, gait speed, and their interaction as
study monitoring methods, including the Statistical Analysis continuous predictors. To obtain esti-
National Death Index and individual Descriptive statistics summarized par- mates of median survival (further life ex-
study follow-up. Time from gait speed ticipantcharacteristics,follow-upperiod, pectancy), we fit Weibull accelerated fail-
baseline to death was calculated in days. and median survival from baseline. A ure–time models separately for each with
Five-year survival status was confirmed study-wide a priori P value of .002 pro- time to death as the response variable,
for more than 99% of participants. vides a conservative Bonferroni correc- and age, gait speed, and their interac-
Additional variables include sex, age, tion accounting for at least 25 individual tion as continuous predictors. To com-
race/ethnicity (white, black, Hispanic, statistical comparisons. Kaplan-Meier pare ability to predict survival among
other,definedbyparticipant),height(cen- product-limit survival curves graphically candidate variables and to determine
timeters), weight (kilograms), body mass summarize lifetimes for each gait speed whether gait speed improves predictive
index (BMI), calculated as weight in ki- category.29 For graphical purposes, gait accuracy beyond other clinical mea-
lograms divided by height in meters speed was categorized into 0.2-m/s in- sures, we fit logistic regression models
squared (!25, 25-30, and #30), smok- crements with lower and upper extremes with dichotomized 5-year or 10-year sur-
ing (never, past, current), use of mobil- being grouped as less than 0.4 m/s and vival as the response variable and vari-
ityaids(none,cane,walker),systolicblood higher than 1.4 m/s. ous combinations of predictors as inde-
pressure, self-reports of health (excellent Cox proportional hazards regres- pendent variables with both linear and
or very good vs good, fair, or poor), hos- sion models were used to assess asso- squared terms for BMI. The area under
pitalization in the past year (yes/no), and ciations between gait speed and sur- the receiver operating characteristic
physician-diagnosed medical conditions vival, adjusting for age at baseline, for (ROC) curve or C statistic was used as
(cancer, arthritis, diabetes, and heart dis- which hazard ratios (HRs) correspond a measure predictive of accuracy for mor-
ease,allyes/no).Measuresofself-reported to a 0.1-m/s difference in gait speed. tality. All study-specific statistical analy-
functional status were not collected in all The analyses were repeated adjusting ses were performed using SAS version 9.2
studies and varied in content and form. for height, sex, race, BMI, smoking (SAS Institute Inc, Cary, North Carolina).
Wecreatedadichotomousvariablereflect- history systolic blood pressure, dis- Age-adjustedHRswerepooledfromall
ing dependence in basic activities of daily eases, prior hospitalization, and self- studies using standard meta-analytic sta-
living(ADLs)basedonreportofbeingun- reported heath. Proportionality of tistical methodology. Heterogeneity of
able or needing help from another person hazards was verified by examining HRs across studies was assessed using the
to perform any basic activity, including Schoenfeld residual plots.30 Appropri- Q and I2 statistics.31,32 We used a random-
eating, toileting, hygiene, transfer, bath- ateness of using gait speed as a con- effects model to appropriately pool the
ing, and dressing. For individuals inde- tinuous predictor was confirmed by ob- HRs on the log scale while incorporating
pendent in ADLs, we created a dichoto- serving linearity in Cox models with any heterogeneity among study estimates
52 JAMA, January 5, 2011—Vol 305, No. 1 (Reprinted) ©2011 American Medical Association. All rights reserved.

Downloaded From: http://jama.jamanetwork.com/ on 06/07/2014

 GAIT SPEED AND SURVIVAL IN OLDER ADULTS

andthentransformbacktoobtainanover-
Figure 1. Age-Adjusted Hazard Ratio for Death per 0.1-m/s Higher Gait Speed
all HR, along with a 95% confidence in-
terval (CI) and P value.33 Sensitivity of the No. of
Deaths
Total Sample
Size
results was assessed by fitting a shared Cardiovascular Health 3851 5801
frailty34 (unrelated to the geriatric syn- Study,22 1991
Established Populations for the 1955 2128
dromefrailty)modeltoindividualpartici- Epidemiologies Study of the Elderly,23 1985
pant data with a &-distributed frailty pa- Health, Aging, and Body Composition 848 3048
rametertoaccountforstudyeffect(results Study,11,12 2009, 2005
Hispanic Established Populations for 972 1905
similar; not shown).34,35 Five- and 10-year Epidemiological Study of the Elderly,13 1999
pointwise survival rates from the Kaplan- Invecciare in Chianti,17 2000 187 972
Meier curves for each sex, age-group, and Osteoporotic Fractures in Men,20 2005 1073 5833
gait speed category combination were Third National Health and Nutrition 2837 3958
Examination Study,21 2004
pooled across studies using a random-
Predicting Elderly Performance,28 2003 293 491
effects model on the complementary log- Study of Osteoporotic Fractures,26 1990 5512 10 349
logscale36 andthenappropriatelyinverted
to obtain overall estimates of survival, as Pooled (random effects)
Pooled (shared frailty model)
presented in the tables. We further used
thestandardrandomeffectsmeta-analytic 0.7 1.0 2.0
model to combine sex-specific regression Adjusted Hazard Ratio
coefficients for age, gait speed, and their The size of the data markers is proportional to the square root of the number of participants. The error bars
interaction from logistic regression mod- indicate 95% confidence intervals. The Q statistic for heterogeneity is 45.2 (P!.001; I2, 82.3). Pooled using
els for 5- and 10-year survival and used random effects and shared frailty models.

theoverallestimatestoconstructclinically
usable survival probability nomograms; substantial numbers of African American tional status (independent, difficulty
combine sex-specific regression coeffi- (n=3852) and Hispanic (n=2650) par- with instrumental ADLs, and depen-
cients for age, gait speed, and their inter- ticipants. The studies had a wide age dent in ADLs), the pooled HR per 0.1-
action from accelerated failure time mod- range, including 1765 persons older m/s increase in gait speed for those who
els for time to death and used the overall than 85 years. Similarly, there was a wide were independent was 0.92 (P =.005),
estimatestoconstructclinicallyusablelife- range of gait speeds, from less than 0.4 for those with difficulty in instrumen-
expectancy nomograms; and combine m/s (n=1247) to more than 1.4 m/s tal activities was also 0.92 (P!.001) but
areas under ROC curves obtained from (n=1491). Study follow-up time ranged was 0.94 (P = .02) among those depen-
9 studies. An increase of 0.025 in overall from 6.0 to 21.0 years, with participants dent in ADLs. Because physical activ-
area under ROC curve was interpreted as followed up for a mean of 12.2 and a me- ity measures were not sufficiently con-
clinicallyrelevantbetteraccuracy.37 Toap- dianof13.8years.Therewere17 528total sistent across studies, effects could not
propriatelycombineentiresurvivalcurves deaths across all studies, with rates vary- be pooled. The Osteoporotic Frac-
across the 9 studies, we used the gener- ing from 18.40% to 91.87% in individual tures in Men (MrOS)19 and Hispanic Es-
alizedleastsquaresmethodforjointanaly- studies. Mortality rates appear to be re- tablished Populations for Epidemio-
sisofsurvivalcurves.38 Weusedarandom- lated to length of follow-up (Table 1). logic Studies of the Elderly (EPESE)8
effects model with weights obtained by To assess consistency across studies, used the Physical Activity Scale for the
inverseofthevarianceofthesurvivalfunc- risk of death was estimated per 0.1-m/s Elderly (PASE). When dichotomized at
tionatthemedianlifetimestopooltheme- higher gait speed. Age-adjusted HRs by a score of 100 into low and high activ-
diansurvivaltimesforeachsex,agegroup, study ranged from 0.83 to 0.94 and all ity, MrOS had consistent and statisti-
and gait speed category. We used Com- were significant (P!.001; FIGURE 1). We cally significant HRs for low (HR, 0.85;
prehensiveMetaAnalysisversion2.2(Bio- also examined the survival HRs for gait 95% CI, 0.81-0.88) and high (HR, 0.87;
stat Inc, Englewood, New Jersey) for all speed by study in subgroups, including 95% CI, 0.84-0.90) physical activity. In
meta-analytic methods and Stata SE 8 age, sex, race/ethnicity, BMI, smoking the Hispanic EPESE, the HR for low
(StataCorp,CollegeStation,Texas)forfit- history, use of mobility aids, prior hos- physical activity was significant (0.92;
ting shared frailty models. pitalization, self-reported health, func- 95% CI, 0.88-0.96) but the HR for
tional status, and selected chronic dis- higher physical activity was not (0.99;
RESULTS eases. There were consistent associations 95% CI, 0.95-1.04).Pooled HRs for all
The 9 participating studies contributed across studies, although given the large subgroups except functional status were
a total of 34 485 participants (Table 1). sample sizes, Q statistics were often sta- consistently in the range of 0.81 to 0.92
Although most studies included men and tistically significant (details available in and all were significant (P !.002).
women, 2 were sex specific.19,22 Of the eFigure 1A-M available at http://www The overall HR for survival per each
total, 59.6% were women. There were .jama.com). For the 3 levels of func- 0.1 m/s faster gait speed was 0.88 (95%
©2011 American Medical Association. All rights reserved. (Reprinted) JAMA, January 5, 2011—Vol 305, No. 1 53

Downloaded From: http://jama.jamanetwork.com/ on 06/07/2014

 GAIT SPEED AND SURVIVAL IN OLDER ADULTS

CI, 0.87-0.90; P!.001) when pooled sented with age as a continuous vari- tall participants, so results presented are
across all studies using a random- able). Gait speed was associated with dif- not stratified by height. Stratification by
effects meta-analytic statistical ap- ferences in the probability of survival at race/ethnicity (non-Hispanic white,
proach (Figure 1 and eFigure 1 avail- all ages in both sexes, but was espe- black, Hispanic) suggested generally
able at http://www.jama.com). Further cially informative after age 75 years. In similar survival rates by gait speed among
adjustment for sex, BMI, smoking sta- men, the probability of 5-year survival age and sex groups. Confidence inter-
tus, systolic blood pressure, diseases, at age 85 ranged from 0.3 to 0.88 (eFig- vals were often wide. In some subsets of
prior hospitalization, and self-reported ure 3A) and the probability of 10-year slow walkers of Hispanic descent, sur-
health did not change the results (over- survival at age 75 years ranged from 0.18 vival rates were 10% to 20% higher than
all HR, 0.90; 95% CI, 0.89-0.91; to 0.86 (eFigure 4A). In women, the in other groups (eTable 2).
P!.001). Using data from all studies, we probability of 5-year survival remained We also used our analyses to estimate
created for each sex, 5- and 10-year sur- greater than 0.5 until advanced age (eFig- median years of remaining life based on
vival tables (TABLE 2, data derived from ure 3B), but 10-year survival at age 75 sex, age, and gait speed. (FIGURE 2, pre-
pooled Kaplan-Meier estimates evalu- years ranged from 0.34 to 0.92 and at age dicted survival data are based on an
ated at 5 and 10 years, presented in 3 age 80 years from 0.22 to 0.86 (eFigure 4B). accelerated failure time model with
groups) and graphs (eFigure 3 and eFig- Stratification by sex-specific median Weibull distribution, with age as a con-
ure 4 predicted survival based on pooled height failed to show systematic differ- tinuous variable, and eTable 3, data are
logistic regression coefficients, data pre- ences in survival rates between short and derived from pooled Kaplan-Meier es-

Table 2. Five- and 10-Year Survival in Men and Women by Age and Gait Speed Group
5-Year Survival (95% CI), % a 10-Year Survival (95% CI), %

Men Women Men Women

Gait Speed, Age Age Age Age Age Age Age Age
m/s 65-74 75-84 Age "85 65-74 75-84 Age "85 65-74 75-84 Age "85 65-74 75-84 Age "85
Speed !0.4 68 (47-82) 60 (38-76) 25 (15-36) 80 (71-86) 69 (58-78) 47 (40-54) 56 (23-80) 15 (4-33) 8 (3-18) 58 (46-69) 35 (24-47) 11 (5-19)
"0.4 to !0.6 77 (72-81) 57 (49-64) 31 (24-39) 88 (85-90) 75 (68-80) 61 (50-70) 53 (41-64) 23 (15-31) 6 (3-11) 67 (61-72) 42 (36-48) 18 (9-30)
"0.6 to !0.8 79 (74-83) 65 (57-71) 49 (35-61) 91 (89-93) 82 (78-86) 74 (69-78) 57 (52-62) 31 (24-38) 11 (3-28) 74 (71-77) 52 (46-57) 23 (18-28)
"0.8 to !1.0 85 (82-88) 75 (69-79) 54 (43-64) 93 (91-95) 89 (86-91) 73 (59-83) 67 (62-71) 43 (36-50) 14 (7-25) 80 (75-83) 62 (56-68) 39 (22-56)
"1.0 to !1.2 90 (85-93) 83 (76-87) 68 (57-77) 96 (94-98) 91 (87-94) 61 (35-79) 69 (63-74) 53 (46-59) 50 (6-84) 86 (82-89) 73 (70-77) 33 (13-54)
"1.2 to !1.4 93 (86-96) 85 (79-89) 62 (46-74) 96 (94-97) 93 (87-96) 67 (5-95) 75 (40-91) 51 (16-78) NE 83 (38-96) 80 (72-86) NE
Speed "1.4 95 (89-97) 93 (86-96) 91 (51-99) 97 (94-99) 95 (72-99) NE 93 (81-98) 50 (6-84) NE 87 (71-95) 92 (71-98) NE
All gait 87 (82-91) 74 (65-81) 46 (39-53) 93 (91-94) 84 (80-87) 64 (58-70) 62 (58-66) 36 (30-42) 10 (8-13) 77 (71-82) 54 (46-60) 22 (15-29)
speeds
Abbreviations: CI, confidence interval; NE, not estimable due to small number of participants in categories.
a Survival estimates are derived from individual study Kaplan-Meier survival estimates that are pooled across studies using random-effects models with inverse variance weighting.

Figure 2. Predicted Median Life Expectancy by Age and Gait Speed

Men Women
Gait
45 45 speed, m/s
1.6
40 40
1.5
Gait
35 speed, m/s 35 1.4

1.6 1.3

30 30 1.2
Median Survival, y

Median Survival, y

1.5
1.1
1.4
1.0
25 1.3 25
0.9
1.2 0.8
20 1.1 20 0.7
1.0 0.6
0.9 0.5
0.4
15 0.8
0.7
15 0.3
0.6 0.2
0.5
0.4
10 0.3 10
0.2

5 5

0 0
65 70 75 80 85 90 95 65 70 75 80 85 90 95
Age, y Age, y

A PDF of enlarged graphs is available at http://www.jama.com.

54 JAMA, January 5, 2011—Vol 305, No. 1 (Reprinted) ©2011 American Medical Association. All rights reserved.

Downloaded From: http://jama.jamanetwork.com/ on 06/07/2014

 GAIT SPEED AND SURVIVAL IN OLDER ADULTS

timates evaluated at 5 and 10 years in (Figure 2; a PDF of enlarged graphs is To compare the 5-year survival pre-
3 age groups.) In the pooled sample, available at http://www.jama.com). Gait dictiveabilitybetweendemographicsand
median survival in years for the age speeds of 1.0 m/s or higher consis- gait speed vs other combinations of vari-
groups 65 through 74 years was 12.6 tently demonstrated survival that was ables,weusedareasundertheROCcurve
for men and 16.8 for women; for 75 longer than expected by age and sex (C statistics) in logistic regression mod-
through 84 years, 7.9 for men and 10.5 alone. In this older adult population, els for individual studies and pooled
for women; and for 85 years or older, the relationship of gait speed with re- across studies (TABLE 3). Gait speed
4.6 for men and 6.4 years for women maining years of life was consistent added substantially37 to age and sex in
(eTable 3 available at http://www.jama across age groups, but the absolute 7 of the 9 studies and in the pooled analy-
.com). Predicted years of remaining life number of expected remaining years of sis. C statistics for age, sex, and gait speed
for each sex and age increased as gait life was larger at younger ages. For 70- were greater than those for age, sex, and
speed increased, with a gait speed of year-old men, life expectancy ranged chronic diseases in 4 of 9 studies, ap-
about 0.8 m/s at the median life expec- from 7 to 23 years and for women, from proximately equivalent in 5 studies and
tancy at most ages for both sexes 10 to 30 years. inferior in no studies. C statistics for age,

Table 3. Predictive Accuracy for 5- and 10-Year Survival by Individual Study and Pooled Data Presented as Area Under the Receiver
Operating Characteristic Curves
C Statistic (95% Confidence Interval)

Osteoporotic Study of
Outcome and Health, Hispanic Invecciare Fractures Osteoporotic
Predictors CHS8 EPESE15 ABC16,17 EPESE8 in Chianti18 in Men19 NHANES III20 PEP21 Fractures22 Pooled
5-Year Mortality
Age, sex 0.705 0.685 0.606 0.694 0.797 0.700 0.710 0.674 0.646 0.690
(0.685-0.725) (0.658-0.712) (0.575-0.637) (0.662-0.725) (0.754-0.841) (0.677-0.723) (0.691-0.729) (0.616-0.732) (0.625-0.667) (0.662-0.717)
Age, sex, 0.711 0.692 0.616 0.703 0.793 0.704 0.719 0.694 0.662 0.698
diseases (0.692-0.731) (0.665-0.719) (0.586-0.647) (0.671-0.725) (0.747-0.838) (0.681-0.727) (0.700-0.737) (0.737-0.750) (0.639-0.684) (0.673-0.723)
Age, sex, 0.736 0.702 0.650 0.728 0.808 0.728 0.744 0.728 0.665 0.719
diseases, (0.717- (0.676- (0.620- (0.698- (0.765- (0.706- (0.727- (0.674- (0.643- (0.693-
BMI, 0.755) b,c 0.728) 0.680) b,c 0.755) b,c 0.850) 0.749) b 0.762) b,c 0.781) b,c 0.686) 0.745) b
systolic
BP, prior
hospitalization
Age, sex, use NA NA NA 0.735 0.803 NA 0.738 0.720 NA 0.747
of mobility (0.705-0.765) (0.756-0.851) (0.720-0.757) (0.663-0.776) (0.720-0.774)
aid, functional
status a
Age, sex, gait 0.734 0.711 0.642 0.710 0.803 0.729 0.737 0.718 0.682 0.717
speed (0.716- (0.685- (0.612- (0.679-0.741) (0.760- (0.707- (0.719- (0.664- (0.662- (0.694-0.740)
0.753) c 0.737) c 0.673) b,c 0.846) 0.751) b,c 0.755) b 0.771) c 0.703) b,c 0.741
(0.706-0.775) d
10-Year Mortality
Age, sex 0.721 0.725 NA 0.700 NA NA 0.741 0.674 0.689 0.712
(0.707-0.734) (0.704-0.746) (0.677-0.724) (0.726-0.757) (0.627-0.721) (0.676-0.703) (0.692-0.731)
Age, sex, 0.728 0.738 NA 0.709 NA NA 0.749 0.698 0.706 0.724
diseases (0.715-0.742) (0.716-0.759) (0.685-0.733) (0.734-0.764) (0.652-0.744) (0.692-0.719) (0.707-0.740)
Age, sex, 0.745 0.749 NA 0.733 NA NA 0.768 0.723 0.709 0.739
diseases, (0.732- (0.729- (0.710- (0.754- (0.678- (0.696- (0.719-
BMI, systolic 0.759) 0.770) 0.756) b 0.783) b 0.727) b,c 0.722) 0.759) b
BP, prior
hospitalization
Age, sex, NA NA NA 0.722 NA NA 0.761 0.702 NA 0.732
functional (0.699-0.746) (0.746-0.776) (0.655-0.748) (0.698-0.767)
status,
walking aid
use a
Age, sex, gait 0.740 0.753 NA 0.709 NA NA 0.766 0.723 0.719 0.737
speed (0.727 (0.733- (0.685- (0.751- (0.679- (0.706- (0.718-0.755) b,
-0.754) 0.774) b 0.732) 0.780) b 0.768) b,c 0.731) b 0.734
(0.692-0.777) d
Abbreviations: ABC, Aging and Body Composition; BMI, body mass index, calculated as weight in kilograms divided by height in meters squared; BP, blood pressure; CHS, Cardiovas-
cular Health Study; CI, confidence interval; EPESE, Established Populations for the Epidemiological Study of the Elderly; NA, not applicable; NHANES III, Third National Health and
Nutrition Examination Survey; PEP, Predicting Elderly Performance.
a Functional status was operationally defined for 3 levels: (1) activities of daily living (ADLs) dependence is defined as report of needing help from another person or being unable to perform
any of 6 basic ADLs, 2) Instrumental ADL difficulty is defined as report of no ADL dependence but difficulty performing shopping, meal preparation, or heavy housework, and (3) In-
dependent is defined as no report of ADL dependence or instrumental ADL difficulty.
b Value is the pooled estimate of the area under the receiver operator characteristic curve for age, sex, and gait speed for the studies that were used in the comparisons of gait speed with
use of mobility aids and functional status. Four studies were included in the estimates of 5 y mortality and three in the estimates of 10 y mortality. Values are reported as the C statistic
representing area under the receiver operator characteristic curve; values that differ by 0.025 or more are considered substantially different.37
c C statistic is greater than for age and sex alone.
d C statistic is greater than for age, sex, and diseases.

©2011 American Medical Association. All rights reserved. (Reprinted) JAMA, January 5, 2011—Vol 305, No. 1 55

Downloaded From: http://jama.jamanetwork.com/ on 06/07/2014

 GAIT SPEED AND SURVIVAL IN OLDER ADULTS

sex, and gait speed were approximately vecciare in Chianti (gait speed, 0.727; lished area under the curve range, 0.66-
equivalent to those for age, sex, chronic 95% CI, 0.678-0.776; SPPB, 0.738; 95% 0.8261 vs this study, 0.717 and 0.737).
diseases, BMI, systolic blood pressure, CI, 0.690-0.735); Predicting Elderly Per- The strengths of this study are the very
and prior hospitalization in all 9 stud- formance study18 (gait speed, 0.667; 95% largesampleofindividualparticipantdata
ies and in the pooled analysis. There were CI, 0.610-0.724; SPPB, 0.691; 95% CI, from multiple diverse populations of
4 studies that had sufficiently consistent 0.637-0.744); and worse than SPPB in community-dwellingelderswhowerefol-
data on functional status to create 3 cat- the Established Populations for the Epi- lowed up for many years and use of con-
egories: dependent in ADLs, difficulty demiological Study of the Elderly15 (gait sistent measures of performance and out-
with instrumental ADLs, and indepen- speed, 0.638; 95% CI, 0.610-0.777; come. We provide survival estimates for
dent. For these studies, gait speed, age, SPPB, 0.663; 95% CI, 0.636-0.691). a broad range of gait speeds and calcu-
and sex yielded a C statistic (0.741) that late absolute rates and median years of
was not significantly different (P = .78) COMMENT survival.Comparedwithpriorstudiesthat
from age, sex, mobility aids, and func- Gait speed, age, and sex may offer the cli- were too small to assess potential effect
tional status (P=.75; Table 3). nician tools for assessing expected sur- modification by age, sex, race/ethnicity,
For 10-year survival, 6 studies had suf- vival to contribute to tailoring goals of and other subgroups, we were able to
ficient follow-up time to perform many care in older adults. The accuracy of pre- assessmultiplesubgroupeffectswithsub-
of the analyses (Table 3). Gait speed dictions based on these 3 factors ap- stantial power. This study has the limi-
added predictive ability to age and sex pears to be approximately similar to more tations of observational research; it can-
in 4 of 6 studies and in the pooled analy- complex models involving multiple other not establish causal relationships and is
sis. C statistics for age, sex, and gait speed health-related factors, or for age, sex, use vulnerable to various forms of healthy
were not significantly different from C of mobility aids, and functional status. volunteer bias. The participating study
statistics with all the other factors for any Gait speed might help refine survival es- cohorts, while large and diverse, do not
study nor for the pooled analysis. Three timates in clinical practice or research be- represent the universe of possible data.
studies had sufficiently consistent data cause it is simple and informative. Oursurvivalestimatesshouldbevalidated
on functional status at baseline to allow Whywouldgaitspeedpredictsurvival? in additional data sets. Only 1 of the 9
pooling. Gait speed, age, and sex yielded Walkingrequiresenergy,movementcon- studies was based in clinical practice,21
a C statistic (0.734) that was not signifi- trol, and support and places demands on and advanced dementia is rare in popu-
cantly different from age, sex, mobility multiple organ systems, including the lations who are competent to consent for
aids, and functional status (0.732; heart, lungs, circulatory, nervous, and research. However, median years of sur-
(P=.95; Table 3). musculoskeletalsystems.Slowinggaitmay vival in this study resemble estimates for
In addition, we used C statistics to as- reflect both damaged systems and a high- US adults across the sex and age range
sess the ability of usual gait speed to pre- energy cost of walking.13,39-54 Gait speed assessed.62 We were unable to assess
dict survival compared with other physi- could be considered a simple and acces- the association of physical activity with
cal performance measures, such as fast siblesummaryindicatorofvitalitybecause survival in pooled analyses because mea-
gait speed and the Short Physical Per- itintegratesknownandunrecognizeddis- sures of activity were highly variable
formance Battery (SPPB), a brief mea- turbancesinmultipleorgansystems,many across studies. Also, participants in these
sure that includes walk speed, chair rise of which affect survival. In addition, de- studies had no prior knowledge about the
ability, and balance. We assessed usual creasing mobility may induce a vicious meaning of walking speed. In clinical use,
vs fast gait speed in the single study with cycle of reduced physical activity and de- participants might walk differently if they
both measures (Invecciare in Chianti18 conditioning that has a direct effect on are aware of the implications of the re-
study: usual, 0.727 [95% CI, 0.678- health and survival.6 sults. Although this study provides infor-
0.776]; fast, 0.684 [95% CI, 0.630- The association between gait speed mationonsurvival,furtherworkisneeded
0.739]), suggesting that fast walks did and survival is known.6,7,9-12,55,56 Prior to examine associations of other impor-
not have an advantage in survival pre- analyses used single cohorts and pre- tant pooled outcomes such as disability
diction over usual-paced walks. Gait sented results as relative rather than ab- and health care use and to examine ef-
speed was superior to the SPPB in the solute risk, as done herein. Similarly, fects in populations more completely
Hispanic Established Populations for the mortality prediction models have been based in clinical practice.
Epidemiological Study of the Elderly8 developed.3-5,57-60 Some models use self- Because gait speed can be assessed by
(gait speed, 0.617; 95% CI, 0.585- reported information but others also in- nonprofessional staff using a 4-m walk-
0.649; SPPB, 0.574; 95% CI, 0.539- clude physiological or performance data, way and a stopwatch,21 it is relatively
0.649); was equivalent in the following for a total of 4 to more than 10 predic- simple to measure compared with many
3 studies: Health, Aging, and Body Com- tive factors. Only a few models assess medicalassessments.Nevertheless,meth-
position (ABC) study and ABC16 (gait overall predictive capacity using C sta- odologicalissuessuchasdistanceandver-
speed, 0.579; 95% CI, 0.548-0.610; tistics; the reported values are in the bal instructions remain.63,64 Self-report is
SPPB, 0.560; 95% CI, 0.528-0.592); In- range found in the present study (pub- an alternative to gait speed for reflecting
56 JAMA, January 5, 2011—Vol 305, No. 1 (Reprinted) ©2011 American Medical Association. All rights reserved.

Downloaded From: http://jama.jamanetwork.com/ on 06/07/2014

 GAIT SPEED AND SURVIVAL IN OLDER ADULTS

function.However,significantchallenges mortality, perhaps those with gait speeds Nevitt, Visser, Kritchevsky, Harris, Newman, Cauley,
Ferrucci, Guralnik.
remain in the use of self-report as well, slower than 0.6 m/s. In these patients, fur- Statistical analysis: Studenski, Perera, Patel, Inzitari,
such as choice of items and reliability, ther examination is targeted at potentially Chandler, Guralnik.
Obtained funding: Studenski, Chandler, Nevitt,
someofwhichcanbeaddressedbyemerg- modifiable risks to health and survival. A Kritchevsky, Badinelli, Harris, Newman, Ferrucci,
ing techniques such as computer adap- recommended evaluation and manage- Guralnik.
Administrative, technical, or material support:
tive testing based on item-response ment of slow walking includes cardiopul- Studenski, Patel, Rosano, Faulkner, Cawthon, Nevitt,
theory.65 The results found herein sug- monary, neurological and musculoskel- Visser, Badinelli, Harris, Newman, Cauley
gest that gait speed appears to be espe- etal systems.6,18 Third, gait speed might Study supervision: Perera, Nevitt, Newman.
Conflict of Interest Disclosures: All authors have com-
cially informative in older persons who promote communication. Primary clini- pleted and submitted the ICMJE Form for Disclosure
report either no functional limitations or cians might characterize an older adult as of Potential Conflicts of Interest. Dr Studenski re-
ported receiving institutional grant support, travel ex-
only difficulty with instrumental ADLs likely to be in poor health and function penses, and consultancy fees from Merck; consult-
and may be less helpful for older adults because the gait speed is 0.5 m/s. In re- lancy fees from Novartis and GTX; and royalties from
Hazzart Text McGraw Hill. Dr Perera reported receiv-
who already report dependence in basic search manuscripts, baseline gait speed ing institutional grant support from Merck Research
ADLs. The research studies analyzed might help to characterize the overall Lab. Dr Inzitari reported receiving insitutional grant
herein used trained staff to measure gait health of older research participants. support from Merck. Dr Brach reported receiving in-
stitutional grant support from Merck. Dr Cawthon re-
speed.Staffinclinicalsettingswouldneed Fourth, gait speed might be monitored ported receiving consultancy fees from Amgen and
initial training and may produce more over time, with a decline indicating a new Merck. Dr Cauley reported receiving consultancy fees
and institutional grant support from Novartis. Drs Pa-
variableresults.Long-distancewalkshave health problem that requires evaluation. tel, Faulkner, Barett-Connor, Nevitt, Visser, Bandel-
become accepted in some medical fields Fifth, gait speed might be used to stratify lini, Harris, Newman, Ferrucci, and Gurlanik reported
no disclosures.
and may contribute information beyond risks from surgery or chemotherapy. Fi- Funding/Support: Additional support for the pooled
short walks.66-68 However, the longer dis- nally, medical and behavioral interven- analyses was provided by grants AG023641, AG024827
tance and time to perform the test may tions might be assessed for their effect on and the Intramural Research Program, National Insti-
tute on Aging, NIH NIA Professional Services Contract
limit feasibility in many clinical settings. gait speed in clinical trials. Such true ex- Health and Human Services number 11200800292P.
Although the sample size of very slow perimentscouldthenevaluatecausalpath- Dr Studenski received grant support from Merck to per-
form this work.
walkers was small, our data suggest that ways to determine whether interventions Role of the Sponsor: The role of Merck and Co in the de-
there may be a subpopulation who walk that improve gait speed lead to improve- sign and conduct of the study; collection, management,
analysis, and interpretation of the data; and preparation,
very slowly but survive for long periods. ments in function, health, and longevity. review, or approval of the manuscript is as follows: Merck
It would be valuable to further charac- The data provided herein are in- and Co reviewed and approved an initial proposal to con-
duct the study, which included gathering existing data
terize this subgroup. tended to aid clinicians, investigators, and andpooledstatisticalanalyses.RepresentativesfromMerck
Although the gait speed–survival re- health system planners who seek simple reviewed the initial manuscript draft.
lationship seems continuous across the indicators of health and survival in older Online-Only Material: A PDF of enlarged Figure 2
graphs; eTables 1-3, and eFigures 1A-M, 2, 3, and 4
entire range, cut points may help inter- adults. Gait speed has potential to be are available at http://www.jama.com.
pretation. Several authors have proposed implemented in practice, using a stop
that gait speeds faster than 1.0 m/s sug- watch and a 4-m course. From a stand- REFERENCES
gest healthier aging while gait speeds ing start, individuals are instructed to 1. Reuben DB. Medical care for the final years of life:
slower than 0.6 m/s increase the likeli- walk at their usual pace, as if they were “When you’re 83, it’s not going to be 20 years.” JAMA.
2009;302(24):2686-2694.
hood of poor health and function.7,21 Oth- walking down the street, and given no 2. Lubitz J, Cai L, Kramarow E, Lentzner H. Health,
ers propose one cutoff around 0.8 m/s.13 further encouragement or instructions. life expectancy, and health care spending among the
In our data, predicted life expectancy at The data in this article can be used to help elderly. N Engl J Med. 2003;349(11):1048-1055.
3. Fried LP, Kronmal RA, Newman AB, et al. Risk fac-
themedianforageandsexoccursatabout interpret the results. Gait speed may be tors for 5-year mortality in older adults. JAMA. 1998;
0.8 m/s; faster gait speeds predict life ex- a simple and accessible indicator of the 279(8):585-592.
4. Lee SJ, Lindquist K, Segal MR, Covinsky KE. De-
pectancy beyond the median. Perhaps a health of the older person. velopment and validation of a prognostic index for
gaitspeedfasterthan1.0m/ssuggestsbet- 4-year mortality in older adults. JAMA. 2006;295
Author Contributions: Dr Studenski had full access to (7):801-808.
terthanaveragelifeexpectancyandabove all of the data in the study and takes responsibility for 5. Schonberg MA, Davis RB, McCarthy EP, Marcantonio
the integrity of the data and the accuracy of the data ER. Index to predict 5-year mortality of community-
1.2 m/s suggests exceptional life expec- analysis. dwelling adults aged 65 and older using data from the
tancy, but additional research will be nec- Study concept and design: Studenski, Faulkner, NationalHealthInterviewSurvey.JGenInternMed.2009;
Chandler, Nevitt, Kritchevsky, Ferrucci, Guralnik.
essary to determine this relationship. Acquisition of data: Studenski, Faulkner, Nevitt, Visser,
24(10):1115-1122.
6. Cesari M, Kritchevsky SB, Newman AB, et al; Health,
How might gait speed be used clini- Kritchevsky, Badinelli, Harris, Newman, Cauley, Aging and Body Composition Study. Added value of
cally? First, gait speed might help iden- Ferrucci, Guralnik. physical performance measures in predicting adverse
Analysis and interpretation of data: Studenski, Perera, health-related events. J Am Geriatr Soc. 2009;
tify older adults with a high probability Patel, Rosano, Faulkner, Inzitari, Brach, Chandler, 57(2):251-259.
of living for 5 or 10 more years, who may Cawthon, Barrett-Connor, Nevitt, Visser, Harris, 7. Cesari M, Kritchevsky SB, Penninx BW, et al. Prog-
Newman, Cauley, Guralnik. nostic value of usual gait speed in well-functioning
be appropriate targets for preventive in- Drafting of the manuscript: Studenski, Perera, Inzitari, older people. J Am Geriatr Soc. 2005;53(10):1675-
terventions that require years for benefit. Badinelli. 1680.
Critical revision of the manuscript for important in- 8. Markides KS, Stroup-Benham C, Black S, Satis S,
Second, gait speed might be used to iden- tellectual content: Studenski, Perera, Patel, Rosano, Perkowski L, Ostir G. The health of Mexican Ameri-
tifyolderadultswithincreasedriskofearly Faulkner, Brach, Chandler, Cawthon, Barrett-Connor, can elderly: Selected findings from the Hispanic EPESE.

©2011 American Medical Association. All rights reserved. (Reprinted) JAMA, January 5, 2011—Vol 305, No. 1 57

Downloaded From: http://jama.jamanetwork.com/ on 06/07/2014

 GAIT SPEED AND SURVIVAL IN OLDER ADULTS

In: Wykle ML, Ford A, ed. Serving Minority Elders in 27. Washburn RA, Smith KW, Jette AM, Janney CA. in nondiabetic older men. BMC Geriatr. 2009;
the 21st Century. New York, NY: Springer; 1999: The Physical Activity Scale for the Elderly (PASE): de- 9:49.
72-90. velopment and evaluation. J Clin Epidemiol. 1993; 50. Nebes RD, Pollock BG, Halligan EM, Kirshner MA,
9. OstirGV,KuoYF,BergesIM,MarkidesKS,Ottenbacher 46(2):153-162. Houck PR. Serum anticholinergic activity and motor
KJ. Measures of lower body function and risk of mortal- 28. Martin FC, Hart D, Spector T, Doyle DV, Harari performance in elderly persons. J Gerontol A Biol Sci
ity over 7 years of follow-up. Am J Epidemiol. 2007; D. Fear of falling limiting activity in young-old women Med Sci. 2007;62(1):83-85.
166(5):599-605. is associated with reduced functional mobility rather 51. Rosano C, Aizenstein HJ, Studenski S, Newman
10. Rolland Y, Lauwers-Cances V, Cesari M, Vellas B, than psychological factors. Age Ageing. 2005;34 AB. A regions-of-interest volumetric analysis of mo-
Pahor M, Grandjean H. Physical performance measures (3):281-287. bility limitations in community-dwelling older adults.
aspredictorsofmortalityinacohortofcommunity-dwelling 29. Lawless JF. Statistical Models and Methods for J Gerontol A Biol Sci Med Sci. 2007;62(9):1048-
older French women. Eur J Epidemiol. 2006;21(2): Lifetime Data. New York, NY: Wiley; 2002. 1055.
113-122. 30. Thernau TM, Grambsch PM. Modeling Survival 52. Volpato S, Blaum C, Resnick H, Ferrucci L, Fried
11. Rosano C, Newman AB, Katz R, Hirsch CH, Kuller Data: Extending the Cox Model. New York, NY: LP, Guralnik JM; Women’s Health and Aging Study.
LH. Association between lower digit symbol substi- Springer; 2000. Comorbidities and impairments explaining the asso-
tution test score and slower gait and greater risk of 31. Cochran WG. The combination of estimates from ciation between diabetes and lower extremity disability.
mortality and of developing incident disability in well- different experiments. Biometrics. 1954;10(1): Diabetes Care. 2002;25(4):678-683.
functioning older adults. J Am Geriatr Soc. 2008; 101-129. 53. Jones LM, Waters DL, Legge M. Walking speed
56(9):1618-1625. 32. Higgins JP, Thompson SG. Quantifying hetero- at self-selected exercise pace is lower but energy cost
12. Woo J, Ho SC, Yu AL. Walking speed and stride geneity in a meta-analysis. Stat Med. 2002;21 higher in older versus younger women. J Phys Act
length predicts 36 months dependency, mortality, and (11):1539-1558. Health. 2009;6(3):327-332.
institutionalization in Chinese aged 70 and older. J Am 33. DerSimonian R, Laird N. Meta-analysis in clinical 54. Mian OS, Thom JM, Ardigò LP, Narici MV, Minetti
Geriatr Soc. 1999;47(10):1257-1260. trials. Control Clin Trials. 1986;7(3):177-188. AE. Metabolic cost, mechanical work, and efficiency
13. Abellan van Kan G, Rolland Y, Andrieu S, et al. 34. Klein JP, Moeschberger M. Survival Analysis: Tech- during walking in young and older men. Acta Physiol
Gait speed at usual pace as a predictor of adverse out- niques for Censored and Truncated Data. New York, (Oxf ). 2006;186(2):127-139.
comes in community-dwelling older people. J Nutr NY: Springer; 1997. 55. Cesari M, Onder G, Zamboni V, et al. Physical func-
Health Aging. 2009;13(10):881-889. 35. Katsahian S, Latouche A, Mary JY, Chevret S, tion and self-rated health status as predictors of
14. Hall WJ. Update in geriatrics. Ann Intern Med. Porcher R. Practical methodology of meta-analysis of mortality. BMC Geriatr. 2008;8:34.
2006;145(7):538-543. individual patient data using a survival outcome. Con- 56. Markides KS, Black SA, Ostir GV, Angel RJ, Guralnik
15. Lavsky-Shulan M, Wallace RB, Kohout FJ, Lemke temp Clin Trials. 2008;29(2):220-230. JM, Lichtenstein M. Lower body function and mor-
JH, Morris MC, Smith IM. Prevalence and functional 36. Kalbfleisch JD, Prentice RL. The Statistical Analy- tality in Mexican American elderly people. J Gerontol
correlates of low back pain in the elderly: the Iowa sis of Failure Time Data. New York, NY: Wiley; 1980. A Biol Sci Med Sci. 2001;56(4):M243-M247.
65$ Rural Health Study. J Am Geriatr Soc. 1985; 37. Apfel CC, Kranke P, Greim CA, Roewer N. What 57. Inouye SK, Peduzzi PN, Robison JT, Hughes JS,
33(1):23-28. can be expected from risk scores for predicting post- Horwitz RI, Concato J. Importance of functional mea-
16. Visser M, Deeg DJ, Lips P, Harris TB, Bouter LM. operative nausea and vomiting? Br J Anaesth. 2001; sures in predicting mortality among older hospital-
Skeletal muscle mass and muscle strength in relation 86(6):822-827. ized patients. JAMA. 1998;279(15):1187-1193.
to lower-extremity performance in older men and 38. Dear KB. Iterative generalized least squares for 58. Keeler E, Guralnik JM, Tian H, Wallace RB, Reuben
women. J Am Geriatr Soc. 2000;48(4):381-386. meta-analysis of survival data at multiple times. DB. The impact of functional status on life expec-
17. Visser M, Goodpaster BH, Kritchevsky SB, et al. Biometrics. 1994;50(4):989-1002. tancy in older persons. J Gerontol A Biol Sci Med Sci.
Muscle mass, muscle strength, and muscle fat infil- 39. Atkinson HH, Rosano C, Simonsick EM, et al; 2010;65(7):727-733.
tration as predictors of incident mobility limitations in Health ABC study. Cognitive function, gait speed de- 59. Mazzaglia G, Roti L, Corsini G, et al. Screening
well-functioning older persons. J Gerontol A Biol Sci cline, and comorbidities. J Gerontol A Biol Sci Med of older community-dwelling people at risk for death
Med. 2005;60(3):324-333. Sci. 2007;62(8):844-850. and hospitalization: the Assistenza Socio-Sanitaria in
18. Ferrucci L, Bandinelli S, Benvenuti E, et al. Sub- 40. Baezner H, Blahak C, Poggesi A, et al; LADIS Study Italia project. J Am Geriatr Soc. 2007;55(12):1955-
systems contributing to the decline in ability to walk. Group. Association of gait and balance disorders with 1960.
J Am Geriatr Soc. 2000;48(12):1618-1625. age-related white matter changes. Neurology. 2008; 60. Ostbye T, Steenhuis R, Wolfson C, Walton R, Hill
19. Orwoll E, Blank JB, Barrett-Connor E, et al. De- 70(12):935-942. G. Predictors of five-year mortality in older Canadians.
sign and baseline characteristics of the osteoporotic 41. Buchman AS, Boyle PA, Leurgans SE, Evans DA, J Am Geriatr Soc. 1999;47(10):1249-1254.
fractures in men (MrOS) study. Contemp Clin Trials. Bennett DA. Pulmonary function, muscle strength, and 61. Keeler E, Guralnik JM, Tian H, Wallace RB, Reuben
2005;26(5):569-585. incident mobility disability in elders. Proc Am Thorac DB. The impact of functional status on life expec-
20. Plan and operation of the Third National Health Soc. 2009;6(7):581-587. tancy in older persons. J Gerontol A Biol Sci Med Sci.
and Nutrition Examination Survey, 1988–94. Hyatts- 42. Callisaya ML, Blizzard L, Schmidt MD, McGinley 2010;65(7):727-733.
ville, MD: National Center for Health Statistics, US Dept JL, Lord SR, Srikanth VK. A population-based study 62. United States Life Tables, 2010. http://www.cdc
of Health and Human Services Services; 2004. PHS of sensorimotor factors affecting gait in older people. .gov/nchs/data/nvsr/nvsr58/nvsr_21.pdf). Accessed
94–1308. Age Ageing. 2009;38(3):290-295. November 29, 2010.
21. Studenski S, Perera S, Wallace D, et al. Physical 43. Cham R, Studenski SA, Perera S, Bohnen NI. Stria- 63. Graham JE, Ostir GV, Fisher SR, Ottenbacher KJ.
performance measures in the clinical setting. J Am Geri- tal dopaminergic denervation and gait in healthy adults. Assessing walking speed in clinical research. J Eval Clin
atr Soc. 2003;51(3):314-322. Exp Brain Res. 2008;185(3):391-398. Pract. 2008;14(4):552-562.
22. Cummings SR, Black DM, Nevitt MC, et al; The 44. Cuoco A, Callahan DM, Sayers S, Frontera WR, 64. Graham JE, Ostir GV, Kuo YF, Fisher SR, Ottenbacher
Study of Osteoporotic Fractures Research Group. Ap- Bean J, Fielding RA. Impact of muscle power and KJ. Relationship between test methodology and mean
pendicular bone density and age predict hip fracture force on gait speed in disabled older men and women. velocity in timed walk tests: a review. Arch Phys Med
in women. JAMA. 1990;263(5):665-668. J Gerontol A Biol Sci Med Sci. 2004;59(11):1200- Rehabil. 2008;89(5):865-872.
23. Fried LP, Borhani NO, Enright P, et al. The Car- 1206. 65. Gill TM. Assessment of function and disability in
diovascular Health Study. Ann Epidemiol. 1991; 45. Fitzpatrick AL, Buchanan CK, Nahin RL, et al; longitudinal studies. J Am Geriatr Soc. 2010;58
1(3):263-276. Ginkgo Evaluation of Memory (GEM) Study (suppl 2):S308-S312.
24. Guralnik JM, Ferrucci L, Pieper CF, et al. Lower Investigators. Associations of gait speed and other mea- 66. Newman AB, Simonsick EM, Naydeck BL, et al.
extremity function and subsequent disability: consis- sures of physical function with cognition in a healthy Association of long-distance corridor walk perfor-
tency across studies, predictive models, and value of cohort of elderly persons. J Gerontol A Biol Sci Med mance with mortality, cardiovascular disease, mobil-
gait speed alone compared with the short physical per- Sci. 2007;62(11):1244-1251. ity limitation, and disability. JAMA. 2006;295(17):
formance battery. J Gerontol A Biol Sci Med Sci. 2000; 46. Fried LF, Lee JS, Shlipak M, et al. Chronic kidney 2018-2026.
55(4):M221-M231. disease and functional limitation in older people. J Am 67. Solway S, Brooks D, Lacasse Y, Thomas S. A
25. Bohannon RW. Comfortable and maximum walk- Geriatr Soc. 2006;54(5):750-756. qualitative systematic overview of the measurement
ing speed of adults aged 20-79 years: reference val- 47. Holtzer R, Verghese J, Xue X, Lipton RB. Cogni- properties of functional walk tests used in the
ues and determinants. Age Ageing. 1997;26(1): tive processes related to gait velocity. Neuropsychology. cardiorespiratory domain. Chest. 2001;119(1):
15-19. 2006;20(2):215-223. 256-270.
26. Guralnik JM, Simonsick EM, Ferrucci L, et al. A 48. Kerrigan DC, Lee LW, Collins JJ, Riley PO, Lipsitz 68. Simonsick EM, Newman AB, Visser M, et al; Health,
short physical performance battery assessing lower ex- LA. Reduced hip extension during walking. Arch Phys Aging and Body Composition Study. Mobility limita-
tremity function: association with self-reported dis- Med Rehabil. 2001;82(1):26-30. tion in self-described well-functioning older adults.
ability and prediction of mortality and nursing home 49. Kuo CK, Lin LY, Yu YH, Wu KH, Kuo HK. Inverse J Gerontol A Biol Sci Med Sci. 2008;63(8):841-
admission. J Gerontol. 1994;49(2):M85-M94. association between insulin resistance and gait speed 847.

58 JAMA, January 5, 2011—Vol 305, No. 1 (Reprinted) ©2011 American Medical Association. All rights reserved.

Downloaded From: http://jama.jamanetwork.com/ on 06/07/2014