midlife fitness and the development of chronic conditions in later lifemidlife fitness and chronic...

Scan for AuthorAudio Interview

ORIGINAL INVESTIGATION

Midlife Fitness and the Developmentof Chronic Conditions in Later LifeBenjamin L. Willis, MD, MPH; Ang Gao, MS; David Leonard, PhD;Laura F. DeFina, MD; Jarett D. Berry, MD, MS

Background: The association between cardiorespira-tory fitness (fitness) and mortality is well described. How-ever, the association between midlife fitness and the de-velopment of nonfatal chronic conditions in older age hasnot been studied.

Methods: To examine the association between midlifefitness and chronic disease outcomes in later life, par-ticipant data from the Cooper Center Longitudinal Studywere linked with Medicare claims. We studied 18 670healthy participants (21.1% women; median age, 49 years)who survived to receive Medicare coverage from Janu-ary 1, 1999, to December 31, 2009. Fitness estimated byBalke treadmill time was analyzed as a continuous vari-able (in metabolic equivalents [METs]) and accordingto age- and sex-specific quintiles. Eight common chronicconditions were defined using validated algorithms, andassociations between midlife fitness and the number ofconditions were assessed using a modified Cox propor-tional hazards model that stratified the at-risk popula-tion by the number of conditions while adjusting for age,body mass index, blood pressure, cholesterol and glu-cose levels, alcohol use, and smoking.

Results: After 120 780 person-years of Medicare expo-sure with a median follow-up of 26 years, the highest quin-

tile of fitness (quintile 5) was associated with a lower in-cidence of chronic conditions compared with the lowestquintile (quintile 1) in men (15.6 [95% CI, 15.0-16.2]vs 28.2 [27.4-29.0] per 100 person-years) and women(11.4 [10.5-12.3] vs 20.1 [18.7 vs 21.6] per 100 person-years). After multivariate adjustment, higher fitness (inMETs) was associated with a lower risk of developingchronic conditions in men (hazard ratio, 0.95 [95% CI,0.94-0.96] per MET) and women (0.94 [0.91-0.96] perMET). Among decedents (2406 [12.9%]), higher fit-ness was associated with lower risk of developing chronicconditions relative to survival (compression hazard ra-tio, 0.90 [95% CI, 0.88-0.92] per MET), suggesting mor-bidity compression.

Conclusions: In this cohort of healthy middle-agedadults, fitness was significantly associated with a lowerrisk of developing chronic disease outcomes during 26years of follow-up. These findings suggest that highermidlife fitness may be associated with the compressionof morbidity in older age.

Arch Intern Med. 2012;172(17):1333-1340.Published online August 27, 2012.doi:10.1001/archinternmed.2012.3400

H EALTHY AGING HAS BEEN

well studied, with mul-tiple reports1-7 describ-ing an association of tra-ditional cardiovascular

risk factors, such as smoking and hyper-tension. Although physical activity (PA)likely represents an important determi-nant of healthy aging, studies5-7 have re-ported inconsistent results; therefore, theincremental contribution of PA to healthyaging beyond other healthy lifestyle char-acteristics remains unclear. The inverseassociation between cardiorespiratoryfitness (fitness) and mortality after adjust-ment for other risk factors is well estab-lished.8-16 In addition, compared with self-reported measures of PA, fitness levels are

more strongly associated with mortality,reflecting, at least in part, the objective na-ture of their measurement.15

Therefore, we hypothesized that highermidlife fitness levels would be strongly as-sociated with healthy aging as defined bya low burden of chronic condition (CC)outcomes. To test this hypothesis, wemerged individual-level data from the Coo-per Center Longitudinal Study (CCLS)with Medicare claims files from the Cen-ter for Medicare and Medicaid Services(CMS).

See Invited Commentaryat end of article

Author AffilCooper Instiand DeFina)Cardiology,Internal MedDr Berry), anClinical ScieBiostatisticsThe UniversSouthwesterDallas.

Author Affiliations: TheCooper Institute (Drs Willisand DeFina), Division ofCardiology, Department ofInternal Medicine (Ms Gao andDr Berry), and Department ofClinical Science, Division ofBiostatistics (Dr Leonard),The University of TexasSouthwestern Medical Center,Dallas.

ARCH INTERN MED/ VOL 172 (NO. 17), SEP 24, 2012 WWW.ARCHINTERNMED.COM1333

©2012 American Medical Association. All rights reserved.

METHODS

STUDY SAMPLE

The study sample was derived from the CCLS, which is a largecohort of individuals who have completed a preventive medi-cine examination at the Cooper Clinic in Dallas, Texas, from1970 to 2009. Patients seen at the Cooper Clinic are generallywell-educated non-Hispanic whites from middle to upper so-cioeconomic strata. Patients who are part of the CCLS receivea comprehensive clinical examination that includes self-reported personal and family history, standardized medical ex-amination by a physician, anthropometric measurements, fast-ing laboratory studies, and a maximal treadmill exercise test.Participants provide written informed consent for inclusion inthe research database. The study was reviewed and approvedannually by the institutional review board of The CooperInstitute.

Among 73 439 participants in the CCLS who had completedata for analysis, 24 809 became 65 years or older between Janu-ary 1, 1999, and December 31, 2009, and were eligible to re-ceive Medicare coverage. After excluding 2973 participants(12.0%) lacking traditional Medicare fee-for-service coverage(ie, Medicare Advantage participants lacking claims files), weexcluded an additional 2559 participants (10.3%) with a self-reported history of myocardial infarction, stroke, cancer, chroniclung disease, or diabetes mellitus as defined by self-report orfasting blood glucose level of 126 mg/dL or higher (to convertto millimoles per liter, multiply by 0.0555) at study entry. Anadditional 607 individuals (2.4%) whose CCLS examination oc-curred after enrollment into a Medicare fee-for-service plan orwere receiving early (younger than 65 years) eligibility cover-age because of disability or renal dialysis were excluded, leav-ing a study sample of 18 670 (21.1% women) CCLS partici-pants for the present analysis.

CLINICAL VARIABLES

Fitness was assessed by maximal effort using the modified Balkeprotocol, as previously described.17 The test time using this pro-tocol correlates highly with directly measured maximal oxy-gen uptake (r=0.92).18,19

In accordance with standard approaches to the analysis offitness data,8,9 each participant’s treadmill time was classifiedinto age- and sex-specific quintiles of fitness, with low fitnessrepresented as quintile 1. Details of the treadmill times acrossfitness quintiles are reported by age and sex groups in eTable1 (http://www.archinternmed.com). Using well-characterizedregression equations, treadmill times from the Balke protocolalso allow for estimation of fitness level in metabolic equiva-lents (METs).18,20

The measurement of other baseline variables in the CCLShas been well described.8,9,21 Body mass index (calculated asweight in kilograms divided by height in meters squared) wasdetermined using a standard clinical scale and stadiometer.Seated resting blood pressure was obtained with a mercurysphygmomanometer. Venous blood obtained when the par-ticipants were fasting was assayed for serum cholesterol andglucose using standardized, automated techniques. Physical ac-tivity was measured using the Physical Activity Index, a 5-levelPA questionnaire that has been reported22,23 (0, no regular PA;1, some PA other than walking, running, or jogging; 2, walk-ing, jogging, or running �16 km/wk; 3, walking, jogging, orrunning 16-32 km/wk; and 4, walking, jogging, or running �32km/wk).

OUTCOME MEASURES

Medicare inpatient claims data were obtained from the CMSfor surviving participants who were 65 years or older and werethus eligible for Medicare benefits from 1999 through 2009.The CMS data contain 100% of claims paid by Medicare for cov-ered health care services. Chronic condition diagnoses used inthis study were determined from the Chronic Condition Ware-house included in the Beneficiary Annual Summary File.24

Chronic conditions are defined within the Chronic ConditionWarehouse from well-established algorithms for research pur-poses.24-26 A panel of 8 CCs was used for the present analysis:congestive heart failure, ischemic heart disease, stroke, diabe-tes mellitus, chronic obstructive pulmonary disease, chronickidney disease, Alzheimer disease, and colon or lung cancer.These conditions were chosen from the Chronic ConditionWarehouse in an effort to define a broad panel across multipleorgan systems in accordance with definitions of healthy agingas defined by others.2,5,6,27 Minor conditions (ie, cataracts) andsex-specific outcomes (ie, breast or prostate cancer) were ex-cluded to define a consistent set of conditions between menand women. To create a summary measure of the burden ofCCs, the combined number was used as the outcome measure(eg, 0 CCs, 1 CC).

STATISTICAL ANALYSIS

We determined the overall burden of CCs at ages 70, 75, 80,and 85 years by classifying all participants alive at these agethresholds according to the presence or absence of each of the8 CCs. This required that the earliest indication of the condi-tion occurred before or at the attained age among survivors.We calculated the incidence of CCs by dividing the number ofdiagnoses by the total observation time. Because patients coulddevelop more than 1 CC during the surveillance period, a modi-fied multivariate failure time model was used.28,29 Death dur-ing the surveillance period and survival to the end of the sur-veillance period were considered censoring events. We usedattained age as the time scale, relegating all age effects to thebaseline hazard characterizing each event stratum. Midlife fit-ness was entered as a continuous variable (in METs). Men andwomen were modeled separately. We accommodated possibledepartures from proportional hazards by testing for and re-taining significant covariate by attained age effects. Main ef-fects were estimated at the mean attained age.

To assess the association between fitness and (1) the devel-opment of a CC or (2) death, we conducted additional analy-ses among the subset of participants who died during the ob-servation period, using a similar modeling approach.28,29 In thismodel, we treated death as an outcome of interest rather thana censoring event. This model therefore allowed transitions todeath as well as to higher CC states. The estimates for the ef-fect of fitness on the transition to (1) an additional CC or (2)death were compared using the Wald �2 statistic. This com-parison takes the form of a ratio of relative risks in which eachrelative risk is for CCs relative to death, and the numerator isevaluated at 1-MET higher fitness than the denominator. Werefer to this ratio as the morbidity compression ratio; values sig-nificantly less than 1 offer evidence of compression of morbid-ity. Finally, to compare descriptively the association betweenfitness and CC burden at the end of life, for each level of fit-ness we partitioned the remaining years of follow-up accord-ing to the time spent at each level of CC burden.

P� .05 (2-sided) was considered statistically significant. Allstatistical analyses were performed using commercial soft-ware (SAS for Windows, release 9.2; SAS Institute, Inc).



RESULTS

Baseline characteristics for 14 726 men and 3944 womenin the study sample are reported in Table 1, demon-strating low levels of traditional risk factors at study en-try. As expected, fitness levels were higher in men com-pared with women, with higher levels of traditional riskfactors in the lower fitness strata. After 120 780 person-years of Medicare follow-up, there was considerable varia-tion in the prevalence of CC burden by attained age andacross conditions (Figure).

The association between midlife fitness and the inci-dence of CCs is provided in Table 2, demonstrating ahigher incidence of CCs across levels of fitness mea-sured in midlife. The highest level of midlife fitness (quin-tile 5) was associated with a lower incidence of CCs com-

pared with low midlife fitness (quintile 1) in men (15.6[95% CI, 15.0-16.2] vs 28.2 [27.4-29.0] per 100 person-years) and women (11.4 [10.5-12.3] vs 20.1 [18.7 vs 21.6]per 100 person-years). After multivariate adjustment,higher fitness was associated with a lower risk of devel-oping CCs (men: hazard ratio [HR], 0.95 [0.94-0.96] perMET; women: HR, 0.94 [0.91-0.96] per MET; P � .001for all comparisons) (Table 3). In men, higher bloodpressure, higher total cholesterol, higher body mass in-dex, higher glucose, and smoking prevalence were asso-ciated with a higher risk of developing CC outcomes(Table 3). Overall, a comparable pattern of results wasobserved for women but with wider CIs.

When each of the 8 CCs was removed from the listof CCs in separate sensitivity analyses, the associationbetween lower fitness and the risk of CC outcomes re-

Table 1. Baseline Characteristics by Sex and Fitness Quintilesa

Characteristic

Men

Q1(n = 2632)

Q2(n = 2986)

Q3(n = 3209)

Q4(n = 3033)

Q5(n = 2866)

Age, y 46.0 (8.2) 47.7 (8.5) 49.3 (8.2) 50.3 (8.6) 50.7 (8.2)BMI 28.3 (4.3) 26.8 (3.3) 26.3 (2.9) 25.6 (2.6) 24.4 (2.3)Systolic BP, mm Hg 124.0 (14.4) 122.3 (13.7) 121.9 (13.9) 122.0 (14.0) 121.7 (14.2)Total cholesterol level, mg/dL, % 221.2 (40.5) 216.8 (39.1) 215.7 (38.8) 211.5 (37.8) 207.8 (36.9)Glucose level, mg/dL, % 100.5 (10.6) 100.3 (10.1) 99.5 (9.7) 99.1 (9.2) 98.3 (8.9)Fitness, METs 8.5 (1.2) 9.9 (1.0) 10.9 (1.0) 12.0 (1.1) 14.1 (1.7)Treadmill time, min 11.1 (2.5) 14.2 (2.2) 16.3 (2.3) 18.8 (2.5) 23.2 (3.0)Physical Activity Index, median (IQR)b 0.0 (0.0-1.0) 0.0 (0.0-2.0) 1.0 (0.0-2.0) 2.0 (1.0-2.0) 2.0 (2.0-3.0)No physical activity, No. (%) 1851 (70.3) 1614 (54.1) 1133 (35.3) 586 (19.3) 232 (8.1)Smoker, No. (%) 827 (31.4) 680 (22.8) 540 (16.8) 322 (10.6) 194 (6.8)Family history of premature CHD,

No. (%)68 (2.58) 74 (2.48) 112 (3.49) 132 (4.35) 135 (4.71)

Alcohol use (drinks/wk), median (IQR) 4.0 (0.0-12.0) 5.0 (0.0-12.0) 4.0 (0.0-12.0) 5.0 (0.0-12.0) 5.0 (0.0-12.0)Educational level, yc 15.4 (2.64) 15.9 (2.4) 16.0 (2.6) 16.3 (2.5) 16.5 (2.6)Nonwhite, No. (%) 49 (1.9) 47 (1.6) 60 (1.9) 52 (1.7) 34 (1.2)

Women

Q1(n = 552)

Q2(n = 767)

Q3(n = 730)

Q4(n = 944)

Q5(n = 951)

Age, y 48.4 (9.2) 50.0 (9.3) 51.3 (8.8) 52.2 (8.6) 53.3 (7.8)BMI 25.2 (5.3) 23.7 (3.9) 23.2 (3.6) 22.9 (3.2) 21.8 (2.5)Systolic BP, mm Hg 118.4 (15.5) 116.1 (15.0) 116.0 (15.9) 116.4 (15.8) 116.1 (15.5)Total cholesterol level, mg/dL, % 213.9 (39.2) 212.9 (40.3) 213.9 (39.2) 210.7 (38.7) 208.2 (36.8)Glucose level, mg/dL, % 95.6 (10.4) 94.7 (9.7) 94.9 (9.5) 94.7 (8.9) 94.1 (8.7)Fitness, METs 6.4 (0.9) 7.6 (0.9) 8.4 (0.9) 9.2 (1.0) 11.0 (1.5)Treadmill time, min 6.6 (1.8) 9.2 (1.9) 10.9 (2.0) 12.8 (2.2) 16.6 (3.1)Physical Activity Index, median (IQR)b 0.0 (0.0-1.0) 1.0 (0.0-2.0) 1.0 (0.0-2.0) 2.0 (1.0-3.0) 2.09 (1.10)No physical activity, No. (%) 355 (64.3) 367 (47.8) 220 (30.1) 198 (21.0) 83 (8.7)Smoker, No. (%) 101 (18.3) 85 (11.1) 65 (8.9) 68 (7.2) 36 (3.8)Family history of premature CHD,

No. (%)22 (4.00) 50 (6.52) 47 (6.44) 79 (8.37) 78 (8.20)

Alcohol use (drinks/wk), median (IQR) 2.0 (0.0-5.0) 2.0 (0.0-6.0) 2.0 (0.0-6.0) 2.0 (0.0-7.0) 2.0 (0.0-6.0)Educational level, yc 14.0 (2.5) 14.3 (2.2) 14.6 (2.4) 14.6 (2.4) 15.1 (2.2)Nonwhite, No. (%) 12 (2.2) 21 (2.7) 15 (2.1) 18 (1.9) 14 (1.5)

Abbreviations: BMI, body mass index (calculated as weight in kilograms divided by height in meters squared); BP, blood pressure; CHD, coronary heart disease;IQR, interquartile range; METs, metabolic equivalents; Q, quintile.

SI conversion factors: To convert cholesterol to millimoles per liter, multiply by 0.0259; glucose to millimoles per liter, multiply by 0.0555.aQ1 to Q5 represents age- and sex-specific quintiles of fitness based on Balke treadmill times, with Q1 being low fitness. Data are given as mean (SD) unless

otherwise indicated.bPhysical Activity Index is a self-reported scale, for which 0 indicates no regular physical activity; 1, some physical activity other than walking, running, or jogging;

2, walking, jogging, or running less than 16 km/wk; 3, walking, jogging, or running 16 to 32 km/wk; and 4, walking, jogging, or running more than 32 km/wk.cOnly 20% of participants reported educational level.



mained unchanged, suggesting that our findings are in-sensitive to any particular set of CCs (eTable 2). In anadditional sensitivity analysis, when these data were strati-fied by the median age (49 years) at examination, we ob-served a similar pattern of results for fitness levels mea-sured in younger participants (men: HR, 0.94 [0.92-

0.95]; women: 0.92 [0.88-0.96]) and older participants(men: HR, 0.95 [0.93-0.96]; women: 0.94 [0.91-0.97];P � .001 for all comparisons).

By the end of the follow-up period, there were 2406deaths (12.9%) with 13 759 person-years of Medicare fol-low-up, representing approximately 5 years of fol-

80

60

40

20

0

Prev

alen

ce, %

IHD (n = 7306) CHF (n = 2833) Stroke (n = 1852) DM (n = 2713)

Chronic Condition (No. of Cases)

Men

Women

COPD (n = 1714) CKD (n = 2018) ALZ (n = 777) CA (n = 545)

Age 70 y (n = 9988)Age 75 y (n = 6055)Age 80 y (n = 3101)Age 85 y (n = 1198)

Age 70 y (n = 2499)Age 75 y (n = 1510)Age 80 y (n = 841)Age 85 y (n = 385)

A

80

60

40

20

0

Prev

alen

ce, %

IHD (n = 1331) CHF (n = 490) Stroke (n = 406) DM (n = 412)

Chronic Condition (No. of Cases)COPD (n = 380) CKD (n = 311) ALZ (n = 242) CA (n = 89)

B

Figure. Prevalence of selected chronic conditions in men and women by attained age (N = 18 670). Prevalence of chronic conditions at ages 70, 75, 80, and 85years was determined by classifying all participants alive at these age thresholds according to the presence or absence of each of the 8 chronic conditions. Thepresence of a condition required that the earliest indication of the condition occurred before or at the attained age among survivors. Because some participantssurvived across multiple age thresholds and are represented in more than 1 age category, the numbers listed for the age groups total more than 18 670.ALZ indicates Alzheimer disease; CA, cancer of the colon or lung; CHF, congestive heart failure; CKD, chronic kidney disease; COPD, chronic obstructivepulmonary disease; DM, diabetes mellitus; and IHD, ischemic heart disease.

Table 2. Rate of CC Burden by Midlife Fitness Measurement in Men and Womena

Characteristic Q1 Q2 Q3 Q4 Q5

MenNo. 2632 2986 3209 3033 2866Rate of CC burden (95% CI) 28.2 (27.4-29.0) 22.4 (21.7-23.1) 20.1 (19.5-20.7) 17.8 (17.2-18.4) 15.6 (15.0-16.2)No. of conditions 4797 4376 4263 3466 2856Person-years 17 002 19 506 21 256 19 483 18 300

WomenNo. 552 767 730 944 951Rate of CC burden (95% CI) 20.1 (18.7-21.6) 16.6 (15.6-17.8) 14.3 (13.2-15.4) 12.3 (11.4-13.3) 11.4 (10.5-12.3)No. of conditions 747 876 673 713 652Person-years 3724 5251 4721 5793 5744

Abbreviations: CC, chronic condition; Q, quintile.aRates per 100 person-years reported in age- and sex-specific quintiles of fitness based on Balke treadmill times, with Q1 being low fitness and Q5 being high

fitness.



low-up in each fitness group before death. Among dece-dents with higher levels of midlife fitness, the morbiditycompression ratio comparing CC development relativeto that of death was significantly lower (0.90 [0.88-0.92] per MET). Thus, higher midlife fitness is associ-ated with the delay in the development of CCs to a greaterextent than the extension of the lifespan, suggesting thecompression of morbidity nearer the end of life. The as-sociation of fitness with the compression of CC burdencan also be seen in the proportion of time spent with dif-ferent numbers of CCs before death. Compared with par-ticipants with lower midlife fitness, those with highermidlife fitness appeared to spend a greater proportionof their final 5 years of life with a lower burden of CCs(Table 4).

COMMENT

In the present study, higher fitness measured in midlifewas strongly associated with a lower incidence of CCsdecades later. Furthermore, higher midlife fitness wasmore strongly associated with a delay in the onset of CCsthan with overall survival, suggesting that higher fitnessin midlife is associated with the compression of morbid-ity in later life.

We observed clinically significant associations be-tween midlife fitness levels and chronic disease burdenin later life. At lower fitness levels, where the associa-

tion was strongest (Table 2), our data suggest that a mod-est increase in fitness could translate into marked reduc-tion of CCs in older age. For example, a 1- to 2-METimprovement in fitness resulting in promotion from thefirst to the second fitness quintile at age 50 years was as-sociated with a 20% reduction in the incidence of CCsat ages 65 and older. Previous PA intervention stud-ies30-32 have achieved mean fitness gains of this magni-tude using a 6-month program of 150 minutes per weekof moderate-intensity exercise.

Studies have examined the association between riskfactors in midlife and healthy aging,1,3,4,6,7,27 demonstrat-ing consistent associations between obesity, smoking, andhypertension and subsequent healthy aging decades later.However, the association between midlife PA patterns andhealthy aging has been inconsistent.1,4,6,7,15,27,33-35 To ourknowledge, the association between midlife fitness andhealthy aging has not been reported. In contrast to themore inconsistent associations with PA, we observed thathigher midlife fitness was strongly and consistently as-sociated with a lower rate of CC outcomes in later life inmen and women. This discordance between PA and fit-ness is not unexpected given reports15,16,35 demonstrat-ing stronger associations between fitness and mortalitycompared with PA.

The inverse association between fitness and mortal-ity has been well studied,8-10,12-14,21 but less is known re-garding the association between midlife fitness and CCs

Table 3. Risk for Developing a Chronic Condition by Fitness and Risk Factor Levels Measured in Midlifea

Characteristic

Men(n = 14 726)

Women(n = 3944)

HR (95% CI) �2b P Value HR (95% CI) �2b P Value

Fitness, per 1 METc 0.95 (0.94-0.96) 144.3 �.001 0.94 (0.91-0.96) 27.3 �.001BMI, per 3 U 1.36 (1.13-1.63) 10.9 .001 1.63 (1.15-2.31) 7.5 .01SBP, per 20 mm Hgd 1.09 (1.07-1.11) 55.8 �.001 1.05 (1.00-1.10) 3.4 .07Total cholesterol level, per 40 mg/dLd 1.68 (1.38-2.04) 26.2 �.001 1.53 (1.02-2.30) 4.3 .04Smoking, no vs yesd 0.84 (0.80-0.87) 73.0 �.001 0.76 (0.68-0.85) 23.8 �.001Glucose level, per 1 mg/dL 1.05 (1.03-1.07) 19.7 �.001 1.03 (0.99-1.08) 1.7 .20Alcohol intake, per 1 drink/wk 1.00 (1.00-1.00) 1.7 .19 1.00 (1.00-1.01) 0.3 .58

Abbreviations: BMI, body mass index (calculated as weight in kilograms divided by height in meters squared); HR, hazard ratio; MET, metabolic equivalent;SBP, systolic blood pressure.

SI conversion factors: To convert cholesterol to millimoles per liter, multiply by 0.0259; glucose to millimoles per liter, multiply by 0.0555.aAll models were adjusted for all covariates listed in the table as well as follow-up entry age, year of examination, and time-dependent interaction terms (see

“Methods” section for details).b�2 with 1 df.cOne MET equals 3.50 mL of oxygen per kilogram per minute.dStatus at baseline.

Table 4. Proportion of Final 5 Years of Life Spent With Chronic Conditions, Stratified by Midlife Fitness Level

Fitness Level

Chronic Conditions, % (95% CI)a

0 or 1 2 or 3 �4

Q1 (n = 649), 3399 person-years before death 43.5 (41.8-45.2) 38.1 (36.5-39.7) 18.3 (17.0-19.6)Q2-Q3 (n = 1046), 6242 person-years before death 51.0 (49.8-52.2) 35.7 (34.5-36.9) 13.4 (12.6-14.2)Q4-Q5 (n = 711), 4118 person-years before death 58.3 (56.8-59.8) 32.1 (30.7-33.5) 9.5 (8.6-10.4)

Abbreviation: Q, quintile.aData represent the proportion (given as the percentage [95% CI]) of follow-up time based on person-years spent at each level of chronic condition burden

according to strata of midlife fitness levels.



later in life. This reflects limited available data for nonfa-tal clinical events among established fitness cohorts withlong-term follow-up. In the present study, we combinedMedicare administrative claims data with a large cohortof healthy men and women with objectively measured fit-ness levels, providing an efficient strategy to examine theassociation between midlife fitness and the developmentof a diverse set of CCs decades later. Our findings sup-port the hypothesis that fitness in midlife is associated witha lower burden of chronic disease in later life.

In addition to reducing the burden of CCs, we alsoobserved that higher midlife fitness was associated morestrongly with the delay in the onset of CCs than with sur-vival, suggesting that higher midlife fitness may pro-mote the compression of morbidity in later life.36-39 Be-cause of the strong correlation between morbidity andmortality, lifestyle patterns that equally delay the onsetof both morbidity and mortality could result in more yearsof life lived with chronic disease.40 In contrast, lifestylecharacteristics that delay the onset of chronic disease toa greater extent than they prolong the lifespan could theo-retically compress life-years lived with chronic diseaseand hence increase the years with improved quality oflife and lower health care expenditures.38,40-43 Our find-ings have important implications for public health andprevention practice by extending our knowledge of thehealth benefits of exercise in midlife.

Several studies37-39,44,45 suggested the importance ofphysical exercise as a potential source of morbidity com-pression. However, most were relatively small with fewoverall deaths, limiting the ability to test for the pres-ence of this phenomenon. In a cohort of middle-aged run-ners and control individuals monitored for 21 years(N = 961), runners were found to have less disability asassessed by the Health Assessment Questionnaire Dis-ability Index, suggesting that regular physical exercisemight delay the onset of disability.46 However, the au-thors were not able to assess for the presence or absenceof morbidity compression because of the overall smallnumber of deaths in the study sample (n = 225).46 In thepresent study, we included 18 670 participants with 2406deaths, allowing a comparison of the associations be-tween fitness, CCs, and mortality in the final 5 years oflife.

Several limitations of this study should be noted. First,outcome data were derived from administrative data fromthe CMS rather than adjudicated clinical diagnoses. Nev-ertheless, Medicare data have been shown to be a reli-able source of information across multiple clinical out-comes.25,26,47-52 Furthermore, Medicare data represent aunique and cost-effective resource, providing an oppor-tunity to assess the association between midlife fitnesslevels and long-term chronic disease outcomes that wouldbe prohibitively expensive, if not impossible, to repli-cate in a prospective cohort study of comparable size andduration.

Second, we linked individual-level data with Medi-care claims files to compare the association between fit-ness and chronic disease outcomes at age 65 years or older.We were not able to capture outcomes that occurred be-tween study entry and the onset of Medicare eligibility.For example, participants who died before achieving Medi-

care eligibility were not included in the present analy-sis. However, merging individual-level data with Medi-care claims files has been used by other investigators ina parallel context, providing novel insight into the con-tribution of traditional risk factors and other Medicareoutcomes.2,27-29 We observed a similar pattern of resultsfor fitness levels measured in earlier life (ie, age �49 years)as well as in later life (ie, age �49 years) and closer toMedicare eligibility.

Third, we created an a priori panel of CCs across mul-tiple organ systems and in accordance with definitionsof healthy aging defined by others.4-7,27,53 Although thismay have influenced our findings, we observed a simi-lar pattern of results after multiple sensitivity analysesthat sequentially excluded one of these CCs (eTable 2).Therefore, we believe that our findings are insensitive tothe choice of CC panel and reflect a reasonable estimateof the association between midlife fitness and CC out-comes in older age.

Fourth, the CCLS is a unique cohort with a higher so-cioeconomic and educational status and a lower preva-lence of traditional risk factors when compared with thegeneral population. However, although the level of riskfactors is lower than in the general population, prior worksuggests that the effects are similar.54

Finally, additional factors, such as life stress and di-etary patterns, were not included in our analyses be-cause these data are limited in the CCLS. Additional co-variates could alter the observed associations of CCs andfitness. Follow-up data on fitness and PA are not uni-formly available in the CCLS; therefore, we are not ableto estimate the effect of changes in fitness with chronicdisease burden in older age. However, our primary pur-pose was to determine the contribution of a single mea-sure of midlife fitness with surrogates for healthy agingdecades later. Furthermore, although fitness measuresmore proximate to the outcome would be of interest, itwould also raise concerns regarding the possibility of re-verse causation, in which low fitness reflected undiag-nosed chronic disease burden. The healthy nature of ourcohort and the long duration of follow-up make the pres-ence of undiagnosed CCs at study entry unlikely and fur-ther support the hypothesis that greater midlife exerciseis associated with a lower burden of chronic disease acrossthe life span.

In summary, midlife fitness was associated with a lowerrisk of common chronic health conditions in men andwomen older than 65 years enrolled in Medicare. The find-ing that higher fitness was more strongly associated withCCs than with overall survival suggests that higher midlifefitness may be associated with the compression of mor-bidity in older age.

Accepted for Publication: May 30, 2012.Published Online: August 27, 2012. doi:10.1001/archinternmed.2012.3400Correspondence: Jarett D. Berry, MD, MS, Division ofCardiology, The University of Texas Southwestern Medi-cal Center, 5323 Harry Hines Blvd, Dallas, TX 75390([email protected]).Author Contributions: Dr Willis had full access to all thedata in the study and takes responsibility for the integ-



rity of the data and the accuracy of the data analysis. DrBerry had final responsibility for the decision to submitfor publication. All authors have read and agree to themanuscript as written. Study concept and design: Willis,DeFina, and Berry. Acquisition of data: Willis. Analysis andinterpretation of data: Willis, Gao, Leonard, DeFina, andBerry. Drafting of the manuscript: Willis and Berry. Criti-cal revision of the manuscript for important intellectual con-tent: Willis, Gao, Leonard, DeFina, and Berry. Statisticalanalysis: Willis, Gao, and Leonard. Obtained funding: Berry.Administrative, technical, and material support: Willis andDeFina. Study supervision: Willis and Berry.Financial Disclosure: Dr Berry has received financial com-pensation from Merck for being a member of its speak-ers’ bureau.Funding/Support: The Cooper Institute is a 501(c)(3)nonprofit research institute and provided internal fund-ing for this study. Dr Berry receives funding from the Ded-man Family Scholar in Clinical Care endowment at TheUniversity of Texas Southwestern Medical Center; grantK23 HL092229 from the National Heart, Lung, and BloodInstitute; and grant 10BG1A4280091 from the Ameri-can Heart Association.Role of the Sponsors: The University of Texas South-western Medical Center; the National Heart, Lung, andBlood Institute; and the American Heart Association madeno contribution to the design of the study. In addition,the sponsors made no contribution to the collection, man-agement, analysis, or interpretation of the data, or prepa-ration, review, or approval of the manuscript.Online-Only Material: The eTables are available at http://www.archinternmed.com.Additional Contributions: The authors thank KennethH. Cooper, MD, MPH, for establishing the Cooper Cen-ter Longitudinal Study in 1970 and the Cooper Clinic phy-sicians and staff for data collection.

REFERENCES

1. Britton A, Shipley M, Singh-Manoux A, Marmot MG. Successful aging: the con-tribution of early-life and midlife risk factors. J Am Geriatr Soc. 2008;56(6):1098-1105.

2. Burke GL, Arnold AM, Bild DE, et al; CHS Collaborative Research Group. Factorsassociated with healthy aging: the Cardiovascular Health Study. J Am Geriatr Soc.2001;49(3):254-262.

3. Daviglus ML, Liu K, Pirzada A, et al. Favorable cardiovascular risk profile in middleage and health-related quality of life in older age. Arch Intern Med. 2003;163(20):2460-2468.

4. Sun Q, Townsend MK, Okereke OI, Franco OH, Hu FB, Grodstein F. Adiposity andweight change in mid-life in relation to healthy survival after age 70 in women:prospective cohort study. BMJ. 2009;339:b3796. doi:10.1136/bmj.b3796.:b3796.

5. Terry DF, Pencina MJ, Vasan RS, et al. Cardiovascular risk factors predictive forsurvival and morbidity-free survival in the oldest-old Framingham Heart Studyparticipants. J Am Geriatr Soc. 2005;53(11):1944-1950.

6. Willcox BJ, He Q, Chen R, et al. Midlife risk factors and healthy survival in men.JAMA. 2006;296(19):2343-2350.

7. Yates LB, Djousse L, Kurth T, Buring JE, Gaziano JM. Exceptional longevity inmen: modifiable factors associated with survival and function to age 90 years.Arch Intern Med. 2008;168(3):284-290.

8. Blair SN, Kohl HW III, Paffenbarger RS Jr, Clark DG, Cooper KH, Gibbons LW.Physical fitness and all-cause mortality: a prospective study of healthy men andwomen. JAMA. 1989;262(17):2395-2401.

9. Blair SN, Kohl HW III, Barlow CE, Paffenbarger RS Jr, Gibbons LW, Macera CA.Changes in physical fitness and all-cause mortality: a prospective study of healthyand unhealthy men. JAMA. 1995;273(14):1093-1098.

10. Ekelund LG, Haskell WL, Johnson JL, Whaley FS, Criqui MH, Sheps DS. Physi-

cal fitness as a predictor of cardiovascular mortality in asymptomatic North Ameri-can men: the Lipid Research Clinics Mortality Follow-up Study. N Engl J Med.1988;319(21):1379-1384.

11. Kodama S, Saito K, Tanaka S, et al. Cardiorespiratory fitness as a quantitativepredictor of all-cause mortality and cardiovascular events in healthy men andwomen: a meta-analysis. JAMA. 2009;301(19):2024-2035.

12. Kokkinos P, Myers J, Kokkinos JP, et al. Exercise capacity and mortality in blackand white men. Circulation. 2008;117(5):614-622.

13. Mora S, Redberg RF, Sharrett AR, Blumenthal RS. Enhanced risk assessment inasymptomatic individuals with exercise testing and Framingham risk scores.Circulation. 2005;112(11):1566-1572.

14. Myers J, Prakash M, Froelicher V, Do D, Partington S, Atwood JE. Exercise ca-pacity and mortality among men referred for exercise testing. N Engl J Med. 2002;346(11):793-801.

15. Myers J, Kaykha A, George S, et al. Fitness versus physical activity patterns inpredicting mortality in men. Am J Med. 2004;117(12):912-918.

16. Talbot LA, Morrell CH, Metter EJ, Fleg JL. Comparison of cardiorespiratory fit-ness versus leisure time physical activity as predictors of coronary events in menaged �65 years and �65 years. Am J Cardiol. 2002;89(10):1187-1192.

17. Willis BL, Morrow JR Jr, Jackson AW, DeFina LF, Cooper KH. Secular change incardiorespiratory fitness of men: Cooper Center Longitudinal Study. Med Sci SportsExerc. 2011;43(11):2134-2139.

18. Pollock ML, Bohannon RL, Cooper KH, et al. A comparative analysis of four pro-tocols for maximal treadmill stress testing. Am Heart J. 1976;92(1):39-46.

19. Pollock ML, Foster C, Schmidt D, Hellman C, Linnerud AC, Ward A. Comparativeanalysis of physiologic responses to three different maximal graded exercise testprotocols in healthy women. Am Heart J. 1982;103(3):363-373.

20. FranklinBA,WhaleyMH,HowleyET,BaladyGJ.ACSM’sGuidelines forExerciseTest-ing and Prescription. 6th ed. Philadelphia, PA: Lippincott Williams & Wilkins, 2000.

21. Blair SN, Kampert JB, Kohl HW III, et al. Influences of cardiorespiratory fitnessand other precursors on cardiovascular disease and all-cause mortality in menand women. JAMA. 1996;276(3):205-210.

22. Blair SN, Kannel WB, Kohl HW, Goodyear N, Wilson PW. Surrogate measures ofphysical activity and physical fitness: evidence for sedentary traits of resting tachy-cardia, obesity, and low vital capacity. Am J Epidemiol. 1989;129(6):1145-1156.

23. Kohl HW, Blair SN, Paffenbarger RS Jr, Macera CA, Kronenfeld JJ. A mail sur-vey of physical activity habits as related to measured physical fitness. AmJ Epidemiol. 1988;127(6):1228-1239.

24. Chronic Condition Data Warehouse User Guide. CMS Chronic Condition Warehouse.2011:1-22. http://www.ccwdata.org. Accessed September 19, 2011.

25. Gorina Y, Kramarow EA. Identifying chronic conditions in Medicare claims data:evaluating the Chronic Condition Data Warehouse algorithm. Health Serv Res.2011;46(5):1610-1627.

26. Virnig BA, McBean M. Administrative data for public health surveillance andplanning. Annu Rev Public Health. 2001;22:213-230.

27. Sun Q, Townsend MK, Okereke OI, Franco OH, Hu FB, Grodstein F. Physical ac-tivity at midlife in relation to successful survival in women at age 70 years orolder. Arch Intern Med. 2010;170(2):194-201.

28. Wei LJ, Lin DY, Weissfeld L. Regression analysis of multivariate incomplete fail-ure time data by modeling marginal distribution. J Am Stat Assoc. 1989;84:1065-1073. doi:10.1080/01621459.1989.10478873.

29. Lin DY, Wei LJ. The robust inference for the Cox proportional hazards model.J Am Stat Assoc. 1989;84:1074-1078. doi:10.1080/01621459.1989.10478874.

30. Duncan JJ, Gordon NF, Scott CB. Women walking for health and fitness: howmuch is enough? JAMA. 1991;266(23):3295-3299.

31. Oja P. Dose response between total volume of physical activity and health andfitness. Med Sci Sports Exerc. 2001;33(6)(suppl):S428-S437.

32. Skinner JS, Jaskolski A, Jaskolska A, et al; HERITAGE Family Study. Age, sex,race, initial fitness, and response to training: the HERITAGE Family Study. J ApplPhysiol. 2001;90(5):1770-1776.

33. Roos NP, Havens B. Predictors of successful aging: a twelve-year study of Mani-toba elderly. Am J Public Health. 1991;81(1):63-68.

34. Guralnik JM, Kaplan GA. Predictors of healthy aging: prospective evidence fromthe Alameda County study. Am J Public Health. 1989;79(6):703-708.

35. Lee DC, Sui X, Ortega FB, et al. Comparisons of leisure-time physical activity andcardiorespiratory fitness as predictors of all-cause mortality in men and women.Br J Sports Med. 2011;45(6):504-510.

36. Fries JF. Aging, natural death, and the compression of morbidity. N Engl J Med.1980;303(3):130-135.

37. Fries JF. Physical activity, the compression of morbidity, and the health of theelderly. J R Soc Med. 1996;89(2):64-68.

38. Fries JF. Measuring and monitoring success in compressing morbidity. Ann In-tern Med. 2003;139(5, pt 2):455-459.

39. Fries JF. Frailty, heart disease, and stroke: the Compression of Morbidity paradigm.Am J Prev Med. 2005;29(5)(suppl 1):164-168.



40. Gruenberg EM. The failures of success: 1977. Milbank Q. 2005;83(4):779-800.41. Vogeli C, Shields AE, Lee TA, et al. Multiple chronic conditions: prevalence, health

consequences, and implications for quality, care management, and costs. J GenIntern Med. 2007;22(suppl 3):391-395.

42. Graham P, Blakely T, Davis P, Sporle A, Pearce N. Compression, expansion, ordynamic equilibrium? the evolution of health expectancy in New Zealand. J Epi-demiol Community Health. 2004;58(8):659-666.

43. Fries JF, Koop CE, Beadle CE, et al; Health Project Consortium. Reducing healthcare costs by reducing the need and demand for medical services. N Engl J Med.1993;329(5):321-325.

44. Hubert HB, Bloch DA, Fries JF. Risk factors for physical disability in an agingcohort: the NHANES I Epidemiologic Follow-up Study. J Rheumatol. 1993;20(3):480-488.

45. Hubert HB, Bloch DA, Oehlert JW, Fries JF. Lifestyle habits and compression ofmorbidity. J Gerontol A Biol Sci Med Sci. 2002;57(6):M347-M351.

46. Chakravarty EF, Hubert HB, Lingala VB, Fries JF. Reduced disability and mortal-ity among aging runners: a 21-year longitudinal study. Arch Intern Med. 2008;168(15):1638-1646.

47. Wolff JL, Starfield B, Anderson G. Prevalence, expenditures, and complicationsof multiple chronic conditions in the elderly. Arch Intern Med. 2002;162(20):2269-2276.

48. Virnig B, Durham SB, Folsom AR, Cerhan J. Linking the Iowa Women’s HealthStudy cohort to Medicare data: linkage results and application to hip fracture.Am J Epidemiol. 2010;172(3):327-333.

49. Daviglus ML, Liu K, Yan LL, et al. Relation of body mass index in young adult-hood and middle age to Medicare expenditures in older age. JAMA. 2004;292(22):2743-2749.

50. Daviglus ML. Health care costs in old age are related to overweight and obesityearlier in life. Health Aff (Millwood). 2005;24(suppl 2):W5R97-W5R100.

51. Daviglus ML, Liu K, Pirzada A, et al. Cardiovascular risk profile earlier in life andMedicare costs in the last year of life. Arch Intern Med. 2005;165(9):1028-1034.

52. Daviglus ML, Liu K, Pirzada A, et al. Relationship of fruit and vegetable con-sumption in middle-aged men to Medicare expenditures in older age: the Chi-cago Western Electric Study. J Am Diet Assoc. 2005;105(11):1735-1744.

53. Britton A, Shipley M, Singh-Manoux A, Marmot MG. Successful aging: the con-tribution of early-life and midlife risk factors. J Am Geriatr Soc. 2008;56(6):1098-1105.

54. Berry JD, Willis B, Gupta S, et al. Lifetime risks for cardiovascular disease mor-tality by cardiorespiratory fitness levels measured at ages 45, 55, and 65 yearsin men: the Cooper Center Longitudinal Study. J Am Coll Cardiol. 2011;57(15):1604-1610.

INVITED COMMENTARY

Thriving of the Fittest

I n 1980, James F. Fries, MD, published his seminalthesis on the limit of the average human life span—which he pinpointed at 85 years—and the com-

pression morbidity into the final period of life, which heoffered as both observation of a societal trend and a pub-lic health goal.1 He whimsically invoked this ideal by cit-ing a poem by Oliver Wendell Holmes Sr about a “one-hoss shay,” a carefully constructed carriage that functionsbeautifully and then breaks down “to pieces all at once”after exactly 100 years. Fries contrasts this vision to thenotion that ever-increasing life span afforded by improve-ments in public health and medicine would lead to in-creasing proportions of the population spending manyof their final years with significant infirmity. Fries andhis colleagues2 have continued to find evidence for thecompression of morbidity and the “rectangularization”of the survival curve.

The concept of healthy or “successful” aging, once con-sidered almost an oxymoron, was introduced in the 1960sand 1970s and has since been an area of active investiga-tion. Healthy aging has been found to be related to severallifestyle factors, includingabstinence fromsmoking,physi-cal activity, maintenance of weight within normal ranges,and moderate alcohol consumption.3 In recent years, cluesto the genetics of longevity have begun to emerge,4 and ge-netics undoubtedly plays an important role in maintaininggoodhealthaswellasavoidingdisease.Severalstudies5 havealsofoundthatcardiorespiratoryfitnessmeasuredatasingletime is stronglyassociatedwithboth longevityandreducedrisk of chronic disease, particularly cardiovascular disease,and is abetterpredictor thanphysical activity levels, likely,in part, because of the objectivity of the measurement andability to capture the cumulative effects of exercise.

Willis and colleagues6 provide further evidence forphysical fitness as a contributor to healthy aging and thecompression of morbidity. By linking the large clinicaldatabase comprising men and women who visited theCooper Clinic from 1970 to 2009 and underwent stan-dardized treadmill fitness testing to Medicare claims whenthey reached age 65 years or older, the authors identi-fied 18 670 persons who were free from chronic diseaseat the time of their examinations and were covered byMedicare from 1999 to 2009. After adjusting for age, bodymass index, systolic blood pressure, total cholesterol level,smoking, fasting blood glucose level, and alcohol in-take, there was a strong graded relationship of fitness tothe rate of development of a set of common chronic con-ditions, including ischemic heart disease, congestive heartfailure, stroke, diabetes mellitus, chronic obstructive pul-monary disease, chronic kidney disease, Alzheimer dis-ease, and colon or lung cancer. The authors found an ap-proximate 6% reduction in the risk of a chronic diseasefor every MET achieved, with a range of 5 to 6 METs acrossquintiles of fitness or an approximate doubling of riskbetween quintiles 1 and 5. This relationship was robustand similar when individual diseases were eliminated fromthe common set of conditions. Fitness appeared stronglyprotective against each condition, and there were simi-lar relationships in men and women.

Furthermore, in an analysis of 2406 decedents, therewas a clear relationship between fitness and the propor-tion of the final 5 years of life spent with a chronic con-dition. The highest of 3 fitness groups spent approxi-mately 50% as much time with 4 or more chronic diseasesas the lowest fitness group and 34% more time with noor 1 chronic disease.

Scan for AuthorAudio Interview



midlife fitness and the development of chronic conditions in later lifemidlife fitness and chronic...

Documents