American Journal of Epidemiology
© The Author 2013. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of
Public Health. All rights reserved. For permissions, please e-mail: [email protected].
Vol. 178, No. 9
DOI: 10.1093/aje/kwt157
Advance Access publication:
September 5, 2013
Original Contribution
The “Obesity Paradox,” Frailty, Disability, and Mortality in Older Men:
A Prospective, Longitudinal Cohort Study
Timo E. Strandberg*, Sari Stenholm, Arto Y. Strandberg, Veikko V. Salomaa, Kaisu H. Pitkälä, and
Reijo S. Tilvis
* Correspondence to Dr. Timo E. Strandberg, Department of Health Sciences/Geriatrics, University of Oulu, Aapistie 5, PO Box 5000,
Oulun Yliopisto 90014, Finland (e-mail: [email protected]).
Initially submitted April 27, 2013; accepted for publication June 11, 2013.
An inverse relationship between overweight and mortality (the “obesity paradox”) is well documented, but there
are scarce data on how body weight during the life course affects this relationship. In the Helsinki Businessmen
Study, we examined the effect of weight trajectories on incident disability, frailty, and mortality by stratifying 1,114
men (mean age of 47 years in 1974) into the following 4 groups based on body mass index (weight (kg)/height (m)2)
values measured twice, in 1974 and 2000: 1) constantly normal weight (n = 340, reference group); 2) constantly
overweight (n = 495); 3) weight gain (n = 136); and 4) weight loss (n = 143). Twelve-year mortality rates (from 2000 to
2012) and frailty and mobility-related disability in late life were determined. Compared with constantly normal weight,
weight loss was associated with disability (odds ratio (OR) = 2.4, 95% confidence interval (CI): 1.1, 4.9) and frailty
(OR = 3.7, 95% CI: 1.3, 10.5) in late life. Constant overweight was associated with increased disability (OR = 1.9,
95% CI: 1.1, 3.2). Men with constantly normal weight had the fewest comorbidities in late life (P < 0.001). Higher 12year mortality rates were observed both with weight loss (hazard ratio = 1.8, 95% CI: 1.3, 2.3) and with constant
overweight (hazard ratio = 1.3, 95% CI: 1.03, 1.7). Those with constantly normal weight or weight gain had similar
outcomes. We observed no obesity paradox in late life when earlier weight trajectories were taken into account.
aged; disability; frailty; life course; mortality; obesity paradox
Abbreviations: BMI, body mass index; CI, confidence interval; OR, odds ratio.
The “obesity paradox” involves the notion that overweight,
or even obesity, is somehow protective in regard to chronic disease, such as cardiovascular disease, and during late life (1–8).
A recent, large systematic review has further strengthened the
view of beneficial overweight (9, 10) and contradicts prevailing
health advice that leanness is optimal for health. In our earlier
analysis explaining the obesity paradox in the Helsinki BusinessmenStudy (11),wenoted that,inlatelife,bothnormalweight
men and overweight men had had different weight trajectories during their life courses. We observed a clearly higher mortality rate among those overweight men who had become normal
weight since midlife. Because this higher mortality rate was
independent of clinical diseases, we speculated that the development of frailty—a geriatric syndrome characterized by reduced
physiological reserves and increased vulnerability, but not
necessarily clinical disease or disability (12, 13)—could be the
background mechanism (11). We anticipated that longer followup would reveal differences between the weight trajectory groups
and further elucidate the existence of the obesity paradox.
The aim of this study was to extend our previous report by
examining the relationship between weight trajectory and incident mobility-related disability and frailty in late life. In addition,
by using the prolonged mortality follow-up data from the previous 7 years, and up to 12 years, we examined the association
between weight trajectory and death, assuming that longer follow-up would result in more separation of the mortality curves.
MATERIALS AND METHODS
The cohort and examinations in the Helsinki Businessmen
Study have been described earlier (11, 14). A flow chart pertaining to the present analyses is provided in Figure 1. Briefly,
1452
Am J Epidemiol. 2013;178(9):1452–1460
Obesity Paradox in Older Men 1453
1,815 Clinically healthy men, born 1919–1934, participating in the
Helsinki Businessmen Study
(In 1974, provided recall of weight at 25 years of age)
425 Died
1,390 Contacted with mailed questionnaires in 2000
276 Did not respond or had missing
variables
1,114 Divided into 4 groups according to body mass index trajectories in 2000
2 0 9 D i ed
283 and 295 Missing variables for
phenotypic frailty or mobility disability,
respectively
622 and 610 Assessed for phenotypic frailty or mobility disability, respectively, in 2007
Vital status verified from the Population Information System in 2000–2012
Figure 1. Flow chart of the Helsinki Businessmen Study, 1974–2012. Body mass index (weight (kg)/height (m)2) values were available at an
average age of 40 years for 1,114 subjects, 52 years for 1,043 subjects, and 58 years for 932 subjects.
the present cohort consisted of men who were healthy in midlife (in 1974, at a mean age of 47 years) and alive in late life (in
2000, at a mean age of 73 years) and whose body mass index
(BMI) (weight (kg)/height (m)2) trajectories could be followed into late life. The BMI trajectory groups were assessed
with regard to the development of frailty and to death within
a follow-up period of 12 years. The follow-up studies of the
Helsinki Businessmen Study have been approved by the ethics
committee of the Department of Medicine, University of Helsinki, Finland.
hypertension) with no history or clinical signs of chronic diseases including diabetes; no electrocardiographic abnormalities pertaining to heart disease; and no use of regular medications
such as antihypertensive, antilipidemic, or antidiabetic drugs.
BMI measurements in 1979–1980 and 1985–1986
The participants were clinically examined in 1979–1980
and 1985–1986, with BMI measured in 1,043 men (93.6%
of the cohort defined below) and 932 men (83.7%), respectively.
Examinations in 1974
Initially healthy men, who were mostly business executives
born in 1919–1934, participated in structured health examinations (including measurement of BMI at an average age of
40 years) during the 1960s and early 1970s at the Institute of
Occupational Health in Helsinki, Finland. In 1974, subjects
were thoroughly assessed with questionnaires and clinical and
laboratory (including serum lipid and 1-hour postload blood
glucose)examinations (11).Weight andheight were measured,
and BMI was calculated. Overweight was defined as a BMI
value of 25 or greater, and normal weight was defined as a BMI
value of less than 25. In 1974, the men were asked to recall their
weight at 25 years of age, and their BMI values were calculated. Global, self-rated health status was assessed by asking,
“What do you think about your present state of health; is it very
good, fairly good, average, fairly poor, or very poor?” Because
few men perceived their health status as “very poor,” they were
combined with the “fairly poor” group. The wording of selfrated health questions in our study was similar to that used in
the Whitehall II Study (15). A total of 1,815 men were found
to be clinically healthy (albeit often with risk factors such as
Am J Epidemiol. 2013;178(9):1452–1460
Examinations in 2000
In 2000, we mailed a questionnaire to the 1,390 surviving subjects (remailed once for nonrespondents), and 1,231 (88.6%)
responded. The distribution of overweight in 1974 was not significantly different (P = 0.62) between respondents and nonrespondents who were alive in 2000. In addition to questions
about their present weight, lifestyle, and history of diseases,
the questionnaire included the Finnish version of the RAND
36-item health survey, version 1.0 (RAND Corporation, Santa
Monica, California). This survey is practically identical to the
RAND 36-item short form health survey (SF-36), and has been
validated in the Finnish population (16). Data on self-reported
diseases were used to compose the summary comorbidity measure (17). Mobility-related disability was defined as any reported
difficulty walking 500 m. Frailty was defined by using a modification of the “Fried criteria” (12) as previously described
(14) and on the basis of 4 criteria available from the questionnaire and data on weight change during follow-up. Participants were classified as frail or prefrail if 3–4 or 1–2 of
the criteria were met, respectively, and nonfrail if none of
1454 Strandberg et al.
the criteria was present. Our modification of the definition of
frailty has been shown to predict mobility-related disability
and death in this cohort (14).
RESULTS
Weight trajectory groups in 2000
The weight trajectories over the life course from age 25
years are presented in Figure 2. Whereas the constant weight
groups largely maintained their BMI values, there was an opposite development in the other 2 groups during the seventh decade
of life.
BMI values ranged from 18.7 to 39.7 in 1974 and from 17.3
to 44.3 in 2000. Consequently, no man in our study was underweight (BMI <18.5) in midlife. The prevalence of obesity
(BMI ≥30) was 6.0% and 9.8% in 1974 and 2000, respectively. Cardiovascular risk factors were common in midlife. Of
the total cohort in 1974, 26.4% (n = 294) were smokers, 31.6%
(n = 352) were hypertensive (blood pressure >140/90 mm Hg),
44.3% (n = 492) had hypercholesterolemia (serum cholesterol
>6.2 mmol/L), and 14.3% (n = 159) were glucose intolerant
(1-hour postload blood glucose >9 mmol/L).
Table 1 shows the characteristics of the groups in 1974,
2000, and 2007 according to weight trajectory. Cardiovascular
risk factors (blood pressure, serum lipids, and blood glucose)
were measured at baseline only, and those values should be
considered in light of contemporary (in 1974) BMI values.
Self-rated health in midlife was significantly different (P =
0.003) among the weight trajectory groups and tended to be
better among those whose weight was normal in midlife. In
those who were in late life in 2000, the comorbidity indexes
were significantly different (P < 0.001) between the groups
and lowest among those with constantly normal weight.
Among the respondents in 2000, there were 1,114 men for
whom BMI values at various ages were available (80.1% of
eligible men), and they formed the present cohort for the frailty,
mobility-related disability, and mortality follow-up from 2000
onward. Participants were divided into the following 4 weight
trajectory groups according to BMI values in 1974 and 2000
(7): 1) Constantly normal weight group (n = 340) with BMI
values less than 25 in 1974 and 2000; 2) constantly overweight
group (n = 495) with BMI values of 25 or greater in 1974 and
2000; 3) weight gain group (n = 136) with BMI values less
than 25 in 1974 and 25 or greater in 2000; and 4) weight loss
group (n = 143) with BMI values of 25 or greater in 1974 and
less than 25 in 2000. The prevalence of obesity (BMI ≥30)
was less frequent than in contemporary cohorts (18), and therefore we did not divide overweight further.
Survey in 2007
In 2007, we mailed a questionnaire including the RAND36 survey instrument to all survivors (n = 905), and 666 (73.6%)
responded. The phenotype of frailty was determined according
to 4 criteria in 2000, and weight loss was defined as greater than
5% loss between 2000 and 2007. Mobility-related disability
was defined in 2000 as any difficulty walking 500 m.
Characteristics of the groups according to weight
trajectory from 1974 to 2000
Twelve-year mortality follow-up
Data on death within the study population through July 31,
2012, were retrieved from the Finnish Population Information System (Population Register Centre, Helsinki, Finland),
which keeps a registry of all Finnish citizens; thus, our determination of vital status is virtually 100% complete.
30
Dotted = Constantly overweight
Dashed-dotted = Constantly normal
Dashed = Weight loss
Solid = Weight gain
N=495
Number Crunching Statistical System, version 2007, software (NCSS Statistical Software, Kaysville, Utah) was used
for the statistical analyses. The weight change groups defined
in 2000 were categorized as described above. Student’s t tests,
nonparametric tests, and analysis of covariancewere used, where
appropriate, to compare continuous variables. χ2 and trend tests
were used to compare proportions. Kaplan-Meier survival curves
were constructed to compare survival rates within weight trajectory groups, and differences were analyzed with the logrank test. Cox regression analysis was used to calculate adjusted
hazard ratios with 95% confidence intervals for mortality associated with the weight trajectory groups. Multinominal logistic
regression was used to assess the relationship between weight
trajectories and frailty and mobility-related disability in 2000
and 2007, and odds ratios with 95% confidence intervals were
calculated. Covariates in the regression analyses are reported
for respective models. In statistical analyses, 2-sided P values
of less than 0.05 were considered significant.
Mean BMI
Statistical analysis
N=136
25
20
25
40
47
52
58
73
Age, years
Figure 2. Weight trajectories during the life course from age 25
years to an average of 73 years in 2000 in the Helsinki Businessmen
Study. The constantly overweight group (n = 495) had body mass
index (BMI) (weight (kg)/height (m)2) values of 25 or greater at ages
47 and 73 years. The constantly normal weight group (n = 340) had
BMI values of less than 25 at ages 47 and 73 years. The weight loss
group (n = 143) had BMI values of 25 or greater at age 47 years and
less than 25 at age 73 years. The weight gain group (n = 136) had
BMI values less than 25 at age 47 years and 25 or greater at age 73
years.
Am J Epidemiol. 2013;178(9):1452–1460
Obesity Paradox in Older Men 1455
Table 1. Characteristics of 1,114 Men in the Helsinki Businessmen Study by Weight Status Group in 1974, 2000, and 2007
Weight Status Group
Characteristica
Constantly Normal Weightb
(n = 340)
Mean (SE)
No.
%
Constantly Overweightc
(n = 495)
Mean (SE)
No.
%
Weight Gaind
(n = 136)
Mean (SE)
No.
Weight Losse
(n = 143)
%
Mean (SE)
No.
P
Value
%
Age, years
In 1974
47.6 (0.2)
46.9 (0.2)
46.4 (0.3)
48.9 (0.3)
<0.001
In 2000
73.6 (0.2)
72.9 (0.2)
72.4 (0.3)
74.9 (0.3)
<0.001
In 2007
79.1 (0.5)
78.3 (0.2)
78.0 (0.4)
79.9 (0.5)
0.003
Smokers
In 1974
80
23.5
127
25.7
49
36.0
38
26.6
0.04
In 2000
36
10.6
29
5.9
10
7.4
19
13.3
0.01
In 2007
11
5.3
10
3.3
5
5.6
2
3.1
0.61
Height in 1974, cm
177 (0.3)
176 (0.3)
177 (0.5)
177 (0.5)
0.19
In 1974
73.0 (0.4)
86.5 (0.3)
75.1 (0.7)
82.7 (0.7)
<0.001
In 2000
72.1 (0.4)
87.9 (0.4)
83.1 (0.7)
74.4 (0.7)
<0.001
In 2007
71.6 (0.6)
85.5 (0.5)
80.0 (0.5)
73.2 (1.1)
<0.001
Weight change
between 1974
and 2000, kg
−0.9 (0.3)
+1.4 (0.3)
+8.0 (0.5)
−8.3 (0.5)
<0.001
Weight change
between 2000
and 2007, kg
−1.1 (0.4)
−2.2 (0.3)
−2.3 (0.6)
−1.8 (0.7)
0.14
In 1974
23.2 (0.1)
27.7 (0.1)
23.9 (0.1)
26.4 (0.1)
<0.001
In 2000
22.9 (0.1)
28.2 (0.1)
26.5 (0.2)
23.7 (0.2)
<0.001
In 2007
22.7 (0.2)
27.5 (0.1)
25.5 (0.3)
23.4 (0.3)
<0.001
Systolic
137.4 (1.0)
145.2 (0.8)
137.6 (1.6)
144.1 (1.5)
<0.001
Diastolic
87.7 (0.6)
94.2 (0.5)
87.4 (1.0)
92.3 (0.9)
<0.001
Body weight, kg
BMIf
Blood pressure in
1974, mm Hg
Serum lipids in
1974, mmol/L
Cholesterol
6.0 (0.05)
6.2 (0.05)
5.7 (0.09)
6.3 (0.09)
Triglycerides
1.3 (0.04)
1.8 (0.04)
1.3 (0.07)
1.6 (0.07)
<0.001
<0.001
One-hour postload
glucose in
1974, mmol/L
6.6 (0.1)
7.3 (0.09)
6.5 (0.2)
6.8 (0.2)
<0.001
Self-rated health
in 1974
global
0.003
Very good
28
8.3
26
5.3
13
9.6
5
3.5
Fairly good
183
54.0
231
46.7
84
61.8
72
50.3
Average
121
35.7
222
44.8
37
27.2
61
42.7
7
2.1
16
3.2
2
1.5
5
3.5
Fairly poor or
very poor
Comorbidity index
In 2000
1.1 (0.07)
1.5 (0.06)
1.4 (0.1)
1.5 (0.1)
<0.001
In 2007
2.0 (0.1)
2.1 (0.09)
1.9 (0.2)
2.1 (0.2)
0.64
Abbreviations: BMI, body mass index; SE, standard error.
a
Age-adjusted.
b
Constantly normal weight represents BMI values of less than 25 in 1974 and 2000.
c
Constantly overweight represents BMI values of 25 or greater in 1974 and 2000.
d
Weight gain represents BMI values of less than 25 in 1974 and 25 or greater in 2000.
e
Weight loss represents BMI values of 25 or greater in 1974 and less than 25 in 2000.
f
Weight (kg)/height (m)2.
Am J Epidemiol. 2013;178(9):1452–1460
1456 Strandberg et al.
Mortality during follow-up
Between 1974 and 2000, the total mortality rate was significantly higher among overweight men than among men of
normal weight in 1974 (25.8% vs. 20.1%, P = 0.003; 425 total
deaths). Between 2000 and 2012, there were 463 deaths among
the 1,114 men whose weight trajectories could be determined
in 2000. There was no statistically significant difference for the
12-year mortality rate between normal-weight and overweight
men (42.2% vs. 41.1%, P = 0.70).
However, division by weight trajectory provided important
new information (Figure 3). The 12-year mortality ratewas clearly
highest in the weight loss group. However, with the longer followup, there was also an increasing trend of death in the constantly
overweight group compared with the constantly normal weight
group. Overall, the log-rank P value was significant between the
groups, suggesting that more detailed and adjusted analyses
are needed.
Adjusted analyses of mortality rates by using Cox regression are presented in Table 2. With the constantly normal weight
group as the reference group, the mortality rates between 2000
and 2012 were 30% higher for those who were constantly
overweight and 80% higher for the weight loss group, but similar among those men who had gained weight after midlife.
These differences in mortality rate were insensitive to adjustments for self-rated health and smoking in midlife and prevalent diseases in late life. The results were virtually unchanged
when men with BMI values of less than 20 in 1974 were
excluded. However, when frailty status in 2000 was added to
the model, the mortality risk associated with the weight loss
group was clearly attenuated (Table 2).
Development of frailty and mobility-related disability
Frailty status could be determined for 1,058 men (94.8%)
in 2000 and 622 men (68.7% of survivors) in 2007. The prevalence of frailty in 2000 was significantly higher (P < 0.001)
among nonresponders (15.9%) than responders (5.3%) in 2007,
suggesting that the risk of death was higher among frail men.
Frailty status in 2000 and 2007 in the weight change groups
is shown in Table 3. Frailty was especially prevalent in the
weight loss group, both in 2000 and 2007. The constantly
overweight group also tended to be frailer than the constantly
normal weight and weight gain groups. Global differences
between the groups were clearly significant in 2000 (P < 0.001)
and were close to reaching significance in 2007 (P = 0.06).
Similarly, mobility-related disability (determined for 610 men
in 2007) was most prevalent in the weight loss and constantly
overweight groups both in 2000 (P < 0.001) and 2007 (P =
0.005).
Independent predictors of prefrailty or frailty and mobilityrelated disability in 2007 were tested by using multinominal
logistic regression (Table 4). These analyses were adjusted for
age, smoking, comorbidity index, and mobility-related disability in 2000. Compared with the constantly normal weight group,
the odds ratios for both frailty and mobility-related disability
were significantly higher in the weight loss group, whereas in
the constantly overweight group, only the risk for developing
mobility-related disability was significantly higher. The weight
gain group did not differ significantly from the normal weight
group.
DISCUSSION
This long-term prospective study of older men indicates that
a good overall health prognosis is associated with maintaining normal weight over the life course. It also demonstrates a
dualistic association between overweight and risk of death.
Midlife overweight is associated with a higher mortality rate,
whereas in late life, the associations become more complex.
Those who lose weight after being overweight in midlife not
only have a higher risk of death, but also have a higher risk
of developing frailty and incident mobility-related disability
in late life. During the long follow-up period, men who survived to older age despite being constantly overweight also
had a 30% higher risk of death and almost double the risk of
mobility-related disability compared with men who maintained
a normal weight. Accordingly, we did not observe any obesity
paradox in our cohort and consequently challenge this concept as an excuse for not advocating normal weight and sensible weight reduction.
Strengths and limitations
The strengths of our study include the long follow-up, several points of BMI measurement, and the socioeconomically
homogenous cohort. The homogeneity reduces confounding
but also limits the generalizability of the results to other socioeconomic groups and to women. No men were underweight
(BMI <18.5), nor did any have chronic diseases or regularly
take medications for chronic conditions at baseline in 1974,
which reduces confounding and the potential for reverse causality. On the other hand, subjects were not “superhealthy,”
because many had risk factors such as hypertension, hypercholesterolemia, smoking, or glucose intolerance.
The response rates to the questionnaire surveys were good,
and nonrespondents were not different than respondents in
terms of earlier body weight or weight gain (11). The reliability of the mortality data is supported by the use of national registers. The body weight, morbidity, and frailty and disability
status in 2000 and 2007 were based on questionnaire data, but
these have been used repeatedly in epidemiologic studies and
are considered appropriate methods (19, 20). In our cohort of
men with high social status, the correlation between reported
and measured weight was 0.93 (unpublished observation).
It can be argued that the groups representing weight trajectories in the present study are a crude measure for studying
the obesity paradox, because they may not gather all information. For example, a participant’s BMI value may decrease substantially between midlife and late life, and yet the participant
may be defined as constantly overweight. However, the BMI
change groups also differed according to average BMI values
at various time points (Figure 2). It is of note that our primary
goal was to study the impact of overweight as currently defined
(BMI cutpoint of 25) on prognosis.
An important limitation may be that the nature of weight
loss (intentional or unintentional) is not known. Accordingly,
we do not know whether intentional weight loss is associated with poorer prognosis in late life. Among older people,
Am J Epidemiol. 2013;178(9):1452–1460
Obesity Paradox in Older Men 1457
weight loss is considered to be mainly due to an endogenous process, associated with physiological and/or pathological changes
(21). This is supported by the finding that midlife cardiovascular risk, overweight, and less physical activity predict frailty
in late life (22–24). Nevertheless, one might also argue the opposite, that is, recognized cardiovascular risk in midlife provokes intentional weight reduction, which predisposes some
individuals to sarcopenia, frailty, and poorer prognosis (25–27)
unless care is taken to include enough protein in the diet and to
engage in muscle strengthening activities. However, we believe
that endogenous weight loss should be distinguished from “exogenous” weight loss. This is supported by the good prognosis of
normal-weight subjects in our cohort and by studies suggesting that well-controlled weight reduction is not detrimental and
may have health benefits among older adults (28, 29).
We did not study factors related to body composition or fitness (e.g., waist circumference, waist-to-hip ratio, percent body
fat or muscle, visceral fat, muscle strength, cardiorespiratory
fitness) (4, 5, 30–32), but various combinations of fitness, leanness, and fatness in weight change groups would only tend to
dilute prognostic differences between the groups. The Helsinki
Businessmen Study provides detailed data of midlife physical
activity in a subgroup, but we did not report this here. In a separate report, we showed that, in this subgroup, lower physical
activity predicted the development of frailty in late life (24).
Finally, our cohort was relatively small, and the findings must
be verified in other data sets.
12-Year Survival, %
100
60
Dashed = Weight gain
Solid = Constantly normal weight
Dotted = Constantly overweight
Dashed-dotted = Weight Loss
20
0
2
4
6
8
10
12
Survival Time, years
Figure 3. Kaplan-Meier survival curves for different weight trajectories in the Helsinki Businessmen Study, 2000–2012. Mortality followup was from 2000 through July 31, 2012. The constantly normal
weight group (n = 340) had body mass index (BMI) (weight (kg)/
height (m)2) values of less than 25 at ages 47 and 73 years. The constantly overweight group (n = 495) had BMI values of 25 or greater at
ages 47 and 73 years. The weight gain group (n = 136) had BMI
values less than 25 at age 47 years and 25 or greater at age 73
years. The weight loss group (n = 143) had BMI values of 25 or
greater at age 47 years and less than 25 at age 73 years. Log-rank,
P < 0.001 between groups.
Problems with the obesity paradox
During recent years, the obesity paradox has been a popular
topic in the research of chronic diseases, such as cardiovascular
Table 2. Multivariate-Adjusted Hazard Ratios of Total Mortality During Follow-up of the Helsinki Businessmen
Study by Weight Status Group, 2000–2012
Weight Status Group
Model
Constantly
Overweightb
(n = 495)
Constantly
Normal Weighta
(n = 340)
Weight Gainc
(n = 136)
Weight Lossd (n = 143)
HRe
HR
95% CI
HR
95% CI
HR
95% CI
Referent
1.4
1.08, 1.7
1.0
0.7, 1.5
1.8
1.4, 2.4
Bg
1.0
1.3
1.07, 1.7
0.96
0.7, 1.4
1.8
1.4, 2.4
h
C
1.0
1.3
1.06, 1.7
0.96
0.7, 1.4
1.8
1.4, 2.4
Di
1.0
1.3
1.03, 1.7
1.0
0.7, 1.5
1.8
1.3, 2.3
Ej
1.0
1.3
1.05, 1.7
1.0
0.7, 1.4
1.4
1.0, 1.9
A
f
Abbreviations: BMI, body mass index; CI, confidence interval; HR, hazard ratio.
a
Constantly normal weight represents BMI (weight (kg)/height (m)2) values of less than 25 in 1974 and 2000.
b
Constantly overweight represents BMI values of 25 or greater in 1974 and 2000.
c
Weight gain represents BMI values of less than 25 in 1974 and 25 or greater in 2000.
d
Weight loss represents BMI values of 25 or greater in 1974 and less than 25 in 2000.
e
Hazard ratios were calculated by using Cox regression analysis.
f
Model A is adjusted for age.
g
Model B is adjusted for age and smoking in 1974.
h
Model C is adjusted for model B variables plus self-rated health in 1974.
i
Model D is adjusted for model C variables plus reported diseases (memory disturbances, cerebrovascular disorders, coronary heart disease, congestive heart failure, pulmonary disease, musculoskeletal disease, and cancer)
in 2000.
j
Model E is adjusted for age, smoking in 1974, and frailty status in 2000.
Am J Epidemiol. 2013;178(9):1452–1460
1458 Strandberg et al.
Table 3. Frailty and Mobility-Related Disability Status in the Helsinki Businessmen Study by Weight Status Group
in 2000 and 2007
Weight Status Group
Constantly
Overweightb
Constantly
Normal Weighta
Disability Status
No.
%
No.
Weight Gainc
%
No.
Weight Lossd
%
No.
%
Total subjects
In 2000
340
495
136
143
In 2007
196
284
82
60
Frailty statuse
Nonfrail
In 2000
145
45.3
201
41.9
65
51.2
8
6.2
In 2007
94
48.0
118
41.5
36
43.9
20
33.3
Prefrail
In 2000
156
48.4
239
49.8
57
44.9
86
66.7
In 2007
89
45.4
137
48.2
42
51.2
29
48.3
In 2000
20
6.2
40
8.3
5
3.9
35
27.1
In 2007
13
6.6
29
10.2
4
4.9
11
18.3
16
12.4
38
28.8
18
21.7
20
35.7
Frail
P value in 2000
<0.001
P value in 2007
0.065
f
Mobility-related disability
In 2000
30
9.4
113
23.3
P value
<0.001
In 2007
32
17.2
84
29.5
P value
0.005
Abbreviation: BMI, body mass index.
a
Constantly normal weight represents BMI (weight (kg)/height (m)2) values of less than 25 in 1974 and 2000.
b
Constantly overweight represents BMI values of 25 or greater in 1974 and 2000.
c
Weight gain represents BMI values of less than 25 in 1974 and 25 or greater in 2000.
d
Weight loss represents BMI values of 25 or greater in 1974 and less than 25 in 2000.
e
Defined according to modified Fried criteria (14, 22).
f
Any difficulty walking 500 m.
Table 4. Multivariate-Adjusted Odds Ratios of Frailty and Mobility-Related Disability During Follow-up of the
Helsinki Businessmen Study by Weight Status Group, 2000–2007
Weight Status Group
Disability Status
Constantly
Normal Weighta
(n = 340)
ORe
f
Prefrail
Constantly
Overweightb
(n = 495)
OR
95% CI
Weight Gainc
(n = 136)
OR
95% CI
Weight Lossd
(n = 143)
OR
95% CI
Referent
1.2
0.8, 1.8
1.2
0.7, 2.1
1.5
0.7, 2.9
Frailf
1.0
1.6
0.7, 3.5
0.6
0.2, 2.5
3.7
1.3, 10.5
Mobility-related
disabilityg
1.0
1.9
1.1, 3.2
1.1
0.5, 2.3
2.4
1.1, 4.9
Abbreviations: BMI, body mass index; CI, confidence interval; OR, odds ratio.
a
Constantly normal weight represents BMI (weight (kg)/height (m)2) values of less than 25 in 1974 and 2000.
b
Constantly overweight represents BMI values of 25 or greater in 1974 and 2000.
c
Weight gain represents BMI values of less than 25 in 1974 and 25 or greater in 2000.
d
Weight loss represents BMI values of 25 or greater in 1974 and less than 25 in 2000.
e
Odds ratios were calculated by using multinominal linear regression adjusted for age, smoking in 1974, selfrated health in 1974, comorbidity index in 2000, and mobility-related disability in 2000.
f
Defined according to modified Fried criteria (14, 22).
g
Any difficulty walking 500 m.
Am J Epidemiol. 2013;178(9):1452–1460
Obesity Paradox in Older Men 1459
disease and diabetes (1–8). The suggestion that overweight
or obesity could have beneficial effects may even have raised
questions about the need for weight control. In addition, several explanations related to both differences in care and pathophysiological aspects (10) have been presented. However, the
apparent paradox may be due to relatively short follow-up times
and, especially, the inability of most studies to account for weight
trajectories during the life course. Also, unintentional weight loss
brought about by the development of intervening factors—
such as frailty or smoking—can be misleading. Weight loss
seems to be associated with poorer cardiovascular prognosis
even before late life (33), and this may contribute to the obesity paradox in cardiovascular diseases. One important consideration is that higher BMI values may also be due to higher
lean body mass. This notion is supported by findings that “fat
and fit” is preferable to “lean and unfit,” at least for cardiovascular risk (30–32).
A recent review questioned the existence of any obesity paradox (34) and called for a life course approach to studying this
phenomenon. Our results concur with this and suggest that
the development of frailty may be an intervening mechanism
between weight loss after midlife and higher mortality risk in
later life (Table 2). In addition, those who are constantly overweight have poorer long-term prognoses (and more disability, 35, 36) in late life, and this supports conventional views
of overweight as a risk factor.
Conclusions
These 12-year mortality follow-up results do not support
the existence of an obesity paradox in late life when life-course
trajectories of BMI are taken into account. Men with constantly
normal weight over the life course had a good prognosis in
late life. Men who were either constantly overweight or who
changed from overweight in midlife to normal weight in late
life had poorer prognosis and more frailty and disability in late
life. Even those who developed overweight after midlife did
not fare better than those of constantly normal weight. Our
findings support the view that a healthy lifestyle, including
weight control, is worthwhile to maintain throughout life.
ACKNOWLEDGMENTS
Author affiliations: Geriatric Clinic, Department of Medicine,
University of Helsinki and Helsinki University Central Hospital, Helsinki, Finland (Timo E. Strandberg, Arto Y. Strandberg,
Reijo S. Tilvis); Institute of Health Sciences/Geriatrics, University of Oulu, Oulu, Finland (Timo E. Strandberg); THLNational Institute for Health and Welfare, Helsinki/Turku, Finland (Sari Stenholm, Veikko Salomaa); Department of Public
Health, University of Turku, Turku, Finland (Sari Stenholm);
and Unit of General Practice, Helsinki University Central Hospital, Helsinki, Finland (Kaisu H. Pitkälä).
This work was supported by the Konung Gustaf V:s och
Drottning Victorias Frimurarestiftelse; the Jahnsson Foundation; the University Central Hospital of Oulu; the University
Central Hospital of Helsinki; the Uulo Arhio Foundation;
Am J Epidemiol. 2013;178(9):1452–1460
the Päivikki and Sakari Sohlberg Foundation; and the Academy of Finland (grant 273850 to S.S.). The funding sources
had no role in the design and conduct of the study, the
collection, analysis, and interpretation of the data, or in the
preparation, review, or approval of the manuscript.
We thank Andy Langdon for proofreading the manuscript.
Conflict of interest: none declared.
REFERENCES
1. Kalantar-Zadeh K, Horwich TB, Oreopoulos A, et al. Risk
factor paradox in wasting diseases. Curr Opin Clin Nutr Metab
Care. 2007;10(4):433–442.
2. Carnethon MR, De Chavez PJ, Biggs ML, et al. Association of
weight status with mortality in adults with incident diabetes.
JAMA. 2012;308(6):581–590.
3. Nilsson G, Hedberg P, Öhrvik J. Survival of the fattest:
unexpected findings about hyperglycemia and obesity in a
population based study of 75-year-olds. BMJ Open. 2011;
1:e000012.
4. Lavie CJ, De Schutter A, Patel D, et al. Body composition and
coronary heart disease mortality—an obesity or a lean
paradox? Mayo Clin Proc. 2011;86(9):857–864.
5. Lavie CJ, De Schutter A, Patel DA, et al. Body composition
and survival in stable coronary heart disease: impact of lean
mass index and body fat in the “obesity paradox”. J Am Coll
Cardiol. 2012;60(15):1374–1380.
6. McAuley PA, Artero EG, Sui X, et al. The obesity paradox,
cardiorespiratory fitness, and coronary heart disease. Mayo
Clin Proc. 2012;87(5):443–451.
7. Clark AL, Chyu J, Horwich TB. The obesity paradox in men
versus women with systolic heart failure. Am J Cardiol.
2012;110(1):77–82.
8. Azimi A, Charlot MG, Torp-Pedersen C, et al. Moderate
overweight is beneficial and severe obesity detrimental for
patients with documented atherosclerotic heart disease. Heart.
2013;99(9):655–660.
9. Flegal KM, Kit BK, Orpana H, et al. Association of all-cause
mortality with overweight and obesity using standard body
mass index categories: a systematic review and meta-analysis.
JAMA. 2013;309(1):71–82.
10. Heymsfield SB, Cefalu WT. Does body mass index adequately
convey a patient’s mortality risk? JAMA. 2013;309(1):87–88.
11. Strandberg TE, Strandberg AY, Salomaa VV, et al. Explaining
the obesity paradox: cardiovascular risk, weight change, and
mortality during long-term follow-up in men. Eur Heart J.
2009;30(14):1720–1727.
12. Fried LP, Tangen CM, Walston J, et al. Frailty in older adults:
evidence for a phenotype. J Gerontol A Biol Sci Med Sci.
2001;56(3):M146–M156.
13. Strandberg TE, Pitkälä KH, Tilvis RS. Frailty in older people.
Eur Geriatr Med. 2011;2(6):344–355.
14. Sirola J, Pitkala KH, Tilvis RS, et al. Definition of frailty in
older men according to questionnaire data (RAND-36/SF-36):
the Helsinki Businessmen Study. J Nutr Health Aging.
2011;15(9):783–787.
15. Marmot MG, Shipley MJ. Do socioeconomic differences in
mortality persist after retirement? 25 Year follow up of civil
servants from the first Whitehall II Study. BMJ. 1996;
313(7066):1177–1180.
16. Aalto A-M, Aro AR, Teperi J. RAND-36 as a measure of
health-related quality of life. Reliability, construct validity,
1460 Strandberg et al.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
and reference values in the Finnish general population.
Helsinki, Finland: Stakes, Research Reports 101; 1999.
Charlson ME, Pompei P, Ales KL, et al. A new method of
classifying prognostic comorbidity in longitudinal studies:
development and validation. J Chron Dis. 1987;40(5):373–383.
Flegal KM, Carroll MD, Kit BK, et al. Prevalence of obesity
and trends in the distribution of body mass index among US
adults, 1999–2010. JAMA. 2012;307(5):491–497.
Bergmann MM, Jacobs EJ, Hoffmann K, et al. Agreement of
self-reported medical history: comparison of an in-person
interview with a self-administered questionnaire. Eur
J Epidemiol. 2004;19(5):411–416.
Tamakoshi K, Yatsuya H, Kondo T, et al. The accuracy of
long-term recall of past body weight in Japanese adult men. Int
J Obes Relat Metab Disord. 2003;27(2):247–252.
Kotz C, Billington C, Levine A. Obesity and aging. Clin
Geriatr Med. 1999;15(2):391–412.
Strandberg TE, Sirola J, Pitkälä KH, et al. Association of
midlife obesity and cardiovascular risk with old age frailty: a
26-year follow-up of initially healthy men. Int J Obes (Lond).
2012;36(9):1153–1157.
Stenholm S, Strandberg TE, Pitkälä K, et al. Midlife obesity
and risk of frailty in old age during a 22-year follow-up in men
and women: the Mini-Finland Follow-up Survey [ published
online ahead of print May 2, 2013]. J Gerontol A Biol Sci Med
Sci. (doi:10.1093/gerona/glt052).
Savela SL, Koistinen P, Stenholm S, et al. Leisure-time
physical activity in midlife is related to old age frailty
[ published online ahead of print March 22, 2013]. J Gerontol
A Biol Sci Med Sci. (doi:10.1093/gerona/glt029).
Snih S, Raji MA, Markides KS, et al. Weight change and
lower body disability in older Mexican Americans. J Am
Geriatr Soc. 2005;53(10):1730–1737.
Rejeski WJ, Marsh AP, Chmelo E, et al. Obesity, intentional
weight loss and physical disability in older adults. Obes Rev.
2010;11(9):671–685.
27. Arnold AM, Newman AB, Cushman M, et al. Body weight
dynamics and their association with physical function and
mortality in older adults: the Cardiovascular Health Study.
J Gerontol A Biol Sci Med Sci. 2010;65(1):63–70.
28. Villareal DT, Chode S, Parimi N, et al. Weight loss, exercise,
or both and physical function in obese older adults. N Engl J
Med. 2011;364(13):1218–1229.
29. Shea MK, Nicklas BJ, Houston DK, et al. The effect of
intentional weight loss on all-cause mortality in older adults:
results of a randomized controlled weight-loss trial. Am J Clin
Nutr. 2011;94(3):839–846.
30. McAuley P, Myers J, Abella J, et al. Body mass, fitness and
survival in veteran patients: Another obesity paradox? Am
J Med. 2007;120(6):518–524.
31. Fogelholm M. Physical activity, fitness and fatness: relations to
mortality, morbidity and disease risk factors. Obes Rev.
2010;11(3):202–221.
32. McAuley PA, Smith NS, Emerson BT, et al. The obesity
paradox and cardiorespiratory fitness [ published online ahead
of print February 20, 2012]. J Obes. (doi:10.1155/2012/
951582).
33. Bowman TS, Kurth T, Sesso HD, et al. Eight-year change in
body mass index and subsequent risk of cardiovascular disease
among healthy non-smoking men. Prev Med. 2007;45(6):
436–441.
34. Ferreira I, Stehouwer CDA. Obesity paradox or inappropriate
study designs? Time for life-course epidemiology.
J Hypertens. 2012;30(12):2271–2275.
35. Stenholm S, Rantanen T, Alanen E, et al. Obesity history as a
predictor of walking limitation at old age. Obesity. 2007;
15(4):929–938.
36. Stenholm S, Sallinen J, Koster A, et al. Association between
obesity history and hand grip strength in older adults—
exploring the roles of inflammation and insulin resistance as
mediating factors. J Geront A Biol Sci Med Sci. 2011;
66A(3):341–348.
Am J Epidemiol. 2013;178(9):1452–1460