confidential: for review only - bmj · confidential: for review only 1 1 the combined associations...
TRANSCRIPT
Confidential: For Review O
nly
The combined associations of body weight and lifestyle
factors with all-cause and cause-specific mortality in men and women
Journal: BMJ
Manuscript ID BMJ.2016.034265
Article Type: Research
BMJ Journal: BMJ
Date Submitted by the Author: 30-Jun-2016
Complete List of Authors: Veronese, Nicola; University of Padova, Li, Yanping; HSPH Manson, JoAnn; Brigham and Women's Hospital, Dept. of Epidemiology Willett, Walter; Harvard School of Public Health, Nutrition Fontana, Luigi; Washington University in Saint Louis Hu, Frank; Harvard School of Public Health
Keywords: body weight, mortality, lifestyle
https://mc.manuscriptcentral.com/bmj
BMJ
Confidential: For Review O
nly
1
The combined associations of body weight and lifestyle factors with all-cause and cause-1
specific mortality in men and women 2
3
Running title: BMI, lifestyle factors and mortality 4
5
Nicola Veronese1,2,*
, Yanping Li3,*
, JoAnn E. Manson4,5,6
, Walter C. Willett3,4,5
, Luigi Fontana1,7,8
, 6
Frank B. Hu3,4,5
7
1Division of Geriatrics and Nutritional Science, Washington University, St. Louis, MO, USA; 8
2Division of Geriatrics, Department of Medicine, University of Padova, Italy; 9
3Department of Nutrition, Harvard T.H. Chan School of Public Health, Boston, MA, USA; 10
4Department of Epidemiology, Harvard T.H. Chan School of Public Health, Boston, MA, USA; 11
5Channing Division of Network Medicine, Department of Medicine, Brigham and Women’s 12
Hospital, Harvard Medical School, Boston, MA, USA 13
6Division of Preventive Medicine, Department of Medicine, Brigham and Women’s Hospital, 14
Boston, Massachusetts 15
7Department of Clinical and Experimental Sciences, Brescia University Medical School, Brescia, 16
Italy; 17
8CEINGE Biotecnologie Avanzate, Napoli, Italy; 18
*Equally contributed to the paper. 19
20
Page 1 of 40
https://mc.manuscriptcentral.com/bmj
BMJ
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Confidential: For Review O
nly
2
Correspondence to: 21
Frank B. Hu, Department of Nutrition, Harvard T.H. Chan School of Public Health, 655 Huntington 22
Avenue, Boston, MA 02115, USA Channing Division of Network Medicine, Brigham and 23
Women's Hospital and Harvard Medical School, Boston Department of Epidemiology, Harvard T.H. 24
Chan School of Public Health, Boston; Phone: 617.432.0113. Email: 25
27
Luigi Fontana, Division of Geriatrics and Nutritional Sciences and Center for Human Nutrition, 28
Washington University School of Medicine, St. Louis, MO 63110, USA; Department of Clinical 29
and Experimental Sciences, Brescia University, 25121 Brescia, Italy; CEINGE Biotecnologie 30
Avanzate, 80122 Napoli, Italy. Email: [email protected]. 31
32
33
Page 2 of 40
https://mc.manuscriptcentral.com/bmj
BMJ
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Confidential: For Review O
nly
3
ABSTRACT 34
BACKGROUND: Some epidemiological data suggest that overweight and obese individuals have 35
lower mortality than normal-weight individuals. We aimed to evaluate the combined associations of 36
diet, physical activity, moderate alcohol consumption and smoking with body weight on risk of all-37
cause and cause-specific mortality. 38
METHOD AND FINDINGS: 74,582 women from the Nurses' Health Study (1980-2012) and 39
39,284 men from the Health Professionals Follow-up Study (1986-2012) free of cardiovascular 40
disease and cancer at baseline were followed in alongitudinal study with up to 32 years of follow-up. 41
Exposures included body mass index (BMI), Alternate Healthy Eating Index (AHEI) score, physical 42
activity level (PAL), smoking habits and alcohol drinking while outcome was mortality (all-cause, 43
cardiovascular, cancer). During up to 32 years of follow-up, 30,013 death cases were documented 44
(10,808 of cancer-mortality and 7,189 of CVD-mortality). In each of the 4 BMI categories studied 45
(18.5-22.4, 22.5-24.9, 25-29.9, ≥30 kg/m2), people with 1 or more healthy lifestyle factors had a 46
significantly lower risk of total, cardiovascular and cancer mortality than individuals with no low-47
risk lifestyle factors. A combination of at least 3 low-risk lifestyle factors and BMI between 18.5-48
22.4 kg/m2 was associated with the lowest risk of all-cause (HR: 0.39; 95%CI: 0.35-0.43), cancer 49
(HR: 0.40; 95%CI: 0.34-0.47) and cardiovascular (HR: 0.37; 95%CI: 0.29-0.46) mortality, 50
compared to those with BMI between 22.5-24.9 and none of the 4 low-risk lifestyle factors. 51
CONCLUSION: People in the 18.5-22.4 BMI range with high AHEI, PAL, moderate alcohol 52
drinking and non-smoking had substantially lower mortality rate. It is important to consider diet and 53
lifestyle factors when evaluating the relationship between BMI and mortality. 54
55
Page 3 of 40
https://mc.manuscriptcentral.com/bmj
BMJ
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Confidential: For Review O
nly
4
INTRODUCTION 56
Data from experimental, epidemiological and clinical studies indicate that leanness is associated 57
with greater metabolic health and a lower risk of developing type 2 diabetes, cardiovascular disease 58
(CVD) and cancer.(1–3) However, data from some epidemiological studies are challenging this 59
concept and suggest that being overweight, and possibly even mildly obese, is associated with a 60
reduced mortality risk.(4,5) Normal weight, defined as a body mass index (BMI) between 18.5-61
24.9 kg/m2, is a broad reference group containing not only non-smoking individuals who are lean 62
because they are physically active and eating healthy diets, but also those who have fallen in the 63
normal BMI range for other reasons which accelerate the aging process (e.g. chronic inflammation, 64
chronic obstructive pulmonary disease, and neurodegenerative diseases), probably leading to an 65
artificially reduced mortality among the overweight and moderately obese groups.(6–10) Therefore, 66
when evaluating the relationship between BMI and mortality, it might be important to consider 67
lifestyle factors such as smoking, diet, moderate alcohol intake and physical activity levels. 68
In this large prospective cohort study, we examined the joint associations of BMI and lifestyle 69
factors with all-cause, cardiovascular and cancer mortality among 74,582 women from the Nurses' 70
Health Study (NHS 1980-2012) and 39,284 men from the Health Professionals Follow-up Study 71
(1986-2012) who were free of cardiovascular disease (CVD) and cancer at baseline. We 72
hypothesized that nonsmoking individuals who were lean, and also exercised regularly and 73
followed a healthy diet with a moderate alcohol intake had the lowest mortality risk. 74
75
METHODS 76
Study Population 77
The Health Professions’ Follow-up Study (HPFS) was established in 1986, when 51,529 male US 78
health professionals (dentists, optometrists, osteopaths, podiatrists, pharmacists, and veterinarians) 79
aged 40–75 years completed a mailed questionnaire about their medical history and lifestyle.(11) 80
Page 4 of 40
https://mc.manuscriptcentral.com/bmj
BMJ
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Confidential: For Review O
nly
5
The Nurses’ Health Study (NHS) began in 1976, when 121,701 female nurses 30 to 55 years of age 81
responded to a questionnaire regarding medical, lifestyle, and other health-related information.(11)
82
Both cohorts were designed with the similar questions on diet, exercise, smoking status and other 83
factors (the use of hormone replacement therapies and reproductive related questions were asked in 84
the NHS only). In both cohorts, questionnaires have been sent every two years to update this 85
information and identify newly diagnosed cases of various diseases. As shown in S1 Table, we 86
excluded subjects with previously diagnosed cancer and cardiovascular diseases at baseline. We 87
also excluded participants with baseline missing value of BMI, PAL, food frequency questionnaire 88
(FFQ) or smoking. After these exclusions, we included 74,582 women and 39,284 men. Excluded 89
participants and those included in this study were similar in terms of baseline age, BMI and 90
Alternate Healthy Eating Index (AHEI) (S2 Table). 91
The study was approved by the institutional review board of Brigham and Women’s Hospital in 92
Boston; completion of the self-administered questionnaire was considered to imply informed 93
consent. 94
95
Patient involvement 96
No patients were involved in setting the research question or the outcome measures, nor were they 97
involved in the design and implementation of the study. There are no plans to involve patients in 98
dissemination. 99
100
Assessment of exposure 101
Diet was assessed using a validated FFQ asking how often, on average, a participant had consumed 102
a particular amount of a specific type of food during the previous year.(12) The intake of nutrients
103
was computed by multiplying the frequency of consumption of each unit of food by its nutrient 104
content and accounting also for the consumption of alcoholic drinks. Diet quality was assessed 105
using the AHEI score that is strongly associated with the onset of metabolic and cardiovascular 106
Page 5 of 40
https://mc.manuscriptcentral.com/bmj
BMJ
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Confidential: For Review O
nly
6
diseases in the general population.(13,14) Briefly, points were assigned for intake of each 107
component on a scale from 0 to 10, with 10 indicating adherence to the recommended levels of 108
servings per day and 0 the worst intake. Multivitamin use was scored as 2.5 points for 0 to 4 years 109
of use and 7.5 points for 5 or more years of use. We included 10 components of the index in our 110
diet score: higher intakes of vegetables, fruit, nuts, whole grains, polyunsaturated fatty acids, and 111
long chain omega 3 fatty acids and lower intakes of red and processed meats, sugar sweetened 112
beverages, trans fats, and sodium. In this analysis, AHEI was divided in quintiles, classifying the 113
two upper quintiles as healthier diets.(15–17) The information about diet was updated every 4 years. 114
Physical activity levels were investigated using a validated questionnaire and updated every 2 115
years.(18) We estimated the amount of hours per week spent in moderate-to-vigorous activities 116
(including brisk walking) requiring at least the expenditure of 3 MET or more per hour. The validity 117
and reproducibility of the questionnaire were evaluated and reported previously. Briefly, the 118
correlation between physical activity as reported in one week recalls and that reported on the 119
questionnaires was 0.79. The correlation between moderate to vigorous activity reported in diaries 120
and that reported on the questionnaires was 0.62.(19) 121
Smoking habits were categorized as never smoked, former smoker, or current smoker (including the 122
number of cigarettes smoked per day) and updated every two years. 123
Body mass index was calculated as self-reported weight (kg)/height (m)2. In our validation study, 124
the correlation between self-reported and technician-measured weight was 0.97.(20) 125
126
Definition of Low-risk Lifestyle Group 127
Low-risk lifestyle factors included never-smoking, physical activity levels ≥30 min/day at moderate 128
to vigorous intensity, moderate alcohol consumption defined as 5 to 15 g alcohol/d in women and 129
5–30 g alcohol/d in men (21), and the AHEI in upper two quintiles. Each factor was coded as 0 or 1, 130
Page 6 of 40
https://mc.manuscriptcentral.com/bmj
BMJ
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Confidential: For Review O
nly
7
and the sum of these four scores together gave a final score of 0, 1, 2, 3 or 4 with higher scores 131
indicating a healthier lifestyle. 132
133
Assessment of Other Confounders 134
Information on potential confounders, such as age, ethnicity, use of multi-vitamin, regular use of 135
aspirin, postmenopausal hormone use (NHS only), family history of diabetes, cancer, or myocardial 136
infarction (in first-degree relatives) was collected via biennial questionnaires.
137
138
Ascertainment of deaths 139
Deaths were identified from state vital statistics records, the National Death Index, reported by the 140
families, and the postal system.(24) The follow-up for death in both cohorts was at least 98% 141
complete. Cause of death was identified from death certificates or review of medical records. In the 142
current analysis, we evaluated all-cause mortality, and also death from CVD (International 143
Classification of Diseases, eighth revision (ICD-8) codes 390 through 458) or cancer (ICD-8 codes 144
140 through 207). 145
146
Statistical Analysis 147
Participants contributed person time from the return of the baseline questionnaire (NHS, 1980; 148
HPFS, 1986) until the date of death, loss to follow-up, or the end of the follow-up period (30 June 149
2012 for NHS and 30 January 2012 for HPFS), whichever came first. 150
Proportional hazards assumption was evaluated with a likelihood-ratio test comparing the model 151
with and without an interaction term between time period and the joint categories of BMI and 152
lifestyles. The P value for the proportional hazards assumption was 0.44, indicating that the 153
proportional-hazards assumption was not violated in our analyses. We therefore used Cox 154
proportional hazard models to calculate the adjusted hazard ratios (HRs) of all-cause, cancer and 155
Page 7 of 40
https://mc.manuscriptcentral.com/bmj
BMJ
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Confidential: For Review O
nly
8
cardiovascular mortality with their 95% confidence intervals (CIs) across BMI categories (kg/m2: 156
<18.5, 18.5-22.4, 22.5-24.9, 25.0-27.4, 27.5-29.9, 30.0-34.9, 35.0-39.9, and ≥40). (22) We defined 157
a BMI of 22.5-24.9 as the referent category, which has been used in previous studies.(23). In 158
addition, this was the range of BMI associated with the lowest mortality in our cohorts (S1 Figure). 159
We updated the information during follow-up by using the most recently available information of 160
BMI and lifestyle factors. To minimize potential reverse causality bias, we stopped updating body 161
weight once the participants were diagnosed with chronic diseases such as cancer, cardiovascular 162
diseases, or diabetes. 163
The covariates included in the multivariable model were: age, ethnicity, current multivitamin use, 164
current aspirin use, status with regard to a family history of diabetes mellitus, myocardial infarction, 165
or cancer; and, for women, menopausal status and hormone use. If data were missing at a given 166
time point, the last observation was carried forward for 1 cycle. 167
We calculated the HRs across joint classification of number of low-risk lifestyle factors (0, 1, 2, or 168
≥3) and BMI (kg/m2: 18.5-22.4, 22.5-24.9, 25.29.9, and ≥30) because of too few cases for cause-169
specific mortality in some groups as shown in S3 Table. A similar analysis was performed 170
considering only the effect of AHEI, moderate alcohol consumption, and physical activity levels for 171
each BMI category after excluding ever-smokers. 172
Several sensitivity analyses were run to test the robustness of our results. First, we kept all missing 173
values at baseline and during follow-up, and imputed missing values of BMI and lifestyle factors by 174
applying a multiple imputation approach. Second, we calculated an expanded low-risk lifestyle 175
score by assigning weights to each low-risk factor based on the beta-coefficients from the 176
multivariable-adjusted Cox model with all-cause mortality as the outcome. We then summed up the 177
products of each binary lifestyle score multiplied by its weight, divided it by the sum of all beta 178
coefficient values, and then multiplied by 4, to make the low-risk lifestyle score easier to interpret. 179
Page 8 of 40
https://mc.manuscriptcentral.com/bmj
BMJ
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Confidential: For Review O
nly
9
In this way, the expanded low-risk lifestyle score ranged from 0 to 4, and each unit of low-risk 180
lifestyle score represented the change in one risk factor. Third, we conducted analyses that assessed 181
the effects of specific combination of lifestyle factors on mortality. Finally, we conducted additional 182
analyses by combining the two normal weight groups. 183
We used SAS version 9.3 to analyze the data. Statistical significance was set at a two tailed P<0.05. 184
185
RESULTS 186
The characteristics of the study population according to baseline BMI are shown in Table 1 for 187
women and men. As reported in S4 Table, those with ≥ 3 low risk factors were 8.9% in women and 188
19.6% in men. 189
During up to 32 years of follow-up, 30,013 death cases were documented (10,808 of cancer-190
mortality and 7,189 of CVD-mortality). As shown in Table 2, compared to those with BMI 191
between 22.5 and 24.9 kg/m2, most other BMI categories had an increased risk of all-cause 192
mortality after adjusting for potential confounders. Among participants without any low risk 193
lifestyle factor, the association between BMI and mortality was U-shaped (P for non-linear trend < 194
0.0001), while among participants with one or more low risk lifestyle factors, the association 195
between BMI and mortality was in a more linear shape (Ps for linear trend < 0.0001) (P for 196
interaction < 0.0001, Table 3). 197
The mortality risk significantly decreased with the increasing numbers of low-risk lifestyle factors. 198
A combination of at least 3 low-risk lifestyle factors and BMI between 18.5-22.4 kg/m2 was 199
associated with the lowest risk of total (HR: 0.39; 95%CI: 0.35-0.43), cancer (HR: 0.40; 95%CI: 200
0.34-0.47) and cardiovascular (HR: 0.37; 95%CI: 0.29-0.46) mortality, compared to those with 201
BMI between 22.5-24.9 and none of the 4 low-risk lifestyle factors (Figure 1). These findings were 202
confirmed in never-smokers (HR=0.45; 95%CI: 0.33-0.61 for all-cause mortality; HR=0.47; 95%CI: 203
Page 9 of 40
https://mc.manuscriptcentral.com/bmj
BMJ
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Confidential: For Review O
nly
10
0.29-0.77 for cancer mortality; HR=0.41; 95%CI: 0.21-0.80 for CVD mortality) (Figure 2). 204
Participants with all four low-risk factors (high AHEI scores, regular physical activity, never 205
smoking, with a moderate alcohol consumption) combined with a BMI between 18.5 and 24.9 206
Kg/m2, as compared with all others in the study population, had a reduced risk for all-cause 207
(HR=0.39; 95%CI: 0.32-0.47), cancer (HR=0.40; 95%CI: 0.29-0.54) or CVD (HR=0.41; 95%CI: 208
0.28-0.59) mortality (Table 4). 209
The results of sensitivity analyses using multiple imputations (S2 Figure), applying expanded low-210
risk lifestyle scores (S3 Figure), specific combinations of lifestyle factors in the low-risk category 211
(S4 Figure) or combining the two normal weight categories (S5 Figure) were similar to those from 212
the main analysis. 213
214
DISCUSSION 215
In these large prospective cohorts, we found after taking into account the effect of high PAL and 216
AHEI scores, moderate alcohol drinking and never-smoking, men and women in the 18.5-217
22.4 BMI range experienced substantially lower CVD, cancer and total mortality risk. 218
Our findings are in agreement with a recent meta-analysis, including about 30 millions participants, 219
showing that overweight and obesity are associated with increased risk of all-cause mortality, with 220
the nadir of the curve observed at BMI 23-24 among never smokers, 22-23 among healthy never 221
smokers, and 20-22 in studies with a follow-up over 20 years. (25) Of note, in a recent study 222
conducted among three Danish cohorts, the BMI associated with the lowest all-cause mortality 223
increased by 3.3 unit from cohorts enrolled from 1976-1978 through 2003-2013. (26) However, the 224
average follow-up duration for the 2003-2013 cohort was only 4.6 years, and thus the shift to higher 225
BMI associated with the lowest mortality in this cohort is likely to be due to reserve causation bias 226
(i.e. individuals with pre-existing or preclinical diseases have increased mortality but lower BMI 227
Page 10 of 40
https://mc.manuscriptcentral.com/bmj
BMJ
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Confidential: For Review O
nly
11
due to weight loss). It should be noted that these studies did not take into account diet and lifestyle 228
factors. 229
Many epidemiological studies have shown a U-shape relationship between BMI and mortality, with 230
some studies showing the lowest mortality around 22.5-25, and others around 25-29.9.(4,27) Our 231
data are confirming that there is a U-shape relationship between BMI and mortality, with an 232
increased mortality risk in the <22.4 and ≥30 BMI range, as recently reported in a large meta-233
analysis including 189 studies without smokers and almost 4 millions participants.(28) However, 234
our data showed for the first time that among never-smoking men and women eating healthier diets, 235
with moderate alcohol intake and exercising regularly ≥30 min/d at moderate intensity, there was a 236
linear relation between BMI and mortality risk with the nadir in the 18.5-22.4 BMI category. The 237
problem is that only a small fraction of the North American and European populations is both 238
normal weight and practicing a healthy lifestyle (i.e., not smoking, exercising regularly, moderate 239
alcohol consumption, and eating a healthy diet).(29) In this study of US health professionals, only 240
19.6% of men and 8.9% of women, respectively, were lean and had ≥3 low-risk lifestyle factors. 241
Our findings suggest that in a substantial proportion of the US population, leanness is driven by 242
other factors that may increase mortality risk, including existing or preclinical chronic diseases that 243
may cause weight loss and cigarette smoking, which is also associated with lower body weight. The 244
reverse causation problem (i.e. low BMI is a consequence of chronic diseases) may be particularly 245
pronounced among those who were sedentary and consuming unhealthy diets, conditions which are 246
important determinants of chronic disease morbidity and mortality. This may explain why an 247
increased mortality was observed in the low BMI range of 18.5-22.4 only among those who did not 248
follow a healthy lifestyle, but not among those with a relatively healthy lifestyle. Increasing basic 249
and translational research suggests that smoking, a sedentary lifestyle and unhealthy hyper-caloric 250
diets cause an overstimulation of multiple mitogenic, inflammatory and oxidative pathways. (30–32) 251
These factors may accelerate the accumulation of unrepaired cellular damage and senescent cells, 252
Page 11 of 40
https://mc.manuscriptcentral.com/bmj
BMJ
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Confidential: For Review O
nly
12
stimulate the DNA damage response, inhibit stem cells regenerative function, and promote 253
involuntary weight loss, organ dysfunction and frailty.(7,10,33–35) 254
BMI is a crude, but practical and inexpensive tool to measure adiposity in large population 255
studies.(22,36) However, it does not measure the lean/fat ratio or the distribution of fat and, most 256
importantly, the relative contribution of lifestyle and other metabolic factors on the risk of 257
developing a chronic disease and dying prematurely. In this study, we showed that, in each of the 258
four BMI categories, men and women with healthy lifestyles were powerfully protected against 259
CVD, cancer and total mortality. Interestingly, even within the obese category, individuals who do 260
not smoke, eat healthier diets, drink a moderate quantity of alcohol and exercise regularly, have a 261
much lower risk of mortality than those who have unhealthy lifestyles. The mechanisms responsible 262
for the protective effects of healthy lifestyles in these individuals are not completely known, but 263
likely encompass a series of metabolic and molecular alterations induced by exercise training, 264
healthier diets rich in vitamins and phytochemicals, and avoidance of smoking. These conditions 265
probably inhibit insulin resistance, inflammation, and oxidative stress, and slow the accumulation of 266
cellular and organ damage.(3,32–34). These results suggest that the health benefit by engaging in 267
even one healthy lifestyle behavior was seen in all BMI categories, especially among those in the 268
unhealthy BMI categories. 269
Our study had several strengths. The analysis is based on a large sample size and a long follow-up 270
with a large number of incident cases, which provide high statistical power for the analysis. Another 271
important strength is the high follow-up rate. In each two or four year cycle of the survey in our 272
cohorts, follow-up rates have averaged 94%. Other strengths include the repeated measurements of 273
diet, physical activity, smoking and other factors. Our study addresses not only the relation between 274
BMI, major lifestyle factors and all-cause mortality, but also cause-specific mortality. 275
The current study has some potential limitations. Our cohorts included only health professionals, 276
mostly white men and women, which limits the generalizability of the findings. However, the 277
Page 12 of 40
https://mc.manuscriptcentral.com/bmj
BMJ
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Confidential: For Review O
nly
13
relative homogeneity of the study population in educational attainment and socioeconomic status 278
enhances the internal validity. Also, measurement errors in self-reported lifestyle factors are 279
inevitable. However, because of the prospective design, these measurement errors are typically 280
unrelated to the outcomes of the study and thus are likely to have attenuated the observed 281
associations. For this analysis, we did not have necessary validation data to correct for measurement 282
errors with repeated assessments of dietary patterns and other lifestyles over time as our current 283
validation data and methodology allow us to correct for measurement errors in baseline 284
measurements only. However, we used time-varying repeated assessments of lifestyles, which 285
would have substantially reduced within-person variations in the assessment of long-term dietary 286
and lifestyle patterns. 287
In conclusion, people in the 18.5-22.4 BMI range with high PAL, AHEI, moderate alcohol 288
consumption and non-smoking had substantially lower mortality rate. We also found that one or 289
more low-risk lifestyle behaviors significantly reduced mortality risk also in overweight and obese 290
individuals. Our data suggest that the U-shaped relationship between BMI and mortality observed in 291
many epidemiological studies is driven by an overrepresentation in our societies of individuals who 292
are lean because of chronic metabolic and pathological conditions caused by exposure to smoke, a 293
sedentary lifestyle and/or unhealthy diets. Finally, our findings indicate that leanness induced by 294
healthy lifestyles is the optimal way to promote healthy longevity and to reduce the risk of death. 295
296
297
Page 13 of 40
https://mc.manuscriptcentral.com/bmj
BMJ
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Confidential: For Review O
nly
14
What is already known on this topic 298
Leanness is associated with greater metabolic health and a lower risk of developing type 2 diabetes, 299
cardiovascular disease, cancer and all-cause mortality. 300
However, data from some epidemiological studies suggest that being overweight, and possibly even 301
mildly obese, is associated with a reduced mortality risk. 302
What this study adds 303
One or more low-risk lifestyle behaviors significantly reduced mortality risk also in overweight and 304
obese individuals. 305
The U-shaped relationship between BMI and mortality observed in many epidemiological studies is 306
driven by an overrepresentation in our societies of individuals who are lean because of chronic 307
metabolic and pathological conditions caused by exposure to smoke, a sedentary lifestyle and/or 308
unhealthy diets. 309
310
Page 14 of 40
https://mc.manuscriptcentral.com/bmj
BMJ
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Confidential: For Review O
nly
15
Acknowledgement: We would like to thank the participants and staff of the Nurses’ Health Study 311
and the Health Professionals Study that contributed data for their valuable contributions as well as 312
the following state cancer registries for their help: AL, AZ, AR, CA, CO, CT, DE, FL, GA, ID, IL, 313
IN, IA, KY, LA, ME, MD, MA, MI, NE, NH, NJ, NY, NC, ND, OH, OK, OR, PA, RI, SC, TN, 314
TX, VA, WA, WY. 315
Funding: The cohorts were supported by grants of UM1 CA186107, P01 CA87969, R01 316
HL088521, UM1 CA167552, and R01 HL35464 from the National Institutes of Health. L.F. and 317
N.V. were supported by grants from the Bakewell Foundation and the Longer Life Foundation (an 318
RGA/Washington University Partnership). The funders play no-role in the design of the study; the 319
collection, analysis, and interpretation of the data; and the decision to approve publication of the 320
finished manuscript. The authors assume full responsibility for analyses and interpretation of these 321
data. 322
Contributors: NV, YL, LF, and FBH were involved in the study conception and design. WCW, 323
FBH, and JEM obtained funding. WCW, FBH, and JEM provided study materials or patients and 324
collected and collated data. All authors were involved in analysis and interpretation of the data. NV 325
drafted the article; NV, YL, LF, JEM, WCWCand FBH revised it critically for important 326
intellectual content; all authors approved the final version. LF and FBH is the guarantors. 327
Competing interests: All authors have completed the ICMJE uniform disclosure form at 328
www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and declare: 329
no support from any organization for the submitted work other than those detailed above; no 330
financial relationships with any organizations that might have an interest in the submitted work in 331
the previous three years; no other relationships or activities that could appear to have influenced the 332
submitted work. 333
Page 15 of 40
https://mc.manuscriptcentral.com/bmj
BMJ
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Confidential: For Review O
nly
16
Ethical approval: The study protocol was approved by the institutional review boards of the 334
Brigham and Women’s Hospital and the Harvard School of Public Health. Completion of the self 335
administered questionnaire was considered to imply informed consent. 336
Transparency statement: The lead authors (the manuscript’s guarantors) affirm that the 337
manuscript is an honest, accurate, and transparent account of the study being reported; that no 338
important aspects of the study have been omitted; and that any discrepancies from the study as 339
planned (and, if relevant, registered) have been explained. 340
Data sharing: No additional data available. 341
342
Page 16 of 40
https://mc.manuscriptcentral.com/bmj
BMJ
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Confidential: For Review O
nly
17
REFERENCES 343
1. Fontana L, Hu FB. Optimal body weight for health and longevity: bridging basic, clinical, 344
and population research. Aging Cell [Internet]. 2014 Jun [cited 2016 Jan 30];13(3):391–400. 345
Available from: 346
http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=4032609&tool=pmcentrez&rend347
ertype=abstract 348
2. Clinical Guidelines on the Identification, Evaluation, and Treatment of Overweight and 349
Obesity in Adults--The Evidence Report. National Institutes of Health. Obes Res [Internet]. 350
1998 Sep [cited 2015 May 16];6 Suppl 2:51S – 209S. Available from: 351
http://www.ncbi.nlm.nih.gov/pubmed/9813653 352
3. Rizza W, Veronese N, Fontana L. What are the roles of calorie restriction and diet quality in 353
promoting healthy longevity? Ageing Res Rev [Internet]. 2014 Jan [cited 2015 Feb 354
13];13(1):38–45. Available from: http://www.ncbi.nlm.nih.gov/pubmed/24291541 355
4. Flegal KM, Kit BK, Orpana H, Graubard BI. Association of all-cause mortality with 356
overweight and obesity using standard body mass index categories: a systematic review and 357
meta-analysis. JAMA [Internet]. 2013 Jan 2 [cited 2014 Dec 17];309(1):71–82. Available 358
from: http://www.ncbi.nlm.nih.gov/pubmed/23280227 359
5. Veronese N, Cereda E, Solmi M, Fowler SA, Manzato E, Maggi S, et al. Inverse relationship 360
between body mass index and mortality in older nursing home residents: a meta-analysis of 361
19,538 elderly subjects. Obes Rev [Internet]. 2015 Aug 7 [cited 2015 Aug 8]; Available from: 362
http://www.ncbi.nlm.nih.gov/pubmed/26252230 363
6. Manson JE, Stampfer MJ, Hennekens CH, Willett WC. Body weight and longevity. A 364
reassessment. JAMA [Internet]. 1987 Jan 16 [cited 2015 Dec 30];257(3):353–8. Available 365
Page 17 of 40
https://mc.manuscriptcentral.com/bmj
BMJ
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Confidential: For Review O
nly
18
from: http://www.ncbi.nlm.nih.gov/pubmed/3795418 366
7. López-Otín C, Blasco MA, Partridge L, Serrano M, Kroemer G. The hallmarks of aging. Cell 367
[Internet]. 2013 Jun 6 [cited 2014 Jul 10];153(6):1194–217. Available from: 368
http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3836174&tool=pmcentrez&rend369
ertype=abstract 370
8. Russell SJ, Kahn CR. Endocrine regulation of ageing. Nat Rev Mol Cell Biol [Internet]. 2007 371
Sep [cited 2016 Jan 30];8(9):681–91. Available from: 372
http://www.ncbi.nlm.nih.gov/pubmed/17684529 373
9. Schumacher B, Garinis GA, Hoeijmakers JHJ. Age to survive: DNA damage and aging. 374
Trends Genet [Internet]. 2008 Mar [cited 2016 Jan 30];24(2):77–85. Available from: 375
http://www.ncbi.nlm.nih.gov/pubmed/18192065 376
10. Fontana L, Partridge L, Longo VD. Extending healthy life span--from yeast to humans. 377
Science [Internet]. 2010 Apr 16 [cited 2015 Jun 11];328(5976):321–6. Available from: 378
http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3607354&tool=pmcentrez&rend379
ertype=abstract 380
11. Colditz GA, Rimm EB, Giovannucci E, Stampfer MJ, Rosner B, Willett WC. A prospective 381
study of parental history of myocardial infarction and coronary artery disease in men. Am J 382
Cardiol [Internet]. 1991 May 1 [cited 2016 Jan 30];67(11):933–8. Available from: 383
http://www.ncbi.nlm.nih.gov/pubmed/2018010 384
12. Willett WC, Sampson L, Stampfer MJ, Rosner B, Bain C, Witschi J, et al. Reproducibility 385
and validity of a semiquantitative food frequency questionnaire. Am J Epidemiol [Internet]. 386
1985 Jul [cited 2016 Jan 30];122(1):51–65. Available from: 387
http://www.ncbi.nlm.nih.gov/pubmed/4014201 388
Page 18 of 40
https://mc.manuscriptcentral.com/bmj
BMJ
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Confidential: For Review O
nly
19
13. Tobias DK, Hu FB, Chavarro J, Rosner B, Mozaffarian D, Zhang C. Healthful dietary 389
patterns and type 2 diabetes mellitus risk among women with a history of gestational diabetes 390
mellitus. Arch Intern Med [Internet]. 2012 Nov 12 [cited 2016 Jan 30];172(20):1566–72. 391
Available from: 392
http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3774782&tool=pmcentrez&rend393
ertype=abstract 394
14. McCullough ML, Willett WC. Evaluating adherence to recommended diets in adults: the 395
Alternate Healthy Eating Index. Public Health Nutr [Internet]. 2006 Mar [cited 2016 Jan 396
30];9(1A):152–7. Available from: http://www.ncbi.nlm.nih.gov/pubmed/16512963 397
15. Chiuve SE, Rexrode KM, Spiegelman D, Logroscino G, Manson JE, Rimm EB. Primary 398
prevention of stroke by healthy lifestyle. Circulation [Internet]. 2008 Aug 26 [cited 2016 Mar 399
4];118(9):947–54. Available from: 400
http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2730914&tool=pmcentrez&rend401
ertype=abstract 402
16. Hu FB, Manson JE, Stampfer MJ, Colditz G, Liu S, Solomon CG, et al. Diet, lifestyle, and 403
the risk of type 2 diabetes mellitus in women. N Engl J Med [Internet]. 2001 Sep 13 [cited 404
2016 Apr 24];345(11):790–7. Available from: 405
http://www.ncbi.nlm.nih.gov/pubmed/11556298 406
17. Reis JP, Loria CM, Sorlie PD, Park Y, Hollenbeck A, Schatzkin A. Lifestyle factors and risk 407
for new-onset diabetes: a population-based cohort study. Ann Intern Med [Internet]. 2011 408
Sep 6 [cited 2016 May 3];155(5):292–9. Available from: 409
http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3491359&tool=pmcentrez&rend410
ertype=abstract 411
18. Hu FB, Sigal RJ, Rich-Edwards JW, Colditz GA, Solomon CG, Willett WC, et al. Walking 412
Page 19 of 40
https://mc.manuscriptcentral.com/bmj
BMJ
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Confidential: For Review O
nly
20
compared with vigorous physical activity and risk of type 2 diabetes in women: a prospective 413
study. JAMA [Internet]. 1999 Oct 20 [cited 2016 Jan 30];282(15):1433–9. Available from: 414
http://www.ncbi.nlm.nih.gov/pubmed/10535433 415
19. Wolf AM, Hunter DJ, Colditz GA, Manson JE, Stampfer MJ, Corsano KA, et al. 416
Reproducibility and validity of a self-administered physical activity questionnaire. Int J 417
Epidemiol [Internet]. 1994 Oct [cited 2016 Jan 30];23(5):991–9. Available from: 418
http://www.ncbi.nlm.nih.gov/pubmed/7860180 419
20. Rimm EB, Stampfer MJ, Colditz GA, Chute CG, Litin LB, Willett WC. Validity of self-420
reported waist and hip circumferences in men and women. Epidemiology [Internet]. 1990 421
Nov [cited 2016 Jan 30];1(6):466–73. Available from: 422
http://www.ncbi.nlm.nih.gov/pubmed/2090285 423
21. Solomon CG, Hu FB, Stampfer MJ, Colditz G a, Speizer FE, Rimm EB, et al. Moderate 424
alcohol consumption and risk of coronary heart disease among women with type 2 diabetes 425
mellitus. Circulation. 2000;102(5):494–9. 426
22. Obesity: preventing and managing the global epidemic. Report of a WHO consultation. 427
World Health Organ Tech Rep Ser [Internet]. 2000 Jan [cited 2016 Jan 10];894:i – xii, 1–253. 428
Available from: http://www.ncbi.nlm.nih.gov/pubmed/11234459 429
23. Berrington de Gonzalez A, Hartge P, Cerhan JR, Flint AJ, Hannan L, MacInnis RJ, et al. 430
Body-mass index and mortality among 1.46 million white adults. N Engl J Med [Internet]. 431
2010 Dec 2 [cited 2015 Dec 15];363(23):2211–9. Available from: 432
http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3066051&tool=pmcentrez&rend433
ertype=abstract 434
24. Rich-Edwards JW, Corsano KA, Stampfer MJ. Test of the National Death Index and Equifax 435
Page 20 of 40
https://mc.manuscriptcentral.com/bmj
BMJ
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Confidential: For Review O
nly
21
Nationwide Death Search. Am J Epidemiol [Internet]. 1994 Dec 1 [cited 2016 Jan 436
30];140(11):1016–9. Available from: http://www.ncbi.nlm.nih.gov/pubmed/7985649 437
25. Aune D, Sen A, Prasad M, Norat T, Janszky I, Tonstad S, et al. BMI and all cause mortality: 438
systematic review and non-linear dose-response meta-analysis of 230 cohort studies with 439
3.74 million deaths among 30.3 million participants. BMJ [Internet]. 2016 May 4 [cited 2016 440
May 6];353(may04_6):i2156. Available from: 441
http://www.bmj.com/content/353/bmj.i2156?utm_content=bufferd8358&utm_medium=socia442
l&utm_source=twitter.com&utm_campaign=buffer 443
26. Afzal S, Tybjærg-Hansen A, Jensen GB, Nordestgaard BG. Change in Body Mass Index 444
Associated With Lowest Mortality in Denmark, 1976-2013. JAMA [Internet]. American 445
Medical Association; 2016 May 10 [cited 2016 May 11];315(18):1989. Available from: 446
http://jama.jamanetwork.com/article.aspx?articleID=2520627 447
27. Tobias DK, Hu FB. Does being overweight really reduce mortality? Obesity (Silver Spring) 448
[Internet]. 2013 Sep [cited 2016 Jan 30];21(9):1746–9. Available from: 449
http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3806201&tool=pmcentrez&rend450
ertype=abstract 451
28. Angelantonio E Di, Bhupathiraju SN, Wormser D, Kaptoge S, Gonzalez AB De, Cairns BJ, 452
et al. Body-mass index and all-1 cause mortality: Individual-participant-data meta-analysis of 453
239 prospective studies in four continents. Lancet. 2016;in press. 454
29. Stampfer MJ, Hu FB, Manson JE, Rimm EB, Willett WC. Primary prevention of coronary 455
heart disease in women through diet and lifestyle. N Engl J Med [Internet]. 2000 Jul 6 [cited 456
2016 Jan 26];343(1):16–22. Available from: http://www.ncbi.nlm.nih.gov/pubmed/10882764 457
30. Chiolero A, Faeh D, Paccaud F, Cornuz J. Consequences of smoking for body weight, body 458
Page 21 of 40
https://mc.manuscriptcentral.com/bmj
BMJ
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Confidential: For Review O
nly
22
fat distribution, and insulin resistance. Am J Clin Nutr [Internet]. 2008 Apr 1 [cited 2016 Jan 459
30];87(4):801–9. Available from: http://www.ncbi.nlm.nih.gov/pubmed/18400700 460
31. Messner B, Bernhard D. Smoking and cardiovascular disease: mechanisms of endothelial 461
dysfunction and early atherogenesis. Arterioscler Thromb Vasc Biol [Internet]. 2014 Mar 462
[cited 2015 Nov 23];34(3):509–15. Available from: 463
http://www.ncbi.nlm.nih.gov/pubmed/24554606 464
32. Rom O, Kaisari S, Aizenbud D, Reznick AZ. Sarcopenia and smoking: a possible cellular 465
model of cigarette smoke effects on muscle protein breakdown. Ann N Y Acad Sci [Internet]. 466
2012 Jul [cited 2016 Jan 30];1259:47–53. Available from: 467
http://www.ncbi.nlm.nih.gov/pubmed/22758636 468
33. Booth FW, Roberts CK, Laye MJ. Lack of exercise is a major cause of chronic diseases. 469
Compr Physiol [Internet]. 2012 Apr [cited 2015 Jun 15];2(2):1143–211. Available from: 470
http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=4241367&tool=pmcentrez&rend471
ertype=abstract 472
34. Campisi J. Aging, cellular senescence, and cancer. Annu Rev Physiol [Internet]. 2013 Jan 473
[cited 2014 Jul 11];75:685–705. Available from: 474
http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=4166529&tool=pmcentrez&rend475
ertype=abstract 476
35. Conboy IM, Rando TA. Heterochronic parabiosis for the study of the effects of aging on 477
stem cells and their niches. Cell Cycle [Internet]. 2012 Jun 15 [cited 2016 Jan 478
30];11(12):2260–7. Available from: 479
http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3383588&tool=pmcentrez&rend480
ertype=abstract 481
Page 22 of 40
https://mc.manuscriptcentral.com/bmj
BMJ
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Confidential: For Review O
nly
23
36. Wannamethee SG. Obesity Epidemiology. Frank B Hu. Int J Epidemiol [Internet]. 2008 Oct 482
29 [cited 2016 Jan 30];38(1):325–6. Available from: 483
http://ije.oxfordjournals.org/content/38/1/325.full 484
485
486
Page 23 of 40
https://mc.manuscriptcentral.com/bmj
BMJ
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Confidential: For Review O
nly
24
Figure legends 487
Figure 1. Hazard ratios* for all-cause and cause-specific mortality according to joint classification 488
of a low-risk lifestyle score**
and body mass index based on pooled results from NHS and HPFS 489
cohorts 490
*Adjusted for age; race; current multivitamin use, and current aspirin use; status with regard to a family history of 491
diabetes mellitus, myocardial infarction, or cancer; and, for women, menopausal status and hormone use. 492
**Low risk lifestyles include never-smoking, exercise ≥30 min/d at moderate or vigorous intensity, the Alternate 493
Healthy Eating Index score in the upper two fifth (quintiles), and moderate alcohol consumption (moderate: 5 to 15 g 494
alcohol/d in women, 5–30 g alcohol/d in men). Each factor was coded as 0 or 1 and sum four scores together, so the 495
healthy lifestyles scores were 0, 1, 2, 3 or 4 (healthiest). 496
497
Figure 2. Hazard ratios* for all-cause and cause-specific mortality according to joint classification 498
of a low-risk lifestyle score**
and body mass index among never smoking men and women 499
*Adjusted for age; race; current multivitamin use, and current aspirin use; status with regard to a family history of 500
diabetes mellitus, myocardial infarction, or cancer; and, for women, menopausal status and hormone use. 501
**Low risk lifestyles include exercise ≥30 min/d at moderate or vigorous intensity, the Alternate Healthy Eating Index 502
score in the upper two fifth (quintiles) and moderate alcohol consumption (moderate: 5 to 15 g alcohol/d in women, 5–503
30 g alcohol/d in men). Each factor was coded as 0 or 1 and sum three scores together, so the healthy lifestyles scores 504
were 0, 1, 2, or 3 (healthiest). 505
506
Page 24 of 40
https://mc.manuscriptcentral.com/bmj
BMJ
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Confidential: For Review O
nly
25
Supporting information captions 507
S1 Table. Sample sizes before and after baseline exclusions in the NHS and HPFS cohorts 508
S2Table Comparison of baseline characteristics between included and excluded study participants 509
(Mean (SD) or %) 510
S3 Table: The number of deaths during the follow-up period according to joint classifications of 511
body mass index and healthy lifestyle scores 512
S4 Table: Distribution proportion of low-risk lifestyle factors (%) according to the BMI categories 513
S1 Figure. Hazard ratios for all-cause mortality according to BMI by each unit 514
S2 Figure. Hazard ratios for all-cause and cause-specific mortality according to joint classification 515
of low-risk lifestyle score and body mass index based on multiple imputations for missing value at 516
baseline and during follow-up period 517
S3 Figure. Hazard ratios for all-cause and cause-specific mortality according to joint classification 518
of body mass index and low-risk lifestyle score using an expanded low-risk lifestyle score 519
S4 Figure. Hazard ratios for all-cause mortality and cause-specific mortality by specific 520
combinations of lifestyle factors in the low-risk category 521
S5 Figure. Hazard ratios for all-cause and cause-specific mortality according to joint classifications 522
of low-risk lifestyle score and body mass index by combining the two normal weight groups 523
524
525
Page 25 of 40
https://mc.manuscriptcentral.com/bmj
BMJ
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Confidential: For Review O
nly
26
Table 1: Age-adjusted baseline Characteristics in the Nurses’ Health Study (NHS, 1980) and 526
Health Professions’ Follow-up Study (HPFS, 1986) 527
528
BMI (kg/m2)
<
18.5
18.5-
22.4
22.5-
24.9
25.0-
27.4
27.5-
29.9
30.0-
34.9
35.0-
39.9
>=
40
NHS, 1980
n 1492 29125 19105 11563 5362 5552 1724 659
Age, years* 44.1 44.6 46.4 47.3 47.4 47.1 46.3 46.7
BMI, kg/m2 17.7 20.9 23.7 26.2 28.8 32.0 37.0 43.6
Alternate Healthy Eating Index 30.6 31.3 31.2 30.9 30.6 30.2 29.8 29.5
Physical activity, hours/week 3.8 4.2 3.9 3.7 3.5 3.3 2.9 2.7
Current smoking, % 45.6 32.1 28.5 26.9 25.2 21.9 21.2 18.2
Alcohol consumption, gram/day 7.6 7.6 6.7 5.7 4.8 3.8 2.8 2.6
Caucasian, % 97.9 98.0 97.6 97.3 96.8 96.9 97.4 96.0
Post-menopausal, % 46.2 42.3 42.8 43.2 43.2 43.3 42.1 40.7
Aspirin use, % 57.9 55.0 53.1 51.5 51.3 49.3 48.1 44.4
Multivitamin use, % 61.4 63.5 66.1 67.8 69.1 70.5 73.4 71.8
Family history of diabetes, % 21.2 24.1 28.1 30.6 33.5 38.1 40.2 39.8
Family history of MI, % 22.8 22.9 24.7 25.5 27.0 27.2 27.1 27.9
Family history of cancer, % 14.2 13.8 13.8 13.5 13.2 13.6 13.0 14.2
HPFS, 1986
n 139 5639 13092 12585 4805 2564 339 121
Age, years* 55.3 52.2 52.6 53.2 53.3 53.0 51.6 52.5
BMI, kg/m2 17.4 21.5 23.8 26.1 28.6 31.7 36.7 47.5
Alternate Healthy Eating Index 45.1 47.9 47.5 46.2 45.5 45.1 43.4 47.1
Physical activity, hours/week 2.2 3.3 3.2 2.7 2.2 1.7 1.2 2.1
Current smoking, % 24.1 11.2 8.9 9.9 10.4 9.8 8.9 9.7
Alcohol consumption, gram/day 12.5 10.6 11.6 11.6 11.8 10.4 9.6 10.9
Caucasian, % 89.3 92.8 94.2 94.5 94.5 94.6 96.0 93.8
Aspirin use, % 76.7 75.3 73.5 71.4 70.5 70.2 71.3 72.2
Multivitamin use, % 34.4 33.6 36.7 39.8 42.1 43.8 45.5 37.2
Family history of diabetes, % 12.6 18.2 19.6 21.7 22.7 26.6 28.9 23.4
Family history of MI, % 31.1 31.9 31.3 32.2 32.5 33.6 33.3 43.5
Family history of cancer, % 31.3 34.4 34.8 34.3 33.8 34.2 28.9 23.6
BMI: Body Mass Index; MI: Myocardial Infarction 529
530
531
Page 26 of 40
https://mc.manuscriptcentral.com/bmj
BMJ
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Confidential: For Review Only27
Table 2: Hazard ratios of all-cause mortality according to body mass index in the Nurses’ Health Study (NHS, 1980-2012) and Health
Professions’ Follow-up Study (HPFS, 1986-2012)
BMI (kg/m2)
<18.5 18.5-22.4 22.5-
24.9 25.0-27.4 27.5-29.9 30.0-34.9 35.0-39.9 >=40
NHS, 1980-2012
Person years 32722 653018 572034 416990 213264 217770 67570 25594
# deaths 528 4789 4246 3450 1909 2103 816 387 Crude incidence (105PY) 1614 733 742 827 895 966 1208 1512
Age-adjusted HR (95%CI) 2.22 (2.03-2.43) 1.13 (1.08-1.17) 1.0 1.06 (1.02-1.11) 1.14 (1.08-1.21) 1.32 (1.25-1.39) 1.86 (1.73-2.01) 2.52 (2.27-2.79)
Multivariate-adjusted 1* 2.11 (1.93-2.31) 1.13 (1.08-1.17) 1.0 1.04 (0.99-1.09) 1.09 (1.04-1.16) 1.23 (1.17-1.30) 1.71 (1.58-1.84) 2.21 (1.99-2.46)
Multivariate-adjusted 2** 1.92 (1.75-2.10) 1.13 (1.08-1.18) 1.0 1.01 (0.97-1.06) 1.04 (0.99-1.10) 1.15 (1.09-1.21) 1.54 (1.43-1.66) 1.94 (1.74-2.15)
HPFS, 1986-2012
Person years 2781 113480 282946 290160 125288 67988 9756 2843
# deaths 87 1594 3457 3689 1700 1021 172 65
Crude incidence (105PY) 3128 1405 1222 1271 1357 1502 1763 2287
Age-adjusted 1.69 (1.36-2.09) 1.07 (1.01-1.13) 1.0 1.06 (1.01-1.11) 1.26 (1.18-1.33) 1.52 (1.42-1.63) 2.12 (1.82-2.47) 2.77 (2.17-3.54) Multivariate-adjusted 1* 1.56 (1.26-1.93) 1.06 (1.00-1.13) 1.0 1.03 (0.99-1.08) 1.22 (1.15-1.30) 1.45 (1.35-1.55) 2.00 (1.72-2.34) 2.10 (1.64-2.69)
Multivariate-adjusted 2** 1.50 (1.21-1.86) 1.08 (1.02-1.15) 1.0 0.98 (0.94-1.03) 1.10 (1.04-1.16) 1.22 (1.14-1.31) 1.61 (1.38-1.87) 1.71 (1.34-2.20)
Pooled results Age-adjusted 2.13 (1.96-2.32) 1.12 (1.08-1.15) 1.0 1.06 (1.02-1.09) 1.19 (1.15-1.24) 1.38 (1.32-1.44) 1.91 (1.79-2.04) 2.56 (2.33-2.82)
Multivariate-adjusted 1* 2.03 (1.87-2.20) 1.11 (1.08-1.15) 1.0 1.03 (1.00-1.07) 1.15 (1.10-1.20) 1.30 (1.24-1.35) 1.76 (1.65-1.88) 2.21 (2.01-2.43)
Multivariate-adjusted 2**
1.85 (1.70-2.01) 1.12 (1.08-1.16) 1.0 0.99 (0.96-1.03) 1.06 (1.02-1.11) 1.17 (1.12-1.22) 1.55 (1.45-1.66) 1.92 (1.74-2.11)
Pooled results among never smokers
Age-adjusted 1.63 (1.37-1.94) 0.97 (0.91-1.03) 1.0 1.09 (1.03-1.15) 1.26 (1.17-1.35) 1.54 (1.43-1.65) 2.27 (2.04-2.52) 2.96 (2.54-3.45)
Multivariate-adjusted 1* 1.54 (1.29-1.83) 0.97 (0.91-1.03) 1.0 1.06 (1.00-1.12) 1.19 (1.11-1.28) 1.44 (1.34-1.55) 2.08 (1.87-2.31) 2.65 (2.27-3.09)
Multivariate-adjusted 2** 1.53 (1.29-1.82) 0.99 (0.93-1.06) 1.0 1.01 (0.95-1.07) 1.09 (1.02-1.17) 1.27 (1.19-1.37) 1.78 (1.60-1.98) 2.26 (1.94-2.64)
HR: hazard ratio; CI: confident interval;
* Adjusted for age; race; current multivitamin use, and current aspirin use; family history of diabetes mellitus, myocardial infarction, or cancer; and,
for women, menopausal status and hormone use. **
Further adjusted for smoking status (never smoker, former smoker, current smoker: 1-14, 15-24 or ≥ 25 cigarettes/d), exercise (hours/week: 0, 0.01-
1.0, 1.0-3.5, 3.5-6.0, ≥6), alternate healthy eating index (quintile) and alcohol drinking (g/d: 0, 0.1-4.9, 5.0-14.9, 15.0-19.9, 20.0-29.9, and ≥30).
Page 27 of 40
https://mc.manuscriptcentral.com/bmj
BMJ
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Confidential: For Review Only28
Table 3: Association between body mass index and all-cause mortality stratified by the number of low-risk lifestyle factors*,§
Numbers of BMI (kg/m2) P for
Low-risk factors <18.5 18.5-22.4 22.5-24.9
25.0-27.4 27.5-29.9 30.0-34.9 35.0-39.9 >=40 Trendǂǂǂǂ
None # deaths 266 2163 2257 2169 1133 1058 342 161
HR (95% CI)**
2.37 (2.08-2.69) 1.24 (1.17-1.32) 1.0 1.00 (0.94-1.06) 1.03 (0.96-1.10) 1.11 (1.03-1.19) 1.50 (1.34-1.68) 1.80 (1.53-2.11) 0.81
One
# deaths 237 2485 3091 3028 1604 1423 459 207
HR (95% CI)** 1.89 (1.65-2.16) 1.11 (1.06-1.17) 1.0 1.00 (0.95-1.05) 1.11 (1.05-1.18) 1.22 (1.15-1.30) 1.59 (1.44-1.75) 2.13 (1.84-2.45) <0.0001
Two
# deaths 92 1307 1746 1535 682 556 164 71
HR (95% CI)** 1.57 (1.27-1.94) 1.05 (0.97-1.13) 1.0 1.03 (0.96-1.10) 1.13 (1.03-1.23) 1.35 (1.22-1.48) 2.09 (1.78-2.46) 2.40 (1.89-3.05) <0.0001
Three or Four
# deaths 20 428 609 407 190 87 23 13
HR (95% CI)**
1.21 (0.77-1.91) 0.98 (0.86-1.11) 1.0 0.96 (0.84-1.09) 1.26 (1.07-1.48) 1.40 (1.11-1.76) 2.32 (1.52-3.54) 3.50 (2.00-6.13) <0.0001
P for interaction ƫ <0.0001
HR: hazard ratio; CI: confident interval;
* Low-risk lifestyle factors include never-smoking, exercise ≥30 min/d at moderate or vigorous intensity, the Alternate Healthy Eating Index score
in the upper two fifth (quintiles), and moderate alcohol consumption (moderate: 5 to 15 g alcohol/d in women, 5–30 g alcohol/d in men). Each
factor was coded as 0 or 1 and sum three scores together, so the healthy lifestyles scores were 0, 1, 2, 3 or 4 (healthiest). **
Adjusted for age; race; current multivitamin use, and current aspirin use; family history of diabetes mellitus, myocardial infarction, or cancer;
and, for women, menopausal status and hormone use. ǂ P for linear trend tested by treating BMI as a continuous variable.
§Significant non-linear relation of BMI with risk of all-cause mortality was only significant among people with none low-risk lifestyle factor (P for
non-linear trend <0.0001),which was tested using likelihood ratio method by comparing model with linear term only with model with both linear
and cubic spline terms stratified by numbers of low-risk lifestyle factors. ƫWe tested the multiplicative interaction between BMI (8 categories) and numbers of low-risk factors by comparing the −2 log likelihood of the
multivariate adjusted models with and without the cross product interaction term.
Page 28 of 40
https://mc.manuscriptcentral.com/bmj
BMJ
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Confidential: For Review Only29
Table 4 Multivariate-adjusted hazard ratios (95%CI) of all-cause and cause-specific mortality by specific combinations of lifestyle
factors in the low-risk category
Participants All-cause mortality Cancer mortality CVD mortality
Factors in low-risk category % #
deaths
HR (95% CI) #
deaths
HR (95% CI) #
deaths
HR (95% CI)
All other categories 38.7 9906 1.0 (reference) 3785 1.0 (reference) 2423 1.0 (reference)
Diet in top 2 quintiles 61.3 20107 0.85 (0.83-0.87) 7023 0.91 (0.87-0.94) 4766 0.89 (0.84-0.93)
All other categories 12.5 1899 1.0 (reference) 835 1.0 (reference) 431 1.0 (reference)
Plus regular physical activity 87.5 28114 0.62 (0.59-0.65) 9973 0.70 (0.65-0.75) 6758 0.61 (0.55-0.67)
All other categories 5.6 680 1.0 (reference) 283 1.0 (reference) 147 1.0 (reference)
Plus never-smoking 94.4 29333 0.47 (0.44-0.51) 10525 0.51 (0.45-0.57) 7042 0.44 (0.37-0.52)
All other categories 1.7 174 1.0 (reference) 64 1.0 (reference) 46 1.0 (reference)
Plus moderate alcohol consumption 98.3 29839 0.43 (0.37-0.50) 10744 0.41 (0.32-0.52) 7143 0.46 (0.35-0.62)
All other categories 1.1 105 1.0 (reference) 41 1.0 (reference) 28 1.0 (reference)
Plus body mass index 18.5-24.9 kg/m2 98.9 29908 0.39 (0.32-0.47) 10767 0.40 (0.29-0.54) 7161 0.41 (0.28-0.59)
HR: hazard ratio; CI: confident interval; *
Adjusted for age; race; current multivitamin use, and current aspirin use; family history of diabetes mellitus, myocardial infarction, or cancer;
and for women, menopausal status and hormone use, and each lifestyle risk factors not already included in the model.
Page 29 of 40
https://mc.manuscriptcentral.com/bmj
BMJ
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Confidential: For Review O
nly
Figure 1
59x79mm (300 x 300 DPI)
Page 30 of 40
https://mc.manuscriptcentral.com/bmj
BMJ
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Confidential: For Review O
nly
Figure 2
59x78mm (300 x 300 DPI)
Page 31 of 40
https://mc.manuscriptcentral.com/bmj
BMJ
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Confidential: For Review O
nly
1
Supporting information
S1 Table. Sample sizes before and after baseline exclusions in the NHS and HPFS cohorts
NHS HPFS
Year of baseline 1980 1976
Baseline sample 98,712 51,529
Baseline exclusion
Cancer 3972 2160
Cardiovascular diseases 2308 3996
Missing BMI 591 1032
Missing physical activity 12307 2237
Missing smoking 142 1664
Missing FFQ (AHEI and alcohol) 4810 1156
Sample size for study population 74,582 39,284
The Nurses’ Health Study (NHS) began in 1976, when 121,701 female nurses responded. In 1980,
98,712 nurses response the long questionnaire including questions with detail food items (FFQ)
and physical activities, which was treated as the baseline for NHS in the current study.
Page 32 of 40
https://mc.manuscriptcentral.com/bmj
BMJ
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Confidential: For Review O
nly
2
S2 Table. Comparison of baseline characteristics between included and excluded study
participants (Mean (SDDDD) or %)
Excluded Included
NHS, 1980
N 24,130 74,582
Age, years 48.4(7.0) 45.9(7.2)
BMI, kg/m2 24.7(4.7) 24.3(4.5)
Alternate Healthy Eating Index 31.7(6.4) 31.0(6.2)
Physical activity, hours/week 3.5(2.9) 3.9(2.9)
Alcohol consumption, gram/day 6.3(10.5) 6.4(10.6)
Current smoking, % 30.7 28.8
High blood pressure, % 21.1 15.5
Hypercholesterolemia, % 8.4 5.1
Diabetes, % 3.8 2.1
HPFS, 1986
N 12,245 39,284
Age, years 59.1(9.7) 52.9(9.5)
BMI, kg/m2 25.8(3.6) 25.5(3.3)
Alternate Healthy Eating Index 48.3(10.9) 46.7(10.8)
Physical activity, hours/week 2.8(4.7) 2.8(4.2)
Alcohol consumption, gram/day 11.2(15.8) 11.4(15.4)
Current smoking, % 25.5 9.8
High blood pressure, % 31.9 19.7
Hypercholesterolemia, % 20.4 10.5
Diabetes, % 5.9 2.4
Page 33 of 40
https://mc.manuscriptcentral.com/bmj
BMJ
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Confidential: For Review O
nly
3
S3 Table: The number of deaths during the follow-up period according to joint classifications of
body mass index and healthy lifestyle scores*
BMI (kg/m2)
Cases, HR (95% CI) <18.5 18.5-22.4 22.5-24.9 25.0-29.9 >=30
All-cause mortality
None healthy lifestyle 266 2163 2257 3302 1561
One healthy lifestyle 237 2485 3091 4632 2089
Two healthy lifestyles 92 1307 1746 2217 791
Three healthy lifestyles 19 385 547 536 116
Four healthy lifestyles 1 43 62 61 7
Cancer mortality
None healthy lifestyle 69 765 805 1133 479
One healthy lifestyle 70 913 1207 1634 672
Two healthy lifestyles 22 518 702 819 271
Three healthy lifestyles 5 156 239 227 38
Four healthy lifestyles 1 17 24 20 2
CVD mortality
None healthy lifestyle 36 424 483 863 460
One healthy lifestyle 45 466 678 1230 590
Two healthy lifestyles 20 248 412 593 241
Three healthy lifestyles 0 75 125 129 25
Four healthy lifestyles 0 9 19 15 3
Low risk lifestyles include never-smoking, exercise ≥30 min/d at moderate or vigorous intensity,
the Alternate Healthy Eating Index score in the upper two fifth (quintiles), and moderate alcohol
consumption (moderate: 5 to 15 g alcohol/d in women, 5–30 g alcohol/d in men). Each factor was
coded as 0 or 1 and sum three scores together, so the healthy lifestyles scores were 0, 1, 2, 3 or 4
(healthiest).
Page 34 of 40
https://mc.manuscriptcentral.com/bmj
BMJ
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Confidential: For Review O
nly
4
S4 Table: Distribution proportion of low-risk lifestyle factors (%) according to the BMI
categories
BMI Number of low-risk factors (PYs, %)
(kg/m2) None One Two Three Four
NHS
<18.5 26.0 38.2 25.6 9.0 1.1
18.5-22.4 20.3 38.7 29.2 10.5 1.3
22.5-24.9 21.2 40.7 28.2 8.9 1.1
25.0-27.4 22.6 43.3 26.3 7.2 0.7
27.5-29.9 24.1 44.9 25.0 5.6 0.5
30.0-34.9 25.5 46.8 23.2 4.3 0.2
35.0-39.9 27.0 49.3 20.6 3.0 0.1
>=40 30.5 47.4 19.3 2.7 0.1
Total 22.2 41.9 26.9 8.0 0.9
HPFS
<18.5 21.5 36.7 28.3 10.7 2.8
18.5-22.4 11.9 28.8 32.5 21.2 5.6
22.5-24.9 11.9 31.2 33.2 19.1 4.6
25.0-27.4 15.3 35.2 31.5 15.0 3.0
27.5-29.9 18.2 37.8 30.1 12.0 2.0
30.0-34.9 22.8 40.1 27.1 8.8 1.2
35.0-39.9 25.8 43.0 23.8 6.7 0.7
>=40 25.6 34.3 27.0 11.8 1.3
Total 14.9 33.9 31.6 16.1 3.5
Page 35 of 40
https://mc.manuscriptcentral.com/bmj
BMJ
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Confidential: For Review O
nly
5
S1 Figure. Hazard ratios for all-cause mortality according to BMI by each unit
Adjusted for age; race; current multivitamin use, and current aspirin use; family history of diabetes mellitus, myocardial
infarction, or cancer; and, for women, menopausal status and hormone use, smoking status (never smoker, former
smoker, current smoker: 1-14, 15-24 or ≥ 25 cigarettes/d), exercise (hours/week: 0, 0.01-1.0, 1.0-3.5, 3.5-6.0, ≥6),
alternate healthy eating index (quintile) and alcohol drinking (g/d: 0, 0.1-4.9, 5.0-14.9, 15.0-19.9, 20.0-29.9, and ≥30).
Page 36 of 40
https://mc.manuscriptcentral.com/bmj
BMJ
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Confidential: For Review O
nly
6
S2 Figure. Hazard ratios for all-cause and cause-specific mortality according to joint
classification of low-risk lifestyle score and body mass index based on multiple imputations
for missing value at baseline and during follow-up period
Low risk lifestyles include never-smoking, exercise ≥30 min/d at moderate or vigorous intensity, the Alternate Healthy
Eating Index score in the upper two fifth (quintiles), and moderate alcohol consumption (moderate: 5 to 15 g alcohol/d
in women, 5–30 g alcohol/d in men). Each factor was coded as 0 or 1 and sum three scores together, so the healthy
lifestyles scores were 0, 1, 2, 3 or 4 (healthiest).
Adjusted for age; race; current multivitamin use, and current aspirin use; family history of diabetes mellitus, myocardial
infarction, or cancer; and, for women, menopausal status and hormone use.
Page 37 of 40
https://mc.manuscriptcentral.com/bmj
BMJ
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Confidential: For Review O
nly
7
S3 Figure. Hazard ratios for all-cause and cause-specific mortality according to joint
classification of body mass index and low-risk lifestyle score using an expanded low-risk
lifestyle score
Low risk lifestyles include never-smoking, exercise ≥30 min/d at moderate or vigorous intensity, the Alternate Healthy
Eating Index score in the upper two fifth (quintiles), and moderate alcohol consumption (moderate: 5 to 15 g alcohol/d
in women, 5–30 g alcohol/d in men). Each factor was coded as 0 or 1 and sum three scores together, so the healthy
lifestyles scores were 0, 1, 2, 3 or 4 (healthiest).
To calculate the expanded low-riskscore, we assigned weights to each binary low-risk factor based on the beta-
coefficients from the multivariable-adjusted Cox model with all-cause mortality as the outcome. We then summed up
the products of each binary lifestyle score multiplied by its weight, divided it by the sum of all beta coefficient values,
and then multiplied by 4, to make the low-risk lifestyle score easier to interpret. In this way, the expanded low-risk
lifestyle score ranged from 0 to 4, and each unit of low-risk lifestyle score represented the change in one risk factor
Adjusted for age; race; current multivitamin use, and current aspirin use; family history of diabetes mellitus, myocardial
infarction, or cancer; and, for women, menopausal status and hormone use.
Page 38 of 40
https://mc.manuscriptcentral.com/bmj
BMJ
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Confidential: For Review O
nly
8
S4 Figure. Hazard ratios for all-cause mortality and cause-specific mortality by specific
combinations of lifestyle factors in the low-risk category
Not listed factors in each combination were 0, such as participants in “A: Diet in top 2 quintiles” group did not had any
other low risk factors, only “AHEI score in score in the upper two fifth”; “B: A + regular physical activity” group had
“AHEI score in the upper two fifth” and “exercise ≥30 min/d at moderate or vigorous intensity”, without any other
factors.
Not all potential combinations of lifestyle factors were presented.
Adjusted for age; race; current multivitamin use, and current aspirin use; family history of diabetes mellitus, myocardial
infarction, or cancer; and, for women, menopausal status and hormone use.
Page 39 of 40
https://mc.manuscriptcentral.com/bmj
BMJ
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Confidential: For Review O
nly
9
S5 Figure. Hazard ratios for all-cause and cause-specific mortality according to joint
classifications of low-risk lifestyle score and body mass index by combining the two normal
weight groups
Low risk lifestyles include never-smoking, exercise ≥30 min/d at moderate or vigorous intensity, the Alternate Healthy
Eating Index score in the upper two fifth (quintiles), and moderate alcohol consumption (moderate: 5 to 15 g alcohol/d
in women, 5–30 g alcohol/d in men). Each factor was coded as 0 or 1 and sum three scores together, so the healthy
lifestyles scores were 0, 1, 2, 3 or 4 (healthiest).
Adjusted for age; race; current multivitamin use, and current aspirin use; family history of diabetes mellitus, myocardial
infarction, or cancer; and, for women, menopausal status and hormone use.
Page 40 of 40
https://mc.manuscriptcentral.com/bmj
BMJ
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960