Issuu on Google+

International Journal of Obesity (2011) 35, 1193–1199 & 2011 Macmillan Publishers Limited All rights reserved 0307-0565/11

ORIGINAL ARTICLE Lifelong doubling of mortality in men entering adult life as obese E Zimmermann1,2, C Holst1 and TIA Sørensen1 1 Institute of Preventive Medicine, Copenhagen University Hospital, Copenhagen, Denmark and 2Institute of Biomedical Sciences, University of Copenhagen, Copenhagen, Denmark

Background: The association between obesity in adults and excess morbidity and mortality is well established, but the impact of being obese in early adulthood on health throughout adult life needs elucidation. We investigated the all-cause mortality until 80 years of age in men starting adult life as obese. Methods: Among 362 200 Danish young men, examined for military service between 1943 and 1977, all obese (defined as body mass index (BMI)X31.0 kg m2), and, as controls, a random 1% sample of the remaining population were identified. A total of 1862 obese, corresponding to all men above the 99.5 percentile in this population, and 3476 controls were included, at a median age of 19 years (range: 18–25 years of age). They were followed until 2007 and Cox regression models were used to estimate the mortality in the obese relative to the controls. In addition, two reference groups were used: normal weight men (BMI: 18.5–24.9 kg m2) and the men with the lowest mortality in this cohort (BMI: 22.0–24.9 kg m2). Results: During the 65 years of follow-up, 1191 men died. At all ages from 18 to 80 years, the mortality in the obese was twice that of the controls (hazard ratio (HR): 2.10; 95% confidence interval (CI): 1.84–2.39). The median survival proportion (0.5) was reached about 8 years earlier in the obese than in either of the reference groups. Relative to the normal weight and men with the lowest mortality HRs of 2.14 (95% CI: 1.86–2.45) and 2.38 (95% CI: 2.00–2.85), respectively, were estimated for the obese. Neither year of birth nor education significantly influenced the excess mortality. Conclusion: Men entering adult life as obese experience a lifelong doubling of mortality, a finding that strongly supports the continued need to avoid beginning adult life as obese. International Journal of Obesity (2011) 35, 1193–1199; doi:10.1038/ijo.2010.274; published online 15 February 2011 Keywords: obesity in early adulthood; mortality; epidemiology

Introduction The prevalence of obesity has been increasing worldwide and in all age groups for some decades.1 However, the impact of the epidemic on children and adolescents seems greater than in adults, reflected by relatively higher increase in overweight and obesity in the younger age groups.1,2 The effect of obesity on mortality has been investigated thoroughly for decades. A recent report by the Prospective Studies Collaboration, with data from 900 000 participants and 57 prospective studies, found that obesity is associated with a nearly twofold increased all-cause mortality in both sexes and in all ages from 35 to 89 years

Correspondence: Dr E Zimmermann, Institute of Preventive Medicine, Øster Søgade 18, 1, Copenhagen DK1357K, Denmark. E-mail: Received 25 March 2010; revised 4 December 2010; accepted 16 December 2010; published online 15 February 2011

compared with people of normal weight.3 Similar results have been found in another recent report with data from 19 prospective studies.4 A common characteristic of these previous cohort studies is that the effect of obesity is investigated in predominantly middle-aged populations, which do not allow a distinction between recent or early development of obesity. Large-scale studies of the association between obesity in early adulthood and mortality has been conducted,5–8 but the lifelong effect of starting adult life as obese is poorly elucidated. In view of the still progressing epidemic of childhood obesity throughout the world,2 and the positive association between childhood obesity and obesity in young adulthood, it is important to get solid estimates of the lifelong health impact of this condition. The aim of this study is to examine the association between obesity present in early adulthood and morbidity, including fatal morbidity, and all-cause mortality throughout adult life in men. Here, we report the results of the study of all-cause mortality.

Obesity in young adulthood and mortality E Zimmermann et al


Population and methods Study participants The present analyses are based on studies, initiated in the early 1970s,9–11 of obesity in a population of Danish young men examined at the draft boards since weighing was introduced in the examination. This started in 1943 in District 1, the greater Copenhagen area, and in 1964 in District 2, the remainder of the Eastern part of Denmark. The young men were enforced by law to appear before the board at age 18 or shortly thereafter. The only exemption was severe medical disabilities (5% of the study population), so obesity was not a condition allowing exemption.11 Another 2% did not appear as they volunteered for service before the age of 18. Our study population comprised the 362 200 Danish men examined at the boards in District 1 from 1943 to 1977 and in District 2 from 1964 to 1977. Among these, all obese men (n ¼ 1930) and a 1% random sample (n ¼ 3601) of the remainder of the underlying population were manually identified in the archived paper files during the 1970s. Obesity was defined as 35% overweight relative to a Scandinavian standard in use at the time of sampling. This threshold later proved to correspond to a body mass index (BMI)X31.0 kg m2, which is above the 99.5th percentile of BMI in this population (Figure 1). The random sample, which serves as a control cohort, does not include the extremely overweight men already included in the obese group. Thus, a study population was sampled, which represented the underlying BMI distribution of the population through the control cohort, and where the right end of the distribution was 100-fold enriched relative to the controls. This design allowed us to investigate the mortality in the obese group in detail, and compare it to the mortality in the population from which they were sampled.

Covariates A priori, draft board district, year of birth and educational attainment at draft board examination were identified as potential confounders. Draft board district is associated with obesity prevalence,9 and may also be associated with mortality, because of both regional variation in mortality and the different sampling periods in the two districts.11 Year of birth is a potential confounder because of decline in general mortality and the rise in obesity prevalence in more recent time.10 Education is associated with both obesity12 and mortality.13 Information on educational attainment was recorded at the draft board examinations from 1956 onwards, therefore this information was missing in the 973 men examined before this year. Education was grouped into four categories according to years of schooling (see Table 1), which has shown to be predictive of social position later in adult life.14

Follow-up The Danish Civil Registration System was established on 2 April 1968 (see ref. 15) All Danish residents alive at that day

Table 1 Characteristics of the two cohorts of men given as median (range)

BMI, kg m2 BMI, N (deaths)b Underweight Normal weight Overweight Obese Year of birth Age, years Education, N (%)c p7 years p7 years þ apprenticeship 8–10 years 410 years



BMI The medical personnel at the Danish military service measured the men’s height without shoes and weight with only underwear. BMI was computed as weight in kilograms divided by square of height in meters.


N at risk at ages 30 years 40 years 50 years 60 years 70 years


0 15







Body mass index (kg/m2) Figure 1 Histogram of the distribution of body mass index at draft board examination for the 5338 men. Histograms illustrating the substantial enrichment of the right end of the tail by the obese cohort (BMIp31.0).

International Journal of Obesity

Obese cohort

Control cohort a

32.7 (31.0–51.8)

21.4 (15.7–30.9)

F F F 1862 (435) 1951 (1918–1959) 19 (18–25)

209 (44) 2983 (660) 284 (52) F 1946 (1918–1959) 19 (18–25)

753 581 280 131

(43.2) (33.3) (16.0) (7.5)

1811 1741 1500 352 44

559 939 654 468

(21.3) (35.8) (25.0) (17.9)

3368 3283 2990 1587 545

Abbreviations: BMI, body mass index; N, number. aAs the control cohort was a random sample of the total study population, we show the contribution from underweight, normal weight and overweight men in this sample. bBMI was categorised as underweight: BMI below 18.5; normal weight: BMI ranging between 18.5 and 24.9; overweight: BMI ranging between 25.0 and 30.9; obesity: BMI equal to or above 31.0. cData on education was missing in 117 obese and 856 randomly selected men.

Obesity in young adulthood and mortality E Zimmermann et al

1195 or born thereafter have been assigned a unique identification number. Information on vital status was obtained by linking the personal identification number to the Civil Registration System until 31 October 2007. Information about deaths occurring before 1968 were obtained from local population registers. One man could not be found in these registers.

Analytical strategy To keep the focus on men entering adult life as obese, only men between 18 and 25 years of age at draft board examination were included in the analyses. Hereby, 192 men not examined in this age range were excluded, leaving 1862 obese and 3476 controls for the analyses. We investigated the lifelong mortality in the obese versus the control cohort while controlling for the potential confounders. To facilitate comparisons with other studies, we also estimated the lifelong mortality in the obese versus a normal weight reference group as defined by World Health Organization (BMI: 18.5–24.9 kg m2) (see ref. 1) as well as a reference group consisting of the men with the lowest mortality in this population. Moreover, the sampling design allows data to be analyzed as cohort data with two nested sub-cohorts based on the different sampling fraction across the range of BMI at draft board examination.16 Thus, we investigated the BMI-mortality association throughout the broad range of BMI as well. All analyses were repeated in the sub-cohorts with data on education. Statistical methods Data were analysed with Stata (version 9.2, Stata Corporation, College Station, TX, USA). The observation time for each participant was the period from draft board examination (between 1943 and 1977) until the date of death (n ¼ 1191), emigration (n ¼ 194), loss to follow-up (n ¼ 5), or 31 October 2007, whichever came first. An individual was classified as lost to follow-up, if he had disappeared from the Civil Registration Registry for unknown reason. Cox proportional hazard regression was used to estimate hazard ratios (HRs) of mortality. Age was used as the underlying time axis with delayed entry to ensure that the estimation procedures were on the basis of comparisons of individuals of the same age. To take into account the putative confounding by draft board district, we used stratified estimation allowing the baseline hazards to differ by district and sampling period; the stratification variable had four classes defined as District 1 with the examination periods split into 1943–1953, 1954–1963 and 1964–1977, and District 2 with the examination period 1964–1977. To investigate whether the hazard of mortality changed across age, we plotted the cumulative hazard at corresponding ages for the obese group versus the control cohort (or the additional reference groups), a so-called double Nelson–Aalen plot. In this plot, the slope of the curve is equal to the HR at a given age. The Kaplan–Meier method was used for calculating survival curves. A smoothing spline

was used to explore the BMI-mortality association without a priori assumptions about the shape.17 A smoothing parameter of 5 degrees of freedom was chosen, which represents the amount of smoothing in terms of effective degrees of freedom for the spline. Year of birth was modelled as a continuous variable, after testing that the assumption of linearity to the log of the hazard function was fulfilled using restricted cubic splines. Education was modelled as a categorical variable as described above. Statistical interaction between the obese and the reference groups, and year of birth and education, respectively, was assessed by deviance tests based on comparisons of 2 log likelihood in nested models with and without cross-product terms.

Results Out of the 5338 young men, 1191 died during follow-up covering the age span from 18 to 80 years of age, with a mean follow-up of 37 years. Table 1 shows the characteristics for the obese and control cohorts. The majority of the men in the control cohort had a BMI within the normal range (18.5–24.9 kg m2). The median year of birth was 5 years later for the obese than the controls, because of the increase in obesity prevalence in more recent time.10 Also, a larger percentage of the obese had a low level of education compared with the controls. Finally, the numbers at risk at different ages throughout adult life is also illustrated in Table 1. Figure 2 illustrates the HR between the obese and control cohorts at equivalent ages. A curve concordant with the line of equality would show a constant HR of 1.00. The solid line illustrates the constant excess mortality in the obese cohort, based on the HRs calculated from Cox regression model without consideration of the covariates, and it verifies that the requirement of proportional hazards was fulfilled throughout the follow-up period. When adjusted for year of birth a HR of mortality of 2.10 (95% confidence interval (CI): 1.84–2.39, Po104) was observed in the obese group relative to the control cohort. The analysis was repeated in the sub-cohorts with data on education, but adjusting for education only attenuated the association marginally. The entire BMI range was investigated using a smoothing spline, which showed a curvilinear, J-shaped relation, with increasing mortality at both the lower and upper end of the BMI range (Figure 3). By visual inspection of the smoothing function the nadir of the curve was identified, with the lowest mortality in those with a BMI between 22.0 and 24.9 kg m2. The HR of death was 1.09 (95% CI: 1.07–1.12) per the increase of one BMI unit equal to or above 25.0 kg m2 when adjusted for year of birth.

Analyses using different reference groups Cumulative hazard plots similar to Figure 2 were obtained for the obese plotted against the group of normal weight men and for the obese plotted against the group of men International Journal of Obesity

Obesity in young adulthood and mortality E Zimmermann et al


Proportion surviving

The estimated cumulative hazards in obese group






75 y 0.00 1









Age 65 y

Figure 4 Kaplan–Meier plots for the obese cohort (dark grey line), the control cohort (black line) and the reference group with a BMI between 22 and 25 kg m2 (light grey line).

30 y 0 0

0.5 1 2 The estimated cumulative hazards in the controls

Figure 2 Cumulative hazard plot of mortality for the obese versus the control cohort at corresponding ages from 18 to 80 years. To investigate whether the hazard ratio changed across age, we created a graph of the cumulative hazard from the obese group versus the controls at corresponding ages. For every age the cumulative hazard in the obese group is plotted against the cumulative hazard in the control cohort. The arrows denote corresponding ages in the two cohorts. The grey, broken line is the line of equality. The interpretation of the black line is that the excess risk in the obese group is constant across the age range of observation from 18 to 80 years of age, and thus the assumption of proportionality in the Cox model is fulfilled.

HR for mortality

4 3


1 20

25 31 35 Body mass index (kg/m2)


Figure 3 Smoothing spline (5 degrees of freedom) of the association between BMI and all-cause mortality with 95% confidence limits and adjusted for year of birth.

with the lowest mortality risk and showed no signs of non-proportionality (not shown). The obese had a HR of mortality of 2.14 (95% CI: 1.86–2.45) compared with the normal weight reference group when year of birth was taken into account, and a HR of 2.38 (95% CI: 2.00–2.85) when compared with the group of men with the lowest mortality. International Journal of Obesity

The proportions surviving from age 18 to 80 years are described by the Kaplan–Meier plot in Figure 4. The figure illustrates that the obese were followed until 80 years of age, whereas the reference groups were followed until 90 years of age. Thus, we were able to estimate the difference in mortality between the obese and the reference groups until 80 years of age. The figure shows the shorter life expectation in the obese group compared with both the control cohort and those with the lowest mortality in this population. At age 70 years, about 70% in either of the reference groups versus only about 50% of the obese were expected to be alive. The plot also shows that the median survival proportion (0.5) was reached about 8 years earlier in the obese than in either of the reference groups. The plot for the group of normal weight men was almost identical to the plot for the control cohort, and thus not shown.

Interaction analyses There were no interactions between obesity status (obese versus reference group) and year of birth (P40.27 for all three reference groups). We found a stepwise decline in mortality with increasing educational attainment in both the obese and the different reference groups (P for trend ¼ 0.001 in the obese; P for trendo0.001 in the control cohort and the group of normal weight, respectively; P for trend ¼ 0.11 if the reference group with the lowest mortality). Further, the association between obesity status and mortality was investigated in strata of the educational categories as presented in Table 2, and no interactions were observed.

Discussion The present study show that obesity in young adulthood is associated with a doubling in mortality from ages 18 to 80 compared with the expected mortality estimated from the

Obesity in young adulthood and mortality E Zimmermann et al

1197 Table 2 Hazard ratios (95% confidence intervals) of all-cause mortality for the obese cohort (vs the three reference groups) in strata of educational attainment N (deaths)

HR (95% CIs)

P for Interaction a

Reference: the underlying population Educationb p7 years p7 years þ apprenticeship 8–10 years 410 years

1312 (278) 1520 (238)

1.71 (1.33–2.22) 2.11 (1.62–2.76)

934 (121) 599 (59)

1.97 (1.34–2.91) 1.97 (1.10–3.53)


Reference: normal weight group Educationb p7 years p7 years þ apprenticeship 8–10 years 410 years

1204 (262) 1367 (219)

1.60 (1.22–2.10) 2.12 (1.61–2.81)

836 (108) 535 (54)

2.06 (1.37–3.08) 1.94 (1.07–3.52)


Reference: lowest mortality groupc Educationb p7 years p7 years þ apprenticeship 8–10 years 410 years

915 (199) 851 (148)

2.43 (1.54–3.85) 2.70 (1.76–4.15)

483 (71) 281 (26)

2.01 (1.19–3.38) 2.86 (1.21–6.77)


Abbreviations: HR, Hazard ratio; CIs, Confidence Intervals; N, number. Information on educational attainment was missing for 973 men. aInteraction between the obese/reference group and educational attainment. bAdjusted for year of birth. cBMI between 22 and 25 kg m2.

control cohort, which represents the underlying population. This led to a difference in estimated proportions surviving through 70 years of age, where about 50% of the obese were expected to be alive compared with 70% controls. Equal survival proportions, that is, for example the survival of 50%, were reached 8 years earlier in the obese than in the controls, a difference which is comparable with the detrimental effect on mortality seen in smokers.18 When the mortality in the obese was compared with either the group of normal weight men, or with the group with the lowest mortality (BMI between 22.0–24.9 kg m2), the risk estimates increased slightly. With the obese cohort representing the most obese of all young menFabove the 99.5th percentile among 362 200 menFour study design provides great advantages for the analysis of this group. Conventional cohort studies usually lack sufficient numbers of obese individuals to get solid estimates of the impact of this condition. Another strength of the present study was the very long follow-up, which allows detection of associations occurring late in life. Further, in the time period investigated, examination for suitability for military service was mandatory leaving no suspicion of selection bias. The objective measures of height and weight diminishes the potential bias due to misreporting.19

Since 1943, the year from which the earliest data were available, the prevalence of obesity has increased about 50-fold in Denmark,10,20 and similar trends have been observed in most parts of the Western world.1 As the prevalence of obesity increases, the public health concerns have also increased. A recent large-scale study of Swedish male draftees followed through to middle age found that the obese had more than twice the mortality compared with a normal weight reference group.6 Likewise, a Norwegian study of male and female adolescents between 14 and 19 years of age followed through 32 years, reported a 80–100% excess mortality in those with BMI above the 95th percentile relative to the 25–74th percentile.5 We found similar effect sizes in the present study, and the excess mortality was constant at all ages from 18 to 80 years. Moreover, the association was not influenced by year of birth or education. The World Health Organization recommends an optimum BMI ranging from 18.5 to 24.9 kg m2 in regard to morbidity and mortality.1 However, this classification has been debated as Flegal et al.21 found a decreased mortality in the overweight (relative to the normal weight category), while obesity was associated with increased mortality in the NHANES surveys I-III, which are representative of the U.S. population. When investigating the BMI-mortality relation throughout the very broad range of BMI, we observed an optimum BMI in relation to mortality of 22.0–24.9 kg m2. Both the before mentioned Swedish study on male draftees and an American study on women (using recalled BMI at age 18) found that the minimum risk of death was observed for participants with BMI between 18.5–21.9 kg m2, with an increasing mortality seen already in the upper range of normal weight (BMI 22.0–24.9) (see refs. 6, 22). Whether these small differences in optimum BMI are true differences or ascribed to the different study populations is, however, difficult to determine. Nevertheless, the current study as well as the previous studies of young adults, all indicate that mortality increases in individuals with BMI above 25 or 26 kg m2, respectively.6,22–25 The lack of association between overweight and mortality may therefore particularly be a phenomenon observed in older age groups, as suggested by other studies,26,27 while overweight young adults seems to be a serious health issue with harmful effects throughout life. Smoking is a potential confounder and modifier of the mortality risk, as smoking is associated with lower BMI and is strongly associated with cardiovascular diseases that are likely to be increased among those who are overweight and obese.3,28,29 The above-mentioned Swedish study of draftees found no significant interactions between smoking status and overweight or obesity in relation to mortality and also no evidence of confounding by smoking.6 Hence, using these results from a neighbouring population as basis, we consider our results as valid even though we did not have access to information about smoking among the young men examined at the draft board examination. On the basis of previous studies, it might be expected that the effect of obesity on subsequent mortality would International Journal of Obesity

Obesity in young adulthood and mortality E Zimmermann et al

1198 decline with age, because those most susceptible to the detrimental effects of obesity would have died prematurely.30 Stevens et al.31 found that the relative risk associated with greater body weight was highest in the younger groups. A recent review of 13 studies concluded that age modifies the relation between obesity and mortality, resulting in a lower relative risk of mortality associated with obesity among individuals aged 65 years or older compared with young and middle-aged populations.31,32 It is, however, possible that the health risk of obesity established early in life may be underestimated in cohort studies recruiting individuals later in life. These studies may not include those who have been obese since young adulthood and who have developed serious chronic diseases, either because these individuals are deliberately excluded or because they decline participation. Moreover, cohorts starting later in life will include individuals who have developed obesity later in life, which implies that the observed mortality does not pertain to the group in focus here. Overall, there is a considerable risk that the mortality observed in cohorts started during adult life may not reveal the detrimental effects of obesity present since start of adult life. In this study, we found no decline with age in the impact of obesity present when entering adulthood on mortality. It is difficult to determine whether the lifelong excess mortality associated with obesity in early adulthood is due to obesity in this young age, or whether it is due to obesity tracking into adulthood. A study on a sub-cohort of the earlier mentioned Norwegian adolescents who also had adult BMI measurements, found that the association between adolescent obesity and mortality was explained by adult BMI in men, whereas the effect was not entirely explained by adult BMI in women.33 An American study of 13–18-year olds of both sexes found an increased mortality in the overweight adolescents, which was independent of BMI at 55 years of age.34 Another American study on female nurses found that the association between recalled BMI at age 18 and mortality was only partly explained by adult BMI.22 Either way, the present study demonstrates that obesity in early adulthood obviously has a key role in relation to adult mortality. Next, we need to analyse the patterns of morbidities, including the morbidity apparently causing death, to gain more knowledge about the mechanisms through which obesity works, and whether they differ with age. In conclusion, we found that young men who enter adulthood as obese have more than twice the risk of mortality compared with the mortality observed in the underlying population. The association did not decline with age or alongside the increasing obesity epidemic, nor was it influenced by education. Our results support the continued need to prevent development of obesity before entrance to adulthood.

Conflict of interest E Zimmermann and C Holst both declare no conflicts of interest. TIA Sørensen has various industrial collaborations International Journal of Obesity

on obesity research as indicated on the website http://

Acknowledgements The work was supported by grants from the Cluster for Endocrinology and Metabolism, University of Copenhagen, Denmark. This work is also part of the project ‘Hepatic and adipose tissue and functions in the metabolic syndrome’ (HEPADIP, see, which is supported by the European Commission as an Integrated Project under the 6th Framework Programme (Contract LSHM-CT-2005018734), and part of the research activities in the Danish Obesity Research Centre (DanORC, see, supported by the Danish Council for Strategic Research.

References 1 Obesity: preventing managing the global epidemic. Report of a WHO consultation. World Health Organ Tech Rep Ser 2000; 894: i-253. 2 Wang Y, Lobstein T. Worldwide trends in childhood overweight and obesity. Int J Pediatr Obes 2006; 1: 11–25. 3 Prospective Studies Collaboration. Body-mass index and causespecific mortality in 900 000 adults: collaborative analyses of 57 prospective studies. Lancet 2009; 373: 1083–1096. 4 Berrington de Gonzalez A, Hartge P, Cerhan JR, Flint AJ, Hannan L, MacInnis RJ et al. Body-mass index and mortality among 1.46 million white adults. N Engl J Med 2010; 363: 2211–2219. 5 Engeland A, Bjorge T, Sogaard AJ, Tverdal A. Body mass index in adolescence in relation to total mortality: 32-year follow-up of 227 000 Norwegian boys and girls. Am J Epidemiol 2003; 157: 517–523. 6 Neovius M, Sundstrom J, Rasmussen F. Combined effects of overweight and smoking in late adolescence on subsequent mortality: nationwide cohort study. BMJ 2009; 338: b496. 7 Waaler HT. Hazard of obesityFthe Norwegian experience. Acta Med Scand Suppl 1988; 723: 17–21. 8 Sonne-Holm S, Sørensen TIA, Christensen U. Risk of early death in extremely overweight young men. Br Med J (Clin Res Ed) 1983; 287: 795–797. 9 Christensen U, Sonne-Holm S, Sørensen TIA. Constant median body mass index of Danish young men, 1943–1977. Hum Biol 1981; 53: 403–410. 10 Sonne-Holm S, Sørensen TIA. Post-war course of the prevalence of extreme overweight among Danish young men. J Chronic Dis 1977; 30: 351–358. 11 Sørensen TIA, Sonne-Holm S. Mortality in extremely overweight young men. J Chronic Dis 1977; 30: 359–367. 12 Teasdale TW, Sørensen TIA, Stunkard AJ. Intelligence and educational level in relation to body mass index of adult males. Hum Biol 1992; 64: 99–106. 13 Lynch JW, Kaplan GA. Socioeconomic position. In: Berkman LF KI, (ed). Social epidemiology. Oxford University Press: New York, 2000. pp 13–35. 14 Sonne-Holm S, Sørensen TIA. Prospective study of attainment of social class of severely obese subjects in relation to parental social class, intelligence, and education. Br Med J (Clin Res Ed) 1986; 292: 586–589. 15 Pedersen CB, Gotzsche H, Moller JO, Mortensen PB. The Danish Civil Registration System. A cohort of eight million persons. Dan Med Bull 2006; 53: 441–449. 16 Rothman K, Greenland S, Lash T. Modern Epidemiology. Third ed. Lippincott, Williams & Wilkins: Philadelphia, PA, 2008.

Obesity in young adulthood and mortality E Zimmermann et al

1199 17 Hastie TJ, Tibshirani RJ. Generalized Additive Models. Chapman & Hall: New York, NY, 1990. 18 Doll R, Peto R, Boreham J, Sutherland I. Mortality in relation to smoking: 50 years’ observations on male British doctors. BMJ 2004; 328: 1519. 19 Niedhammer I, Bugel I, Bonenfant S, Goldberg M, Leclerc A. Validity of self-reported weight and height in the French GAZEL cohort. Int J Obes Relat Metab Disord 2000; 24: 1111–1118. 20 Sørensen HT, Sabroe S, Gillman M, Rothman KJ, Madsen KM, Fischer P et al. Continued increase in prevalence of obesity in Danish young men. Int J Obes Relat Metab Disord 1997; 21: 712–714. 21 Flegal KM, Graubard BI, Williamson DF, Gail MH. Excess deaths associated with underweight, overweight, and obesity. JAMA 2005; 293: 1861–1867. 22 van Dam RM, Willett WC, Manson JE, Hu FB. The relationship between overweight in adolescence and premature death in women. Ann Intern Med 2006; 145: 91–97. 23 Bjorge T, Engeland A, Tverdal A, Davey Smith G. Body mass index in adolescence in relation to cause-specific mortality: a follow-up of 230 000 Norwegian adolescents. Am J Epidemiol 2008; 168: 30–37. 24 Hoffmans MD, Kromhout D, de Lezenne CC. The impact of body mass index of 78 612 18-year old Dutch men on 32-year mortality from all causes. J Clin Epidemiol 1988; 41: 749–756. 25 Jeffreys M, McCarron P, Gunnell D, McEwen J, Smith GD. Body mass index in early and mid-adulthood, and subsequent mortality: a historical cohort study. Int J Obes Relat Metab Disord 2003; 27: 1391–1397.

26 McGee DL. Body mass index and mortality: a meta-analysis based on person-level data from twenty-six observational studies. Ann Epidemiol 2005; 15: 87–97. 27 Strawbridge WJ, Wallhagen MI, Shema SJ. New NHLBI clinical guidelines for obesity and overweight: will they promote health? Am J Public Health 2000; 90: 340–343. 28 Lawlor DA, Hart CL, Hole DJ, Davey Smith G. Reverse causality and confounding and the associations of overweight and obesity with mortality. Obesity 2006; 14: 2294–2304. 29 Manson JE, Bassuk SS, Hu FB, Stampfer MJ, Colditz GA, Willett WC. Estimating the number of deaths due to obesity: can the divergent findings be reconciled? J Womens Health (Larchmt) 2007; 16: 168–176. 30 Zamboni M, Mazzali G, Zoico E, Harris TB, Meigs JB, Di Francesco V et al. Health consequences of obesity in the elderly: a review of four unresolved questions. Int J Obes (London) 2005; 29: 1011–1029. 31 Stevens J, Cai J, Pamuk ER, Williamson DF, Thun MJ, Wood JL. The effect of age on the association between body-mass index and mortality. N Engl J Med 1998; 338: 1–7. 32 Heiat A, Vaccarino V, Krumholz HM. An evidence-based assessment of federal guidelines for overweight and obesity as they apply to elderly persons. Arch Intern Med 2001; 161: 1194–1203. 33 Engeland A, Bjorge T, Tverdal A, Sogaard AJ. Obesity in adolescence and adulthood and the risk of adult mortality. Epidemiology 2004; 15: 79–85. 34 Must A, Jacques PF, Dallal GE, Bajema CJ, Dietz WH. Long-term morbidity and mortality of overweight adolescents. A follow-up of the Harvard Growth Study of 1922–1935. N Engl J Med 1992; 327: 1350–1355.

International Journal of Obesity

Lifelong doubling of mortality in men entering adultlife as obese