Increasing prevalence of overweight and obesity has raised major public health concerns.1 This concern stems primarily from a large number of observational studies linking high body mass index (BMI), as a measure of overweight and obesity, to morbidity and mortality.2–8 Recently, a pooled analyses based on 57 prospective studies showed that a 5-unit increase in BMI is associated with a 30% increase in risk of all-cause mortality and a 40% increase in risk of CVD mortality.8 However, the interpretation of these findings is complicated by the fact that both BMI and mortality in part are inherited: twin studies indicate heritability of 50%–90% for BMI and 20%–30% for longevity.9–12 If some of the genes that influence mortality also lead to overweight, the associations observed between overweight and mortality could be because of genetic factors. As far as we know, the role of genetic factors in the association between BMI and death has not been explored previously.
The twin design is an invaluable tool to control for genetic confounding, taking advantage of the fact that monozygotic (MZ) twins are presumed to be genetically identical and dizygotic (DZ) twins are presumed to share about half of their segregating genes. Differences in mortality or disease rate within MZ pairs are probably due to environmental causes. We use data from the Swedish Twin Registry—the largest twin cohort in the world—to investigate to what extent the association between BMI and all-cause mortality, CVD and CHD mortality is confounded by genetic factors. If genetic factors were the main explanation for the association between BMI and mortality, we would expect to find an association within DZ twins, where genetic confounding is only partially controlled, whereas no association (or a relatively weak association) between BMI and mortality would be expected within MZ pairs.
The Swedish Twin Registry
Our analysis is based on data from 2 cohorts within the Swedish Twin Registry. The first cohort consists of all same-sex twin pairs born 1886–1925 who were resident in Sweden in 1961. The second cohort consists of all same-sex twins born from 1926 through 1958 who were living in Sweden in 1970.13
The first cohort received a questionnaire in 1967 including questions on height and weight, health, and lifestyle factors. In 1969/1970, an additional questionnaire was sent to those who did not respond to the previous survey. Corresponding information for the second cohort was collected in 1972. The response rates were 85% and 88%, respectively. Responses were received from 55,911 twins who were alive at baseline; among those, 44,258 subjects (79%) had complete information on BMI, smoking, and zygosity.
The data collection was approved by the Swedish Data Inspection Authority and the Ethical Committee at Karolinska Institutet.
Zygosity assessment was based on the question: “were you as children as alike as 2 peas in a pod?” When both twins answered affirmatively, they were defined as monozygotic and when both answered no they were determined as dizygotic. Fewer than 5% of twins disagreed in their answers; those were excluded. This method has been demonstrated to correctly classify more than 95% of the twins.13
Body Mass Index
BMI (kg/m2) was calculated based on self-reported height and weight obtained from questionnaire responses in 1969/1970 or in 1967 (if information was missing in the 1969/1970 questionnaire) for the first cohort and in 1972 for the second cohort. We categorized subjects as underweight (BMI <18.5), lean/normal (18.5 ≤ BMI < 25), overweight (25 ≤ BMI < 30), and obese (BMI ≥30) according to WHO classification. Subjects with BMI higher than 45 were considered outliers and excluded (n = 29).
Subjects were classified as never, former, or current smokers. Alcohol consumption was assessed by separate questions on the amount of beer, wine, and spirits consumed during a month. There was only baseline information on self-reported diabetes and high blood pressure for the first cohort. Physical activity was assessed only in the second cohort, by a question on average physical activity during leisure time in the past year, with 7 response options ranging from “almost never” to “very much.” This variable has previously been linked to mortality.14
Mortality was determined by linkage to the National Causes of Death Registry, using the personal identification number assigned to every Swedish citizen. This registry includes information on all deaths in Sweden, with cause of death classified according to the International Classification of Diseases (ICD). All deaths to 2004 were identified, including deaths with CVD (ICD 8–9, codes 390–459, and later ICD 10, codes I00–I99) or CHD (ICD 8–9, codes 410–414, and later ICD 10, codes I20–I21, I24–I25) as the primary or secondary cause of death.
Heredity of BMI
We used data from MZ and DZ twins simultaneously in a structural equation model to estimate how much of the variance in BMI could be explained by additive genetic effects, common environmental effects, and unique environmental effects.15 Shared environmental effects reflect twin similarity not explained by heritable genetic influence. These could include the intrauterine environment or early shared family factors. Unique environmental effects are those explained by individual lifestyle or other environmental factors specific to the individual. These analyses were based on the assumptions that twins are representative of the general population and that the genetic similarity is half as great for DZ as for MZ twin pairs, whereas shared environmental influences contribute equally to the similarity of MZ and DZ pairs.15
We analyzed twin data without adjustment for genetic factors in an ordinary cohort analysis. The association between BMI and mortality was estimated by hazards ratios (HRs) using Cox proportional hazards models (SURVIVAL 2.16 Package with Coxph in R 2.0). This was done for total mortality, CVD mortality, and CHD mortality. Person-years were accumulated from age at start of the study (1 January 1972 for the first cohort and 1 January 1975 for the second cohort) until age of death or age at end of follow-up (30 June 2004), whichever came first. Follow-up was lagged by 2 years to exclude subjects who died shortly after filling out the questionnaire. Age was the underlying timescale in the Cox model, and a frailty component was included to handle within-pair dependences.16 BMI was analyzed categorically, as 4 categories, with normal weight (BMI; 18.5–25) as reference, and continuously (excluding subjects with BMI less than 18.5). To adjust for confounding, we included smoking in the Cox model. Additional adjustment for alcohol consumption, physical activity, socioeconomic status, diabetes, and high blood pressure did not affect the HRs and were therefore not included in the final models.
Co-twin Control Analyses
To investigate the association between body mass index and mortality while taking genetic factors into account, we used multivariate conditional logistic regression to estimate the odds ratio (OR) of the heavier twin dying before the leaner co-twin. The design resembles a matched case-control design. An advantage of this approach is that the association between BMI and mortality can be controlled for confounding factors shared by the twins in a pair, such as common genes, sex, age, and environment shared by the twins early in life. The analyses were done separately in MZ and DZ twins. In the first set of analyses, we calculated ORs separately for co-twins with a difference of 1, 2, 3, 4, or ≥5 BMI units within the pair. The results of these analyses indicated that a linear model was appropriate, and we therefore used BMI as a continuous variable, in a second set of co-twin control analyses. Because DZ twins share 50% of their segregating genes, the estimated within-pair effect of BMI on mortality risk in DZ twins is only partly controlled for genetic factors. In contrast, analyses within MZ twin pairs fully control for genetic factors. Therefore, if the effect of BMI on the risk of mortality is smaller within MZ than DZ twins, we infer that the association is confounded by genetic factors. To assess the difference between the association between BMI and mortality by zygosity, we calculated 95% confidence intervals (CIs) for the ratio of ORs within MZ and DZ twins. A more detailed description of the twin methods used here is provided elsewhere.17
In this cohort of all eligible Swedish twins born in Sweden 1886–1958, the age range was 16–86 years at baseline, and mean BMI at baseline was 24.5 kg/m2 in women and 24.6 kg/m2 in men. The prevalence of overweight was 22% in women and 23% in men; 4% of women and 2% of men were obese (eFigure, http://links.lww.com/EDE/A432).
During follow-up, there were 14,217 deaths overall (6831 in men and 7386 in women) among the 44,258 twins. Among those, 9009 (63%) were deaths from CVD and 3564 (25%) were deaths from CHD. Mean follow-up time was 25.7 years.
Heredity of BMI
Structural equation analyses indicated that 78% (95% CI = 75%–80%) of the variance in BMI was attributable to genetic factors and less than 2% to common environmental factors. Consequently, the influence of unique environmental factors on BMI variance was 21% (95% CI = 20%–21%).
Overweight and obesity was associated with increased mortality in both men and women (Table 1). The HR of all-cause mortality was estimated as 1.22 (95% CI = 1.15–1.30) in overweight men and 1.30 (1.23–1.38) in overweight women; it was 1.64 (1.42–1.88) in obese men and 1.66 (1.49–1.84) in obese women. HRs were even higher for death from CVD and CHD.
When assessing BMI as a continuous variable for men and women combined, a one-unit increase in BMI was associated with an HR of 1.05 (95% CI = 1.05–1.06) for all-cause mortality, 1.07 (1.07–1.09) for CVD mortality, and 1.09 (1.08–1.10) for CHD mortality (Table 1). Another way of expressing these results is to say that a 15-kg weight difference between subjects 175-cm tall (corresponding to 4.9 BMI units) is associated with a 27% increased risk of all-cause mortality, a 39% increased risk of CVD mortality, and a 53% increased risk of CHD mortality. The association between BMI and mortality was weaker at higher ages; when the analyses were stratified by baseline age (≤50, 50–64, and ≥65 years), the HR of death per kg/m2 was 1.08 (1.06–1.09), 1.05 (1.04–1.06), and 1.02 (1.01–1.03), respectively.
To take physical activity into account we made separate analyses of the association between BMI and mortality in the younger cohort, for which information on physical activity was available. After adjustment for physical activity, the HR of all-cause mortality per unit BMI was marginally changed from 1.076 to 1.074. For CVD and CHD mortality, the change in HR was even less.
Co-twin Control Analyses
The co-twin control analysis was carried out on pairs with BMI between 18.5 and 45, for whom at least 1 co-twin had died during follow-up. These included 32% of MZ twins and 35% of DZ twins. Characteristics of these pairs are given in Table 2. As expected, these were older than twins used in the cohort analyses (mean age was 55 compared with 42 years). There was no difference in mean BMI between MZ and DZ twins, and only very small differences in other characteristics. When these twins were analyzed as an ordinary cohort (disregarding twinship), similar associations between BMI and mortality were seen in MZ and DZ twins. For example, for all-cause mortality, the HR associated with a one-unit increase in BMI was 1.01 (1.00–1.02) in MZ twins and 1.02 (1.01–1.03) in DZ twins.
Co-twin control analyses in DZ pairs indicated that the twin with higher BMI had an increased risk for death from all-causes, from CVD, and from CHD, compared with the leaner co-twin (Table 3). In MZ pairs, risk was increased for death from CHD (OR = 1.06 [1.00–1.12]) but the association with all-cause mortality and CVD was weak. To explore the association between BMI and mortality in addition to what is attributed to CHD, cases of CHD death were excluded from the analyses of all-cause and CVD mortality. After this exclusion, there was no association between BMI and all-cause (OR = 0.99 [0.96–1.02]) or CVD mortality (1.00 [0.95–1.04]). Consistent with this, we found that the ratio of OR in DZ and MZ twins was close to 1.0 for CHD, which implies that genetic factors do not influence the association between BMI and CHD, whereas for all-cause and CVD mortality (excluding CHD), the DZ/MZ ratio was 1.05 (95% CI = 1.01–1.09) and 1.05 (0.99–1.11), respectively, indicating a role for genetic factors. After stratification by age, similar results were seen, ie, the association between BMI and mortality was consistently stronger in DZ than in MZ pairs; the DZ/MZ ratio of OR associated with all-cause mortality (excluding CHD) was 1.06 (95% CI = 1.00–1.14) in subjects ≤50 years, 1.03 (0.97–1.10) in subjects 50–64 years, and 1.05 (0.97–1.13) in subjects ≥65 years.
We also compared mortality separately in twins with intrapair differences of 1, 2, 3, 4, and 5 BMI units (eTable, http://links.lww.com/EDE/A432), which included 62% of MZ pairs and 74% of DZ pairs in which at least 1 co-twin had died during follow-up. The analyses were hampered by small numbers, especially for CHD mortality, but the results were similar to those obtained when BMI was analyzed as a continuous variable, ie, the association between BMI and mortality appeared to be stronger in DZ twins than in MZ twins.
Numerous studies have demonstrated an association between BMI and mortality.2–8 Our aim was to explore the extent to which this association might be confounded by genetic factors. To confirm previous findings, we first analyzed the twins as an ordinary cohort, disregarding the influence of genetic factors. In these analyses, the association between BMI and mortality was almost identical to findings based on pooled data from 57 different cohort studies, indicating an RR of 1.28 for all-cause mortality and 1.40 for CVD mortality (per 5 kg/m2).8 In a second set of analyses, genetic factors were taken into account in co-twin control analyses of DZ and MZ twins. The striking finding of these analyses was that BMI was not associated with all-cause mortality or CVD mortality in MZ twins, ie, in genetically identical individuals, the heavier twin did not tend to die before his leaner sibling. This implies that genetic factors contribute to the association between BMI and mortality, and that subjects genetically prone to develop high BMI may also carry a genetically increased risk of death. In contrast, the association between BMI and death from CHD persisted in co-twin control analyses, which supports the existence of a causal relationship. This is an important finding because CHD is the leading cause of death in the western world, accounting for 25% of all deaths in this cohort. The reason that BMI may be causally related to CHD but not CVD mortality could be that factors through which BMI may exert an influence (such as dyslipidaemia and hyperlipidemia, hyperinsulinemia, metabolic disturbances and hypertension) are more associated with the development of coronary artery disease and subsequent CHD, rather than to CVD in general.18 On the other hand, it is also important to note that there were more than twice as many CVD deaths as CHD deaths, making the results more precise for CVD than for CHD.
This study was based on a single assessment of BMI from almost 40 years earlier. Misclassification of exposure is therefore likely to be substantial, which has presumably diluted the association between BMI and mortality. Still, it is noteworthy that despite its crudeness, our measure of BMI was linked in cohort analyses to mortality in a similar way as in other studies with more detailed measures and shorter follow-up time.8 Also, the crude measure of BMI does not explain the different results for MZ and DZ twins.
Some studies indicate that obesity rather than overweight carries an increased risk of premature death.19 Using BMI measured in the late 1960s, fewer than 4% of our twins were obese at baseline. Because of the low prevalence of obesity, and also to the similarity of BMI in twins (only 21% of MZ and 36% of DZ differed by more than 2 BMI units), we could not investigate mortality in pairs where 1 twin was normal weight and the other obese. The extent to which the association between obesity and mortality is explained by genetic factors is therefore unclear; it is possible that genetic factors underlying extremely high BMI are different from those associated with more moderate BMI. There are alternative measures of adiposity such as waist-hip ratio that are linked to mortality and also partly inherited.5,20 Information on waist-hip ratio was not available in the present study, the role of genetic factors in that context is a topic for future studies.
Our conclusions are based on the assumption that twins are representative of singletons. Even though this seems to be true for most lifestyle factors and health outcomes, MZ twins tend to be slightly lighter than both DZ twins and singletons.21,22 Still, we have no reason to believe that this would have influenced the effects of BMI on mortality. Notably, there were no differences in mortality rates by zygosity. Furthermore, when the cohort analyses were stratified by zygosity, the association between BMI and mortality was similar in MZ and DZ twins. The generalizability of our findings is supported by the fact that the cohort results were consistent with previous findings in unrelated singletons.8 However, if, for example, smoking, illness, or some other factor related to mortality were the cause of the BMI discordance within MZ pairs but not in DZ pairs, this could explain why no increased all-cause or CVD mortality risk was seen in the heavier MZ co-twins. In a study of Finnish twins, smoking did contribute to BMI discordance within MZ pairs.23 Still, our results persisted after adjustment for a number of factors (including smoking, alcohol consumption, physical activity, socioeconomic status, diabetes, and high blood pressure), but there may be other factors related to both BMI and mortality (such as prenatal factors) for which we did not have information. If MZ and DZ twins differ with regard to reliability of self-reported weight and height, the results may be confounded. In this context, it should be noted that the association between BMI and death from CHD was similar in MZ and DZ pairs, which we would not expect if BMI difference had different etiology in MZ than in DZ twins, or if misclassification of BMI were more pronounced in MZ twins.
Both BMI and mortality are in part inherited; in the present study, 78% of the individual variance in BMI could be attributed to genetic influence. It has previously been shown in this cohort that about 25% of the variance in longevity is inherited.9–12 Specific genes that might affect both high weight and mortality are not known; it is likely that a wide range of genes play a role in both pathways. A variant of the FTO gene has been linked to obesity and type 2 diabetes24 and to the metabolic syndrome.25 Because this is a common genotype (present in about half of the studied population) and because it is linked to both obesity and morbidity, this could be an example of a potentially interesting genetic component.24 There are also heritable traits such as IQ and depression, which are linked to both BMI and mortality and that therefore may be of interest in this context.26–31
1. WHO. Obesity; preventing and managing the global epidemic. Geneva: WHO; 2000.
2. Adams KF, Schatzkin A, Harris TB, et al. Overweight, obesity, and mortality in a large prospective cohort of persons 50 to 71 years old. N Engl J Med.
3. Jee SH, Sull JW, Park J, et al. Body-mass index and mortality in Korean men and women. N Engl J Med.
4. Calle EE, Teras LR, Thun MJ. Obesity and mortality. N Engl J Med.
5. Pischon T, Boeing H, Hoffmann K, et al. General and abdominal adiposity and risk of death in Europe. N Engl J Med.
6. Yan LL, Daviglus ML, Liu K, et al. Midlife body mass index and hospitalization and mortality in older age. JAMA.
7. Manson JE, Colditz GA, Stampfer MJ, et al. A prospective study of obesity and risk of coronary heart disease in women. N Engl J Med.
8. Prospective Studies Collaboration; Whitlock G, Lewington S, et al. Body-mass index and cause-specific mortality in 900 000 adults: collaborative analyses of 57 prospective studies. Lancet.
9. Maes HH, Neale MC, Eaves LJ. Genetic and environmental factors in relative body weight and human adiposity. Behav Genet.
10. Ljungquist B, Berg S, Lanke J, McClearn GE, Pedersen NL. The effect of genetic factors for longevity: a comparison of identical and fraternal twins in the Swedish Twin Registry. J Gerontol A Biol Sci Med Sci.
11. Skytthe A, Pedersen NL, Kaprio J, et al. Longevity studies in GenomEUtwin. Twin Res.
12. Herskind AM, McGue M, Holm NV, Sørensen TI, Harvald B, Vaupel JW. The heritability of human longevity: a population-based study of 2872 Danish twin pairs born 1870–1900. Hum Genet.
13. Lichtenstein P, De Faire U, Floderus B, Svartengren M, Svedberg P, Pedersen NL. The Swedish Twin Registry: a unique resource for clinical, epidemiological and genetic studies. J Intern Med.
14. Carlsson S, Andersson T, Lichtenstein P, Michaëlsson K, Ahlbom A. Physical activity and mortality—is the association explained by genetic selection? Am J Epidemiol.
15. Neale MC, Boker SM, Xie G, Maes HH. MX: Statistical Modelling
. 5th ed. Richmond, VA: Virginia Commonwealth University; 2002:112.
16. Therneau TM, Grambsch PM, Pankratz VS. Penalized survival models and frailty. Rochester, MN: Department of Health Science Research, Mayo Clinic; 2000. Technical report no. 66.
17. Neale MC, Cardon LR, eds. Methodology for Genetic Studies of Twins and Families
. Dordrecht, Netherlands: Kluwer Academic Publishers; 1992.
18. Kannel WB. Hazards, risks, and threats of heart disease from the early stages to symptomatic coronary heart disease and cardiac failure. Cardiovasc Drugs Ther.
19. Flegal KM, Graubard BI, Williamson DF, Gail MH. Cause-specific excess deaths associated with underweight, overweight and obesity. JAMA.
20. Nelson TL, Vogler GP, Pedersen NL, Miles TP. Genetic and environmental influences on waist-to-hip ratio and waist circumference in an older Swedish twin population. Int J Obes Relat Metab Disord.
21. Andrew T, Hart DJ, Snieder H, de Lange M, Spector TD, MacGregor AJ. Are twins and singletons comparable? A study of disease-related and lifestyle characteristics in adult women. Twin Res.
22. Evans DM, Martin NG. The validity of twin studies. GeneScreen.
23. Hakala P, Rissanen A, Koskenvuo M, Kaprio J, Rönnemaa T. Environmental factors in the development of obesity in identical twins. Int J Obes Relat Metab Disord.
24. Frayling TM, Timpson NJ, Weedon MN, et al. A common variant in the FTO gene is associated with body mass index and predisposes to childhood and adult obesity. Science.
25. Freathy RM, Timpson NJ, Lawlor DA, et al. Common variation in the FTO gene alters diabetes-related metabolic traits to the extent expected given its effect on BMI. Diabetes.
26. Bartels M, Rietveld MJ, Van Baal GC, Boomsma DI. Genetic and environmental influences on the development of intelligence. Behav Genet.
27. Kendler KS, Gatz M, Gardner CO, Pedersen NL. A Swedish national twin study of lifetime major depression. Am J Psychiatry.
28. Chandola T, Deary IJ, Blane D, Batty GD. Childhood IQ in relation to obesity and weight gain in adult life: the National Child Development (1958) Study. Int J Obes.
29. Lager A, Bremberg S, Vågerö D. The association of early IQ and education with mortality: 65 year longitudinal study in Malmö, Sweden. BMJ
30. Cuijpers P, Smit F. Excess mortality in depression: a meta-analysis of community studies. J Affect Disord.
31. Zhao G, Ford ES, Dhingra S, Li C, Strine TW, Mokdad AH. Depression and anxiety among US adults: associations with body mass index International. Int J Obes.