Body mass index and cancer: Results from the Northern Sweden Health and Disease Cohort



Excess weight has been associated with increased risk of cancer. The effect of body mass index (BMI, kg/m2) on overall cancer risk and on risk of developing several common cancer types was examined in a population-based cohort study. Height and weight measurements were available for 35,362 women and 33,424 men recruited in the Northern Sweden Health and Disease Cohort between 1985 and 2003. Among cohort members, 2,691 incident cancer cases were identified. The association of BMI with cancer risk was examined using Poisson regression. Women with BMI > 27.1 (top quartile) had a 29% higher risk of developing any malignancy compared to women with BMI of 18.5–22.2 (lowest quartile), which increased to 47% in analysis limited to nonsmokers. Analyses according to WHO cut-off points showed that obese women (BMI ≥ 30) had a 36% higher risk of cancer than women with BMI in the normal range (18.5–25). Individual cancer sites most strongly related to obesity were endometrium (risk for top quartile = 3.53, 95% confidence interval 1.86–7.43), ovary (2.09, 1.13–4.13) and colon (2.05, 1.04–4.41). BMI was inversely related to breast cancer occurring before age 49 (0.58, 0.29–1.11, ptrend < 0.04). In men, there was no association of BMI with overall cancer risk. Obese men (BMI ≥ 30), however, were at increased risk of developing kidney cancer (3.63, 1.23–10.7) and, after exclusion of cases diagnosed within 1 year of recruitment, colon cancer (1.77, 1.04–2.95). Our study provides further evidence that BMI is positively associated with cancer risk. In women from northern Sweden, up to 7% of all cancers were attributable to overweight and obesity and could be avoided by keeping BMI within the recommended range. © 2005 Wiley-Liss, Inc.

Excess body weight develops as a consequence of increased energy intake relative to energy expenditure. In contemporary life, a positive energy balance is favored by the wide availability of calorie-dense foods (rich in saturated fats and readily digestible carbohydrates) combined with a reduced level of physical activity and increased prevalence of a sedentary lifestyle.1 As a result, the obesity epidemic is spreading at an alarming rate, affecting adults, adolescents and children from both industrialized and developing countries.1, 2

Overweight and obesity have long been recognized as an important cause of dyslipidemia, type 2 diabetes mellitus, hypertension, coronary heart disease, stroke and nonalcoholic fatty liver disease.3, 4 In 2001, a Working Group on Evaluation of Cancer Preventive Strategies at IARC concluded that high body weight also confers elevated risk of developing cancer of the esophagus (adenocarcinoma), colorectum, endometrium, postmenopausal breast and kidney (renal cell).1 More recent epidemiologic evidence indicates that obesity may increase the incidence and/or mortality of liver, gallbladder, pancreatic and stomach cancers.2 Major mediators of the adverse effect of obesity and overweight on disease risk are believed to be alterations in the synthesis of endogenous hormones such as insulin, adipokines, sex steroids and insulin-like growth factors.1, 2, 5, 6

A reasonably good and practical approximation of body adiposity is provided by indices based on weight and height.7 By far the most widely used weight-for-height measure is body mass index (BMI), defined as weight (in kilograms) divided by height (in meters squared).7 BMI can be easily compared across studies and populations, and since 1980, it is the most common index of adiposity used to define normal weight, overweight and obesity.1, 8

Our aim was to assess the effect of BMI on overall cancer risk and on risk of several common cancer types in a population-based cohort study.

Material and methods

The Northern Sweden Health and Disease Cohort

The Northern Sweden Health and Disease Cohort (NSHDC) is a long-term population-based interventional study intended for the promotion of health in the population of Västerbotten County, northern Sweden, with approximately 256,000 inhabitants. The principal aim of the study is screening for primary prevention of cardiovascular disease and diabetes and to advocate a healthy diet and lifestyle to the general public. The study was launched in 1985 and is still ongoing.

All persons residing in the county of Västerbotten are invited to participate in a health survey in the calendar year in which they become 30 (up to 1996), 40, 50 or 60 years old. Administrative and logistic difficulties, rather than subjects' refusal to participate, lowered the recruitment rate, especially during the first years of the study. Comparison of participants vs. nonparticipants in the study has shown relatively small differences in social characteristics and overall health status.9

At initial recruitment, subjects are asked to complete a self-administered questionnaire to collect demographic, medical and lifestyle information, including smoking status. Body weight is measured in light indoor clothing and recorded to the closest kilogram. Height is measured without shoes to the nearest centimeter. A medical examination follows, including withdrawal of a 20 cc blood sample for standard biochemical analysis (blood glucose and lipids) and for research purposes. The visit ends with a counseling session on healthy diet and physical activity conducted by specifically trained nurses.

Since 1994, at 10-year intervals, cohort members are invited to a second health examination and asked to donate a new blood sample and update questionnaire information.

For the purpose of this study, data collected up to December 31, 2003, are presented.

Study population

In October 2003, 74,207 subjects (38,530 women, 35,677 men) had been recruited in the NSHDC. From these, we excluded participants with missing values for height or weight (n = 1,180), those who had previous history of cancer except nonmelanoma skin cancer (n = 1,435), those without information on smoking status (n = 2,896) and those with BMI < 18.5 (n = 562). In total, 68,786 members of the NSHDC (35,362 women, 33,424 men) were included (Table I).

Table I. Characteristics of NSHDC Participants Included in the Study
  • 1

    According to WHO criteria for normal weight (BMI 18.5–24.9), overweight (BMI 25–29.9) and obesity (BMI ≥30).8

  • 2

    Current smokers were those who reported regular smoking of cigarettes, cigarillos, cigars or pipe. Nonsmokers and occasional smokers were classified as nonsmokers, and subjects who had stopped regular smoking were classified as ex-smokers.

Cohort participants35,36233,424
Age at entry (years)
 Mean (SD)46.1 (9.7)46.0 (9.8)
Mean (SD) years of follow-up8.3 (3.5)8.2 (3.6)
Number of cancer cases1,4401,251
Person-years at risk292,652273,715
BMI (kg/m2) at baseline
 Mean (SD)25.3 (4.2)26.0 (3.5)
 Normal weight156%43%
Smoking at baseline2
 Current smoker21%18%

Ascertainment of vital status and migration

The cohort was linked to several national registries using the unique personal identification numbers assigned to all Swedish citizens. Vital status and date of death were ascertained by linkage with the Swedish Person Register (for deaths occurring until the end of 2001) and with the Northern Sweden Region Person Register (for deaths occurring in 2002 and 2003). Information on emigration from the country for the whole study period or change of residence outside of northern Sweden after 2002 was obtained from the Northern Sweden Region Person Register. During the study period, 2,487 cohort members have died, left the county or moved outside the northern region after 2002.

Identification of incident cancer cases

Invasive cancers (codes 140–209 of the International Classification of Diseases, 7th Revision) occurring among cohort members during the study period were identified by linkage with the Swedish Cancer Registry. The Swedish Cancer Registry was founded in 1958. Registration of newly detected cancers is based on mandatory reports from all physicians serving out- and inpatient departments in all public and private hospitals. Reporting is also mandatory for all pathologists involved with surgical biopsies, cytologic specimens and autopsies. Overall reporting to the registry is estimated to be 96–98% of all diagnosed cancers.10, 11 To ensure inclusion of all cases occurring in 2002 and 2003 in the county, linkage with the local Northern Sweden Cancer Registry was also performed. In total, 2,691 eligible cancer cases were identified (1,251 men, 1,440 women; Table I). Two hundred and fifty-two cancer cases (94 men, 158 women) were diagnosed within 1 year of recruitment into the cohort.

Total numbers of person-years at risk, after an average follow-up of 8.2 years, were 273,715 for men and 292,652 for women (Table I).

The study was reviewed and approved by the research ethical committee of Umeå University Hospital.

BMI categories

The main results for BMI are presented according to sex-specific BMI quartiles (to quantify risk according to BMI distribution in the study population) and according to the widely accepted BMI categories proposed by the WHO. The BMI quartile cut-off points were 18.5–23.4, 23.5–25.3, 25.4–27.6 and ≥27.7 for men and 18.5–22.1, 22.2–24.2, 24.3–27.0 and ≥27.1 for women (Table II). When the number of cases for a sex-specific site was lower than 40, analyses in tertiles were conducted with the following cut-off points: 18.5–24.1, 24.2–26.7 and ≥26.8 for men and 18.5–22.8, 22.9–25.9 and ≥26.0 for women. The WHO BMI categories used were 18.5–24.9 kg/m2 (normal weight), 25–29.9 kg/m2 (grade 1 overweight, herein referred to as overweight), 30–39.9 kg kg/m2 (grade 2 overweight or obesity) and ≥40 kg/m2 (grade 3 overweight or morbid obesity).8 As very few cohort members had obesity grade 3 [100 men (0.3%), 232 women (0.7%)], we merged those with obesity grades 2 and 3 into a single category, herein referred to as obesity.

Table II. Risk of Cancer According to BMI Quartiles (Or Tertiles) in Men and Women from The NSHDC1
  • 1

    Quartile cut-off points in men: 18.5–23.4, 23.5–25.3, 25.4–27.6 and ≥27.7. Tertile cut-off points in men: 18.5–24.1, 24.2–26.7 and ≥26.8. Quartile cut-off points in women: 18.5–22.1, 22.2–24.2, 24.3–27.1 and ≥27 for all women; 18.5–21.5, 21.6–23.3, 23.4–25.9 and >26.0 for women < age 49; 18.5–22.7, 22.8–24.9, 25.0–27.8 and ≥27.9 for women ≥ age 49. Tertile cut-off points in women: 18.5–22.8, 22.9–25.9 and ≥26.0 for all women: 18.5–22.1, 22.2–24.9 and ≥25.0 for women < age 49; 18.5–23.5, 23.6–26.8 and ≥26.9 for women ≥ age 49.

  • 2

    Directly standardized rate/10,000 person-years.

  • 3

    RR estimated from Poisson model with likelihood ratio CIs, adjusted for age, calendar year and smoking, all subjects.

  • 4

    RR estimated from Poisson model with likelihood ratio CIs, nonsmokers only.

  • 5

    Test for trend was performed using class medians as scores.

All cancers
  Number of cases266292335358 
  Rate2 (95% CI)48.2 (42.7–54.4)44.8 (40.0–50.3)44.5 (40.0–49.6)46.5 (41.8–51.6) 
  RR3 (95% CI)10.95 (0.81–1.13)0.95 (0.81–1.12)0.99 (0.84–1.16)0.99
  Number of cases122129147161 
  RR4 (95% CI)10.97 (0.82–1.15)0.95 (0.80–1.12)0.97 (0.82–1.15)0.80
  Number of cases254353364469 
  Rate2 (95% CI)43.4 (38.2–49.4)50.3 (45.3–55.9)46.2 (41.7–51.2)54.3 (49.5–59.6) 
  RR3 (95% CI)11.18 (1.00–1.39)1.09 (0.92–1.28)1.29 (1.11–1.51)0.003
  Number of cases133213218289 
  RR4 (95% CI)11.34 (1.08–1.67)1.20 (0.97–1.50)1.47 (1.19–1.81)0.002
Colorectal cancer
  Number of cases28273249 
  Rate2 (95% CI)5.1 (3.5–7.5)4.2 (2.9–6.1)4.2 (3.0–5.9)6.3 (4.7–8.3) 
  RR3 (95% CI)10.81 (0.47–1.37)0.81 (0.49–1.36)1.20 (0.76–1.94)0.25
  Number of cases18222246 
  Rate2 (95% CI)3.0 (1.9–4.9)3.2 (2.1–4.9)2.7 (1.8–4.1)5.0 (3.7–6.7) 
  RR3 (95% CI)10.97 (0.52–1.83)0.82 (0.44–1.56)1.54 (0.90–2.74)0.04
Colon cancer
  Number of cases13162024 
  Rate2 (95% CI)2.3 (1.4–4.1)2.5 (1.5–4.1)2.6 (1.7–4.0)3.1 (2.1–4.6) 
  RR3 (95% CI)11.03 (0.50–2.19)1.11 (0.56–2.30)1.28 (0.66–2.60)0.42
  Number of cases10171435 
  Rate2 (95% CI)1.8 (0.9–3.5)2.5 (1.6–4.0)1.7 (1.0–2.9)3.8 (2.8–5.4) 
  RR3 (95% CI)11.32 (0.61–3.01)0.92 (0.41–2.16)2.05 (1.04–4.41)0.02
Rectal cancer
  Number of cases13111123 
  Rate2 (95% CI)2.5 (1.4–4.2)1.7 (0.9–3.1)1.5 (0.8–2.6)2.9 (2.0–4.4) 
  RR3 (95% CI)10.72 (0.31–1.60)0.61 (0.27–1.37)1.23 (0.63–2.51)0.36
  Number of cases85810 
  Rate2 (95% CI)1.2 (0.6–2.5)0.7 (0.3–1.7)1.0 (0.5–2.0)1.0 (0.6–1.9) 
  RR3 (95% CI)10.53 (0.16–1.60)0.76 (0.28–2.08)0.86 (0.33–2.30)0.93
Stomach cancer
  Number of cases1061614 
  Rate2 (95% CI)1.8 (0.9–3.3)0.9 (0.4–2.0)2.1 (1.3–3.6)2.0 (1.1–3.4) 
  RR3 (95% CI)10.57 (0.19–1.54)1.41 (0.64–3.22)1.22 (0.54–2.83)0.32
  Number of cases7136  
  Rate2 (95% CI)0.9 (0.4–1.9)1.3 (0.8–2.2)0.5 (0.2–1.0)  
  RR3 (95% CI)11.45 (0.59–3.87)0.57 (0.18–1.74) 0.20
Pancreatic cancer
  Number of cases886  
  Rate2 (95% CI)1.1 (0.5–2.2)0.8 (0.4–1.7)0.6 (0.3–1.3)  
  RR3 (95% CI)10.87 (0.32–2.37)0.61 (0.20–1.75) 0.35
  Number of cases871115 
  Rate2 (95% CI)1.6 (0.7–3.3)1.0 (0.5–2.2)1.4 (0.7–2.4)1.7 (1.0–2.9) 
  RR3 (95% CI)10.71 (0.25–1.97)0.98 (0.39–2.55)1.23 (0.53–3.10)0.37
Urinary tract cancers
  Number of cases15171725 
  Rate2 (95% CI)2.6 (1.6–4.4)2.7 (1.7–4.4)2.3 (1.4–3.6)3.3 (2.2–4.9) 
  RR3 (95% CI)11.06 (0.53–2.15)0.96 (0.47–1.94)1.39 (0.74–2.71)0.30
  Number of cases5910  
  Rate2 (95% CI)0.7 (0.3–1.7)0.9 (0.5–1.8)0.8 (0.4–1.5)  
  RR3 (95% CI)11.35 (0.46–4.42)1.21 (0.42–3.97) 0.60
Kidney cancer
  Number of cases31012  
  Rate2 (95% CI)0.4 (0.1–1.4)1.1 (0.6–2.0)1.2 (0.7–2.0)  
  RR3 (95% CI)12.86 (0.87–12.8)3.20 (1.01–14.1) 0.09
  Number of cases5510  
  Rate2 (95% CI)0.7 (0.3–1.7)0.5 (0.2–1.2)0.8 (0.4–1.5)  
  RR3 (95% CI)10.76 (0.21–2.75)1.26 (0.44–4.10) 0.52
Malignant melanoma
  Number of cases6111314 
  Rate2 (95% CI)1.0 (0.4–2.2)1.6 (0.9–2.9)1.8 (1.0–3.1)1.9 (1.1–3.3) 
  RR3 (95% CI)11.73 (0.66–5.02)1.92 (0.76–5.48)2.04 (0.81–5.80)0.35
  Number of cases616917 
  Rate2 (95% CI)0.8 (0.3–1.8)2.3 (1.4–3.8)1.2 (0.6–2.3)2.4 (1.4–3.9) 
  RR3 (95% CI)12.55 (1.05–7.13)1.39 (0.50–4.19)2.56 (1.04–7.18)0.16
Non-Hodgkin, T cell (mm)
  Number of cases10151415 
  Rate2 (95% CI)1.8 (1.0–3.5)2.3 (1.4–3.8)1.9 (1.1–3.2)2.0 (1.2–3.3) 
  RR3 (95% CI)11.29 (0.59–2.98)1.07 (0.48–2.49)1.11 (0.50–2.56)0.97
  Number of cases61712  
  Rate2 (95% CI)0.8 (0.3–1.8)1.7 (1.1–2.8)1.0 (0.5–1.7)  
  RR3 (95% CI)12.20 (0.91–6.12)1.29 (0.49–3.75) 0.95
Respiratory tract cancers
  Number of cases16141613 
  Rate2 (95% CI)3.1 (1.9–5.1)2.1 (1.3–3.6)2.1 (1.3–3.4)1.6 (0.9–2.8) 
  RR3 (95% CI)10.84 (0.41–1.73)0.88 (0.44–1.78)0.72 (0.34–1.49)0.41
  Number of cases11111419 
  Rate2 (95% CI)2.3 (1.2–4.2)1.6 (0.9–3.0)1.8 (1.1–3.0)2.1 (1.3–3.3) 
  RR3 (95% CI)10.85 (0.36–1.98)0.97 (0.44–2.20)1.26 (0.60–2.75)0.38
Prostate cancer
 Number of cases93114129125 
  Rate2 (95% CI)17.9 (14.6–22.0)17.7 (14.8–21.3)16.8 (14.1–20.0)15.5 (13.0–18.5) 
  RR3 (95% CI)11.00 (0.76–1.32)0.96 (0.74–1.26)0.89 (0.68–1.16)0.31
Breast cancer
  Number of cases109142127136 
  Rate2 (95% CI)17.6 (14.4–21.4)19.7 (16.7–23.3)16.5 (13.8–19.6)15.7 (13.3–18.7) 
  RR3 (95% CI)11.14 (0.89–1.47)0.94 (0.72–1.22)0.95 (0.74–1.23)0.36
Breast cancer < age 49
  Number of cases23312414 
  Rate2 (95% CI)5.8 (3.8–8.8)7.6 (5.3–10.8)5.5 (3.7–8.3)3.3 (2.0–5.6) 
  RR3 (95% CI)11.31 (0.77–2.28)0.99 (0.56–1.76)0.58 (0.29–1.11)0.04
Breast cancer ≥ age 49
 Number of cases10810697111 
 Rate2 (95% CI)256 (211–310)256 (212–310)231 (189–282)263 (218–318) 
 RR3 (95% CI)10.99 (0.76–1.30)0.90 (0.68–1.18)1.04 (0.80–1.36)0.83
Ovarian cancer
 Number of cases13172337 
 Rate2 (95% CI)25 (14–45)25 (16–41)29 (19–44)48 (34–67) 
 RR3 (95% CI)11.13 (0.55–2.38)1.39 (0.71–2.84)2.09 (1.13–4.13)0.007
Ovarian cancer < age 49
 Number of cases359  
 Rate2 (95% CI)5 (2–17)9 (4–22)16 (8–31)  
 RR3 (95% CI)11.65 (0.40–8.03)2.95 (0.88–13.3) 0.076
Ovarian cancer ≥ age 49
 Number of cases212329  
 Rate2 (95% CI)42 (28–65)42 (28–62)52 (36–76)  
 RR3 (95% CI)11.06 (0.58–1.92)1.29 (0.74–2.30) 0.35
Endometrial cancer
  Number of cases10243153 
  Rate2 (95% CI)1.8 (0.9–3.4)3.6 (2.4–5.3)3.8 (2.7–5.4)6.1 (4.7–8.1) 
  RR3 (95% CI)12.01 (0.98–4.42)2.29 (1.16–4.94)3.53 (1.86–7.43)0.0001

Statistical analysis

Person-years of follow-up accrued from the date of entry into the NSHDC until a first cancer diagnosis, emigration, death or the end of follow-up (31 October 2003). The end points in our analyses were overall and site-specific cancer incidences. Generally, only cancer sites with more than 60 cases were included in the site-specific analyses, with the exception of kidney cancer because of an intermediate number of cases and a consistent positive association with obesity reported in the literature.2

Standardized incidence ratios (SIRs) were calculated to compare the pattern of cancer occurrence in our study sample to that of the underlying source population in Västerbotten. The observed number of cancer cases in NSHDC was compared with the expected numbers, based on age-, gender- and calendar year-specific cancer incidence rates in Västerbotten County.12 Reproducibility of BMI measurements, taken on average about 10 years apart, was estimated by calculating Pearson correlation coefficients.

Directly standardized cancer rates across quartiles of BMI were calculated using the age distribution of the entire male or female study population as the standard population, and 95% confidence intervals (CIs) were calculated using an approximated standard error of the logarithm of the rate.13 Relative risks (RRs) and 95% CIs associated with increasing BMI were estimated using Poisson models, adjusting for age, calendar year and smoking. Age (in 5-year categories) and calendar year (before or after 1998) were analyzed as time-dependent variables, although calendar year had only a minor effect on risk estimates. Tests for linear trend were calculated using the median of each BMI category as a score and entering the score as a continuous term in the regression model. All tests were 2-sided, with an α of 0.05.

Additionally, analyses in subgroups based on smoking habits, age and lag time between recruitment and cancer diagnosis (≥1 year) were conducted. Smoking status was defined as current, ex- or nonsmoker. Current smokers were those who reported regular smoking of cigarettes, cigarillos, cigars or pipe; occasional smokers were classified as nonsmokers; and those who had stopped regular smoking were classified as ex-smokers. All analyses relating BMI to cancer risk were adjusted for smoking. No individual information about age at menopause was available, but previous studies, nested within the NSHDC, have shown the mean age at menopause in this population to be about 49–50 years; thus, age 49 was used as a cut-off point in subgroup analyses based on age in women. However, for most cancer sites, there were very few case women in the younger subgroup.

The proportion of cancer cases attributable to BMI was calculated as previously described.14, 15 For consistency with other reports,1, 16 calculations were based on multivariate-adjusted RRs obtained in analyses according to the WHO BMI categories of overweight. The population-attributable fraction was calculated only for cancer sites for which BMI was significantly associated with risk.

All statistical analyses were performed using SAS (Cary, NC), version 8.


Selected characteristics of the study population are presented in Table I. Mean age at recruitment was 46.1 years (median 49.0, range 29–61), and follow-up time was on average 8.2 years, very similar for men and women. About 50% of the study population had BMI within the normal limits, about 40% were overweight and 12% were obese according to the WHO criteria. Men tended to be more frequently overweight than women (46% vs. 31%). The distribution according to smoking status was similar in men and women, with about 20% of the participants reporting smoking at baseline (Table I). Over 90% of NSHDC participants were recruited between 1989 and 2000.

Close to 10,000 NSHDC participants (n = 9,572) had repeated weight and height measurements taken on average 10 years apart. In both women and men, there was very good agreement between the repeated BMI measurements as estimated by Pearson's correlation (r = 0.81). After 10 years of follow-up, >70% of subjects with BMI at recruitment in the extreme quartiles preserved their initial BMI classification (72% and 76% in the lowest and highest BMI quartiles, respectively). On average, women gained about 1.8 BMI unit and men about 1.4 BMI unit over 10 years. A significant trend toward an increase in mean BMI with calendar time (from 1990 to 2002) in both men and women from all ages was observed. On average, the annual increase in BMI was about 0.1 BMI units among men and 0.06 BMI units among women.

SIRs for cancer cases observed in NSHDC vs. the expected numbers according to the incidence in Västerbotten County for overall and site-specific cancers were very close to unity, giving reassurance that our study population is representative of the underlying population. SIRs for overall cancer incidence were 0.98 (95% CI 0.92–1.04) for men and 1.00 (0.95–1.05) for women. The only site-specific cancer incidence that was lower than expected was the incidence of pancreatic cancer in men (0.67, 0.42–1.00; p < 0.05).

BMI and cancer risk

Women with BMI in the top quartile (>27.1) had a 29% higher risk of developing any cancer compared to women with BMI in the lowest quartile (<22.2) (Table II). The association of BMI with cancer risk was somewhat stronger in nonsmoking women, reaching a 47% increase in risk for the top BMI quartile. Analysis conducted according to the WHO BMI categories indicated that the risk associated with body adiposity was confined to women who were obese (BMI ≥ 30), while overweight women had risk largely similar to that of normal-weight women (RR = 1.06, 95% CI 0.94–1.19, and 1.36, 1.18–1.57, for overweight and obese women, respectively) (Table III).

Table III. Risk of Cancer According to BMI in Predefined Classes in Men and Women from The NSHDC
  • 1

    Directly standardized rate/10,000 person-years.

  • 2

    RR estimated from Poisson model with likelihood ratio CIs, adjusted for age, calendar year and smoking, all subjects.

  • 3

    RR estimated from Poisson model with likelihood ratio CIs, non-smokers only.

  • 4

    Test for trend with BMI divided into 3 classes was performed using class medians as scores. With BMI in 2 classes, the effect of RR was evaluated using the likelihood ratio statistic, testing the difference in log likelihoods between models with and without BMI.

  • If the number of cancer cases available for analyses was <5 in the obese category, overweight and obese subjects were compared to normal-weight participants in the NSHDC.

All cancers
  Number of cases487631133 
  Rate1 (95% CI)45.8 (41.9–50.1)46.3 (42.8–50.1)43.3 (36.5–51.4) 
  RR2 (95% CI)11.03 (0.91–1.16)0.95 (0.78–1.15)0.79
  Number of cases23626657 
  RR3 (95% CI)10.94 (0.79–1.12)0.94 (0.70–1.25)0.55
  Number of cases699483258 
  Rate1 (95% CI)46.0 (42.7–49.6)47.9 (43.7–52.4)61.8 (54.5–70.1) 
  RR2 (95% CI)11.06 (0.94–1.19)1.36 (1.18–1.57)0.0001
  Number of cases402287164 
  RR3 (95% CI)11.05 (0.90–1.23)1.47 (1.22–1.76)0.0002
Colorectal cancer
  Number of cases456922 
  Rate1 (95% CI)4.3 (3.2–5.8)5.0 (3.9–6.3)7.1 (4.7–10.8) 
  RR2 (95% CI)11.17 (0.80–1.71)1.61 (0.95–2.65)0.08
  Number of cases433926 
  Rate1 (95% CI)2.9 (2.1–3.9)3.7 (2.7–5.1)5.6 (3.8–8.3) 
  RR2 (95% CI)11.27 (0.82–1.97)2.01 (1.22–3.27)0.005
Colon cancer
  Number of cases21439 
  Rate1 (95% CI)2.0 (1.3–3.1)3.1 (2.3–4.2)2.9 (1.5–5.6) 
  RR2 (95% CI)11.57 (0.94–2.71)1.43 (0.62–3.02)0.23
  Number of cases292720 
  Rate1 (95% CI)2.0 (1.4–2.9)2.6 (1.7–3.8)4.3 (2.8–6.7) 
  RR2 (95% CI)11.28 (0.78–2.18)2.25 (1.25–3.98)0.008
Rectal cancer
  Number of cases222313 
  Rate1 (95% CI)2.1 (1.4–3.2)1.7 (1.1–2.5)4.2 (2.4–7.2) 
  RR2 (95% CI)10.80 (0.44–1.45)1.96 (0.96–3.86)0.13
  Number of cases14125 
  Rate1 (95% CI)0.9 (0.5–1.5)1.1 (0.6–2.0)1.1 (0.4–2.6) 
  RR2 (95% CI)11.31 (0.59–2.87)1.30 (0.42–3.45)0.54
Stomach cancer
  Number of cases1630  
  Rate1 (95% CI)1.5 (0.9–2.4)1.9 (1.3–2.7)  
  RR2 (95% CI)11.36 (0.75–2.57) 0.31
  Number of cases179  
  Rate1 (95% CI)1.2 (0.7–1.9)0.6 (0.3–1.1)  
  RR2 (95% CI)10.53 (0.22–1.18) 0.12
Pancreatic cancer
  Number of cases1210  
  Rate1 (95% CI)1.2 (0.7–2.0)0.6 (0.3–1.1)  
  RR2 (95% CI)10.58 (0.24–1.34) 0.20
  Number of cases18176 
  Rate1 (95% CI)1.3 (0.8–2.0)1.6 (1.0–2.7)1.3 (0.6–3.1) 
  RR2 (95% CI)11.40 (0.71–2.74)1.20 (0.43–2.87)0.55
Urinary tract cancers
  Number of cases273710 
  Rate1 (95% CI)2.5 (1.7–3.7)2.7 (1.9–3.8)3.3 (1.8–6.3) 
  RR2 (95% CI)11.17 (0.71–1.94)1.39 (0.64–2.79)0.36
  Number of cases1167 
  Rate1 (95% CI)0.8 (0.4–1.5)0.6 (0.2–1.3)1.6 (0.7–3.4) 
  RR2 (95% CI)10.76 (0.26–2.02)2.12 (0.77–5.43)0.19
Kidney cancer
  Number of cases7117 
  Rate1 (95% CI)0.7 (0.3–1.5)0.8 (0.4–1.5)2.3 (1.1–4.9) 
  RR2 (95% CI)11.30 (0.51–3.56)3.63 (1.23–10.66)0.02
  Number of cases965 
  Rate1 (95% CI)0.6 (0.3–1.2)0.5 (0.2–1.2)1.1 (0.5–2.8) 
  RR2 (95% CI)10.92 (0.31–2.58)1.79 (0.55–5.27)0.37
Malignant melanoma
  Number of cases1529  
  Rate1 (95% CI)1.3 (0.8–2.1)1.8 (1.3–2.6)  
  RR2 (95% CI)11.44 (0.78–2.78) 0.24
  Number of cases241014 
  Rate1 (95% CI)1.5 (1.0–2.3)1.1 (0.6–2.0)4.2 (2.4–7.2) 
  RR2 (95% CI)10.74 (0.34–1.52)2.55 (1.27–4.93)0.02
Non-Hodgkin T cell (mm)
  Number of cases23265 
  Rate1 (95% CI)2.1 (1.4–3.2)1.9 (1.3–2.8)1.7 (0.7–4.1) 
  RR2 (95% CI)10.90 (0.51–1.60)0.76 (0.25–1.84)0.55
  Number of cases16145 
  Rate1 (95% CI)1.1 (0.6–1.7)1.3 (0.8–2.2)1.1 (0.4–2.6) 
  RR2 (95% CI)11.23 (0.59–2.54)1.03 (0.34–2.66)0.84
Respiratory tract cancers
  Number of cases27257 
  Rate1 (95% CI)2.6 (1.8–3.8)1.8 (1.2–2.7)2.3 (1.1–4.8) 
  RR2 (95% CI)10.80 (0.46–1.39)1.04 (0.42–2.27)0.83
  Number of cases261910 
  Rate1 (95% CI)1.8 (1.3–2.7)1.9 (1.2–2.9)2.2 (1.2–4.2) 
  RR2 (95% CI)11.14 (0.62–2.06)1.48 (0.68–3.00)0.31
Prostate cancer
  Number of cases17923844 
  Rate1 (95% CI)17.6 (15.2–20.3)17.0 (15.0–19.3)13.6 (10.1–18.2) 
  RR2 (95% CI)10.99 (0.81–1.20)0.78 (0.55–1.08)0.20
Breast cancer
  Number of cases28515772 
  Rate1 (95% CI)18.1 (16.1–20.4)16.0 (13.7–18.8)16.9 (13.4–21.5) 
  RR2 (95% CI)10.89 (0.73–1.08)1.00 (0.76–1.29)0.70
Breast cancer < age 49
  Number of cases7220  
  Rate1 (95% CI)6.4 (5.1–8.1)3.6 (2.3–5.6)  
  RR2 (95% CI)10.57 (0.34–0.92) 0.02
Breast cancer ≥ age 49
  Number of cases21314069 
  Rate1 (95% CI)22.7 (19.8–25.9)20.8 (17.6–24.6)24.0 (18.9–30.4) 
  RR2 (95% CI)10.92 (0.74–1.14)1.09 (0.83–1.43)0.70
Endometrial cancer
  Number of cases424135 
  Rate1 (95% CI)2.9 (2.2–4.0)3.9 (2.8–5.3)8.4 (6.0–11.8) 
  RR2 (95% CI)11.45 (0.93–2.24)2.93 (1.85–4.61)0.0001
Ovarian cancer
  Number of cases373617 
  Rate1 (95% CI)2.6 (1.9–3.6)3.8 (2.7–5.3)4.5 (2.8–7.4) 
  RR2 (95% CI)11.53 (0.96–2.43)1.74 (0.95–3.07)0.04
Ovarian cancer < age 49
  Number of cases89  
  Rate1 (95% CI)0.7 (0.4–1.4)1.7 (0.9–3.2)  
  RR2 (95% CI)12.36 (0.90–6.28) 0.08
Ovarian cancer ≥ age 49
  Number of cases293113 
  Rate1 (95% CI)3.8 (2.7–5.5)5.3 (3.7–7.5)5.2 (3.0–9.1) 
  RR2 (95% CI)11.43 (0.86–2.39)1.42 (0.71–2.69)0.22

Analyses according to both BMI quartiles and WHO BMI categories showed a significant adverse effect of increasing body weight on risk of endometrial, colon, colorectal and ovarian cancers (Tables II, III). Obese women had a >2.5-fold higher risk of developing melanoma (RR = 2.55, 95% CI 1.27–4.93; ptrend < 0.02), but the results according to BMI quartiles were less stable; although women in the top quartile had RR = 2.56 (1.04–7.18), the test for trend across BMI quartiles was not significant. BMI was inversely associated with breast cancer in women less than age 49 at the time of cancer diagnosis, in both quartile and WHO categories analyses. Risk of ovarian cancer with increasing BMI appeared to be stronger in women diagnosed before age 49 in analyses according to both study-specific tertiles and WHO BMI categories, but this was based on only 17 ovarian cancers in the younger subgroup.

Among men, no effect of BMI on overall cancer risk was observed in models based on quartile distribution of BMI, according to the WHO BMI classification or in analysis limited to nonsmokers (Tables II, III). The strongest association for individual cancer site was with kidney cancer (RR for the top tertile = 3.20, 95% CI 1.01–14.1; p < 0.09). In general, the results were stronger when obese men were compared to normal-weight men with RRs of 3.63 (1.23–10.7) for kidney cancer (p < 0.02) and 1.61 (0.95–2.65) for colorectal cancer (p < 0.08). Analysis in quartiles indicated a tendency for increased risk for malignant melanoma, but only a nonsignificant 44% increase in risk was observed when overweight and obese men were compared to normal-weight men (1.44, 0.78–2.78; p < 0.24).

Analyses restricted to cancer cases diagnosed at least 1 year after index measurement of body weight and height were very similar for most of the cancer sites, suggesting little effect of weight loss that may accompany cancer diagnosis on the study results (inverse causation bias). However, in the restricted analyses, obese men had significantly higher risk of developing colorectal cancer than normal-weight men (1.77, 1.04–2.95; p < 0.04).

The proportion of cancers in Västerbotten that could be attributed to overweight and obesity in women was estimated to be 7%, with a larger effect on the incidence of endometrial (30%), ovarian (22%), colon (20%) and colorectal (16%) cancers. Because excess weight entails major adverse health effects other than cancer, no attributable fraction was estimated for breast cancer occurring before age 49. In men, about 30% of all kidney cancer cases could be attributed to overweight and obesity.


In this population-based prospective cohort study, we investigated the association of BMI with overall and site-specific cancer risks in women and men. In women, a positive association between BMI and overall cancer risk emerged, mainly driven by the strong effects of elevated BMI on the incidence of endometrial, ovarian and colon cancers as well as melanoma. In men, BMI was not related to overall cancer risk. The lack of association was probably driven by the results for prostate and respiratory tract cancers, which accounted for >40% of all malignancies. As in our data, most epidemiologic studies suggest little,17, 18 if any, effect of increased body adiposity on risk of developing prostate and respiratory tract cancers.1, 2, 19 A greater adverse effect of elevated BMI on overall cancer incidence and/or mortality in women than in men has been previously reported in other studies from Western countries.16, 20

As demonstrated by numerous epidemiologic studies,1, 2 the strongest adverse effect of elevated BMI on individual cancer site was observed for endometrial cancer. Only 10 women (8%) had BMI in the lowest quartile, and although based on small numbers, analysis across quartiles indicated that risk starts to increase within the range of BMI considered normal: women with BMI of 22.2–24.2 kg/m2 had about 2-fold higher risk (RR = 2.01, 95% CI 0.98–4.42) compared to women with BMI of 18.5–22.1. This remarkably strong adverse effect of BMI on risk of endometrial cancer is believed to be due to alterations in the synthesis of sex steroids caused by increased body adiposity:21 both obesity and endometrial cancer risk have been associated with decreased synthesis of progesterone in premenopausal women and with increased circulating estrogens after menopause.1, 5, 21, 22 Another major metabolic consequence of obesity that may influence risk of endometrial cancer, both before and after menopause, is development of insulin resistance and hyperinsulinemia.2, 23

The effects of obesity on sex-steroid metabolism are also believed to underlie the differential effect of increasing body weight on risk of breast and possibly ovarian cancer before and after menopause.24, 25 Similar to other studies,26 an inverse association of BMI with breast cancer was observed among women diagnosed before age 49. The lack of effect of BMI on risk of breast cancer in older women may be due to failure to adjust for use of exogenous hormones and other reproductive characteristics, as observed in other studies.27

There was about a 2-fold increase in risk of ovarian cancer over quartiles of BMI, which is consistent with the adverse effect observed in several large epidemiologic studies.17, 20, 28 It has also been shown that high BMI during adolescence confers an increased risk of ovarian cancer,29, 30, 31, 32 while weight gain throughout adulthood has not been related to risk.29, 30, 32, 33, 34, 35 There were only 17 ovarian cancer cases diagnosed before age 49 in our study; thus, we did not have sufficient statistical power to address a differential effect of elevated BMI by age at cancer diagnosis. Nevertheless, our study suggests that the association may be stronger in younger women, as also reported by a Danish record linkage study.36

Among cancers that affect both men and women, the overwhelming majority of epidemiologic studies have reported a direct association of BMI with colon and colorectal cancer.1 In most studies, obesity has been associated with a stronger adverse effect in men (about 50–100% increase in risk) than in women (about 20–50% increased risk).1, 2, 37 However, several large studies have indicated that BMI during women's fertile life is associated with about a 2-fold increase in colorectal cancer risk, while the adverse effect is substantially weaker after menopause.38, 39, 40 The 1.5- to 2-fold higher risk of colorectal cancer associated with obesity observed in men and women from our study, half of whom were recruited before age 49, is consistent with these previous observations. Screening for colorectal cancer (by either fecal occult blood test or colonoscopy) is not ongoing in Västerbotten, nor is it recommended during the counseling session when subjects are recruited in the NSHDC; thus, the association of BMI with colorectal cancer observed in our study was not influenced by the existence of an established screening program.

Kidney cancer, and specifically renal-cell cancer, has been consistently associated with obesity.1, 2 Although a small number of kidney cancers were available for our study (25 in men, 20 in women), obese men had a significant >3-fold higher risk of kidney cancer compared to normal-weight men.

Obese women had about a 2.5-fold higher risk of developing malignant melanoma than normal-weight women. Analyses across BMI quartiles in both men and women also indicated a higher risk of melanoma with increasing BMI, although the results did not achieve statistical significance. Several other studies have reported elevated melanoma risk in obese men and women,18, 41, 42 but generally the epidemiologic evidence lacks consistency and there is no established mechanism that may link obesity with melanoma.

Major strengths of our study are its population-based prospective design, minimizing selection bias, and the reliance on national and regional person and cancer registers in Sweden, renowned for their quality and completeness.10, 11 Overall and site-specific cancer incidence rates in the NSHDC were very similar to those in the underlying county population. Another major strength of the study is the availability of height and weight measurements, taken in a standardized fashion by trained nurses for close to 70,000 men and women, virtually eliminating the possibility of reporting bias. More than 70% of subjects in extreme BMI quartiles preserved their initial quartile classification after an average period of 10 years, indicating that, although not ideal, the baseline measurements provided a good indicator of the long-term BMI of cohort members. The prevalence of obesity among NSHDC participants was 12% on average, substantially lower than the prevalence reported in the United States but very close to that previously reported in Swedish populations of similar age.43, 44 However, the small percentage of cohort members with BMI >40 (0.3% men, 0.7% women) precluded analyses across a wider BMI range and possibly attenuated the risk estimates, especially in men.

A weakness of the study is that we could not control for diabetes and, in women, for reproductive characteristics and exogenous hormone use. Overweight and obesity are associated with development of diabetes, which in turn has been independently associated with elevated cancer risk, specifically of endometrial and possibly colon cancers.6, 12 Failure to adjust for diabetes could have inflated our risk estimates for these 2 cancer sites. In contrast, not controlling for use of exogenous hormones for menopausal symptoms could have limited our ability to detect an association of high BMI with postmenopausal breast cancer. Another concern is the possibility of residual negative confounding by smoking in our analyses. Adjustment for smoking (current, ex- and nonsmoker) was included in all analyses, but we lacked sufficient detail to adjust for smoking intensity. In women, the association of BMI with cancer risk was stronger in analyses limited to nonsmokers, as expected; but in men, the estimates remained virtually unchanged.

Despite an average follow-up of 8.2 years of almost 70,000 members of the NSHDC, during which a total of 2,691 new cancers were diagnosed, a relatively limited number of site-specific cancer cases were available for analyses, resulting in unstable risk estimates and precluding more detailed analyses according to cancer subtypes. We also could not examine the association of BMI with relatively infrequent malignancies for which obesity has already been implicated as a risk factor, e.g., esophageal and gallbladder cancers.

As members of the NSHDC are representative of the underlying population of Västerbotten, we estimated the proportion of cancers that could be avoided if all women and men in the county maintained BMI within the WHO-recommended range (18.5–25.0). In women, up to 7% of all cancers in the county are attributable to overweight and obesity, with a larger effect on endometrial (up to 30%), colon (20%) and ovarian (22%) cancers. Up to 18% of melanoma cases among women and 30% of kidney cancers in men could also be attributed to BMI above the normal range.

In conclusion, the results from the population-based NSHDC provide further evidence that BMI is positively associated with risk of developing cancer. The NSHDC data also confirmed the tendency for increase in BMI with calendar time observed in Sweden and other countries.1, 45, 46 Thus, the fraction of cancer and other diseases attributable to obesity is expected to increase in the future. Development and implementation of efficient public health policies and initiatives for promotion of lifelong healthy eating and physical activity are urgently needed, together with research targeted at understanding the precise mechanisms that link obesity with cancer.