The authors have declared that no competing interests exist.
Conceived and designed the experiments: SU AOO KA. Performed the experiments: SU AOO. Analyzed the data: SU AOO AJ. Contributed reagents/materials/analysis tools: WPK KA. Wrote the paper: SU AOO. Consultant on experimental design: WPK. Consultant on experimental design and outcome: JMY.
Few studies have examined the association between body mass index (BMI: kg/m2) and pancreatic cancer risk in Asian populations. We examined this relationship in 51,251 Chinese men and women aged 45–74 who enrolled between 1993 and 1998 in the population based, prospective Singapore Chinese Health Study. Data were collected through in-person interviews. By December 31, 2011, 194 cohort participants had developed pancreatic cancer. A Cox proportional hazards model was used to estimate hazard ratios (HR) and their 95% confidence intervals (95% CI). We hypothesized the association between BMI and pancreatic cancer risk may vary by smoking status (ever v. never) and there was evidence for this as the interaction between BMI and smoking status was significant (p = 0.018). Among ever smokers, being classified as underweight (BMI <18.5 kg/m2), was associated with a significantly elevated risk of pancreatic cancer relative to smokers with a BMI of 21.5–24.4 kg/m2 (HR = 1.99, 95% CI = 1.03–3.84). This association was strengthened after exclusion of the first three years of follow-up time. Among never smokers, there was no association between BMI and pancreatic cancer risk. However, after excluding pancreatic cancer cases and person-years in the first three years of follow-up, never smokers with a BMI ≥ 27.5 kg/m2 showed a suggestive increased risk of pancreatic cancer relative to never smokers with a BMI of 21.5–24.4 kg/m2 (HR = 1.75, 95% CI = 0.93–3.3). In conclusion, Singaporean Chinese who were underweight with a history of smoking had an increased risk of developing pancreatic cancer, whereas there was no significant association between BMI and pancreatic cancer in never smokers.
Pancreatic cancer kills more than 250,000 people each year worldwide. The malignancy is very aggressive; the mortality/morbidity ratio is 0.99 to 1
Despite a large number of studies on the etiology of pancreatic cancer, few consistent risk factors have been identified
The generalizability of the Western findings to Asian populations is uncertain
The institutional review boards at the National University of Singapore, the University of Minnesota, and the University of Pittsburgh approved this study. Written informed consent was obtained from all participants.
The design of the Singapore Chinese Health Study has been previously described
For this study we excluded 1936 subjects of the original 63,257 participants with a history of invasive cancer at baseline (except non-melanoma skin cancer or superficial, papillary bladder cancer), because they did not meet study inclusion criteria and we excluded an additional 10,070 participants who were missing measures of either, or both, height and weight. Hence, the present study included 51, 251 participants. Participants excluded due to missing BMI (N = 10,070) were not materially different from participants included in the analysis in the distributions of noted demographic and lifestyle characteristics.
Self-reported height and weight were collected at the baseline interview. BMI was calculated as weight (kg) divided by height squared (m2). Self-report of body weight has been shown to be highly valid across many populations, as well as specifically in Asians
A semi-quantitative food-frequency questionnaire specifically developed for this population to assess 165 commonly consumed food items was administered during the baseline interview assessing usual dietary intake of the previous year. The questionnaire has subsequently been validated against a series of 24-hour dietary recall interviews
Identification of incident pancreatic cancer cases and deaths among cohort members was accomplished by record-linkage analysis of the cohort database with respective databases from the population-based Singapore Cancer Registry and the Singapore Registry of Births and Deaths. As of December 31, 2011, 194 members of this study had developed pancreatic cancer. The nationwide cancer registry has been in place since 1968 and has been shown to be comprehensive in its recording of cancer cases
Study participants were initially grouped into eight categories of BMI, as reported at the baseline interview (<18.5 kg/m2, 18.5–19.9 kg/m2, 20.0–21.4 kg/m2, 21.5–22.9 kg/m2, 23.0–24.4 kg/m2, 24.5–25.9 kg/m2, 26.0–27.4 kg/m2, ≥27.5 kg/m2). These categories were created to allow for a detailed examination of the association between BMI and pancreatic cancer based on the distribution of BMI in the study population, with the consideration of BMI cut points recommended by the World Health Organization (WHO) working group for Asian populations (BMI <18.5 kg/m2 = underweight, 18.5–22.9 kg/m2 = normal weight, 23.0–27.4 kg/m2 = overweight, ≥27.5 kg/m2 = obese)
Proportional hazards (Cox) regression methods were used to estimate multivariate-adjusted hazard ratios (HR), together with the corresponding 95% confidence intervals (95% CI), and p-values of pancreatic cancer incidence by BMI group. In the primary Cox regression model, age was categorized as years at the baseline examination (<50 years, 50–54 years, 55–59 years, 60–64 years, and ≥65 years). Other variables included were sex, year of interview (1993–95 and 1996–98), dialect (Hokkien vs. Cantonese), level of education (no formal schooling, primary school, secondary school or above), baseline physician diagnosed diabetes mellitus (no, yes), smoking (no, yes), alcohol intake (none, light-moderate, heavy drinkers), vigorous work or strenuous physical activity (≥1.5 hours/week vs. <1.5 hours/week), moderate physical activity (≥2 hours/week vs. <2 hours/week), sleep (<6 hours or ≥9 hours per night vs. 6–8 hours per night), and a vegetable, fruit, and soy-rich dietary pattern score. For each subject, person-years were counted from the date of baseline interview to the date of cancer diagnosis, date of death, date of last contact or December 31, 2011, whichever occurred first.
We tested for a potential quadratic association since the data suggested that this may be a possibility. In order to test whether or not the BMI-cancer relationship was linear or quadratic (J- or U-shaped curve), linear and quadratic terms with values corresponding to the median value for each BMI category were included in the model and statistical significance evaluated using the Wald chi-square test. Because the results provided no evidence of a quadratic association, only the p-linear trend values are included in the results. The proportionality assumption was met, as indicated by the lack of significant interaction between BMI as a function of survival time in the model.
To reduce potential bias due to pre-clinical disease or cancer, illness-related weight loss, as well as accounting for the anorectic effects of smoking at the time of the baseline interview, we performed analyses with exclusion of cases diagnosed in the first three years of follow up
N | 3,850 | 13,705 | 17,730 | 10,650 | 5,316 |
Age (SD) | 57.2 (8.2) | 55.9 (8.0) | 55.7 (7.8) | 55.9 (7.7) | 55.8 (7.9) |
Sex (% Women) | 51.8 | 54.2 | 54.0 | 53.5 | 59.6 |
Education (% Secondary) | 29.7 | 34.0 | 33.2 | 29.8 | 25.3 |
Diabetes Mellitus (%) | 4.1 | 6.3 | 9.0 | 11.2 | 14.0 |
Ever smoked (%) | 41.3 | 32.5 | 28.8 | 27.6 | 26.6 |
Alcohol (% Daily) | 5.8 | 4.3 | 3.1 | 2.5 | 2.3 |
18.9 | 20.2 | 20.0 | 20.1 | 20.0 | |
Energy intake (kcal) | 1542 (546) | 1586 (562) | 1582 (555) | 1581 (580) | 1564 (613) |
Sleep (hours) | 7.0 (1.2) | 7.0 (1.1) | 7.0 (1.1) | 7.0 (1.2) | 7.0 (1.1) |
Moderate activity (hr/week) | 0.8 (2.5) | 0.8 (2.6) | 1.0 (2.9) | 1.0 (3.0) | 0.8 (2.6) |
Vigorous/Strenuous Activity | 0.7 (3.2) | 0.8 (3.4) | 0.8 (3.4) | 0.9 (3.6) | 0.7 (3.3) |
Dietary pattern (top 20%) = Dietary pattern score of greatest conformity in the population to a diet rich in vegetables, fruit and soy foods.
Data for age represent mean (SD).
BMI (kg/m2) | <18.5 | 18.5–21.4 | 21.5–24.4 | 24.5–27.4 | ≥ 27.5 | p-trend (linear) |
No. cases/N | 23/3850 | 55/13,705 | 53/17,730 | 47/10,650 | 16/5316 | |
41.9 | 27.07 | 19.97 | 29.59 | 20.44 | 0.08 | |
HR (95% CI) | 1.89 (1.15–3.09) | 1.34 (0.92–1.96) | 1.0 | 1.46 (0.99–2.17) | 1.02 (0.58–1.79) | |
No. cases/N | 8/2260 | 31/9255 | 30/12,632 | 31/7716 | 15/3903 | |
23.52 | 21.83 | 15.47 | 26.31 | 25.62 | 0.58 | |
HR (95% CI) | 1.52 (0.70–3.33) | 1.47 (0.89–2.43) | 1.0 | 1.64 (0.99–2.71) | 1.57 (0.84–2.92) | |
No. cases/N | 15/1590 | 24/4450 | 23/5098 | 17/4347 | 0.012 | |
72.06 | 39.26 | 32.19 | 27.98 | |||
HR (95% CI) | 1.99 (1.03–3.84) | 1.20 (0.68–2.13) | 1.0 | 0.91 (0.48–1.70) |
SCHS = Singapore Chinese Health Study.
Standardized rate = Age and sex standardized cancer rate per 100,000 person years using person year time, age & sex distributions of SCHS.
HR (95% CI) = Hazard Ratio; 95% confidence interval: Model adjusted for age, sex, year of enrollment, dialect, education, diabetes status,
smoking (in whole population analysis), age of initiation of smoking habits, number of cigarettes per day, years of smoking, alcohol intake, dietary pattern score,
physical activity, sleep and energy intake.
BMI categories combined for ever-smokers into ≥ 24.5 kg/m2.
Because potential underlying disease or poor health of the subjects at baseline might have an impact on the risk of developing pancreatic cancer, we excluded pancreatic cancer cases and person-years that occurred within the first three years post-enrollment (
BMI (kg/m2) | <18.5 | 18.5–21.4 | 21.5–24.4 | 24.5–27.4 | ≥ 27.5 | p-trend (linear) |
No. cases/N | 20/3701 | 45/13,364 | 48/17,367 | 38/10,408 | 16/5193 | |
36.63 | 22.21 | 18.13 | 23.98 | 20.49 | 0.25 | |
HR (95% CI) | 1.83 (1.08–3.09) | 1.21 (0.81–1.82) | 1.0 | 1.31 (0.85–2.0) | 1.13 (0.64–2.0) | |
No. cases/N | 5/2205 | 26/9109 | 27/12,457 | 25/7591 | 15/3831 | |
14.74 | 18.33 | 13.94 | 21.26 | 25.67 | 0.22 | |
HR (95% CI) | 1.06 (0.41–2.76) | 1.36 (0.79–2.33) | 1.0 | 1.47 (0.85–2.54) | 1.75 (0.93–3.3) | |
No. cases/N | 15/1496 | 19/4255 | 21/4910 | 14/4175 | 0.007 | |
72.58 | 31.25 | 29.52 | 23.15 | |||
HR (95% CI) | 2.25 (1.15–4.4) | 1.05 (0.56–1.96) | 1.0 | 0.82 (0.41–1.61) |
SCHS = Singapore Chinese Health Study.
Standardized rate = Age and sex standardized cancer rate per 100,000 person years using person year time, age & sex distributions of SCHS.
HR (95% CI) = Hazard Ratio; 95% confidence interval: Model adjusted for age, sex, year of enrollment, dialect, education, diabetes status,
smoking (in whole population analysis), age of initiation of smoking habits, number of cigarettes per day, years of smoking, alcohol intake, dietary pattern score,
physical activity, sleep and energy intake.
BMI categories combined for ever-smokers into ≥ 24.5 kg/m2.
We observed a differential association between BMI and incidence of pancreatic cancer in never vs. ever smoking Chinese men and women in Singapore. In never smokers there was no evidence of any association between BMI and incident pancreatic cancer until excluding participants with potential underlying disease or poor health, when a suggestive association between obesity and incident pancreatic cancer appeared. On the other hand, underweight ever smokers with a BMI <18.5 kg/m2 were at an increased risk for pancreatic cancer even after accounting for potential underlying disease and poor health. A number of lifestyle and behavioral factors were evaluated as covariates and did not alter the risk of pancreatic cancer observed for BMI and smoking.
Previous reviews have concluded that an obese BMI is associated with increased pancreatic cancer risk
With respect to Asian populations, the association between BMI and pancreatic cancer has been inconsistent in the published data. Results from the Japan Collaborative Cohort pooled study
A review of 21 prospective studies with a total of 8,062 pancreatic cancer cases revealed that all studies adjusted for, but did not stratify by smoking
Some inconsistencies in the literature might be due to variations in how BMI was modeled across studies and to the small number of cases in a number of studies. Additional differences between our findings and those of others could be due to either different BMI criteria being used to define obesity (e.g., Western population cut points), different adjustments of covariates, different reference categories, different classes of exposure (e.g., study- and specific quartiles) and/or different study populations. Further investigations are needed to clarify how BMI across the life span is associated with pancreatic cancer in Asian populations.
Insulin regulation may be the mechanism linking obesity and pancreatic cancer. A hyperinsulinemic state, leading to an increased bioavailability of insulin-like growth factor-I, could stimulate cell proliferation and lead to tumorigenesis
related DNA adducts and DNA damage
Additional studies have shown that smokers typically have a lower BMI and body weight than do non-smokers
Limitations include the use of self-reported height, weight and other demographic and lifestyle data. Although the use of self-reported body weight and height could be prone to non-differential misclassification and thus lead to under-estimation of the BMI-pancreatic cancer risk association, self-report of body weight has been shown to be highly valid across many populations
Strengths of our study include the assumption that case ascertainment was complete, given that Singapore is a small city-state where there is thorough specialized medical care available and the nationwide cancer registry has been in place since 1968 and shown to be comprehensive in its recording of cases
In summary, we observed an increased risk of pancreatic cancer in ever smokers with a BMI <18.5 kg/m2, and a suggestive increased risk in never smokers with a BMI ≥ 27.5 kg/m2 in analyses that accounted for potential biases and confounding factors. Thus, the data from this study underscore the consideration of smoking status, one of the few recognized risk factors for pancreatic cancer, when examining BMI in relation to pancreatic cancer risk. Continued thorough investigation of this topic in varied populations will further increase the understanding of the relationship between BMI and pancreatic cancer risk.
We would also like to thank Siew-Hong Low of the National University of Singapore for supervising the field work of the Singapore Chinese Health Study and Renwei Wang for maintenance of the cohort study database. Finally, we acknowledge the founding, long-standing Principal Investigator of the Singapore Chinese Health Study – Mimi C. Yu.