br z Adjusted for variables in b previous history of

z Adjusted for variables in “b” þ previous history of diabetes (yes, no). x Adjusted for variables in “b” þ waist circumference. k Adjusted for variables in “b” þ waist to hip ratio.
Analyses were performed using Stata 14.1 (StataCorp, College Station, TX). All P-values are two-sided.
Results
Compared with noncases, women subsequently diagnosed with pancreatic cancer were older at enrollment and had significantly greater mean WC, WHR, WHtR, alcohol intake, and pack-years of smoking (Table 1). The proportion of women with a history of diabetes was nonsignificantly higher in cases compared with noncases (6.8 vs. 5.9 percent, P ¼ 0.44).
In multivariable-adjusted analyses (excluding diabetes as a covariate), compared with the lowest quartile, the highest quartile of BMI was positively associated with risk of pancreatic cancer (HR 1.25, 95% CI 1.00e1.56, P for trend ¼ 0.016) (Table 2). Women in the highest quartile of WC, WHR, and WHtR also had increased 
included in the models, the corresponding HRs were slightly
attenuated. When the exposures of interest were examined as continuous variables, the HR per 5 kg/m2 increment in BMI was
There was no evidence to suggest that smoking modifies the associations between the metabolic factors and risk of pancreatic cancer (Table 3).
When the first 3 years of follow-up were excluded, all 3 anthropometric variables remained positively associated with risk of pancreatic cancer, but, the association with diabetes was no longer apparent (Table S1).
Table 3
Association of diabetes and measured anthropometric variables with risk of pancreatic cancer in the Women's Health Initiative by smoking status
Exposures
Smoking
No
Yes
Cases/noncases HR 95% CI
Cases/noncases HR 95% CI
Diabetes
P for heterogeneity
Per 5 kg increase
P for heterogeneity
Waist circumference (cm)
Per 10 cm increase
P for trend
P for heterogeneity
Waist-hip ratio
Per 0.1 unit increase
P for trend
P for heterogeneity
Waist-height ratio
Per 0.1 unit increase
P for trend
Adjusted for age (continuous), alcohol intake (servings/day), previous history of diabetes (yes, no), metabolic equivalent task hrs/week (MET-hrs/weekecontinuous), educational level (less than high school grad, high school grad/some college, college grad, post-college), Veratridine (white, black, other), and allocation to the OS or treatment/ placebo/control arm of clinical trials unless included as the main exposure.
Table 4
Association of self-reported BMI at ages 18, 35, and 50 years with risk of pancreatic cancer in the Women's Health Initiative observational study
Per 5 kg increase
Per 5 kg increase
Per 5 kg increase
Models examining BMI at earlier ages were limited to the observational study (377 participants were missing information on BMI at earlier times.). * Adjusted for age only.
y Adjusted for age (continuous), smoking status (never smoked, former smoker, current smoker), alcohol intake (servings/day), previous history of diabetes (yes, no), metabolic equivalent task hrs/week (MET-hrs/weekdcontinuous), lost more than 10 lbs. (yes, no), educational level (less than high school grad, high school grad/some college, college grad, post-college), race (white, black, other).
z Adjusted for variables in “b” þ previous history of diabetes (yes, no).
Discussion
The results of this study suggest that a history of diabetes, overall obesity, and, in particular, central adiposity measured at enrollment have modest positive associations with risk of pancre-atic cancer. The associations with measured anthropometric factors persisted in sensitivity analyses, whereas the association with diabetes was somewhat attenuated when the early years of follow-up were excluded. Except for a positive association between self-reported BMI at age 50 years and risk of pancreatic cancer, self-reported BMI at earlier points in life showed no association with risk.
In the present study, the association between diabetes and pancreatic cancer appeared to be explained partly by level of cen-tral adiposity, as adjusting for WC significantly attenuated the HR for the association between diabetes and pancreatic cancer. Nevertheless, as in several previous studies, diabetes was observed to have a modest positive association with risk of pancreatic cancer independent of level of adiposity and other confounders [21e29]. Pooled results from a recent meta-analyses also showed that dia-betes was associated with an almost two-fold increased risk of pancreatic cancer (RR: 1.97, 95% CI: 1.78e2.18) [30]. A recent um-brella review of meta-analyses of diabetes and cancer [15] judged the evidence regarding pancreatic cancer to be uncertain because of heterogeneity between studies and raised the possibility of biases, including publication bias and selective reporting of positive results.