Abstract
Purpose
Although obesity is associated with increased incidence of pancreatic cancer, studies have not prospectively evaluated prediagnostic body mass index (BMI) and survival.
Patients and Methods
We analyzed survival by prediagnostic BMI assessed in 1986 among 902 patients from two large prospective cohorts diagnosed from 1988 to 2010. We estimated hazard ratios (HRs) for death using Cox proportional hazards models, with adjustment for age, sex, race/ethnicity, smoking, diagnosis year, and stage. We evaluated the temporal association of BMI with survival by grouping reported BMI by 2-year lag-time intervals before diagnosis.
Results
The multivariable-adjusted HR for death was 1.53 (95% CI, 1.11 to 2.09) comparing patients with BMI ≥ 35 kg/m2 with those with BMI < 25 kg/m2 (P trend = .001), which was similar after adjustment for stage. The association of BMI with survival was stronger with longer lag times between reported BMI and cancer diagnosis. Among patients with BMI collected 18 to 20 years before diagnosis, HR for death was 2.31 (95% CI, 1.48 to 3.61; P trend < .001), comparing obese with healthy-weight patients. No statistically significant differences were seen by cohort, smoking status, or stage, although the association was stronger among never-smokers (HR, 1.61; 95% CI, 1.01 to 2.57; P trend = .002) than ever-smokers (HR, 1.36; 95% CI, 0.86 to 2.15; P trend = .63), comparing BMI ≥ 35 kg/m2 with BMI < 25 kg/m2. Higher prediagnostic BMI was associated with more advanced stage at diagnosis, with 72.5% of obese patients presenting with metastatic disease versus 59.4% of healthy-weight patients (P = .02).
Conclusion
Higher prediagnostic BMI was associated with statistically significantly decreased survival among patients with pancreatic cancer from two large prospective cohorts.
INTRODUCTION
Pancreatic cancer is the fourth-leading cause of cancer-related death in the United States.1 Among patients who develop pancreatic adenocarcinoma, only 5% will survive 5 years after diagnosis, and most patients live < 12 months. The length of patient survival is greatly influenced by disease stage at presentation, but few other markers of survival have been well characterized.2
Multiple studies have demonstrated that high body mass index (BMI) is associated with increased risk of developing pancreatic cancer.3–5 Subsequent work has suggested that alterations in systemic metabolism as a consequence of peripheral insulin resistance may underlie the association of excess weight with disease risk.6–9 Additionally, chronic inflammation that can be induced by obesity may be related to the observed associations.10,11 However, few studies have evaluated the impact of BMI on survival of patients with pancreatic cancer, and no studies have used prospectively collected data on BMI.
Early manifestations of subclinical pancreatic cancer include weight loss and hyperglycemia, and these symptoms are commonly seen 12 to 24 months before the clinical diagnosis of pancreatic cancer.12,13 Therefore, measurement of BMI at the time of diagnosis does not provide accurate representation of a patient's long-term exposure to obesity and insulin resistance. Furthermore, retrospective studies are subject to biases related to patient recall and proxy respondents. Prospective cohort studies provide a unique opportunity to evaluate repeated BMI evaluations collected over many years before the diagnosis of pancreatic cancer. In the current study, we prospectively evaluated the association of prediagnostic BMI with survival among participants diagnosed with pancreatic cancer from two large prospective cohort studies.
PATIENTS AND METHODS
Study Population
We assessed the association of prediagnostic BMI with survival among participants with pancreatic cancer from the Nurses' Health Study (NHS) and Health Professionals Follow-Up Study (HPFS). NHS was initiated in 1976 when 121,700 female registered nurses age 30 to 55 years returned a mailed questionnaire describing demographics, lifestyle choices, and medical history.14 HPFS began in 1986 when 51,529 men age 40 to 75 years working as dentists, veterinarians, pharmacists, optometrists, osteopathic physicians, or podiatrists returned a mailed questionnaire on health-related behaviors and medical history.15 Participants have updated information on exposures and medical history through biennial follow-up questionnaires. The studies are approved by the Human Research Committee at Brigham and Women's Hospital (Boston, MA), and participants provided informed consent.
Exposure Data
The primary exposure of baseline prediagnostic BMI (weight [kg] divided by square of height [m]) was calculated from weight reported on the 1986 questionnaire in NHS and HPFS and height reported on the 1976 questionnaire in NHS and 1986 questionnaire in HPFS. For participants in NHS without weight available on the 1986 questionnaire, we carried forward the weight from 1980 to 1984, if available. For secondary analyses, we used weight reported on biennial follow-up questionnaires to calculate BMI at additional time points before diagnosis. BMI was categorized using WHO criteria for healthy weight (18.5 to 24.9 kg/m2), overweight (25.0 to 29.9 kg/m2), and obese (≥ 30 kg/m2). For some analyses, sample size was adequate to further classify the obese group into two categories: BMI 30.0 to 34.9 kg/m2 and BMI ≥ 35 kg/m2. In a validation subsample, questionnaire-reported and staff-measured weights were highly correlated (r = 0.96; mean difference, 3.3 lbs).16
Detailed covariate data, including age, race/ethnicity (white, black, other, unknown), smoking status (never, past, current, missing), multivitamin use (yes, no), physical activity (MET-hr/wk), alcohol intake (g/d), energy intake (kcal/d), and history of diabetes mellitus (yes, no), were obtained from 1986 and subsequent questionnaires. Date and pancreatic cancer stage at diagnosis were obtained from medical record review. Cancer stage was classified as: local disease amenable to surgical resection; locally advanced disease with extrapancreatic extension rendering it unresectable, but without distant metastases; distant metastatic disease; or unknown.
Patients With Pancreatic Cancer
Incident patient cases of pancreatic cancer were identified through May 31, 2010, in NHS and January 31, 2010, in HPFS by self-report or during follow-up of a participant's death. Deaths were ascertained from next of kin or postal service and by searching the National Death Index; this method has been shown to capture > 98% of deaths.17 Diagnoses were confirmed by review of medical records, death certificates, and/or tumor registry data. Those with nonadenocarcinoma histology were excluded.
During 24 years of follow-up, we identified 958 patient cases of pancreatic cancer in NHS and HPFS with available BMI from the baseline questionnaires. Fifty-eight participants who developed pancreatic cancer did not report BMI on baseline questionnaires and therefore were not included in the analysis. Because pancreatic cancer can lead to significant weight loss and metabolic derangements in 12 to 24 months before diagnosis,9,10 we excluded 48 participants diagnosed with pancreatic cancer within 2 years of return of the 1986 questionnaire. We also excluded eight participants with BMI < 18.5 kg/m2 because of concerns regarding coexistent illness that might limit survival. Therefore, the current analysis included 902 patients with pancreatic cancer, including 561 participants from NHS and 341 from HPFS.
Statistical Analysis
We examined the association of prediagnostic BMI with survival using Cox proportional hazards regression to calculate hazard ratios (HRs) and 95% CIs for mortality by category of BMI. Overall survival time was calculated from pancreatic cancer diagnosis to death or May 31, 2010, in NHS and to death or January 31, 2010, in HPFS, whichever came first. Two-sided tests for trend across BMI categories were calculated by entering median values from each BMI category into Cox models as continuous variables. The proportionality of hazards assumption was satisfied by evaluating a time-dependent variable, which was the cross-product of BMI and time (P = .73). Survival curves were generated using the Kaplan-Meier method, and statistical significance was measured using the log-rank test.
A priori, we included in multivariable-adjusted models the covariates with the greatest likelihood for confounding, including age at diagnosis, race/ethnicity, sex, smoking status, and cancer stage. We also included the time period of diagnosis to account for changes in survival over time. We evaluated physical activity, alcohol intake, and multivitamin use as additional possible confounders, but none were associated with survival (all P > .1). History of diabetes mellitus was not included in our multivariate model, because it may lie within the causal pathway of BMI and survival. However, in additional analyses, our results were similar after inclusion of diabetes as a covariate in our multivariate model and after removal of patients with diabetes at baseline or who developed diabetes before diagnosis. Given a prior study suggesting an association between metabolic syndrome and pancreatic cancer stage at diagnosis,18 we created two multivariable-adjusted models with and without stage to investigate whether adjustment for stage might alter an association between BMI and pancreatic cancer survival. We noted similar results in a sensitivity analysis in which we removed patients with unknown stage.
The aim of this analysis was to characterize the association between pancreatic cancer survival and chronic overweight/obesity, before weight loss that often accompanies subclinical or newly diagnosed pancreatic malignancy. Therefore, primary analyses evaluated baseline BMI in 1986 without cumulative updating during follow-up. We also excluded patients diagnosed within 2 years of responding to the 1986 questionnaire to achieve a minimum 2-year lag time between assignment of exposure and diagnosis. We subsequently performed two secondary analyses using full available data for BMI collected from participants every 2 years. First, we grouped reported BMI by 2-year time intervals before cancer diagnosis and evaluated whether the association of BMI with pancreatic cancer survival differed by lag time between BMI measurement and cancer diagnosis. Second, we investigated a cumulative average BMI for each participant, which was average BMI weighted by exposure time from 1986 until the most recent questionnaire returned > 2 years before diagnosis.
We assessed heterogeneity in the association between BMI and pancreatic cancer survival in NHS and HPFS using Cochran's Q statistic19; the Cochran's Q statistic P value was .79 for the comparison of BMI ≥ 30 kg/m2 with BMI < 25 kg/m2 across the cohorts. We also evaluated stratified analyses by cohort (also stratified by sex), smoking status, disease stage, and diagnosis period. We assessed statistical interaction by entering main effect terms and a cross-product term of continuous BMI and stratification variable into the model, evaluating likelihood ratio tests. All statistical analyses were performed using SAS 9.3 statistical package (SAS Institute, Cary, NC), and all P values are two sided.
RESULTS
Baseline characteristics of 902 patients with pancreatic cancer are listed in Table 1. Mean age at baseline BMI measurement was 57.8 years, and median time between baseline BMI and pancreatic cancer diagnosis was 14.7 years. Mean baseline BMI was 25.9 kg/m2, with 36% of patients classified as overweight and 15% as obese. Among those with known disease stage, 19.7% had localized disease, 15.3% had locally advanced disease, and 65.0% had metastatic disease. Median survival by cancer stage was 16 months for those with localized disease, 8 months for those with locally advanced disease, and 3 months for those with metastatic disease. At the end of follow-up, 868 patients (96%) were deceased.
Table 1.
Baseline Characteristics of Patients With Pancreatic Cancer
| Characteristic | NHS (n = 561) |
HPFS (n = 341) |
Total (N = 902) |
|||
|---|---|---|---|---|---|---|
| No. | % | No. | % | No. | % | |
| Age at baseline BMI, years | ||||||
| Mean | 56.5 | 60.1 | 57.8 | |||
| SD | 6.2 | 8.3 | 7.3 | |||
| Age at cancer diagnosis, years | ||||||
| Mean | 71.2 | 72.8 | 71.8 | |||
| SD | 7.6 | 8.7 | 8.1 | |||
| Race/ethnicity | ||||||
| White | 547 | 97.5 | 308 | 90.3 | 855 | 94.8 |
| Black | 6 | 1.1 | 7 | 2.1 | 13 | 1.4 |
| Other | 8 | 1.4 | 11 | 3.2 | 19 | 2.1 |
| Unknown | 0 | 0.0 | 15 | 4.4 | 15 | 1.7 |
| BMI, kg/m2 | ||||||
| Mean | 26.1 | 25.6 | 25.9 | |||
| SD | 5.2 | 3.1 | 4.5 | |||
| < 25 | 290 | 51.7 | 153 | 44.9 | 443 | 49.1 |
| 25 to 29.9 | 162 | 28.9 | 161 | 47.2 | 323 | 35.8 |
| ≥ 30 | 109 | 19.4 | 27 | 7.9 | 136 | 15.1 |
| Physical activity, MET-hr/wk | ||||||
| Mean | 13.6 | 17.8 | 15.4 | |||
| SD | 18.3 | 20.8 | 19.6 | |||
| Smoking status | ||||||
| Never | 234 | 41.7 | 126 | 37.0 | 360 | 39.9 |
| Past | 162 | 28.9 | 151 | 44.3 | 313 | 34.7 |
| Current | 159 | 28.3 | 53 | 15.5 | 212 | 23.5 |
| Missing | 6 | 1.1 | 11 | 3.2 | 17 | 1.9 |
| History of diabetes mellitus | 24 | 4.3 | 17 | 5.0 | 41 | 4.6 |
| Multivitamin use | 172 | 30.7 | 149 | 43.7 | 321 | 35.6 |
| Alcohol intake, g/d | ||||||
| Mean | 6.9 | 12.0 | 9.2 | |||
| SD | 11.8 | 17.2 | 14.7 | |||
| Cancer stage | ||||||
| Localized | 75 | 13.4 | 58 | 17.0 | 133 | 14.8 |
| Locally advanced | 48 | 8.6 | 55 | 16.1 | 103 | 11.4 |
| Metastatic | 272 | 48.5 | 167 | 49.0 | 439 | 48.7 |
| Unknown | 166 | 29.6 | 61 | 17.9 | 227 | 25.2 |
| Diagnosis period | ||||||
| 1988 to 1995 | 108 | 19.3 | 104 | 30.5 | 212 | 23.5 |
| 1996 to 2000 | 125 | 22.3 | 98 | 28.7 | 223 | 24.7 |
| 2001 to 2005 | 185 | 33.0 | 94 | 27.6 | 279 | 30.9 |
| 2006 to 2010 | 143 | 25.5 | 45 | 13.2 | 188 | 20.8 |
| Median time baseline BMI to diagnosis, years | 15.7 | 12.9 | 14.7 | |||
| Median survival time, months | 5 | 5 | 5 | |||
Abbreviations: BMI, body mass index; HPFS, Health Professionals Follow-up Study; NHS, Nurses' Health Study; SD, standard deviation.
Higher baseline BMI was associated with reduced survival (Table 2). In Cox regression models adjusted for age, cohort (also adjusted for sex), race/ethnicity, smoking status, and time period of diagnosis, the HR for death was 1.53 (95% CI, 1.11 to 2.09), comparing BMI ≥ 35 kg/m2 with BMI < 25 kg/m2 (P trend = .001). In multivariate models further adjusted for disease stage, the association was mildly attenuated, with HR for death of 1.45 (95% CI, 1.05 to 1.99), comparing BMI ≥ 35 kg/m2 with BMI < 25 kg/m2 (P trend = .01). Kaplan-Meier curves are shown in Figure 1 (log-rank P = .03). Furthermore, we divided participants into quintiles based on survival time to examine the proportion of healthy-weight individuals across categories of survival time (Appendix Table A1, online only). Notably, 44.1% of those in the bottom 20% of survival time (< 2 months) and 55.7% of those in the top 20% of survival time (> 13 months) were of healthy weight.
Table 2.
HRs for Death Resulting From Pancreatic Cancer by Baseline BMI
| Variable | BMI (kg/m2) |
P* | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| < 25.0 HR | 25.0 to 26.9 |
27.0 to 29.9 |
30.0 to 34.9 |
≥ 35.0 |
||||||
| HR | 95% CI | HR | 95% CI | HR | 95% CI | HR | 95% CI | |||
| Person-months | 4,592 | 1,820 | 1,326 | 831 | 357 | |||||
| No. of deaths | 421 of 443 | 156 of 167 | 155 of 156 | 92 of 92 | 44 of 44 | |||||
| Age adjusted | 1.0 | 1.00 | 0.84 to 1.21 | 1.24 | 1.03 to 1.50 | 1.16 | 0.93 to 1.46 | 1.47 | 1.07 to 2.00 | .003 |
| MV adjusted I† | 1.0 | 0.98 | 0.82 to 1.18 | 1.23 | 1.02 to 1.49 | 1.22 | 0.96 to 1.54 | 1.53 | 1.11 to 2.09 | .001 |
| MV adjusted II‡ | 1.0 | 0.88 | 0.73 to 1.06 | 1.04 | 0.86 to 1.26 | 1.21 | 0.96 to 1.53 | 1.45 | 1.05 to 1.99 | .01 |
Abbreviations: BMI, body mass index; HR, hazard ratio; MV, multivariable.
P for trend; tests calculated by entering stratum-specific median values for BMI as continuous variables in Cox proportional hazards models.
Adjusted for age at diagnosis, cohort (also adjusted for sex), race/ethnicity (white, nonwhite, unknown), smoking status (current, past, never, missing), and year of diagnosis (1988 to 1995, 1996 to 2000, 2001 to 2005, 2006 to 2010).
Adjusted for age at diagnosis, cohort (also adjusted for sex), race/ethnicity (white, nonwhite, unknown), smoking status (current, past, never, missing), year of diagnosis (1988 to 1995, 1996 to 2000, 2001 to 2005, 2006 to 2010), and stage at diagnosis (localized, locally advanced, metastatic, unknown).
Fig 1.
Kaplan-Meier curves of overall survival by baseline body mass index (BMI).
Higher baseline BMI was associated with more advanced stage of disease at diagnosis (Table 3). Patients had mean baseline BMI of 25.2 kg/m2 when presenting with localized disease, 25.5 kg/m2 with locally advanced disease, and 26.3 kg/m2 with metastatic disease (analysis of variance P = .03). This corresponded to 72.5% of patients with baseline BMI ≥ 30 kg/m2 having metastatic disease at presentation compared with only 59.4% of patients with BMI < 25 kg/m2 (Table 3; χ2 P = .02).
Table 3.
Baseline BMI and Pancreatic Cancer Stage at Diagnosis
| Baseline BMI (kg/m2) | Stage* |
|||||
|---|---|---|---|---|---|---|
| Localized |
Locally Advanced |
Metastatic |
||||
| No. | % | No. | % | No. | % | |
| < 25.0 | 79 | 24.3 | 53 | 16.3 | 193 | 59.4 |
| 25.0 to 29.9 | 39 | 15.1 | 40 | 15.4 | 180 | 69.5 |
| ≥ 30.0 | 15 | 16.5 | 10 | 11.0 | 66 | 72.5 |
Abbreviation: BMI, body mass index.
χ2 P = .02.
In secondary analyses, cumulatively updated BMI was similarly associated with pancreatic cancer survival. In our multivariate model adjusted for disease stage, HR for death was 1.37 (95% CI, 1.00 to 1.88), comparing those with cumulatively updated BMI ≥ 35 kg/m2 with BMI < 25 kg/m2 (P trend = .06). To further investigate the temporal association of prediagnostic BMI with pancreatic cancer survival, we grouped reported BMI by 2-year time intervals before cancer diagnosis. In these analyses, we noted a stronger association with greater time between BMI assessment and cancer diagnosis (Table 4). Among participants with BMI reported between 18 and 20 years before cancer diagnosis, HR for death was 2.31 (95% CI, 1.48 to 3.61), comparing those with BMI ≥ 30 kg/m2 with BMI < 25.0 kg/m2 (P trend < .001).
Table 4.
Temporal Association of Prediagnostic BMI and HR for Death Resulting From Pancreatic Cancer
| Years Between Reported BMI and Diagnosis Date | No. of Patients | BMI (kg/m2) |
P* | ||||
|---|---|---|---|---|---|---|---|
| < 25.0 HR† | 25.0 to 29.9 |
≥ 30.0 |
|||||
| HR† | 95% CI | HR† | 95% CI | ||||
| 2 to < 4 | 731 | 1.0 | 0.99 | 0.84 to 1.17 | 1.29 | 1.04 to 1.60 | .04 |
| 4 to < 6 | 696 | 1.0 | 0.96 | 0.81 to 1.14 | 1.29 | 1.02 to 1.61 | .06 |
| 6 to < 8 | 664 | 1.0 | 1.08 | 0.90 to 1.28 | 1.18 | 0.94 to 1.49 | .14 |
| 8 to < 10 | 611 | 1.0 | 1.03 | 0.86 to 1.24 | 1.26 | 0.99 to 1.60 | .08 |
| 10 to < 12 | 548 | 1.0 | 1.08 | 0.89 to 1.31 | 1.16 | 0.90 to 1.49 | .23 |
| 12 to < 14 | 496 | 1.0 | 1.04 | 0.84 to 1.27 | 1.28 | 0.99 to 1.67 | .07 |
| 14 to < 16 | 401 | 1.0 | 1.33 | 1.05 to 1.68 | 1.40 | 1.04 to 1.89 | .01 |
| 16 to < 18 | 300 | 1.0 | 1.08 | 0.82 to 1.41 | 1.60 | 1.12 to 2.27 | .02 |
| 18 to < 20 | 202 | 1.0 | 1.36 | 0.96 to 1.93 | 2.31 | 1.48 to 3.61 | < .001 |
Abbreviations: BMI, body mass index; HR, hazard ratio.
P for trend; tests calculated by entering stratum-specific median values for BMI as single variables in Cox proportional hazards models.
Adjusted for age at diagnosis, cohort (also adjusted for sex), race/ethnicity (white, nonwhite, unknown), smoking status (current, past, never, missing), year of diagnosis (1988 to 1995, 1996 to 2000, 2001 to 2005, 2006 to 2010), and stage at diagnosis (localized, locally advanced, metastatic, unknown).
No statistically significant effect modification was seen by cohort (also stratified by sex), disease stage, smoking status, or diagnosis period (Table 5), although the association did seem stronger among never-smokers than ever-smokers (P interaction = .06).
Table 5.
HRs for Death Resulting From Pancreatic Cancer by Baseline BMI Stratified by Covariates
| Stratification Covariate | No. of Patients | BMI (kg/m2) |
P* | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| < 25.0 HR† | 25.0 to 26.9 |
27.0 to 29.9 |
30.0 to 34.9 |
≥ 35.0 |
|||||||
| HR† | 95% CI | HR† | 95% CI | HR† | 95% CI | HR† | 95% CI | ||||
| Cohort study | .59 | ||||||||||
| HPFS (male) | 341 | 1.0 | 1.12 | 0.85 to 1.48 | 1.07 | 0.79 to 1.45 | 1.22 | 0.76 to 1.95 | 2.23 | 0.88 to 5.66 | |
| NHS (female) | 561 | 1.0 | 0.67 | 0.50 to 0.88 | 1.10 | 0.86 to 1.41 | 1.17 | 0.89 to 1.54 | 1.37 | 0.97 to 1.93 | |
| Smoking status | .06 | ||||||||||
| Never-smoker | 360 | 1.0 | 0.95 | 0.70 to 1.29 | 1.65 | 1.21 to 2.24 | 1.48 | 1.02 to 2.13 | 1.61 | 1.01 to 2.57 | |
| Ever-smoker | 525 | 1.0 | 0.81 | 0.63 to 1.03 | 0.80 | 0.62 to 1.03 | 1.07 | 0.78 to 1.46 | 1.36 | 0.86 to 2.15 | |
| Diagnosis period | .09 | ||||||||||
| 1988 to 2000 | 435 | 1.0 | 0.72 | 0.56 to 0.94 | 0.90 | 0.68 to 1.19 | 0.88 | 0.63 to 1.24 | 1.23 | 0.80 to 1.91 | |
| 2001 to 2010 | 467 | 1.0 | 1.04 | 0.79 to 1.38 | 1.16 | 0.89 to 1.52 | 1.53 | 1.10 to 2.14 | 1.49 | 0.92 to 2.40 | |
| Cancer stage | .85 | ||||||||||
| Localized | 133 | 1.0 | 0.97 | 0.56 to 1.68 | 1.09 | 0.60 to 1.95 | 1.01 | 0.56 to 1.82 | |||
| Locally advanced | 103 | 1.0 | 0.39 | 0.22 to 0.70 | 0.77 | 0.39 to 1.54 | 1.42 | 0.68 to 2.97 | |||
| Metastatic | 439 | 1.0 | 1.01 | 0.78 to 1.32 | 1.14 | 0.88 to 1.48 | 1.32 | 0.98 to 1.78 | |||
Abbreviations: BMI, body mass index; HPFS, Health Professionals Follow-up Study; HR, hazard ratio; NHS, Nurses' Health Study.
P for interaction.
Adjusted for age at diagnosis, cohort (also adjusted for sex), race/ethnicity (white, nonwhite, unknown), smoking status (current, past, never, missing), year of diagnosis (1988 to 1995, 1996 to 2000, 2001 to 2005, 2006 to 2010), and stage at diagnosis (localized, locally advanced, metastatic, unknown), excluding stratification covariate.
DISCUSSION
In this large study with prospectively collected BMI, we noted reduced survival for patients with pancreatic cancer who were obese years before their diagnosis. We also observed a statistically significant association between prediagnostic BMI and cancer stage, with more obese patients presenting with metastatic disease than patients with healthy weight. Nevertheless, the association of prediagnostic BMI was independent of other predictors of survival, including cancer stage. Furthermore, the association of prediagnostic BMI with pancreatic cancer survival was strongest when BMI was measured many years before diagnosis, suggesting the importance of chronic exposure to elevated BMI in the association with survival. In aggregate, these results suggest that chronic alterations in systemic metabolism may affect survival of patients with pancreatic adenocarcinoma.
Three hospital-based studies have evaluated BMI and survival among patients with pancreatic cancer.20–22 In these studies, patients were queried about height and weight at the time of diagnosis and were asked to recall their usual weight during adulthood or at different time points during their life. At MD Anderson Cancer Center,20 patients with recalled usual adult BMI ≥ 30 kg/m2 had HR of death of 1.75 (95% CI, 1.23 to 2.50), compared with those with a healthy weight from ages 30 to 39 to 50 to 59 years. In a study of patients from Memorial Sloan-Kettering Cancer Center,21 those with recalled BMI classified as obese had a nonstatistically significant elevated risk of death. Among patients from Mayo Clinic who completed a questionnaire at diagnosis,22 BMI was inversely associated with survival (P trend = .009). Patients with recalled usual adult BMI of 35 to 39.9 kg/m2 and ≥ 40 kg/m2 had HRs for death of 1.32 (95% CI, 1.02 to 1.72) and 1.67 (95% CI, 1.19 to 2.33), respectively, compared with those with healthy weight.
The associations of prediagnostic BMI with pancreatic cancer survival noted in these retrospective case-control studies are generally consistent with results from our study. However, prospective cohort studies have several design advantages over hospital-based, retrospective case-control studies, lending additional validity to their results. In retrospective case-control studies, patients are asked to recall their weight from many years before diagnosis or provide their usual weight during adulthood. Therefore, misclassification can occur because of incorrect recall of weight. Furthermore, if systematic differences in recall exist by the patient's performance status at diagnosis, recall bias can be introduced. Prospective study designs, such as that in our analysis, limit issues with misclassification and recall bias by soliciting BMI years before diagnosis, such that participants provide their current weight and are not influenced by their degree of illness at diagnosis.
Another limitation of hospital-based case-control studies relates to generalizability to the full population of patients presenting with pancreatic cancer. Tertiary centers attract patients who are younger, have fewer comorbidities, and present with earlier-stage disease.20–23 Furthermore, patients who consent to participate in large hospital-based case-control studies may not represent the full patient panel of the center. Some patients may be too ill or pass away before completing the necessary consents and questionnaires, therefore leading to an under-representation of patients with more aggressive tumor biology. This is of particular importance in studies of rapidly fatal malignancies, such as pancreatic cancer. Prospective study designs, such as that in our analysis, more fully capture the spectrum of patients with pancreatic cancer in terms of disease aggressiveness and stage of disease, because individuals are enrolled before their diagnosis. Notably, survival times for patients in the current study were highly similar to those of 121,713 patients included in the National Cancer Data Base, which is thought to capture 76% of patient cases of pancreatic cancer diagnosed in the United States each year.23 Median survival time for all patients with pancreatic cancer in the National Cancer Data Base was 4.4 months; by stage, survival times were 2.5 months for metastatic disease, 7.7 months for locally advanced disease, and 12.7 to 24.1 months for localized disease. In the current study, median overall survival was 5 months among all patients, and survival times by stage were 3 months for metastatic disease, 8 months for locally advanced disease, and 16 months for localized disease.
Because of the prospective study design, we also could evaluate the association of biennially updated data on BMI with pancreatic cancer survival. Interestingly, BMI reported a greater number of years before cancer diagnosis was more strongly associated with pancreatic cancer survival. These data suggest that chronic exposure to the consequences of obesity may be important in affecting patient survival. Recent experimental data suggest that pancreatic tumors develop and spread over years, providing opportunities for chronic alterations in systemic metabolism and inflammation to influence malignant characteristics and growth.24,25 In this regard, multiple metabolic and inflammatory alterations related to obesity have been shown to have an impact on tumor growth in preclinical models,11,26–28 providing experimental rationale for these findings in patients. Additionally, fatty infiltration of the pancreas may supply growth-promoting signals locally to adjacent malignant cells.29,30 Nevertheless, we cannot rule out that survival in patients with chronic obesity may also be affected by increased complication rates from surgery or systemic therapies or greater degree of comorbidity that limits aggressiveness of care.31,32 Importantly, stratified analyses by stage at diagnosis did not suggest that our results were primarily driven by those with localized disease, the population of patients for whom surgical resection is generally pursued. We also noted that obese patients were more likely to be diagnosed with advanced-stage disease, which may contribute to their worse survival. However, even after adjustment for stage, prediagnostic BMI remained associated with survival, suggesting additional mechanisms aside from more advanced stage at diagnosis.
Limitations of our study also require consideration. Among study participants, treatment programs likely varied, and we could not control for differences in treatment, because this information was not collected in our cohorts. Nevertheless, chemotherapy and irradiation have had only a modest impact on patient survival,2 and treatment programs were unlikely to have varied greatly by baseline BMI measured years before diagnosis. We used overall mortality data in our analyses, as opposed to pancreatic cancer–specific mortality. However, pancreatic cancer is a highly lethal malignancy with cure rates < 5%, such that overall mortality is a good surrogate for cancer-specific mortality. This approach has been used in other studies of pancreatic cancer survival.20–22 We cannot rule out that our findings may have been influenced in part by residual confounding. Nonetheless, we collected information on multiple other exposures and included possible confounding covariates in multivariate models without observing meaningful changes in risk estimates. Finally, our study participants were predominantly individuals of European descent, and additional studies in other populations are warranted.
In conclusion, higher prediagnostic BMI was associated with statistically significantly decreased survival among patients with pancreatic cancer from two large prospective cohort studies. This association was similar across patient subgroups and seemed stronger for BMI measured a greater number of years before cancer diagnosis. These data emphasize the link between chronic alterations in systemic metabolism and pancreatic cancer survival and suggest obesity-related metabolic pathways for possible therapeutic intervention in patients with pancreatic adenocarcinoma.
Acknowledgment
We thank the participants and staff of the Health Professionals Follow-up Study and Nurses' Health Study for their valuable contributions as well as the following state cancer registries for their help: Alabama, Arizona, Arkansas, California, Colorado, Connecticut, Delaware, Florida, Georgia, Idaho, Illinois, Indiana, Iowa, Kentucky, Louisiana, Maine, Maryland, Massachusetts, Michigan, Nebraska, New Hampshire, New Jersey, New York, North Carolina, North Dakota, Ohio, Oklahoma, Oregon, Pennsylvania, Rhode Island, South Carolina, Tennessee, Texas, Virginia, Washington, and Wyoming.
Appendix
Table A1.
Proportion of Patients With Healthy Weight by Quintiles of Survival Time
| Quintile of Survival Time | Survival Time Range (months) | No. of Patients | No. of Patients With Healthy Weight* |
|
|---|---|---|---|---|
| No. | % | |||
| 1 | < 2 | 211 | 93 | 44.1 |
| 2 | 2 to 3 | 157 | 75 | 47.8 |
| 3 | 4 to 7 | 199 | 96 | 48.2 |
| 4 | 8 to 13 | 156 | 79 | 50.6 |
| 5 | > 13 | 167 | 93 | 55.7 |
Abbreviation: BMI, body mass index.
Healthy weight defined as baseline BMI of 18.5 to 24.9 kg/m2.
Footnotes
Supported in part by Grant No. K07 CA140790 from the National Cancer Institute of the National Institutes of Health and by the American Society of Clinical Oncology Conquer Cancer Foundation, Howard Hughes Medical Institute, Lustgarten Foundation, and Promises for Purple (B.M.W.). The Nurses' Health Study and Health Professionals Follow-Up Study were supported by Grants No. P01 CA87969, P01 CA55075, P50 CA127003, R01 CA124908, R01 CA49449, and 1UM1 CA167552 from the National Cancer Institute, National Institutes of Health.
Authors' disclosures of potential conflicts of interest and author contributions are found at the end of this article.
AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST
The author(s) indicated no potential conflicts of interest.
AUTHOR CONTRIBUTIONS
Conception and design: Chen Yuan, Brian M. Wolpin
Financial support: Kimmie Ng, Brian M. Wolpin
Collection and assembly of data: Chen Yuan, Ying Bao, Chen Wu, Kimmie Ng, Meir J. Stampfer, Edward L. Giovannucci, Brian M. Wolpin
Data analysis and interpretation: All authors
Manuscript writing: All authors
Final approval of manuscript: All authors
REFERENCES
- 1.Siegel R, Naishadham D, Jemal A. Cancer statistics, 2013. CA Cancer J Clin. 2013;63:11–30. doi: 10.3322/caac.21166. [DOI] [PubMed] [Google Scholar]
- 2.Hidalgo M. Pancreatic cancer. N Engl J Med. 2010;362:1605–1617. doi: 10.1056/NEJMra0901557. [DOI] [PubMed] [Google Scholar]
- 3.Michaud DS, Giovannucci E, Willett WC, et al. Physical activity, obesity, height, and the risk of pancreatic cancer. JAMA. 2001;286:921–929. doi: 10.1001/jama.286.8.921. [DOI] [PubMed] [Google Scholar]
- 4.Arslan AA, Helzlsouer KJ, Kooperberg C, et al. Anthropometric measures, body mass index, and pancreatic cancer: A pooled analysis from the Pancreatic Cancer Cohort Consortium (PanScan) Arch Intern Med. 2010;170:791–802. doi: 10.1001/archinternmed.2010.63. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Larsson SC, Orsini N, Wolk A. Body mass index and pancreatic cancer risk: A meta-analysis of prospective studies. Int J Cancer. 2007;120:1993–1998. doi: 10.1002/ijc.22535. [DOI] [PubMed] [Google Scholar]
- 6.Stolzenberg-Solomon RZ, Graubard BI, Chari S, et al. Insulin, glucose, insulin resistance, and pancreatic cancer in male smokers. JAMA. 2005;294:2872–2878. doi: 10.1001/jama.294.22.2872. [DOI] [PubMed] [Google Scholar]
- 7.Bao Y, Giovannucci EL, Kraft P, et al. A prospective study of plasma adiponectin and pancreatic cancer risk in five US cohorts. J Natl Cancer Inst. 2013;105:95–103. doi: 10.1093/jnci/djs474. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Wolpin BM, Bao Y, Qian ZR, et al. Hyperglycemia, insulin resistance, impaired pancreatic beta-cell function and risk of pancreatic cancer. J Natl Cancer Inst. 2013;105:1027–1035. doi: 10.1093/jnci/djt123. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Grote VA, Rohrmann S, Nieters A, et al. Diabetes mellitus, glycated haemoglobin and C-peptide levels in relation to pancreatic cancer risk: A study within the European Prospective Investigation into Cancer and Nutrition (EPIC) cohort. Diabetologia. 2011;54:3037–3046. doi: 10.1007/s00125-011-2316-0. [DOI] [PubMed] [Google Scholar]
- 10.Li N, Grivennikov SI, Karin M. The unholy trinity: Inflammation, cytokines, and STAT3 shape the cancer microenvironment. Cancer Cell. 2011;19:429–431. doi: 10.1016/j.ccr.2011.03.018. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Gukovsky I, Li N, Todoric J, et al. Inflammation, autophagy, and obesity: Common features in the pathogenesis of pancreatitis and pancreatic cancer. Gastroenterology. 2013;144:1199.e4–1209.e4. doi: 10.1053/j.gastro.2013.02.007. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Pannala R, Leibson CL, Rabe KG, et al. Temporal association of changes in fasting blood glucose and body mass index with diagnosis of pancreatic cancer. Am J Gastroenterol. 2009;104:2318–2325. doi: 10.1038/ajg.2009.253. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Pannala R, Basu A, Petersen GM, et al. New-onset diabetes: A potential clue to the early diagnosis of pancreatic cancer. Lancet Oncol. 2009;10:88–95. doi: 10.1016/S1470-2045(08)70337-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Colditz GA, Hankinson SE. The Nurses' Health Study: Lifestyle and health among women. Nat Rev Cancer. 2005;5:388–396. doi: 10.1038/nrc1608. [DOI] [PubMed] [Google Scholar]
- 15.Giovannucci E, Ascherio A, Rimm EB, et al. Physical activity, obesity, and risk for colon cancer and adenoma in men. Ann Intern Med. 1995;122:327–334. doi: 10.7326/0003-4819-122-5-199503010-00002. [DOI] [PubMed] [Google Scholar]
- 16.Mozaffarian D, Hao T, Rimm EB, et al. Changes in diet and lifestyle and long-term weight gain in women and men. N Engl J Med. 2011;364:2392–2404. doi: 10.1056/NEJMoa1014296. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Rich-Edwards JW, Corsano KA, Stampfer MJ. Test of the National Death Index and Equifax Nationwide Death Search. Am J Epidemiol. 1994;140:1016–1019. doi: 10.1093/oxfordjournals.aje.a117191. [DOI] [PubMed] [Google Scholar]
- 18.Wang DS, Wang ZQ, Zhang L, et al. Are risk factors associated with outcomes in pancreatic cancer? PLoS ONE. 2012;7:e41984. doi: 10.1371/journal.pone.0041984. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Cochran WG. The combination of estimates from different experiments. Biometrics. 1954;10:101–129. [Google Scholar]
- 20.Li D, Morris JS, Liu J, et al. Body mass index and risk, age of onset, and survival in patients with pancreatic cancer. JAMA. 2009;301:2553–2562. doi: 10.1001/jama.2009.886. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Olson SH, Chou JF, Ludwig E, et al. Allergies, obesity, other risk factors and survival from pancreatic cancer. Int J Cancer. 2010;127:2412–2419. doi: 10.1002/ijc.25240. [DOI] [PubMed] [Google Scholar]
- 22.McWilliams RR, Matsumoto ME, Burch PA, et al. Obesity adversely affects survival in pancreatic cancer patients. Cancer. 2010;116:5054–5062. doi: 10.1002/cncr.25465. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Bilimoria KY, Bentrem DJ, Ko CY, et al. Validation of the 6th edition AJCC Pancreatic Cancer Staging System: Report from the National Cancer Database. Cancer. 2007;110:738–744. doi: 10.1002/cncr.22852. [DOI] [PubMed] [Google Scholar]
- 24.Yachida S, Jones S, Bozic I, et al. Distant metastasis occurs late during the genetic evolution of pancreatic cancer. Nature. 2010;467:1114–1117. doi: 10.1038/nature09515. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Haeno H, Gonen M, Davis MB, et al. Computational modeling of pancreatic cancer reveals kinetics of metastasis suggesting optimum treatment strategies. Cell. 2012;148:362–375. doi: 10.1016/j.cell.2011.11.060. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Zyromski NJ, Mathur A, Pitt HA, et al. Obesity potentiates the growth and dissemination of pancreatic cancer. Surgery. 2009;146:258–263. doi: 10.1016/j.surg.2009.02.024. [DOI] [PubMed] [Google Scholar]
- 27.Vansaun MN. Molecular pathways: Adiponectin and leptin signaling in cancer. Clin Cancer Res. 2013;19:1926–1932. doi: 10.1158/1078-0432.CCR-12-0930. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Pollak M. The insulin and insulin-like growth factor receptor family in neoplasia: An update. Nat Rev Cancer. 2012;12:159–169. doi: 10.1038/nrc3215. [DOI] [PubMed] [Google Scholar]
- 29.White PB, True EM, Ziegler KM, et al. Insulin, leptin, and tumoral adipocytes promote murine pancreatic cancer growth. J Gastrointest Surg. 2010;14:1888–1893. doi: 10.1007/s11605-010-1349-x. discussion 1893-1894. [DOI] [PubMed] [Google Scholar]
- 30.Mathur A, Zyromski NJ, Pitt HA, et al. Pancreatic steatosis promotes dissemination and lethality of pancreatic cancer. J Am Coll Surg. 2009;208:989–994. doi: 10.1016/j.jamcollsurg.2008.12.026. discussion 994-996. [DOI] [PubMed] [Google Scholar]
- 31.House MG, Fong Y, Arnaoutakis DJ, et al. Preoperative predictors for complications after pancreaticoduodenectomy: Impact of BMI and body fat distribution. J Gastrointest Surg. 2008;12:270–278. doi: 10.1007/s11605-007-0421-7. [DOI] [PubMed] [Google Scholar]
- 32.Khan S, Sclabas G, Reid-Lombardo K, et al. Does body mass index/morbid obesity influence outcome in patients who undergo pancreatoduodenectomy for pancreatic adenocarcinoma? J Gastrointest Surg. 2010;14:1820–1825. doi: 10.1007/s11605-010-1285-9. [DOI] [PubMed] [Google Scholar]