Anthropometric Risk Factors for Cancers of the Biliary Tract in the Biliary Tract Cancers Pooling Project

Abstract

Biliary tract cancers are rare but highly fatal with poorly understood etiology. Identifying potentially modifiable risk factors for these cancers is essential for prevention. Here we estimated the relationship between adiposity and cancer across the biliary tract, including cancers of the gallbladder (GBC), intrahepatic bile ducts (IHBDC), extrahepatic bile ducts (EHBDC), and the ampulla of Vater (AVC). We pooled data from 27 prospective cohorts with over 2.7 million adults. Adiposity was measured using baseline body mass index, waist circumference, hip circumference, waist-to-hip and waist-to-height ratios. Hazard ratios (HR) and 95% confidence intervals (95%CI) were estimated using Cox proportional hazards models adjusted for sex, education, race, smoking, and alcohol consumption with age as the time metric and the baseline hazard stratified by study. During 37,883,648 person-years of follow-up, 1,343 GBC cases, 1,194 EHBDC cases, 784 IHBDC cases, and 623 AVC cases occurred. For each 5 kg/m2 increase in body mass index there were risk increases for GBC (HR: 1.27 [95% CI: 1.19, 1.36]), IHBDC (HR: 1.32 [95% CI: 1.21, 1.45]), and EHBDC (HR: 1.13 [95% CI: 1.03, 1.23]), but not AVC (HR: 0.99 [95% CI: 0.88, 1.11]). Increasing waist circumference, hip circumference, waist-to-hip ratio, and waist-to-height ratio were associated with GBC and IHBDC but not EHBDC or AVC. These results indicate that adult adiposity is associated with an increased risk of biliary tract cancer, particularly GBC and IHBDC. Moreover, they provide evidence for recommending weight maintenance programs to reduce the risk of developing these cancers.

Introduction:

Biliary tract cancers (BTC) include cancers of the gallbladder (GBC), intrahepatic bile duct (IHBDC), extrahepatic bile duct (EHBDC), and the ampulla of Vater (AVC). Worldwide BTCs account for 3% of all adult cancers with marked variations in incidence by geography and ethnicity (1). The highest rates of BTC are seen among women in Latin America, South Asia, and Eastern Europe, though rates for men are also elevated in China and Japan (2, 3). In most countries, five-year survival is less than 20% (2), largely attributable to lack of early signs or symptoms of disease until the cancers are well advanced (4). Thus, understanding the etiology of BTCs and identifying potentially modifiable risk factors are critical for primary prevention.

Globally, the prevalence of obesity has been steadily rising, increasing three-fold among men, and more than doubling among women between 1975 and 2014 (5). If this trajectory persists, global obesity prevalence will be 18% among men and over 21% among women by 2025 (6). These trends in adiposity are translating into increasing incidence of obesity-related diseases among children and adolescents, and likely elevate the risk of developing several obesity-related cancers later in adulthood. Anthropometric parameters, such as body mass index (BMI) and waist circumference, have been associated with higher all-cause and cancer-specific mortality rates across racial and ethnic groups (7-11). Furthermore, obesity is associated with increasing years of life lost, and being overweight is also associated with increased cancer risk (12, 13).

Previous research suggests that overweight and obesity is associated with GBC (14), but the literature on obesity and other BTC sites has been less consistent. Studies of obesity and cholangiocarcinoma have found either an increased risk (15, 16) or no association (17, 18). Similar inconclusive results were obtained when IHBDC has been analyzed separately from EHBDC (17, 19, 20), though a recent meta-analysis found strong associations between obesity and IHBDC (21). The rarity of BTCs often results in studies that are insufficiently powered, highlighting the need for large consortium pooling projects. We examined associations between different anthropometric measures of adiposity and cancer across the biliary tract in the largest study to date involving prospective data from 27 cohorts with over 2.7 million individuals from North America, Europe, Asia, and Australia.

Methods:

Study Population:

Data were analyzed from the Biliary Tract Cancers Pooling Project (BiTCaPP), which consists of 27 studies, including 22 prospective cohort studies, and observational follow-up of participants enrolled in four randomized controlled prevention trials, and one cancer screening trial (Table 1), from North America, Europe, Asia, and Australia. BiTCaPP was determined to be exempt from Institutional Review Board review by the National Cancer Institute’s Office of Human Subjects Research. All component cohort studies within BiTCaPP received IRB approval at their respective institutions.

Table 1:

Summary of study characteristics contributing to the Biliary Tract Cancers Pooling Project^a

Study (Acronym)	Study Population	Follow-Up Period	Baseline  Sample N (%)	Total Person- Time	GBC Cases N (%)	IHBDC Cases N (%)b	EHBDC Cases N (%)	AVC Cases N (%)
AgHealth	U.S.A.	1993-2013	69,422 (2.6)	1,129,815	18 (1.3)	15 (1.9)	14 (1.5)	10 (1.6)
AHS-2	U.S.A.	2002-2015	93,264 (3.4)	971,114	15 (1.1)	9 (1.1)	11 (1.0)	6 (1.0)
ATBC	Finland	1985-2010	29,101 (1.1)	443,724	17 (1.3)	38 (4.6)	42 (3.5)	16 (2.6)
BCDDP	U.S.A.	1980-1998	42,874 (1.6)	329,205	8 (0.6)	4 (0.5)	7 (0.6)	8 (1.3)
COSM	Sweden	1998-2008	43,430 (1.6)	404,822	11 (0.8)	6 (0.7)	13 (1.1)	3 (0.5)
CPS-II NC	U.S.A.	1992-2011	152,771 (5.6)	2,055,047	69 (5.1)	54 (6.9)	53 (4.4)	34 (5.4)
EPIC	Europe	1992-2010	485,465 (17.9)	6,762,629	132 (9.8)	118 (15.1)	109 (9.1)	85 (13.6)
HPFS	U.S.A.	1986-2012	50,178 (1.9)	931,984	11 (0.8)	14 (1.8)	22 (1.8)	10 (1.6)
IWHS	U.S.A.	1986-2013	37,977 (1.4)	723,154	69 (5.1)	14 (1.8)	30 (2.5)	12 (1.9)
JPHC I JPHC II	Japan	1990-2011 1993-2011	98,031 (3.6)	1,673,119	167 (12.4)	120 (15.3)	196 (16.4)	38 (6.1)
MCCS	Australia	1990-2009	39,961 (1.5)	679021	35 (2.6)	20 (2.6)	22 (1.8)	6 (1.0)
MEC	U.S.A.	1993-2010	185,398 (6.9)	2,670,391	109 (8.1)	59 (7.5)	117 (9.8)	63 (10.1)
NHS	U.S.A.	1980-2012	95,609 (3.5)	2,382,004	52 (3.9)	16 (2.0)	31 (2.6)	18 (2.9)
NIH-AARP	U.S.A.	1995-2011	542,356 (20.0)	6,867,629	210 (15.6)	132 (16.8)	216 (18.8)	148 (23.8)
NYUWHS	U.S.A.	1985-2007	13,299 (0.5)	271,155	6 (0.5)	6 (0.7)	9 (0.8)	0
PHS	U.S.A.	1982-2009 1997-2009	28,419 (1.1)	517,908	7 (0.5)	9 (1.2)	9 (0.8)	7 (1.1)
PLCO	U.S.A.	1993-2009	146,688 (5.4)	1,624,424	45 (3.4)	20 (2.6)	49 (4.1)	34 (5.5)
RERF	Japan	1950-2005	47,953 (1.8)	1,099,161	140 (10.4)	N/A	112 (9.4)	23 (3.7)
REVEAL	Taiwan	1991-2012	23,640 (0.9)	447, 458	7 (0.5)	44 (5.6)	12 (1.0)	8 (1.3)
SCHS	Singapore	1993-2008	61,251 (2.3)	739,282	29 (2.2)	26 (3.3)	14 (1.2)	22 (3.5)
SCS	China	1986-2012	18,075 (0.7)	339,282	15 (1.1)	13 (1.7)	21 (1.8)	15 (2.4)
Sisters	U.S.A.	2003-2012	47,782 (1.8)	395,658	4 (0.3)	4 (0.5)	0	4 (0.6)
SMC	Sweden	1997-2008	36,393 (1.3)	352,405	42 (3.1)	3 (0.4)	6 (0.5)	1 (0.2)
VITAL	U.S.A.	2000-2009	73,301 (2.7)	540,831	16 (1.2)	15 (1.9)	15 (1.3)	5 (0.8)
WHI	U.S.A.	1993-2014	159,801 (5.9)	2,024,899	87 (6.5)	19 (2.4)	41 (3.4)	35 (5.6)
WHS	U.S.A.	1992-2010	38,974 (1.4)	573,970	10 (0.7)	6 (0.8)	1 (0.1)	11 (1.8)
WLHS	Norway, Sweden	1991-2011 2003-2011	45,429 (1.7)	933,557	12 (0.9)	6 (0.8)	4 (0.4)	1 (0.2)
Total			2,707,448	37,883,648	1,343	784	1,194	623

Abbreviations: Agricultural Health Study (AgHealth), Seventh-day Adventist Health Study 2 (AHS-2), Alpha-Tocopherol, Beta-Carotene Cancer Prevention Study (ATBC), Breast Cancer Detection Demonstration Project (BCDDP), Cohort of Swedish Men (COSM), Cancer Prevention Study-II, Nutrition Cohort (CPS-II NC), European Prospective Investigation into Cancer and Nutrition (EPIC), Health Professionals Follow-Up Study (HPFS), Iowa Women’s Health Study (IWHS), Japan Public Health Center-based prospective Study I & II (JPHC), Melbourne Collaborative Cohort Study (MCCS), Multiethnic Cohort Study (MEC), National Institutes of Health-American Association of Retired Persons Diet and Health Study (NIH-AARP), Nurses’ Health Study (NHS), New York University Women’s Health Study (NYUWHS), Physicians’ Health Study I & II (PHS), Prostate, Lung, Colorectal and Ovarian (PLCO) Cancer Screening Trial, Risk Evaluation of Viral Load Elevation and Associated Liver Disease/Cancer-Hepatitis B Virus Study (REVEAL), Radiation Effects Research Foundation Life Span Study (RERF), Shanghai Cohort Study (SCS), Singapore Chinese Health Study (SCHS), Sister Study (Sisters), Swedish Mammography Cohort (SMC), VITAL (VITamins And Lifestyle), Women’s Health Initiative (WHI), Women’s Health Study (WHS), Women’s Lifestyle and Health Study (WLHS).

^aATBC, PHS, WHI, and WHS are randomized controlled trials and PLCO is a screening trial. The remaining studies included in BiTCaPP are prospective cohort studies.

^bIntrahepatic bile duct cancer cases not reported by RERF.

Outcomes:

Incident BTC for each anatomical site was defined by the International Classification of Diseases codes and/or medical and death record text and classified as primary GBC, IHBDC, EHBDC, and AVC (Supplementary Table 1). A diagnosis of BTC was verified by linkage to local, state, or national cancer registries (Agricultural Health Study [AgHealth], Adventist Health Study 2 [AHS-2], the Alpha-Tocopherol, Beta-Carotene Cancer Prevention Study [ATBC], Cohort of Swedish Men [COSM], Iowa Women’s Health Study [IWHS], Melbourne Collaborative Cohort Study [MCCS], Multiethnic Cohort Study [MEC], National Institutes of Health-American Association of Retired Persons Diet and Health Study [NIH-AARP], Radiation Effects Research Foundation Life Span Study [RERF], Singapore Chinese Health Study [SCHS], Swedish Mammography Cohort [SMC], and VITamins and Lifestyle Study [VITAL]; medical record, pathology report, or death certificate (Health Professionals Follow-Up Study [HPFS], Nurses’ Health Study [NHS], Prostate, Lung, Colorectal and Ovarian Cancer Screening Trial [PLCO], Physicians’ Health Study [PHS], Women’s Health Initiative [WHI], Women’s Health Study [WHS]), and Women’s Lifestyle and Health Study [WLHS]); or a combination of methods (Breast Cancer Detection Demonstration Project [BCDDP], Cancer Prevention Study-II Nutrition Cohort [CPS-II NC], European Prospective Investigation into Cancer and Nutrition [EPIC], Japan Public Health Center Study [JPHC], New York University Women’s Health Study [NYUWHS], Risk Evaluation of Viral Load Elevation and Associated Liver Disease/Cancer-Hepatitis B Virus Study [REVEAL], Shanghai Cohort Study [SCS], and the Sister Study [Sisters]).

Exclusion Criteria:

BiTCaPP is comprised of 2,847,787 individuals. Those under the age of 18 (n=132), missing age at baseline or exit (n=5,048), with prior cancer diagnoses at baseline (n=61,356), with incident cancers categorized as being at other/other unknown sites or overlapping lesion of biliary tract (n=323), and unknown biliary tract cancer status (n=9). We also excluded those with missing height or weight (n=73,471) data, which represents 2.6% of the total study population. Data from the remaining 2,707,448 individuals comprised the analytic dataset.

Exposures:

BMI was calculated from weight in kilograms (kg) divided by squared height in meters (m²). Baseline height and weight were directly measured by study staff in seven of the cohorts and self-reported in the other 20 studies (see Supplementary Table 2 for details on data collection for anthropometric and other measures by study). BMI was categorized according to the World Health Organization International Classification for Western adults as follows: underweight (15.0 to <18.5 kg/m²), normal weight (18.5 to <25 kg/m²), overweight (25 to <30 kg/m²), and obese (≥30kg/m²). Obesity was further classified as: classes I (30 to <35 kg/m²), II (35 to <40 kg/m²), and III (≥40 kg/m²) (22). BMI for JPHC I and II, RERF, REVEAL, SCHS, and SCS was classified according to World Health Organization recommendations for Asian adults as follows: underweight (15.0 to <18.5 kg/m²), normal weight (18.5 to <23 kg/m²), overweight (23 to <27.5 kg/m²), obese (≥27.5 kg/m²), with obesity further categorized as: classes I (27.5 to <32.5 kg/m²), II (32.5 to <37.5 kg/m²), and III (≥37.5 kg/m²) (23). Self-reported weight at ages 18 or 20 was available from seven studies and was divided by baseline adult height (in m²) to calculate young adult BMI (categorized as above). Adult weight change was calculated by subtracting young adult BMI from baseline adult BMI. Weight change was categorized as: adult weight loss (change of <⁻2 kg/m²), stable weight (⁻2 to <2 kg/m²), and weight gains (2 to <5 kg/m², 5 to <10 kg/m², or ≥10 kg/m²) (24). Adult and young adult BMI were also modeled continuously per 5kg/m² increase after confirming the linear relationship between BMI and development of a BTC. When analyzed continuously, individuals with a BMI <15.0 or >60 kg/m² were excluded from analysis (n=52,410). When analyzed categorically, the BMI categories included all individuals with non-missing values. Analyses excluding extreme values from BMI categories showed no difference in the estimates.

Baseline waist circumference and hip circumference were measured in centimeters (cm) directly by study staff (n=5 studies) or using participant self-measurements (n=9 studies). Individuals with waist or hip circumferences <45 cm or >190 cm were excluded from the analysis (n=35,252 and 31,427, respectively). Waist-to-hip ratio was calculated by dividing waist circumference by hip circumference, and waist-to-height ratio was calculated by dividing waist circumference by height (both in cm). We modeled waist and hip circumferences (both per 5 cm increase) and waist-to-hip and waist-to-height ratios (both per 0.1-unit increase) as continuous variables.

Information on sex, race (white, black, Asian/Pacific Islander, and other), education level (some college, high school graduate or GED, or less than high school graduate), smoking (ever/ never), and alcohol consumption (ever/never) were collected by self-report at baseline (25). Due to a high number of missing values, education was set at “some college” for all missing in the cohorts made up of health professionals (HPFS, NHS, and PHS), and missing race set to “white” in HPFS. History of gallstone information was captured by 18 studies, and cholecystectomy was collected by 9 studies.

Statistical Analyses:

Participant baseline demographic characteristics and history of relevant medical conditions were summarized using descriptive statistics. We evaluated associations between anthropometric measures and incident BTC using Cox proportional hazards regression models with age as the time scale and left truncation at baseline to estimate site-specific hazard ratios (HRs) and 95% confidence intervals (95% CIs). Confounding was assessed using directed acyclic graphs to identify the minimally sufficient set of covariates for control (26). All models of adult adiposity were adjusted for sex, education level, race, smoking, and alcohol consumption with baseline hazard stratified by study. We tested for linear trends across BMI categories with the Wald test with 1 degree of freedom. Models examining weight change from young to middle adulthood were additionally adjusted for young adult BMI.

To examine whether the association between continuous BMI and incidence of biliary tract cancer differed by cancer subtype, we used Cox proportional hazards regression models with a duplication method, testing for heterogeneity using the Wald test (27). To assess statistical heterogeneity of results between the studies we performed a random-effects meta-analysis using Cochrane’s I² (28). Study-specific models were adjusted for sex, race, education level, smoking, and alcohol, where appropriate. The proportional hazards assumption was assessed visually by plotting the scaled Schoenfeld residuals against time and by testing for independence between the residuals and time in the models. The proportional hazards assumption was met for all models.

In the subset of cohorts (n=19) with information on prior gallstone diagnoses, we compared the estimated risk with and without adjustment for gallstones to assess whether the effect of adiposity on BTC is mediated by gallstones, the major risk factor for BTCs (17). GBC analyses were repeated for the subset of studies (n=9) that collected cholecystectomy history, comparing the estimated risk of GBC when restricted to individuals with a gallbladder.

To test for effect measure modification, an interaction term for BMI by sex was included in all models. As there was no evidence of effect measure modification by sex, this term was omitted from final models. We repeated the main analyses stratified by Western and Asian countries as the etiology of BTCs may differ between these areas. As a sensitivity analysis, we re-ran the analyses excluding events within that occurred within one year of exposure assessment to account for possible reverse causation. We saw no difference in the associations in this analysis. For example, the HR for GBC in obese vs. normal-weight individuals was 1.71 (95% CI: 1.40, 2.08) with the one-year lag, compared to 1.72 (95% CI: 1.41, 2.08) without the lag. We also assessed associations between continuous BMI and BTC in the same participants who provided young adult BMI. Due to concerns about residual confounding by smoking and alcohol, we also conducted a sensitivity analysis for continuous BMI restricted to the studies that collected smoking in packyears and drinking in drinks per day.

Statistical tests were two-sided with a type I error rate of α=0.05. Stata (v.14) software was used for meta-analyses; R Studio (v. 3.5.0) was used to test the proportional hazards assumptions; and SAS (v9.4; Cary, NC) software was used for pooled and study specific association estimates.

Results:

In this analysis 1,343 GBC cases, 1,194 EHBDC cases, 784 IHBDC cases, and 623 AVC cases occurred during 37,883,648 person-years of follow-up time as shown in Table 1. Participant characteristics are presented by cohort in Table 2. The median age at baseline was 59 years (standard deviation [SD] =10), 59% were women, 80% were white, 59% had some college education, 52% were ever smokers, and 57% were ever drinkers. The mean BMI for each of the cohorts ranged from 21.9 kg/m² (SD = 3) in RERF to 29.9 kg/m² (SD = 7) in AHS-2. Within the cohorts that collected relevant medical history, 9% of participants had gallstones and 11% reported a cholecystectomy. Anthropometric characteristics at baseline are presented by cohort and sex in Supplementary Table 3.

Table 2:

Summary of Participant Characteristics by Cohort Included in the Biliary Tract Cancer Pooling Project

Study	Womena %	Age Mean   (SD)	Race/Ethnicityb %				Some Collegec %	BMI Mean (SD)	Gallstonesd %	Cholecyst- ectomye %	Ever Smokerf %	Ever Drinkerg %
			White	Black	Asian/ Pacific Islander	Other
Ag Health	42	47 (13)	98	1	<1	<1	48	26.7 (5)	N/A	N/A	39	18
AHS-2	65	58 (14)	68	28	3	2	78	29.9 (7)	3	N/A	20	36
ATBC	0	57 (5)	100	0	0	0	0	26.3 (4)	6	5	100	79
BCDDP	100	62 (8)	90	4	5	1	46	25.1 (5)	N/A	N/A	43	44
COSM	0	61 (10)	100	0	0	0	17	25.9 (3)	11	N/A	64	80
CPS-II NC	53	63 (6)	98	1	<1	<1	68	26.0 (4)	12	13	56	61
EPIC	70	51 (10)	100	0	0	0	48	25.4 (4)	8	N/A	50	84
HPFS	0	54 (10)	97	<1	1	5	100	25.5 (3)	N/A	3	54	77
IWHS	100	62 (4)	99	<1	<1	<1	39	25.9 (5)	N/A	N/A	34	12
JPHC	52	53 (8)	0	0	100	0	12	23.5 (3)	3	N/A	40	43
MCCS	59	55 (9)	100	0	0	0	25	26.9 (4)	9	8	43	71
MEC	54	60 (9)	25	17	36	23	26	26.5 (5)	7	6	56	25
NIH-AARP	40	62 (5)	95	4	1	<1	74	27.1 (5)	10	14	64	48
NHS	100	47 (7)	94	1	1	4	100	24.4 (5)	2	8	56	81
NYUWHS	100	51 (9)	84	12	1	3	69	24.9 (5)	5	N/A	53	19
PHS	0	55 (10)	94	1	5	1	100	24.9 (3)	4	N/A	47	100
PLCO	51	63 (5)	89	5	4	2	70	27.3 (5)	12	N/A	54	72
RERF	60	52 (14)	0	0	100	0	13	21.9 (3)	N/A	N/A	45	47
REVEAL	50	47 (10)	0	0	100	0	3	24.0 (3)	4	N/A	29	11
SCHS	55	56 (8)	0	0	100	0	28	23.1 (3)	N/A	N/A	31	19
SCS	0	56 (6)	0	0	100	0	72	22.2 (3)	N/A	N/A	57	43
Sisters	100	55 (9)	87	9	1	4	85	27.8 (6)	15	13	43	37
SMC	100	62 (9)	100	0	0	0	19	25.3 (4)	20	N/A	46	36
VITAL	51	61 (7)	94	1	3	2	80	27.2 (5)	N/A	N/A	53	34
WHI	100	63 (7)	83	9	3	6	77	27.9 (6)	16	13	49	58
WHS	100	55 (7)	96	2	1	<1	100	25.8 (5)	10	N/A	49	20
WLHS	100	40 (6)	100	0	0	0	41	23.5 (4)	N/A	N/A	59	86
Total	59	57 (10)	80	4	13	3	59	26.2 (5)	9	11	52	57

Abbreviations: Agricultural Health Study (AgHealth), Seventh-day Adventist Health Study 2 (AHS-2), Alpha-Tocopherol, Beta-Carotene Cancer Prevention Study (ATBC), Breast Cancer Detection Demonstration Project (BCDDP), Cohort of Swedish Men (COSM), Cancer Prevention Study-II, Nutrition Cohort (CPS-II NC), European Prospective Investigation into Cancer and Nutrition (EPIC), Health Professionals Follow-Up Study (HPFS), Iowa Women’s Health Study (IWHS), Japan Public Health Center-based prospective Study I & II (JPHC), Melbourne Collaborative Cohort Study (MCCS), Multiethnic Cohort Study (MEC), National Institutes of Health-American Association of Retired Persons Diet and Health Study (NIH-AARP), Nurses’ Health Study (NHS), New York University Women’s Health Study (NYUWHS), Physicians’ Health Study I & II (PHS), Prostate, Lung, Colorectal and Ovarian (PLCO) Cancer Screening Trial, Risk Evaluation of Viral Load Elevation and Associated Liver Disease/Cancer-Hepatitis B Virus Study (REVEAL), Radiation Effects Research Foundation Life Span Study (RERF), Shanghai Cohort Study (SCS), Singapore Chinese Health Study (SCHS), Sister Study (Sisters), Swedish Mammography Cohort (SMC), VITAL (VITamins And Lifestyle), Women’s Health Initiative (WHI), Women’s Health Study (WHS), Women’s Lifestyle and Health Study (WLHS).

Variables are missing for the following numbers of participants out of the studies reporting these variables:

^aSex – 35;

^bRace – 31,185;

^cEducation – 115,157;

^dHistory of gallstones – 684,518;

^eCholecystectomy – 1,722,714;

^fEver smoker–46,979;

^gEver drinker – 741,107. N/A indicates these data were not available.

Associations of adiposity with BTC risk from the pooled analysis are shown in Table 3. An increase in BMI was associated with increased risk for each anatomic specific cancers except AVC. For each 5 kg/m² increase in BMI there was an increased risk of GBC (HR: 1.27 [95% CI: 1.19, 1.36]), IHBDC (HR: 1.32 [95% CI: 1.21, 1.45]), and EHBDC (HR: 1.13 [95% CI: 1.03, 1.23]). These associations differed by anatomic site when GBC was compared to AVC and EHBDC (P for heterogeneity=0.0005 and 0.04, respectively), but not when compared to IHBDC (P=0.89). There was no evidence that these associations differed by sex or of between-study heterogeneity for adult BMI (I²: 0% for all cancer sites; Supplementary Figures 1a-d). These associations did not change substantially when adjusting for smoking packyears and drinks per day among the subset of studies that collected these measures.

Table 3:

Associations between anthropometric characteristics and biliary tract cancers in the Biliary Tract Cancer Pooling Project^a

Characteristic	Non- casesb	GBC Cases	GBC HR (95% CI)	IHBDC Cases	IHBDC HR (95% CI)	EHBDC Cases	EHBDC HR (95% CI)	AVC Cases	AVC HR (95% CI)
BMIc (kg/m2): per 5 kg/m2	2,383,716	1,101	1.27 (1.19, 1.36)	624	1.32 (1.21, 1.45)	937	1.13 (1.03, 1.23)	538	0.99 (0.88, 1.11)
P-heterogeneity (compared to AVC)d			0.0005		0.001		0.308
P-heterogeneity (compared to IHBDC)d			0.886				0.009
P-heterogeneity (compared to EHBDC)d			0.004
BMI Category (kg/m2)e:
Underweight	39,400	29	0.98 (0.60, 1.60)	6	0.91 (0.43, 1.95)	15	0.92 (0.53, 1.59)	14	1.25 (0.65, 2.39)
Normal	1,036,939	402	1.00 (reference)	215	1.00 (reference)	352	1.00 (reference)	205	1.00 (reference)
Overweight	912,487	431	1.31 (1.11, 1.54)	266	1.34 (1.09, 1.64)	407	1.14 (0.96, 1.35)	232	1.09 (0.87, 1.35)
Obese	442,878	261	1.72 (1.41, 2.08)	144	2.06 (1.62, 2.61)	175	1.33 (1.06, 1.65)	92	1.08 (0.81, 1.35)
Obese I	338,889	181	1.62 (1.31, 2.00)	110	2.03 (1.57, 2.61)	131	1.31 (1.03, 1.67)	73	1.07 (0.78, 1.48)
Obese II	97,456	50	1.54 (1.06, 2.22)	24	2.17 (1.39, 3.38)	31	1.27 (0.81, 2.00)	17	1.33 (0.78, 2.28)
Obese III	45,101	30	3.32 (2.15, 4.80)	10	2.16 (1.05, 4.45)	13	1.67 (0.88, 3.18)	2	0.49 (0.12, 1.94)
P-trendf			<0.0001		<0.0001		0.008		0.78
Waist Circumferenceg (cm): per 5 cm	1,127,471	434	1.15 (1.11, 1.20)	245	1.09 (1.04, 1.16)	327	1.03 (0.98, 1.09)	223	1.03 (0.97, 1.10)
Hip Circumferenceh (cm): per 5 cm	1,011,338	398	1.16 (1.11, 1.22)	216	1.13 (1.05, 1.21)	298	1.02 (0.95, 1.09)	198	1.02 (0.94, 1.11)
Waist-to-Hip Ratiog,h: per 0.1	1,037,853	445	1.11 (1.06, 1.17)	229	1.09 (1.01, 1.18)	323	1.04 (0.93, 1.16)	207	0.99 (0.80, 1.22)
Waist-to-Height Ratiog: per 0.1	1,104,877	418	1.47 (1.33, 1.63)	242	1.27 (1.07, 1.51)	317	1.08 (0.94, 1.25)	216	1.07 (0.87, 1.32)

Abbreviations: AVC, ampulla of Vater cancer; BMI, body mass index; cm, centimeter; EHBDC, extrahepatic bile duct cancer; GBC, gallbladder cancer; IHBDC, intrahepatic bile duct cancer; kg, kilogram; and m, meter.

^aAll models use age as the time scale and were adjusted for sex, race, (white, black, Asian/Pacific Islander, other), education (<high school graduate, high school graduate, some college/post-high school training), smoking (ever versus never), and alcohol consumption (ever versus never) and the baseline hazard is stratified by study.

^bNon-cases: The same non-case group was used for all analyses, except for analyses of intrahepatic bile duct cancer. RERF did not provide information on intrahepatic bile duct cancer diagnoses, so this study was excluded from these analyses.

^cExcluded those with a BMI less than 15 kg/m² or greater than 60 kg/m²

^dThe Wald test was used to test for the heterogeneity of the associations between continuous BMI and biliary tract cancer subtype.

^eThe BMI categories for Western adults were defined as follows: underweight (15.0 to <18.5 kg/m²), normal weight (18.5 to <25 kg/m²), overweight (25 to <30 kg/m²), and obese (≥30kg/m²), obese class I (30 to <35 kg/m²), obese class II (35 to <40 kg/m²), and obese class III (≥40 kg/m²). The BMI categories for Asian adults were defined as follows: underweight (15.0 to <18.5 kg/m²), normal weight (18.5 to <23 kg/m²), overweight (23 to <27.5 kg/m²), obese (≥27.5 kg/m²), obese class I (27.5 to <32.5 kg/m²), obese class II (32.5 to <37.5 kg/m²), and obese class III (≥37.5 kg/m²).

^fThe Wald test was used to test for a linear trend across categories of BMI and biliary tract cancer site.

^gRestricted to cohorts that collected waist circumference at baseline (BCDDP, COSM, CPS-II NC, EPIC, IWHS, MCCS, NIH-AARP, NHS, NYUWHS, REVEAL, Sister, SMC, WHI, and WLHS.)

^hRestricted to cohorts that collected hip circumference at baseline (BCDDP, COSM, EPIC, IWHS, MCCS, NIH-AARP, NHS, NYUWHS, REVEAL, Sister, SMC, WHI, and WLHS.)

When BMI was categorized according to WHO classifications, compared to normal weight, those in the overweight and obese classes had an increased risk of GBC (HR for overweight: 1.31 [95% CI: 1.11, 1.54] and HR for obese: 1.72 [95% CI: 1.41, 2.08]) and IHBDC (HR for overweight: 1.34 [95% CI: 1.09, 1.64] and HR for obese: 2.06 [95% CI: 1.62, 2.61]). Increasing BMI category was also associated with an increased risk in EHBDC (HR for overweight: 1.14 [95% CI: 0.96, 1.35] and HR for obese: 1.33 [95% CI: 1.06, 1.65] compared to normal weight). The risk of cancer was particularly pronounced in individuals in the highest obesity category (obese III) compared to normal weight for GBC (HR: 3.32 [95% CI: 2.15, 4.80]) and IHBDC (HR: 2.16 [95% CI: 1.05, 4.45]). There was no evidence of an association between BMI classification and AVC.

Waist and hip circumference, waist-to-hip ratio, and waist-to-height ratios were associated with increased risk of developing GBC and IHBDC, but not EHBDC or AVC, (Table 3). The risk of GBC increased 15%, and the risk of IHBDC increased 9% for each 5 cm increase in waist circumference. A 5 cm increase in hip circumference conferred a 16% and 13% increased risk for GBC and IHBDC, respectively. Waist-to-hip ratio was associated with the risk of GBC (HR per 0.1 unit: 1.11 [95% CI: 1.06, 1.17]) and IHBDC (HR per 0.1 unit: 1.09 (1.01, 1.18). Stronger associations were seen with increases in waist-to-height ratios than in waist-to-hip ratios for GBC (HR per 0.1unit: 1.47 [95% CI: 1.33, 1.63]) and IHBDC (HR per 0.1 unit: 1.27 [95% CI: 1.07, 1.51].

Seven studies collected from participants their recalled weight at ages 18 or 20. Among these participants, increasing BMI in young adulthood was associated with increased risk later in adult life of GBC (HR per 5 kg/m²: 1.26 [95% CI: 1.04, 1.52]), IHBDC (HR per 5 kg/m²: 1.34 [95% CI: 1.03, 1.73]), and AVC (HR per 5 kg/m²: 1.33 [95% CI: 1.01, 1.75]), but not EHBDC (HR per 5 kg/m²: 0.87 [0.68, 1.12]) (Supplementary Table 4). Adult weight gain (≥10 kg/m²) compared to maintaining stable weight was associated with increased risk of developing GBC later in life (HR: 1.77 [95% CI: 1.16, 2.70]) and EHBDC (HR: 2.04 [95% CI: 1.17, 3.55]). Being overweight or obese as a young adult compared to normal weight was associated with an increased risk of AVC later in life (HR for overweight: 1.88 [95% CI: 1.10, 3.21] and HR obese: 2.07 [95% CI: 0.65, 6.60]). The results of the main analysis were similar when we repeated the analysis of adult adiposity restricting to participants who provided data on young adult BMI.

The associations between adiposity and BTCs did not change substantially when models were additionally adjusted for gallstones (Supplementary Table 5). However, when restricting the analysis of GBC to those without a history of cholecystectomy, the associations with cancer in the obese classes I, II, and III compared to normal weight were stronger than the associations where cholecystectomy was ignored (Supplementary Table 6). For example, compared to normal weight the HR for obese class III was 3.45 (95% CI: 1.95, 6.11) when restricted to those without cholecystectomy and 2.57 (95% CI: 1.58, 4.19) when not restricting inclusion in GBC analysis based on history of cholecystectomy. There were also consistently stronger associations between waist and hip circumferences, waist-to-hip ratio, and waist-to-height ratio with GBC when restricting the analysis to those without a history of cholecystectomy.

When stratifying the results by region (Supplementary Table 7), the association between BMI and GBC remained for individuals in Western countries (BMI continuous HR: 1.29 [95% CI: 1.20, 1.39]), but this association was not observed in Asian countries (BMI continuous HR: 1.00 [95% CI: 0.79, 1.28]).

Discussion:

In the largest prospective study of BTCs by site to date, we found that adult adiposity was associated with increased risk of GBC and IHBDC, and to a lesser extent EHBDC, but not AVC. Compared with being in the normal weight range, being in the overweight and obese classifications was associated with a 31% and 72% increased risk of GBC, respectively, and with a 34% and 106% increased risk of IHBDC. Being in the obese category was associated with a 33% increased risk of EHBDC. These results imply that interventions to maintain a healthy weight in adulthood may be beneficial in the prevention of BTCs. The robustness of our findings is strengthened by the consistency across multiple measures of adiposity, and several sensitivity analyses.

Obesity may increase risk of carcinogenesis of the biliary tract through disruption in the metabolism of hormones and inflammatory mediators, such as insulin and cytokines (17). Obesity may also affect cancer risk indirectly by increasing the risk for gallstones (29). However, these effects appear to vary across the biliary tract. The overweight and obese categories were strongly associated with GBC and IHBDC, while obesity was more modestly associated with EHBDC, and BMI category was not associated with AVC. A prior case-control study found that increasing BMI over time was associated with GBC, but not EHBDC or AVC; however power was limited to assess risk at these sites (17). In addition, strong associations with GBC and IHBDC risk were seen across multiple measures of adult adiposity, while, adiposity was not consistently associated with EHBDC or AVC risk in this study and in other studies (17, 30). These findings suggest that the etiologic mechanism by which adiposity causes cancer differs by biliary tract site.

The lack of an association between AVC and adult adiposity may be due in part because the ampulla may be epidemiologically more similar to duodenal cancers for which there is mixed evidence for the effect of obesity on cancer development (31). Excess adipose tissue can contribute to low-grade systemic inflammation, and thus, a condition like non-alcoholic fatty liver disease might be expected to have the strongest effect on cancer development within the intrahepatic bile ducts and the gallbladder little to no impact on AVC (21, 32). However, we did see an association between young adult BMI, but not adult weight gain, and increased risk AVC. Yet, the small number of AVC cases that reported their young adult weight makes it difficult to draw conclusions from these data.

Gallstones are a major risk factor for GBC and, to a lesser extent, other BTCs. In a previous study from Shanghai, China, 80% of gallbladder, 59% of bile duct, and 42% of ampulla of Vater cancers were attributed to gallstones (17). Given that obesity increases the risk of gallstones (3), it is unclear whether the association between obesity and BTCs is due to the increased risk of gallstones, and to what extent it is independent of gallstones. That our risk estimates were not substantially different from the main analyses when we adjusted for gallstones, suggests that BMI increases the risk of BTCs through mechanisms other than gallstones. Further, our finding that increasing BMI was not associated with GBC risk in Asia suggest that the etiology of GBC may differ between Asian and Western countries. For example, compared to Western countries, the prevalence of gallstones is lower in Asian countries where pigment stones associated with infection have historically been more common than cholesterol gallstones. The lack of association between obesity and GBC in Asian countries may suggest that the association in Western countries is related to cholesterol gallstones, highlighting the importance of gallstones in gallbladder carcinogenesis.(3) However, it is important to note that gallstone data were based on self-report at baseline; many people with gallstones are unaware that they have them (3), and some individuals who reported no gallstones may have developed them later on. Ideally, an analysis of BMI within a cohort of individuals with gallstones is needed to better assess the mechanisms by which obesity increases the risk of BTCs, especially GBC, independently of gallstones.

Our study expands on previously pooled analyses, which also found that adiposity was associated with GBC and IHBDC risk (14, 21, 33). Most research on adiposity has focused on BMI and do not always included measures abdominal obesity, such as waist circumference, hip circumference, waist-to-hip ratio, and waist-to-height ratio. Abdominal fat, independent of BMI, has been associated with the risk of cancers of the colon and rectum (34), endometrium (35), and esophagus (36). Our analysis not only confirmed that central adiposity increases the risk of GBC, but also identified differences in the risk of cancer across the biliary tract. We found that GBC and IHBDC were more strongly associated with waist-to-height ratio than waist circumference, hip circumference, or waist-to-hip ratio, similar to a study of hepatocellular carcinoma, which found waist-to-height ratio, compared to waist-to-hip ratio, had a stronger association with cancer risk (37). The association between waist-to-height ratio was weaker for EHBDC. In addition, we saw no evidence that the risk of any of the BTCs differed by sex, though other studies have (38, 39). Our study was largely comprised of postmenopausal women and is consistent with previous research that shows the sex disparity for BTCs narrows in older ages (14, 21, 40).

This study has several strengths, including its prospective design and sample size; with nearly 4,000 BTC cases, we were able to analyze associations by anatomic site within the biliary tract. BiTCaPP also includes one of the largest collections of AVCs, a relatively unstudied cancer. Our results indicate that obesity may not be a risk factor for the development of AVC, which has also been suggested by smaller studies (17, 30). Such site-specific analysis is important given that BTC risk factors vary by site, resulting in differential clinical health management. By pooling data across existing cohorts, we were able to study a rare cancer in these relatively lower risk populations. We were also able to account for two important factors that have largely been ignored in previous studies: gallstones and cholecystectomy. Our analysis of associations with GBC restricted to people not reporting a history of cholecystectomy are particularly important as they allowed us to assess associations between measures of adiposity and GBC among only those people who were truly at risk of developing GBC. Additionally, this pooling project is not subject to publication bias where null findings are often excluded from analysis, as is often the case in publication-based meta-analyses (41).

This study also has some limitations. Seven of the 27 studies collected anthropometric measures by trained study staff and 20 relied on participant self-report. Women tend to underreport their weight and men are more likely to overreport their height (42). Thus, bias towards lower BMI when measured by self-report can occur. However, as this bias is non-differential by case status, it likely skews our associations between the obese weight category and BTCs towards the null. We also lacked data on physical activity and had a large amount of missing data on important covariates, which may have resulted incomplete adjustment for confounding. Our analysis on young adult BMI should also be interpreted with caution because these variables were not collected by all the cohort studies and the number of events, especially for AVC, was small.

In conclusion, findings from this pooled analysis of 27 prospective studies support the hypothesis that adiposity is associated with an increased risk of BTC, particularly GBC and IHBDC. The differences found in the associations with several measures of adiposity across the biliary tract point to a unique etiology for cancer at these sites. These results provide evidence for recommending weight maintenance programs to patients, especially those at high risk for biliary tract cancers (e.g. those with a family history or a history of gallstones). Further research is warranted to see if these relationships between adult adiposity and cancer risk across the biliary tract persist in higher-risk geographic regions, such as Latin America and South Asia.

Significance: Findings identify a correlation between adiposity and biliary tract cancers, indicating that weight management programs may help minimize the risk of these diseases.

Abbreviations

AVC

Ampulla of Vater cancer

BMI

Body mass index

BTC

Biliary tract cancer

CI

Confidence interval

EHBDC

Extrahepatic bile duct cancer

IHBDC

Intrahepatic bile duct cancer

GBC

Gallbladder cancer

HR

Hazard ratio