Associations between autoimmune conditions and hepatobiliary cancer risk among elderly US adults

Abstract

Growing evidence suggests that people with autoimmune conditions may be at increased risk of hepatobiliary tumors. In the present study, we evaluated associations between autoimmune conditions and hepatobiliary cancers among adults aged ≥66 in the United States. We used Surveillance, Epidemiology, and End Results (SEER)-Medicare data (1992-2013) to conduct a population-based, case-control study. Cases (n=32,443) had primary hepatobiliary cancer. Controls (n=200,000) were randomly selected, cancer-free adults frequency-matched to cases by sex, age, and year of selection. Using multivariate logistic regression, we calculated odds ratios (ORs) and 95% confidence intervals (CIs) for associations with 39 autoimmune conditions identified via Medicare claims. We also conducted separate analyses for diagnoses obtained via inpatient versus outpatient claims. Sixteen conditions were associated with at least one hepatobiliary cancer. The strongest risk estimates were for primary biliary cholangitis with hepatocellular carcinoma (OR: 31.33 [95% CI: 23.63-41.56]) and primary sclerosing cholangitis with intrahepatic cholangiocarcinoma (7.53 [5.73-10.57]), extrahepatic cholangiocarcinoma (5.59 [4.03-7.75]), gallbladder cancer (2.06 [1.27-3.33]), and ampulla of Vater cancer (6.29 [4.29-9.22]). Associations with hepatobiliary-related conditions as a group were observed across nearly all cancer sites (ORs ranging from 4.53 [95% CI: 3.30-6.21] for extrahepatic cholangiocarcinoma to 7.18 [5.94-8.67] for hepatocellular carcinoma). Restricting to autoimmune conditions diagnosed via inpatient claims, 6 conditions remained associated with at least one hepatobiliary cancer, and several risk estimates increased. In the outpatient restricted analysis, 12 conditions remained associated. Multiple autoimmune conditions are associated with hepatobiliary cancer risk in the US Medicare population, supporting a shared immuno-inflammatory etiology to these cancers.

INTRODUCTION

Hepatobiliary cancers, which include hepatocellular carcinoma, intrahepatic cholangiocarcinoma, extrahepatic cholangiocarcinoma, gallbladder cancer, and ampulla of Vater cancer, are highly lethal malignancies characterized by aggressive growth, frequent metastases, late detection, and poor survival.^1-3 Increasing incidence rates of hepatocellular carcinoma, intrahepatic cholangiocarcinoma, and extrahepatic cholangiocarcinoma have been reported in the United States (US)^4-6, and liver cancer is the second leading cause of cancer death globally.⁷ Gallbladder and ampulla of Vater cancers account for a relatively small proportion of cancer incidence and mortality both in the US and globally,^{6, 7} but gallbladder cancer disproportionately affects certain subgroups of the US population, such as American Indians, Alaska Natives, and Hispanics.^{3, 6} Hepatobiliary cancers are also most commonly diagnosed among older adults.^{4, 6}

Immune and inflammatory responses are known contributors to tumorigenesis, and inflammation-inducing risk factors have been associated with increased cancer risk at various sites.⁸ Hepatobiliary cancers, though diverse in their biological and clinical features, share a host of immune-related risk factors, including obesity, chronic inflammation, and viral infection, providing evidence for a common immune component to their etiology.^{1, 9}

Immune response is dysregulated in individuals with autoimmune diseases, and the association between prior history of autoimmune diseases and increased risk of certain cancers, particularly lymphomas, is well established.¹⁰ Growing evidence suggests that patients diagnosed with autoimmune conditions are also at increased risk of solid tumors, including those of the gastrointestinal tract,¹¹ but limited research has been conducted on hepatobiliary cancers. In a recent population-based study using administrative health data from Sweden,¹² a large number of associations between autoimmune conditions and hepatobiliary cancers was reported. However, these results were based on autoimmune diseases identified via inpatient claims only (i.e. patients who were hospitalized). Replication in other populations is needed to evaluate the validity and generalizability of these findings.

To address this issue, we evaluated associations between autoimmune conditions and hepatobiliary cancers by anatomic site in a large, US population-based case-control study of Medicare-aged adults. We ascertained the presence of autoimmune conditions using both inpatient and outpatient claims combined as well as in separate analyses for inpatient versus outpatient only diagnoses, allowing us to extend upon prior work that relied solely on inpatient claims.

METHODS

Data Source

The National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER) Program collects information on all incident cancers that occur in a defined geographic area representing approximately 28% of the US population.¹³ Medicare, administered by the US Centers for Medicare and Medicaid Services, is the primary health insurer for 97% of the US population aged 65 years and older. All Medicare beneficiaries are entitled to Part A benefits (coverage of hospital or inpatient care), and approximately 96% also subscribe to Part B benefits (coverage of physician and outpatient care). The Medicare database contains International Classification of Diseases 9th revision (ICD-9) diagnosis codes, including diagnoses made for both inpatients and outpatients.

SEER-Medicare is an electronic linkage database that matches SEER cancer registry data with Medicare enrollment and claims. To date, 94% of all people aged 65 or older in the SEER database have been successfully matched to the Medicare enrollment database.¹⁴ Additional details on the SEER-Medicare database have been described previously.¹⁵

Study Design and Participant Selection

We conducted a population-based, nested case-control study within the SEER-Medicare population.¹⁵ Hepatobiliary cancer cases were identified in SEER by anatomic site and histological code using International Classification of Disease for Oncology (ICD-O) codes (Supplementary Table 1). Eligible cases for this analysis were primary hepatobiliary cancers occurring as first malignancies between 1992 and 2013. Individuals with cancer diagnosed by death certificate only were excluded because research has shown these data sources are often inaccurate.¹⁶ Cancers diagnosed at autopsy only were also excluded because these cancers may have different characteristics than cancers that present clinically.¹⁵ To avoid surveillance bias (i.e. that diagnosis of an autoimmune disease led to increased surveillance for cancer), we excluded the 12-month period prior to case diagnosis from the ascertainment of autoimmune diseases. Thus, cases were required to have a minimum of 13 months of Medicare Part A and B, non-health maintenance organization coverage prior to cancer diagnosis, making the minimum age at diagnosis 66 years. Cases were also required to have at least 1 Medicare claim (MEDPAR, NCH, or OUTPATIENT) more than 12 months prior to selection, to ensure patients included in the study were using Medicare benefits. Individuals over the age of 99 were excluded.

Two hundred thousand cancer-free population-based controls were randomly selected from a 5% sub-cohort representing a random sample of Medicare recipients residing in the geographic region covered by SEER. Controls were required to be alive and cancer-free as of July 1 of the calendar year of their selection and were subject to the same exclusion criteria as those used for cases. Controls were frequency-matched to cases by sex, age category (66-69, 70-74, 75-79, 80-84, and 85-99), and calendar year of selection. The present study was conducted simultaneously with a pancreatic cancer study; therefore, controls were matched to both hepatobiliary and pancreatic cancer cases.

Ascertainment of Autoimmune Diseases and Other Conditions

We used ICD-9 codes to identify 39 autoimmune conditions. Since diabetes is a well-established risk factor for these cancers¹⁷ and is associated with various autoimmune diseases,¹⁸ we also identified diabetes diagnoses (types I and II) using ICD-9 codes (Supplementary Table 2). To improve specificity, autoimmune diseases and diabetes diagnoses were defined as 1 inpatient (MEDPAR) claim or 2 physician/outpatient (NCH or OUTPATIENT) claims at least 30 days apart. We required autoimmune conditions to be present at least 12 months prior to cancer diagnosis or control selection to minimize reverse causation. We also analyzed associations with the following groups of conditions: 1) any autoimmune condition, 2) any digestive system-related autoimmune condition (ankylosing spondylitis, celiac disease, Crohn’s disease, discoid lupus erythematosus, immune thrombocytopenic purpura, morphea, pernicious anemia, polyarteritis nodosa, reactive arthritis, sarcoidosis, systemic lupus erythematosus, systemic sclerosis, or ulcerative colitis), and 3) any hepatobiliary-related autoimmune condition (autoimmune hepatitis, primary biliary cholangitis, or primary sclerosing cholangitis).

Statistical Analysis

Associations between claims for autoimmune diseases and hepatobiliary cancers were assessed using multivariate, unconditional logistic regression models. Models were adjusted for case-control matching factors (sex, age category, and year of diagnosis/selection [divided into quartiles based on the controls and treated categorically]) and a priori selected confounders: race/ethnicity (non-Hispanic white, non-Hispanic black, Hispanic, Asian/Pacific Islander, American Indian/Alaska Native, and other), average number of physician claims per year >12 months before cancer diagnosis or control selection (as a measure of healthcare use [divided into quartiles based on the controls and treated categorically]), and diabetes (ever versus never diagnosed). All analyses were stratified by cancer site. Adjusted odds ratios (ORs) with 95% confidence intervals (CIs) were calculated to assess the strength of these associations. OR variance estimates were adjusted to account for the fact that controls could be sampled more than once in different calendar years and could also later become cases. Due to the large number of statistical tests performed, we considered associations to be significant if they passed a 10% false discovery rate according to the Benajmini-Hochberg procedure.¹⁹ For associations that were based on 0 exposed cases or where models did not converge, we additionally performed Fisher’s exact tests to assess any associations that may have been missed in the logistic regression analysis. All Fisher’s exact tests were non-significant (p>0.05). In accordance with the SEER-Medicare data use agreement, we suppressed cells with <11 observations and any cell that could be used to calculate the value of a cell with <11 observations.

To identify differences in associations between patients diagnosed with an autoimmune disease via inpatient versus outpatient claims, we performed separate analyses restricting models to 1) autoimmune diseases diagnosed with at least one inpatient claim and 2) autoimmune diseases diagnosed using outpatient claims only. All statistical analyses were performed using SAS Version 9.1 (SAS Institute, Cary, NC).

RESULTS

A total of 32,443 hepatobiliary cancers were included in our study; by cancer site, there were 16,423 (50.6%) hepatocellular carcinomas, 3,007 (9.3%) intrahepatic cholangiocarcinomas, 4,222 (13.0%) extrahepatic cholangiocarcinomas, 6,023 (18.6%) gallbladder cancers, and 2,768 (8.5%) ampulla of Vater cancers (Table 1). The mean age among controls was 77.7 years (standard deviation [SD]: 7.2) and ranged by site among cancer cases from 75.7 (SD: 6.4) to 78.4 years (SD: 7.1). The majority of controls were female (53.5%), while the proportion of females ranged by cancer site from 34.4 to 72.0%. Most of the participants were non-Hispanic whites (83.1% of controls; 67.1 to 82.0% of cancer cases) and had never received a diabetes diagnosis (74.6% of controls; 52.8 to 71.1% of cancer cases). The average number of physician claims per year among controls was 7.5 (SD: 7.3) and ranged by cancer site from 7.3 (SD: 7.2) to 9.2 (SD: 9.3).

Table 1.

Demographic characteristics of SEER-Medicare hepatobiliary cancer cases and matched controls, 1992 -2013.

Characteristics	Controls n (%)	Hepatobiliary cancer cases* n (%)
		HCC	IHCC	EHCC	GBC	AVC
Total	200,000	16,423	3,007	4,222	6,023	2,768
Age (years)
66-69	28,025 (14.0)	3,156 (19.2)	422 (14.0)	545 (12.9)	716 (11.9)	376 (13.6)
70-74	46,143 (23.0)	4,613 (28.1)	679 (22.6)	895 (21.2)	1,286 (21.4)	638 (23.1)
75-79	47,503 (23.8)	4,150 (25.3)	736 (24.5)	942 (22.3)	1,445 (24.0)	657 (23.7)
80-84	40,297 (20.2)	2,814 (17.1)	598 (19.9)	919 (21.8)	1,290 (21.4)	570 (20.6)
85-99	38,032 (19.0)	1,690 (10.3)	572 (19.0)	921 (21.8)	1,286 (21.4)	527 (19.0)
mean (SD)	77.7 (7.2)	75.7 (6.4)	77.7 (7.2)	78.3 (7.2)	78.4 (7.1)	77.8 (7.1)
Sex
Male	92,923 (46.5)	10,771 (65.6)	1,317 (43.8)	2,054 (48.7)	1,686 (28.0)	1,370 (49.5)
Female	107,077 (53.5)	5,652 (34.4)	1,690 (56.2)	2,168 (51.4)	4,337 (72.0)	1,398 (50.5)
Year of diagnosis/selection †
1992-2000	43,776 (21.9)	3,096 (18.9)	835 (27.8)	719 (17.0)	1,584 (25.7)	664 (24.0)
2001-2005	53,143 (26.6)	4,131 (25.2)	592 (19.7)	1,158 (27.4)	1,667 (27.7)	789 (28.5)
2006-2009	50,446 (25.2)	4,302 (26.2)	679 (22.6)	1,134 (26.9)	1,402 (23.3)	650 (23.5)
2010-2013	52,635 (26.3)	4,894 (29.8)	901 (30.0)	1,211 (28.7)	1,406 (23.3)	665 (24.0)
Race/ethnicity
White (non-Hispanic)	166,176 (83.1)	11,024 (67.1)	2,386 (79.4)	3,332 (78.9)	4,639 (77.0)	2,270 (82.0)
Black (non-Hispanic)	14,617 (7.3)	1,410 (8.6)	182 (6.1)	271 (6.4)	548 (9.1)	174 (6.3)
Hispanic	4,854 (2.4)	737 (4.5)	82 (2.7)	129 (3.1)	272 (4.5)	84 (3.0)
Asian/Pacific Islander	8,593 (4.3)	2,078 (12.7)	192 (6.4)	288 (6.8)	295 (4.9)	>150 (5.5)
American Indian/Alaska Native	672 (0.3)	139 (0.9)	21 (0.7)	17 (0.4)	59 (1.0)	<11 (0.3)
Other‡	5,088 (2.5)	1,035 (6.3)	144 (4.8)	185 (4.4)	210 (3.5)	80 (2.9)
Average number of physician claims per year §
0-2.5	49,979 (25.0)	3,449 (21.0)	741 (24.6)	978 (23.2)	1,543 (25.6)	706 (25.5)
>2.5-5.6	50,003 (25.0)	3,397 (20.7)	746 (24.8)	959 (22.7)	1,518 (25.1)	653 (23.6)
>5.6-10.0	50,036 (25.0)	3,962 (24.1)	721 (24.0)	1,048 (24.8)	1,530 (25.4)	679 (24.5)
>10.0	49,982 (25.0)	5,615 (34.2)	799 (26.6)	1,237 (29.3)	1,432 (23.8)	730 (26.4)
mean (SD)	7.5 (7.3)	9.2 (9.3)	7.6 (7.2)	8.1 (7.8)	7.3 (7.2)	7.9 (9.2)
Diabetes ∥
No	149,219 (74.6)	8,675 (52.8)	2,065 (68.7)	2,840 (67.3)	4,191 (69.6)	1,968 (71.1)
Yes	50,781 (25.4)	7,748 (47.2)	942 (31.3)	1,382 (32.7)	1,832 (30.4)	800 (28.9)

^*HCC=hepatocellular carcinoma; IHCC=intrahepatic cholangiocarcinoma; EHCC=extrahepatic cholangiocarcinoma; GBC=gallbladder cancer; AVC=ampulla of Vater cancer.

^†Selection year: Calendar year of cancer diagnosis or control selection.

^‡Other: Includes other, missing, and unknown.

^§Average number of physician claims per year: Calculated excluding the 12-month period immediately before cancer diagnosis or control selection.

^∥Diabetes: Defined as 1 inpatient claim or 2 physician/outpatient claims ≥30 days apart, excluding the 12-month period prior to cancer diagnosis or control selection.

Across all cancer sites, 12.5 to 16.1% of cancer patients had a prior autoimmune disease diagnosis in either inpatient or outpatient claims: 16.1% among hepatocellular carcinomas, 15.0% among intrahepatic cholangiocarcinomas, 14.6% among extrahepatic cholangiocarcinomas, 14.0% among gallbladder cancers, and 12.5% among ampulla of Vater cancers. Among controls, 12.2% had an autoimmune disease diagnosis prior to their selection date. After Benajmini-Hochberg correction, a total of 16 individual autoimmune conditions were significantly associated with hepatobiliary cancer at one or more sites, using both inpatient and outpatient claims to define autoimmune diseases (Table 2).

Table 2.

Associations between autoimmune (AI) conditions and hepatobiliary cancers in the SEER-Medicare database, overall and restricted based on AI condition claim status (inpatient versus outpatient), 1992-2013.^*

	OVERALL				AI CONDITION DIAGNOSED VIA INPATIENT CLAIM(S) †			AI CONDITION DIAGNOSED VIA OUTPATIENT CLAIMS ONLY ‡
	Total affected by AI condition		OR (95% CI)§	p-value	Total affected by AI condition	OR (95% CI)§	p-value	Total affected by AI condition	OR (95% CI)§	p-value
	n	%			n			n
Hepatocellular carcinoma (n=16,423)
Any AI condition	2,636	16.1	1.33 (1.27, 1.40)	<0.0001	857	1.61 (1.49, 1.75)	<0.0001	1779	1.23 (1.16, 1.30)	<0.0001
Any digestive system-related AI condition	659	4	1.19 (1.09, 1.30)	<0.0001	140	1.33 (1.10, 1.61)	0.003	519	1.16 (1.05, 1.28)	0.002
Any hepatobiliary-related AI condition	257	1.6	7.18 (5.94, 8.67)	<0.0001	109	3.66 (2.83, 4.73)	<0.0001	148	20.01 (14.80, 27.06)	<0.0001
Aplastic anemia	31	0.2	3.06 (1.99, 4.69)	<0.0001	<11	1.92 (0.74, 5.00)	0.2	>20	3.52 (2.19, 5.66)	<0.0001
Chronic rheumatic heart disease	347	2.1	1.18 (1.05, 1.33)	0.006	224	1.25 (1.08, 1.45)	0.003	123	1.08 (0.88, 1.32)	0.5
Immune thrombocytopenic purpura	59	0.4	5.02 (3.49, 7.23)	<0.0001	35	5.91 (3.58, 9.74)	<0.0001	24	4.14 (2.44, 7.02)	<0.0001
Myasthenia gravis	<11	0.1	0.45 (0.24, 0.86)	0.02	<11	0.54 (0.24, 1.22)	0.1	<11	0.37 (0.13, 1.01)	0.05
Polymyalgia rheumatica	90	0.5	0.69 (0.56, 0.86)	0.0009	24	0.78 (0.51, 1.18)	0.2	66	0.67 (0.52, 0.86)	0.002
Primary biliary cholangitis	186	1.1	31.33 (23.63, 41.56)	<0.0001	52	34.55 (20.87, 57.22)	<0.0001	134	30.20 (21.54, 42.33)	<0.0001
Primary sclerosing cholangitis	68	0.4	1.99 (1.48, 2.67)	<0.0001	55	1.75 (1.27, 2.41)	0.0006	13	4.42 (2.06, 9.45)	0.0001
Psoriasis	347	2.1	1.56 (1.38, 1.77)	<0.0001	71	2.12 (1.62, 2.79)	<0.0001	276	1.47 (1.28, 1.68)	<0.0001
Pure red cell aplasia	475	2.9	6.79 (5.94, 7.76)	<0.0001	277	9.01 (7.50, 10.82)	<0.0001	198	5.02 (4.13, 6.11)	<0.0001
Systemic lupus erythematosus	59	0.4	1.33 (1.01, 1.76)	0.04	27	1.82 (1.20, 2.76)	0.005	32	1.08 (0.74, 1.57)	0.7
Vitiligo	22	0.1	2.26 (1.37, 3.73)	0.001	<11	2.24 (0.66, 7.60)	0.2	>11	2.26 (1.31, 3.92)	0.004
Intrahepatic cholangiocarcinoma (n=3,007)
Any AI condition	450	15	1.24 (1.11, 1.38)	0.0001	124	1.24 (1.03, 1.50)	0.03	326	1.23 (1.09, 1.38)	0.001
Any digestive system-related AI condition	138	4.6	1.31 (1.10, 1.57)	0.002	30	1.47 (1.02, 2.13)	0.04	108	1.27 (1.04, 1.55)	0.02
Any hepatobiliary-related AI condition	44	1.5	6.45 (4.69, 8.87)	<0.0001	>30	6.69 (4.68, 9.56)	<0.0001	<11	5.69 (2.84, 11.39)	<0.0001
Aplastic anemia	<11	0.2	3.39 (1.48, 7.72)	0.004	<11	4.20 (1.00, 17.65)	0.05	<11	3.08 (1.12, 8.45)	0.03
Crohn’s disease	19	0.6	2.47 (1.55, 3.93)	0.0001	<11	2.11 (0.98, 4.52)	0.06	>11	2.75 (1.54, 4.92)	0.0007
Primary biliary cholangitis	<11	0.3	5.63 (2.72, 11.63)	<0.0001	<11	8.07 (2.43, 26.77)	0.0006	<11	4.77 (1.92, 11.88)	0.0008
Primary sclerosing cholangitis	40	1.3	7.53 (5.37, 10.57)	<0.0001	>30	6.95 (4.83, 10.01)	<0.0001	<11	NA	NA
Ulcerative colitis	31	1	2.24 (1.55, 3.22)	<0.0001	<11	2.16 (1.14, 4.09)	0.02	>20	2.27 (1.46, 3.53)	0.0003
Vitiligo	<11	0.2	3.29 (1.32, 8.20)	0.01	0	NA	NA	<11	4.14 (1.64, 10.42)	0.003
Extrahepatic cholangiocarcinoma (n=4,222)
Any AI condition	617	14.6	1.15 (1.05, 1.26)	0.002	179	1.20 (1.03, 1.41)	0.02	438	1.13 (1.02, 1.25)	0.02
Any digestive system-related AI condition	192	4.5	1.25 (1.08, 1.45)	0.004	34	1.17 (0.83, 1.65)	0.4	158	1.27 (1.07, 1.49)	0.005
Any hepatobiliary-related AI condition	45	1.1	4.53 (3.30, 6.21)	<0.0001	>30	4.66 (3.26, 6.65)	<0.0001	<11	4.10 (2.06, 8.15)	<0.0001
Aplastic anemia	<11	0.2	2.67 (1.24, 5.75)	0.01	<11	1.38 (0.19,10.12)	0.7	<11	3.15 (1.37, 7.23)	0.007
Crohn’s disease	20	0.5	1.79 (1.14, 2.82)	0.01	<11	1.45 (0.68, 3.11)	0.3	>11	2.05 (1.17, 3.60)	0.01
Primary biliary cholangitis	<11	0.1	2.09 (0.76, 5.73)	0.2	<11	5.87 (1.76, 19.60)	0.004	<11	0.71 (0.10, 5.20)	0.7
Primary sclerosing cholangitis	44	1	5.59 (4.03, 7.75)	<0.0001	>30	4.84 (3.37, 6.95)	<0.0001	<11	13.67 (6.15, 30.37)	<0.0001
Ulcerative colitis	34	0.8	1.67 (1.18, 2.37)	0.004	12	1.78 (0.99, 3.20)	0.06	22	1.62 (1.05, 2.49)	0.03
Gallbladder cancer (n=6,023)
Any AI condition	843	14	1.13 (1.05, 1.23)	0.002	230	1.15 (1.00, 1.33)	0.05	613	1.12 (1.03, 1.23)	0.01
Any digestive system-related AI condition	263	4.4	1.24 (1.09, 1.41)	0.001	52	1.28 (0.96, 1.70)	0.09	211	1.23 (1.06, 1.42)	0.005
Crohn’s disease	27	0.4	1.83 (1.23, 2.71)	0.003	13	1.99 (1.12, 3.52)	0.02	14	1.70 (0.99, 2.93)	0.05
Pernicious anemia	152	2.5	1.28 (1.09, 1.52)	0.003	13	1.32 (0.75, 2.32)	0.3	139	1.28 (1.07, 1.52)	0.006
Primary sclerosing cholangitis	19	0.3	2.06 (1.27, 3.33)	0.003	>11	1.77 (1.03, 3.01)	0.04	<11	5.50 (1.73, 17.54)	0.004
Sjögren’s syndrome	<11	0.1	0.39 (0.18, 0.81)	0.01	<11	0.54 (0.13, 2.21)	0.4	<11	0.35 (0.14, 0.84)	0.02
Ampulla of Vater cancer (n=2,768)
Any AI condition	345	12.5	1.03 (0.91, 1.16)	0.6	93	1.02 (0.83, 1.26)	0.8	252	1.04 (0.91, 1.19)	0.6
Any hepatobiliary-related AI condition	30	1.1	4.97 (3.41, 7.24)	<0.0001	30	6.40 (4.36, 9.38)	<0.0001	0	NA	NA
Ankylosing spondylitis	<11	0.3	2.96 (1.38, 6.35)	0.005	<11	4.05 (1.27, 12.92)	0.02	<11	2.47 (0.90, 6.74)	0.08
Celiac disease	<11	0.3	3.26 (1.60, 6.64)	0.001	<11	2.82 (0.69, 11.57)	0.1	<11	3.44 (1.51, 7.81)	0.003
Discoid lupus erythematosus	<11	0.2	1.81 (0.74, 4.41)	0.2	<11	5.56 (1.69, 18.24)	0.005	<11	0.90 (0.22, 3.64)	0.9
Primary sclerosing cholangitis	30	1.1	6.29 (4.29, 9.22)	<0.0001	30	6.85 (4.66, 10.07)	<0.0001	0	NA	NA

^*A total of 39 individual autoimmune conditions and 3 groups of conditions were evaluated. Shown here are the individual conditions and groups of conditions that were significantly associated with hepatobiliary cancer in at least one of the analyses (overall, inpatient, or outpatient), after accounting for multiple comparisons using Benjamini-Hochberg correction. In addition, associations with any autoimmune condition are shown for all cancer sites.

^†Inpatients: Restricted to autoimmune conditions diagnosed with at least 1 inpatient claim.

^‡Outpatients: Restricted to autoimmune conditions diagnosed with at least 2 physician/outpatient claims ≥30 days apart and no inpatient claims.

^§Adjusted for age, sex, year of cancer diagnosis/control selection, average number of physician visits per year, race, and diabetes.

NA= No cancer cases affected or models did not converge.

NOTE: Associations that passed Benjamini-Hochberg correction are shown in bold.

Eight autoimmune diseases (aplastic anemia, chronic rheumatic heart disease, immune thrombocytopenic purpura, primary biliary cholangitis, primary sclerosing cholangitis, psoriasis, pure red cell aplasia, and vitiligo) were significantly associated with increased risk of hepatocellular carcinoma, with ORs ranging from 1.18 (95% CI: 1.05-1.33) for chronic rheumatic heart disease to 31.33 (95% CI: 23.63-41.56) for primary biliary cholangitis. On the other hand, myasthenia gravis and polymyalgia rheumatica were associated with a decreased risk of hepatocellular carcinoma (OR: 0.45 [95% CI: 0.24-0.86] and OR: 0.69 [95% CI: 0.56-0.86], respectively).

Six autoimmune diseases (aplastic anemia, Crohn’s disease, primary biliary cholangitis, primary sclerosing cholangitis, ulcerative colitis, and vitiligo), were associated with intrahepatic cholangiocarcinoma, with ORs ranging from 2.24 (95% CI: 1.55-3.22) for ulcerative colitis to 7.53 (95% CI: 5.73-10.57) for primary sclerosing cholangitis. Additionally, patients diagnosed with aplastic anemia, Crohn’s disease, primary sclerosing cholangitis, and ulcerative colitis were at higher risk for extrahepatic cholangiocarcinoma (OR ranging from 1.67 [95% CI: 1.18-2.37] for ulcerative colitis to 5.59 [95% CI: 4.03-7.75] for primary sclerosing cholangitis).

For gallbladder cancer, an increased risk was detected among patients diagnosed with Crohn’s disease (OR: 1.83 [95% CI: 1.23-2.71]), pernicious anemia (OR: 1.28 [95% CI: 1.09-1.52]), and primary sclerosing cholangitis (OR: 2.06 [95% CI: 1.27-3.33), while patients diagnosed with Sjögren’s syndrome were at decreased risk (OR: 0.39 [95% CI: 0.18-0.81]). Increased risk of ampulla of Vater cancer was observed among patients diagnosed with ankylosing spondylitis (OR: 2.96 [95% CI: 1.38-6.35]), celiac disease (OR: 3.26 [95% CI: 1.60-6.64]), and primary sclerosing cholangitis (OR: 6.29 [95% CI: 4.29-9.22]).

In the analysis of grouped exposures, hepatocellular carcinoma, intrahepatic cholangiocarcinoma, extrahepatic cholangiocarcinoma, and gallbladder cancer were all associated with a history of any autoimmune condition (ORs ranging from 1.13 [95% CI: 1.05-1.32] for gallbladder cancer to 1.33 [95% CI: 1.27-1.40] for hepatocellular carcinoma). These same four cancers were also associated with any digestive system-related condition, with ORs ranging from 1.19 (95% CI: 1.09-1.30) for hepatocellular carcinoma to 1.31 (95% CI: 1.10-1.57) for intrahepatic cholangiocarcinoma. In addition, a significant increase in risk of hepatocellular carcinoma, intrahepatic cholangiocarcinoma, extrahepatic cholangiocarcinoma, and ampulla of Vater cancer was detected among patients with a history of any hepatobiliary-related autoimmune condition (OR: 7.18 [95% CI: 5.94-8.67]), OR: 6.45 [95% CI: 4.69-8.87], OR: 4.53 [95% CI: 3.30-6.21], and OR: 4.97 [95% CI: 3.41-7.24], respectively). Risk estimates that did not pass Benjamini-Hochberg correction are presented in Supplementary Table 3.

After restricting the analyses to autoimmune conditions diagnosed with at least one inpatient claim, nearly 40% of the autoimmune conditions identified in the combined analysis (6 out of 16) remained significantly associated with at least one hepatobiliary cancer and many risk estimates increased (Table 2). New associations were observed between systemic lupus erythematosus and hepatocellular carcinoma (OR: 1.82 [95% CI: 1.20-2.76]), primary biliary cholangitis and extrahepatic cholangiocarcinoma (OR: 5.87 [95% CI: 1.76-19.61]), and discoid lupus erythematosus and ampulla of Vater cancer (OR: 5.56 [95% CI: 1.69-18.24]). Systemic lupus erythematosus was also associated with hepatocellular carcinoma in the combined analysis (OR: 1.33 [95% CI: 1.01-1.76]) but this association did not pass Benajmini-Hochberg correction. Associations with the grouped exposures were fairly consistent. Other associations did not pass Benjamini-Hochberg correction (Supplementary Table 4).

In the analyses restricted to autoimmune conditions diagnosed via outpatient claims only, 75% of the conditions identified in the combined analysis (12 out of 16) remained significantly associated with at least one cancer site, and nearly all the associations with grouped exposures were consistent (Table 2). Associations that did not pass Benjamini-Hochberg correction are shown in Supplementary Table 5.

DISCUSSSION

In the largest study to date on this topic, we found a 1.1- to 31.3-fold increased risk of hepatobiliary cancer following diagnosis of a range of different autoimmune diseases. As expected, several autoimmune conditions with gastrointestinal tract involvement (Crohn’s diseases, primary biliary cholangitis, primary sclerosing cholangitis, and ulcerative colitis) were associated with increased risk at multiple hepatobiliary cancer sites, and the strongest risk estimate (OR: 31.33) was observed for primary biliary cholangitis and hepatocellular carcinoma. When we analyzed associations with grouped exposures of digestive system-related conditions and hepatobiliary-related conditions, we observed associations across nearly all cancer sites. The associations with hepatobiliary-related autoimmune conditions were especially strong, with ORs ranging from 4.53 to 7.18 in the combined inpatient and outpatient analysis. In addition, even more generalized systemic conditions, such as ankylosing spondylitis and systemic lupus erythematous, were associated with increased risk at some hepatobiliary cancer sites.

We observed a few associations that were particularly strong, for example, primary biliary cholangitis and primary sclerosing cholangitis with multiple hepatobiliary cancers and immune thrombocytopenic purpura and pure red cell aplasia with hepatocellular carcinoma. Hepatobiliary cancer screening for primary biliary cholangitis and primary sclerosing cholangitis is currently recommended by multiple medical associations, including the American Association for the Study of Liver Diseases.^{20, 21} Our results support these recommendations and suggest that further investigation into the utility of hepatobiliary cancer screening among individuals with other autoimmune conditions may be warranted.

Our study is the first to our knowledge to evaluate associations between a large number of autoimmune diseases and all hepatobiliary cancers by site, while assessing variations in associations based on autoimmune condition claim status. Nearly two-thirds of the risk estimates observed in the combined analyses were higher in the inpatient restricted analyses and lower in the outpatient restricted analyses; however, most of these differences were small and confidence intervals often overlapped. A few risk estimates were much higher in the inpatient analysis as compared to the outpatient analysis: pure red cell aplasia with hepatocellular carcinoma (OR: 9.01 [95% CI: 7.50, 10.82] for inpatients versus 5.02 [4.13, 6.11] for outpatients); primary biliary cholangitis with both intrahepatic cholangiocarcinoma (8.07 [2.43, 26.77] versus 4.77 [1.92, 11.88]) and extrahepatic cholangiocarcinoma (5.87 [1.76, 19.60] versus 0.71 [0.10, 5.20]); and discoid lupus erythematosus with ampulla of Vater cancer (5.56 [1.69, 18.24] versus 0.90 [0.22, 3.64]). Associations observed among adults hospitalized for certain autoimmune conditions may be stronger because hospitalization may indicate a more severe disease presentation, resulting in increased hepatobiliary cancer risk. This is biologically plausible given the role that chronic immune stimulation or suppression has been shown to play in carcinogenesis.^{8, 10} In addition, increasing severity of chronic inflammation in patients with autoimmune diseases has been associated with an increased risk of other cancers, such as lymphomas.²² Alternatively, the observed increase in some risk estimates may reflect greater accuracy among inpatient versus outpatient diagnoses.

There were also several cases where risk estimates were much higher in the outpatient analysis than in the inpatient analysis: the association of primary sclerosing cholangitis with hepatocellular carcinoma (OR: 4.42 [95% CI: 2.06, 9.45] for outpatients versus 1.75 [1.27, 2.41] for inpatients), extrahepatic cholangiocarcinoma (13.67 [6.15. 30.37] versus 4.84 [3.37, 6.95]), and gallbladder cancer (5.50 [1.73, 17.54] versus 1.77 [1.03, 3.01]). Primary sclerosing cholangitis also contributed to the association of any hepatobiliary-related condition with hepatocellular carcinoma (OR: 20.01 [95% CI: 14.80, 27.06] for outpatients versus 3.66 [2.83, 4.73] for inpatients). It is possible that competing morbidity or mortality among patients with severe cases of primary sclerosing cholangitis accounts for the lower risk estimates observed among inpatients.

We observed 5 novel associations with hepatocellular carcinoma (3 conditions associated with increased risk and 2 associated with decreased risk). Aplastic anemia, pure red cell aplasia, and vitiligo were associated with an increased risk of hepatocellular carcinoma, with particularly strong associations observed for pure red cell aplasia. In previous studies, aplastic anemia has been associated with lymphoproliferative neoplasms, and possible mechanisms for this association include immunological changes, genetic characteristics, or exposure to environmental factors (e.g., infections and toxic chemicals) that predispose individuals to both aplastic anemia and lymphomas.²³ Although associations with other cancers remain uncertain, similar mechanisms may account for the observed association with hepatocellular carcinoma. For pure red cell aplasia, mechanisms may include infectious or inflammatory pathways.²⁴ Possible mechanisms for vitiligo include innate immune activation²⁵ and associated liver X receptor α gene polymorphisms,²⁶ which may influence lipid dysregulation and consequently hepatocellular carcinoma risk.

Two conditions, myasthenia gravis and polymyalgia rheumatica, were associated with decreased risk of hepatocellular malignancy. In most prior studies, myasthenia gravis has not been associated with an increased risk of cancer among patients without thymoma,²⁷ and one prior study found a potential protective effect against extrathymic cancer.²⁸ However, the association we observed was based on a low number of exposed cancer cases and requires further investigation. The association with polymyalgia rheumatica was based on a larger number of exposed cases, but previous studies investigating cancer risk among polymyalgia rheumatica patients have been conflicting. One study investigating risk of all cancers found no significant association,²⁹ while another study investigating risk of hepatocellular carcinoma observed an increased risk; however, this study relied on inpatient diagnoses of polymyalgia rheumatica only.¹² In our study, the risk estimate for patients diagnosed via inpatient claims was closer to the null than both the overall and outpatient only risk estimates and was not statistically significant, suggesting that disease claim status may contribute to these conflicting results.

We also confirmed several previously reported associations between hepatocellular carcinoma and autoimmune diseases such as chronic rheumatic heart disease,¹² immune thrombocytopenic purpura,¹² primary biliary cholangitis,^{12, 30} primary sclerosing cholangitis,^31-33 psoriasis,^{12, 34} and systemic lupus erythematosus.^{12, 35} Notably, all of these associations were observed in a recent, population-based study conducted in Sweden,¹² except primary sclerosing cholangitis, which was not assessed in the Swedish study. Replication of these prior findings suggests that these associations are robust. A modest increase in hepatocellular carcinoma risk has been observed among primary sclerosing cholangitis patients in several other studies,^31-33 however these studies have all had a relatively small number of cancer cases, most have been cross-sectional, and some analyzed all hepatobiliary cancers together. Our study confirms the modest association with hepatocellular carcinoma in a much larger population using a longitudinal design and site-specific risk estimates. Our results are consistent with prior research, which has illustrated increased incidence of hepatocellular carcinoma among patients with primary biliary cholangitis,³⁰ whereas the incidence among patients with primary sclerosing cholangitis appears to be relatively low.³⁶ The stronger association we observed may be particularly true for patients infected with hepatitis B virus, as there is evidence to suggest that primary biliary cholangitis related liver damage may be more severe in the context of hepatitis B infection.³⁷

Although physio-pathological mechanisms for liver carcinogenesis have been posited for primary biliary cholangitis and primary sclerosing cholangitis,^{30, 32} much less is known about the role of chronic rheumatic heart disease, immune thrombocytopenic purpura, polymyalgia rheumatica, psoriasis, and systemic lupus erythematosus in the development of hepatocellular carcinoma. Possible hypotheses include liver congestion and subsequent damage for chronic rheumatic heart disease,³⁸ changes in immune and inflammation responses as a result of low platelet counts for immune thrombocytopenic purpura,³⁹ alterations in cell proliferation and carcinogenic effects of treatments for psoriasis,³⁴ and altered clearance of viruses for systemic lupus erythematosus.³⁵ Finally, a few conditions that have been associated with hepatocellular carcinoma in prior studies, including celiac disease and ulcerative colitis,¹² were not observed in our study.

We also observed associations with intrahepatic cholangiocarcinoma, extrahepatic cholangiocarcinoma, gallbladder cancer, and ampulla of Vater cancer. Because these malignancies are rare, few prior studies have investigated the role of autoimmune diseases in their etiology. Nevertheless, we replicated previously reported associations of Crohn’s disease,^{9, 12, 40} primary biliary cholangitis,^{40, 41} primary sclerosing cholangitis,^{32, 40-42} and ulcerative colitis^{12, 43} with intrahepatic and extrahepatic cholangiocarcinomas. In addition, we replicated a few previously observed associations with gallbladder cancer, including Crohn’s disease,¹² pernicious anemia,¹² and primary sclerosing cholangitis.⁴⁴ Some of these conditions, such as ulcerative colitis and primary sclerosing cholangitis, are strongly associated with one another and may share an etiologic pathway.⁴⁵ Primary biliary cholangitis, a chronic cholestatic liver disease, may increase biliary tract cancer risk through local inflammation, destruction of small intrahepatic bile ducts, or progression to liver cirrhosis.³⁰ The associations of primary biliary cholangitis with intrahepatic and extrahepatic cholangiocarcinomas were strong (OR >5 in at least one of the analyses); however, the association with hepatocellular carcinoma was approximately 6 times stronger than those with cholangiocarcinoma. Although the reason for this difference in magnitude is unclear, it may reflect augmentation of the effects of primary biliary cholangitis by viral hepatitis,³⁷ which is much more strongly associated with hepatocellular carcinoma than both intrahepatic and extrahepatic cholangiocarcinoma.⁴⁶ Although we did not observe an association between primary biliary cholangitis and gallbladder or ampulla of Vater cancer, the small number of cases with these conditions may have limited our ability to detect associations. Using both inpatient claims only and inpatient claims combined with outpatient claims, intrahepatic cholangiocarcinoma, extrahepatic cholangiocarcinoma, and ampulla of Vater cancer were all strongly associated with the grouped exposure of hepatobiliary-related conditions, suggesting that organ-specific autoimmune diseases may increase hepatobiliary cancer risk through local inflammation.

In addition, we observed novel associations of aplastic anemia and vitiligo with intrahepatic cholangiocarcinoma; aplastic anemia with extrahepatic cholangiocarcinoma; and ankylosing spondylitis, celiac disease, discoid lupus erythematosus, and primary sclerosing cholangitis with ampulla of Vater cancer. Several potential mechanisms might explain our findings. Primary sclerosing cholangitis is characterized by chronic bile duct destruction and inflammation and has been consistently associated with other liver and biliary tract cancers.³² Mechanisms for the associations with ankylosing spondylitis, aplastic anemia, celiac disease, discoid lupus erythematosus, and vitiligo may include immunological, genetic, and inflammatory pathways.^{23, 25, 47-49} For example, celiac disease, which causes inflammation in the small intestine,⁴⁹ could potentially lead to carcinogenic changes in the ampulla of Vater, located in the duodenal wall. A novel, inverse association was also observed between Sjögren’s syndrome and gallbladder cancer, however potential mechanisms for an inverse association remain unclear.

Strengths of this study include its population-based design and large sample size. Using SEER-Medicare data, we were able to systematically assess associations in a nationally representative population of US adults aged 66 years and older. SEER registries employ strict quality-control measures to ensure the reliability of cancer case ascertainment. The large size of our study also allowed us to assess associations of rare autoimmune conditions with hepatobiliary cancers by site and to perform separate analyses by autoimmune claim status.

Despite these strengths, some limitations exist. Our study is limited in its ability to accurately assess some potential confounders, such as co-infections, obesity, and smoking. In addition, we did not have information on treatments for autoimmune diseases (e.g., immunosuppressive medications), which may increase cancer risk.⁵⁰ The generalizability of our findings to a younger population is also limited. Medical claims data are inherently limited in their diagnostic accuracy; because we relied on Medicare claims data, we were unable to ascertain the presence of medical conditions diagnosed before age 65 and some autoimmune conditions may have been under-ascertained (i.e. present but undiagnosed) or misreported by physicians. We expect that misreporting would be attenuated by our strict definition of autoimmune diagnoses (1 inpatient claim or 2 physician/outpatient claims), and under-ascertainment of autoimmune conditions would have driven the observed associations toward the null. In addition, the rarity of some hepatobiliary cancers and autoimmune conditions may have limited our ability to detect associations with modest effect sizes. Some of our findings have not been observed in prior studies or vice versa, which may be due to differences in study design, sample size, characteristics of the study population, and/or the low prevalence of some autoimmune conditions. Finally, multiple comparisons may have led to some spurious findings. However, we used statistical correction to mitigate chance findings.

In this US, population-based study of autoimmune conditions and hepatobiliary cancers, we found an increased risk of several hepatobiliary cancers among patients with a prior autoimmune condition diagnosis, particularly for conditions with gastrointestinal involvement. We replicated prior findings from a nationally representative Swedish study in a new population and observed novel associations with several autoimmune conditions (e.g., aplastic anemia, pure red cell aplasia). If confirmed by other studies, these novel findings may assist in developing strategies for monitoring cancer risk in individuals with these autoimmune conditions. In fact, hepatobiliary cancer surveillance is already recommended in some instances, (e.g., in patients with primary sclerosing cholangitis, annual surveillance for gallbladder abnormalities via abdominal ultrasound is recommended).²⁰ Our findings also support previous studies positing a shared immuno-inflammatory pathway for these cancers and suggest a possible modifying effect based on autoimmune condition severity. Further research is needed to elucidate biological mechanisms responsible for autoimmune condition-mediated carcinogenesis and to identify optimal strategies for hepatobiliary cancer surveillance and prevention.

Novelty and Impact:

In the largest study to date on this topic, we investigated associations between hepatobiliary cancers and 39 autoimmune conditions. Novel associations with several autoimmune conditions were observed. The magnitude of associations was particularly strong for primary biliary cholangitis, primary sclerosing cholangitis, pure red cell aplasia, and immune thrombocytopenic purpura. Findings support an immuno-inflammatory etiology and suggest further research may be warranted to assess the need for additional cancer monitoring in individuals with certain autoimmune conditions.

Abbreviations:

US

United States

SEER

Surveillance, Epidemiology, and End Results

ICD

International Classification of Diseases

OR

odds ratio

CI

confidence interval

SD

standard deviation