Article Text

Download PDFPDF

Risk of cancer in patients with bile acid diarrhoea: a Danish nationwide matched cohort study
  1. Nynne Nyboe Andersen1,
  2. Signe Wildt2,
  3. Aske Thorn Iversen3,
  4. Gry Poulsen3,
  5. Tine Jess3,
  6. Lars Kristian Munck1,
  7. Christian Borup1
  1. 1Department of Gastroenterology, Zealand University Hospital Koge, Koge, Denmark
  2. 2Department of Gastroenterology and Hepatology, Hvidovre Hospital, Hvidovre, Denmark
  3. 3Center for Molecular Prediction of Inflammatory Bowel Disease, Department of Clinical Medicine, The Faculty of Medicine, Aalborg University, Copenhagen, Denmark
  1. Correspondence to Dr Nynne Nyboe Andersen; nynne{at}nyboeandersen.com

Abstract

Objective Bile acid diarrhoea is a common cause of chronic diarrhoea. Increased levels of potentially carcinogenic bile acids in faeces, theoretically, may increase the risk of colorectal cancer in particular, but the long-term disease course is unknown. We aimed to investigate the overall and site-specific cancer risk in bile acid diarrhoea.

Design Adult patients with bile acid diarrhoea were identified using nationwide Danish registries from 2003 to 2020 by a diagnostic gold-standard 75-selenium tauroselcholic acid procedure followed within 6 months by sequestrant prescription. The risk of overall and site-specific cancers in cases with bile acid diarrhoea was compared with sex, age and comorbidity-adjusted matched controls. A competing risk model estimated cumulative incidence functions and cause-specific HRs.

Results We identified 2260 patients with bile acid diarrhoea with a mean follow-up of 5.5 years (SD 4.2). The overall cancer risk was increased by an HR of 1.32 (95% CI 1.12 to 1.54). The risk of site-specific cancer was increased in 3 of 10 cancer groups: haematological, HR 2.41 (1.36 to 4.02); skin, HR 1.33 (1.01 to 1.71); and male genital cancers, HR 1.85 (1.11 to 2.92). No increased risk of colorectal cancer was detected in patients with bile acid diarrhoea, HR 0.73 (0.34 to 1.63).

Conclusions Bile acid diarrhoea was associated with an increased overall risk of cancer, especially haematological cancers, but the risk of colorectal cancer was not increased. The lack of a diagnostic code for bile acid diarrhoea and potential residual confounding are limitations, and the findings should be replicated in other cohorts.

  • CANCER
  • BILE ACID
  • DIARRHOEA
  • IRRITABLE BOWEL SYNDROME

Data availability statement

All data relevant to the study are included in the article or uploaded as supplemental information.

http://creativecommons.org/licenses/by-nc/4.0/

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.

Statistics from Altmetric.com

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

WHAT IS ALREADY KNOWN ON THIS TOPIC

  • One-third of patients diagnosed with diarrhoea-type irritable bowel syndrome actually have bile acid diarrhoea.

  • Levels of potentially carcinogenic bile acids are elevated in faeces.

  • This may, in theory, increase the risk, particularly of colorectal cancer.

  • The long-term risk of cancer is unknown.

WHAT THIS STUDY ADDS

  • Patients with bile acid diarrhoea had a 32% increased risk of overall cancer driven by haematological, skin, and male genital cancers.

  • The risk of colorectal cancer was not increased.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

  • The observed increased risk of three site-specific cancers needs replication.

  • Routine screening for colorectal cancer seems sufficient in bile acid diarrhoea.

Introduction

Bile acid diarrhoea (BAD) is found in an estimated 1% of the general population.1 Increased concentrations of dihydroxy bile acids in the colon trigger peristalsis and watery secretion, leading to watery diarrhoea.2–4 This condition may be primary, due to hepatic overproduction of bile acids exceeding the small bowel capacity to reabsorb bile acids, or secondary to dysfunction or resection of the terminal ileum.3 5 Additionally, changes in the bile acid profile from the interaction with the gut microbiota may have a role.4 6–8 The gold-standard diagnostic test for BAD is the 75-selenium tauroselcholic acid (SeHCAT) test, widely available in Denmark.9 BAD is treated with sequestrants that bind bile acids to hinder their perturbation of the colonic mucosa.10

The epidemiology of BAD has been studied using retrospective registers from single or a few centres combined, but these studies provide no information on long-term outcomes.11–13 As bile acids are amphiphilic molecules, alterations in the composition of the bile acid profile and distribution are, in theory, carcinogenic.14–16 Therefore, conditions such as post-cholecystectomy and BAD with altered distribution and enterohepatic circulation of bile acids may be associated with elevated cancer risk, particularly of the digestive system. This hypothesis is supported by a Korean population-based study of patients with a history of cholecystectomy showing an increased incidence of colorectal and liver cancer after cholecystectomy compared with the general population.17

We, therefore, aimed to investigate the risk of cancer in patients with BAD using nationwide population-based Danish registers.

Methods

Study population

The source population included all Danish citizens identified through the Danish Civil Registration System containing registration by a unique personal identification number (CPR number). The CPR number allows linkage to other national registers, including the Danish National Patient Registry, the Danish National Prescription Registry, and the Danish Cancer Registry.18

Exposure information

An International Classification of Disease (ICD) diagnostic code for BAD was not included in the Danish National Patient Registry before 2022. Consequently, we used a proxy exposure definition based on the SeHCAT test procedure code (WGLGSSEXX; data available from 2003 and onwards) followed within 6 months by a prescription of a sequestrant (identified by Anatomical Therapeutic Chemical code C10AC in the Danish National Prescription Registry) to indicate the SeHCAT test was positive. Data from the Danish National Prescription Registry are valid and complete from 1995 onwards. The prescription date was defined as the time of diagnosis and used as the index time. We included patients with a SeHCAT test performed between 2003 and 2020. Patients 18–100 years old who lived in Denmark 5 years before the SeHCAT test were eligible.

Patients with a prevalent diagnosis of cancer were excluded. To avoid misclassification bias, we excluded patients with other potential indications for sequestrant treatment including primary biliary cholangitis (ICD-10 codes: K74.3, K74.4, K74.5), primary sclerosing cholangitis (ICD-10 code: K83) or intrahepatic cholestasis of pregnancy (ICD-10 code: O26.6). Inflammatory bowel disease potentially increases the risk of cancer, and patients with bowel resection due to Crohn’s disease have a high prevalence of BAD.14 Cholecystectomy may also increase the risk of certain cancers and is associated with BAD.14 18 19 Hence, both patients with inflammatory bowel diseases (ICD-8: 563.01, 563.02, 563.08, 563.09, 563.19, 569.04; ICD-10 codes: K50, K51) and patients with a history of cholecystectomy (procedure codes: 47360, 47365 until 1995 and KJKA20, KJKA21 after 1995) were excluded.

Matching

We used exact matching of individuals with BAD in a 1:10 ratio without replacement to individuals in Denmark who were alive on the index date, of the same sex, born within 10 days of the case and with no registered sequestrant use.

The matched controls were given the same index date as the corresponding patient.

Outcome information

Cancer diagnoses were identified by the Danish Cancer Registry and classified according to ICD-10. Any incident cancer was analysed as a primary outcome and also in 1 of 10 site-specific cancer groups, including digestive organs (C15-17, C21-26), colorectal (C18-20), respiratory (intrathoracic organs (C30-39, C450)), non-melanoma skin cancer and malignant melanoma (C43–44), breast (C50), female genital organs (C51–58), male genital organs (C60–63), urinary tract (C64–68), lymphoid/haematopoietic tissue (C81–90, C91–96, D46), and other (C40–41, C45 (not C450), C46–49, C69–80, C97, C00–14). Only the first diagnosed cancer was included in the analyses.

Confounder assessment

We adjusted for concomitant disease burden by Charlson’s Comorbidity Index (CCI).20 Information on diagnoses prior to the index date was extracted from the Danish National Patient Register to calculate the CCI.

Data on important potential confounders, obesity, smoking habits, and alcohol consumption were unavailable. The impact of these confounders is to some extent captured by adjusting for the CCI. However, in an attempt to further adjust for smoking, the final analyses were adjusted for chronic obstructive pulmonary disease (COPD) as a proxy for cumulative smoking exposure.

Additionally, we calculated an e-value to address the impact of unmeasured confounding.21 The e-value is the minimum strength of association on the relative risk scale that a confounder would need with both treatment and outcome to fully explain the observed association, conditional on the measured covariates. Accordingly, the e-value helps to interpret the strength of evidence of a true association. The higher the e-value, the more difficult it is to attribute the observed results to an unmeasured covariate.

Statistical analyses

All BAD cases and their corresponding matches were followed from 180 days after the index date until a diagnosis of any cancer, death, emigration, or end of follow-up on 31 December 2020, whichever occurred first. In addition, matches were censored if they received a prescription for a sequestrant. A lag period of 180 days was used to avoid including incipient cancers not likely associated with the exposure.

Descriptive analyses were performed to characterise the populations, and Χ2 tests evaluated differences.

Our analyses were challenged by competing risk of death and other site-specific cancers. The methods may lead to biased or non-interpretable results when using the Kaplan-Meier estimator in the presence of a significant and related competing risk. Consequently, we used cumulative incidence functions and the effect of covariates was modelled using a cause-specific HR where death before cancer and other site-specific cancers were censoring events. Differences in the cumulative risk functions were evaluated by Gray’s test for equality.

The cause-specific HRs were calculated using Cox regression models with time since the index as the underlying variable calculating HRs for cancer with 95% CIs.

The assumption of proportional hazards was tested graphically and indicated no violation of the assumption. All analyses were performed using SAS V.9.4 (SAS Institute) and all statistical tests were two sided and considered significant if 95% CIs did not overlap.

Patient and public involvement

Patient groups and the public were not involved in the planning of this research.

Results

In total, 11 386 SeHCAT tests were performed between 2003 and 2020, and 4576 (40%) patients had a sequestrant prescription within 180 days. We excluded 2295 patients aged less than 18 years, not residing in Denmark, or having a history of cancer, biliary disorders, inflammatory bowel disease or cholecystectomy (figure 1). This left 2260 cases with BAD matched with 22 600 controls from the background population. The mean age at diagnosis was 49.3 (SD 16.4) years, 56% were female, and the mean follow-up duration was 5.5 (SD 4.2) years. 53% of cases were indexed in the most recent time period of 2015–2020 (table 1).

Figure 1

Study cohort formation. SeHCAT, 75-selenium tauroselcholic acid.

Table 1

Characteristics of patients with bile acid diarrhoea and their matched controls

Risk of overall cancer

In patients with BAD, 183 cancer events occurred during 12 072 person-years of follow-up compared with 1417 events during 124 114 person-years of follow-up in corresponding matches. When only adjusting for the demographic characteristics of sex, age, and time period, the resulting demographically adjusted overall cancer risk was increased with an HR of 1.34 (95% CI 1.14 to 1.55). The estimate did not change when further adjusting for CCI and COPD, giving a fully adjusted HR of 1.32 (95% CI 1.12 to 1.54). See details in table 2 and figure 2 for the cumulative incidence function curve for overall cancer. In a sensitivity analysis omitting non-melanoma skin cancer in the main estimate, the risk of any cancer in BAD had a fully adjusted HR of 1.32 (1.09, 1.58) compared with the matched controls. Based on the p values, no significant effect modification by sex was detected regarding the risk of any cancer (table 3).

Figure 2

Cumulative incidence function for any cancer comparing patients with bile acid diarrhoea with matched controls. Due to few observations, patient numbers at risk after 15 years of follow-up are not presented.

Table 2

Number of cancer events, person-years (pyrs) of follow-up, incidence rate, and adjusted HR for cancer in individuals with BAD compared with matched controls

Table 3

Effect modification analysis according to sex

Site-specific cancers

The risk of site-specific cancer was significantly increased in 3 of 10 cancer groups: cancers of the haematological system (fully adjusted HR 2.41; 95% CI 1.36 to 4.02), skin (fully adjusted HR 1.33; 95% CI 1.01 to 1.71), and cancers of the male genital organs (fully adjusted HR 1.85; 95% CI 1.11 to 2.92). The risk of cancer in the respiratory system was significantly increased in demographically adjusted analyses (HR 1.70; 95% CI 1.07 to 2.56), but when further adjusted for CCI and COPD, the risk was attenuated and no longer significantly increased (fully adjusted HR 1.39; 95% CI 0.87 to 2.11).

Of note, no significantly increased nor decreased risk of colorectal cancer was observed with a fully adjusted HR of 0.73 (95% CI 0.34 to 1.36).

No significant associations were detected between BAD and the remaining site-specific cancer groups (table 2).

The impact of unmeasured confounding

Using the main estimate, 1.32 (95% CI 1.12 to 1.54), gave an e-value for the point estimate of 1.97 and 1.49 for the CI. Thus, an unmeasured confounder could negate the main estimate with a risk ratio of 1.97 for the exposure–confounder relationship and a risk ratio of 1.97 for the confounder–outcome association. An unmeasured confounder with a risk ratio of 1.49 for the exposure–confounder and confounder–outcome association could change the lower CI bound to include 1 after adjustment.

Discussion

This nationwide matched cohort study of 2260 Danish patients with BAD compared with 22 600 matched controls found a 32% increased risk of any cancer in patients with BAD. Analyses of site-specific cancers showed an increased risk of cancer in the haematological system, skin, and male genital organs. No increased risk of colorectal cancer was observed.

The study of bile acids and their interaction with diet and microbiota in relation to cancer is not new and has been reviewed.16 19 22 A combination of mechanisms seems to be responsible for the bile acid-induced carcinogenesis in organs with elevated specific bile acids, including oxidative damage to DNA, inflammation, activation of NF-κB and enhanced cell proliferation.16 22 However, the physiological functions of both primary and secondary bile acids are complex and not fully understood. Recent data even suggest that specific bile acids may have biphasic effects with carcinogenic properties in high concentrations as opposed to anti-cancer effects in low concentrations.23 In patients with BAD, faecal concentrations of primary bile acids are significantly elevated; however, an altered serum profile of bile acids has also been shown to a much lesser degree with reduced levels of secondary bile acids and increased levels of primary bile acids.8 24 The observed increase in overall cancer rate associated with BAD could be a consequence of this altered bile acid composition. Another theoretical explanation could be that sequestrant treatment, which was part of our case definition, may in itself be carcinogenic. However, to the authors’ knowledge, there is no study in humans to support a putative carcinogenic effect of sequestrants, and a study in mice did not show increased rate of tumours.25 Vitamin D deficiency has been suggested as a risk factor for cancer and sequestrant-induced deficiency of fat-soluble vitamins including vitamin D in patients with BAD may play a role.26 However, this hypothesis is not supported by studies of vitamin D supplementation and benefit on cancer risk.27 The same is true for deficiency of vitamin E.28

The observed increased cancer risk was, surprisingly, not driven by an increased risk of colorectal cancers. A recent meta-analysis based on 18 cohort studies suggested that cholecystectomy-driven alterations in bile acid composition was associated with right-sided colon cancer but not colorectal cancer overall.29 Unfortunately, our study was underpowered to stratify according to colonic segments and we were thus unable to detect any potential altered risk in site-specific colorectal cancers. Additionally, our patient population with BAD were inherently treated with sequestrants, as part of the case definition, which bind bile acids, potentially preventing their carcinogenic effect on the colonic epithelium.

A skewed distribution of colonoscopy could have an impact on study findings regarding the incidence of colorectal cancer. A total of 1171 (51.8%) of the 2260 patients with BAD underwent colonoscopy in the 2 years preceding time of index compared with only 2.8% of the matched controls (table 1). This surveillance bias could reduce the number of observed cases with colorectal cancer in the patients with BAD; the question is by how much. A pragmatic randomised controlled trial reports the effect of screening colonoscopy in 84 585 individuals in Sweden, Norway, Holland, and Poland. Screening colonoscopy gave an 18% relative risk reduction and a mean of 455 (95% CI 270 to 1429) colonoscopies were needed to prevent one incident case of colorectal cancer.30 Based on the lower CI bound of 270, an estimate of about four cases of colorectal cancer may have been prevented by the 1171 colonoscopies performed in the patients with BAD. Following this line of thought, our observed total BAD cases with colorectal cancer would increase to 13, and eventually the crude estimated incidence rate of colorectal cancer would increase to 1.07, similar to the rate of 1.04 observed in the matched controls. Although our data on colorectal cancer are confounded by the skewed prevalence of preceding colonoscopy, we consider the assessed impact too small to significantly change the finding of no increased risk of colorectal cancer in patients with BAD.

The findings of increased risk of cancers in the haematological system, skin, and male genital organs associated with BAD were unexpected. The altered bile acid composition in patients with BAD may play a role, but the link is unclear and these findings should be replicated in other cohorts.

The diagnosis of BAD is commonly missed or mistaken for irritable bowel syndrome. A systematic review and meta-analysis found a 32% prevalence of BAD among patients with diarrhoea-type irritable bowel syndrome.10 The increased risk of cancer that we found emphasises that recognition and diagnosis of BAD to be discerned from irritable bowel syndrome are warranted. Recent research in blood biomarkers could, if validated, enable a shift towards a screening strategy for BAD among patients presenting with relevant symptoms, as many patients without such screening could be classified as having irritable bowel syndrome or functional diarrhoea.24 31 This would be welcomed, as the evidence base for treatments targeting BAD is growing.32 33 Finally, recognising BAD as a disease, including routine clinical diagnostic coding reaching registries, would advance efforts to understand the long-term disease course.

The primary strength of this study is the nationwide register-based design, providing a large cohort of patients with BAD with minimal loss to follow-up and high external validity. Based on the prescription of sequestrants after SeHCAT testing, we found a 40% rate of positive tests, similar to the positive rate in previously published SeHCAT cohorts.10 13 24

Several limitations should be acknowledged. Most importantly, our case definition of BAD was based on a SeHCAT test followed by the prescription of a sequestrant within 6 months. This proxy definition was needed because no diagnostic code for BAD was available in Denmark before 2022. Attempts were made to strengthen the definition by excluding patients treated with sequestrants for other indications, and with the temporal relation of a SeHCAT test and the prescription of a sequestrant required, we consider the case definition specific for BAD. Some physicians with poor access to SeHCAT testing consider a positive empirical response to sequestrant diagnostic, and we could have included such cases. However, we recently reported the false positive rate of such an approach as high as 35% and such misclassification would dilute out differences between cases and controls.24 Lastly, although we used nationwide registries, the number of patients with BAD and associated cancer events was relatively low. An expert opinion estimates a 1% prevalence of BAD by extrapolating data from cohorts of diarrhoea-predominant irritable bowel syndrome undergoing SeHCAT testing.1 Our included number of cases reflects a prevalence about 0.1% in Denmark using the same case definition.34 Some patients with undiagnosed BAD could have been included as controls of this study, but we deem the impact of such misclassification bias negligible. Increasing awareness of BAD means a higher number of diagnosed cases in recent years, so more data should be available for epidemiological studies of BAD in the near future.34 We encourage researchers with access to similar data from other countries to investigate long-term outcomes of BAD.

Several efforts were made in the interest of the internal validity of the analyses to minimise confounding, among others, using a 6-month lag period excluding incipient cancers detected during diagnostic work-up, excluding patients with a history of cholecystectomy or inflammatory bowel disease and lastly, applying an age and sex-matched design adjusted for comorbidities. Unfortunately, data on lifestyle factors, smoking, and obesity were not available. Obesity is prevalent in patients with BAD,24 35 and an altered bile acid homeostasis has been described in non-alcoholic fatty liver disease.36 However, adjusting for the CCI as an overall proxy would account for some of the burden of these specific disease entities, and an additional analysis with adjustment for COPD as a proxy for cumulative smoking exposure did not significantly change the results. To assess the needed magnitude of unmeasured confounding to affect the primary result, we calculated an e-value. No data are currently available on the magnitude of known confounders between cancer and BAD to put this in context, but future data may be applied. Overall, our strategies to address confounding factors and sensitivity analyses testing the robustness of our results do not suggest that the observed results are notably biased.

In conclusion, based on Danish nationwide population-based data, BAD was associated with an increased risk of overall cancer. No increased risk of colorectal cancer was detected. The lack of a diagnostic code for BAD and the potential for residual confounding are important limitations, and these novel findings should be replicated in other large cohorts.

Data availability statement

All data relevant to the study are included in the article or uploaded as supplemental information.

Ethics statements

Patient consent for publication

Ethics approval

The study followed the rules of the Danish Data Protection Agency. According to Danish law, approval from the Ethics Committee is not required for registry-based research.

References

Footnotes

  • X @ChrisBorup

  • Contributors NNA—study concept and design; analysis and interpretation of data; drafting of the manuscript; critical revision of the manuscript for important intellectual content. CB—study concept and design; analysis and interpretation of data; drafting of the manuscript; critical revision of the manuscript for important intellectual content. LKM—study concept and design; interpretation of data; critical revision of the manuscript for important intellectual content. ATI—study concept and design; acquisition of data; analysis and interpretation of data; critical revision of the manuscript for important intellectual content. GP—study concept and design; acquisition of data; analysis and interpretation of data; critical revision of the manuscript for important intellectual content. TJ—study concept and design; acquisition of data; interpretation of data; critical revision of the manuscript for important intellectual content. SW—study concept and design; interpretation of data; drafting of the manuscript; critical revision of the manuscript for important intellectual content. All authors approved the final version of the article, including the authorship list. Data and study material will not be made available to other researchers. NNA—Guarantor of article .

  • Funding Danish Cancer Society (grant/award number: R327-A19007).

  • Competing interests SW has served as an advisory board member for AbbVie and Bristol-Myers Squibb and received financial fees for teaching and consulting from Tillotts and Bristol-Myers Squibb. TJ has received financial fees for consulting from Ferring.

  • Provenance and peer review Not commissioned; externally peer reviewed.