https://orcid.org/0000-0002-9819-8179
Abstract
Numerous studies have assessed the association of gallstones or cholecystectomy (CE) with risk of colorectal cancer (CRC). However, the findings are mixed.
To systematically review and meta-analyse the association between the presence of gallstone disease (GD), or CE and the incidence of CRC. Secondary endpoints were the risk based on type of exposure, study design, tumour subsites and sex.
PubMed and EMBASE were searched from September 2020 to May 2021. The protocol was registered on the Open Science Foundation Platform. We identified and classified studies according to their design into prospective cohort, population-based case-control, hospital-based case-control and necropsy studies reporting CRC incidence among individuals with diagnosed GD or after CE (or both). Among 2157 retrieved studies, 65 (3%) met the inclusion criteria. We followed the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) reporting guidelines. Data were extracted by two independent reviewers. We evaluated the quality of the study according to the Newcastle-Ottawa Scale and only studies with a score of 6 and above were included in the final analyses. We pooled log-transformed odds ratios/risk ratios from the available adjusted models to estimate a summary relative risk (RR) and 95% confidence interval (CI) in a random-effects model. The primary outcome was overall CRC incidence. We also conducted secondary analyses according to sex and CRC subsites (proximal colon, distal colon and rectum). The outcome was measured by RRs with 95% CIs.
The overall association of GD and/or CE with CRC was RR = 1.15 (1.08; 1.24), primarily driven by hospital-based case-control studies [RR = 1.61 (1.29; 2.01)], whereas a more modest association was found in population-based case-control and cohort studies [RR = 1.10 (1.02; 1.19)]. Most hospital-based case-control and necropsy studies reported estimates that were adjusted for age and sex only, leaving room for residual confounding; therefore we restricted to population-based case-control and cohort studies for our subsequent analyses. Similar associations were found for women [RR = 1.21 (1.05; 1.4) and men (RR = 1.24 (1.06; 1.44)]. When assessed by CRC subsites, GD and CE were primarily associated with higher risk of proximal colon cancer [RR = 1.16 (1.07; 1.26)] but not distal colon cancer [RR = 0.99 (0.96; 1.03)] or rectal cancer [RR = 0.94 (0.89; 1.00)].
Gallstones are associated with a modestly increased risk of colon cancer, primarily in the proximal colon.
There is a modest association between the presence of gallstones or cholecystectomy and the risk of colorectal cancer, and it is present in both women and men.
The proximal colon is the site with the highest risk; there is a no association with the risk for rectal cancer.
Introduction
Most people with gallstone disease (GD) never experience symptoms; nonetheless, for symptomatic patients this disease contributes to a considerable number of outpatient and emergency visits.1 Associations of GD with various conditions such as diabetes and cardiovascular disease have been previously described and gallstones have been associated with numerous gastrointestinal cancers.2–6 A considerable number of studies has been conducted during the past 40 years to ascertain the associated risks of cholecystectomy (CE) and symptomatic or screen-detected gallstones (GD) with colorectal cancer (CRC). Some but not other studies have reported a higher risk for CRC among women after CE or higher risk of cancer in the proximal colon.7–11 Whereas systematic reviews on the subject have been conducted earlier, none has included recently published data from multicentre studies, cohort collaborations with millions of participants, such as the European Prospective Investigation into Cancer and Nutrition (EPIC), or data from underrepresented populations.12–15 Furthermore, prior reviews have examined CE or GD as separate conditions but have not yet assessed their combined associations with CRC. Therefore, we conducted an updated systematic review and meta-analysis to comprehensively characterise the association of gallstones with CRC risk, thereby elucidating possible pathogenesis and informing clinical practice for better prevention of CRC.
Methods
Search strategy and data collection
We performed a systematic literature search in PubMed/MEDLINE and Embase with the search terms [gallstones OR cholelithiasis OR choledocholithiasis OR cholecystitis OR cholangitis OR cholecystectomy] AND [colon cancer OR colorectal cancer] with no date or language limits, between 1 September 2020 and 5 March 2021. We also identified studies cited by the included studies and previous systematic reviews, by hand-searching the relevant references. Study authors and institutions were examined to determine any possibility of dataset overlap. Full-text manuscripts were available for all abstracts; therefore no attempt was made to contact study authors to obtain further data. The protocol was registered with the Open Science Foundation (Registration DOI : 10.17605/OSF.IO/6XVPA). We used the open access online tool CADIMA for offline data extraction.16 Inclusion criteria were observational studies which: (i) reported the exposure of interest as screen-detected gallstones, symptomatic gallstones, gallstone disease or cholecystectomy; (ii) reported the outcome of interest as incidence of CRC; and (iii) provided odds ratios (OR) and/or risk ratios (RR) and the corresponding confidence intervals (CI) or sufficient data to calculate them. The primary outcome was incidence of CRC and the planned secondary analyses were to be based on type of exposure, study design, tumour subsites and sex-related risk. We excluded case series and studies where the outcome was absent within the exposed group (e.g. studies that reported adenomas17–20). The flowchart of the systematic review can be found in Figure 1.
PRISMA diagram of included studies. PRISMA, Preferred Reporting Items for Systematic Reviews and Meta-Analyses
Three co-authors (H.S., F.A.A. and G.P., with previous experience in systematic reviews) independently extracted data on study characteristics from each study, such as the number of the exposed, the number of CRC cases, country name or geographical region, last name of the first author, publication year, study period, follow-up time and ORs/RRs and 95% CIs for the association between each exposure category and CRC. Articles published in languages other than English (German and French) were evaluated after translation. G.P. and H.S. independently extracted data and assessed studies using the Newcastle-Ottawa Scale; and in case of discrepancy, we performed joint re-evaluation. We followed the MOOSE (Meta-analyses Of Observational Studies in Epidemiology) reporting checklist for this review.21,22
Statistical analysis
Quantitative meta-analyses were performed in R software, version 4.1.2 (R Foundation for Statistical Computing) and RStudio version 1.4.1717 (RStudio, PBC). Under the assumption that CRC is a rare outcome, we considered RR and OR to be equivalent. We estimated the random-effect binomial proportions, 95% confidence intervals, and prediction intervals.23 We assessed heterogeneity using I2 and tau, as estimated by the maximum likelihood, and tested heterogeneity using the inverse variance method with the metafor, meta and dmetar R packages.24–26 The I2 cutoffs of 0–25%, 25–50%, 50–75% and 75–100% were used to define low, low-moderate, moderate-high and high heterogeneity, respectively. Publication bias was assessed using funnel plots and Egger regression testing.27,28 Unless otherwise specified, all statistical tests were two-tailed. P-values in the context of heterogeneity measures were based on Wald tests for I2 and DerSimonian-Laird tests for tau, unless otherwise specified. For the primary analyses of overall CRC risk, we only included studies with Newcastle-Ottawa Scale score of ≥6. As mentioned in the registered protocol, we performed tumour subsite- and sex- stratified analyses, due to the increased incidence of GD for women and the often-reported association of CE with proximal CRC. To assess the influence of individual studies and outliers, we conducted leave-one-out meta-analyses plots and meta-analyses without outliers using the dmetar package for R.24,29 To further assess heterogeneity, we ran meta-regression and multiple-meta-regression (sex and subsite) analyses using the metafor R-package.25,29 Taking into account correlated measures in the sex and subsite analyses, we applied robust variance estimation (RVE) with the robumeta R package.30,31
Results
Characteristics of included studies
We identified a total of 65 studies, including 15 population-based case-control studies, 21 cohort studies, 21 hospital-based case-control and eight necropsy studies (prevalence of gallstones at autopsy) published between 1967 and 2020. The total number of CRC cases was 203 799 among 3 509 412 participants (5.8%). The risk of overall CRC was reported in 51 studies; other studies reported tumour subsite- or sex-specific results only; those were not used for the analysis of the overall association with CRC. The majority of studies (n = 48, 74%) described European and US population data, most studies (n = 49, 75%) investigated cholecystectomy as exposure and eight studies examined multiple exposures. The study quality varied considerably, with the Newcastle-Ottawa Scale (NOS) mean of 6.7 (SD: ±1.53, range: 4–9). The characteristics and aggregated outcomes for the included studies are reported in Supplementary Table S1 (available as Supplementary data at IJE online).
Overall risk of colorectal cancer
The overall RR of CRC in relation to any gallstone exposure from the 51 included studies was 1.15 (95% CI: 1.08; 1.24) (Figure 2). There was substantial between-study heterogeneity, with an estimated tau2 of 0.03 and I2 of 80.5% (95% CI: 74.9%; 84.8%). The prediction interval ranged from 0.81 to 1.64. GD showed a stronger association with CRC risk (13 studies, RR = 1.20, 95% CI: 1.03; 1.40, I2 = 91.7%) than CE (33 studies, RR = 1.12, 95% CI: 1.03; 1.21, I2 = 65.3%) or combined exposure (five studies, RR = 1.19, 95% CI: 1.01; 1.39, I2 = 37.7%). The estimates remained at RR >1 in the analysis without outliers for both CE (30 studies, RR = 1.13, 95% CI: 1.05; 1.22, I2 = 71%) and GD (12 studies, RR = 1.22, 95% CI: 1.08; 1.38, I2 = 71%) (Supplementary Figures S1 and S6, available as Supplementary data at IJE online).
Forest plot of relative risk (RR) of colorectal cancer associated with gallstone disease, cholecystectomy, or combined exposures. CE, cholecystectomy; comb., combined exposure; GD, gallstone disease; NOS, Newcastle-Ottawa Scale
When assessed by study design, hospital-based case-control studies showed a stronger association (six studies, RR = 1.61, 95% CI: 1.29; 2.01, I2 = 0%) than necropsy studies (10 studies, RR = 1.22, 95% CI: 1.06; 1.41, I2 = 30.4%) and cohort studies (20 studies, RR = 1.20, 95% CI: 1.09; 1.31, I2 = 76%), whereas no associations were found in population-based case-control studies (15 studies, RR = 0.95, 95% CI: 0.91; 1.00, I2 = 23.4%) (Figure 3).
Forest plot of relative risk (RR) of colorectal cancer associated with gallstone disease, cholecystectomy or combined exposures, according to different study. PCC, population case-control study; HCC, hospital case-control study; CE, cholecystectomy; comb., combined exposure; GD, gallstone disease; NOS, Newcastle-Ottawa Scale
When we analysed population-based case-control studies and cohort studies only (n = 35) we obtained an RR of 1.11 (95% CI: 1.03; 1.20, I2 = 84.6%). We considered data from these two designs only for the subsequent analyses. Furthermore, when analysing cohort studies with a follow-up of at least 10 years, we saw a much higher point estimate for GD (1.32, 95% CI: 1.17; 1.49) than CE (RR = 1.06, 95% CI: 0.97; 1.15) (Supplementary Figure S7, available as Supplementary data at IJE online).
Risk according to subsite
Several studies reported results according to tumour subsites but not in a homogeneous way. Of total of 84 outcomes on different subsites, we managed to obtain estimates for proximal colon cancer (n = 22), distal colon cancer (with or without rectal cancer) (n = 25), all colon cancer (n = 16) and rectal cancer (n = 21) (Figure 4). The RR was 1.19 (95% CI: 1.10; 1.30, I2 = 64.4%) for proximal colon cancer and 1.17 (95% CI: 1.05; 1.30, I2 = 78.7%) for all colon (proximal and distal) cancer, whereas a null association was found for distal colon and rectal cancer combined (RR = 0.99, 95% CI: 0.96; 1.03, I2 = 0%) and an inverse association was found for rectal cancer (RR = 0.94, 95% CI: 0.88; 0.99, I2 = 38.8%, p for heterogeneity among groups <0.0001). After considering the correlation between the studies, the results of the RVE analysis were similar (Supplementary Table S2, available as Supplementary data at IJE online).
Forest plot of relative risk (RR) of colorectal cancer in different subsites associated with gallstone disease, cholecystectomy or combined exposures. PCC, population case-control study; HCC, hospital case-control study; CE, cholecystectomy; com.b, combined exposure; GD, gallstone disease; NOS, Newcastle-Ottawa Scale
Sex-specific CRC risk
When analysed according to sex, we found a similar association of gallstones with CRC risk for women (16 studies, RR = 1.17, 95% CI: 1.00; 1.37, I2 = 81.6%) and men (15 studies, RR = 1.22, 95% CI: 1.04; 1.43, I2 = 67.4%) (meta-regression p for interaction = 0.84, R2 = 0.19% heterogeneity accounted for). When proximal colon cancer was assessed separately, we found that gallstones were more strongly associated with proximal colon cancer for women (RR = 1.28) than for men (RR = 1.14) (multivariable meta-regression p for interaction = 0.14, R2 = 34.3% heterogeneity accounted for). For both sexes, the association of gallstones with distal colon and rectal cancer was null (Supplementary Figure S2a and b, available as Supplementary data at IJE online).
Population-specific CRC risk
We analysed both exposures according to the study population and observed a similar trend in studies including European (12 studies, RR = 1.19, 95% CI: 1.00; 1.41, I2 = 67%) and Asian participants (three studies, RR = 1.19, 95% CI: 0.89; 1.58, I2 = 56%) for cholecystectomy; there were not enough studies of Asian populations for analysis of trends in gallstones. This was not the case for studies of US populations, which presented an RR = 1 (six studies, 95% CI: 0.85; 1.18, I2 = 76%) (Supplementary Figure S8, available as Supplementary data at IJE online).
Publication bias
We detected a considerable publication bias when assessing overall CRC, as indicated by the Egger’s test (intercept of 0.857, 95% CI: 0.04; 1.68, P = 0.046). Similar findings were observed in the analysis according to subsites (intercept of 0.826, 95% CI: 0.21; 1.44, P = 0.01). The funnel plots for both analyses are presented in Supplementary Figure S3 (available as Supplementary data at IJE online). The result of the outlier-excluded analysis for CRC was RR = 1.17 (50 studies, 95% CI: 1.11; 1.23, P <0.0001, I2 = 24.3%), and the leave-one-out meta-analysis did not show any considerable change in the point estimate or the 95% CI for either overall risk or according to subsite, when each of the studies was left out (Supplementary Figure S4, available as Supplementary data at IJE online).
Discussion
In our systematic review we showed a modest increase in risk of CRC associated with the presence of gallstones and/or cholecystectomy. The risk elevation was higher for colon cancer, particularly proximal colon cancer, whereas a marginally inverse association was found for rectal cancer. Gallstones were similarly associated with CRC risk for men and women, although the point estimate for proximal colon cancer was higher for women than men. We observed considerable discrepancies in the methodology of the included studies, where the study designs and follow-up times varied substantially. We also observed an evident publication bias, and a potential for selection bias. Notably, most hospital-based case-control and necropsy studies reported estimates that were adjusted for age and sex only, leaving room for residual confounding.
Our findings consolidate the reported evidence on the modest association of both GD and CE with CRC. Gallstones can be considered as an indirect marker of dysregulation in metabolic pathways, contributing to or even initiating inflammatory processes in disease development, including CRC. Nevertheless, the implied causation mechanisms have yet to be deciphered. In a Mendelian randomization study of the China Kadoorie Biobank, Pang et al. have inferred causal effects of gallstones on various gastrointestinal (GI) cancers, demonstrating a hazard ratio of 1.13 (95% CI: 1.01; 1.29) for CRC associated with GD.32 Higher risk of GD has been correlated with extremely high circulating bilirubin levels, a biomarker associated with GD but also other biliary pathologies.33 Additionally, findings from a case-control study nested in the European Prospective Investigation into Cancer and Nutrition (EPIC) showed positive association of prediagnostic unconjugated bilirubin with CRC for men. In this study, the single nucleotide polymorphism (SNP) mostly associated with an increase in total bilirubin (rs6431625) was also associated with a higher CRC risk for men (OR = 1.07, 95% CI: 1.02; 1.12).34 Contrary to these findings, Culliford et al. in a Mendelian randomization study of the UK Biobank did not find any causal association with CRC risk of either gallstones or circulating bilirubin levels.35 Time-varying associations of GD with CRC have not been completely addressed in these biobank studies. For example, in high-quality cohort studies such as Shabanzadeh et al., the risk of CRC for asymptomatic GD patients increased after 15 years of follow-up whereas Schernhammer et al. reported increased risk even in the first 4 years after cholecystectomy.8,36 This time-varying association is possibly related to a multitude of factors including metabolic disturbances and lifestyles. Furthermore, GD is likely to emerge as a result of dietary and lifestyle exposures, many of which are shared risk factors for CRC, whereas cholecystectomy is an intervention that may affect both local and systemic microbial and metabolic balance. In our subgroup analysis of cohorts with a longer follow-up, we saw that the association with GD was much stronger than that with CE. With many exposures such as lifestyle factors being associated both with GD and CRC, the question remains whether metabolic disturbances causing GD also cause CRC, or if the processes are independent and there is a non-causal explanation for the associations shown.
In our meta-analysis, we found a stronger association of gallstones with proximal colon compared with distal colon CRC and an inverse association with rectal cancer. This finding is compelling, based on the data from the more recent and larger cohort studies, and resonates with the heterogeneity of CRC subtypes across different subsites where proximal tumours have been found to have distinct molecular alterations compared with distal and rectal cancers.37,38 The findings suggest a potential role of gallstones-related pathogenesis in the development of proximal colon cancer. The formation of cholesterol gallstones has already been linked to altered intestinal microbiome.39 Several microbial features (e.g. biofilms and Fusobacterium nucleatum) have been associated with increased risk of CRC, particularly proximal colon cancer.40
Secondary bile acids [predominantly deoxycholic acid (DCA) and lithocholic acid (LCA)] are produced from the bacterial metabolism of bile acids formed in the liver and41–44 have been linked to CRC, particularly in the proximal colon in certain experimental models.40,45–48 Both through their enterohepatic circulation and locally on the intestinal mucosa, these acids can induce tumorigenesis by means of DNA damage via oxidative stress and by immune modulation, particularly of Treg cells.49–52 Studies of post-CE and GD patients’ microbiota have shown higher levels of bile acids and elevated presence of specific bacteria responsible for the production of secondary bile acids.53 Furthermore, individuals exhibit reduced diversity of the microbiome after CE, with Wang et al. also reporting an increase in specific E. Coli strains that have been associated with CRC.54,55 Taken together, these data support the hypothesis that individuals with gallstones may be more susceptible to CRC, particularly proximal colon cancer, due to the changes in the gut microbial composition and function. It should be noted that if the main causal driver is bile acid flow on the intestinal mucosa, its effect might differ between GD and CE patients, as the exposure to bile acids in the colon can be different after gallbladder removal.
Although some studies reported a stronger association of gallstones with CRC risk for women than men,10,56–58 we did not find such a sex difference in the current meta-analysis. Interestingly, in the subsite-stratified analysis, we indeed observed that CE and/or GD were more strongly associated with higher risk of proximal colon cancer for women than men. Although the reasons for such a sex difference remain unknown, women are known to be more susceptible to proximal colon cancer than men and sex-dependent patterns of the gut microbiome have been reported, with sex hormones being a potent driver of differences in the gut microbiome.59,60 Furthermore, in our analysis according to study populations, we observed that US-based studies had an association closest to 1. This finding could be attributed to the higher proportion of cohort versus hospita-based case-control (HCC) studies in the USA, since the association tended to be more modest for the former than latter.
Our study has several strengths. First, we conducted a comprehensive review of the literature, added recent and well-conducted cohort studies and performed detailed analysis according to study design, tumour subsites, and sex. Second, we evaluated and accounted for the heterogeneity across studies through rigorous analysis and thorough reporting. Our study’s limitations are primarily related to the high heterogeneity of the included studies and potential publication bias. Although the heterogeneity among the hospital case-control studies was low, these as a group reported much higher estimates than the other types of studies. To minimize the influence of these issues, we carefully assessed the quality of the included studies and focused on the high-quality studies in our analysis. Another limitation of our study is that most of the included studies relied on self-reported data and primarily at baseline only, leaving room for exposure misclassification and precluding the assessment of the time-varying effect of gallstones and their treatment on CRC risk.
Conclusion
In conclusion, results of this systematic review and meta-analysis demonstrate that there is a modest association between gallstones and higher risk of colon cancer, particularly proximal colon cancer. In the presence of significant heterogeneity, these results should be interpreted with caution and serve as an impetus for further research.
Ethics approval
Not required for this study.
Data availability
The data underlying this article will be shared on reasonable request to the corresponding author.
Supplementary data
Supplementary data are available at IJE online.
Author contributions
GP: study design, systematic review, data extraction, risk of bias analysis, quantitative analysis, manuscript drafting and editing. HS: systematic review, data extraction, risk of bias analysis, manuscript editing. FAA: systematic review, data extraction, risk of bias analysis, manuscript editing. SP: quantitative analysis, manuscript drafting and editing. EG: study design, manuscript drafting and editing. MS: study design, quantitative analysis, manuscript drafting and editing.
Funding
This work was supported by the German Research Foundation [Deutsche Forschungsgemeinschaft (DFG) project number 426308975 to GP]. The funding agency had no role in the design and conduct of the study.
Acknowledgements
The authors would like to thank Dr Ilias Dodos for his assistance with the cross-evaluation of articles in French.
Conflict of interest
None declared.
