Introduction
Diabetes mellitus (DM) is an established carcinogenesis risk for biliary tract cancers, i.e., cholangiocarcinoma (CCA) and carcinoma of the gallbladder.1,2 The increased risk of CCA, both intrahepatic (iCCA) and extrahepatic (eCCA), and gallbladder malignancy are reported to be associated with DM in several observational studies, as well as in meta-analyses.3,4 The association between DM and increased risk of CCA is speculated to result from the effects of hyperglycemia, insulin, insulin-like growth factor, and their receptors.5–7 Recently, some anti-diabetic medications have been reported to potentially modify CCA risk in patients with DM.8,9 Several observational studies have been carried out to determine the effect of different anti-diabetic drugs on CCA risk modification. While exogenous insulin and sulfonylurea usage are not associated with an increased risk of CCA, incretin-based therapy was shown in one cohort study in the United Kingdom to have a modest association with the increased CCA risk in DM patients.8 In contrast, the effects of incretin-based therapy are still controversial. Another case-control study conducted in Italy reported a null effect of incretin-based therapy on CCA development, neither in the patients using dipeptidyl peptidase-4 inhibitor nor in those using glucagon-like peptide 1 (GLP-1) receptor agonist.10 In addition, preclinical studies of exendin-4, a GLP-1 receptor agonist, indeed inhibited CCA cell growth both in vitro and in vivo.11,12 Contrariwise, DM patients using metformin were found likely to benefit from decreased CCA risk in both case-control and cohort studies.9,13 These conflicting data have led to extensive epidemiological and molecular-level studies of metformin’s effects on CCA development.
Metformin (N, N-Dimethylbiguanide), classified as a biguanide anti-diabetic drug and previously recommended as a first-line drug in type 2 DM treatment by the American Association of Diabetes, is extensively used worldwide.14 Metformin’s primary effect is to activate the AMP-activated protein kinase thus in turn inhibiting the gluconeogenesis of hepatocytes. This effect prevents hyperglycemia in patients with DM and also helps sensitize insulin receptor signaling.15
The association between metformin use and decreased risk of hepatobiliary cancers has been studied in both hepatocellular carcinoma and BTC. One case-control study in the United States found that metformin was associated with a 60% reduced risk of iCCA in DM patients.9 Molecular studies, both in vitro and in vivo, also support findings that metformin exerts a potent effect on the growth and aggressive phenotypes of CCA cells, and might be associated with prolonging survival in DM patients treated with metformin.16,17 The anti-tumor effects of metformin appear broad and effective for both liver fluke- and non-liver fluke-associated CCA. Currently, metformin is registered for a clinical trial study on isocitrate dehydrogenase (IDH)-1 and IDH-2 mutation solid tumors, including iCCA.18 However, over a wider range of BTCs, including eCCA and gallbladder carcinoma, metformin is not consistently effective in the prevention of BTC in DM patients.19–24 In addition, BTC patients taking metformin for their DM treatment showed a discrepant effect among different populations.25–27 Available systematic reviews and meta-analyses also show inconsistent results.28,29 Thus, the benefit of using metformin for BTC prevention and as an add-on treatment, remains debatable and inconclusive.
This meta-analysis aimed to determine the effect of metformin on the prevention of BTC development among DM patients, as well as, the therapeutic benefit for patients who had BTC and were receiving metformin concurrently for DM. This clarification of metformin’s effect will guide further translational and clinical studies as well as a prescribing regimen for anti-diabetic medication in patients with BTC.
Materials and methods
Data source and search strategies
All articles were searched for in PubMed, Web of Science, and Embase from their respective inceptions up to February 28, 2023, using the following search terms; [(metformin) or (antidiabetic)] and [(cholangiocarcinoma) or (bile duct cancer) or (biliary tract cancer) or (biliary carcinoma)] without language restriction. An additional manual search was also performed.
Inclusion and exclusion criteria
The epidemiological observational studies, both cohort and case-control, were included if they met the following criteria; 1) patients with diagnosed CCA or BTC, 2) metformin was prescribed for DM treatment in patients and, 3) relative risk; including Risk Ratio (RR), odds ratio (OR), hazard ratio (HR), were reported for BTC development, or overall survival of patients with BTC. The following studies were excluded; 1) non-clinical studies, in vitro and animal studies, conference abstracts, and reviews, 2) studies involving only gallbladder carcinoma which has a different etiology from other BTC and, 3) RR, OR, or HR and 95% confidence intervals (95% CI) were not reported for CCA, BTC risk development, or overall survival.
Data extraction and quality assessment
Screening and selecting of eligible studies were performed by two independent researchers (SS and CS) using Covidence software (Melbourne, Australia). First author, publication year, region, the subtype of CCA and other BTC, sample size, RR, OR and HR, were obtained. The discrepancies between the process of study selection were solved by consensus discussion with CCA research expert co-authors (SW and WS). The quality of included studies was evaluated using the Newcastle-Ottawa Scale.
Statistical analysis
The meta-analysis of all included studies was performed using RevMan 5.4 (Nordic Cochrane Centre, Copenhagen, Denmark) by using an inverse variance method, and the pooled RRs were then approximated. OR, RR, or HR, and 95% CI were used to compare outcomes between metformin and non-metformin groups. Statistical heterogeneity was assessed by I2. The combined estimates were calculated and pooled under a random-effects model regardless of heterogeneity. Statistical significance was set at p < 0.05.
Results
Literature search and eligible studies
The literature search identified a total of 130 references from PubMed, 3,615 from Web of Science, and 2,097 from Embase. After removing duplications, 1,005 reports were included for screening. After in-depth evaluation 9 studies were eligible for meta-analysis. The PRISMA literature selection process is depicted in Figure 1. Table 1 shows the characteristics of studies included in the meta-analysis.9,13,19,21,22,24–27
Table 1Characteristics of included studies
Study (Ref) | Region | Study design | Study period | No. of subjects |
---|
Chaiteerakij et al, 20139 | USA | Case-control | 2000–2010 | 1206 |
de Jong et al, 201719 | Netherland | Cohort | 1998–2011 | 57,621 |
Oh and Song, 202024 | South Korea | Cohort | 2011–2015 | 66,627 |
Tseng, 202013 | Taiwan | Cohort | 1999–2005 | 304,224 |
Sookaromdee and Wiwanitkit, 202021 | Thailand | Case-control | NR | 18,547,869 |
Marcano-Bonilla et al, 202222 | Sweden | Cohort | NR | 5,760,482 |
McNamara et al, 201526 | Canada | Cohort | 1987–2013 | 913 |
Yang et al, 201625 | USA | Cohort | 2001–2012 | 250 |
Casadei-Gardini et al, 202127 | Italy | Cohort | 2005–2020 | 537 |
Risk of bias for methodology assessment
All included studies were evaluated with Newcastle-Ottawa Scale tool and ranked with more than 6 stars, indicating the low bias risk in each study. The results for the risk of bias for the methodology are shown in Table 2.9,13,19,21,22,24–27
Table 2New Castle-Ottawa Scale (NOS) for the assessment of the included studies
| Risk
|
---|
Author (year) | Selection | Comparability | Outcome | Total |
---|
Case-control study | | | | | |
1 | Chaiteerakij (2013)9 | *** | ** | *** | 8 |
2 | Sookaromdee and Wiwanitkit (2020)21 | *** | ** | ** | 7 |
Cohort study | | | | | |
1 | de Jong (2017)19 | **** | ** | *** | 9 |
2 | Oh and Song (2020)24 | *** | ** | *** | 8 |
3 | Tseng (2020)13 | **** | ** | *** | 9 |
4 | Marcano-Bonilla (2022)22 | **** | ** | *** | 9 |
Effects of metformin on the prevention of BTC
Among 9 included studies, 6 studies (2 case-control and 4 cohort studies) reported the effect of metformin on the development of BTC in DM patients. All case-control and cohort studies were included in a final meta-analysis with a pooled sample of 24,743,526 subjects. One study by de Jong et al included 2 different sub-cohorts according to designs for their analyses and thus both were included.19 Significant heterogeneity among all studies was observed with an I2 of 98%, p < 0.001, thus the random-effect model for meta-analysis was subsequently used. The results showed that metformin did not possess any preventive effect on BTC development, with pooled RR of 0.82 (95% CI: 0.42–1.59, p = 0.56). A forest plot of this meta-analysis is shown in Figure 2a.
As a highly significant heterogeneity was observed among all studies, a sub-group analysis based on the research designs of the original studies was also conducted. Meta-analysis of 4 cohort studies was done with a pooled population of 6,188,954 subjects. A heterogeneity among the studies remained with an I2 of 91%, p < 0.001, with the random effect model being used for meta-analysis. Metformin showed a null effect on risk modification of BTC with a pooled RR of 0.81 (95% CI: 0.48–1.37, p = 0.44) (Fig. 2b). Sub-group analysis of 2 case-control studies also showed a great heterogeneity with an I2 of 98%, p < 0.001. Random effect model meta-analysis was used to analyze the effects of metformin on the risk of BTC in a pooled population of 18,554,572 subjects. No effect of metformin on the risk of BTC was observed (RR: 0.69, 95% CI: 0.06–7.47, p = 0.76) (Fig. 2c).
Effects of metformin on survival outcome of patients with DM
The effect of metformin on the survival of patients with BTC was further analyzed in another 3 studies with a pooled population of 1,700 individuals. A minimal and non-significant heterogeneity among the studies was observed with an I2 of 7%, p = 0.34. By the random effect meta-analysis, metformin showed a marginal benefit for patients with BTC, who were prescribed this medication for their concurrent DM, with a pooled RR of 0.83 (95% CI: 0.68–1.01, p = 0.07) (Fig. 3a). The marginal beneficial effect of metformin on BTC patients’ survival was also consistent in the fixed effect model meta-analysis with a pooled RR of 0.83 (95% CI: 0.68–1.00, p = 0.05) (Fig. 3b).
Discussion
The association of DM and increased BTC risk, especially CCA, has been consistently reported in several studies over the past decade.3,4,9 However, the effect of anti-diabetic medication on the modification of CCA risk is still controversial among different DM drug treatment groups.2 The authors of the current study have reviewed research that showed metformin has a promising role in CCA prevention in DM patients as well as a potential role for add-on treatment in patients with CCA.2 A breakthrough case-control study in the United States by Chaiteerakij et al reported an association between taking metformin and a 60% reduced CCA risk in patients with DM.9 In addition, the results of many preclinical studies on the molecular mechanisms underlying the inhibitory effects of metformin on CCA cells also support the epidemiological observations.16,17,30–32 However, observational studies in other regions are inconsistent with Chaiteerakij et al’s findings.19,20 Since metformin seemed highly promising for repurposing as a CCA chemoprevention agent, as well as, an add-on medication for CCA treatment, this systematic review and meta-analysis also investigated metformin’s effectiveness globally.
Our literature search across 3 databases, identified 6 observational studies (2 case-controls and 4 cohorts) that reported the relative risk of BTC and/or CCA development. Two case-control studies reported the ORs and showed the opposite outcomes of metformin on CCA risk to each other.9,21 Notably, these 2 studies were conducted in different regions where the known risk factors are different.33,34 The other studies were cohort studies that reported the HR for the development of a group of BTC, including all subtypes of CCA (iCCA and eCCA) as well as gallbladder carcinoma, based on International Classification of Disease (ICD) coding systems used in the primary databases.13,19,20 All 6 reports (case-control and cohort studies), with one study possessing 2 sub-cohorts, were included in the final meta-analysis.19 This meta-analysis showed that metformin was not associated with a modified risk for BTC development. Sub-group analyses classified by the research designs of the original studies (case-control vs. cohort studies) also consistently showed that metformin was not associated with a modification of BTC risk among the included population. Significant heterogeneity was observed among all studies. This could be due to the biological heterogeneity of cancer subtypes all grouped as biliary tract cancers in the original studies, e.g., gallbladder carcinoma has a different etiology to CCA and is more aggressive.35,36 Even in the CCA group, the iCCA and eCCA could also originate from different cell types and be associated with different risk factors.37 These factors are potential confounders in our analysis and need to be considered in interpreting the results. As the original studies did not report relative risk for each BTC subtype, sensitivity analysis, could not be carried out in the present meta-analysis.
The add-on therapeutic effects of metformin in BTC patients receiving standard treatment were also meta-analyzed. From 3 included studies with minimal and non-statistically significant heterogeneity, metformin showed null effects on overall survival in 2 cohorts, and another cohort showed a benefit on overall survival in BTC patients who were prescribed metformin.25–27 The meta-analysis thus showed a marginal benefit of metformin on the overall survival of BTC patients. Although metformin is very promising for BTC treatment in preclinical studies, the lack of efficacy and discrepancies in metformin’s effects on BTC between preclinical and human studies may result from several factors as discussed in a recent review by the current authors.38 First, the in vitro studies used a relatively high dose of metformin at a millimolar scale which could increase the risk of adverse effects in humans at the same concentration. Thus in vitro dosage may not be directly translatable. Second, metformin seems beneficial for cancers that originated from the tissues with high expression of its transporters and in tissues with high accumulation capacity, e.g., the liver and small intestine. Conversely, BTC and CCA are desmoplastic by nature, thus this factor could be a barrier to metformin’s activity. Third, all patients in the included studies were at the late stage of BTC and metformin dosages in clinical practice also vary across the patients depending on their glycemic status, renal function, and other indications or contraindications. These could be confounding factors that make metformin less beneficial in a setting of observational clinical studies. Last, observational studies are limited by the nature of the study designs. To affirm whether metformin is beneficial for BTC treatment, randomized controlled trials need to be carried out. At the time of our search, one phase Ib clinical trial of metformin and chloroquine in 12 patients with isocitrate dehydrogenase-1 mutated iCCA was located.39 It reported a poor clinical response of the tumor to both drugs, though they were well tolerated by the patients.39 Due to great heterogeneity among BTC subtypes, more experimental research is needed for more definitive conclusions.
Even though this meta-analysis did not support a beneficial effect of metformin on CCA prevention, the estimated RR of 0.82 favored an 18% lower risk of patients with DM who used this drug over the other medications. Moreover, metformin also shows a high potential for improving survival outcomes in BTC patients by favoring a reduced risk of mortality with an RR of 0.83 and with a marginal statistical significance (p = 0.05) in a fixed model meta-analysis. Metformin has been consistently reported in a series of population-based cohort studies to be associated with a significantly lower risk of CCA and cancers that are highly malignant and located in organs associated with the biliary tract system, including hepatocellular carcinoma, pancreatic cancer, and gastric cancer.13,40–42 Due to lowering the risk for such cancers, unless contraindicated metformin should remain in a major position in the treatment of type 2 DM in clinical practice both in the general population and patients who have BTC.
This systematic review and meta-analysis are the updated report on the effects of metformin on BTC, both on a carcinogenesis risk and a benefit on survival outcome. However, several limitations need consideration. First, the number of studies was limited at the time this meta-analysis was done. Second, sensitivity analysis for subtypes of BTC with heterogenous biological backgrounds could not be done due to the limitation in the data. Finally, CCA and BTC are associated with different risk factors in different regions. Therefore, when the numbers of primary studies are sufficient, further meta-analyses with sensitivity analysis of BTC subgroups classified by various factors are needed.
Conclusion
This meta-analysis of observational studies suggests that metformin does not provide any chemopreventive effects against BTC development in patients who have DM as an underlying disease. Neither does metformin for DM treatment appear to confer any benefit for the survival of BTC patients with DM. To affirm the results of the present study, a meta-analysis of a greater number of studies as well as randomized control trials are needed.
Abbreviations
- BTC:
biliary tract cancer
- CCA:
cholangiocarcinoma
- CI:
confidence interval
- DM:
diabetes mellitus
- eCCA:
extrahepatic cholangiocarcinoma
- GLP-1:
glucagon-like peptide 1
- HR:
hazard ratio
- iCCA:
intrahepatic cholangiocarcinoma
- IDH:
isocitrate dehydrogenase
- OR:
odds ratio
- RR:
relative risk
Declarations
Acknowledgement
We would like to thank Professor John F Smith for editing the English presentation of this manuscript via KKU publication Clinic, Khon Kaen University, Thailand.
Data sharing statement
The data that support the findings of this study are available from the corresponding author upon reasonable request.
Funding
This work was supported by the research grant for a young talented researcher, National Research Council of Thailand (Grant No. N41A640108) and the Fundamental Fund 65, Thailand Science Research and Innovation. SS and CS are awardees of the Prince Mahidol Award Youth Program of the Year 2018 from the Prince Mahidol Award Foundation under the royal patronage of His Majesty the King of Thailand (SS: PMAYP-10615, and CS: PMAYP-10612).
Conflict of interest
The authors have no conflicts of interest related to this publication.
Authors’ contributions
Conceptualization (SS, WS, SW, CS), methodology (SS, LW, SN, CS), formal analysis (LW, SN, CS), supervision (WS, SW), writing first draft (SS, CS), All authors have made a significant contribution to this study and have approved the final manuscript.