Using data from two large population-based studies which differed in design (case–control vs prospective cohort) and method of exposure ascertainment (self-report vs prescription records), we found a consistent association between PPI use and liver cancer risk. Conversely, there was little evidence of association with use of H2RAs. PPI use was more strongly associated with the risk of IBDC than for HCC, and the association was not attenuated after adjustments for available confounders. However, the associations were slightly attenuated when controlling for potential reverse causation (using lags) and, using data from PCCIU, we found no evidence of a dose response based upon duration of use.
A previous case–control study in Taiwan found a statistically significant association between use of H2RAs and the risk of HCC (adjusted OR 1.46 95% CI 1.30, 1.64) but not PPIs (adjusted OR 0.94, 95% CI 0.78, 1.13), but did not investigate IBDC. Although we observed stronger associations between PPIs and liver cancer risk, in analyses by subtype these associations were only apparent for IBDC. The difference in the findings of our study and the Taiwan study are unclear but could reflect differences in the underlying populations with respect to other liver cancer risk factors or genetic factors, differences in PPI prescribing patterns, or study specific differences (for instance they frequency matched controls to cases whereas we individually matched).
Our study has a number of strengths. We observed consistent findings for PPIs and liver cancer risk across two independent datasets. There was minimal risk of recall bias as PCCIU analyses were based upon GP prescription records whilst UK Biobank was a prospective cohort study in which medications were recorded at least 1 year prior to liver cancer onset. In both datasets, we adjusted for a wide range of confounders and, particularly, in UK Biobank, we had detailed information on lifestyle risk factors including smoking and alcohol. Also, data on number of prescriptions were available in PCCIU.
The main limitation is that we cannot rule out confounding by incomplete or unknown exposures. Although we adjusted for cirrhosis and liver disease, cirrhosis patients are commonly prescribed PPIs and therefore any misclassification within PCCIU or UK Biobank could lead to residual confounding. Furthermore we cannot rule out confounding by indication, for example, individuals with GERD, for which they receive PPIs, have been shown to have increased risk of nonalcoholic fatty liver disease which is a risk factor for primary liver cancer. There was an indication of reverse causation as the associations were slightly attenuated when medication use in the period prior to onset was removed, which could be influenced by liver cancer symptoms. A further weakness was that histological subtype was not available in PCCIU but we did have these data in UK Biobank. Finally, adherence to medications was unknown in either dataset, but this seems more likely to dilute associations.
The cause of the observed increased risk of liver cancer with PPIs use is unknown. If real, our findings are consistent with an experimental study which showed that PPIs can promote liver tumors in rats. Various potentially harmful mechanisms of PPIs have been proposed. In particular, long-term PPIs use can lead to hypergastrinemia which has been shown to have a carcinogenic effect, particularly on liver cells. Also PPIs reduce gastric acid secretion increasing the survival of various microbes in the stomach.[7,32] The resulting bacterial overgrowth could greatly contribute to the transformation of primary bile acid in the intestine to secondary bile acid and has been shown to impact upon the liver exacerbating various liver diseases in mice. A high level of secondary bile acid has been shown to cause toxic, inflammatory, and DNA damaging effects on liver cells and bile duct cells, leading to HCC and cholangiocarcinoma. The generally weaker associations observed for H2RAs could reflect the weaker acid suppression and the less marked effect on gastrin associated with these medications. Alternatively, various features of the observed association do not support a causal interpretation including the lack of dose–response (the most marked association was seen for less than 6 prescriptions), the possibility of confounding by indication, and the possibility of reverse causation (suggested by the attenuation of associations when prescriptions in the period prior to diagnosis where removed). However, the widespread use of PPIs, and particularly their use without clear indication, and the high mortality of liver cancer highlight the need for further research. Specifically, further studies should contain sufficiently long follow-up to investigate reverse causation, high quality liver cancer outcome data and detailed and complete information on liver cancer risk factors.
In conclusion, our study provides some evidence of an association between PPI use and the risk of liver cancer; however, this association requires confirmation in other studies due to the possibility of residual confounding and/or reverse causation.
The analysis of UK Biobank has been conducted using the UK Biobank Resource under Application Number 34374. We acknowledge collaboration with the Research Applications and Data Management Team lead by Ms Katie Wilde, University of Aberdeen in conducting this study. KTT is supported by the Vietnam International Education Cooperation Department. Access to PCCIU data was provided by Queen's University Belfast and the Centre for Academic Primary Care, University of Aberdeen. Access to the UK Biobank was funded by a Cancer Research UK Population Research Postdoctoral Fellowship awarded to ÚCMcM. HGC is a co–investigator of the UKCRC Centre of Excellence for Public Health Northern Ireland.
Declaration of personal interests
Declaration of funding interests
This study was funded by Queen's University Belfast and the University of Aberdeen (providing funding to access PCCIU data), a CRUK Population Research Postdoctoral Fellowship for ÚCMcM (providing funding to access UK Biobank) and the Vietnam International Education Cooperation Department (providing funding for a PhD for KTT).
Guarantor of the article
Chris R Cardwell.
Queen's University Belfast; University of Aberdeen; Vietnam International Education Cooperation Department
Aliment Pharmacol Ther. 2018;48(1):55-64. © 2018 Blackwell Publishing