Proton Pump Inhibitors and Odds of Cholangiocarcinoma

A Retrospective Case-control Study

Jianping Xiong; Yaqin Wang; Weiyu Xu; Zhisong Liu; Haochen Wang; Zhiyuan Zhang; Yanjing Han; Chao Yin; Shasha Cao; Zheran Yang; Tianhao Su; Jian Wei; Guang Chen; Long Jin


Liver International. 2020;40(11):2848-2857. 

In This Article

Patients and Methods

Study Population and Design

We recruited 4404 subjects (including 1468 pathologically diagnosed CCA cases and 236 controls) from the Affiliated Beijing Friendship Hospital of Capital Medical University (Beijing, China) into the present hospital-based case-control study. Beijing Friendship Hospital is among the top hospitals for treating CCA cases in Beijing.

The information of patients diagnosed with CCA from February 2002 to October 2018 was summarized by using the patient informatics database of Beijing Friendship Hospital. Patients diagnosed by histopathology were enrolled in this study. CCA was categorized into ICC and ECC. Radiologic images (CT, MRI, and/or endoscopic retrograde cholangiopancreatography) were reviewed to delineate the anatomic tumour location.[22] Tumours originating from intrahepatic bile ducts were defined as ICC, while the remaining tumours were defined as ECC. Hilar cholangiocarcinoma (ie Klatskin tumours) is typically considered extrahepatic.[23]

The medical records of patients were manually and retrospectively analysed to extract demographic and clinical data and risk factors. Related information was acquired by interviewing the cases or their family members, and recorded by doctors using the structured data collection form. All information was collected routinely in our hospital and stored as part of the medical record. Moreover, data regarding age, gender, drinking, smoking, cirrhosis, choledocholithiasis or hepatolithiasis, hepatitis B virus (HBV) infection, hepatitis C virus (HCV) infection, parasitic infections (hepatobiliary flukes, opisthorchis viverrini and clonorchis sinensis), H pylori infection, cholecystolithiasis, fatty liver disease, hypertension, dyslipidemia, diabetes mellitus, obesity, coronary artery disease, peptic ulcer disease, GERD, metformin use, statin use, aspirin use and PPI use, were statistically analysed.

The control group was randomly recruited from the Health Screening Center of Beijing Friendship Hospital, which might not fully reflect the general population. The case group was recruited from patients in Beijing Friendship Hospital, rather than from the general population, so selection bias was minimized by paired control selection of the Health Screening Center of Beijing Friendship Hospital. Cases were propensity score-matched one to two with controls at the Health Screening Center of Beijing Friendship Hospital from May 2012 through January 2019. The propensity score was calculated using logistic regression to estimate the probability of the CCA assignment based on baseline variables including age, gender, drinking, smoking, cirrhosis, choledocholithiasis or hepatolithiasis, HBV, HCV infection, parasitic infections (hepatobiliary flukes, opisthorchis viverrini and clonorchis sinensis), H pylori infection, cholecystolithiasis, fatty liver disease, hypertension, dyslipidemia, diabetes mellitus, obesity, coronary artery disease, peptic ulcer disease, GERD, metformin use, statin use and aspirin use. Altogether, 2936 participants were enrolled in the control group. The chosen age, gender, medication and comorbidities can help to clarify the significant influence of PPI on CCA odds.

Clinical Information

Cirrhosis was evidenced by imaging examinations (nodular liver, portal hypertension and caudate hypertrophy) or confirmed by pathology. HBV and HCV infections were confirmed by the positive hepatitis B surface antigen (HBsAg) and HCV RNA respectively. In addition, in the doctor's records, the diagnosis of HBV and HCV confirmed virus infections. People who had ever smoked were identified with a history of smoking. Hepatolithiasis was diagnosed in the case of stones in the intrahepatic duct. Cholecystolithiasis and choledocholithiasis were diagnosed in the presence of stones in the gallbladder and common bile duct respectively. Cystic duct stones were assigned to cholecystolithiasis. A diagnosis of fatty liver was made in the presence of related radiological or histopathological findings showing liver fat infiltration as well as increased liver enzyme levels. Hypertension was diagnosed in accordance with the definition and diagnosis standards in the 2004 "Chinese hypertension prevention and treatment guide" commentary.[24] According to the Asia-Pacific region obesity standards, obesity was diagnosed when the body mass index (BMI) was ≥ 25.0 kg/m2.[25] Diabetes mellitus (DM) was diagnosed following the World Health Organization (WHO).[26] The diagnosis of dyslipidemia was made in the case of HDL-C < 40 mg/dL, whereas TC ≥ 200 mg/dL, LDL-C ≥ 130 mg/dL, and TG ≥ 130 mg/dL were assessed separately, in line with the National Cholesterol Education Program.[27] Peptic ulcer referred to an open ulcer formed on the stomach wall or the upper small intestine, which was diagnosed according to the upper digestive tract series (barium swallowing) or endoscopy.[28] GERD, a digestive system disorder, has a certain influence on the lower oesophageal sphincter as well as the oesophagus-stomach muscle ring, and it is diagnosed based on symptoms and whether the symptoms are improved with medical therapy.[29]

Exposure to PPIS

PPI applications referred to the application of the Anatomical Therapeutic Chemical classification system (ATC) codes A02BC, such as omeprazole, pantoprazole, lansoprazole, rabeprazole or esomeprazole (with the ATC codes of A02BC01 to A02BC05, separately). The defined daily dose (DDD) according to the ATC/DDD system of the WHO collaborating centre for drug statistics and methodology, was used as a unit for measuring a prescribed amount of a given drug; it was the assumed average maintenance dose per day of a drug consumed for its main indication in adults (omeprazole: 20 mg; pantoprazole: 40 mg; lansoprazole: 30 mg; rabeprazole 20 mg; esomeprazole 30 mg) (WHO International Working Group for Drug Statistics Methodology, 2003). As reported in most previous works,[30] patients who took PPIs < 28 cumulative limited daily dose (cDDD) were classified as the reference group (nonuser group). Subjects who received PPIs ≥ 28 cDDD were defined as PPI users. This definition was not changed if those medical record data indicated that the PPI was retained during the diagnosis of CCA, or the healthcare provider advised patients to avoid taking a PPI because of CCA-related symptoms. This definition prevented us from exaggerating the effect of PPI application by underestimating the frequency of PPI application in case groups.

For determining how the PPI dose affected CCA risk, the cumulative DDD (cDDD) was estimated as the sum of the dispensed DDDs of any PPI, which can be divided into 3 levels, including 28–90 cDDD, 91–180 cDDD or > 180 cDDD.

Statistical Analysis

Categorical variables were analysed by the chi-square test, while continuous variables were analysed by t-tests. Variables showing P < .05 upon univariate analysis were subsequently adjusted to potential confounders and incorporated into the multivariate logistic regression model. Factors associated with CCA development were determined by logistic regression, while odds ratios (ORs) and adjusted odds ratios (AORs) together with the corresponding 95% confidence intervals (CIs) were calculated. Additionally, for the control group, AORs for one factor along with the related occurrence were used to calculate the attributable risk ratio of the population with statistically significant risk factors. Bilateral P < .05 indicated statistical significance. Since a considerable number of variables are taken as covariates, multicollinearity test were conducted. SPSS20.0 (SPSS Inc) and SAS (SAS Institute) statistical software were employed for all statistical analyses. The present work gained approval from the Ethics Review Committee of Beijing Friendship Hospital.

To achieve robust analytic results, a sensitivity analysis was carried out to strengthen the validity of the results. Firstly, the 2-year washout period was adopted to exclude cases with PPI exposure prior to participation in this study, which avoided possible drug carryover effects at the beginning of the observation period. Secondly, we used lung cancer as a negative cancer case to ensure that any association observed in the initial analysis was not random. This is also a tool to detect confounding variables in observational studies.[31] We further added gastric cancer and hepatocellular carcinoma as an outcome control to supporting the validity of our findings. Furthermore, special attention and a subgroup analysis focused on H pylori infection. In addition, we assessed how different categorizations of PPI effects (pantoprazole, omeprazole, esomeprazole lansoprazole and rabeprazole) has an impact on CCA odds.