Association between clinical use of lansoprazole and the risk of type 2 diabetes mellitus: a pharmacoepidemiological cohort study

The current study was a population-based retrospective cohort study using medical 85 claims dataset from the National Health Insurance (NHI) Program in Taiwan, the NHIRD. The NHI is a publicly funded single-payer health insurance program for all Taiwanese residents. The NHIRD contains comprehensive medical care information, including demographic data of insured individuals, data of clinical visits, diagnostic codes in the format of International Classification of Diseases, Ninth revision, Clinical Modification (ICD-9- CM), and prescription details, as described previously [19]. The NHIRD can deliver as a basis for the procurement of high-quality epidemiological studies [20] with a good validity on information regarding diagnoses and drug prescription [21, 22]. The data of this study was obtained from the Longitudinal Health Insurance Database 2000 (LHID 2000), a subset of NHIRD. The LHID 2000 data set contains historical ambulatory and inpatient care data for 1 million randomly sampled beneficiaries enrolled in the NHI system in 2000. There were no significant differences in the distributions of age, sex, and healthcare costs between the individuals in LHID and NHIRD [19]. Since the data set was released for research purposes and included only scrambled information on patient identification, the study was exempt from informed consent from the subjects. Meanwhile, the present study has been approved by the Institutional Review Board of Fu-Jen Catholic University (FJU-IRB No:C104014).

The source population was patients who were randomly selected into LHID 2000 and were aged between 20 and 80 years from January 1, 2000 to December 31, 2005. To be eligible, patients needed to be continuously covered by NHI between 2000 and 2005. In the present study, we used incident user design to define lansoprazole and other PPIs (including omeprazole, esomeprazole, pantoprazole, and rabeprazole) exposures. That is, patients who did undergo lansoprazole or other PPIs treatments prior to 2000 were excluded. Patients who received lansoprazole≧90 days and without any co-prescription of other PPIs between January 1, 2000 and December 31, 2005 (the exposure period) were considered as the exposed cohort. In comparison, patients who received other PPIs prescriptions≧90 days and without use of lansoprazole in the study period were considered as the comparison cohort. In this study, we used patients who received lansoprazole or other PPIs≧90 days (equal to three times of prescriptions in outpatient visits) in the study period as the stable lansoprazole or PPIs users. Patients who received lansoprazole or other PPIs with less than 90 days in the study period were excluded. In the current study, the temporal period associated with drug exposures was referred to the exposure period, which was equal for each patient. The date of initial prescriptions of lansoprazole for each patient was assigned as their index date. Initiation was defined as being free from any lansoprazole therapy for 12 months prior to the first prescription (index date). Patients in both exposed and comparison cohorts had no diagnosis of T2DM or prescriptions of anti-diabetic agents prior to the index date. To control for potential confounders between the two cohorts, we applied propensity score matching at a ratio of 1:2 for exposed and comparison cohorts. The propensity score was calculated for each patient by using a logistic regression model with covariates of age, sex, index date, baseline comorbidities, including heart failure (ICD-9-CM code 428.0), malignant neoplasms (ICD-9-CM codes:140-208), cardiovascular disease (ICD- 9-CM codes: 410-414, 425, 428, 674, and 678), hypertension (ICD-9-CM codes:401-405), hyperlipidemia (ICD-9-CM code:272.4), chronic liver disease (ICD-9-CM codes 570-572), and chronic kidney disease (ICD-9-CM code:585) and use of co-medications, including beta blocking agents (Anatomic Therapeutic Chemical (ATC) code: C07), statins (ATC codes:C10AA01, C10AA02, C10AA03, C10AA04, C10AA05, and C10AA07), corticosteroids (ATC code: R01AD), and thiazide (ATC codes:C03AA03). Cohort members were excluded from the study if they were aged < 20 years or > 80 years (n = 1819), had been diagnosed with T2DM and/or use of anti-diabetic agents prior to the index date (n = 7648). We finally included 1,668 patients as the exposed cohort and 3,336 patients as the comparison cohort (Fig. 1). All of the study participants were followed from the index date to the onset of T2DM, death (as indicated by disenrollment from the NHI) or the end of the study date (December 31, 2013), whichever occurred first.

The primary outcome was the occurrence of new-onset T2DM. To ensure the diagnostic validity of T2DM, we determined patients having one hospital admission or at least three outpatient diagnoses of T2DM based on the ICD-9-CM codes: 250.0, 250.1, 250.2, 250.3, 250.4, 250.5, 250.6, 250.7, 250.8, and 250.9 combined with treated with insulin or diabetes- specific hypoglycemic agents for > 90 cumulative defined daily doses (DDDs) within 365 days.

Patients’ demographics, baseline comorbidities, and use of co-medications were identified as covariates. We used outpatient files to ascertain whether study subjects had comorbidities. Comorbidities were determined in a patient if he or she was diagnosed for any of the aforementioned diseases on at least two outpatient claims during the study period. In addition, data on the use of concomitant medications were extracted from The NHIRD prescription database by using the code of the ATC classification system.

Chi-square and t-tests were used to evaluate the differences in distributions of categorical and continuous variables between the study cohorts. In addition, we used the Kaplan–Meier method to estimate the cumulative incidence of T2DM for study cohorts. The log-rank test was used to evaluate differences in cumulative incidence of T2DM between the cohorts. Furthermore, the multivariable Cox proportional hazards regression models were performed to compute hazard ratios (HRs) with 95% confidence intervals (CIs) to assess the association of use of lansoprazole and the risk of incident T2DM after adjusting for potential confounders.The log minus log plot of survival was used to verify that explanatory variables analyzed satisfy the proportionality assumption of the Cox regression model [23]. All statistical tests were two-sided, and a level of 0.05 was considered statistically significant. All data analyses were performed using SAS software, version 9.1 (SAS Institute, Cary, NC).