Data source
In this study, we retrieved data from the medical claim database of Taiwan’s NHIRD—an extensive database provided by a single-payer, universal, compulsory health care system, National Health Insurance (NHI), for nearly all 23.7 million residents of Taiwan. To date, the NHIRD has been comprehensively used for high-quality epidemiological studies [13, 14] and information on diagnoses, prescriptions, and hospitalizations have been shown to be of good validity [15, 16]. In the present study, we obtained data from the Longitudinal Health Insurance Database (LHID 2000), a subset of NHIRD that comprises historical ambulatory and inpatient care data of 1 million randomly sampled beneficiaries enrolled in the NHI system in 2000. Moreover, the LHID 2000 database has facilitated access to the medical service use history of patients under investigation. Notably, no marked differences were observed in terms of the distributions of age, sex, and health care costs between individuals in the LHID and NHIRD [13, 14].
Participants
In this nationwide, retrospective, population-based cohort study, we enrolled patients with a primary diagnosis of ESRD undergoing hemodialysis between January 1, 2000, and December 31, 2008, as per the LHID. This duration was defined as the exposure period to determine hemodialysis patients who received KT. Figure 1 summarizes the subject selection process. First, we identified patients diagnosed with ESRD [International Classification of Disease, Revision 9, Clinical Modification (ICD-9-CM) code: 585] undergoing hemodialysis (ICD-9-CM code: 39.95) and received KT (ICD-9-CM codes: V42.0, 996.81, and V58.69) as the exposed cohort (n = 1375). Comparatively, patients with ESRD who were undergoing hemodialysis but not receiving KT were identified as the non-exposed cohort (n = 24,703). Patients were excluded if they were younger than 30 years of age, had incomplete demographic data, or were ever diagnosed with CRC prior to the beginning of follow-up (January 1, 2000). Overall, a total of 3739 hemodialytic patients receiving KT (exposed group) and 42,324 not receiving KT (non-exposed group) were included in data analyses.
Ascertainment of CRC
The primary outcome was the primary diagnosis of CRC (ICD-9-CM codes: 153, 153.0, 153.1, 153.2, 153.3, 153.6, 153.7, 153.8, 153.9154, 154.0, 154.1, 154.2, 154.3, 154.8, and 159.0). In this study, CRC diagnosis was defined according to the Registry for Catastrophic Illness Patient Database (RCIPD), a subpart of the NHIRD. The diagnosis of CRC needs histologic confirmation to be reported in the RCIPD. We followed both cohorts from January 1, 2000, to the date of CRC diagnosis, death (indicated by withdrawal from the NHI), or the end of 2013, whichever occurred first.
Potential confounders and comorbidities
We noted the covariates that were the potential confounders in the association between KT and CRC. The potential confounders considered in this study included age, sex, and comorbidities, [17] including chronic obstructive pulmonary disease (ICD-9-CM codes: 490, 491, 492, 493, 494, 495, and 496), diabetes mellitus (ICD-9-CM code: 250), coronary artery disease (ICD-9-CM codes: 410, 411, 412, 413, and 414), hypertension (ICD-9-CM codes: 401, 402, 403, 404, and 405), alcohol-related conditions (ICD-9-CM codes: alcoholic liver disease, 571.0, 571.1, 571.2, and 571.3 and alcohol dependence, 303), hypercholesterolemia (ICD-9-CM codes: 272.0, 272.1, 272.2, and 272.4), peptic ulcer (ICD-9-CM codes: 531, 532, and 533), liver cirrhosis and chronic hepatitis (ICD-9-CM code: 571), and inflammatory bowel disease (ICD-9-CM codes: 555 and 556).
Statistical analysis
We tested differences in descriptive statistics on demographic characteristics and baseline comorbidities between the exposed and non-exposed cohorts using chi-square tests or Student’s t-test when appropriate. In addition, we used Kaplan–Meier method to estimate the cumulative incidence of CRC. The log-rank test was performed to assess the difference in the cumulative incidence of CRC between the curves of the cohorts. In addition, we used Cox proportional hazards models to compute hazards ratios (HRs) with 95% confidence intervals (CIs) to determine the association between KT and CRC risk after adjusting for the potential confounders. All statistical analyses were performed using SAS version 9.4 (SAS Institute Inc., Cary, NC), and we set the statistical significance at 0.05 for two-tailed tests.
Ethics, consent and permission
Since the dataset was released for research purposes and included only scrambled and anonymous information on patient identification, the study was exempt from informed consent from the subjects. Further, the execution of the present study and a waiver of obtaining informed consent from subjects have been approved by the Institutional Review Board of Fu-Jen Catholic University (FJU-IRB NO:C104014).