All data were collected from the National Health Insurance Service (NHIS) using their customized database service. In Korea, nearly all citizens have joined the National Health Insurance program as a mandatory social insurance since 1989. The NHID is a public database formed by the NHIS and contains sociodemographic characteristics, dates of birth and death, health care utilization, and health screening for the whole population of Korea [5]. Therefore, we used the NHID to obtain the most accurate epidemiological data of the entire Korean population.

This study was approved by the Institutional Review Board of Hanyang University Guri Hospital (2017-10- 003). Because personal information in this national dataset was de-identified for research use, the committee waived the requirement of written informed consent.

We initially selected end stage renal disease patients who had started maintenance hemodialysis between January 1, 2003 and December 31, 2005. Patients who started hemodialysis were identified as those having a new claim of hemodialysis-related code, including V001, O7020, O7021, or O9991. Among them, we defined ESRD patients on maintenance hemodialysis as receiving hemodialysis for more than 90 days. The key exclusion criteria were a history of any cancer and/or transplantation before dialysis, age of 18 years or younger, claims of hemodialysis only as an inpatient condition, and patients who survived for less than 90 days from initiation of dialysis. Patients were followed from the time of initiation of hemodialysis to renal transplantation, death, or December 31, 2016, whichever came first.

In Korea, patients with newly diagnosed malignant disease were registered with a special certification for cancer, code V193, from September 2005 to reduce the cost. Cancer was defined as a new claim for major diagnosis of C code (ICD 10: C00-C96) with the inpatient condition until August 2005; thereafter (from September 2005), it was defined as a new claim for major diagnosis of C code and V193. The cancer occurrence date was defined as the date of a new claim. Patients who had two or more codes for cancer were considered diagnosed with the first claimed code. The definition of newly diagnosed cancer was a diagnosed malignant neoplasm at least 90 days after initiation of hemodialysis.

To analyze risk factors associated with cancer occurrence, we obtained data on age at initiation of hemodialysis, sex, insurance type, and comorbidities. Insurance type was used as an indirect indicator of income. The ICD-10 codes used to define potential confounding comorbidities were as follows: diabetes mellitus (E101, E102, E103, E104, E105, E109, E111, E112, E113, E114, E115, E119, E131, E132, E133, E134, E135, E139, E141, E142, E143, E144, E145, E149), hypertension (I10), liver disease (B18, K702, K703, K717, K73, K740, K742, K743, K744, K745, K746, K721, K729, K766, K767), pulmonary disease (J40–J47, J60–J67), and cardiovascular disease, including angina, acute myocardial infarction, congestive heart failure, peripheral vascular disease, or cerebral vascular accident (I20, I21, I22, I252, I50, I71, I790, I739, R02, Z958, Z959, I60, I61, I62, I63, I64, I65, I66, I670, I671, I672, I674, I675, I676, I677 I678, I679, I681, I682, I688, I69, G450, G451, G452, G454, G458, G459, G46). We referred to new ICD-10 version of the Charlson comorbidity index for defining these comorbidities [6]. We added chronic viral hepatitis (B18) to liver disease because Korea is endemic area of hepatitis B. These comorbidities were identified in more than two hospital outpatient visits or more than one hospital inpatient visit using diagnostic codes within 1 year before the start of dialysis.

Statistical analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC, USA). Data are presented as the mean ± standard deviation (SD). The Pearson chisquare test was performed to compare covariates between the groups with and without a new cancer diagnosis. Incidence rate was calculated. Hazard ratio (HR) and 95% confidence interval (CI) were calculated.

Overall survival was measured from the date of cancer diagnosis to the date of death from any cause or the date of last follow-up. Survival curves were calculated using the Kaplan-Meier method and compared with other prognostic variables using the log-rank test. The Cox proportional hazards model was performed to identify prognostic factors for survival. A p < 0.05 was considered statistically significant.

We initially selected a total of 49,770 ESRD patients with hemodialysis-related codes, including V001, O7020, O7021, or O9991, between January 1, 2003 and December 31, 2005. Among them, we excluded 35,388 patients: 24,333 with less than 90 days duration between the initial and last claims of hemodialysis, 10,341 who started hemodialysis before 2003, 346 who had a history of cancer, 193 who had undergone transplantation before dialysis, 94 with missing data on date of birth or sex, and 81 who were 18 years or younger. Finally, we identified 14,382 ESRD patients who started maintenance hemodialysis between 2003 and 2005. The mean follow-up duration was 77.51 ± 53.99 months.

Patient characteristics are summarized in Table 1. The mean age at the time of initial hemodialysis was 58.22 ± 12.15 years. Hypertension (70.0%) was the most common comorbidity, followed by diabetes mellitus (42.9%), cardiovascular disease (33.7%), pulmonary disease (15.4%), and liver disease (7.1%). A total of 2,186 patients (15.2%) underwent renal transplantation at a mean of 55.56 ± 43.13 months after initiation of hemodialysis. Among the 14,382 ESRD patients, 8,883 (61.76%) died during follow-up.

Of the total 14,382 ESRD patients who started maintenance hemodialysis between 2003 and 2005, 1,124 (7.82%) were diagnosed with cancer during the study period. Among them, 97 patients (8.63%) had two or more codes for cancer, and they were considered diagnosed with the first claimed code. Cancer patients included 728 men and 396 women, and the mean age at cancer diagnosis was 63.43 ± 11.56 years. The mean duration from the start of hemodialysis to new cancer identification was 64.40 ± 41.81 months, and the median was 56.92 months (range, 3.07 to 168.23). The significant risk factors for the development of new cancer were old age, male sex, and liver disease. On the contrary, patients with diabetes showed low risk for new cancer (Table 2).

Cancer incidence according to the duration of hemodialysis is shown in Fig. 1. Cumulative incidence of any cancer from the start of hemodialysis in ESRD patients, stratified by sex is described in Fig. 2.

The primary site of cancer classified by sex is reported in Table 3. In the 728 male patients, the colorectum (17.31%) was the most common primary site of cancer, followed by the liver (15.8%), stomach (14.29%), lung (13.6%), kidney (10.3%), bladder (5.35%), and prostate (4.26%). The mean time from start of hemodialysis to cancer development was 60 ± 41.01 months. Hematologic malignancy had the shortest mean duration of cancer development (45.29 ± 46.53 months). In the 396 female cancer patients, the colorectum (14.65%) was also the most common primary site of cancer, followed by the breast (12.88%), thyroid (12.63%), stomach (10.86%), lung (8.08%), uterus (7.58%), and kidney (6.57%). The mean duration from start of hemodialysis to cancer development was 70.83 ± 42.54 months in women.

The median survival duration of 14,382 ESRD patients from the initiation of dialysis was 86.37 months (range, 0.5 to 170.23). The median survival duration of ESRD patients with cancer was significantly shorter than that of ESRD patients without cancer (72.63 months vs. 87.5 months, p < 0.001) (Fig. 3).

Of the 1,124 ESRD patients with cancer, the median survival duration from date of cancer diagnosis was 26.63 months. According to the primary site of cancer, breast cancer showed the longest median survival duration (130.93 months), followed by thyroid (105.87), kidney (78.73), colorectum (35.33), bladder (30.9), stomach (23.33), liver (9.4), and lung (6.7).

On univariate analysis using the log-rank test, significant risk factors for overall survival in ESRD patients with cancer were age ≥ 50, male, diabetes mellitus, cardiovascular disease, and lung disease (Fig. 4). Although women had a longer survival duration than men, there was no significant survival difference between women and men for major cancers, including stomach, colon, liver, kidney, bladder, thyroid, and lung cancer. On multivariate analyses, overall survival was affected by age and diabetes (Table 4).