This was a single-center, retrospective cohort study conducted at Ramathibodi Hospital, a tertiary care university hospital that performs the highest volume of kidney transplants in the country, accounting for one-quarter of the cumulative kidney transplant surgeries in Thailand. This study included consecutive adult patients aged ≥18 years who received a first kidney transplant between 1 January 1986 and 30 July 2018 and had ≥1 month of follow-up. All recipients were followed up from the date of kidney transplantation to the date of a diagnosis of de novo incident cancer, all-cause death, graft failure, loss to follow-up, or administrative censoring on 31 December 2019, whichever occurred first. The exclusion criteria were cancer occurring after graft failure and cancer occurring before or ≤1 month following transplantation. The study was approved by the Institutional Review Board, Faculty of Medicine, Ramathibodi Hospital (MURA2022/503). Informed consent was not required due to the de-identification of patient data. This study was conducted in accordance with the tenets of the Declaration of Helsinki.

Data were collected from three sources. First, incidence data for the Thai general population were obtained from the Thai National Cancer Institute’s (NCI) nationwide cancer data registry. We chose to compare incident malignancy rates between our KTRs and the Thai general population between 2013 and 2015. We only selected the 2013–2015 period because the place of residence to be matched with that of our KTRs in the registry had been validated by cross-checking with the Ministry of Interior’s National Civil Registration records [9]. Additionally, data after 2016 were not available when we conducted the analysis. Moreover, the national coverage of the Thai NCI has gradually grown since the establishment of the first population-based only on data from Chiangm Mai province in 1986 [10]. Thus, adjusting for calendar year using historical subnational coverage would have reduced our sample size and power to detect any significant results. Diagnoses were initially coded by registry personnel using the International Classification of Diseases for Oncology (ICD-O, third edition), and the coding demonstrated good reliability [9]. These codes were then transformed to the International Classification of Diseases, Tenth Revision (ICD 10) for analysis. Second, we collected baseline patient data from the Ramathibodi KTR database, including age, sex, region of residence, and type and date of transplantation. Third, we manually reviewed hospital electronic medical records to collect outcomes and the date of outcomes. Diagnoses of de novo PTMs were ascertained by histopathologic confirmation or compatible radiographic studies with serologic testing. These were then categorized according to the ICD-10.

Immunosuppressive regimens for induction and maintenance therapy in KTRs were carefully selected based on the immunologic risks of the patients. Induction regimens included methylprednisolone with or without an interleukin-2 receptor antagonist or anti-thyroglobulin. The maintenance medications were mainly calcineurin inhibitor (CNI)-based regimens. A typical combination of these regimens before 2000 was cyclosporine-azathioprine-prednisolone, which gradually changed to tacrolimus-mycophenolate mofetil-prednisolone. Other regimens were used much less frequently, and they included CNI-mTOR inhibitor-prednisolone and CNI-prednisolone.

At our center, age-appropriate pre-transplant cancer screening and abdominal ultrasound have been performed since January 2013. Post-transplant cancer surveillance was opportunistic and consistent with standard cancer screening recommended for the general population [11].

The outcome was the standardized incidence ratio (SIR), which represents the excess risk of malignancy compared to the Thai general population.

Patient characteristics were analyzed descriptively. Continuous data are presented as mean (SD) or median (interquartile range), as appropriate. Categorical data are presented as frequencies (%).

The excess risk of de novo PTM in KTRs compared to the general population was estimated by SIR by indirect standardization of the Thai general population. Confidence intervals were estimated using the Poisson distribution. We selected the 2013–2015 data from the Thai NCI, which used the projected Thai population in mid-2014 based on the 2010 national population census. We defined the start of the time at risk as 1 month after the date of the kidney transplant to exclude non-de novo PTMs. We defined the end of the time at risk as incident de novo PTM, loss to follow-up, all-cause mortality, or administrative censoring at the end of the study period, whichever happened first. We also reported the incidence and SIRs stratified by sex, age at kidney transplant, duration of follow-up after kidney transplant, and transplant era of the top three cancers with the most excess risk. To avoid underestimating the incidence due to multiple de novo PTMs, if a second or third de novo PTM occurred, we generated time at risk from 1-month post-transplant to the date of the second or third incident PTM. Statistical analyses were performed using Stata 15.1 (StataCorp. 2017. College Station, TX). A non-null 95% confidence interval (CI) was considered significant.

During the 32-year period from January 1986 to July 2018, there were 2,448 kidney transplantations. After exclusions, a total of 2,024 cases were included in the final analysis, contributing to 16,495 patient-years of risk time in the follow-up (Figure 1; Table 1). The mean age at transplantation was 42.4 years (SD 11.4), and 38.6% of the subjects were women (Table 1). The median follow-up time was 6.4 years (interquartile range 3.21, 11.5) (Table 1).

A total of 133 de novo PTMs were identified in 125 (6.2%) patients. In total, eight (6%) patients developed secondary cancer, including 2 non-melanoma skin cancers, 2 liver and biliary cancers, 1 prostate cancer, 1 kidney cancer, 1 urothelial cancer, and 1 colorectal cancer. The incidence by type of PTMS, age at diagnosis, and time from transplantation to diagnosis are shown in Table 2. The most common PTM was urothelial cancer (n = 28), accounting for one-fifth of all PTMs in the cohort, followed by non-Hodgkin’s lymphoma (NHL) at 22% and non-melanoma skin cancer at 16%.

Table 3 shows the SIRs adjusted for age, sex, and place of residence, comparing the KTRs to the Thai general population. The excess risk of de novo PTMs in KTRs was nearly fourfold compared with the general population. The top three cancers with significantly increased risk were consistent across genders. This risk was more pronounced for urothelial cancer, NHL, and non-melanoma skin cancer, all of which had more than a 20-fold excess risk. In particular, urothelial cancer showed the highest significant excess risk in women with a SIR of 114.7 (95% CI 66.8, 183.6). The risk of the most common cancers in the Thai general population, including lung, colorectal, breast, ovarian, and cervical cancers was comparable. The SIR for prostate cancer was significantly increased [SIR 8.11 (95% CI 3.71, 15.4)]. The SIR calculations considering only the first de novo PTM are shown in Supplementary Table S1.

Table 4 shows the SIRs for all PTMs stratified by age at transplant, duration of follow-up, and era of transplantation. The excess risk in KTRs compared to the general population by age at transplant was greatest in KTRs aged 45–64 years, and there was a comparable risk in KTRs transplanted at <20 years of age. By duration of follow-up, there was a decreasing trend over the duration of follow-up until the risk was comparable to the general population at 15 years or more of follow-up.

Table 5 shows the incidence rates and SIRs for the top three most common PTMs in Thai KTRs, which were urothelial cancer, NHL, and non-melanoma skin cancer, stratified by age at transplant, duration of follow-up, and era of transplant. All three cancers showed no excess risk in KTRs aged ≥65 years at transplantation. For urothelial cancer, age at transplant between 20 and 64 years old showed significant excess risk with the greatest excess between 45–64 years old [SIR 28.4 (95% CI 17.1, 44.4)], and duration of follow-up showed the greatest significant excess risk between 1 and 5 years of follow-up. There was still a significant excess risk at ≥20 years of follow-up. NHL had the greatest excess risk by age at transplant in KTRs aged less than 20 years at transplant [SIR 362 (95% 9.15, 2,015)], and the excess risk decreased with increasing age at transplant. There was a significant excess risk of NHL at all follow-up periods, with the risk decreasing over time. For non-melanoma skin cancer, the greatest excess risk according to age at transplantation was between 20 and 44 years old [SIR 1,820 (95% CI 46.1, 10,139)], and there was excess risk up to >20 years of follow-up.

Supplementary Figures S1–S3 show the cumulative incidence functions (CIF) of the three most common cancers adjusted for competing risks. After adjusting for all-cause death, graft failure, loss to follow-up, or administrative censoring, the odds of developing any of the three cancers were mostly low, at less than 5% over a period of nearly 20 years post-kidney transplantation.

There are strengths and limitations to our study. One strength is that this is the first study to demonstrate the excess de novo cancer risk among the Thai KTRs in a large sample size with a long period of follow-up for most of the KTRs. In addition, all pathologic diagnoses were confirmed by retrospective chart review, and the stratification of the excess incidence based on subnational regional risk in the Thai general population adjusted for different underlying risk factors in each subnational region. However, there are also limitations. One limitation is the use of the 2013–2015 Thai general population. We would have preferred to adjust our statistics for excess risk in KTRs compared to the Thai general population incidence by calendar year. However, there is a limitation as we have mentioned in the methodology section. The reasons for potential bias in our estimated standardized incidence ratio due to our choice of Thai general population comparison group can be divided into differences in age and sex distribution in the Thai population in mid-2014 and differences in cancer incidence between 2013 and 2015 independent of age and sex (e.g., changes in the prevalence of other cancer risk factors or national/subnational systematic cancer screening) compared with an ideal comparison group adjusting for age, sex, and cancer incidence by calendar year. The age structure of the Thai population has changed in recent years toward an aging society [46]. Due to the association of older age with higher cancer incidence in most PTMs, recent calendar year data with an older population may have higher cancer incidence, creating a bias in our estimates toward lower SIRs. The incidence of different types of cancer has changed at different rates. For example, cancers of the liver, lung, stomach, and Hodgkin’s lymphoma have had declining incidence trends, while cancers of the colon, breast, ovary, prostate, kidney, and non-Hodgkin’s lymphoma have had increasing trends. Hence, the expected directions of bias are upward bias in SIRs for the former group of cancers with decreasing trends and downward bias for those with increasing trends. Given the strong magnitudes of the strength of association at the lower bounds of the 95% CIs for cancers with known viral etiologies, which are known to be more likely in KTRs due to long-term maintenance immunosuppression, we suggest that there are truly significantly greater risks of these types of PTMs, regardless of bias due to choice of population for indirect standardization and other methods of adjusting for confounders [47]. We also acknowledge that the lack of standardized mortality ratios in our study may not allow for the interpretation of the data as to whether Thai KTRs are genuinely an appropriate, special, and very high-risk population that requires service provision for systematic screening for some cancers. Only cancers with significant excess risk, adjusted for confounders, compared to the Thai general population are attractive targets for systematic screening service provision. This issue should be investigated in future studies, but such studies may be challenging due to historical vital statistics data quality issues [9].