The Association Between Renal Function Decline and the Incidence of Urothelial Carcinoma: A 16-year Retrospective Cohort Study in Taiwan

Background The incidence of cancer is higher among patients with end-stage renal disease but it remains uncertain whether a mild decrease in renal function affects cancer. Objective To measure the effect of impaired renal function, represented by the estimated glomerular filtration rate (eGFR), personal health behaviors, and long-term exposure to fine particulate matter (PM2.5) on the risk of urothelial carcinoma (UC) incidence. Design, setting, and participants We performed a population-based cohort study of 372 008 participants aged ≥30 yr with no prior cancer history using the MJ health examination database (2000–2015) and UC diagnosis data from the Taiwan Cancer Registry database. Outcome measurements and statistical analysis Cox proportional hazards models were used to quantify the association between eGFR and UC incidence. Results and limitations We detected 383 UC cases during a median follow-up of 10.3 yr. Low eGFR was significantly associated with UC (p value for trend <0.01): compared to eGFR ≥90 ml/min/1.73 m2, the adjusted hazard ratio (HR) was 1.36 (95% confidence interval [CI] 0.98–1.88), 1.86 (95% CI 1.22–2.84), and 1.95 (95% CI 1.06–3.56) for eGFR strata of 60–89, 45–59, and <45 ml/min/1.73 m2, respectively. The risk remained elevated after stratifying the follow-up duration to check for reverse causality, and the dose-response relationship was stronger for women than for men. Current smoking (HR 1.34, 95% CI 1.02–1.77) and long-term exposure to PM2.5 concentrations ≥25.1 μg/m3 (HR 1.54, 95% CI 1.14–2.09) both significantly increased the risk of UC incidence. A significant dose-response relationship between PM2.5 and UC was also noted (ptrend < 0.01). Limitations include the retrospective design and limited information on medical history. Conclusions Lower renal function showed a dose-response relationship in elevating UC risk. Long-term exposure to PM2.5 is also a possible UC risk factor. Patient summary People with kidney function that is lower than normal should monitor the health of their kidneys and other organs in the urinary system. Our study confirmed that as well as smoking, exposure to fine particulate matter in the air may be a risk factor for cancers of the urinary system.


Introduction
The association of chronic kidney disease (CKD)-a wellknown public health burden with its high risk of progression to end-stage renal disease (ESRD), high medical expenses, high morbidity, and poor mortality prognosis [1,2]-with cancer has recently been discussed [3][4][5][6]. Deteriorating renal function leads to retention of metabolic waste products. These uremic toxins can increase DNA damage [7] and induce a state of oxidative stress. Chronic oxidative stress contributes to immune dysfunction and systemic inflammation, both of which play a role in cancer development [8,9]. It has been well established that cancer incidence increases after renal replacement therapy in patients with ESRD. However, only a few recent studies have investigated whether renal function is associated with cancer risk among patients with early-stage CKD and healthy individuals [4][5][6]. The results are also less robust for site-specific cancers except for urinary tract cancer. Therefore, the primary aim of this study was to elucidate the association between renal function and urothelial carcinoma (UC).
In order to measure the real size of an effect, it is important to control all possible confounding factors. Recent studies found that long-term exposure to fine particulate matter (diameter <2.5 mm, denoted PM 2.5 ) can increase the risk of CKD [10,11] and a plausible positive correlation between PM 2.5 and UC was identified in previous studies [12,13]. We therefore believe that PM 2.5 should be considered as a confounder when exploring the association between renal function and UC. We added PM 2.5 to the statistical model for adjustment to investigate the relationship between PM 2.5 and UC as a secondary study aim.
We hypothesized that lower renal function elevates UC risk, and that long-term exposure to PM 2.5 might also be associated with the development of UC.

2.
Patients and methods

Study population and design
From 2000 to 2015 (16-yr study period), 471 669 people participated in the MJ health examination program [14], yielding 1 093 479 visits in total.
At each visit, information on lifestyle behavior was gathered using a standard self-administered questionnaire, and anthropometric measurements and biological test data were collected during the health examination. Each participant provided written consent authorizing use of the data.  Participants with no proper ID number for which a link could be established (n = 5197), implying they gave the incorrect ID or were a foreigner, and those who illogically had a death date before their health examination date (n = 21) were excluded. Other exclusion criteria were any prior cancer history at the time of enrollment (n = 1071), age <30 yr, or missing or inadequate renal function information (estimated glomerular filtration rate [eGFR] !200 or <2 ml/min/1.73 m 2 , as these values suggested the measurements were probably incorrect because of technical errors) [15]. The remaining participants (n = 372 008) were included in the subsequent analyses. The study was approved by the Academia Sinica review board for biomedical science research (AS-IRB-BM-17044).

Assessment of variables
The primary predictor was renal function at baseline, represented by eGFR calculated using the Chronic Kidney Disease Epidemiology Collaboration equation [15]. We classified eGFR into four categories: !90, 60-89, 45-59, and <45 ml/min/1.73 m 2 . Normal eGFR for adults is !90 ml/min/1.73 m 2 , so this served as the reference.
The primary outcome of interest was the first incident UC after baseline.
According to location, UC in the renal pelvis or ureter is called upper tract UC (UTUC) and UC in the bladder or urethra is called lower tract UC (LTUC).
To adjust for potential confounders, the following covariates were included in the statistical models. PM 2.5 exposure at each participant's address reported in the questionnaire was estimated using a satellitebased spatiotemporal model [16] with high spatial resolution of 1 km Â 1 km on the basis of National Aeronautics and Space Administration aerosol optical thickness data [17]. We calculated a 2-yr average Conclusions: Lower renal function showed a dose-response relationship in elevating UC risk. Long-term exposure to PM 2.5 is also a possible UC risk factor. Patient summary: People with kidney function that is lower than normal should monitor the health of their kidneys and other organs in the urinary system. Our study confirmed that as well as smoking, exposure to fine particulate matter in the air may be a risk factor for cancers of the urinary system.

Sensitivity analysis
To explore the possibility of reverse causality, we split the follow-up duration into two periods ( 10 yr and >10 yr) and estimated separate HRs for each interval as a sensitivity analysis. We also constructed another competing-risk model considering death before UC incidence as a competing event using the Fine and Gray [19] approach. An additional sensitivity analysis was performed by including only participants who provided residential coordinates in the questionnaire.

Baseline characteristics
During a median follow-up of 10.3 yr, 383 cases of UC occurred ( Table 1). The median age at UC diagnosis was 70.0 yr. The age-standardized incidence rate for UC was 8.87 per 10 5 population overall, and 12.32 and 5.53 for men and women, respectively. The mean age was 43.2 yr. Compared to individuals with normal renal function, participants with low eGFR tended to be older and have higher body mass index, lower education level, and higher comorbidity (all p < 0.0001). Individuals with normal renal function were more likely to have healthy lifestyle behaviors and less likely to be taking long-term medications than the other eGFR groups (all p < 0.0001).

Association between renal function and UC incidence
Cumulative hazard curves for UC are shown in Figure 2. Significant differences were noted across the four eGFR strata (p < 0.0001). 1.14-2.09) were identified as significant risk factors for UC. A significant dose-response relationship between PM 2.5 concentration and UC risk was also noted (p trend < 0.01).

Sensitivity analysis
To explore the possibility of reverse causality, we performed a sensitivity analysis in which separate HRs were estimated for follow-up durations of 10 yr and >10 yr ( Table 3). The HR remained elevated beyond 10 yr from enrollment, suggesting that reverse causality explains a limited part of the relationship and strengthening the notion that lower eGFR may be a risk factor for UC. The other two sensitivity analyses are shown in Supplementary Table 2: one treated death before cancer incidence as a competing event using the subdistribution hazard approach; the other model was constructed after excluding participants who provided coordinates for their company and not their residential location. The results from both analyses were not considerably different from those for the main model.

Stratified analysis
Given the knowledge that men have higher risks for both lower renal function and UC incidence, we conducted a gender-stratified analysis ( Taiwan uniquely has a higher proportion of UTUCs among all UCs in comparison to Western countries, and the proportion of women with UTUC is surprisingly higher than for men [20]. Supplementary Table 3 shows the proportions of UC events for upper and lower urinary tract sites by gender in this study. Owing to the unique incidence pattern for Taiwan, we performed an analysis stratified by cancer site (Supplementary Table 4). A dose-response relationship between eGFR and UC was observed for both UTUC and LTUC. However, male gender was not a significant risk factor for UTUC (HR 1.16, 95% CI 0.52-2.59). In addition, smoking was only a risk factor for LTUC (HR 1.36, 95% CI 1.02-1.81) and the risk associated with PM 2.5 was much greater for

Discussion
In this retrospective cohort study, we identified a strong and significant dose-response relationship between low renal function and UC incidence. Besides the well-established risk factor of smoking, we also identified PM 2.5 exposure as a possible risk factor for UC.
Mechanisms that may contribute to tumorigenesis include accumulation of carcinogenic compounds, a weakened immune system, systemic inflammation, and impaired DNA repair mechanisms [7][8][9]. Some other possible mechanisms that have been discussed are alterations of the renin-angiotensin system [21] and some medications used by CKD patients [1,8]. All the aforementioned mechanisms may result in an increase in the risk of cancer.
Smoking is the most well-established risk factor for bladder cancer, and smoking cessation may help in decreasing bladder cancer risk [22]. We also found that individuals who lived in areas with higher PM 2.5 concentrations were more likely to develop UC (p trend < 0.01), consistent with previous studies [12,13]. The most common hypothesis for the mechanism underlying this association is that PM 2.5 exposure may lead to systemic inflammation and oxidative stress [23]. In addition, PM 2.5 particles can enter the circulatory system through the alveoli and have adverse effects on remote organs [23]. The toxicity of chemical compounds such as heavy metals and polycyclic aromatic hydrocarbons that adhere to PM may also pose risks [24]. However, a metaanalysis indicated that PM 2.5 exposure has no association with UC [25]. It is possible that because of lower PM concentrations in Western countries where the studies were carried out, they did not detect this link. Thus, more studies on different ambient PM 2.5 concentrations are warranted to validate this relationship.

Strengths and limitations
This study has several strengths. First, it is a large, populationbased cohort study with favorable generalizability. This is confirmed by the fact that our findings for UC incidence (data   shown in Supplementary Table 6) are similar to those in the 2017 Taiwan Cancer Registry Annual Report (https://www.hpa. gov.tw/Pages/Detail.aspx?nodeid=269&pid=12235), which is 10.15 per 10 5 population overall, and 12.99 and 7.66 for men and women, respectively. Second, the study had longer follow-up than similar studies, and we conducted a sensitivity analysis to explore the possibility of reverse causality. There are also some limitations. Patients with ESRD have different characteristics from others and have a higher cancer risk [3]. Owing to the lack of information on medical history, we were not able to exclude patients with ESRD before the analysis. Checking of eGFR data revealed that there were only 305 individuals with eGFR <15 ml/min/1.73 m 2 at baseline among the entire study population, of whom five developed UC before the end of the follow-up period. In addition, as our study population was from a database for routine health examinations, we presume that only a few people may have had ESRD. Another limitation is the lack of detailed information about the clinical condition of patients at the time of their UC diagnosis. It would be better if we could take into consideration complications caused by UC that could also impair renal function and adjust for these potential confounders. However, we found that very few people (0.01%) reported ever having had nephrosis before baseline. Combined with the view that UC has a relatively short lead-time and the long follow-up duration for the UC groups (9.4 yr; Supplementary Table 8), we believe that it is unlikely that complications caused by UC could affect renal function in our cohort. Moreover, it is theoretically plausible that patients with poorer renal function have a more intense follow-up schedule and may have ended up with a higher UC detection rate compared to healthy patients. The data and biological samples from the MJ Health Database were voluntarily donated by participants, who were apparently healthy and receiving physical examination services from the MJ Health Management Institution. The number of Analyses were restricted to participants for whom complete data for all covariates in the model was available. * p < 0.05. donations, as well as the time interval between each donation, varies from one participant to another. However, from the similar rates for UC incidence (Supplementary Table 6) and tumor stage (Supplementary Table 7) we observed in the MJ database compared to Taiwan, we would optimistically speculate that the disease risk for participants in the MJ cohort is close to that of the general population in Taiwan. In addition, previous evidence suggests that even short-term exposure or exposure in recent years can serve as a favorable surrogate for long-term PM exposure because of the high correlation in annual exposure between different geographical locations over time [26]. Thus, we conducted a sensitivity analysis restricted to participants who gave their residential address, and the results were similar to the original results.

Conclusions
In conclusion, lower renal function is an independent risk factor for UC incidence. Even a mild reduction in renal function leads to higher UC risk, indicating that improving the management of CKD is critical in the prevention of UC. This finding may also aid in the development of a complete cancer screening strategy: individuals with lower renal function should monitor their health not only in terms of the kidneys themselves but also the other organs in the urinary system. We also found that smoking and long-term exposure to PM 2.5 are both risk factors for UC. Tobacco and air pollution control policies should be strictly implemented for prevention of UC and many other adverse health outcomes.
Author contributions: Ta-Chien Chan had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Chan, Chuang, I-F. Lin.
Analysis and interpretation of data: Chuang, Chan, I-F. Lin.
Drafting of the manuscript: Chuang, Chan.
Critical revision of the manuscript for important intellectual content: Chan.
Administrative, technical, or material support: Chan.