Bladder and upper urinary tract cancers as first and second primary cancers

Abstract Background Previous population‐based studies on second primary cancers (SPCs) in urothelial cancers have focused on known risk factors in bladder cancer patients without data on other urothelial sites of the renal pelvis or ureter. Aims To estimate sex‐specific risks for any SPCs after urothelial cancers, and in reverse order, for urothelial cancers as SPCs after any cancer. Such two‐way analysis may help interpret the results. Methods We employed standardized incidence ratios (SIRs) to estimate bidirectional relative risks of subsequent cancer associated with urothelial cancers. Patient data were obtained from the Swedish Cancer Registry from years 1990 through 2015. Results We identified 46 234 urinary bladder cancers (75% male), 940 ureteral cancers (60% male), and 2410 renal pelvic cancers (57% male). After male bladder cancer, SIRs significantly increased for 9 SPCs, most for ureteral (SIR 41.9) and renal pelvic (17.2) cancers. In the reversed order (bladder cancer as SPC), 10 individual FPCs were associated with an increased risk; highest associations were noted after renal pelvic (21.0) and ureteral (20.9) cancers. After female bladder cancer, SIRs of four SPCs were significantly increased, most for ureteral (87.8) and pelvic (35.7) cancers. Female bladder, ureteral, and pelvic cancers associated are with endometrial cancer. Conclusions The risks of recurrent urothelial cancers were very high, and, at most sites, female risks were twice over the male risks. Risks persisted often to follow‐up periods of >5 years, motivating an extended patient follow‐up. Lynch syndrome‐related cancers were associated with particularly female urothelial cancers, calling for clinical vigilance.

Conclusions: The risks of recurrent urothelial cancers were very high, and, at most sites, female risks were twice over the male risks. Risks persisted often to follow-up periods of >5 years, motivating an extended patient follow-up. Lynch syndrome-related cancers were associated with particularly female urothelial cancers, calling for clinical vigilance.
K E Y W O R D S cancer etiology, relative risk, renal pelvic cancer, second primary cancer, ureter cancer, urothelial cancer 1 | INTRODUCTION Urothelial carcinomas include bladder cancer (90-95% of all) and cancer of the upper urinary tract (UUT), of which two-thirds are located in the renal pelvis and the remaining in the ureter. 1 Bladder cancer is characterized by male excess, ranging from three-to sixfold. 2,3 International incidence trends have found correlation with the regional smoking prevalence. 2,4 Bladder cancer incidence in Swedish men has been low and relatively stable at about 20/100 000, while for women, the rate has increased, reaching an incidence of about 6/100 000. 5,6 Smoking prevalence is not the only explanation to the sex difference in bladder cancer incidence in Sweden, because smoking prevalence in men and women has been approximately equal since 1980 and dropping from 30% to below 10%; around 1950, half of men smoked compared to 10% of women 6 (www.pnlee.co.uk/downloads/iss/iss-sweden_ 111024.pdf). UUT cancers show also male excess. 1 All urothelial cancers share a number of risk factors, including smoking, occupational exposures, family history, and association with Lynch syndrome. 1,[7][8][9] However, while smoking appears to be the main risk factor for bladder cancer compared to Lynch syndrome, the opposite may be the case for UUT cancers. 10,11 Colorectal cancer and endometrial cancers are traditional hallmarks of Lynch syndrome (www.lscarisk.org). 12,13 Few guidelines to date suggest urological follow-up with Lynch patients. 1 In Sweden, survival has slightly improved for bladder cancer since about 1980. [14][15][16] In 2012-2016, the relative 5-year survival for female bladder cancer was 72% but it was higher, 77% for men (http://wwwdep.iarc.fr/NORDCAN/english/frame.asp). Early detection, novel imaging technologies, and improvements in treatment have contributed to positive trends in bladder cancer survival. 5 At bladder cancer diagnosis, some 20-25% of patients present with muscle invasive tumors, and the remaining patients have superficial tumors, which can later progress to invasive cancer. 17 Surgery is the main treatment mode for urothelial cancers. For bladder cancer, non-muscle-invasive tumors are transuretherally resected while muscle-invasive tumors are typically treated with cystectomy; both of which can be supplemented with chemotherapy or immunotherapy. 5,17 Radiotherapy may be used for bladder preservation. 18 For UUT cancers, treatment may involve removal of the ipsilateral ureter and kidney. 1 Improved survival implies that the likelihood of second primary cancers (SPCs) increases. SPCs after bladder cancer show typically high risks of tobacco-related cancers of the lung and head and neck. [19][20][21][22] Studies on SPCs in UUT cancer patients are limited, and one of the problems is to distinguish independent SPCs from recurrences. 21 Recurrence of UUT urothelial cancer into the bladder is relatively common, while seeding from the bladder into UUT is rarer. 1,[23][24][25] Plausible etiologies for SPCs are many, but probably the most important ones are intensive medical surveillance after the diagnosis of first primary cancer (FPC), therapy for FPC, shared genetic or nongenetic risk factors between FPC and SPC and immune dysfunction, or interactions between these. [26][27][28] As data on the possible risk factors for SPC are usually limited, we have devised a bidirectional analysis as a tool to help etiological search. [26][27][28][29] In this analysis, we want to define risks for specific subsequent cancers related to bladder and UUT cancers using Swedish nationwide data. Risks for SPCs are assessed pair-wise as FPC and SPC, that is, the standardized incidence ratio (SIR) of lung cancer was assessed after bladder cancer, and, alternatively, SIR of bladder cancer was assessed after lung cancer. The bidirectional analysis will help to distinguish, at least to some extent, the influence of treatment and medical surveillance on SIR because two different cancers are usually treated and diagnosed in different ways. As the previous literature has focused on SPCs after bladder cancer, we hypothesized that the novel type of bidirectional analysis will be able to produce novel qualified data on risks for bladder and UUT cancers.

| Diagnostic codes and nomenclature
We considered bladder and UUT cancers diagnosed from 1990 to 2015 in the Swedish Cancer Registry using International Classification of Diseases (ICD) version 7 and later codes. The project database is located at the Center for Primary Health Care Research in Malmö, Sweden.
Code 1810 was used for bladder cancers and, of these, 98% are transitional cell carcinomas. 30 The code for ureter cancer was 1811. 31 Urethral cancer was not considered because of its rarity and late introduction of a specific diagnostic code. Kidney cancer (ie, renal + pelvis/calyx) was identified with code 180, renal cell cancer (RCC) with code 1800, and pelvic/calyx cancer with 1801. The correctness of classification of UUT cancer in the Cancer Registry has been evaluated, and 93% were found to be correct; the misclassification frequently involved other urinary tract tumors. 32 In this article, we consider a possible "recurrence" when a second urothelial cancer occurred at the same anatomic site or at another urothelial site, following the practice of the European Association of Urology. 1

| Patient follow-up periods and methods
Bladder and UUT cancer patients were followed from year 1990 through year 2015 for diagnosis of any common SPC, and, conversely in a reverse order, any common cancer was FPC and bladder, and UUT cancers were SPCs. The other cancers include any of 21 common male and 22 female primary cancers. For a proper bidirectional analysis, we excluded cancers for which 1-year survival was less than 50% (esophagus, pancreas).
Patients were followed for SPCs from the diagnosis of FPC until the end of 2015 or immigration or death, whichever came earliest. Only discordant (different) FPC-SPC pairs were included without applying any lag time between the two diagnoses. The upper aerodigestive tract (UAT) included the lip, oral cavity, pharynx, and larynx. For skin cancer, only squamous cell carcinoma (SCC) was included. For the risk of "all" cancers, bladder and UUT cancers were excluded. The SIR for "all" cancers was weighed according to person-years at risk. No latency time was apply between diagnoses of FPC and SPC because in the Swedish Cancer Registry, practically all cancers are histologically verified and thus true cancers. 31 The application of a latency time would have caused bias because a large number of true SPCs would have been missed.

| Calculation of relative risk
Sex-specific SIRs were calculated to measure the risk of SPCs as the ratio of observed to expected number of cases. For risk of a certain SPC, the expected number of cases was calculated by strata-specific person years in patients with diagnosis of first primary bladder or UUT cancers, multiplied by strata-specific incidence rates of the same SPC as FPC in the general population. The strata were specified by sex, 5-year age group, 5-year-calendar period, socioeconomic status, and place of residence. In the reverse analysis, SIRs for second bladder and UUT cancers were calculated in the same way. The two-tailed 95% confidence intervals (95% CIs) of SIRs were calculated by assuming a Poisson distribution. The expected numbers can be obtained by dividing observed numbers with SIR. The method of SIR calculation is based on indirect standardization, and it is particularly suitable for datasets with small case numbers because the expected numbers are calculated from the large background population, all Sweden in this case. 33 Bidirectional SIRs for bladder cancer were summarized in forest plots. Pearson correlation coefficient was used to estimate the relation between the bidirectional SIRs for bladder cancer (FPC vs SPC) as well as the SIRs of the common cancers in men and women (men vs women). In addition, we carried out period-specific analysis by calculating SIRs during 1, 2-5, and >5 years after first primary cancer diagnosis. All the statistical analyses were performed in SAS 9.4, and forest plot was generated in R 3.3.6. In order to simplify the tables, we did not show data for cancers with less than five cases in any comparison, unless the SIR was significant.
The difference between two SIRs was considered significant when their 95% CIs did not overlap. Only significant SIRs were commented on. In the tables, some lines for the urothelial cancers were repeated in the reverse analyses, but they were kept to help comparisons.

| RESULTS
During the follow-up (inclusion) period of 1990 to 2015, we identified 46 234 bladder cancers, 940 ureter cancers, and 2410 renal pelvic cancers ( Table 1). The total number of other cancers considered as SPCs or FPCs was 513 693 for men and 496 600 for women in the concurrent Swedish population of 6.2 million men and 6.2 million women.
Bidirectional SIRs for male SPCs for bladder and UUT cancers are shown in Table 2. After bladder cancer, SIRs were significantly increased for 10 SPCs (counting RCC and renal pelvis but not kidney), most for ureteral (SIR 41.9), renal pelvic (17.20), and small intestinal (2.38) cancers. SIR for RCC was 2.20. The overall SIR for any SPC was 1.56. In the reversed order (bladder cancer as SPC), 10 individual FPCs were associated with an increased risk; highest associations were noted after renal pelvic (21.0), ureteral (20.9), and testicular (2.01) cancers. Second bladder cancer risk was 1.45 after RCC. The SIR for all cancers was 1.28.
Only five cancers (ureteral, pelvic, RCC, lung and prostate cancers) were bidirectionally associated. For six cancer pairs, the bidirectional SIRs differed significantly (ie, the 95% CIs did not overlap); for ureteral, stomach, lung, RCC, and prostate cancers, the SIRs were higher when these cancers were SPCs than in the reverse order; for skin SCC, the opposite was the case. The data are summarized in Figure 1A.
Bladder cancer accounted for half of all SPCs. Ureteral cancer as SPC was associated with renal pelvic (129.0, the highest risk recoded in this study) and bladder (41.9) cancers. After renal pelvic cancer, bladder (21.0) and ureteral (129.0) cancers were increased. Bladder cancer accounted for more than half of all SPCs. In the reverse order, the same cancers and RCC (3.94) showed associations. Similar analysis is shown for female cancer in Table 3 The data are summarized in Figure 1B.
After ureteral cancer, increased risks of SPCs were noted for renal pelvic (197.8), bladder (69.9), lung (4.10), and breast (2.33) cancers ( 10 FPCs, only five male cancers showed a bidirectional association. For women, four cancers were bidirectionally increased. Bidirectional associations of bladder cancer were shared for men and women for ureteral, renal pelvic, RCC, and lung cancers. The correlation analysis showed high concordance (P < .00021) between men and women when SPCs were compared after bladder cancer or in the reverse order, bladder cancer as SPC. Although previous studies have identified risks of smoking-related SPCs after bladder cancer, the novel results in this study define bidirectional risks between many sites not related to smoking. [19][20][21][22] In the below discussion, we compare the consistency of the results internally and externally (with published literature) and speculate about the putative mechanisms, keeping in mind the limitations of observational epidemiology.

| Urological sites
Urothelial carcinomas have been described as a pan-urothelial disease with a propensity to recur throughout these sites. 24,36 As with many other cancers, even intratumoral heterogenicity has been described. 37 However, a recent literature review concluded that most recurrent urothelial tumors are monoclonal which would imply that the mechanism of spread would be intraluminal seeding or intraepithelial migration. 24 These in turn would imply that the direction of urine flow and anatomic vicinity would play a role. In a Spanish study, concomitant primary urothelial tumors were found in 17% of the patients. 36 The likelihood of finding a concomitant bladder cancer increased by anatomic location of the primary tumor, being 10, 18, and 33% in patients with primary caliceal/renal pelvic, upper ureteral, and lower ureteral cancers, in line with the above predictions. The panurothelial disease is the likely explanation why the risks among urothelial sites far exceeded those between urothelial and nonurothelial sites.
In the present study, the male risk of second bladder cancer was equally high (SIR 21) when renal pelvic or ureteral cancers were FPCs, but, for women, the risks were slightly higher after ureteral (69.9) than after pelvic (60.5) cancers. Most urothelial cancer associations were higher in women than in men. The female SIRs were more than doubled and significantly higher compared to male rates for FCP-SPC pairs of bladder-pelvis (SIR in women 35.7), pelvis-bladder (60.5), bladderureter (87.8), and ureter-bladder (69.9). Although the risks were highest during the year of diagnosis, they persisted often to follow-up period of >5 years (conditional on patient survival). These risks were exquisitely high and clinically relevant, motivating an extended patient follow-up.
Bladder cancer was by far the most common SPC after first ureteral and renal pelvic cancers, accounting for at least half of all SPCs.
Of urological interest is also the possible SPC risks between bladder,

| Strengths and limitations
This study has a number of strengths, the foremost being a nationwide coverage and access to a high-level cancer registry data. 31,46 Because of linkage to the censuses, we had information on socioeconomic and residential background data covering the whole population, which is unique in nation-wide studies. Socio-economic data are highly correlated with lung cancer incidence in Sweden and thus provide a proxy of smoking level. [47][48][49] This kind of studies with numerous comparisons will produce some significant associations by chance. Assessing both sexes separately and comparing the consistency was helpful in avoiding chance findings. The bidirectional design is another strength in helping to interpret the associations.
Ureteral and renal pelvic cancers are rare, and all related case numbers were low, affording low statistical power. The major limitation was that data on treatment were lacking, but we assume that surgery has been the primary treatment for bladder and UUT cancers.

| CONCLUSIONS
We showed that many cancers were associated with bladder cancer as SPCs or, in the reverse order, bladder cancer as SPC. The risks of recurrence of urothelial cancers were very high, and, at most sites,

CONFLICT OF INTEREST
None.

DATA AVAILABILITY STATEMENT
DATA AVAILABILITY The data that support the findings of this study are available from Lund University but restrictions apply to the availability of these data, which were used under license for the current study and so are not publicly available.

ETHICAL STATEMENT
The study was approved on 25 January 2013, by the Regional Ethical Review Board in Lund without requirement for informed consent (registration number 2012/795). Instead, the Ethical Review Board obliged us to advertise in the newspapers in order to inform the public that their data would be used for secondary purposes. After that, around 40 individuals were excluded (≈40 excluded individuals/≈10 million individuals in the Swedish population).