Calderwood AH, Huo D, Rubin DT. Association Between Colorectal Cancer and Urologic Cancers. Arch Intern Med. 2008;168(9):1003-1009. doi:10.1001/archinte.168.9.1003
Different types of urologic cancers have been associated with colorectal cancer (CRC) in hereditary nonpolyposis CRC, but it is still unknown whether there is an association between urologic cancer and CRC in the general population. We sought to quantify the risk for CRC after urologic cancer and the risk for urologic cancer after CRC in patients without known genetic syndromes.
We performed a retrospective cohort analysis of the Surveillance, Epidemiology, and End Results program database from 1973 to 2000. Standard incidence ratios (SIRs) of observed to expected cases of invasive CRC were calculated for each urologic cancer site based on age, sex, ethnicity, and calendar year of diagnosis. Similar analysis was performed to determine the SIRs of urologic cancers in patients with previous CRC.
Overall, the risk for CRC was increased among patients with previous ureteral cancer (SIR, 1.80; 95% confidence interval [CI], 1.46-2.20) and renal pelvis cancer (SIR, 1.44; 95% CI, 1.20-1.72). This risk was greatest among patients who received the diagnosis of renal pelvis or ureteral cancer before the age of 60 years. There was a minimal increased risk for subsequent CRC in patients with bladder or renal parenchymal cancer. Overall, the risk for urologic cancer was increased after a diagnosis of CRC (SIR, 1.24; 95% CI, 1.20-1.28), with the highest risk for subsequent renal pelvis and ureteral cancers in patients with a CRC diagnosis before the ages of 50 to 60 years or multiple primary CRCs.
Previous renal pelvis and ureteral cancers, particularly when diagnosed at an early age, increase the risk for subsequent CRC. Likewise, a history of CRC, especially in cases with multiple primary tumors, is associated with an increased risk of renal pelvis and ureteral cancers. These findings support a possible common pathogenetic mechanism between CRC and urologic cancers and may have implications for screening guidelines.
Screening guidelines for colorectal cancer (CRC) emphasize the importance of identifying individuals at increased risk compared with the general population.1,2 Among those at increased risk for CRC are patients with a family history of CRC as well as those who meet established criteria for hereditary syndromes.3,4 The most common hereditary colon cancer syndrome, hereditary nonpolyposis CRC (HNPCC), confers an increased risk of colon cancer as well as extracolonic tumors of the uterus, ovaries, stomach, biliary tract, renal pelvis, and ureter. Also, previous studies have shown that a history of ovarian and endometrial cancer increases the risk for subsequent CRC in the general population.5 There are numerous case reports describing patients with both urologic cancer and CRC,6- 10 but the pathogenetic or environmental factors related to such associations have been incompletely characterized outside of knownHNPCC pedigrees.11- 15
The primary aim of our study was to quantify the overall risk for CRC after urologic cancer and the risk for urologic cancer after CRC. A secondary aim was to determine the risk for CRC and urologic cancer in different subgroups defined by type of urologic cancer (or CRC), age at diagnosis, ethnicity, sex, cancer stage at diagnosis, and duration of follow-up.
We performed a retrospective cohort analysis of the Surveillance Epidemiology and End Results (SEER) public database, a network of 14 population-based cancer registries containing data from 1973 to 2000, with 3 registries containing data from 1992 to 2000. The catchment area includes approximately 14% of the US population.16 Approval for this study was obtained from the institutional review board at The University of Chicago Hospitals, Chicago, Illinois.
We used an analytic strategy similar to the approach outlined by Schoenberg and Myers,17 calculating standardized incidence ratios (SIRs) of a second primary cancer in a defined cohort of patients with a first primary cancer. The analysis was performed in 2 directions: (1) the association of urologic cancer with subsequent CRC, and (2) the association of CRC with subsequent urologic cancer.
We identified patients in the SEER database who were diagnosed as having urologic cancer and CRC between 1973 and 2000 and had follow-up information. Follow-up time was defined as time from first primary cancer diagnosis to the date of diagnosis of the second primary cancer, the date of last physician visit, or the date of death, whichever occurred first. Using the unique identifier assigned to these patients, we searched the database for records of subsequent invasive CRC or urologic cancer. If multiple primary tumors of the same site were diagnosed, only the first occurrence was included, because for this study, the per-patient analysis was more clinically relevant than the per-event analysis.
The incidence rate was calculated as the total number of urologic cancers divided by the total years of follow-up in the cohort of patients with CRC and as the total number of CRCs divided by the total years of follow-up in the cohort of patients with urologic cancer. The rate of CRC in the cohort of patients with urologic cancer and the rate of urologic cancer in the cohort of patients with CRC were then compared with the standard rates in the general population, also derived from the SEER program over the same period from 1973 to 2000. An SIR was calculated as the ratio of observed to expected cases on the basis of age-, sex-, ethnicity-, and calendar year–specific standard rates. Therefore, SIRs can be interpreted as the age-, sex-, ethnicity-, and calendar year–adjusted relative risks.
We further calculated the SIR separately for each cancer site, as well as for subgroups within each site defined by age at first primary cancer diagnosis, sex, cancer stage (in situ, local, regional, distant, and unstaged), ethnicity, duration of follow-up, and whether there were multiple first primary cancers. Poisson regression methods incorporating external standard rates were used to estimate the SIR and the corresponding 95% confidence interval (CI).18 When the number of observed events was fewer than 30, we calculated exact 95% CIs based on Poisson distribution. Synchronous CRC was defined as 2 primary CRCs diagnosed within 6 months of each other, whereas metachronous cases were defined as 2 primary nonmetastic CRCs diagnosed more than 6 months apart.
We identified 194 329 patients with a urologic cancer and excluded those in whom the urologic cancer diagnosis was made at autopsy or on death certificate (n = 2536), those with no follow-up (n = 4098), and those whose race was unknown (n = 723). After these exclusions, 186 972 patients remained: 52 449 patients with renal parenchymal cancer, 6403 patients with renal pelvis cancer, 3744 patients with ureteral cancer, and 124 376 patients with bladder cancer.
Table 1 lists the frequency distribution of select baseline characteristics of patients with urologic cancer and the subcohort with subsequent invasive CRC, as reported to the SEER database (1973-2000). The mean (SD) age at diagnosis was 62.0 (16.3) years for patients with renal parenchymal cancer, 68.4 (12.4) years for patients with renal pelvis cancer, 70.4 (10.6) years for patients with ureteral cancer, and 69.2 (12.4) years for patients with bladder cancer. The majority of patients in our analysis were white men, with an underrepresentation of women and minority races. The characteristics of patients in this cohort reflect the overall population included in the SEER database. The urologic cancer stage of patients at diagnosis was most often stage I, II, or III, with only a very small percentage of patients whose stage was unknown.
The median follow-up time was 2.5 years for patients with renal parenchymal cancer, 3.2 years for those with renal pelvis cancer, 3.4 years for those with ureteral cancer, and 4.1 years for those with bladder cancer. Of the 6758 cases involving patients with multiple urologic cancers (4% of total patients with urologic cancers), 2666 (39.5%) were synchronous and 4083 (60.5%) were metachronous. We identified 2789 people with urologic cancer who developed subsequent CRC, giving an overall unadjusted invasive CRC incidence rate of 272.2 per 100 000 person-years (95% CI, 262.3-282.5). This rate was equivalent to an SIR of 1.13 (95% CI, 1.09-1.17).
Table 2 lists the SIRs in each cohort, both overall and stratified by age at urologic cancer diagnosis, sex, ethnicity, urologic cancer stage, and duration of follow-up. Overall, patients with cancer of the upper urinary tract (ie, renal pelvis cancer and ureteral cancer) had a significant increase in risk of subsequent CRC as compared with the general population (SIR, 1.44; 95% CI, 1.20-1.72 vs SIR, 1.80; 95% CI, 1.46-2.20). Patients who received the diagnosis of renal pelvis cancer before the age of 50 years were about 5 times more likely to have subsequent CRC (SIR, 4.94; 95% CI, 2.47-8.84). Similarly, a diagnosis of ureteral cancer before the age of 60 years conferred a more than 2-fold increased risk for subsequent CRC.
There were no statistically significant differences in CRC rates for patients with a diagnosis of renal pelvis cancer based on race, sex, urologic cancer stage at diagnosis, duration of follow-up, or whether patients had multiple primary urologic cancers (all P values >.05). Certain subgroups, in particular female sex and black race, contained relatively smaller numbers of cases and, as such, their SIRs were not statistically significant. In the subgroup of patients with a diagnosis of ureteral cancer, the incidence of subsequent CRC was not affected by race, sex, stage, duration of follow-up, or whether there were multiple primary urologic cancers.
Patients with kidney parenchymal cancer and bladder cancer had a less strong but still statistically significant increased risk of CRC (SIR, 1.14; 95% CI, 1.04-1.25; and SIR, 1.10; 95% CI, 1.05-1.14, respectively) as compared with the general population. Also, the SIR of patients with renal parenchymal cancer was 1.12 (95% CI, 1.02-1.23) after those who received a diagnosis within the first 6 months were excluded, and the SIR of patients with bladder cancer was 1.08 (95% CI: 1.04-1.13) after those who received a diagnosis within the first 6 months were excluded, demonstrating significance even after possible screening bias was controlled for. Finally, whether or not patients had multiple primary urologic cancers did not influence the incidence of subsequent CRC.
We identified 375 931 patients with CRC and excluded those in whom CRC was diagnosed at autopsy or on death certificate (n = 4634), those with no follow-up (n = 13 513), and those whose race was unknown (n = 1187). After these exclusions, 357 597 patients remained: 251 946 patients with colon cancer and 105 651 patients with rectal cancer. As mentioned above, the final analysis included herein combines the data for colon and rectal cancers (as denoted by CRC) because the results when analyzed separately were similar (data on file, not shown).
Table 3 lists the frequency distribution of select baseline characteristics of patients with CRC and the subcohort with subsequent invasive urologic cancers, as reported to the SEER database (1973-2000). The mean (SD) age at diagnosis was 70.0 (12.5) years for patients with colon cancer and 67.5 (12.6) years for patients with rectal cancer. There was a predominance of white men in our patient sample, with an underrepresentation of women and minorities. Most patients were diagnosed as having local or regional CRC. The median follow-up time was 2.8 years (interquartile range [IQR], 0.9-7.3 years). Of the 18 656 cases involving patients with multiple primary CRC (5% of the total number of patients with CRC), 9406 (50.4%) were synchronous and 9250 (49.6%) were metachronous.
Our analysis of the differences in characteristics between the full cohort and the subcohort in which urologic cancer developed revealed that the average follow-up time in the subcohort was 1 year longer than in the full cohort. As expected, the proportion of men in the subcohort was higher than in the full cohort, as men have a higher risk of urologic cancer than women. Also, the proportion of whites was higher in the subcohort than in the full cohort.
We identified 3026 patients with CRC who received a subsequent diagnosis of urologic cancer. The overall unadjusted incidence rate for invasive urologic cancer after CRC was 168.9 per 100 000 person-years (95% CI, 162.9-175.1). This rate translated into an SIR of 1.24 (95% CI, 1.20-1.28). The median lag time from a diagnosis of CRC to a diagnosis of urologic cancer was 3.8 years (IQR, 1.3-8.1 years). This lag time varied among the different subcohorts divided by type of urologic cancer: 2.6 years (IQR, 0.5-6.9 years) for patients with renal parenchymal cancer, 4.1 years (IQR, 1.8-7.2 years) for patients with renal pelvis cancer, 5.9 years (IQR, 2.5-10.5 years) for patients with ureteral cancer, and 2.6 years (IQR, 1.1-9.9 years) for patients with bladder cancer.
Table 4 lists the SIRs for each cancer outcome stratified by age at CRC diagnosis, sex, ethnicity, CRC stage at diagnosis, and duration of follow-up. Patients with CRC had a 59% increased risk of subsequent renal pelvis cancer (SIR, 1.59; 95% CI, 1.31-1.91) and a 100% increased risk of subsequent ureteral cancer (SIR, 2.00; 95% CI, 1.59-2.47). The risk for the development of renal parenchymal and bladder cancer was also higher, but these increases may be partly the result of intensive screening. After patients who received a diagnosis within the first 6 months were excluded, the SIR was 1.21 (95% CI, 1.12-1.32) for patients with renal parenchymal cancer and 1.10 (95% CI, 1.05-1.15) for patients with bladder cancer.
Overall, the relative risk of subsequent urologic cancer was influenced by several variables, including age at diagnosis, race, duration of follow-up, and presence of multiple primary CRCs. Patients diagnosed as having CRC before the age of 50 years had a 2-fold increase in their risk for any subsequent urologic cancers as compared with their counterparts in the general population (SIR, 2.09; 95% CI, 1.70-2.55). Relative risk by age of diagnosis differed among the different types of urologic cancer. Patients with CRC who received a diagnosis before the age of 50 years had a 2.29-fold (95% CI, 1.65-3.09) increase in subsequent renal parenchymal cancer, a 6.20-fold (95% CI, 2.49-12.78) increase in subsequent renal pelvis cancer, a 6.37-fold (95% CI, 1.31-18.62) increase in subsequent ureteral cancer, and a 1.66-fold (95% CI, 1.21-2.23) increase in subsequent bladder cancer. Also, the relative risk of subsequent ureteral and bladder cancer was stronger in nonwhite patients than in white patients.
The elevated risks for renal pelvis and ureteral cancer were more dramatic in patients with multiple primary CRCs than in patients with a single CRC. Patients with multiple primary CRCs had a 3-fold increased risk of subsequent renal pelvis cancer (SIR, 3.20; 95% CI, 1.86-5.12) and a 5-fold increased risk of subsequent ureteral cancer (SIR, 5.30; 95% CI, 3.09-8.49) compared with the general population.
In this study of a national database, patients with a diagnosis of urologic cancer were found to be at a higher risk for developing subsequent CRC than the general population, and patients with a previous diagnosis of CRC were at a higher risk for developing subsequent urologic cancer. Specifically, patients with previous cancer of the ureter or renal pelvis were at an 80% and a 44% increased risk, respectively, for subsequent CRC compared with the general population, while patients with bladder or renal parenchymal cancer had a minimally increased risk that is statistically significant but unlikely to be clinically meaningful. Diagnosis of these 2 types of urologic cancers before the age of 60 years significantly increased the risk of subsequent CRC. On the other hand, patients with CRC had a 59% increased risk of developing renal pelvis cancer and a 100% increased risk of developing ureteral cancer. The relative risks for these 2 urologic cancers were particularly pronounced in patients with a diagnosis of CRC before the age of 60 years and in those with multiple primary CRCs.
Possible explanations of this 2-directional association include shared environmental risk factors for the different types of cancer, screening bias, a genetic predisposition common to both cancers, or the effect of treatment of one type of cancer on the other. Shared environmental risk factors between urologic cancer and CRC, such as smoking, diet, and exposure to carcinogens, may increase the risk for the development of primary cancers in multiple organs and may account for some of the concordance of urologic cancer and CRC in this study. Screening bias is another possible explanation, as patients diagnosed as having 1 cancer will likely undergo a more thorough examination, increasing the chance of finding a second, otherwise asymptomatic cancer. Most secondary primary urologic cancers are discovered by computed tomographic scanning, magnetic resonance imaging, and ultrasonography during preoperative examinations.9,10 We performed stratified analyses according to duration of follow-up. After incidence cases diagnosed in the first 6 months were excluded, the conclusion that CRC was associated with renal pelvis and ureteral cancer remained valid, suggesting that surveillance bias was not an alternative explanation for this association. A third potential explanation for the 2-directional association is a shared genetic predisposition toward both urologic cancer and CRC, such as a mismatch repair defect. Hereditary nonpolyposis CRC, an autosomal dominant condition, is the most common genetic colon cancer syndrome and acts by disrupting the functionality of various mismatch repair genes. Individuals carrying a mutation in 1 of these genes have an 80% lifetime risk of developing CRC19 and a well-described increased risk of developing extracolonic tumors, including endometrial, ovarian, ureteral, and renal cancers.20 Finally, therapy of urologic cancer may lead to cellular changes that predispose these patients to the subsequent development of CRC.
We believe that this is the first study to demonstrate an association between urologic cancer and CRC in a cancer database of this size and reflects a true association in the population. The findings in this study, which demonstrate that the diagnosis of colorectal or urologic cancer in a single patient increases the likelihood of a subsequent invasive cancer of one or the other type, are of great interest and have implications for the follow-up of patients with one or the other type of cancer. A previous study has described microsatellite instability testing in patients with multiple tumors in the HNPCC tumor field to be a cost-effective and feasible method for identifying candidates for HNPCC testing,21 indicating that microsatellite instability testing may be a reasonable diagnostic tool for patients with colorectal and urologic cancers. The fact that the association was strongest for patients with specific urologic cancers and in general for patients who received a diagnosis at a younger age is not surprising, since the young age is suggestive of a more likely genetic (and less likely environmental) effect.
Although the strength of our study is the substantial size of the patient population, there are several limitations. Although the SEER database is a well-described database of high quality, there is always the possibility of some misclassifications and miscoding of tumor types, stages, or timing of diagnosis, as this information is dependent on accurate reporting by the clinician. We were unable to account for confounding variables, such as smoking history, family history, or treatment history, that might provide insight into the cancers that were identified. As mentioned, it is unclear how much impact a screening bias may have had, although our analysis attempted to address this limitation.
In conclusion, our study findings show that some patients with specific urologic cancers are at a higher risk for a subsequent invasive CRC and may benefit from earlier CRC screening and more frequent surveillance colonoscopic examinations. Likewise, patients with a primary diagnosis of CRC, especially those of a younger age, should be considered at increased risk for specific urologic cancers. Further studies need to determine whether screening with urinalysis or renal imaging would be beneficial and cost-effective. In the ongoing efforts to stratify colonoscopic resources and prevention strategies, patients with certain urologic cancers should be considered a group at increased risk for CRC.
Correspondence: David T. Rubin, MD, Department of Medicine, Section of Gastroenterology, The University of Chicago Medical Center, 5841 S Maryland Ave, MC 4076, Chicago, IL 60637 (email@example.com).
Accepted for Publication: November 21, 2007.
Author Contributions: Dr Rubin had full access to all of 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: Calderwood and Rubin. Acquisition of data: Calderwood, Huo, and Rubin. Analysis and interpretation of data: Calderwood, Huo, and Rubin. Drafting of the manuscript: Calderwood, Huo, and Rubin. Critical revision of the manuscript for important intellectual content: Calderwood, Huo, and Rubin. Statistical analysis: Huo. Administrative, technical, and material support: Rubin. Study supervision: Rubin.
Financial Disclosure: None reported.
Previous Presentation: This study was presented in part at the American College of Gastroenterology; October 13, 2003; Baltimore, Maryland.
Additional Contributions: Jeremy T. Hetzel assisted in writing and editing the manuscript for this article.