In this population-based case–control study among elderly adults, we found that ESRD was not associated with an overall increased risk of cancer. These results differ from those seen in most previous studies [5
], which have shown a modest overall increased risk of cancer. This difference may be explained by the restriction of our study to elderly adults. As noted above, Maisonneuve et al. reported an attenuation with age in the association between ESRD and cancer risk, such that in the U.S. the overall SIR for cancer declined from 4.6 for those whose age at first dialysis was <35
years to 1.1 for individuals who were age 65
years or older [5
]. This overall attenuation could be due to the increase of background cancer rates with age, but the results likely differ for individual cancer types.
Indeed, although overall cancer risk was not elevated in association with ESRD, we did see increased risks for certain cancers such as cancers of stomach, small intestine, colon, liver, biliary tract, lung, cervix, and kidney, as well as for multiple myeloma and chronic myeloid leukemia, where risks were elevated 1.16 to 2.40-fold. Among cancers of the kidney and pelvis, we found an approximately 3-fold excess of renal cell carcinoma but no increase for transitional cell carcinoma. Previous studies reported an excess risk of several of these malignancies, including multiple myeloma and bladder and kidney cancers [4
]. Some prior studies found increased risks for other cancer that we did not find associated with ESRD, including thyroid cancer, non-Hodgkin lymphoma (for which we found a modest inverse association), Hodgkin lymphoma, and Kaposi sarcoma [5
]. Again, the differing magnitudes of association between prior studies and our study could potentially be attributed to our subjects’ advanced age, while other studies included a wider age range.
Notably, the increased risks that we observed for several infection-related cancers (i.e., liver, stomach, and cervical cancers) could reflect immunodeficiency associated with ESRD. Chronic uremia leads to metabolic abnormalities that alter the immune response such that antigen-presenting cell (APC) function is impaired, lymphocyte survival is shortened, proliferation of T-cells is impaired, together with an increased suppressor T-cell activity, decreased helper T-cell (Th) activity, and increased Th1/Th2 ratio [2
]. Nutritional abnormalities that are prevalent among individuals with ESRD, such as selenium deficiency, reduced glutathione peroxidase activity, and vitamin D deficiency [17
] could also play a role in development of cancer, particularly for colon cancer [21
]. Individuals with ESRD commonly receive erythropoietin for treatment of anemia, which might be implicated in carcinogenesis and could potentially explain some of the observed excess cancer risk among individuals with ESRD [24
However, arguing against a biological effect underlying some of these associations, we did not observe an increase in cancer risk with longer duration of ESRD. Another possibility is that, rather than reflecting the effects of ESRD itself, the high risk of some cancers may arise from a high prevalence of established cancer risk factors (e.g., infections with human papillomavirus [cervical cancer], HBV and HCV [liver cancer], or tobacco use [lung cancer]) [25
]. Indeed, this possibility is supported by our observation that the association between ESRD and liver cancer was attenuated after we adjusted for HCV infection, HBV infection, or diabetes mellitus.
In addition, excess risks were observed for malignancies (i.e., bladder and kidney cancers, multiple myeloma) that could directly lead to ESRD or are related to such conditions. Cancers of the renal pelvis and bladder may be attributed to excess analgesic consumption, a cause of renal failure [9
]. Likewise, polycystic kidney disease, a cause of renal failure, predisposes to renal cell carcinoma [31
]. Nonetheless, ESRD itself probably causes some kidney cancers, as the development of multiple renal cysts during dialysis, with the subsequent occurrence of renal cell carcinoma, is well-described [33
]. This notion is further supported by the study of Maisonneuve et al. [5
] that detected an increased risk of kidney cancer with increasing duration of dialysis.
In contrast, the strong increase in multiple myeloma risk in the first 1–2
years after ESRD diagnosis, and declining risk over longer time intervals, could be attributed to reverse causality, such that multiple myeloma leads to renal failure. This situation could arise if multiple myeloma was undiagnosed at initiation of dialysis, or if the diagnosis date is in error in the cancer registry due to delayed reporting. Maisonneuve et al. also noted the excess of multiple myeloma cases and likewise attributed this observation to prevalent cases [5
]. Finally, treatment of glomerulonephritis with azathioprine or cyclophosphamide may predispose to certain types of cancers such as kidney cancer, bladder cancer, and multiple myeloma [41
Our findings of markedly low risk of prostate cancer deserve comment. A biologically protective effect of ESRD seems unlikely, because the inverse association was present only for localized cancer and not for distant/metastatic cancer. We believe that the inverse association instead may be a manifestation of screening. Recommendations for prostate cancer screening using PSA testing and digital rectal examination are presently in flux, but screening has typically been considered only for men who have an extended life expectancy. One possibility is that, due to their lower life expectancy, men with ESRD do not receive screening with the PSA test as frequently as other men, which could lead to an apparent deficit of prostate cancer. Indeed, it is estimated that 40% of prostate cancer cases in the U.S. are detected through PSA testing [43
]. Among controls, we found a lower prevalence of claims for testing among men with ESRD, which supports that the apparent diminished risk is partly due to lack of screening. However, we also found a lower risk for prostate cancer associated with ESRD among men who received PSA testing, which could potentially be explained by decreased sensitivity of PSA testing in men with ESRD. This notion is supported by Bruun et al. [44
], who reported that individuals with chronic kidney disease have elevated levels of free PSA, which is interpreted (when the total PSA is elevated) to indicate the presence of benign prostate disease. In addition, the association between ESRD and reduced risk of prostate cancer among apparently unscreened men may partly reflect confounding due to unmeasured screening that we did not capture. In summary, we believe that the reduced risk of localized/regional prostate cancer in men with ESRD may be explained by less frequent screening or reduced sensitivity of screening. However, we do not advocate for increasing the frequency of screening among men with ESRD. Recently the U.S. Preventive Services Task Force (USPSTF), opposed the use of PSA for prostate cancer screening regardless of age because the harms overweigh the benefits [45
]. Given, the overall poor survival of men with ESRD, PSA-based prostate cancer screening will not be of additional value.
Our study has several important strengths, including the ability to assess all type of cancers and its large size. Our study is representative of the elderly U.S. population, since it includes all cancers from SEER cancer registries and a random sample of controls with Medicare from the same geographic areas [12
]. The use of SEER allowed comprehensive ascertainment of the occurrence of cancers, since these cancer registries achieve largely complete surveillance and provide high-quality data on incident cancers.
Limitations of our study should also be noted. For instance, SEER does not capture squamous cell and basal cell carcinoma of the skin. In addition, we were unable to assess cancer risk in people younger than 66
years old, so our results are not generalizable to a younger population. Furthermore, because ESRD is somewhat uncommon, we may have missed some associations with cancer. Ascertainment of ESRD in our study was likely complete, because Medicare pays for dialysis care for most U.S. ESRD patients. Our use of Medicare claims is supported by the similar prevalence of ESRD among our controls and reported by the USRDS program for the same age group (0.31% of our controls vs. 0.21-0.44% during 1992–2005) [46
]. However, because we used dialysis claims to identify people with ESRD, we may have inadvertently included some people with less severe kidney disease. In addition, we did not have the capability to fully determine duration of ESRD because Medicare claims were only available starting in 1991. Another limitation that should be considered is the possibility that competing mortality from secondary complications of dialysis would attenuate the association between ESRD and cancer. This effect would occur in any study of the ESRD population, and the results should be interpreted to reflect risk of cancer among surviving patients. A final limitation is that we made multiple comparisons, and some associations might be due to chance. Fewer associations would be statistically significant using a conservative p-value threshold of 0.002 (based on a Bonferroni correction for 31 cancer types, see Table ).
Our results may help inform clinical decisions about cancer screening. Given the lower risk of prostate cancer in men with ESRD, and the lower life expectancy of these men relative to the general population there is no indication that men with ESRD should receive PSA testing. These findings also do not support more aggressive screening for other common cancers, because there was no increased risk for breast cancer, and only a modest increase in colon cancer risk. Similarly, others have suggested that in the circumstances of high morbidity and mortality from other diseases, screening for kidney cervical, colon, and breast cancers is less cost-effective than other health interventions in the ESRD population [47