The objectives of this study were to determine whether bladder cancer (BC) screening in healthy men could lead to earlier detection and reduced BC mortality compared with unscreened men and to determine long-term outcomes of a geographically defined, unscreened population with newly diagnosed BC.
In 1987 and from 1998 to 1992, 1575 men ages 50 years and older who were solicited from well patient rosters in clinics in and around Madison, Wisconsin, tested their urine repetitively with a chemical reagent strip for hemoglobin. Participants who had positive test results underwent standard urologic evaluation. BC grades and stages and the outcomes of men with BC detected by screening were compared with the grades, stages, and outcomes of 87% of men ages 50 years and older with newly diagnosed BC who were reported to the Wisconsin Tumor Registry in 1988 (n = 509 men).
Two hundred fifty-eight screening participants (16.4%) were evaluated for hematuria, and 21 participants (8.1%) were diagnosed with BC. Proportions of low-grade (Grade 1 and 2) superficial (Stage Ta and T1) versus high-grade (Grade 3) superficial or invasive (Stage ≥ T2) cancers in screened men (52.4% vs. 47.7%) and in men from the tumor registry (60.3% vs. 39.7%) were similar (P = .50). The proportion of high-grade superficial or invasive BCs that were invasive were lower in screened men (10%) than in unscreened men (60%; P = .002). At 14 years of follow-up, no men with screen-detected BC had died of BC, whereas 20.4% of men with unscreened BC had died of BC (P = .02).
Bladder cancer (BC) is the sixth most common noncutaneous malignancy in the United States with >63,000 new diagnoses in 2005.1 Virtually all patients who die from BC do so because of metastases, and the vast majority have muscle-invasive cancers (Stage ≥ T2) preceding or concurrent with the diagnosis of metastases.2, 3 Although BCs frequently recur after transurethral resection (TUR) with or without intravesical chemotherapy or immunotherapy,4 the majority of patients with muscle-invasive cancers have that level of invasion at the time of initial diagnosis5–8 and do not have a history of prior superficial (Stage ≤ T1) tumors. It is noteworthy that the treatments for Stage ≥ T2 BCs have greater morbidity and are far less successful than the treatments for superficial cancers.9–12 Thus, the rationale for promoting early BC detection is to diagnose cancers that are destined to become muscle-invasive before they invade. Sufficient information is available, based on tumor architecture, cystoscopic appearance, and histology, to distinguish superficial cancers (Stages Ta, T1, and tumor in situ [TIS]) that are likely to become muscle-invasive from tumors that are not likely to invade deeply.13–17 An additional factor that is important for screening is that BC is very rarely, if ever, found incidentally at autopsy.18–21 Thus, individuals who are diagnosed through screening would not be subject to “over diagnosis,” because their cancers almost certainly would have become evident, and they would have received treatment during their lifetimes.
Based on this rationale and on the recognition that most BCs are diagnosed because they induce hematuria,4, 22 we conducted BC screenings by using repetitive hematuria home testing in healthy men age 50 years and older. We compared the tumor grades, disease stages, and outcomes of men who had screening-detected BCs with those of all men age 50 years and older who had newly diagnosed BC reported to the Wisconsin Cancer Reporting System (WCRS) in 1988 (a year in which the screenings were not conducted). Preliminary findings indicated reduced 2-year BC mortality in the screening group.23 To determine whether lead time bias or other factors were responsible for this seeming benefit from screening and whether overall survival was affected, in the current study, we have updated these data with 14-year follow-up. In addition, for the first time to our knowledge, we present the long-term outcomes of a geographically defined population of patients with this disease.
MATERIALS AND METHODS
The screening cohort consisted of men age 50 years and older who were recruited from well patient clinic rosters in the counties in and around Madison, Wisconsin. The methods and findings of solicitation and evaluations have been reported previously.24–27 Briefly, 3515 men were solicited for both studies, and 1575 men (44.8%) participated. Men with histories of hematuria, hematuria-producing diseases, prior unexplained or unevaluated hematuria, prior treatment with pelvic radiation therapy, or other medical or psychiatric conditions that would impair their ability to comply with testing or evaluation protocols were not solicited or, if such conditions were found on eligibility checks after solicitation, were not allowed to participate in screenings. Two different screening protocols were used. In a 1987 pilot study, 235 men tested their urine at home by using Ames Hemastix daily for 5 days and then weekly thereafter for 51 weeks. From 1989 to 1992, 1340 similar men who were recruited from a broader geographic base were asked to test their urine for 14 consecutive days at the beginning of the study and to repeat this test 9 months later if all tests during the first 14 days were negative. In both studies, if a single test was positive (greater than or equal to “trace”), was not explained by physical or sexual trauma, and was not the first void in the morning, then men were asked to undergo a hematuria evaluation, which included a history and physical examination by a urologist, serum creatinine, complete blood count, prothrombin time, microscopic urinalysis, urine culture, intravenous urography (or comparable upper-tract imaging), office cystoscopy, and voided or bladder wash cytology. Men who had an attributable cause of hematuria (e.g. urinary infection) were not evaluated unless hematuria persisted for at least 6 weeks after the cause was treated successfully. Both screening studies were approved by the University of Wisconsin's Institutional Review Board, and all screening participants provided written informed consent. Because any screening test may not detect all diagnosable BC, we queried the WCRS whether any participant without or with hematuria (in whom BC was not detected) had BC diagnosed within 12 months of what would have been their last home testing date. These men and the screenees with BC who were diagnosed by screening comprised our “screened BC” group (n = 21 men).
Of 590 BCs reported to the WCRS in men age 50 years and older in 1988, pathology materials were requested on all men and were received for 527 men (89.3%). These materials were reviewed by a single pathologist (K.W.G.) for grade and histologic stage. In 16 men, cancer could not be identified on the slides or blocks sent, and it was determined that 2 men had been diagnosed before 1988, which left 509 men for review. The screened BCs detected in the years surrounding 1988 also were reviewed by the same pathologist, who was unaware of the grades and stages attributed by the original pathologists for either screened or unscreened men. Cancers were divided into 3 groups: low-grade superficial (LGS) (Grade 1 and 2, Stage Ta and T1), high-grade superficial (HGS) (Grade 3 or carcinoma-in-situ, Stage TIS, Ta, or T1), and muscle-invasive (INV) (any grade, Stage ≥ T2).28 Primary bladder malignancies that were not predominantly urothelial (transitional cell) cancer (n = 18 men) were categorized as high-grade cancers because of their aggressive behavior.4 WCRS records were used to assign the dates and causes of death and were confirmed by death certificates. All investigators were blinded to the identities of unscreened men and to the code that linked identities to histology or outcomes.
Demographic characteristics of the Wisconsin male population age 50 years and older were obtained from the Wisconsin Behavioral Risk Factor Survey (WBRFS), a population-based telephone interview that was conducted from 1988 to 1990.29 These characteristics were compared with the characteristics of the screening participants.
Significance tests were used to assess the differences in proportions of grades and stages between the screened and unscreened populations. A 2-sided Fisher exact test was used to detect the reduction of deaths (BC, other causes, or overall) by screening. To more efficiently study the efficacy of screening, we considered the 14 year follow-up data as survival data with death as the endpoint and used the stratified log-rank test (stratified by BC stage and grade) to test the difference between the screened and unscreened groups. Two-sided tests of significance are reported.
Results of the pilot study and the larger screening study were similar and, thus, are combined in this report. Two hundred fifty-eight of 1575 participants (16.4%) who had hematuria, as defined by at least 1 trace-positive Hemastix test, were deemed to be appropriate for evaluation and were included. Of these, 25 participants (9.7%) had urinary tract cancer detected, including 21 urothelial cancers of the bladder and 4 renal cell carcinomas. All of those cancers were detected by screening, and none were diagnosed in the year after the last home testing date. Greater than 90% of screenees (19 of 21 men) who had BC detected, compared with 59.9% of all screening participants, were current or former cigarette smokers (P = .0003). The mean age of screenees at the time of BC diagnosis was 69.2 years (median age, 70.0 years), and the mean age of all screenees was 64.8 years (median, 63.0 years; P = .024 for differences between mean ages; P = .001 for differences between median ages). Based on the WBRFS, the mean age of men age 50 years and older in Wisconsin was 64 years, 52% were retired (compared with 51% of screening participants), and 69% were former or current cigarette smokers. These and other characteristics of the general Wisconsin population and screening participants were similar statistically.23 The grades and stages of screened men with BC and deaths at a mean follow-up of 14 years are shown in Table 1. There were no deaths from BC in this group.
Pathology materials were received on 509 unscreened men who were diagnosed with BC in 1988. Because the histology obtained was only from the initial diagnostic TUR, data on depth of invasion beyond the muscularis propria were not available. The grade and stage distributions and the deaths from BC and other causes are shown in Table 2 and Figure 1.
Table 2. Bladder Cancer in Wisconsin, 1988 (14-Year Follow-Up)
The proportions of all newly diagnosed BCs that were LGS, 52.4% in screenees and 60.7% in tumor registry cases, did not differ significantly (P = .50). The proportions of HGS or INV cancers, 39.3% (unscreened) and 47.6% (screened), also were similar. However, 60% of HGS and/or INV unscreened cancers were muscle-invasive, whereas only 10% of HGS and/or INV cancers in screenees were muscle-invasive (P = .002). There were fewer deaths overall (P < .004) and from BC (P = .02) in men with screen-detected cancers. The lower overall mortality in screenees with BC primarily was because of the reduced mortality from BC in that group (Table 3).
Table 3. Outcomes at 14 Years of Follow-Up
BC indicates bladder cancer.
Unscreened men with INV BC who succumbed to their disease did so quickly (median time to death from BC, 1.05 years). However, unscreened men with INV BC died at similar rates from causes other than BC as unscreened men with superficial BCs (both LGS and HGS; P = .64), indicating that the differences in overall survival between the 3 groups with unscreened BC P ≤ .02) were caused exclusively by differences in deaths from BC (P < .0003) (Fig. 1).
To our knowledge, this is the first report of long-term outcomes of BC screening in a general population and also is the first report of long-term outcomes from a geographically defined segment of the general population with BC that has been characterized histologically by central histology review (CHR). CHR is needed, because tumor registries, such as the National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER) Program, do not record accurately the critical distinction between superficial and muscle-invasive disease30 and do not record histologic grade at all. Moreover, because there can be disagreement among pathologists about factors which are critical for prognosis, such as muscle invasion31–33 let alone histologic grade, CHR is needed to compare histologies of screened patients and unscreened patients. The outcomes from BC of unscreened patients based on stage and grade classifications closely reflect the results of large, randomized, prospective clinical trials12, 34 and clinical series16, 17, 35 indicating that, although we are uncertain of complete staging information, treatments received, recurrence rates, and mean lengths of follow-up, men in Wisconsin in 1988 received care that approached the best contemporary standards (in terms of survival from BC) and, unfortunately, that is not different from current standards. It is noteworthy that another contemporary, geographically based tumor registry series from Stockholm, Sweden, indicated a grade and stage distribution of newly diagnosed BC and short-term BC outcomes36 similar to those observed in the current study, confirming that data from our unscreened patients are generalizable to other Western populations.
The observation that unscreened men with INV BCs died of non-BC causes at a rate similar to the rate of unscreened men with HGS and LGS cancers who died of non-BC causes implies that the treatments for these cancers, although quite different between the 3 groups, probably contributed little to long-term mortality (beyond their affects on tumor or acute toxicities). Average survival from non-BC causes in unscreened men (8.72 years [standard deviation, 0.52 years] after the diagnosis of BC) was 2.3 years shorter than that reported for the contemporary Caucasian male population age 72 years in Wisconsin (P < .0001),37 indicating that, if patients with BC survive the lethal effects of their disease (and its treatment), then they still are likely to have a somewhat reduced life expectancy. This is not surprising considering the relatively unhealthy lifestyles (e.g. cigarette smoking)4, 38 and lower levels of education in patients with BC compared with the general male population.39 In addition, the similarity in mortality rates from non-BC causes between the 3 groups of unscreened men with BC and the similarities in ages between the groups (Table 2) implies that risk factors for these different types of urothelial cancer are similar, a finding that is supported by other lines of evidence.4, 38
The similar rates of non-BC causes of death in the 3 groups of unscreened patients with BC highlights the enormous impact of the disease and the timing of its discovery on patient outcome (Fig. 1). Because of similarities in risk factors and demographics between these groups, and because of the similarities in molecular profiles and clinical behaviors of both HGS cancers and INV BCs,3, 4, 10, 13–16 a strong argument can be made for attempting to detect BCs that are destined to become invasive at preinvasive stages.
Our screening cohort was similar to the male population age 50 years and older in Wisconsin in several ways (median age, smoking history, and racial composition) that have widely recognized, significant impacts on BC development and outcome.4, 40 Although we do not have information about risk factors in men with unscreened BC, based on their median age (72 years) and race (97% Caucasian),41 they were similar to the risk factors in men with screened BC (median age, 70 years; 100% Caucasian). Conversely, although the overall proportion of non-BC deaths during follow-up was similar in unscreened men with BC and screened men with BC (P = .38) (Table 3), unscreened men died of non-BC causes sooner after the initial diagnosis of BC than men with screen-detected BC (P = .04) (Fig. 1B). Whether this was caused by overall better health status in screenees with BC than in men with unscreened BC and/or was a reflection of lead time bias for non-BC causes of death (in screenees) is uncertain.
The 1940 solicited men who declined to take part in screening, as reported previously, according to the WCRS, had BC diagnosed at a rate similar to that in our screened population,23 indicating that our screenees had not selected themselves to take part in screening because of a perceived increased BC risk. Similarly, our rate of detecting BC (1.3%) was similar to that reported for a similar BC screening program in Leeds, England,42, 43 and was similar to the reported incidence of BC reported in men who underwent cystoscopy at the time of radical prostatectomy for prostate cancer44 and at the time men entered into a watchful waiting versus TUR prostate study for lower urinary tract voiding symptoms.45 Finally, the finding that the proportions of high-grade and low-grade cancers were similar in men with screen-detected BC and unscreened BC indicates that screening was not detecting a particularly indolent variant of the disease. The data we report here indicate that lead-time issues were not responsible for the initial reduction in BC mortality we observed in men with screened BCs.23 Despite the similarities between our screenees and the general population in terms of demographics, known BC risk factors, BC incidence, and the proportions of low-grade and high-grade cancers, we acknowledge that differences between our screenees and the general population may exist; therefore, a randomized trial of screening versus no screening to determine the efficacy of BC screening is needed.
The means and accuracy of our follow-up may be questioned, because there are limitations to the way tumor registries collect and analyze data, including the reporting of new cases, interpreting histology, and documenting death and its cause. We queried the WCRS for all malignancies that were found by screening (including 7 diagnosed prostate cancers in addition to 25 diagnosed urinary tract cancers) and noted 100% accuracy. Moreover, this registry routinely compares itself with SEER data from Iowa and performs quite favorably in terms of case capture and follow-up accuracy.46 In addition, the WCRS is in direct contact with other federal and state tumor and vital statistics registries; therefore, it is likely that most screened or unscreened BC diagnoses or deaths were captured. Finally, we tried to overcome the limitations and inconsistencies of histologic interpretation by having a CHR.
No screening participant who did not have BC detected by screening had it diagnosed subsequently within 1 year of his last home testing date. This indicates that repetitive hematuria testing is a very sensitive means of detecting BC. Although the majority of BCs continue to be detected because they produce hematuria,22 since the current study was carried out, other noninvasive tests have been developed, several of which are available commercially (BTA Stat, NMP22 BC, UroVysion FISH, and ImmunoCyt). Most tests report sensitivities from 80% to 92% for high-grade cancers, those BCs that are particularly desirable to detect through screening. However, tumor size has a direct influence on the sensitivity of most marker tests that have been evaluated, and several tests had far lower sensitivities for smaller BCs compared with larger BCs, even high-grade tumors.47 Furthermore, the performance characteristics for screening general populations are unknown for any of the marker tests. However, their performances in populations at risk for BC because of hematuria or other signs or symptoms have been reported. This is particularly important, because the majority of screenees with microhematuria do not have BC (the specificity and positive predictive value of hematuria home screening for BC are 84.7% and 8.1%, respectively). A 2-tiered screening may be valuable in which repetitive hematuria testing is performed first, and patients with positive hematuria tests undergo BC marker testing to determine whether cystoscopy should be performed. Such a protocol may be a reasonable model, and, if it is promising, suitable for prospective assessment. This may reduce the number of screenees without BC who undergo cystoscopy, probably would increase acceptance of the screening protocol by individuals who are solicited to participate, and possibly may reduce the cost of case finding.
Finally, based on how infrequently BC is found at autopsy, we assumed that it has a brief preclinical prevalence; thus, screening needs to be performed at moderately frequent intervals if it is to detect most tumors that are destined to become muscle-invasive before they actually invade. However, this has not been assessed prospectively.
In conclusion, repetitive home hematuria reagent strip testing is a sensitive means of detecting BCs that are destined to become muscle-invasive at preinvasive stages. This appears to improve survival from BC. These data provide compelling support for conducting a randomized, prospective screening trial using the test and design reported here or using other screening methods and protocols.