COMPARATIVE EVALUATION OF THE DIAGNOSTIC PERFORMANCE OFTHE BTA STAT TEST, NMP22 AND URINARY BLADDER CANCERANTIGEN FOR PRIMARY AND RECURRENT BLADDER TUMORS

ARIS GIANNOPOULOS, THEODOROS MANOUSAKAS, ANTONIA GOUNARI,CONSTANTINOS CONSTANTINIDES,

HELEN CHOREMI-PAPADOPOULOU AND CONSTANTINOS DIMOPOULOSFrom the Departments of Urology and Immunology, University of Athens Medical School, Laikon Hospital and G. Papanicolaou Research

Center, Saint Savas Hospital, Athens, Greece

ABSTRACT

Purpose: We compared overall sensitivity and specificity of the urinary bladder cancer antigenenzyme-linked immunosorbent assay (UBC, IDL Biotech, Sollentuna, Sweden), BTA stat test(Bion Diagnostic Sciences, Inc., Redmond, Washington) and NMP22 test kit (Matritech, Newton,Massachusetts), and the differential sensitivity regarding the histological pattern of tumors.

Materials and Methods: A total of 213 patients with clinical and/or imaging signs of bladdercancer provided a single voided urine sample for the bladder cancer antigen, BTA stat test andNMP22 before cystoscopy. Of these patients 95 were monitored for superficial bladder cancer,while the remaining 118 had no history of bladder cancer. All detected bladder tumors orsuspicious lesions were resected transurethrally. A group of 21 age and sex matched healthyvolunteers were also evaluated with the same tests.

Results: Bladder cancer was confirmed histologically in 118 patients, of whom primary andrecurrent tumors were in 68 and 50, respectively. The optimal cutoffs calculated with receiveroperating characteristics curves were 8 units per ml. for NMP22 and 12 mg./l. for bladder cancerantigen. Overall sensitivity and specificity were 72.9% and 64.6% for the BTA stat test, 63.5%and 75.0% for NMP22, and 80.5% and 80.2%, respectively, for bladder cancer antigen. Bladdercancer antigen proved significantly more sensitive than NMP22 for detecting bladder cancer (p 50.001) but not more than the BTA stat test, while the specificity of it was significantly higherthan that of the BTA stat test (p 5 0.009). Bladder cancer antigen had a sensitivity of 80.7% forstage Ta tumors, which was significantly higher than NMP22 (52.6%, p 5 0.001) and the BTAstat test (57.9%, p 5 0.01). In grade I tumors the sensitivity of bladder cancer antigen (70%) didnot differ significantly than that of the BTA stat test (50%) and NMP22 (50%, p 5 0.14). Bladdercancer antigen had the least false-positive results in patients with a history of bladder cancer andnegative cystoscopy, and those with urological disease other than bladder cancer.

Conclusions: Our data indicate that bladder cancer antigen may be a more potent diagnosticmarker for bladder cancer than NMP22 and the BTA stat test based on the higher sensitivity fordetecting low stage and low grade tumors, and the higher specificity. The contribution of thesetests for detection of bladder cancer should still be considered adjunctive to cystoscopy.

KEY WORDS: antigens; bladder neoplasms; tumor markers, biological

Approximately 75% of all newly diagnosed transitional cellcarcinomas of the bladder are limited to the mucosa (Ta) andlamina propria (T1).1 Although these superficial tumors areexcised easily with transurethral resection, they are associ-ated with a high recurrence rate up to 70% and 90% in 5 and15 years, respectively.2 Patients in whom recurrent tumordevelops in the first year are at higher risk for multiplerecurrences in the future.3 Another major problem with su-perficial transitional cell carcinoma of the bladder is thattumor progression to a higher stage or grade develops in 40%of the patients in 10 years, with half of them representinginvasive disease.4 Thus, this carcinoma of the bladder is notan “innocent” cancer, and has a mortality rate of 11% forstage Ta and 30% for stage T1 tumors in 20 years.5

The high probability of tumor recurrence has led to therequirement of a thorough followup with frequent cystosco-pies in patients with superficial transitional cell carcinoma.Because carcinoma in situ cannot always be visualized reli-ably, most urologists use cytology as an adjunctive diagnostic

procedure. However, cytology has several disadvantages, in-cluding dependence on the skill of the pathologist,6 a lowsensitivity of 20% to 40% in well differentiated tumors7 andrelatively high cost.8 Despite any information provided bycytology, cystoscopy remains the reference diagnostic proce-dure for bladder cancer but even when performed with aflexible instrument, it is still an invasive method and pa-tients have some discomfort. Special interest has been gen-erated in noninvasive objective urinary biomarkers to indi-vidualize the current followup protocols for patients withsuperficial bladder cancer. Because a biomarker with 100%sensitivity and negative predictive value that could replacecystoscopy, does not exist, the role of urine biomarkers is stilladjunctive to cystoscopy, substituting or eliminating theneed for cytology in certain indications.

The BTA stat test is a single step, qualitative immunochro-matographic assay for detection of bladder tumor associatedantigen in urine. This antigen has been identified as a hu-man complement factor H related protein, which is producedby bladder tumor cells in cell cultures and not by any otherAccepted for publication February 16, 2001.

0022-5347/01/1662-0470/0THE JOURNAL OF UROLOGY® Vol. 166, 470–475, August 2001Copyright © 2001 by AMERICAN UROLOGICAL ASSOCIATION, INC.® Printed in U.S.A.

470

epithelial cell lines.9 This protein protects bladder cancercells from complement mediated lysis.

The NMP22 test is an enzyme-linked immunoassay with 2monoclonal antibodies to measure the levels of complexedand fragmented forms of the nuclear mitotic apparatus pro-tein in stabilized urine. This protein is found in at least25-fold greater intracellular concentration in bladder cancercells compared with normal urothelium and is shed intourine during apoptosis.10 NMP22 was initially presented as aprognostic marker11 but in all recent studies it was evaluatedas a diagnostic procedure.12–16

The new urinary bladder cancer antigen test (UBC) is anenzyme-linked immunoassay that determines the levels ofcytokeratin 8 and 18 urinary fragments. Cytokeratins areintermediate cytoplasmic filaments expressed by all epithe-lial cells and are altered during epithelial differentiation.17

More than 20 different human cytokeratins have been iden-tified and several cytokeratin expression patterns have beendescribed as tissue specific.18 Therefore, serum cytokeratinfragments have already been tested as tumor markers forvarious malignancies, including bladder cancer in which theresults were unsatisfactory.19 The differential expression ofcytokeratins in normal and malignant urothelium, and thefact that bladder tumors are in contact with urine led to thesearch for cytokeratin fragments in urine as well as evalua-tion of their potential role as biomarkers in transitional cellcarcinoma of the bladder. We compare the diagnostic perfor-mance of 3 currently available biomarkers, including theBTA stat test, NMP22 and UBC enzyme-linked immunosor-bent assay, and their combinations for detection of bladdercancer. Furthermore, we compared the sensitivity of eachtest in regard to pathological stage and grade, with cystos-copy and histology as the “gold standard” for definitive diag-nosis.

MATERIALS AND METHODS

A total of 182 male and 31 female patients with a mean ageof 66 years (range 25 to 93) were enrolled in our study.Eligible patients were those who were suspicious for bladdercancer on the basis of clinical signs, symptoms and/or recentintravenous urographic or ultrasonographic results. Patientswith a history of superficial bladder cancer were eligible forenrollment in study. Those patients with known invasivebladder cancer or a history of upper tract transitional cellcarcinoma were not included in the study. Subjects withindwelling catheters or those who had undergone any kind ofgenitourinary manipulation during the 30 days before urinecollection were also excluded. Accordingly, patients on intra-vesical treatment protocols for superficial bladder cancershould have received the last instillation at least 30 daysbefore urine collection. Of our patients 95 were followed forsuperficial bladder cancer, while the remaining 118 had nosuch history. Written informed consent was obtained from allpatients.

All patients provided a fresh, single, midstream urine spec-imen the day before cystoscopy. Part of this specimen wasused for the BTA stat test, which was performed immediatelyby the same investigator for all patients. The remaining partwas divided into 2 aliquots for NMP22 and UBC, respec-tively. Specimens for NMP22 were immediately stabilizedaccording to manufacturer guidelines, centrifuged and sub-sequently stored at 220C until use. Specimens for bladdercancer antigen were also centrifuged and stored at 220C.The technicians who performed the measurements wereblinded to the results of the other tests. All tests were per-formed as previously described.11, 20, 21 All patients under-went cystoscopy, and all detected tumors or suspicious le-sions were resected. The final diagnosis of bladder cancerwas based on histological examination. Tumors were graded

according to the WHO grading system22 and staged accord-ing to TNM classification.23

Another group of 18 male and 3 female healthy volunteerswith a mean age of 65.3 years was enrolled in our study.These subjects were evaluated with the same tests but with-out cystoscopy. To minimize the probability of a healthyvolunteer having “undiagnosed” bladder cancer, we followedthese individuals for a median of 27 months (range 22 to 34).Bladder cancer symptoms did not develop in any of them andnone had any suspicious ultrasonographic examination. Ofthe 234 enrolled subjects 168 have already been reported onin a comparative study of the BTA stat test, NMP22 andcytology.14

The threshold value for optimal sensitivity and specificityof NMP22 and UBC was determined by receiver operatingcharacteristics (ROC) curves. Sensitivity, specificity, positiveand negative predictive values, accuracy and the 95% confi-dence interval were calculated according to standard statis-tical methods. Specificity was recalculated after excludingpatients with calculi, urinary tract infection, prostate cancerand renal cell carcinoma according to exclusion criteria re-ported by Sharma et al.24 The McNemar test was used todetect differences in overall sensitivity and specificity, and instratified stage and grade sensitivity results when the BTAstat test, NMP22 and UBC were evaluated in pairs. Pear-son’s chi-square test, with the Yates correction when needed,was used to estimate differences between proportional sub-groups.

RESULTS

Primary transitional cell carcinoma of the bladder wasidentified in 68 of the 118 patients with no history of bladdercancer, whereas recurrent tumor was found in 50 of the 95followed for superficial bladder cancer. Thus, our study groupwas comprised of 118 patients with detected bladder cancerand 116 with no evidence of bladder neoplastic disease. Thelatter group consisted of 3 subgroups, including 45 patientswith a history of superficial bladder cancer but no evidence ofdisease at cystoscopy, 50 with urological disease other thanbladder cancer (benign prostatic hyperplasia 16, stress incon-tinence 3, retroperitoneal fibrosis 1, urethral stricture 5,neurogenic bladder 2, stone disease 7, urinary tract infection6, prostate cancer 5 and renal cell carcinoma 5) and 21healthy volunteers. The ROC curve analysis indicated anoptimal threshold value greater than 8 units per ml. forNMP22 and greater than 12 mg./l. for UBC to differentiatenegative from positive results (see figure). Interestingly, thecutoff of 12 mg./l. for UBC was also recommended by themanufacturer.

Table 1 shows overall sensitivity, specificity, positive andnegative predictive values, and accuracy of all tests studied.The sensitivity of the BTA stat test, NMP22 and UBC was72.9%, 63.5% and 80.5%, respectively. Bladder cancer anti-gen was significantly more sensitive than NMP22 (p 50.001), while the differences between UBC versus the BTAstat test and the BTA stat test versus NMP22 were notsignificant (p 5 0.19 and 0.16, respectively). UBC had thehighest overall specificity (80.2%) compared with NMP22(75.0%) and the BTA stat test (64.6%). It was significantlymore specific than the BTA stat test (p 5 0.009) but notNMP22 (p 5 0.34). When the 23 patients with calculi, uri-nary tract infection, prostate cancer and renal cell carcinomawere excluded, the specificity of the BTA stat test andNMP22 improved from 64.6% to 73.1% and 75% to 80.6%,respectively. Neither the former nor the latter values differedsignificantly (p 5 0.07 and p 5 0.26, respectively). The spec-ificity of UBC also improved from 80.2% to 87.1%, which wasthe highest of all tests and maintained the significant differ-ence of the BTA stat test (p 5 0.03). In addition, UBC showed

COMPARISON EVALUATION OF 3 URINE DIAGNOSTIC MARKERS FOR BLADDER CANCER 471

higher positive and negative predictive values, and accuracythan both of the other tests.

We also calculated the combined sensitivity of all possibletest pairs and then all 3 together. The combination of theBTA stat test and UBC had the highest paired sensitivity of92.4% (range 86.0 to 96.0), which was significantly higherthan the sensitivity of either the BTA stat test or UBC (p,0.001, respectively). The combination of the BTA stat testand NMP22 had a sensitivity of 89.8% (range 82.9 to 94.2),and NMP22 and UBC had the lowest sensitivity (85.6%,range 78.0 to 90.9). The sensitivity of the triple combination(94.9%, range 89.2 to 97.7) did not differ significantly thanthat of the BTA stat test and UBC combination.

Table 2 shows the sensitivity of the BTA stat test, NMP22and bladder cancer antigen according to stage, grade andhistory of bladder cancer. The BTA stat test was slightlymore sensitive than NMP22 in all stage and grade subgroupsbut no difference was significant. Moreover, bladder cancerantigen was more sensitive than NMP22 in all stage andgrade subgroups, except in stages T2-T4, but the differenceswere statistically significant only in patients with stage Taand grade II tumors (p 5 0.001 and 0.003, respectively).Bladder cancer antigen was also significantly more sensitivethan the BTA stat test in stage Ta tumors (p 5 0.01). In gradeI tumors bladder cancer antigen had 70% sensitivity, rela-tively higher than that of either the BTA stat test or NMP22

(50%, respectively) but this difference did not reach a signif-icant level (p 5 0.14). Accordingly, in 27 stage Ta grade Itumors bladder cancer antigen had 74.1% sensitivity, higherthan that of either the BTA stat test (51.9%) or NMP22(55.6%) but not significantly (p 5 0.14 and p 5 0.17, respec-tively). Combined sensitivity of bladder cancer antigen andthe BTA stat test was 85.2% for stage Ta grade I tumors,which was significantly higher than that of the BTA stat test(p 5 0.003) but not bladder cancer antigen. The sensitivity ofbladder cancer antigen and NMP22 was equally 85.2% forstage Ta grade I tumors, which was significantly higher thanthat of NMP22 (p 5 0.008). All tests had similar sensitivity ininvasive and grade III tumors.

Regarding the presence of a history of bladder cancer, table2 shows that the sensitivity of the BTA stat test and bladdercancer antigen was equal between primary and recurrenttumors. Although NMP22 tended to be more efficient fordiagnosing primary than recurrent tumors (sensitivity 69.1%versus 56.0%, respectively), this difference was not signifi-cant (p 5 0.14). We also evaluated the distribution of primaryand recurrent tumors by stage and grade. The number ofprimary and recurrent tumors did not differ significantly ineach subgroup of stage (p 5 0.15) or grade (p 5 0.16).

We further analyzed the performance of all biomarkers inthe subgroups of patients who did not have bladder cancer(table 3). All 3 tests showed equal specificity in healthyvolunteers. The BTA stat test had 22 false-positive results inpatients with urological disease, whereas NMP22 and blad-der cancer antigen had 16 and 14, respectively. However,after application of the exclusion criteria reported by Sharmaet al,24 the BTA stat test had the highest improvement inspecificity, and the difference of the other tests was reducedbut bladder cancer antigen still had the best specificity.Moreover, UBC had only 8 false-positive results in patientswith no evidence of disease, 3 less than NMP22 and 9 lessthan the BTA stat test.

DISCUSSION

Although invasive and costly, cystoscopy is currently con-sidered the reference diagnostic procedure for bladder can-cer.25 Due to the high recurrence rate for the disease, itshould be performed frequently. Urinary cytology has notbeen proved reliable enough to reduce the number of cysto-scopies because of the well-known shortcomings.6–8 The re-cently described BTA stat test and NMP22 proved to besignificantly more sensitive than voided urine cytology fordetecting bladder cancer at the expense of a modest reductionin specificity.13, 14, 26, 27 We have focused on the comparison ofthe BTA stat test and NMP22 to the newer biomarker, UBC.

UBC is a urine biomarker that traces fragments of cyto-keratins 8 and 18. In our study the sensitivity of UBC fordetection of bladder cancer was 80.5%, with a cutoff of 12mg./l. There are 2 studies that were performed at the samecenter that have shown comparable sensitivity of UBC withour results. In the first study the reported sensitivity of UBCwas 87% with 9.74 mg./l. cutoff,21 while in the second thereported sensitivity was 74% but with 15.5 mg./l. cutoff.28 On

TABLE 1. Overall sensitivity, specificity, positive and negative predictive values, and accuracy of the 3 tests

No. Pts.95% CI

BTA Stat Test NMP22 UBC

Sensitivity (range) 118 72.9 (64.0–80.2) 63.5 (54.4–71.8) 80.5 (72.3–86.7)Specificity (range) 116 64.6 (55.4–72.9) 75.0 (66.2–82.1) 80.2 (71.8–86.5)Specificity (range)* 93 73.1 (63.1–81.2) 80.6 (71.2–87.5) 87.1 (78.6–92.5)Predictive value: –

Pos. 67.7 72.1 80.5Neg. 70.1 66.9 80.2

Accuracy – 68.8 69.2 80.3* After exclusion of 23 patients with stones, urinary tract infection and urological malignancies other than bladder cancer.

ROC curves for determination of UBC and NMP22 reference valueof optimal sensitivity and specificity for detection of bladder transi-tional cell carcinoma. Inflection points are nearest to 12 mg./l. and 8units per ml., respectively.

COMPARISON EVALUATION OF 3 URINE DIAGNOSTIC MARKERS FOR BLADDER CANCER472

the contrary, Mian et al found 64.8% sensitivity of UBC,although they used a cutoff of 12 mg./l.29 This low value mightbe attributed to the small number of patients with bladdercancer that was evaluated. However, the reported low sensi-tivities for stage T1 and grade III tumors in that studyremain unclear.

Comparing the sensitivity of UBC and NMP22 we foundsignificant superiority of UBC over NMP22 (80.5% versus63.5%, respectively, p 5 0.001). Mian et al compared UBCwith NMP22 in a small number of patients and found UBC tobe minimally more sensitive than NMP22.29 Sanchez-Carbayo et al also compared UBC with NMP22 and reportedsimilar sensitivity (73.9% versus 75.7%, respectively).28 Al-though the results of the latter study are comparable to oursas far as UBC is concerned, the sensitivity of NMP22 is muchhigher than that of our sample, with no obvious explanation.However, it is noteworthy that the sensitivity of NMP22 inthat study is the highest reported. The sensitivity of UBC(80.5%) was not significantly higher than that of the BTAstat test (72.9%). To our knowledge there is no other reportedstudy comparing the BTA stat test to UBC enzyme-linkedimmunosorbent assay.

Analyzing the differences in sensitivity by grade and stage,UBC was more sensitive than the other markers in stage Ta(80.7%), and grades I (70%) and II tumors (80%). In stageTa tumors the superiority of UBC over the BTA stat test andNMP22 was statistically significant. In grade II tumors UBChad significantly higher sensitivity than NMP22 (p 5 0.003)but not the BTA stat test (p 5 0.13). The high sensitivity ofUBC in grade I tumors was confirmed in the 2 aforemen-tioned studies (70% and 66.6%, respectively).28, 29 In stage Tagrade I tumors UBC had a relatively higher sensitivity(74.1%) than the BTA stat test (51.9%) and NMP22 (55.6%).However, the combined sensitivity of UBC and either theBTA stat test or NMP22 for stage Ta grade I tumors reached85.2%, which represents a significant facilitation for detec-tion of the most difficult to detect tumors. Nevertheless, all 3biomarkers performed similarly for detection of tumors inhigher stages (T1 and greater) and grade (III).

In our patients the specificity of UBC was 80.2%, signifi-cantly higher than that of the BTA stat test (64.6%, p 50.009) but not NMP22 (75%). When we excluded the patients

with lithiasis, urinary tract infection, prostate cancer andrenal cell carcinoma according to Sharma et al,24 the speci-ficity of the BTA stat test and NMP22 improved to 73.1% and80.6%, respectively. Moreover, the specificity of UBC im-proved to 87.1%, remaining higher than the other tests. Thisresult indicates that the selection of patients to be evaluatedwith UBC has the same limitations of the BTA stat test andNMP22 to minimize test interference. The higher specificityof UBC than the BTA stat test and NMP22 in our study ismainly explained by the fewer false-positive results in pa-tients with no evidence of disease. It is noteworthy that Mianet al reported an extremely low rate of false-positive resultsfor UBC in patients with no evidence of disease (3 of 115).29

There are no reported assumptions for the role of thesefalse-positive results. On the contrary, we have assumed thatthe false-positive results of BTA stat test and NMP22 havesome prognostic value for recurrence in the near future14 butthis has to be confirmed in a larger series. Thus, the highnumber of false-positive results of the BTA stat test in pa-tients with no evidence of disease should still be considered adrawback of the test.

Calculating the combined sensitivity of all possible biomar-ker pairs we found that NMP22 increased the individualsensitivity of UBC to 85.6%. Sanchez-Carbayo et al reportednearly the same combined sensitivity (86.5%) of NMP22 andUBC.28 However, the highest combined sensitivity in ourdata was obtained by UBC and the BTA stat test (92.4%),which significantly improved the individual sensitivity ofeach marker. Moreover, the fact that the sensitivity of thetriple biomarker combination (94.9%) was not significantlyhigher than that of the BTA stat test and UBC indicates thatthe results of the latter biomarkers mainly overlap those ofNMP22.

In our patient group the BTA stat test had an overallsensitivity of 72.9%, which represents an intermediate valueof the sensitivity results in other studies, ranging from 65%to 82.8%.20, 26, 27, 30 Generally, our results were comparable toall of the studies, except for 2 apparently significant differ-ences. Pode et al30 reported a sensitivity of 72.2% for stage Tatumors in contrast to 57.9% in our data. They also reported90% sensitivity for detecting primary tumors in contrast to

TABLE 3. Calculated specificity of the 3 tests for each subgroup of the bladder cancer-free cases

No.% Specificity (No. false-pos. results)

BTA Stat Test NMP22 UBC

Healthy volunteers 21 90.5 (2) 90.5 (2) 95.2 (1)Pts. with urological disease 50 56.0 (22) 68.0 (16) 72.0 (14)Pts. with no evidence of disease 45 62.2 (17) 75.5 (11) 82.2 (8)Pt. subgroup with urological disease* 27 77.8 (6) 81.5 (5) 88.9 (3)

* Does not include 23 patients with stones, urinary tract infection and urological malignancies other than bladder cancer.

TABLE 2. Sensitivity of the 3 tests according to stage, grade and bladder cancer history

No. Primary Tumors/No. RecurrentTumors

% Sensitivity

BTA Stat Test NMP22 UBC

Stage:Ca in situ 2/4 100 83.3 83.3Ta 30/27 57.9 52.6 80.7T1 21/11 78.1 65.6 75.0T2–4 15/5 95.0 90.0 85.0Tx* 0/3 – – –

Grade:I 13/17 50.0 50.0 70.0II 27/18 73.3 55.5 80.0III 28/15 88.4 81.4 88.4

Bladder Ca history: 68/50Primary tumor 73.5 69.1 80.9Recurrent tumor 72.0 56.0 80.0

* Improper specimens with no stage determination.

COMPARISON EVALUATION OF 3 URINE DIAGNOSTIC MARKERS FOR BLADDER CANCER 473

73.5% in our patients. The possible explanation for the latterdifference was that their primary tumors were of a higherstage and grade. There is also a noteworthy difference amongthe aforementioned studies regarding carcinoma in situ.Leyh,26 Raitanen27 and Pode30 et al found a sensitivity of100% for carcinoma in situ in a small number of these casesthat were similar in our study. The total number of carci-noma in situ cases in these 4 studies (including our study)was 20. On the other hand, Sarosdy et al found 61% sensi-tivity in 18 cases of carcinoma in situ.20 Because there is noreported study with a high number of patients with carci-noma in situ for the BTA stat test nor for any other urinebiomarker, cytology is still considered necessary for detectionof this “insidious” form of transitional cell carcinoma. How-ever, the majority of biomarker studies show at least equalresults with cytology for detection of carcinoma in situ.

Although a large number of studies concerning NMP22have been published, there is no agreement on the optimalcutoff. In contrast, there are already 7 proposed cutoffs, rang-ing from 3.6 to 14.6 units per ml.8, 11–16, 28 Thus, to comparethe apparently conflicting results in these studies, we have torefer to a common cutoff. By using a cutoff of 8 units per ml.,we found a 63.5% overall sensitivity for NMP22. We havecommented on all aforementioned studies in our previouswork,14 except those with the most extreme results.8, 16 Ra-makumar et al reported a sensitivity of 53% for NMP22 butwith a cutoff of 3.6 units per ml.8 Although the study wasbased on a small number of patients with bladder cancer, halfof them having stage Ta tumors, the authors could not as-sume any possible explanation for their disappointing re-sults. On the other hand, Sanchez-Carbayo et al reported thehighest sensitivity of 78.2% and 75.7% for NMP22, with acutoff of 13.7 and 14.6 units per ml., respectively, in 2 sepa-rate studies sharing the same patients.16, 28 Applying any ofthe latter cutoffs in our patients we obtained a sensitivity ofless than 50% for NMP22. Although the distribution of pa-tients in our and their studies is similar, the high differencein sensitivity of NMP22 remains unclear.

When comparing biomarkers, it is difficult to state which isthe best. Our data indicate that UBC is more sensitive in welldifferentiated and stage Ta tumors, as opposed to the othertests. This fact combined with the higher specificity suggeststhat UBC performed better than the BTA stat test andNMP22. It has been proposed that the BTA stat test mayreplace cytology as an adjunct to cystoscopy for the primarydiagnosis of bladder cancer. It may also be used as a substi-tute for the many periodic cystoscopies after the initial onefollowing primary resection of a superficial tumor.27 Anotheroption is that for low risk tumors, the BTA stat test can beperformed every 3 months and cystoscopy only annually.30

Both approaches appear reasonable, and our data suggestthat UBC could probably be used instead of the BTA stat test.However, such modifications in the current followup protocolof superficial bladder cancer should be tested in large scalerandomized trials.

CONCLUSIONS

Our data suggest that UBC enzyme-linked immunosorbentassay represents a significant improvement in the researchof bladder cancer diagnostic biomarkers and that it couldstand as a reliable adjunct to cystoscopy. The significantsuperiority over the BTA stat test and NMP22 is focused onthe higher sensitivity of stage Ta, grades I and II tumors aswell as higher specificity. It is also clear that testing with anyof the 3 biomarkers has to be performed under strict patientselection, excluding urological diseases that potentially in-terfere with results. Although more comparative studies arerequired, it seems that the process for identifying the idealbiomarker has been shortened.

Dr. Theodoros P. Vassilakopoulos provided technical ad-vice and critical review of the manuscript.

REFERENCES

1. Messing, E. M., Young, T. B., Hunt, V. B. et al: Comparison ofbladder cancer outcome in men undergoing hematuria homescreening versus those with standard clinical presentations.Urology, 45: 387, 1995

2. Newling, D. W.: Preventing recurrence and progression in su-perficial bladder cancer. Curr Opin Urol, 6: 272, 1996

3. Kurth, K. H., Denis, L., Bouffioux, C. et al: Factors affectingrecurrence and progression in superficial bladder tumours.Eur J Cancer, 31A: 1840, 1995

4. Herr, H. W.: Natural history of superficial bladder tumors: 10 to20 years followup of treated patients. World J Urol, 15: 84,1997

5. Holmang, S., Hedelin, H., Anderstrom, C. et al: The relationshipamong multiple recurrences, progression and prognosis of pa-tients with stages Ta and T1 transitional cell cancer of thebladder followed for at least 20 years. J Urol, 153: 1823, 1995

6. Ooms, E. C., Anderson, W. A., Alons, C. L. et al: Analysis of theperformance of pathologists in the grading of bladder tumors.Hum Pathol, 14: 140, 1983

7. Badalament, R. A., Hermansen, D. K., Kimmel, M. et al: Thesensitivity of bladder wash flow cytometry, bladder wash cy-tology and voided cytology in the detection of bladder carci-noma. Cancer, 60: 1423, 1987

8. Ramakumar, S., Bhuiyan, J., Besse, J. A. et al: Comparison ofscreening methods in the detection of bladder cancer. J Urol,161: 388, 1999

9. Kinders, R., Jones, T., Root, R. et al: Complement factor H or arelated protein is a marker for transitional cell cancer of thebladder. Clin Cancer Res, 4: 2511, 1998

10. Hoffman, M.: The cell’s nucleus shapes up. Science, 259: 1257,1993

11. Soloway, M. S., Briggman, J. V., Carpinito, G. A. et al: Use of anew tumor marker, urinary NMP22, in the detection of occultor rapidly recurring transitional cell carcinoma of the urinarytract following surgical treatment. J Urol, 156: 363, 1996

12. Stampfer, D. S., Carpinito, G. A., Rodriguez-Villanueva, J. et al:Evaluation of NMP22 in the detection of transitional cell car-cinoma of the bladder. J Urol, 159: 394, 1998

13. Landman, J., Chang, Y., Kavaler, E. et al: Sensitivity and spec-ificity of NMP22, telomerase and BTA in the detection ofhuman bladder cancer. Urology, 52: 398, 1998

14. Giannopoulos, A., Manousakas, T., Mitropoulos, D. et al: Com-parative evaluation of the BTAstat test, NMP22 and voidedurine cytology in the detection of primary and recurrent blad-der tumors. Urology, 55: 871, 2000

15. Miyanaga, N., Akaza, H., Ishikawa, S. et al: Clinical evaluationof nuclear matrix protein 22 (NMP22) in urine as a novelmarker for urothelial cancer. Eur Urol, 31: 163, 1997

16. Sanchez-Carbayo, M., Herrero, E., Megias, J. et al: Evaluation ofnuclear matrix protein 22 as a tumour marker in the detectionof transitional cell carcinoma of the bladder. BJU Int, 84: 706,1999

17. Fuchs, E.: Keratins as biochemical markers of epithelial differ-entiation. Trends Genet, 4: 277, 1988

18. Moll, R., Franke, W. W., Schiller, D. L. et al: The catalog ofhuman cytokeratins: patterns of expression in normal epithe-lia, tumors and cultured cells. Cell, 31: 11, 1982

19. Stieber, P., Schmeller, N., Schambeck, C. et al: Clinical rele-vance of CYFRA 21–1, TPA-IRMA and TPA-LIA-mat in uri-nary bladder cancer. Anticancer Res, 16: 3793, 1996

20. Sarosdy, M. F., Hudson, M. A., Ellis, W. J. et al: Improveddetection of recurrent bladder cancer using the Bard BTA stattest. Urology, 50: 349, 1997

21. Sanchez-Carbayo, M., Herrero, E., Megias, J. et al: Initial eval-uation of urinary bladder cancer antigen as a tumor markerfor bladder cancer. J Urol, 161: 1110, 1999

22. Mostofi, F. K., Sobin, L. H. and Torloni, H.: Histological typing ofurinary bladder tumors. International histological classifica-tion of tumours, No. 10. Geneva: World Health Organization,1973

23. American Joint Committee on Cancer: Staging of cancer at gen-itourinary sites. In: Manual for Staging of Cancer, 3rd ed.Edited by O. H. Bears. Philadelphia: J. B. Lippincott, pp.

COMPARISON EVALUATION OF 3 URINE DIAGNOSTIC MARKERS FOR BLADDER CANCER474

194–195, 198824. Sharma, S., Zippe, C. D., Pandrangi, L. et al: Exclusion criteria

enhance the specificity and positive predictive value of NMP22and BTA stat. J Urol, 162: 53, 1999

25. Foresman, W. H. and Messing, E. M.: Bladder cancer: naturalhistory, tumor markers and early detection strategies. SeminSurg Oncol, 13: 299, 1997

26. Leyh, H., Marberger, M., Conort, P. et al: Comparison of the BTAstat test with voided urine cytology and bladder wash cytologyin the diagnosis and monitoring of bladder cancer. Eur Urol,35: 52, 1999

27. Raitanen, M. P., Marttila, T., Kaasinen, E. et al: Sensitivity ofhuman complement factor H related protein (BTA stat) test

and voided urine cytology in the diagnosis of bladder cancer.J Urol, 163: 1689, 2000

28. Sanchez-Carbayo, M., Herrero, E., Megias, J. et al: Comparativesensitivity of urinary CYFRA 21–1, urinary bladder cancerantigen, tissue polypeptide antigen and NMP22 to detect blad-der cancer. J Urol, 162: 1951, 1999

29. Mian, C., Lodde, M., Haitel, A. et al: Comparison of the mono-clonal UBC-ELISA test and the NMP22 ELISA test for thedetection of urothelial cell carcinoma of the bladder. Urology,55: 223, 2000

30. Pode, D., Shapiro, A., Wald, M. et al: Noninvasive detection ofbladder cancer with the BTA stat test. J Urol, 161: 443, 1999

COMPARISON EVALUATION OF 3 URINE DIAGNOSTIC MARKERS FOR BLADDER CANCER 475