Grossman HB, Messing E, Soloway M, Tomera K, Katz G, Berger Y, Shen Y. Detection of Bladder Cancer Using a Point-of-Care Proteomic Assay. JAMA. 2005;293(7):810-816. doi:10.1001/jama.293.7.810
Author Affiliations: Department of Urology
(Dr Grossman) and Department of Biostatistics and Applied Math (Dr Shen),
M.D. Anderson Cancer Center, Houston, Tex; Department of Urology, University
of Rochester Medical Center, Rochester, NY (Dr Messing); Department of Urology,
University of Miami School of Medicine, Miami, Fla (Dr Soloway); Alaska Clinical
Research Center, Anchorage (Dr Tomera); Department of Surgery-Urology Service,
Lake City Veterans Administration Hospital, and LakeShore Urology, Manitowoc,
Wis (Dr Katz); and Associates in Urology, West Orange, NJ (Dr Berger).
Context A combination of methods is used for diagnosis of bladder cancer because
no single procedure detects all malignancies. Urine tests are frequently part
of an evaluation, but have either been nonspecific for cancer or required
specialized analysis at a laboratory.
Objective To investigate whether a point-of-care proteomic test that measures
the nuclear matrix protein NMP22 in voided urine could enhance detection of
malignancy in patients with risk factors or symptoms of bladder cancer.
Design, Setting, and Patients Twenty-three academic, private practice, and veterans’ facilities
in 10 states prospectively enrolled consecutive patients from September 2001
to May 2002. Participants included 1331 patients at elevated risk for bladder
cancer due to factors such as history of smoking or symptoms including hematuria
and dysuria. Patients at risk for malignancy of the urinary tract provided
a voided urine sample for analysis of NMP22 protein and cytology prior to
Main Outcome Measures The diagnosis of bladder cancer, based on cystoscopy with biopsy, was
accepted as the reference standard. The performance of the NMP22 test was
compared with voided urine cytology as an aid to cancer detection. Testing
for the NMP22 tumor marker was conducted in a blinded manner.
Results Bladder cancer was diagnosed in 79 patients. The NMP22 assay was positive
in 44 of 79 patients with cancer (sensitivity, 55.7%; 95% confidence interval
[CI], 44.1%-66.7%), whereas cytology test results were positive in 12 of 76
patients (sensitivity, 15.8%; 95% CI, 7.6%-24.0%). The specificity of the
NMP22 assay was 85.7% (95% CI, 83.8%-87.6%) compared with 99.2% (95% CI, 98.7%-99.7%)
for cytology. The proteomic marker detected 4 cancers that were not visualized
during initial endoscopy, including 3 that were muscle invasive and 1 carcinoma
Conclusion The noninvasive point-of-care assay for elevated urinary NMP22 protein
can increase the accuracy of cystoscopy, with test results available during
the patient visit.
Bladder cancer is the fifth most common malignancy in the United States.1 Early detection improves prognosis, treatment options,
and quality of life. Although the 5-year survival rate is 95%1 when
tumors are detected while they are confined to the mucosa, up to 25% of the
60 240 bladder tumors predicted to be diagnosed this year will be detected
after they have become invasive or metastatic, which lowers 5-year survival
to approximately 48% and 10%, respectively.1 As
a result, 13 000 people in the United States will die of bladder cancer
this year.1 The incidence of bladder cancer
is higher in men, individuals older than 60 years, and those exposed to carcinogens
in their occupation or environment. Cigarette smoking is the most common risk
factor and doubles the risk of bladder cancer, accounting for approximately
50% of the bladder cancer deaths in men and 30% in women.2
Hematuria and irritative voiding symptoms are the most common symptoms
among patients with urinary tract malignancy. Asymptomatic individuals are
frequently diagnosed after routine or screening analysis by their primary
care physicians has demonstrated hematuria. Hematuria in bladder cancer can
be intermittent, and its degree does not correlate with the severity of underlying
disease. Consequently, it is recommended that patients with hematuria undergo
an evaluation after even a single episode.3
A combination of methods is used to evaluate patients at risk for bladder
cancer because no single procedure is 100% sensitive. Flexible cystoscopy
is an excellent tool because it is low risk and generally can be done in the
physician’s office under local anesthesia.
However, accuracy can be reduced by poor visualization caused by inflammatory
conditions or bleeding, and flat urothelial lesions such as severe dysplasias
and carcinoma in situ may be difficult to distinguish from normal bladder
tissue.4,5 For this reason, voided
urine cytology is frequently used as an adjunctive noninvasive test, but it
is expensive, subjective, and has low sensitivity.
We investigated whether a new, noninvasive urine-based test for the
nuclear matrix protein NMP22 proteomic marker, using monoclonal antibodies
in a point-of-care format, has clinical utility as an aid in diagnosis of
bladder cancer and compared its ability to detect cancer with that of voided
urine cytology, which must be analyzed in a clinical laboratory.
Between September 2001 and May 2002, 22 geographically dispersed clinical
sites, including academic, veterans’, and private practice facilities,
prospectively enrolled 1331 consecutive patients with bladder cancer risk
factors or symptoms, such as smoking, hematuria, or dysuria (Figure). No individuals had a prior history of bladder malignancy.
One additional site recruited 26 patients with active malignancies other than
of the bladder to determine the specificity of the NMP22 test for bladder
cancer. Information about race was obtained for Food and Drug Administration
(FDA) submission purposes from patients by clinical staff at each site. Patients
categorized themselves. Institutional review boards reviewed and approved
the study protocol for each site, and all participants provided written informed
Patients with cancers other than of the bladder provided a urine specimen
for NMP22 protein analysis during a routine visit and did not have endoscopy
or voided cytology evaluations. Each patient evaluated for bladder cancer
provided a voided urine sample before undergoing cystoscopy. One portion of
each sample was sent for routine cytological examination, either within the
institution or at a reference laboratory, according to the standard practice
at each participating facility. An aliquot of the remaining specimen was tested
for the presence of NMP22 protein by a member of the clinic staff. Each device
was identified by study identification number so that the physicians who performed
the subsequent cystoscopy were blinded to the NMP22 test results, and the
staff members who performed the NMP22 assay were blinded to cystoscopy test
results. Technicians who conducted the cytological examinations were physically
distant from both the cystoscopy and NMP22 evaluations, and laboratory reports
arrived after the cystoscopies had been completed and documented.
Staff members at each office performed the NMP22 assay per protocol
by adding 4 drops of voided urine to the sample well of the point-of-care
device. Positive or negative results were read 30 to 50 minutes later in the
test window. A built-in control indicated that the assay was complete. There
were no other procedural steps.
The NMP22 point-of-care device (NMP22 BladderChek Test, Matritech Inc,
Newton, Mass) is a lateral flow immunochromatographic qualitative assay. It
detects elevated amounts of the nuclear mitotic apparatus protein, which is
an abundant component of the nuclear matrix. Nuclear matrix proteins make
up the internal structural framework of the nucleus6,7 and
are associated with such functions as DNA replication and RNA synthesis,8,9 as well as regulation and coordination
of gene expression.10- 12 In
tumor cells, nuclear mitotic apparatus protein, which is present in the interphase
nucleus and associated with the organization of mitotic spindles during cell
division,13 is elevated concordant with structural/morphological
changes characteristic of malignant cell nuclei. Nuclear matrix protein expression
varies with cell type of origin.14,15 In
individuals with bladder cancer, nuclear mitotic apparatus protein is released
into the urine during cell death. Unlike cytological examination, its detection
is not dependent on recovery of intact cells. A microtiter plate immunoassay
was developed for this protein previously,16 but
unlike the new test in this investigation, it required that urine be stabilized
and sent to a laboratory for analysis.
Two different monoclonal antibodies are used in the NMP22 point-of-care
assay, one as a capture antibody, and one as a reporter. To perform the test,
fresh unprocessed urine is added to the sample well of the device and allowed
to react with the colloidal gold–conjugated reporter antibody. If NMP22
protein is present in the urine, it will interact with the reporter conjugate
to form an immune complex. The reaction mixture flows through the membrane,
which contains zones of immobilized antibodies. In the test zone, antigen-conjugate
complexes are trapped by the capture antibody, forming a visible line if the
concentration of NMP22 protein in the urine is greater than 10 U/mL. A procedural
control zone contains an immobilized IgG-specific antibody that will capture
the conjugated antibody independently of the presence or absence of the antigen,
thereby always producing a visible control line in the window to demonstrate
that each device is working properly.
A receiver operating characteristic (ROC) analysis, a plot of the true-positive
rate vs the false-positive rate, is a tool for determination of an optimal
decision point for sensitivity and specificity and requires quantitative data.
An ROC analysis of quantitative data from the microtiter plate format of the
NMP22 assay in an evaluation of patients at high risk for bladder cancer had
an area under the curve (AUC) of 73%.17 The
10-U/mL point of determination for the qualitative point-of-care test for
NMP22 protein corresponds to the cutoff previously approved by the FDA for
the quantitative measurement of NMP22 protein.18
All patients with risk factors or symptoms of bladder cancer underwent
cystoscopy. They were considered positive for malignancy if 1 or more tumors
were observed during initial cystoscopy or within the subsequent 3 months.
Nine patients with no malignancy found during their initial cystoscopy had
a subsequent endoscopy due to continued suspicion, such as increased symptoms.
Removed tumors were defined as malignant based on pathological examination.
Tumors that were seen endoscopically but not removed were considered positive
for malignancy and designated stage (TX) and grade (GX). Reasons that neoplasia
were not removed included concurrent health problems that made patients poor
candidates for surgery and advanced age. Patients were considered negative
for cancer if no tumor(s) was seen endoscopically, or if tissue was biopsied
and defined as nonmalignant on the basis of histopathological examination.
Pathological examination of biopsied tissue was done within each institution
or at a reference laboratory, according to the standard practice at each participating
facility. Staging criteria were those established by the American Joint Committee
on Cancer.19 Imaging information was not collected
for comparison because standard practice at participating facilities varied
widely on frequency and type.
Sample size estimate to determine the performance of the NMP22 test
was based on a 1-sample test for binomial proportions using a 1-sided alternative
and was derived from testing the null hypothesis that the observed proportion
of detection is equal to the expected proportion of detection (newly diagnosed
bladder cancers detected by cystoscopy) vs the alternative hypothesis that
the observed proportion of detection is less than the expected proportion
of detection. This was based on a type I error rate of 5% (ie, α = .05),
80% power, and finding a significant difference of 3% in the detection rate.
Assuming 5% to 10% of patients with hematuria or other risk factors for bladder
cancer could be expected to have a pathologically confirmed positive cystoscopic
evaluation, an estimated sample size of 630 to 1300 patients was required.
Sensitivity of the NMP22 test to detect the presence of bladder cancer
was calculated as the number of patients with true-positive test results (positive
NMP22 test result and tumor) divided by the total number of patients with
malignancy, as detected by endoscopy. Specificity was defined as the percentage
of patients with a negative NMP22 test result who were not diagnosed with
tumors. Corresponding 95% confidence intervals (CIs) were calculated for both
sensitivity and specificity. The sensitivity and specificity of voided cytology
were calculated for comparison. A positive cytology test result was defined
as one in which malignant or dysplastic cells were present.
Statistical analysis was performed at the M.D. Anderson Cancer Center
using S-PLUS version 6.1 (Insightful Corp, Seattle, Wash) and StatXact version
4.0 (Cytel Software Corp, Cambridge, Mass) statistical software.
Demographic and baseline characteristics of the individuals with risk
factors or symptoms of bladder cancer are summarized in Table 1. Among the 1331 patients who had cystoscopies, 79 (6%) had
cancer, 685 (51%) were diagnosed with 1 or more benign urological conditions,
and 567 (43%) had no cystoscopic evidence of urinary tract disease. The mean
age of the patients with bladder tumors was 65.8 years (range, 21-86 years),
and they comprised 3 times as many men as women.
Staging information (Table 2)
was available for the 72 cancers that were surgically removed. The 7 tumors
seen during cystoscopy but not excised were categorized as TX. Of the cancers
with pathological staging data, 62 were superficial (stages Ta, Tis, or T1),
and 10 were muscle invasive (T2-T3). Pathological determination of grade was
available for 70 of the 72 removed tumors (Table
2). Of these, 27 were well differentiated (low grade), 18 were moderately
differentiated (medium grade), and 25 were poorly differentiated (high grade).
A total of 27 cancers were muscle invasive (T2 or T3) and/or poorly differentiated
(high grade). No patients had detectable metastases or involvement of regional
lymph nodes. The NMP22 test results were available for all patients with risk
factors (1331), and cytology test results for 1287 of the patients with risk
factors, including 76 of the 79 diagnosed with cancer.
Initial cystoscopy alone detected 88.6% (70/79) of the cancers. The
remaining 9 malignancies were identified during subsequent cystoscopies conducted
due to continued suspicion, such as increased symptoms, within 3 months of
the initial evaluation. The NMP22 assay was positive in 55.7% (44/79), and
cytology test results of malignant or dysplastic cells were found in 15.8%
The NMP22 test was significantly more sensitive than voided urine cytology
when compared using the McNemar χ2 test (χ2 = 24.7, P<.001). This difference remains significant after taking
into account the inherent variability among the investigational sites using
an adjusted McNemar χ2 test (χ2 = 7.0, P = .008).20 This
significant difference is also reflected by the CIs for the sensitivity proportions
since they do not overlap, at 55.7% (95% CI, 44.1%-66.7%) for the NMP22 test
vs 15.8% (95% CI, 7.6%-24.0%) for cytology. The positive predictive values
of the NMP22 assay and cytology were 19.7% (95% CI, 14.5%-25.0%) and 54.6%
(95% CI, 32.2%-75.4%), respectively.
The same methods were used to compare the specificity proportions and
demonstrated that cytology was significantly more specific than the proteomic
assay (χ2 =149.6, P<.001), at
99.2% (95% CI, 98.7%-99.7%) vs 85.7% (95% CI, 83.8%-87.6%), respectively.
The difference remains significant after taking variability among the sites
into account (adjusted McNemar test χ2 = 9.0, P = .003). The negative predictive values of
the NMP22 assay and cytology were 96.8% (95% CI, 95.6%-97.8%) and 94.9% (95%
CI, 93.6%-96.1%), respectively.
Ten of the 79 malignancies were muscle invasive. Initial cystoscopy
visualized 6 (60%) of these, compared with the NMP22 test, which identified
9 (90%) with elevated protein marker. By comparison, voided cytology was positive
in only 2 (22%) of the 9 patients with muscle-invasive disease for whom test
results were available. The NMP22 assay was also positive for a patient diagnosed
with carcinoma in situ after an initial cystoscopic report of benign disease.
Thus, a total of 4 potentially life-threatening tumors (T2 G2 of the ureter;
T2 G3, Tis G3, and T3 G2 of the bladder) were detected by the NMP22 test but
not visualized in the first cystoscopy. One of the 4 tumors was located in
the ureter and therefore outside the viewing area of the cystoscope. Urine
tests are often added to an evaluation to identify urinary tract tumors such
as this. The combination of the NMP22 test and cystoscopy detected 93.7% of
malignancies vs 88.6% for initial cystoscopy alone (P = .26).
Cytology detected 2 of the 4 cancers not seen in the initial endoscopy, but
which were positive by the NMP22 assay.
Among the most aggressive malignancies, those that were poorly differentiated
(high grade) and/or muscle invasive (stage T2 or T3), the NMP22 test result
was positive in 74% (20/27) compared with cytology, which was positive in
39% (10/26). Of the superficial cancers (Ta, Tis, T1) that were moderately
or well differentiated (medium or low grade), the NMP22 assay identified 47%
(20/43), compared with 5% (2/41) for cytology. Overall, the point-of-care
assay detected 32 malignancies missed by cytology: 11 Ta, 10 T1, 4 T2, 2 T3,
1 Cis, and 4 TX. Voided cytology was positive in only 2 cancer patients for
whom the NMP22 test result was negative, both T1 G3.
The specificity of the NMP22 assay was 90.3% among individuals with
symptoms but with no evidence of urinary tract disease seen during cystoscopy,
and 85.7% overall (Table 3). All risk
patients in the study were undergoing an evaluation for bladder cancer that
included cystoscopy, so false-positive test results did not require any additional
procedures. Cytology demonstrated a specificity of 99.2% among patients with
symptoms and was not performed for individuals with nonbladder cancer. Of
the 38 patients with active cancers other than bladder, the NMP22 assay was
negative in 86.8% (33/38) and positive in 13.2% (5/38).
Prognosis and survival of individuals with bladder cancer are related
to the stage of the malignancy at the time of detection. Approximately 50%
of patients with muscle-invasive disease at first diagnosis demonstrate a
recurrence within 2 years of surgery, despite apparently adequate surgical
resection. The majority of these patients will experience a cancer-related
death within 5 years of diagnosis.21 By comparison,
tumors treated while still confined to the epithelium have lower recurrence
rates and progress to higher stages and grades less often, thereby improving
patients’ long-term outcome.22,23 In
addition, early stage disease can be treated by bladder-sparing therapy rather
than cystectomy, the standard for advanced disease, which impacts quality
of life as well as survival.
The direct cost of treatment for patients with metastatic genitourinary
cancer has been estimated to be more than 6 times greater than for those patients
with localized disease for the same period of time.24 The
challenge, therefore, is to improve detection of bladder cancer without adding
increased risk or discomfort to the patient.
Cystoscopy is integral to the diagnosis of bladder cancer, allowing
the physician to visualize the bladder wall directly. The sensitivity of cystoscopy
is very good, but hematuria and other conditions can obscure lesions, and
flat neoplasia can be confused with erythema. As seen in this study, even
later-stage cancers are sometimes missed during endoscopy. The precise rate
of false-negative cystoscopy test results is difficult to determine, but estimates
range from 10% to 40%.25- 27 In
this study it was 11.4%. For this reason, physicians frequently use multiple
tools to aid in diagnosis of bladder cancer, including urinalyses and imaging
of the upper tract.
Voided cytology has been a widely accepted adjunctive test to cystoscopy
because it is noninvasive. This method involves visual assessment of morphological
changes and therefore requires intact cells. Small tumors, well-differentiated
(low-grade) tumors, or both are less likely to exfoliate cells spontaneously
because the strong intercellular attachments are better preserved, and the
degree of morphological departure from normal is less, making recognition
difficult.28 This results in low sensitivity,
approximately 15% to 30% in early stage cancers.29,30 In
addition, some inflammatory conditions cause cellular changes that can be
confused with neoplastic process, contributing to subjective interpretation.
False-positive reports of malignant cells are rare, but ambiguous reports
of atypical cells are common. Collecting and analyzing 3 serial first-morning
specimens for voided cytology is used by some physicians to improve the detection
rate of cancer,31 but it significantly increases
cost, and patient compliance is difficult. Because specially trained technicians
are required for the analysis, samples must be stabilized and sent to a laboratory,
and test results are not available immediately. The high specificity of cytology
is offset by low sensitivity, ambiguous test results, expense, and time lag
to obtain reports.
We found that the NMP22 test is a useful adjunctive tool in the evaluation
of patients at risk for bladder cancer and that it identified several malignancies
missed by initial cystoscopy. Specificity of the NMP22 test was lower than
for cytology (85.7% vs 99.2%), but sensitivity was significantly greater (55.7%
vs 15.8%), with test results available during the patient visit. The NMP22
protein is the only tumor marker approved by the FDA as an aid in the initial
diagnosis of bladder cancer, and the test has been waived under the Clinical
Laboratory Improvement Act so it can be performed in any physician’s
office. The cost of urine tests varies by location. The average Medicare reimbursement
for voided cytology is approximately $56, compared with $24 for the NMP22
In the general US population, bladder cancer occurs in approximately
39 of 100 000 men and 10 of 100 000 women.1 Patients
enrolled in this investigation were at elevated risk for urological cancer
due to behaviors such as smoking or symptoms including hematuria and dysuria.
Consequently, the incidence of malignancy in this group was 6%. Currently,
20% or more of symptomatic patients present with disease that is already invasive
at the time of first diagnosis. Identifying these malignancies earlier could
improve prognosis and reduce the cost of treatment.
Hematuria is the most common symptom of bladder cancer, but it is often
intermittent.33 Repetitive home testing of
high-risk populations has shown good detection results,34 and
the cost of urinary dipstick testing is minimal. However, because hematuria
is not specific to cancer, it is estimated that only 1 in 20 patients with
hematuria will have bladder cancer.34- 36 Positive
predictive value, the percentage of times that a positive test result corresponds
to the presence of the disease in question, is about 5% for hematuria testing
in men for bladder cancer34,35 and
even lower in women.36 Nevertheless, because
the amount of blood in urine is unrelated to stage and grade of cancer, the
American Urological Association Best Practice Policy Panel on Asymptomatic
Microscopic Hematuria and others have concluded that there is no safe lower
limit for hematuria, and high-risk patients should be considered for a urological
evaluation after even a single episode.3,37 Among
study patients with the highest risk for bladder cancer, men older than 60
years with a history of smoking, the positive predictive value of the NMP22
test was 37%. This is higher than the 20% to 30% predictive value typically
reported for prostate-specific antigen in men who have an elevated risk of
prostate cancer, those with levels between 4 to 10 ng/mL.38- 41
In conclusion, the NMP22 assay may be a useful adjunct to cystoscopy
for diagnosing bladder cancer. Studies in different patient populations are
necessary to further define the role of this assay in patients with risk factors
and symptoms suggestive of possible bladder cancer.
Corresponding Author: H. Barton Grossman,
MD, Department of Urology, M.D. Anderson Cancer Center, 1515 Holcombe Blvd,
Unit 446, Houston, TX 77030 (HBGrossman@mdanderson.org).
Author Contributions: As principal investigator,
Dr Grossman 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: Grossman, Messing,
Acquisition of data: Grossman, Messing, Soloway,
Tomera, Katz, Berger, Shen.
Analysis and interpretation of data: Grossman,
Messing, Soloway, Shen.
Drafting of the manuscript: Grossman, Messing.
Critical revision of the manuscript for important
intellectual content: Grossman, Messing, Soloway, Tomera, Katz, Berger,
Statistical analysis: Grossman, Shen.
Obtained funding: Grossman, Messing, Soloway,
Tomera, Katz, Berger.
Administrative, technical, or material support:
Grossman, Messing, Soloway, Tomera, Katz.
Study supervision: Grossman, Messing, Soloway,
Financial Disclosures: None reported.
Funding/Support: Matritech Inc supplied the
experimental assay to the investigators at no cost and reimbursed clinical
sites for the time involved in collection of data related to FDA submission.
This included risk factors, demographic information, and test results.
Role of the Sponsor: Matritech Inc designed
the study, monitored the conduct and collection of data, and reviewed the
manuscript for factual accuracy and approved it.
Independent Statistical Analysis: Independent
statistical analysis was performed by Yu Shen, PhD, at the University of Texas
M. D. Anderson Cancer Center, Department of Biostatistics and Applied Math.
Acknowledgment: We thank statistician Heather
E. Kelley, BA (West Roxbury, Mass), for her efforts analyzing the study data,
as well as the following physicians who contributed to the performance of
the clinical trial: Anthony Abner, MD (Mt. Auburn Hospital, Cambridge, Mass);
David Bock, MD (Kansas City Urology Care, Kansas City, Mo); Jeffrey Brady,
MD (Winter Park Urology Associates, Orlando, Fla); M. Patrick Collini, MD
(Urology Associates of North Texas, Fort Worth); Martin Dineen, MD (Atlantic
Urological Associates, Daytona Beach, Fla); Vahan Kassabian, MD (Georgia Urology,
Atlanta); Shiva Maralani, MD (St Clair Shores, Mich); Raoul Salup, MD (James
A. Haley Veterans Administration Hospital, Tampa, Fla); Barry Stein, MD (Rhode
Island Hospital, Providence); Alan Treiman, MD (Urology Treatment Center,