Purpose: Non–muscle-invasive bladder cancer is a frequently occurring cancer, with an extremely high recurrence risk. Recurrence detection is based on cytology and urethrocystoscopy. A previous study suggested that a single-nucleotide polymorphism (SNP) array may be effective for noninvasive detection of allelic imbalances in urine. We investigated whether this method is suitable to detect allelic imbalance as an indicator of recurrences in non–muscle-invasive bladder cancer follow-up.

Experimental Design: DNA from blood and urine from 158 patients (113 with and 45 without recurrence) was hybridized to the Affymetrix GeneChip Mapping 10K 2.0. Allelic imbalance detection was based on SNPs showing changes from heterozygosity in blood to homozygosity in urine and on automatic analysis of copy number changes using Copy Number Analyser for GeneChip.

Results: Urine samples with tumor showed allelic imbalance at 0.4% of all informative SNPs. In samples without tumors, 0.04% of these SNPs were affected (P = 0.07). In addition, Copy Number Analyser for GeneChip analysis showed more copy number changes in samples with a tumor (P = 0.001). Losses and gains of chromosomal regions showed clustering, overlapping with known bladder cancer loci. However, 25 (22%) patients with a tumor recurrence did not display any regions with copy number changes, whereas 24 (53%) individuals without a recurrence did. Receiver operating characteristic curve analysis using the number of SNPs displaying copy number changes from the Copy Number Analyser for GeneChip analysis resulted in an area under the curve of only 0.67 (95% confidence interval, 0.58-0.76).

Conclusion: Single-nucleotide polymorphism microarray analysis of allelic imbalance in urine cannot replace urethrocystoscopy and cytology for the detection of recurrences in non–muscle-invasive bladder cancer follow-up.

Translational Relevance

More than 70% of all bladder cancers present as non–muscle-invasive tumors and can be managed conservatively. However, the risk of recurrence is extremely high and 10% to 20% of all tumors show progression to life-threatening muscle invasion. For early diagnosis of these recurrences, periodic urethrocystoscopies are necessary. Whereas many urinary tumor markers have been evaluated as possible alternatives for frequent cystoscopies, none seemed to have an adequate negative predictive value. Previously, it was shown that a single-nucleotide polymorphism array for loss of heterozygosity detection in urine can discriminate between patients with bladder cancer and normal controls. In this study, we investigated whether this method can be used in the follow-up of patients treated for non–muscle-invasive bladder cancer. We show that the single-nucleotide polymorphism microarray, which is now widely available, lacks sufficient accuracy for the detection of tumor recurrences.

Bladder cancer is one of the most frequently occurring malignancies in western countries (1). More than 95% of all bladder cancers are urothelial cell carcinomas, and in ∼70% to 80% of urothelial cell carcinoma patients, this disease is confined to the mucosa (stage pTa and pTis) or submucosa (stage pT1; ref. 2). Based on clinical and pathology characteristics [i.e., number of tumors at diagnosis, tumor diameter, previous recurrence rate, stage, grade, and presence of concomitant carcinoma in situ (CIS)], patients can be classified into different prognostic groups. The best prognostic group has a 5-year risk of recurrences of >30%, whereas the worst prognostic group can have a 5-year risk of recurrence as high as 80%. The 5-year risk of progression to a life-threatening muscle-invasive disease is almost negligible in the best prognostic group but as high as 45% in the worst group (3).

The high risk of (frequent) recurrences and progression to muscle invasion in the worst prognostic group necessitates an active follow-up schedule after initial transurethral resection of the tumor and intravesical chemotherapy or immunotherapy instillations. This follow-up is based on urethrocystoscopy and urine cytology. The European Association of Urology recommends as many as 6 to 15 urethrocystoscopy procedures within 5 years, depending on a patient's risk profile.7

Although considered the gold standard for recurrence diagnosis, urethrocystoscopy may miss flat lesions, which are relatively aggressive (4). Urethrocystoscopy is also an expensive, invasive, and uncomfortable procedure. Cytology has fairly good sensitivity for high-grade lesions but a very poor sensitivity for low-grade tumors (5, 6). In addition, the intraobserver and interobserver reproducibility of cytology is known to be poor (7).

Therefore, there is a clear need for noninvasive diagnostic markers that may replace urethrocystoscopy. These markers should be highly sensitivity, which gives a high negative predictive value so that, in the event of a negative marker, urethrocystoscopy may be avoided. Many urinary tumor markers have been investigated for their ability to detect primary bladder tumors and recurrences, but thus far, none of these markers was found to be highly sensitive and reasonably specific (8, 9). Almost all of these markers were based on the detection of specific proteins, antigens, or degradation products, but because even low-risk bladder tumors show genomic instability (10, 11), a different approach to detect bladder tumors may be to identify aberrant nucleic acids in urine.

One way to assess such chromosomal imbalances, without focusing on specific chromosomal regions, is with whole genome single-nucleotide polymorphism (SNP) arrays. Studies using these arrays on tumor material suggest that allelic imbalances can be detected in all bladder tumors (1214). Thus far, only one study has used a microarray harboring 1,494 SNPs (distributed over the whole genome) to detect allelic imbalance in voided urine of patients with a bladder tumor (15). This study showed almost perfect diagnostic efficacy: all 31 tumor patients had ≥24 SNPs showing DNA alterations at the time of urine collection. The urine sample from one control with hematuria showed 10 SNPs with alterations. The other 13, including 4 patients with hematuria, did not show allelic imbalance at any marker. This study suggested that a SNP microarray can determine the presence of a bladder tumor more accurately than urethrocystoscopy or cytology. With a cutoff value between 10 and 24 SNPs showing loss of heterozygosity (LOH), sensitivity and specificity of 100% would be reached. In a diagnostic setting where bladder cancer patients usually present with microscopic or macroscopic hematuria, a tumor marker may not easily replace urethrocystoscopy. We anticipated that the noninvasive SNP microarray may be more useful in the follow-up of patients treated for non–muscle-invasive bladder cancer, although the striking diagnostic efficacy of the array may not hold for this indication. We therefore investigated the accuracy of a 10K SNP microarray for the detection of recurrences in patients under follow-up for non–muscle-invasive bladder cancer. A blind evaluation of the presence of allelic imbalance in urine material was done in 113 patients with and 45 patients without a bladder tumor recurrence.

Patient collection. Patients in follow-up after treatment for non–muscle-invasive bladder cancer were recruited. Despite the high risk of recurrence among these patients, recurrence is detected in only ∼10% of all urethrocystoscopy procedures. Because informed consent had to be obtained before the urethrocystoscopy, it was inefficient to only recruit consecutive patients undergoing urethrocystoscopy because this would have led to many participants without a recurrence (∼90%). A valid assessment of sensitivity (among ∼10% of patients with a recurrence) was considered most important. Therefore, to try and circumvent these issues, we oversampled patients with a recurrence by also recruiting patients who were scheduled to undergo a transurethral resection of a bladder tumor. All but four recurrences diagnosed during urethrocystoscopy were verified by histopathology following transurethral resection of the tumor. The remaining four recurrences were considered very low risk by the urologist and were left in situ. After written informed consent, a 10 mL EDTA blood sample and a 50 mL voided urine sample were collected from 193 patients undergoing urethrocystoscopy or transurethral resection of the tumor at any of the three selected hospitals in the Netherlands: the Radboud University Nijmegen Medical Centre in Nijmegen, the Canisius Wilhelmina Hospital in Nijmegen, or the Twente Hospital Group in Hengelo. Blood and urine samples were taken at the same time on the day of urethrocystoscopy or 1 d before transurethral resection of the tumor. Samples were stored at room temperature, and urine samples were immediately fixed in Carbowax A (50% ethanol and 2% polyethyleneglycol). Twice a week, samples were transported to the Department of Human Genetics, Radboud University Nijmegen Medical Centre, for processing and analysis. Detailed patient and tumor characteristics are depicted in Table 1 and Supplementary Table S1.

Table 1.

Sample characteristics

Collected, nHybridized, nReason for exclusion
Low DNA amount extracted from urine (n)No PCR product (n)Low DNA amount after purification (n)
No tumor/benign      
    No tumor 41 41    
    Papilloma   
    Atypic cells    
    Squamous metaplasia    
    Total 46 45    
Low-risk tumor      
    pTa    
    pTaG1 29 24  
    pTaG2 14 12   
    pTaG2a 18 11  
    No histology    
    Total 66 52    
High-risk tumors      
    pTaG2b 16 11 
    pTaG2 and CIS    
    pTaG3   
    pTaG3 and CIS    
    pT1G2   
    pT1G3  
    pT1G3 + CIS   
    pT2G3 17 16   
    pT2G3 + CIS    
    CIS  
    CIS in ureter    
    pT1 adenocarcinoma    
    Total 75 61    
Total 187 158 15 13 
Collected, nHybridized, nReason for exclusion
Low DNA amount extracted from urine (n)No PCR product (n)Low DNA amount after purification (n)
No tumor/benign      
    No tumor 41 41    
    Papilloma   
    Atypic cells    
    Squamous metaplasia    
    Total 46 45    
Low-risk tumor      
    pTa    
    pTaG1 29 24  
    pTaG2 14 12   
    pTaG2a 18 11  
    No histology    
    Total 66 52    
High-risk tumors      
    pTaG2b 16 11 
    pTaG2 and CIS    
    pTaG3   
    pTaG3 and CIS    
    pT1G2   
    pT1G3  
    pT1G3 + CIS   
    pT2G3 17 16   
    pT2G3 + CIS    
    CIS  
    CIS in ureter    
    pT1 adenocarcinoma    
    Total 75 61    
Total 187 158 15 13 

Approval for the study was obtained from the institutional review boards of the three participating centers.

Sample preparation. Genomic DNA was extracted from leukocytes in peripheral venous blood as previously described (16), followed by a DNA purification step using the QIAamp DNA Mini Kit (Qiagen). Urine was centrifuged for 20 min at 1,400 × g to collect cells. The obtained cell pellet was used for DNA extraction with the QIAamp DNA Mini Kit (Qiagen). DNA concentration and quality were examined using a Nanodrop spectrophotometer (Isogen). DNA was also examined for possible degradation using agarose gel electrophoresis.

Microarray analysis. To detect allelic imbalance, paired hybridization of DNA from urine and blood of the same patient was done on Affymetrix GeneChip Mapping 10K 2.0 Arrays harboring 10,204 SNPs. Sample preparation and hybridization were done according to the manufacturer's protocol (Affymetrix, Inc.; ref. 17). Briefly, 250 ng of DNA were digested using 10 units of XbaI (Westburg) followed by the ligation of universal adaptors to the digested products using T4 DNA ligase (Westburg). Primers complementary to the adaptors were used for the amplification of the digested products. After purification of the products using MinElute plates (Qiagen), a fragmentation was done on 20 μg of purified PCR product. Fragmented products were end labeled and hybridized to the GeneChip overnight. Probe arrays were washed and stained using the Fluidics Station 400 (Affymetrix). The GeneChip Scanner 3000 was used to scan hybridized arrays. Automatic SNP calls were generated using GeneChip DNA analysis software (Affymetrix). SNP calls were AA or BB for homozygous SNPs and AB for heterozygous SNPs. In cases where the software was unable to make a call (AA, BB, or AB), the SNP was scored as NoCall. An overview of the SNP call rates in blood and urine is presented in Supplementary Table S1.

Data analysis. LOH, defined as a SNP score changing from a heterozygous call (AB) in blood to a homozygous call (AA or BB) in urine, was detected by comparing the genotypes from blood and urine of the same patient. This change can be due to the loss of one allele, amplification of an allele, or a genotyping error. However, earlier analyses of duplicate samples with this Affymetrix array showed that the frequency of genotyping errors is negligible. We also defined SNP scores changing from heterozygous calls in blood to NoCall in urine as indicative of LOH because the loss or duplication of an allele is generally not seen in all cancer cells and urine of patients with a recurrence can also contain normal cells. SNPs with a call rate of <90% in blood samples (n = 709) were excluded from further analysis because these SNPs may be difficult to call using this array and might result in SNPs erroneously defined as indicative of LOH.

In addition to the LOH analysis, allelic imbalance was also analyzed using Copy Number Analyser for GeneChip version 2 by comparing blood and urine obtained from the same patient (18). For the copy number analysis, this software automatically calculates the number of alleles present using a hidden Markov model (18). Only regions with five or more consecutive affected SNPs were considered as gained or lost (19, 20).

Mann-Whitney U tests were used to assess the significance of the differences between the numbers of SNPs displaying LOH and between the numbers of SNPs in regions (more than five SNPs) displaying gains and losses in the patients with and without a tumor recurrence. The ability of the SNPs displaying LOH, as well as the number of SNP regions showing gains and losses in the Copy Number Analyser for GeneChip analysis to discriminate between patients with and without a tumor recurrence, was evaluated by receiver operating characteristic curves. The area under the receiver operating characteristic curves and corresponding 95% confidence intervals (95% CI) were calculated using the statistical software SPSS, version 14.0. We repeated all analyses after stratification of recurrences in low- and high-risk tumors. Low risk was defined as pTa grade 1, 2, or 2a. High risk was defined as pTa grade 2b or 3, CIS, pT1, or muscle-invasive tumors.

Patient series. To test the diagnostic efficacy of the microarray approach for bladder tumor recurrence detection, 193 patients were recruited. Six patients were excluded due to missing blood or urine samples. Based on urethrocystoscopy or histology, 46 of the remaining 187 patients were judged to have no tumor recurrence at the time of sample collection. Detailed information about tumor stage and grade is given in Table 1 (additional patient characteristics are provided in Supplementary Table S1). For 14 patients with and 1 without a tumor recurrence, we were unable to extract sufficient DNA (>250 ng) from the cells present in urine for array hybridization. Fourteen additional patients with a tumor recurrence were excluded because of insufficient (purified) PCR product from DNA extracted from urine (Table 1). These 29 samples were processed thrice to exclude technical error and further examination showed that all 29 excluded samples were of low DNA quality based on the 260:230 ratio, but the DNA itself was not degraded (data not shown).

LOH analysis.Table 2 gives an overview of the mean changes in genotypes comparing blood and matching urine samples from the same patient. A mean of one SNP showed a genotyping error (conversion from a homozygous genotype in blood to a heterozygous call in urine), confirming that the influence of this type of error on the results of the study is negligible. LOH in the form of a change from a heterozygous call in blood to a homozygous call in urine was observed for a mean of 39 SNPs (0.4%) in patients with tumor recurrence versus 4 SNPs (0.043%) in patients without recurrence. This difference was borderline significant (Mann-Whitney U test, P = 0.07). Furthermore, we found that the frequency of SNPs changing from heterozygous call to NoCall was increased in urine samples of patients with a tumor recurrence (1.3% versus 0.9%). The number of NoCalls was higher in regions where changes from heterozygous to homozygous calls occurred and might therefore indicate LOH. This phenomenon was also previously observed in an earlier study comparing blood and bladder tumor material (12), and may be attributed to contamination of tumor DNA present in urine with DNA from normal urothelial cells from the bladder wall or normal cells present in a tumor.

Table 2.

Mean numbers of heterozygous and homozygous calls in blood and urine in the groups of patients with (A) and without (B) a recurrence

(A) patients with a tumor (n = 113)
UrineBlood
Heterozygous SNPsHomozygous SNPsNo Call
Heterozygous SNPs 3,010 (32.3) 1 (0) 39 (0.4) 
Homozygous SNPs 39 (0.4) 6,034 (64.7) 24 (0.3) 
No Call 122 (1.3) 33 (0.4) 19 (0.2) 
    
(B) patients without a tumor (n = 45)
 
   
Urine Blood
 
  

 
Heterozygous SNPs
 
Homozygous SNPs
 
No Call
 
Heterozygous SNPs 3,097 (33.2) 1 (0) 44 (0.5) 
Homozygous SNPs 4 (0.043) 6,007 (64.4) 30 (0.3) 
No Call 87 (0.9) 31 (0.3) 24 (0.3) 
(A) patients with a tumor (n = 113)
UrineBlood
Heterozygous SNPsHomozygous SNPsNo Call
Heterozygous SNPs 3,010 (32.3) 1 (0) 39 (0.4) 
Homozygous SNPs 39 (0.4) 6,034 (64.7) 24 (0.3) 
No Call 122 (1.3) 33 (0.4) 19 (0.2) 
    
(B) patients without a tumor (n = 45)
 
   
Urine Blood
 
  

 
Heterozygous SNPs
 
Homozygous SNPs
 
No Call
 
Heterozygous SNPs 3,097 (33.2) 1 (0) 44 (0.5) 
Homozygous SNPs 4 (0.043) 6,007 (64.4) 30 (0.3) 
No Call 87 (0.9) 31 (0.3) 24 (0.3) 

NOTE: This table provides an overview of the mean numbers (and percentage) of calls in blood samples compared with their corresponding urine samples. As expected, most SNPs show consistent calls in blood and urine (homozygous or heterozygous in blood and urine). The mean number (percentage) of SNPs showing allelic imbalance (heterozygous in blood and homozygous in urine) is depicted in bold.

Copy number analysis. Analysis of the samples using Copy Number Analyser for GeneChip showed many gains and losses in urine from patients with a tumor recurrence. Supplementary Table S2 gives an overview of the regions displaying gains and losses in the individual samples analyzed. It seemed that well-known genomic regions affected in bladder tumors also displayed copy number changes in our patient population (21). Eighteen (16%) patients with a tumor recurrence displayed a loss of a region of at least five SNPs based on the Copy Number Analyser for GeneChip analysis at 9p21; 11 (10%) patients showed a loss in region 8p11-p21; and 9 (8%) patients showed a loss in the region of TP53 on chromosome 17. In the region of the retinoblastoma gene (RB1) on chromosome 13 and in the 10q24-q26 region, we identified loss of one allele in three (3%) and four (4%) patients with a tumor recurrence, respectively. Patients without a tumor recurrence did not show losses in these regions. We also observed amplifications in regions known to be amplified in bladder cancer: 1q12-q25 [33 patients (29%)], 6p [23 patients (20%)], 8q21-q22 [31 patients (27%)], and 17q11-13 [18 patients (16%); ref. 18]. In contrast to the regions that were lost, gains of these regions were also observed in patients without a tumor recurrence: 1q12-q25 [2 patients (4%)], 6p [5 patients (11%)], 8q21-q22 [1 patient (2%)], and 17q11-13 [2 patients (4%)].

Although copy number changes were found in many cases, 22% of the samples of recurrence-positive patients did not display any gains or losses of regions with five or more affected SNPs. By contrast, in a substantial number of samples of patients without a bladder tumor recurrence at the time of sampling (53%), we did observe gains or losses of chromosome parts, although the maximum and mean numbers of SNPs affected were smaller in these samples than in samples of patients with a recurrence. A summary of the number of SNPs displaying gains or losses is depicted in Table 3. A Mann-Whitney U test showed that the difference between the two groups in the mean number of SNPs (in regions of more than five affected SNPs) displaying copy number changes was statistically significant (P = 0.001). Among patients with tumor recurrences, the mean number of SNPs displaying gains or losses was higher in those with high-risk tumors than in patients with low-risk tumors, although this difference was not significant (Mann-Whitney U test, P = 0.147).

Table 3.

Number of SNP regions (i.e., regions of more than five affected SNPs) displaying gains and losses in the Copy Number Analyser for GeneChip analysis

Total gains and losses CNAG analysisNo tumor (n = 45)Tumor total (n = 113)Low-risk tumor (n = 52)High-risk tumor (n = 61)
Mean (SD) 140 (283) 577 (974) 349 (658) 770 (1,149) 
Median (minimum-maximum) 13 (0-1,460) 115 (0-4,504) 84 (0-2,966) 263 (0-4,504) 
Samples without regions of five or more affected SNPs, n (%) 21 (47) 25 (22) 11 (21) 14 (23) 
     
Gains CNAG analysis
 
No tumor (n = 45)
 
Tumor total (n = 113)
 
Low-risk tumor (n = 52)
 
High-risk tumor (n = 61)
 
Mean (SD) 139 (282) 452 (741) 243 (443) 631 (888) 
Median (minimum-maximum) 13 (0-1,460) 93 (0-3,277) 58 (0-2,200) 241 (0-3,277) 
Samples without regions of five or more affected SNPs, n (%) 21 (47) 27 (24) 13 (25) 14 (23) 
     
Losses CNAG analysis
 
No tumor (n = 45)
 
Tumor total (n = 113)
 
Low-risk tumor (n = 52)
 
High-risk tumor (n = 61)
 
Mean (SD) 1 (4) 124 (285) 107 (260) 139 (307) 
Median (range) 0 (0-25) 0 (0-1,395) 0 (0-1,107) 0 (0-1,395) 
Samples without regions of five or more affected SNPs, n (%) 41 (91) 80 (71) 37 (71) 43 (70) 
Total gains and losses CNAG analysisNo tumor (n = 45)Tumor total (n = 113)Low-risk tumor (n = 52)High-risk tumor (n = 61)
Mean (SD) 140 (283) 577 (974) 349 (658) 770 (1,149) 
Median (minimum-maximum) 13 (0-1,460) 115 (0-4,504) 84 (0-2,966) 263 (0-4,504) 
Samples without regions of five or more affected SNPs, n (%) 21 (47) 25 (22) 11 (21) 14 (23) 
     
Gains CNAG analysis
 
No tumor (n = 45)
 
Tumor total (n = 113)
 
Low-risk tumor (n = 52)
 
High-risk tumor (n = 61)
 
Mean (SD) 139 (282) 452 (741) 243 (443) 631 (888) 
Median (minimum-maximum) 13 (0-1,460) 93 (0-3,277) 58 (0-2,200) 241 (0-3,277) 
Samples without regions of five or more affected SNPs, n (%) 21 (47) 27 (24) 13 (25) 14 (23) 
     
Losses CNAG analysis
 
No tumor (n = 45)
 
Tumor total (n = 113)
 
Low-risk tumor (n = 52)
 
High-risk tumor (n = 61)
 
Mean (SD) 1 (4) 124 (285) 107 (260) 139 (307) 
Median (range) 0 (0-25) 0 (0-1,395) 0 (0-1,107) 0 (0-1,395) 
Samples without regions of five or more affected SNPs, n (%) 41 (91) 80 (71) 37 (71) 43 (70) 

Abbreviation: CNAG, Copy Number Analyser for GeneChip.

Receiver operating characteristic curve analysis. Receiver operating characteristic curve analysis using the number of SNPs displaying LOH (not shown) resulted in an area under the curve of 59% (95% CI, 49-68%). Evaluation of the ability to detect high-risk recurrences (area under the curve, 62%; 95% CI, 51-72%) and low-risk recurrences (area under the curve, 54%; 95% CI, 43-66%) showed a slightly better discrimination of recurrence positive from recurrence-negative patients for the high-risk recurrences (not shown). Receiver operating characteristic curve analysis using the number of SNPs displaying gains and losses (i.e., regions of five or more affected SNPs) in the Copy Number Analyser for GeneChip analysis (Fig. 1A) resulted in an area under the curve of 67% (95% CI, 58-76%). The analysis performed slightly better, although still unsatisfactory, for high-risk recurrences (area under the curve, 70%; 95% CI, 60-80%; Fig. 1B) than for low-risk recurrences (area under the curve, 64%; 95% CI, 53-75%; Fig. 1C). This analysis shows the limited ability of the microarray approach to discriminate between recurrence-positive and recurrence-negative patients.

Fig. 1.

Receiver operating characteristic curves. Receiver operating characteristic analysis using the mean number of SNPs (in regions of five or more affected SNPs) displaying gains or losses for all tumor recurrences (A), for high-risk recurrences (B), and for low-risk recurrences (C).

Fig. 1.

Receiver operating characteristic curves. Receiver operating characteristic analysis using the mean number of SNPs (in regions of five or more affected SNPs) displaying gains or losses for all tumor recurrences (A), for high-risk recurrences (B), and for low-risk recurrences (C).

Close modal

To our knowledge, this is the first study analyzing the value of the GeneChip Mapping 10K 2.0 Array for the diagnosis of bladder cancer recurrences in urine using a fairly large patient series. We found that the array was not sufficiently sensitive and specific to replace the standard methods of urethrocystoscopy and cytology in the follow-up of bladder tumors (15).

The suitability of a microarray approach in this follow-up setting critically depends on three assumptions: (a) All bladder cancers are genomically unstable; (b) urine contains a high-enough proportion of cancer cells for allelic imbalance detection by microarray; and (c) in patients without a tumor recurrence at the time of sample collection, no allelic imbalance is observed in cells isolated from urine.

At the start of our project, evidence for the first two assumptions was available from the literature. Many studies showed the presence of allelic imbalance in most bladder tumors using polymorphic markers (either repeat polymorphisms or SNPs) or array comparative genomic hybridization (1215, 2232).

The presence of cancer cells in a high-enough proportion in urine for microarray-based detection was also suggested in an earlier study (15) and was partly confirmed by our own findings. The number of SNPs displaying LOH and copy number changes was significantly increased in material from patients with a tumor compared with patients without a tumor at the moment of sample collection. Furthermore, the number of SNPs displaying gains and losses was higher in high-risk recurrences compared with low-risk recurrences. Although this difference did not reach statistical significance (P = 0.147) in our study, it is in line with earlier findings (21). Gains and losses in our study were present in regions known to be affected in bladder tumors (21). However, we were unable to detect allelic imbalance regions in all urine samples from patients with a tumor recurrence; 22% of the samples showed no regions with gains or losses.

We observed a fairly large number of NoCalls in the urine samples, a phenomenon not observed in a study on voided urine analyzed with an array harboring 1,494 SNPs conducted by Hoque et al. (15). NoCalls might be caused, in part, by the presence of many normal cells detaching from the bladder lining, outnumbering tumor cells in urine samples. Different software algorithms to call genotypes in our study and the study by Hoque et al. (15) may also explain these differences. As in our study, the group of Ørntoft from Aarhus found that the number of SNPs scored as NoCall was significantly higher in bladder tumor samples compared with blood samples (12, 14). They attributed this to the presence of normal cells in tumor samples, which interfered with allelic imbalance detection. Another possible explanation for the occurrence of NoCalls might be the presence of incomplete allelic imbalance, with only a fraction of tumor cells showing a specific loss or gain.

Our third assumption is that no allelic imbalance is found in patients without a bladder cancer recurrence at the time of sampling. Indeed, in the case-control study by Hoque et al. (15), no allelic imbalances were detected in urine samples of nine controls without bladder cancer and allelic imbalance was detected in only one of the five samples from patients with hematuria. We detected regions displaying gains and losses in as much as 53% of the samples without tumor recurrence at the time of sample collection. A major difference with the previous study is that, in our non–muscle-invasive bladder cancer follow-up setting, the nonrecurrence group consisted of patients with a history of bladder cancer whereas the previous study used controls without a history of bladder cancer (15). It is possible that cells from the bladder lining of patients with a non–muscle-invasive bladder cancer history harbor allelic imbalances without being detectable by histology or cytology because several studies have shown that histologically normal urothelium and nonmalignant hyperplasias can contain genomic alterations (33, 34). Furthermore, several studies point to genomic aberrations in normal urothelial cells in patients with a tumor history (3538). In addition, allelic imbalance in the urine of patients without recurrence could be indicative of recurrence that may only be macroscopically diagnosed some months later. To check whether “false-positive” urine samples were actually true positives, we followed up 24 patients with a false-positive test result until the next urethrocystoscopy. Only two of these patients seemed to have a recurrence at the next follow-up visit.

The presence of recurrence-positive patients showing no allelic imbalance and recurrence-negative patients showing allelic imbalance suggests that the accuracy of the microarray method for bladder tumor recurrence detection is inadequate for implementation into clinical practice. Receiver operating characteristic analysis confirmed that the GeneChip Mapping 10K 2.0 Array cannot be used as an alternative for urethrocystoscopy in non–muscle-invasive bladder cancer follow-up. In addition to poor diagnostic efficacy, 16% of all collected specimens could not be analyzed because of low DNA content or quality, making it difficult to implement the array as a robust diagnostic tool.

In conclusion, we show that the GeneChip Mapping 10K 2.0 Array is able to assess allelic imbalance in tumor cells in urine. However, the array is not able to accurately discriminate between patients with tumor recurrence and patients without recurrence with high sensitivity and specificity. Therefore, this microarray approach should not be used for the follow-up of patients with non–muscle-invasive bladder cancer.

No potential conflicts of interest were disclosed.

Grant support: Dutch Cancer Foundation grant KUN 2005-3435.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Note: Supplementary data for this article are available at Clinical Cancer Research Online (http://clincancerres.aacrjournals.org/).

B. Franke and L.A.L.M. Kiemeney shared final responsibility.

We thank Dr. Paula Moonen for her assistance with the collection of the samples and clinical data, and the anonymous reviewers of Clinical Cancer Research for their helpful suggestions to improve the article.

1
Murta-Nascimento C, Schmitz-Drager BJ, Zeegers MP, et al. Epidemiology of urinary bladder cancer: from tumor development to patient's death.
World J Urol
2007
;
25
:
285
–95.
2
Heney NM. Natural history of superficial bladder cancer. Prognostic features and long-term disease course.
Urol Clin North Am
1992
;
19
:
429
–33.
3
Sylvester RJ, van der Meijden AP, Oosterlinck W, et al. Predicting recurrence and progression in individual patients with stage Ta T1 bladder cancer using EORTC risk tables: a combined analysis of 2596 patients from seven EORTC trials.
Eur Urol
2006
;
49
:
466
–5.
4
Schmidbauer J, Witjes F, Schmeller N, et al. Improved detection of urothelial carcinoma in situ with hexaminolevulinate fluorescence cystoscopy.
J Urol
2004
;
171
:
135
–8.
5
Koss LG. Errors and pitfalls in cytology of the lower urinary tract.
Monogr Pathol
1997
;
39
:
60
–74.
6
Brown FM. Urine cytology. It is still the gold standard for screening?
Urol Clin North Am
2000
;
27
:
25
–37.
7
Sherman AB, Koss LG, Adams SE. Interobserver and intraobserver differences in the diagnosis of urothelial cells. Comparison with classification by computer.
Anal Quant Cytol
1984
;
6
:
112
–20.
8
Glas AS, Roos D, Deutekom M, et al. Tumor markers in the diagnosis of primary bladder cancer. A systematic review.
J Urol
2003
;
169
:
1975
–82.
9
Lotan Y, Roehrborn CG. Sensitivity and specificity of commonly available bladder tumor markers versus cytology: results of a comprehensive literature review and meta-analyses.
Urology
2003
;
61
:
109
–18.
10
Orntoft TF, Wolf H. Molecular alterations in bladder cancer.
Urol Res
1998
;
26
:
223
–33.
11
Cordon-Cardo C. Molecular alterations in bladder cancer.
Cancer Surv
1998
;
32
:
115
–31.
12
Koed K, Wiuf C, Christensen LL, et al. High-density single nucleotide polymorphism array defines novel stage and location-dependent allelic imbalances in human bladder tumors.
Cancer Res
2005
;
65
:
34
–45.
13
Hoque MO, Lee CC, Cairns P, Schoenberg M, Sidransky D. Genome-wide genetic characterization of bladder cancer: a comparison of high-density single-nucleotide polymorphism arrays and PCR-based microsatellite analysis.
Cancer Res
2003
;
63
:
2216
–22.
14
Primdahl H, Wikman FP, von der MH, et al. Allelic imbalances in human bladder cancer: genome-wide detection with high-density single-nucleotide polymorphism arrays.
J Natl Cancer Inst
2002
;
94
:
216
–23.
15
Hoque MO, Lee J, Begum S, et al. High-throughput molecular analysis of urine sediment for the detection of bladder cancer by high-density single-nucleotide polymorphism array.
Cancer Res
2003
;
63
:
5723
–6.
16
Miller SA, Dykes DD, Polesky HF. A simple salting out procedure for extracting DNA from human nucleated cells.
Nucleic Acids Res
1988
;
16
:
1215
.
17
Matsuzaki H, Loi H, Dong S, et al. Parallel genotyping of over 10,000 SNPs using a one-primer assay on a high-density oligonucleotide array.
Genome Res
2004
;
14
:
414
–25.
18
Nannya Y, Sanada M, Nakazaki K, et al. A robust algorithm for copy number detection using high-density oligonucleotide single nucleotide polymorphism genotyping arrays.
Cancer Res
2005
;
65
:
6071
–9.
19
George RE, Attiyeh EF, Li S, et al. Genome-wide analysis of neuroblastomas using high-density single nucleotide polymorphism arrays.
PLoS ONE
2007
;
2
:
e255
.
20
Harada T, Chelala C, Bhakta V, et al. Genome-wide DNA copy number analysis in pancreatic cancer using high-density single nucleotide polymorphism arrays.
Oncogene
2008
;
27
:
1951
–60.
21
Reznikoff CA, Sarkar S, Julicher KP, et al. Genetic alterations and biological pathways in human bladder cancer pathogenesis.
Urol Oncol
2000
;
5
:
191
–203.
22
Kimura F, Florl AR, Seifert HH, et al. Destabilization of chromosome 9 in transitional cell carcinoma of the urinary bladder.
Br J Cancer
2001
;
85
:
1887
–93.
23
Veltman JA, Fridlyand J, Pejavar S, et al. Array-based comparative genomic hybridization for genome-wide screening of DNA copy number in bladder tumors.
Cancer Res
2003
;
63
:
2872
–80.
24
Cheng L, MacLennan GT, Zhang S, et al. Laser capture microdissection analysis reveals frequent allelic losses in papillary urothelial neoplasm of low malignant potential of the urinary bladder.
Cancer
2004
;
101
:
183
–8.
25
Migaldi M, Sartori G, Rossi G, et al. Prevalence and prognostic significance of microsatellite alterations in young patients with bladder cancer.
Mod Pathol
2005
;
18
:
1176
–86.
26
Hurst CD, Fiegler H, Carr P, et al. High-resolution analysis of genomic copy number alterations in bladder cancer by microarray-based comparative genomic hybridization.
Oncogene
2004
;
23
:
2250
–63.
27
Blaveri E, Brewer JL, Roydasgupta R, et al. Bladder cancer stage and outcome by array-based comparative genomic hybridization.
Clin Cancer Res
2005
;
11
:
7012
–22.
28
Vaish M, Mandhani A, Mittal RD, Mittal B. Microsatellite instability as prognostic marker in bladder tumors: a clinical significance.
BMC Urol
2005
;
5
:
2
.
29
Obermann EC, Junker K, Stoehr R, et al. Frequent genetic alterations in flat urothelial hyperplasias and concomitant papillary bladder cancer as detected by CGH, LOH, and FISH analyses.
J Pathol
2003
;
199
:
50
–7.
30
Berger AP, Parson W, Stenzl A, et al. Microsatellite alterations in human bladder cancer: detection of tumor cells in urine sediment and tumor tissue.
Eur Urol
2002
;
41
:
532
–9.
31
Szarvas T, Kovalszky I, Bedi K, et al. Deletion analysis of tumor and urinary DNA to detect bladder cancer: urine supernatant versus urine sediment.
Oncol Rep
2007
;
18
:
405
–9.
32
Frigerio S, Padberg BC, Strebel RT, et al. Improved detection of bladder carcinoma cells in voided urine by standardized microsatellite analysis.
Int J Cancer
2007
;
121
:
329
–38.
33
Junker K, Kania K, Fiedler W, et al. Molecular genetic evaluation of fluorescence diagnosis in bladder cancer.
Int J Oncol
2002
;
20
:
647
–53.
34
Hartmann A, Moser K, Kriegmair M, et al. Frequent genetic alterations in simple urothelial hyperplasias of the bladder in patients with papillary urothelial carcinoma.
Am J Pathol
1999
;
154
:
721
–7.
35
Muto S, Horie S, Takahashi S, Tomita K, Kitamura T. Genetic and epigenetic alterations in normal bladder epithelium in patients with metachronous bladder cancer.
Cancer Res
2000
;
60
:
4021
–5.
36
Stoehr R, Zietz S, Burger M, et al. Deletions of chromosomes 9 and 8p in histologically normal urothelium of patients with bladder cancer.
Eur Urol
2005
;
47
:
58
–63.
37
Mazzucchelli R, Barbisan F, Stramazzotti D, et al. Chromosomal abnormalities in macroscopically normal urothelium in patients with bladder pT1 and pT2a urothelial carcinoma: a fluorescence in situ hybridization study and correlation with histologic features.
Anal Quant Cytol Histol
2005
;
27
:
143
–51.
38
Junker K, Boerner D, Schulze W, et al. Analysis of genetic alterations in normal bladder urothelium.
Urology
2003
;
62
:
1134
–8.

Supplementary data