SEARCH

SEARCH BY CITATION

Keywords:

  • bladder cancer;
  • urine;
  • chromosomal instability;
  • UroVysion;
  • fluorescence in situ hybridisation

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Subjects and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Conflict of Interest
  9. References

What's known on the subject? and What does the study add?

  • UroVysion™ is a multicolour fluorescence in situ hybridisation assay that detects DNA gain at chromosomes 3, 7 and 17 and loss at the 9p21 locus in exfoliated urothelial cells. This cell-based test is time-consuming and costly compared with voided urine cytology or other molecular markers for the early detection of bladder cancer.
  • We determined copy number changes at chromosomes 3, 7 and 17 and at the 9p21 locus with UroVysion in a prospective screening study among chemical workers. Strong correlations between DNA gains yield a similar performance in detecting bladder cancer with just one of the probes for chromosomes 3, 7 or 17 instead of all, supporting the development of a simpler and cheaper assay.

Objective

  • To explore changes at chromosomes 3, 7, 17 and 9p21 in order to assess associations with bladder cancer for possible improvements of the UroVysion™ assay regarding screening.

Subjects and Methods

  • In all, 1609 men took part in the prospective study UroScreen. Annual screening for bladder cancer was offered to male chemical workers with former exposure to aromatic amines as a voluntary surveillance programme between 2003 and 2010.
  • In all, 191 434 cells in 6517 UroVysion tests were analysed for copy number variations (CNV) at chromosome 3, 7, 17 (gains) and 9p21 (deletions) in 1595 men.
  • We assessed CNVs at single or multiple loci using polysomy indices (PIs, called multiple PI and PI 3, PI 7 and PI 17).
  • We calculated Spearman's rank correlation coefficients (rs) between these PIs and receiver operating characteristic (ROC) curves with areas under the curves (AUCs). We applied Cox regression to estimate hazard ratios (HRs) to assess the risk of developing bladder cancer.

Results

  • Nine out of 21 bladder tumours detected in 20 participants (‘cases’) had a positive UroVysion test, including seven high-grade carcinomas and seven overlapping results with a positive cytology. Four cases with negative test results did not attend screening annually.
  • No case was found because of a complete loss of 9p21 in at least 12 cells.
  • There were strong correlations between pairwise combinations of gains at chromosome 3, 7 or 17, ranging between rs = 0.98 and rs = 0.99 in cases and between rs = 0.84 and rs = 0.88 in non-cases (P < 0.001). Associations were less pronounced with CNVs at 9p21 among cases and were lacking in non-cases.
  • Estimates of the relative risk of DNA gain for developing a bladder tumour assessed with PIs (threshold 10% of cells) were 47.7 (95% confidence interval [CI] 18.3–124.1) for the multiple PI, 44.5 (95%CI 16.5–119.9) for PI 3, 34.7 (95%CI 13.1–92.1) for PI 7 and 52.4 (95%CI 20.7–132.6) for PI 17, as well as 7.9 (95%CI 3.0–20.6) for a complete loss of 9p21 (threshold 2.5% of cells), respectively.
  • ROC analyses showed similar AUCs for multiple PI compared with PIs of single chromosomes 3, 7 and 17 (all AUCs between 0.79 and 0.80) and a lower AUC for a homozygous loss of 9p21 (AUC 0.72).

Conclusions

  • The UroVysion assay showed a reasonable performance in detecting bladder cancer in the present study population and shared positive test results with cytology, which is much cheaper.
  • A simpler, faster and cheaper version of the UroVysion assay might rely on the very strong correlations between gains at chromosomes 3, 7 and 17, resulting in a similar performance in detecting bladder cancer with single-probe PIs compared with the full set of these probes.
  • Loss of 9p21 was less predictive for developing bladder cancer in UroScreen.

Abbreviations
AUC

area under the curve

CIS

carcinoma in situ

CNV

copy number variation

FISH

fluorescence in situ hybridization

HR

hazard ratio

NMP22

nuclear matrix protein 22

NPV

negative predictive value

PI

polysomy index

PPV

positive predictive value

ROC

receiver operating characteristic

rs

Spearman's rank correlation coefficient

UC

urethrocystoscopy

Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Subjects and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Conflict of Interest
  9. References

In patients with bladder cancer the high risk of recurrence makes long-term follow-up a necessity. Early detection of a recurrent cancer is important to prevent progression to muscle-invasive stages. Currently, the standard approach in detecting bladder cancer is a combination of urethrocystoscopy (UC) and urine cytology [1]. UC is an invasive method, whereas cytology is non-invasive but exhibits a lower sensitivity in detecting low-grade urothelial tumours [1-3]. Several non-invasive molecular tumour tests have been developed to improve the sensitivity of cytology or even replace UCs. One promising test, which has been approved by the USA Food and Drug Administration, is a multicolour fluorescence in situ hybridization (FISH) assay, marketed as the UroVysionTM assay (Abbott Laboratories, Abbott Park, IL, USA) [4].

This assay was designed to detect copy number variations (CNVs) of chromosomes 3, 7 and 17, and at the 9p21 locus. CNVs are DNA regions that show gains (e.g. duplications) or losses (deletions) of genetic material. It is assumed that CNVs play an important role in the development of many cancers [5]. UroVysion detects high-grade carcinomas and carcinoma in situ (CIS) with reliable sensitivity, but shows a lower sensitivity for low-grade tumours [6, 7]. Hyperdiploidy may occur in CIS and invasive urothelial bladder carcinomas [8], whereas loss at 9p21 has also been associated with non-invasive papillary carcinomas [8, 9].

The evaluation process of the FISH assay is comparatively time-consuming and restricted to qualified laboratories. Hence, the UroVysion test is more expensive than cytology or other markers, e.g. nuclear matrix protein 22 (NMP22). The assay generates a large set of complex data on CNVs at four loci in different single cells, which can be a valuable source of information on chromosomal instability. Most statistical analyses are based on the final test results as positive or negative, whereas this wealth of genomic information has not been sufficiently explored. An important question is, whether the genomic instability can be assessed with fewer probes, in order to provide evidence for improving the test for bladder cancer screening.

The prospective study UroScreen was aimed at evaluating the performance of tumour markers for early detection of bladder cancer in chemical workers formerly exposed to carcinogenic aromatic amines [10-13]. The results of the performance of the UroVysion test have been published elsewhere [14]. Here, we explored the associations between the CNVs at four chromosomal loci based on data in 191 434 single cells from 6517 urine samples of 1595 men. Our objective was to assess whether the observed CNV changes are correlated and to evaluate the performance of the UroVysion test when reducing the number of probes. This allows suggestions for a future development of a simpler, faster and finally cheaper version of the UroVysion test.

Subjects and Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Subjects and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Conflict of Interest
  9. References

Study Population and Data Collection

UroScreen was a prospective screening study in chemical workers aimed at assessing the predictive values of urinary tumour markers in detecting bladder cancer. The design and implementation of the study was described previously [10, 11, 13]. In brief, participants were recruited from an established annual surveillance programme of the statutory accident insurance of the chemical industry. The study population comprised male active and retired workers from two chemical plants who had been formerly exposed to carcinogenic aromatic amines. Altogether, 2214 men were invited to participate in annual screens for the early detection of bladder cancer between September 2003 and June 2010. In all, 1609 men took part at least once within the study period. The basic surveillance programme included urine status and cytology. UroScreen offered the additional determination of UroVysion and NMP22 on a voluntary basis. Company physicians administered a questionnaire to collect information on smoking status and history of cancer or other relevant diseases. The ethics committee of the University of Tübingen, Germany, approved the study (No. 1/2003V). All participants gave written informed consent.

This analysis was based on data of 1595 men with full information on CNVs in 6517 urine samples. There were 20 cases of bladder cancer among these participants during the study period. Reference pathology was applied according to 2004 WHO classification of bladder tumours [15].

Urine Collection and Determination of Cytology and UroVysion

Urine collection, processing and the determination of the tumour tests have been previously described [10, 13, 14]. In brief, ≈100 mL spot urine was obtained and preserved. Standard urine cytology was performed in urine sediments according to the method by Papanicolaou and Marshall [16].

Chromosomal instability in urothelial cells was assessed at four loci using the UroVysion Bladder Cancer Kit (Abbott Laboratories, Abbott Park, IL, USA) according to the manufacturer's protocol. DNA was denatured for 3 min at 76 °C followed by a hybridization step with a mixture of fluorochrome-marked DNA probes (CEP 3, CEP 7, CEP 17 and LSI 9p21) at 39 °C for 20 h using a HYBriteTM or ThermoBriteTM (Abbott Molecular, Des Plaines, IL, USA). For each sample, about 25 morphologically suspicious cells were evaluated. For this analysis, the CNVs in every analysed cell were documented for all subjects. In total, CNVs at four loci in 191 987 cells from 6517 urine samples from 1595 men were available.

The UroVysion test was considered to be positive when following the decision rule of the manufacturer. A positive test result was given if any of the following criteria was met: if ≥4 of 25 cells showed gains in ≥2 of the chromosomes (3, 7, 17) in the same cell, or if ≥12 of 25 cells had zero 9p21 signals. Otherwise, the analysis was continued until either 4 cells with gain for multiple chromosomes were detected, or 12 cells with zero 9p21 signals were detected, or the entire sample was analysed.

Variables to Assess the Chromosomal Instability

Different indices were computed to define the chromosomal aberration referring to the approach of Bollmann et al. [2]. The frequency of cells with CNVs of >2 (polysomy) in chromosome 3 was defined as polysomy index 3 (PI 3). Analogously, PIs were calculated for chromosomes 7 and 17 (PI 7 and PI 17). In addition, CNVs of >2 of at least two out of these chromosomes was defined as multiple PI. Furthermore, the frequency of cells with a homozygous deletion of 9p21 was computed. All indices were calculated to measure the degree of chromosomal instability.

Statistical Analysis

Time-dependent Cox regression analyses were used to calculate hazard ratios (HRs) and associated 95% confidence intervals (CIs), as risk estimates for developing a bladder tumour during the study period based on UroVysion and the derived indices. Six models were constructed separately for different indices that measure chromosomal instability in order to avoid dependent predictors. Cox proportional hazards models were adjusted for a history of bladder cancer, smoking status and age.

Pairwise Spearman's rank correlation coefficient (rs) and Spearman's tests were computed between the indices (PI 3, 7 and 17, and the frequency of a homozygous deletion at 9p21) to evaluate whether there was a correlation in the increase of aberrations for samples from cases and samples from participants who did not develop a bladder tumour during the study period (‘non-cases’). Sensitivities, specificities, positive predictive values (PPVs) and negative predictive values (NPVs) were calculated for all bladder tumours as well as for the subgroups ‘papilloma’, ‘high-grade’ and ‘low-grade cancer’. In addition, receiver operating characteristic (ROC) curves were constructed, and the areas under the curves (AUCs) were determined to compare the different indices. For non-cases analyses were based on the last analysable test result during the study period. For cases, the last analysable test result before diagnosis was used if the evaluation was performed ≤12 months before diagnosis. This criterion was met by only 16 of the 20 cases.

All statistical tests were performed at the two-tailed α = 0.05 level of statistical significance. Calculations were performed with SAS/STAT and SAS/IML software, version 9.2 (SAS Institute Inc., Cary, NC, USA).

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Subjects and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Conflict of Interest
  9. References

Of 7091 urine samples, 566 samples (8%) could not be analysed mainly due to insufficient cells in the spot urines. There were 107 (1.5%) positive and 6418 (90.5%) negative UroVysion tests. A valid UroVysion result (positive or negative), as well as complete CNV information for the chromosomes 3, 7, 17 and 9p21 was available from 6517 of 6525 screening evaluations. Therefore, the analysis was based on CNV data in 191 434 cells from 1595 active or retired chemical workers.

Table 1 shows the characteristics of 1595 men with UroVysion data. The median age was 62 years in 2010, whereas for cases with bladder tumours the median age was 67 years at the time of diagnosis. More cases had or still did smoke than non-cases (75% vs 65.5%).

Table 1. Characteristics of the UroScreen cohort of male active or retired workers with former exposure to aromatic amines.
 TotalCasesNon-cases
  1. a

    One case with two bladder tumours during UroScreen.

N (%)1609 (100)20a (1.2)1589 (98.8)
Study participants with at least one analysable UroVysion test, n (%)1595 (99.1)20a (100)1575 (99.1)
Median (range) age, years   
Year 201062 (27–90)68 (43–80)62 (27–90)
Age at diagnosis 67 (38–78) 
Smoking status at baseline, n (%)   
Unknown3 (0.2)3 (0.2)
Never548 (34.4)5 (25.0)543 (34.5)
Former614 (38.5)12 (60.0)602 (38.2)
Current430 (27.0)3 (15.0)427 (27.1)
Former bladder cancer at baseline, n (%)21 (1.3)3 (15.0)18 (1.1)

Table 2 describes the cases with test results before diagnosis. UroVysion results were available for 21 tumours in 20 cases. Nine UroVysion tests were positive (seven high-grade and two low-grade tumours). Overall, 11 tests were negative, especially in cases with papilloma. For four out of six negative high-grade cases, the time between urine analysis and cancer diagnosis was >12 months. The negatively tested case #8 had a positive test result in a previous screening round, and case #15 showed chromosomal changes that did not pass the threshold value for positivity. No case was detected because of a complete loss of 9p21 (a loss in ≥12 cells would have been required for a positive diagnosis). Eight cases had a positive cytology. A positive result in both, cytology and UroVysion, was found in seven cases.

Table 2. Characterisation of the cases by histology, screening compliance, cytology and UroVysion results.
CaseTNM stageGradeAll UroVysion tests before diagnosis Npos/Nneg/Nnot analysableLast investigation before diagnosis:
Months before diagnosisCreatinine (g/L)CytologyUroVysionMultiple PI, %PI 3, %PI 7, %PI 17, %Homozygous deletion of 9p21, %
  1. *Case #6 had another positive UroVysion test 1 month before diagnosis but no data on CNVs. Case #16 had another not analysable UroVysion test 2 months before diagnosis. N.D., not determined.

8pT3bHigh1/3/001.53NegativeNegative00000
13pT2High1/1/11N.D.NegativePositive282828248
9pT1High1/1/020.32PositivePositive221922190
15pT1aHigh0/4/020.72SuspiciousNegative55533
19pT1High3/1/03N.D.PositivePositive525252520
20pT3aHigh3/3/060.44SuspiciousPositive2323232018
1pTa-1, CISHigh1/0/071.28PositivePositive382346464
12CISHigh1/0/07N.D.PositivePositive6060606050
6*CISHigh2/0/0110.24PositivePositive16168164
7pT1High0/3/0142.92NegativeNegative00000
10pTa, pNX, pMXHigh0/3/0182.93NegativeNegative00000
4bpTaHigh0/1/0240.87NegativeNegative00000
5pT1, CISHigh0/1/0261.01NegativeNegative00000
3pTaLow0/1/021.59NegativeNegative00000
11pTaLow1/0/020.44PositivePositive524852444
18pTaLow1/8/121.06PositivePositive101015103
14pTaLow0/5/030.74NegativeNegative00000
2pTaLow0/2/0100.23NegativeNegative00000
17 Papilloma0/6/020.77PositiveNegative33330
4a Papilloma0/2/0100.35NegativeNegative00000
16 Papilloma0/3/113N.D.NegativeNegative00000

There were strong pairwise positive correlations between PI 3, PI 7 and PI 17 in samples from cases (Fig. 1), as well as in samples from non-cases (Fig. 2). For example, Spearman's correlation coefficients ranged between 0.98 and 0.99 between PI 3, 7 and PI 17 in cases and between 0.84 and 0.88 in non-cases. The correlations between a homozygous deletion at 9p21 with PI 3, PI 7 and PI 17 were weaker among cases (rs 0.62–0.69). There were no correlations of chromosomal changes in non-cases (rs 0.04–0.07).

figure

Figure 1. Correlations between the indices built from UroVysion data in those cases with screening ≤ 12 months before diagnosis (16 of 20 cases).

Download figure to PowerPoint

figure

Figure 2. Correlations between the indices based on the last analysable test result of 1575 non-cases.

Download figure to PowerPoint

Fig. 3 compares the ROC analyses for the PIs assessed for the single chromosomes 3, 7 and 17, the multiple PI and the frequency of a homozygous deletion at 9p21. The performance of all PIs was similar (AUCs of 0.79 and 0.80), whereas the homozygous deletion at 9p21 had a lower AUC of 0.72. We derived a new threshold of 10% for the fraction of cells with gain for application to all PIs, which resulted in the detection of nine cases with multiple PI, PI 3 and PI 17 (Table 3). These nine cases were also found applying the manufacturer's decision algorithm to the full set of data from four probes. These PIs and the UroVysion results according to the manufacturer's algorithm had a sensitivity of 56.3% for all lesions that increased to 77.8% for high-grade tumours. A homozygous deletion at 9p21 in at least 2.5% of cells reached a sensitivity of 50%. Specificities ranged between 94.6% and 97.5%. In contrast, the manufacturer's decision rule (homozygous loss in ≥12 cells) would not detect any of the cases and has thus resulted in a sensitivity of 0% and a specificity of 100%. Due to the few cases the PPVs were overall low, and NPVs were high (Table 3). The sensitivity for low-grade tumours was lower, but based on only five cases. The two papillomas could not be detected with conventional UroVysion or any of the indices.

figure

Figure 3. ROC analyses for the indices based on the last analysable test result of 16 cases and 1575 non-cases.

Download figure to PowerPoint

Table 3. Cancer-predictive values for UroVysion and PIs based of the last analysable test result from those cases who attended the screening during ≤ 12 months before diagnosis (16 of 20 cases) and 1575 non-cases.
 True positiveFalse negativeFalse positiveTrue negativeSensitivity, %Specificity, %PPV, %NPV, %
UroVysion:        
All lesions9747152856.397.016.199.5
High-grade7247152877.897.013.099.9
Low-grade2347152840.097.04.199.8
Papilloma024715280.097.00.099.9
Multiple PI (threshold 10%):        
All lesions9747152856.397.016.199.5
High-grade7247152877.897.013.099.9
Low-grade2347152840.097.04.199.8
Papilloma024715280.097.00.099.9
PI 3 (threshold 10%):        
All lesions9747152856.397.016.199.5
High-grade7247152877.897.013.099.9
Low-grade2347152840.097.04.199.8
Papilloma024715280.097.00.099.9
PI 7 (threshold 10%):        
All lesions8850152550.096.813.899.5
High-grade6350152566.796.810.799.8
Low-grade2350152540.096.83.899.8
Papilloma025015250.096.80.099.9
PI 17 (threshold 10%):        
All lesions9740153556.397.518.499.5
High-grade7240153577.897.514.999.9
Low-grade2340153540.097.54.899.8
Papilloma024015350.097.50.099.9
Homozygous deletion of 9p21 (threshold 2.5%):        
All lesions8885149050.094.68.699.5
High-grade6385149066.794.66.699.8
Low-grade2385149040.094.62.399.8
Papilloma028514900.094.60.099.9

Table 4 shows the HRs of developing a bladder tumour for the different PIs with a common threshold of 10% gain of the chromosomes 3, 7 or 17, a homozygous deletion at 9p21 in at least 2.5% of cells and for the UroVysion test (decision rule of the manufacturer) estimated in separate models and adjusted for several factors. The highest HR was estimated for UroVysion with 60.4 (95%CI 22.8–159.6), followed by PI 17 with a HR of 52.4 (95%CI 20.7–132.6). The lowest HR was seen for a homozygous loss of 9p21 in at least 2.5% of cells (HR 7.9, 95%CI 3.0–20.6). Furthermore, there was a higher bladder tumour risk for participants with a former bladder tumour and participants aged ≥ 60 years.

Table 4. HRs from Cox regression models with 95% CIs of various indices at all time-points and other factors at baseline for developing bladder cancer (20 tumours) in 1595 men.
 Model 1Model 2Model 3Model 4Model 5Model 6
HR (95%CI)HR (95%CI)HR (95%CI)HR (95%CI)HR (95%CI)HR (95%CI)
UroVysion:      
Negative1     
Positive60.4 (22.8–159.6)     
Multiple PI:      
<10% 1    
≥10% 47.7 (18.3–124.1)    
PI 3:      
<10%      
≥10%  1   
PI 7:  44.5 (16.5–119.9)   
<10%      
≥10%   1  
PI 17:   34.7 (13.1–92.1)  
<10%      
≥10%    1 
Homozygous deletion of 9p21:    52.4 (20.7–132.6) 
<2.5%     1
≥2.5%     7.9 (3.0–20.6)
Age at baseline:      
<60 years111111
≥60 years2.6 (1.0–7.0)2.6 (1.0–6.9)2.9 (1.1–7.8)2.7 (1.1–7.0)2.2 (0.9–5.8)2.0 (0.8–5.1)
Ever a smoker:      
No111111
Yes1.1 (0.4–3.3)1.2 (0.4–3.3)1.2 (0.4–3.5)1.1 (0.4–3.2)1.2 (0.4–3.4)1.1 (0.4–3.1)
Former bladder cancer:      
No111111
Yes7.1 (1.7–28.7)9.7 (2.5–37.6)7.3 (1.8–28.6)11.2 (2.9–42.6)12.6 (3.2–49.9)11.9 (3.2–44.6)

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Subjects and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Conflict of Interest
  9. References

The International Consensus Panel on Bladder Tumour Markers recommended UroVysion as a promising test for detecting bladder cancer [17]. We found a reasonable performance of UroVysion in the UroScreen study that was similar to cytology [14]. This FISH assay detected nine out of 21 bladder tumours in 20 participants cancer-free at baseline, with preference for high-grade carcinomas in the present study population of male chemical workers. The UroVysion test was overlapping with a positive cytology probably due to the pre-selection of morphologically abnormal cells. However, the assay is labour-intensive and expensive. So far, there are no generally accepted criteria for interpreting the complex data generated by the UroVysion test [17-19]. We calculated PIs for each locus referring to Bollmann et al. [2] to explore the associations between the chromosomal changes at the four loci in a large number of nearly 200 000 cells. There were strong correlations between the gains at chromosomes 3, 7 and 17, and a good performance of the single PIs to detect bladder cancer. This offers the opportunity of developing a simpler and thus cheaper FISH assay. In contrast, no case was detected because of a complete loss of 9p21 in a sufficient number of cells according to the manufacturer's decision rule. Loss at 9p21 was less clearly associated with the observed gains at the other loci.

A main advantage of the UroScreen study is the prospective design and the large number of ≈6500 urine samples with documented CNVs. However, there were only a few cases of bladder cancer. This is a general limitation of screening for bladder cancer in asymptomatic subjects due to the low incidence [20]. The few cases of bladder cancer is also reflected in low PPVs for UroVysion as well as for the various PIs. Another important question is if the strong correlations between the gains of these three loci seen in a few cases may be a consistent observation. This seems to be supported by similar although weaker correlations in the large group of cancer-free subjects. Because we found a large fraction of high-grade tumours in this group of chemical workers, it would be important to validate the present results in other study populations.

Previous findings indicated that the combination of all four probes may have a higher sensitivity than each single probe alone [21, 22]. These results were based on 21 patients with urothelial cancer [21] and 74 patients with bladder cancer [22], respectively, and different rules for a positive test. The present ROC analyses showed similar good AUCs for the single PIs of chromosomes 3, 7 and 17 compared with the multiple PI for the combination of chromosomal gain at these loci. The study design, number of cases, proportion of high-grade cancers and the assessment rules applied to the complex data on chromosomal instability may influence the sensitivity estimations. The UroScreen data support a possible reduction of the probes applied to chromosomes 3, 7 or 17 based on strong correlations between the CNVs seen in cases as well as in the 1575 non-cases. Although few urothelial cells are presumed to be undergoing mitosis [23], tetraploidy in proliferating cell populations may partially explain the correlation of the DNA gain at the three loci seen in non-cases. Several groups excluded tetraploid cells from the decision on positivity [2, 24]. We observed similar good correlations in non-cases when excluding tetraploid cells (data not shown). This is in agreement with another investigation of healthy subjects showing a low fraction of cells with DNA gain [23].

The sensitivity of an assay depends upon the choice of a threshold for a positive test result. The UroVysion assay generates complex data for chromosomal changes. We have chosen a threshold value for all PIs of 10% cells with CNV of >2 that showed a performance similar to the decision rule of the manufacturer for all probes. Chromosomal changes assessed with PI 3, as well as with PI 17, detected all nine cases that were tested positively with the manufacturer's decision rule using the full information from four probes. In principle, the addition of other probes might improve the sensitivity of a FISH assay. However, the sensitivity is only improved if CNVs at these putative loci are not correlated with the changes already detected with this test. The sensitivity can be improved in a panel with other markers, e.g. combined with NMP22.

Polysomies in chromosomes 3, 7 and 17 showed high HRs for developing a bladder tumour, comparable to the UroVysion test and to the multiple PI. However, we showed a lower HR of a homozygous deletion at 9p21, when a threshold of 2.5% of cells with such a loss was applied. Also, associations of CNVs at 9p21 with gains at chromosomes 3, 7 and 17 in cases were much weaker and even lacking in cancer-free men. Although loss at 9p21 has been reported as one of the most frequent chromosomal changes in urothelial tumours [8, 9], we found only a small fraction of cells with these chromosomal changes. Therefore, we had to reduce the threshold to only 2.5% of all cells. Sokolova et al. [21] reported that this probe shows a lower sensitivity. Loss of 9p21 can be a feature of non-invasive papillary bladder tumours, whereas polysomies are seen more frequently in CIS and invasive urothelial tumours [8, 9]. However, there were only a few non-invasive tumours in the present study. This role of 9p21 observed in UroScreen needs to be validated.

In addition to developing generally accepted criteria for a positive UroVysion test, it is also important to develop criteria for the evaluation of cells. Because chromosomal instability might occur earlier than morphological changes, investigating all exfoliated cells instead of just morphologically abnormal cells could improve UroVysion, especially in comparison with cytology.

In conclusion, the costly UroVysion assay showed a reasonable performance in detecting bladder cancer but shared positive test results with cytology, a much cheaper test. A simpler, faster and cheaper version of the UroVysion test could rely on the very strong correlations between the observed gains at chromosomes 3, 7 and 17, which yield a similar performance in detecting bladder cancer using single-probe PIs compared with the full set of data from all probes. A loss of 9p21 was less predictive for detecting bladder cancer in male chemical workers, where there were few low-grade cancers.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Subjects and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Conflict of Interest
  9. References

UroScreen Study Group: BASF SE, Department of Occupational Medicine and Health Protection, Ludwigshafen, Germany: Bernd Scheuermann, Friedhelm Eberle, Thomas Mayer, Michael Nasterlack.

German Social Accident Insurance's Institution for the Raw Materials and Chemical Industry (BG RCI), Heidelberg, Germany: Harald Wellhäußer, Matthias Kluckert.

Organisationsdienst für nachgehende Untersuchungen (ODIN), Heidelberg, Germany: Reinhard Detzner.

Institute for Prevention and Occupational Medicine of the German Social Accident Insurance (IPA), Institute of the Ruhr University Bochum, Bochum, Germany: Dirk Taeger, Beate Pesch, Nadine Bonberg, Katarzyna Gawrych, Heike Bontrup, Georg Johnen, Thomas Behrens, Judith Delbanco, Evelyn Heinze, Thomas Brüning.

Currenta GmbH & Co.OHG, Safety – Health Protection, Leverkusen, Germany: Prof. Dr. Gabriele Leng, Dr. Martin Pelster, Christian Bayer, Kay-Gerald Bierfreund, Christian Wiens.

Institute of Urology, Eberhard Karls University, Tübingen, Germany: Gerhard Feil, Karl-Dietrich Sievert, Séverine Banek, Christian Schwentner, Margarete Geiger, Erika Senger, Valentina Gerber, Andrea Hohneder, Gundi Beger, Miriam Dzupinova, Ursula Kuehs, Jörg Hennenlotter, Arnulf Stenzl.

Conflict of Interest

  1. Top of page
  2. Abstract
  3. Introduction
  4. Subjects and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Conflict of Interest
  9. References

Arnulf Stenzl received free assay kits from Abbott Laboratories and Matritech/Alere GmbH. Arnulf Stenzl received travel grants from Matritech/Alere. Georg Johnen received free assay reagents from Fujirebio Diagnostics, Inc. Matthias Kluckert and Harald Wellhäußer are employees of sponsor. Source of funding: German Social Accident Insurance.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Subjects and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Conflict of Interest
  9. References