Role of Tetrasomy for the Diagnosis of Urothelial Carcinoma Using UroVysion Fluorescent In Situ Hybridization

This study was approved by the institutional review board. As a tertiary care hospital, UMass Memorial Health Care (Worcester, Massachusetts) received approximately 23 800 urine cytology cases from 2006 to 2011 with a minimum of 4 years of follow-up. Diagnoses were negative in 17 850 (75%), atypical in 4880 (20.5%), suspicious in 530 (2.2%), and positive in 540 (2.3%). Reflex testing by UroVysion FISH was performed primarily on atypical or suspicious urines and in cases specifically requested by the urologist(s). During this time, FISH was performed on 1532 urines with a cytologic diagnosis of negative in 165 cases (10.8%), atypical in 1204 cases (78.6%), suspicious in 132 cases (8.6%), and positive in 31 cases (2%). One hundred ninety-four patients with urine cytology samples tested by FISH, who had a subsequent biopsy within 3 years, were used for statistical analysis. Patient information is summarized in Table 1.

Unfixed specimens were processed within 24 hours. Specimens fixed in one volume of Saccomanno fixative (Fisher Scientific, Waltham, Massachusetts) at the time of collection were processed within 72 hours. Cells were collected by centrifugation, transferred to PreservCyt (Hologic, Bedford, Massachusetts), and a slide was made using an automated ThinPrep 2000 processor (Hologic). All cases were screened by a cytotechnologist, referred to a cytopathologist for second review, and a diagnosis was made based on the classifications described in Table 1. See supplemental methods for additional details (see supplemental digital content, containing methods and 3 tables, at www.archivesofpathology.org in the June 2016 table of contents).

All slides were prepared for Target-FISH, and brightfield cytomorphology was masked when performing “FISH-only” evaluations. An automated Sakura Tissue-Tek Prisma processor (Sakura Finetek, Torrance, California) was used to stain the FISH slides using a modified Papanicolaou staining protocol (supplemental methods, Table A). An imaging station (BioView Duet, Rehovot, Israel) was used to capture brightfield images of all cellular material. Destaining and FISH pretreatment steps were performed manually or using VP2000 automation (Abbott Laboratories) according to the FDA approved UroVysion protocol with modifications (see supplemental methods for details).

Brightfield images were reviewed by a trained technologist, and a representative sampling of at least 200 urothelial cells were chosen as FISH targets. Squamous or white blood cells with a normal nucleus to cytoplasm ratio were excluded. After the slides were hybridized with FISH probes, the BioView Duet was used to find, match, and classify targets in the fluorescent scan of the same slide.

For FISH analysis, priority was given to cells with abnormal DAPI morphology (large nuclear size, an irregular nuclear shape, or patchy DAPI staining). Polysomy was defined using FDA-approved criteria as cells with gains (3 or more copies) in 2 or more chromosomes (eg, 2:3:4:2). By definition, polysomy includes tetrasomy. To account for artifacts, tetrasomy was defined as having 4 copies of each chromosome 3, 7, and 17 and band 9p21, with up to one missing probe signal (eg, 4:3:4:4). Cells with relative loss of 9p21 (eg, 4:4:4:2) were classified as polysomy and excluded from tetrasomy. In Target-FISH, the cytopathologist can use brightfield morphology to override the FISH classification. Nineteen FISH cases were excluded from the target-FISH analysis because data were not available. Cells with other abnormalities (eg, homozygous loss of 9p21, 4:4:4:0), were excluded from analysis to maintain the focus on tetrasomy and polysomy, which is likely to account for a reduction in assay sensitivity.

The FISH was performed on the concurrent or follow-up formalin-fixed, paraffin-embedded surgical biopsy in all patients (n = 14) with tetrasomy cells (≥10 cells) in the urine cytology.¹²  Pretreatment of specimens selected by the pathologist was performed according to the Abbott PathVysion (Abbott Laboratories) insert. Preparation, hybridization, and posthybridization wash of the UroVysion probe is described above. Diagnostic areas were enumerated using a Nikon E600 fluorescent microscope (Nikon Inc., Melville, New York) equipped with the Abbott filter set by a technologist and pathologist highly experienced in FISH techniques. To control for artifacts of nuclear truncation, signal counts in the tumor were compared with normal urothelium in the same specimen or with a normal database (see supplemental methods for details).

Employing biopsy data as the gold standard, sensitivity and specificity for the 194 cases were calculated for standard FISH and target-FISH using various cell count threshold combinations. Statistical analysis was performed using Fisher exact test as a 2-tailed test with significance at P < .05. The relative accuracy of 3 different FISH categories (polysomy including tetrasomy, polysomy excluding tetrasomy, and tetrasomy only) was compared by receiver operating characteristics (ROC) analysis. Positive thresholds for each category were determined. The DeLong et al method¹³  was used to compare that area under the curve (AUC). The statistical calculations were performed using the IBM SPSS statistical software package (version 19.0 for Windows, SPSS, Chicago, Illinois).

Table 1 shows a summary of demographics and surgical and cytologic diagnoses. Most patients (125 of 194; 64%) were referred for bladder cancer screening because of hematuria (72 of 194; 37.1%), imaging (17 of 194; 8.8%), or reasons not specified (36 of 194; 18.6%). Cytologic diagnosis was available in 191 of 194 cases (Table 2). Most specimens were referred to FISH because the cytology was atypical (118 of 191; 62%).

Three separate categories, consisting of cells with (1) polysomy including tetrasomy, (2) polysomy excluding tetrasomy, and (3) tetrasomy only, were used to construct ROC curves. The ROC analysis was used to determine the cell count threshold needed to achieve maximum accuracy (sensitivity by specificity) when used to diagnose UC (Figure 1). Category 1 (polysomy including tetrasomy) produced a threshold of 8.5 cells, whereas category 2 (polysomy excluding tetrasomy) yielded a threshold of 4.5 cells. The AUC, a measure of diagnostic accuracy, was greatest in category 2 (polysomy excluding tetrasomy), but not significantly different from category 1 (P = .43). By contrast, category 3 (tetrasomy alone) yielded an AUC of 0.501, which was not significantly different from the null hypothesis of AUC = 0.5; therefore, a threshold could not be determined.

The ROC analysis was used to explore whether tetrasomy was associated with tumor grade (Figure 2, a and b). For high-grade tumors, category 2 (polysomy excluding tetrasomy) provided the highest AUC but was not significantly different from category 1 (polysomy including tetrasomy) (P = .94). In low-grade tumors, our analysis showed that category 1 and category 2 had similar AUCs and comparable accuracy at 4 cells (Figure 2, b). Category 1 offered greater sensitivity (66.1% versus 43.5%), whereas category 2 offered greater specificity (59.2% versus 88.7%) for the diagnosis of low-grade tumors.

Based on the ROC analysis described above, a positive threshold of 4 cells was used for categories 1 and 2 in subsequent analyses. A previously published⁹  threshold of 10 cells was used for tetrasomy alone (category 3).

The diagnostic utility of different positive criteria, based on inclusion or exclusion of tetrasomy cells (categories 1 and 2), was investigated. Table 3 describes specificity and sensitivity calculations for standard FISH. We found that excluding tetrasomy from polysomy significantly increases specificity (59.2% versus 78.9%; P = .02), but at a cost in sensitivity (78.9% versus 65.9%; P = .03). Considering tetrasomy alone (category 3) as a criterion for positive FISH, tetrasomy in 10 cells or more had a sensitivity of 13% and a specificity of 90.1%.

Category 4 was created to evaluate the value of tetrasomy cells for positive criterion when polysomy cells failed to reach a positive threshold of 4 cells (tetrasomy when polysomy was <4). The results of category 4 were similar to those of category 3 (tetrasomy only). In an attempt to boost the test performance, we combined category 2 as a first positive criterion, with category 4 as the second positive criterion. This combination (categories 2 and 4) did not yield a significant difference in sensitivity or specificity compared with category 2 alone (P = .30 and P = .56, respectively).

To evaluate the potential benefits of cytomorphology, the results of standard FISH and Target-FISH were compared (Table 4). The use of Target-FISH in conjunction with category 2 significantly increased specificity to 93.7%, compared with standard FISH (78.9%; P = .02) without changing the sensitivity (P = .40). Combining categories 2 and 4 in Target-FISH did not yield a significant difference in sensitivity compared with category 2 alone (P = .54).

We investigated the value of including (category 1) or excluding (category 2) tetrasomy cells in positive FISH criteria for high- and low-grade UC (Table 5). In standard FISH, the exclusion of tetrasomy improved specificity (P = .02) without significantly changing sensitivity for the detection of high-grade UC (P = .28). By contrast, exclusion of tetrasomy improved specificity (P = .03), but significantly reduced sensitivity, for detection of low-grade UC (P = .02). Similar results were obtained in patients referred for screening (new diagnosis), who showed a higher proportion of low-grade tumors (supplemental Table C). Target-FISH further improved specificity (P = .01) without significantly changing sensitivity (P = .45) regardless of tumor grade.

Chromosome tetrasomy was relatively common in this study population. Of specimens with a biopsy, 4 or more tetrasomy cells were observed in 47 of 194 cases (24%), and 23 (12%) of these cases had 10 or more tetrasomy cells (supplemental Table B). We looked more closely at specimens containing tetrasomy and separated cases based on the presence or absence of polysomy cells. Of all FISH cases without polysomy, including those without a biopsy, 4 or more tetrasomy cells were observed in 268 of 1532 cases (17.5%), and 66 (4.3%) of these cases had 10 or more tetrasomy cells. In patients referred for screening (n = 1218), 233 (19%) had tetrasomy in 4 or more cells, and 55 (4.5%) had tetrasomy in 10 cells or more.

Patients in this study were followed from 2011 to 2015, and 7 additional UC diagnoses were reported (3 low grade, 4 high grade). Most patients with 10 or more tetrasomy cells (50 of 55) who were referred for screening (n = 1218) showed no evidence of UC during this time. Three of these 55 patients (5.5%) were diagnosed with UC (1 low grade, 2 high grades). Two of the 55 patients (3.6%) were lost to follow-up after the initial FISH test.

To determine whether the presence of tetrasomy cells in the urine reflected a chromosomal abnormality in the bladder tissue, FISH was performed on paraffin-embedded biopsy tissue of 14 cases with tetrasomy in 10 or more cells. In normal urothelium, truncation effects reduced probe counts in 2% to 28% of cells, producing average probe counts of 1.72 to 1.98. We found scant evidence of tetrasomy or truncated “near-tetrasomy” nuclei in the tumor tissue, despite extensive scanning including review of large nuclei. Three cases were benign (Table 6, cases 1–3) and the associated biopsy specimens were disomic by FISH (Figure 3, a through c). Only 1 of 7 tumor tissues showed chromosome tetrasomy (case 8) from tumors with paired urine specimens showing polysomy in fewer than 4 cells and tetrasomy in 10 or more cells (cases 8–14). The remaining tumors were disomic, showed loss of the 9p21 locus, or exhibited chromosome polysomy in tissue FISH (Figure 4, a through f).

UroVysion FISH is an established adjunct for UC diagnosis. However, controversy remains regarding the inclusion of tetrasomy cells in the polysomy category because this chromosome pattern may be associated with benign conditions. We investigated whether tetrasomy has value as a separate diagnostic category in standard FISH. In our study, ROC analysis was unable to establish a statistically significant threshold for tetrasomy. There was no significant benefit to using tetrasomy in 10 or more cells as a separate category alone or in combination with polysomy in 4 or more cells (Table 3).

Previous studies have suggested that tetrasomy may be more common in benign conditions or certain cell types, (eg, umbrella cells).^9,14,15  In our biopsy population, tetrasomy cells were associated with benign histology (n = 11) more often than it was with low-grade or high-grade UC (n = 10) in specimens lacking polysomy cells (supplemental Table B). In agreement, most patients referred for screening who have tetrasomy cells but lack polysomy cells in FISH did not develop UC.

To determine whether tetrasomic cells observed in cytology specimens are present in the tissue biopsy, we performed FISH on the matching surgical specimens. Of urine FISH specimens with tetrasomy in 10 or more cells, only one low-grade tumor showed tetrasomy on the biopsy. A limitation of this technique is nuclear truncation, which can reduce probe copy number in paraffin sections. However, truncation of tetrasomy cells produce the appearance of imbalanced chromosomal gains (3–4 signals of each probe),¹⁶  which was not observed. In most cases, the origin of tetrasomy cells in the urine remains unclear, but these results support that tetrasomy in the tumor is exceedingly rare. Most tetrasomy cells observed in a FISH specimen may represent benign cells from the upper tract or disomic tumor cells undergoing cell division. Of course, the possibility remains that tetrasomic cells were present in the urothelium but were not sampled by the biopsy.

In our study, exclusion of the tetrasomy cells from the polysomy class in standard FISH yielded the highest accuracy. The ROC analysis demonstrated a positive threshold of 4.5 cells in polysomy without tetrasomy was optimal for detection of UC. Although this threshold appears to be identical to the FDA criteria,¹⁷  previous studies did not exclude tetrasomy cells from the polysomy class. We found a threshold of 8.5 cells was required to achieve similar accuracy when tetrasomy cells were included in the polysomy class. The higher threshold was necessary to preserve specificity and may reflect population differences among studies. However, in our study, factors known to affect test performance, such as tumor prevalence, tumor grade, and cytology diagnoses fell within previously reported ranges.^10,18,19 

At our institution, most FISH cases (1204 of 1532, 78.6%) were reflexed from atypical cytology, which represented 20.5% of all cytology cases (4880 of 23 800). This atypia rate falls within the previously reported range of 1.9% to 32.5%.^20,21  Atypical cytology cases present a greater challenge for FISH (eg, scant tumor cells) and may explain why this study falls in the lower range of reported sensitivity.^10,18,19  The diagnostic sensitivity of UroVysion FISH is higher in suspicious urine cytology than it is in atypical urine cytology.²²  Stricter criteria for cytologic atypia have been proposed (eg, the Paris system²⁰) and are likely to reduce the rate of atypical cytology. Tumor prevalence and the number of patients with UC should increase in the atypical urine category because more-benign conditions will be excluded. This change should reduce FISH testing, increase positive results, and reduce erroneous diagnoses based on tetrasomy cells.

Tumor prevalence also affects the predictive values of UroVysion FISH. For institutions with a low UC prevalence (ie, most patients tested because of hematuria), a diagnostic test with high specificity will provide the best performance. An overall prevalence of 8.5% (130 of 1532) was observed in our FISH population, which is comparable to previous studies (10.8%–13.8%).^18,23  Exclusion of the tetrasomy from the polysomy class increased positive predictive value (PPV) of standard FISH from 0.15 to 0.22 (P = .02) with a marginal impact on negative predictive value (NPV) (0.961 versus 0.968, P = .92). The high NPV increases confidence in a benign diagnosis and avoids the cost and burden of additional surveillance.

For institutions with a high UC prevalence (ie, limiting FISH testing to patients with a history of UC), a diagnostic test with high sensitivity will usually provide the best performance. Patients with biopsy follow-up in this study represent a high risk subset of the entire FISH population and showed high tumor prevalence (63%; 123 of 194). In the biopsy population, exclusion of the tetrasomy cells from the polysomy class had a nonsignificant impact on PPV and NPV of standard FISH (Table 3), increasing PPV from 0.77 to 0.84 (P = .18), but reducing NPV from 0.62 to 0.57 (P = .63). Given the high PPV in this situation, a positive result warrants further investigation even when cystoscopy is negative, as an upper tract tumor cannot be ruled out.

Target-FISH offers the ability to retain tetrasomy as a positive FISH criterion and to detect rare, but genuine, tetrasomy tumors. Compared with standard FISH, Target-FISH improved specificity (93.7% versus 78.9%; P = .01) without adversely affecting sensitivity (65.9% versus 65.2%; P = .45) (Table 4). Target-FISH also provided the highest PPV and NPV in low-risk (prevalence, 8.5%, PPV, 0.49 [P < .001], NPV, 0.967 [P = .93]) and high-risk populations (prevalence, 63%, PPV, 0.948 [P = .03], NPV, 0.602 [P = .66]) when compared with standard FISH (Table 4). This approach helps prevent other diagnostic errors (eg, selecting the correct cells to score), given that an experienced pathologist can identify tetrasomic cells with normal cytomorphology and exclude them from the analysis.

Tetrasomy appears to be a nonspecific finding, typically associated with benign lesions or low-grade tumors. Cells with a tetrasomic chromosome pattern observed in urine cytology specimens are rarely present in the tissue biopsy. Our data support excluding tetrasomy cells in routine practice for the diagnosis of UC. In high-grade UC, this omission significantly improved specificity without affecting sensitivity. For patients with low-grade UC, cystoscopy is likely to compensate for the decrease in sensitivity caused by exclusion of tetrasomy cells. Test results should always be correlated with other clinical findings. Tetrasomy as a positive FISH criterion should be limited to institutions with access to Target-FISH. Target-FISH allows one to separate the tetrasomic cells by morphology and avoid missing the rare tetrasomic tumor.

We thank Stephen P. Baker, MScPH, for his assistance in carrying out the statistical analysis.