UroVysion FISH test for detecting urothelial cancers: Meta-analysis of diagnostic accuracy and comparison with urinary cytology testing

Objectives: Since the introduction of the UroVysion test for detecting urothelial cancers in urine, its reported performance has varied. This article systematically analyzed reported results. Methods: Articles in English conforming to the Oxford EBM criteria were included, with the evaluation focused oil cancers that were histologically confirmed at the time of testing rather than oil any cancers that might develop later. Where applicable, samples with no cells were reclassified as negative so as to further improve the actual estimation of test performance. Where available, cytology data were also analyzed. Meta-DiSc software was used for the statistical analyses. Results: We identified 14 studies involving 2477 FISH tests. The overall prevalence Of urothelial cancers was 35%. The pooledsensitivity and specificity of all studies were 72% (69%-75%) and 83% (82%-85%), respectively. Cytology data were available from 12 Studies. with the overall sensitivity and specificity being 42% (38%-45%) and 96% (95%-97%). Excluding Ta tumors. the sensitivity was 86% (82%-89%) for UroVysion and 61% (56%-66%) for cytology. The overall performance was higher for UroVysion than for cytology: diagnostic odds ratio, 16.8 and 14.1; AUC, 0.867 (SE 0.021) and 0.626 (SE 0.091). These differences in overall test performance measures almost disappeared when superficial cancer cases were excluded from the analysis. Conclusions: The Published trials Suggest that for a general mix of cases. cytology results are highly specific. However, a negative cytology result does not meaningfully change the post-test probability of the presence of urothelial cancer. UroVysion FISH test results should not be considered to provide conclusive evidence for the presence or absence of urothelial cancer, but both positive and negative results do moderately influence the post-test probability of disease. (c) 2008 Elsevier Inc. All rights reserved.