Résumé : Background: The aim of this study was to estimate the efficiency of a recent five-category urinary cytological classification. Methods: A total of 592 bladder washings were fixed immediately with Saccomanno's fixative. All samples were centrifuged in a Hettich cyto-centrifuge. For each sample, the reference standard was the histology when a lesion was present at the time of cystoscopy. A five-category cytological classification was used: negative, suspicious of low (S-Lg) or high (S-Hg) grade neoplasia and consistent with low (Lg) or high (Hg) grade neoplasia. Results: For cytological diagnoses of S-Lg and Lg, sensitivity was 37% and specificity was 95% for the histological diagnosis of low-grade non-invasive urothelial papillary tumour (Lg-UPT), which included papillary urothelial neoplasm of low malignant potential and low-grade urothelial carcinoma. For cytological diagnosis of S-Hg and Hg, sensitivity was 44% for high-grade non-invasive urothelial papillary carcinoma (Hg-UPC), 70% for carcinoma in situ (CIS) and 81% for invasive carcinoma (T1 and higher). Specificity was 99% in each case. Cytological diagnosis of S-Hg or Hg was not found for Lg-UPT (0/59) and no cytological diagnosis of S-Lg or Lg was found for invasive carcinoma, but was seen for Hg-UPC in 10% (3/28) and for CIS in 6% (3/50) of cases. Conclusion: Despite the absence of international consensus, the recent five-category cytological classification for urine is accurate for current urological practice. © 2007 The Authors.