Combined external quality assessment of cytology and histology opinions: a pilot scheme for a cluster of five laboratories.

Abstract

SUMMARY A cluster of five self selected cytodiagnostic laboratories circulated 70 cervical/vaginal smears and 50 sputum smears in a series of five week cycles. Histological sections related to the abnormal smears were also circulated. Proportions of agreement and disagreement were analysed for cervical and sputum smear reports against the original report (and by implication against the consensus); corresponding calculations were made for the histological reports. Cytological and histological correlation was also examined.Agreement for major categories (benign, intraepithelial neoplasia, malignant) was 87% for cervical cytology and 83% for sputum. When the effect of potential random agreement was compensated for by the use of kappa statistics the values of kappa were +0*79 and +0 65, respectively. The corresponding kappa values for gynaecological and respiratory tract histology were +0-78 and +0-82, respectively. Agreement on finer degrees of abnormality was, predictably, less good.Problems arising in this pilot scheme are discussed and suggestions for a simplified scheme are made.To devise an external quality assessment system for cytology analogous to those in other pathology disciplines such as clinical chemistry or haematology is not easy.' 2 The main problem is not, as is often asserted, the difficulty of comparing qualitative as opposed to quantitative results because this also applies in microbiology, where the problem has been solved.3 Nor is it the problem of comparing interpretive opinions where there is no absolute standard of correctness; kappa statistics can be applied to analyse interobserver variation in pattern recognition4 5 and these have been applied satisfactorily in histological observer studies.6 ' The unique difficulty in cytopathology is that it is not possible to produce identical clinical specimens for simultaneous examination by different laboratories.The only solution to this problem is

Reliability of the Routine Cytological Diagnosis in Bladder Cancer

“…In a study with similar characteristics to the present one [25], the weighted ĸ coefficients for the grading of the malignancies detected in urine cytologies were 0.5 and 0.57, respectively, for urine specimens from patients with primary and recurrent tumors. In another study, the concordance between five departments of pathology for the cytological diagnosis of smears of cervical mucus and sputum was excellent (0.78 and 0.82, respectively) [26]. With regard to the performance of the unweighted ĸ statistic, the calculation of the weighted ĸ coefficient better reflects clinical reality: when one evaluates the ordinal scales composed of categories of increasing severity, some discrepancies are more serious than others.…”

Papnet

“…This interactive mode is attractive, but although fatigue is reduced, it does not overcome the subjectivity of the human examination both in review of the dat tape and in the microscopic examination of selected cells on the slide. This is a well-recognized problem in reporting both histology and cytology, 7,14,17 and there can also be a variation, called "drift," when the same observer rereports material after an interval of a few months. 1,10 Table II illustrates interobserver variation in the present series and also the variability that is possible, depending on the background of the observers.…”

“…After all this, however, there is still the human interactive element, which may have the same problems of subjectivity, particularly with low grade squamous intraepithelial lesions (SILs), as are found when cervical smears and cervical biopsies are read manually. 7,14,17 In pursuance of a previous report, 5 we designed a minitrial to investigate the effect of all these factors and the potential variation between different grades of personnel using the system.…”

Interlaboratory Quality Control in Gynecologic Cytopathology Using the Novel CONQUISTADOR Software

“…25 In this study, the ability to discriminate reliably between CIN 2 and CIN 3 was high (mean variability of index B was 0.89). Recently, Husain et al 36 found a crude agreement of 87% and a nonweighted κ of 0.79 in a comparison of the reporting patterns of 5 laboratories using 3 main diagnostic categories (benign, CIN, and malignant). 36 In our approach, using the diagnostic categories of TBS 2001, including ASC and AGC, a crude agreement of 0.67% was obtained among the 4 laboratories, with nonweighted κ of 0.45 (Table VII).…”

“…[32][33][34][35][36][37][38][39][40] On one hand, data provided by performance evaluation systems are subject to potential bias. 32,34,36 According to EQA principles, any EQA system for cytology laboratories should assess the performance of each participant individually, as well as that of the laboratory as a whole. 37 Circulated slide sets among the laboratories is the core of the interlaboratory QC and quality improvement (QI), and it is the only method to evaluate interobserver and interlaboratory reproducibility.…”