Cerebrospinal fluid cytology diagnosis of meningeal carcinomatosis in patients with small-cell carcinoma of the lung. A study of interobserver and intraobserver variability.

The interobserver and intraobserver variation in the cytologic diagnosis of malignancy was determined in 62 cerebrospinal fluid (CSF) specimens from 46 patients with small-cell carcinoma of the lung. In all patients, lumbar puncture was carried out because of suspected central nervous system metastases. Forty CSF specimens from 26 patients with meningeal carcinomatosis and thus with a high probability of a positive CSF cytology were mixed with 22 specimens from 20 patients without meningeal carcinomatosis. The slides were evaluated blindly by two observers, one of whom evaluated all specimens on two separate occasions; only positive, negative and suspicious conclusions were permitted. The consistency of the intraobserver and interobserver conclusions on the initial CSF specimen in each case was 87%. In 13% of the initial CSF specimens in each case, a suspicious conclusion was reached in one of the three evaluations. For all 62 CSFs, the intraobserver and interobserver disagreement was 2% and 3%, respectively. In the first and second evaluations by the one observer and the single evaluation by the other, 17 (65%), 15 (58%) and 12 (46%), respectively, of the 26 "high probability" patients were found to have malignant cells in the CSF. CSF cytology was negative in all 20 patients without meningeal carcinomatosis. Of 10 patients with autopsy-proven meningeal carcinomatosis, 40% were not diagnosed while alive. Multiple CSFs from repeated lumbar punctures increased the number of positive evaluations by 30%. At least 60% of those patients with a suspicious CSF cytology did in fact have meningeal carcinomatosis. On the other hand, 30% of the patients with a positive lumbar puncture had a subsequent negative one.