Diagnostic accuracy of fine-needle aspiration cytology of palpable breast masses: an SROC curve with fixed and random effects linear meta-regression models

Abstract

We used various meta-analytic methods to compare 25 studies describing fine-needle aspiration (FNA) cytologic analyses performed from 1984 to 2007 on palpable breast masses. We found that in the 25 studies examined, the sensitivity ranged from 78% to 100%, the specificity ranged from 76% to 100%, and the diagnostic odds ratio (DOR) ranged from 15.83 to 33 198. The overall diagnostic accuracy was found to be as follows: 0.93 (95% CI: 0.92-0.94) for sensitivity, 0.98 (95% CI: 0.97-0.98) for specificity, and 505.209 (95% CI: 273.08-934.95) for the DOR. The overall diagnostic accuracy according to the results of summary receiving operating characteristic (SROC) curve analysis was 0.95 +/- 0.0032, and the overall weighted area under the (receiving operating characteristic [ROC]) curve (AUC) was 0.99 +/- 0.0014. The DOR values did not show a large variation in the various positivity threshold values. However, the results of those studies had some heterogeneity. The four covariates that were added to the standard SROC model to evaluate variations in the results of the studies were the year of publication, the number of aspirations, the percentage of insufficient material, and the study design. The relative diagnostic accuracy of studies performed after 1990 was 3.98 times higher than that of studies performed before 1990. The relative DOR (RDOR) value was also found to be statistically significant (95% CI: 1.22-13.02). That result may be attributed to the technologic improvements in diagnostic tools over the years. Although it was not statistically significant, an increase in the number of aspirations caused an increase in the RDOR of the FNA cytology (95% CI: 0.52-8.11). In contrast, increasing the percent of insufficient material caused a statistically insignificant but clinically significant decrease in the RDOR of FNA (RDOR = 0.79, 95% CI: 0.21-2.98). In conclusion, our meta-analysis has shown that FNA cytologic analysis of palpable breast masses is highly accurate in the diagnostic differentiation of benign from malignant tumors.