OBJECTIVE To find out the most effective and combined cytomorphologic criteria trying to set up an effective diagnostic model for breast ductal lesion in fine needle aspiration cytology (FNAC). METHODS A total of 400 breast FNAC cases were collected with follow-up information of more than six years. A retrospective analysis including 104 non-proliferative breast diseases, 163 proliferative breast diseases and 133 carcinomas basing on the diagnostic results of surgical biopsies. Altogether, 60 cytomorphologic variables were counted for the evaluation of each case, including 4 main categories: the cellularity and components, natures of background, cellular arrangements and the cellular features. According to the quantity or the classification stage, the variables were semi-quantitatively scored. Multiple step-wise logistic regression (SPSS) and classification tree model (SAS) were performed to determine the significant and combined variables predictive for the diagnosis of non-proliferative lesion, proliferative breast diseases and carcinoma, respectively. RESULTS (1) Among 400 benign and malignant cases studied, and basing on the result of analyses of multiple step-wise logistic regression system, intermingling of myoepithelial cells within the epithelial cluster (P < 0.05), presence of large epithelial cell cluster (P < 0.05), presence of small epithelial cell cluster (P < 0.05), cytoplasmic vacuoles (P < 0.05) and figures of "progressive intussusception" of cells (P < 0.1) were selected as the effectively differential diagnostic criteria for the benign and malignant lesions. However, according to the classification tree model, the most useful variable selected associating with the benign lesion was intermingling of myoepithelial cells within epithelial cluster. The diagnostic accuracy will be increased to 94.4%, if another criterion, presence of a big amount of large epithelial clusters, was used as the second useful variable in combination. Presence of a moderate to large amount of small epithelial cell clusters were indicative of proliferative lesion. If the criterion of myoepithelial cells intermingling within epithelial cluster was not found in the sample and associating with presence of small epithelial cell clusters, cytoplasmic vacuoles and figures of "progressive intussusception" of cells, mostly (81.3%), it would be considered as a case of carcinoma. (2) Among 267 benign non-proliferative and proliferative breast diseases studied, both the multiple step-wise logistic regression and classification tree model, presence of irregular intercellular spaces within the epithelial clusters (P = 0.001), loose epithelial clusters (P < 0.05) and hyperchromasia (P < 0.1) were selected as the significant differential diagnostic criteria for the proliferative lesion. The architectural variables and the amount of the abnormal cell features such as cell cluster formation were considered to be more important. A high frequency of presence of irregular intercellular spaces within the epithelial clusters and the amount of loose epithelial clusters indicated a higher possibility of a proliferative lesion. Presence of a single variable of irregular intercellular spaces within the epithelial clusters had the possibility of a benign lesion diagnosis up to 70.1% in all the proliferative breast disease cases collected in this series. If the frequency of irregular intercellular spaces increased to a moderate degree or even higher, the possibility of a benign lesion would be increased to 82.7%. The possibility of a proliferative breast disease would be reached to 87.5%, if both the criteria of irregular intercellular spaces and loosely arranged epithelial cell clusters were counted in combination. (3) The histological results of 35 lesions with atypical cytological features in FNAC specimens were predominantly a proliferative lesion of the breast (26 cases), and most of them were fibroadenoma with ductal hyperplasia. Occasionally, there might be a few benign cases complicating with lesions of atypical hyperplasia or carcinoma. CONCLUSIONS In breast FNAC diagnosis, a combined evaluation of significant variables and the amount of the variable involved are effective for the differential diagnosis between benign/malignant and non-proliferative/proliferative lesions. Lesion accompanying with atypical cellular features should avoid to be overdiagnosed as carcinoma, and biopsy for a histological diagnosis is indicative.