The prognosis for women with endometrial cancer correlates with stage of disease. Spread to the cervix distinguishes Stage II from Stage I disease. To assess the accuracy of endocervical curettage (ECC) in predicting cervical involvement by endometrial adenocarcinoma, we examined and assigned to one of four groups the ECC from 147 women treated between 1980 and 1985. Ultimate determination of spread to the cervix was based on examination of subsequent hysterectomy specimens. In the hysterectomy specimen 19 of 147 women (13%) had cervical involvement demonstrated. Five women had tumor clearly within the endocervical tissue of the ECC (Group I), and three of these five (60%) had spread of tumor to the cervix in the hysterectomy specimen. At hysterectomy, 13 of 41 women (32%) with tumor present but not contiguous with endocervical tissue within the ECC (Group II) had cervical involvement. Of the 80 women without tumor in the curettage (Group III), one (1.2%) had cervical involvement by tumor at hysterectomy. Two of the 21 women (9.5%) with curettings considered insufficient for diagnosis (Group IV) had spread of adenocarcinoma to the cervix within the hysterectomy specimen. We also compared the results of ECC performed at our institution with those performed at referring hospitals and found a significant difference between them in ability to obtain definitive results. We conclude that (a) the absence of carcinoma in ECC is highly predictive of absence of cervical involvement by endometrial carcinoma; (b) ECC having tumor within endocervical tissue is a good predictor of cervical involvement by tumor.(ABSTRACT TRUNCATED AT 250 WORDS)