Although detection of p53 protein by immunohistochemical testing was originally thought to indicate p53 gene mutation, recent analyses of human malignancies have shown that high expression of p53 protein may occur without detectable gene mutation. Several explanations have been proposed for this phenomenon, including mutation out of "hot spot" regions, overexpression of wild-type protein, sampling error in molecular analyses, and conformational changes of wild-type p53 protein. As discussed, it is unlikely that the first two possibilities contribute significantly to the occurrence of this phenomenon, and the current study examined the possibility that sampling error in molecular analyses might account for a lack of concordance between immunohistochemical and molecular analyses. Such a possibility exists because immunohistochemical studies frequently report high expression when staining is only focal or regional and molecular analyses are based on the polymerase chain reaction, which is highly exponential in nature and may not detect mutation if the target gene segment is not amplified early in the chain reaction. In the current report, p53 protein expression was examined by immunohistochemical testing in 45 cases of endometrioid carcinoma, and all cases showing diffuse positivity were then examined by polymerase chain reaction in combination with single-strand conformational analysis for exons 4 to 9 with the use of a microdissection technique to separate malignant from benign cells. Of the 45 cases, diffuse staining was found in four cases, and only two of the four were found to show evidence of gene mutation.(ABSTRACT TRUNCATED AT 250 WORDS)