1. ¹Department of Pathology The University of Texas MD Anderson Cancer Center, Houston, TX, USA
2. ²Department of Biostatistics, The University of Texas MD Anderson Cancer Center, Houston, TX, USA

Correspondence: Dr J Liu, MD, PhD, Department of Pathology, Unit 85, The University of Texas MD Anderson Cancer Center, 1515 Holcombe Blvd, Houston, TX 77030-4095, USA. E-mail: jliu@mdanderson.org.

Received 18 December 2003; Revised 4 February 2004; Accepted 4 February 2004; Published online 9 April 2004.

Abstract

High-throughput tissue microarray allows many clinical specimens to be analyzed simultaneously on a single slide. One potential limitation of tissue microarray is the correct representation of each tumor with the small tissue core. Because tumors from different organs have different levels of heterogeneity, it requires a validation study for each one of them. We compared immunostaining of Ki-67, estrogen receptors, and p53 in whole sections of 45 cases of high-grade serous ovarian carcinoma with six core samples from those sections with regard to the number of tissue cores needed to reliably represent a whole section. Staining for Ki-67 was graded high or low by automated image analysis of 10 high-power fields; staining for estrogen receptor and p53 was scored on a 0-to-3 scale. Correlation coefficients for whole-section vs core stains were 0.86 for Ki-67, 0.93 for estrogen receptors, and 0.82 for p53. A total of 54 (6.6%) of the cores were inadequate for scoring. The probability that results from one core would correctly represent all three markers in the whole section was 91%; that for two cores was 96%; and that for three cores was 98%. Our results show that analysis of a single readable core matched the staining pattern of a whole section more than 90% of the time, and analysis of two cores increased that value to more than 95%, demonstrating that ovarian carcinoma tissue microarray is a reliable technique to analyze the expression of markers.