Open Access Technical advance

Optimisation of an immunohistochemistry method for the determination of androgen receptor expression levels in circulating tumour cells

Jeffrey Cummings1*, Robert Sloane1, Karen Morris1, Cong Zhou1, Matt Lancashire1, David Moore1, Tony Elliot2, Noel Clarke3 and Caroline Dive1

Author Affiliations

1 Clinical and Experimental Pharmacology Group, Cancer Research UK Manchester Institute, University of Manchester, Manchester Cancer Research Centre, Manchester M20 4BX, UK

2 Department of Clinical Oncology, Christie NHS Foundation Trust, Wilmslow Road, Manchester M20 4BX, UK

3 Urology, Christie NHS Foundation Trust, Wilmslow Road, Manchester M20 4BX, UK

For all author emails, please log on.

BMC Cancer 2014, 14:226  doi:10.1186/1471-2407-14-226

Published: 28 March 2014



AZD3514 inhibits and down regulates the androgen receptor (AR) and has undergone clinical trials in prostate cancer. To provide proof-of-mechanism (POM) in patients, an immunohistochemistry (IHC) method for determination of AR in circulating tumour cells (CTC) was developed and validated.


After an assessment of specificity validation focused on intra- and inter-operator reproducibility utilising a novel modification of incurred sample reanalysis (ISR). β-Content γ-confidence tolerance intervals (BCTI) and Cohen’s Kappa (κ) were employed in statistical analysis of results.


In a first set of IHC reproducibility experiments, almost perfect agreement was recorded (κ=0.94) when two different operators scored CTC as overall positive or negative for AR. However, BCTI analysis identified a specific bias in scoring staining intensity, where one operator favoured moderate over strong assignments, whereas the reverse was the case with the second operator. After a period of additional training involving deployment of a panel of standardised images, a second set of validation experiments were conducted. These showed correction of the inter-operator bias by BCTI with κ for scoring intensity increasing from 0.59 to 0.81, indicative of almost perfect agreement.


By application of BCTI to the validation of IHC, operator bias and therefore poor reproducibility can be identified, characterised and corrected to achieve a level of error normally associated with a quantitative biomarker assay, such as an ELISA. The methodological approach described herein can be applied to any generic IHC technique.

AZD3514; Immunohistochemistry; Method validation; Incurred sample reanalysis; Cohen’s Kappa; β-Content γ-confidence tolerance intervals