Laboratory assessment of HER2neu-testing. Bayes’s Theorem, quality assurance and implications for routine daily practice. Weltevreden E.F. and Kornegoor R. Department of Clinical Pathology, Gelre hospitals, Apeldoorn, The Netherlands
Background: The number of pharmaco-pathologic tests in routine histopathology has increased dramatically over the last decade. To guide adjuvant systemic treatment in breast cancer, the clinico-pathological risk assessment is based on the main prognostic factors; axillary nodal status, size of the primary tumour and histological grade. The choice of (neo-)adjuvant therapy depends on e.g. the HER2 status. Critical appraisal of HER2neu-testing showed discordance rates between immunohistochemistry (IHC) and in situ hybridisation (ISH) and inter-observer variation in pathological examination of breast carcinomas results in significant differences in the HER2 status. It is against this background that the assessment is performed to get insight in our actual performance in HER2neu-testing, to establish a so called outcome-indicator necessary for accreditation programs that guarantee laboratory quality. Design: For this retrospective study we used the rules of Bayesian statistics because of its rationalty, its consistency and logic interpretation. The predictive value of the HercepTest can be derived from Bayes’s theorem in which prevalence (prior odds) and the probability quotient determine the predictive value (posterior odds) of a test. Data have been retrieved from “PALGA”, the nationwide network and registry of histo- and cytopathology in the Netherlands", for a 1 ½ year period immediately prior to analysis. The data comprise 376 consecutive routinely performed semi-automated immunohistochemical HER2neu-tests (IHC), using the HercepTest Kit (DAKO) and the ASCO/APA 2008 guidelines for interpretation.
Additional results (n=103) of manually performed chromogenic in situ hybridisation tests (CISH), were included as the “golden standard” using the ZytoDot SPEC HER2 Probe Kit. All testing was performed at the laboratory for Clinical Pathology, Gelre Hospitals, Apeldoorn, The Netherlands. The numerical relation between IHC and CISH at the threshold IHC-2+ versus IHC+3+ forms the basis of this study.
CISH; amplification HER2neu
Results: At the “protein level” (IHC), the scores in the HercepTest were negative ( 0 or 1+) in respectively 138 (“0”) en 122 (“1+”) cases, strongly positive (“3+”) in 46 cases, while the equivocal score (“2+”) showed up 70 times. So, at the “protein level” 12.2 % scored “positive”, an over rated percentage in case no confirmation is found at the “DNA-level”. Seven percent (n=9) of the IHC-1+ cases and about half of the IHC-3+ cases (n=24) was followed by a CISH confirmation test, while all (n=70) of the IHC-2+ cases were subjected to CISH analysis. No CISH was performed in IHC-0 cases except in the initial validation study that showed no amplification in IHC-0 cases. No amplification of the HER2neu gene was present in IHC1+. IHC-2+ cases showed amplification in 14.3% . The percentage Fals Negative (FN) test outcome was 3% of the total number of IHC-tests. Three IHC-3+ cases ( = 12.5%) showed no sign of amplification, referred to as Fals Positive (FP). This proved to be partly due to polisomy following CISH to the centromere of chromosome 17, or to misinterpretation. The probability quotient = 16.1 and after extrapolation of the CISH findings, the true percentage (= prevalence) of cases with amplification of the HER2neu gene was 13.3 %. Predictive value = prevalence / ( prevalence + (1 – prevalence / probability quotient)) So, derived from Bayes’s theorem, the predictive value of the HercepTest in the prediction of amplification = 0.71 Conclusion: Our findings are in concordance with the literature. The IHC-test is in fact more suited to predict absence of amplification and is less valuable in predicting amplification of the HER2neu gene. We suggest to additionally perform a standard CISH to the centromere of chromosome 17 to detect polisomy restricting the number of falls positive cases and to automation of the CISH procedure to strengthen the “gold standard”. Quality assurance in IHC HER2neu-testing needs a confirmation-test at the DNA-level in all equivocal cases and all positive cases. Outcome indicators like the positive predictive value can simply be assessed on a regular (yearly) bases to monitor laboratory quality.
CISH Confirmation-test obligate ! Adapted from Her2neu algorith M. Bilous et al et al Mod Pathol 2003:16(2):173-182