Intrapatient Tumor Heterogeneity in IHC Interpretation Using PD-L1 IHC 22C3 pharmDx

Abstract

Tumor heterogeneity may impact immunohistochemical (IHC) interpretation, thus potentially affecting decision making by treating oncologists for cancer patient management. Programmed cell death ligand-1 (PD-L1) IHC 22C3 pharmDx is a companion diagnostic used as an aid in identifying patient eligibility for treatment with pembrolizumab (KEYTRUDA). This study aims to investigate tumor heterogeneity impact on IHC staining when evaluating PD-L1 expression using PD-L1 IHC 22C3 pharmDx. The effect of tumor heterogeneity was evaluated based on the PD-L1 diagnostic status of PD-L1 IHC 22C3 pharmDx stained tumor tissue sections at relevant diagnostic cutoffs for non–small cell lung carcinoma, gastric or gastroesophageal junction adenocarcinoma, urothelial carcinoma, head and neck squamous cell carcinoma, esophageal cancer and triple negative breast cancer. Overall agreement for the PD-L1 diagnostic status was assessed for each tumor type within a given specimen block (Intra-Block), between specimen blocks from the same surgical resection (Intra-Case), and between intrapatient primary and metastatic specimens. Intrablock and intracase point estimates were above 75%, and primary versus metastatic point estimates were above 50%. The results suggest that PD-L1 expression is consistent across cut sections through a minimum of 150 µm within a tissue block and between blocks from the same surgical resection and is significantly maintained across primary and metastatic blocks from the same patient despite changes to the tissue microenvironment. These data provide confidence for histopathologists and oncologists that evaluation of PD-L1 expression at clinically relevant cutoffs is reproducible among different assessments (or samplings) of a single tumor specimen.

An artificial intelligence‐powered PD‐L1 combined positive score (CPS) analyser in urothelial carcinoma alleviating interobserver and intersite variability

“…This will help pathologists make accurate evaluations to better match patients to specific treatments, ultimately allowing better prediction in patient prognosis and response to treatment. 22 In particular, when pathologists initially read a borderline CPS score (5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15) and were asked to revisit with the AI model, they were more likely to change the CPS result, so it is expected that the AI model will help pathologists make decisions in these borderline areas when applied to real-world clinical practice in the future. Furthermore, the AI analyser can be especially helpful when seeking remote reading or second opinions, such as through telepathology.…”

“…8,9 Representative methods for evaluating PD-L1 expression level in 22C3 antibody stained immunohistochemistry (IHC) slides include tumour proportion score (TPS) in non-small cell lung cancer (NSCLC) and a combined positive score (CPS) in triple-negative breast cancer (TNBC) and UC. 10 TPS evaluation considers PD-L1-positive or -negative tumour cells, while CPS evaluation additionally considers PD-L1-positive lymphocytes and macrophages. 11 Since PD-L1 expression level evaluation is done by pathologists, there is a problem of interobserver variability.…”

“…The expression level of PD‐L1 in tumour tissue is important for predicting the clinical response of PD‐L1 inhibitors 8,9 . Representative methods for evaluating PD‐L1 expression level in 22C3 antibody stained immunohistochemistry (IHC) slides include tumour proportion score (TPS) in non‐small cell lung cancer (NSCLC) and a combined positive score (CPS) in triple‐negative breast cancer (TNBC) and UC 10 . TPS evaluation considers PD‐L1‐positive or ‐negative tumour cells, while CPS evaluation additionally considers PD‐L1‐positive lymphocytes and macrophages 11 .…”

Esophagogastric Cancers: Integrating Immunotherapy Therapy Into Current Practice

“…However, how to interpret discordant results (making treatment decision on the basis of lowest v highest CPS) is not clear at present time. [9][10][11][12] All of these issues underscore the complex biology of upper GI cancers, as well as underline the need for continued optimization of PD-L1 as a biomarker and for identification of new predictive biomarkers. At present time, DAKO 22C3 clone is the most commonly used antibody to determine…”

Avoiding “False Positive” and “False Negative” Immunohistochemical Results in Breast Pathology

“…False-positive or false-negative PD-L1 testing results are mostly caused by postanalytic interpretations because the reproducibility of PD-L1 scoring has been a concern due to the complexity of scoring systems and the cutoff value of 1%. Although scoring 22C3 labeling has been shown to have good inter-/intra-observer agreements, there is substantial debate on the reproducibility of scoring SP142 labeling [123][124][125][126][127][128][129][130][131][132][133] (Fig. 9).…”