NordiQC EQA Data and Observations for PD-L1 Biomarker Testing
By Søren Nielsen, Director, NordiQC
Immune-checkpoint inhibitors targeting PD-1 or PD-L1 as pembrolizumab, nivolumab, durvalumab and
atezolizumab have since 2016 been approved for multiple cancers as lung carcinoma, urothelial carcinoma,
melanoma, breast carcinoma and many other solid cancers. At present Immunohistochemistry (IHC) for
PD-L1 expression is the most widely used method to guide the selection of patients to receive treatment
with the specific inhibitors and for patient safety a high diagnostic accuracy of the biomarker test is
required.
PD-L1 IHC biomarker testing is based on the “3D-approach” aligning Drug, Disease and Diagnostic assay
inclusive specific read-out methods for thresholds and cells to be scored. This means, that it is compulsory
for EQA programs as NordiQC to align to this approach in the evaluation the IHC analytical test accuracy
among the participants. Especially the Drug-Disease combination will as such determine the choice of
Diagnostic assay to be used to create reference standard results to be used to evaluate the performance of
the IHC results generated by the participants.
NordiQC has since 2019 offered one segment focusing on PD-L1 IHC in non-small cell lung carcinoma
(NSCLC) and urothelial carcinoma (UC) for treatment with KEYTRUDA® (pembrolizumab) and one segment
focusing on PD-L1 IHC in triple negative breast cancer (TNBC) and UC for treatment with TECENTRIQ®
(atezolizumab) and in both segments using clinically validated and CE-IVD approved companions diagnostic
IHC assays as reference standard methods.
The “PD-L1 KEYTRUDA®” IHC segment has accomplished eight assessments with 68-200 participants in each
of these and a pass rate in the range of 50-91%, average 79% being observed. The insufficient PD-L1 IHC
results were most frequently characterized by false negative results changing the PD-L1 status and
consequently affecting the treatment decision in one or more of the carcinomas/patients included in the
assessment materials used. In average in all eight runs performed, 76% of the insufficient results were
caused by false negative results, 10% by false positive results and in the remaining 14% related to
“technical issues” as poor signal-to-noise ratio compromising the read-out and PD-L1 scoring. In this
segment a successful interchangeability of the different companion diagnostic (CDx) assays SK006 (22C3)
Dako/Agilent, SK005 (28-8) Dako/Agilent and SP263 (Ventana/Roche) has been observed providing similar
pass rates and being superior to the benchmark data for all participants and non-CDx based assays. The
performance of laboratory developed (LD) assays have overall been inferior, but in the latest assessment
runs shown improved pass rates indicating a successful development and identification of reproducible LD
assays for PD-L1 in this segment.
The “PD-L1 TECENTRIQ®” segment initiated in 2019 has been performed three times with 84-123
participants and a pass rate ranging from 55-76%, average 65%. In this segment the approved and clinically
validated CDx assay SP142 (Ventana/Roche) has been most successful with pass rates of 83-93%, whereas
other CDx assays and LD assays giving high pass rates in the “PD-L1 KEYTRUDA®” IHC segment were found
to be inferior and overall generated insufficient results. The insufficient results were most frequently
(average 71%) caused by “technical issues” as extensive non-relevant staining reaction in tumour cells
compromising the specific read-out of PD-L1 in immune cells. The inferior performance of non-SP142 CDx
and LD based assays is most likely related to the design and development of these, as they mainly have
been developed for the demonstration and read-out of PD-L1 expression on the membranes of tumour
cells and typically using the Dako/Agilent CDx assay SK006, 22C3 as benchmark reference method.
The SP142 CDx assay differs compared to the other PD-L1 CDx assays both in the protocol set-up being
based on tyramide amplification giving a more binary positive or negative result compared to other CDx
assays with a more “dynamic” range of intensity and proportion of positivity, but also concerning the read-
out criteria for e.g. TNBC and UC only encountering immune cells and excluding tumour cells to establish
the PD-L1 status.
The data observed in the two NordiQC segments for PD-L1 confirms the need to develop and apply PD-L1
IHC biomarker testing as “Fit-For-Purpose” using appropriate validated reference standards. Only by correct
designation of validated PD-L1 IHC methods and associated read-out criteria aligned to cancer type and
intended drug, it seems possible to deliver accurate test-results to stratify patients for the eligibility for
immune-oncology.
Open access to in-depth data for the two NordiQC PD-L1 IHC segments can be found at www.nordiqc.org