Marianne J. Ratcliffe, PhD
Three currently available assays for PD-L1 expression exhibited a high degree of correlation in a comparative study involving archived lung cancer tissue, according to results reported at the 2016 AACR Annual Meeting.
Head-to-head comparisons resulted in 91% to 95% correlation for assessment of the same tissue samples. None of the assays distinguished itself from the other two, according to Marianne J. Ratcliffe, PhD, associate director of diagnostics for AstraZeneca in Alderley Park, UK.
“From a practical standpoint, I think what this means, is that pathologists will need to become familiar with the performance of each of the assays,” said Ratcliffe. “They will need to know how each of the assays performs at different levels of PD-L1 expression.”
Whether the results are generalizable to other types of cancer remains unclear. Until additional studies with different tumor types have been completed, the study’s applicability should be limited to non–small cell lung cancer (NSCLC), she added.
Ongoing investigation of PD-1/PD-L1 inhibitors spans a wide spectrum of tumor types. Both pembrolizumab and nivolumab have worldwide regulatory approvals that include NSCLC. Studies of both monoclonal antibodies demonstrated improved response rates in NSCLC in association with high tumor PD-L1 expression.
Several other PD-1/PD-L1 inhibitors are being investigated in tumor types that include NSCLC, Ratcliffe and colleagues noted in a poster presentation. Companion diagnostic tests have been developed for most of the PD-1/PD-L1 inhibitors developed to date.
“The PD-L1 tests intended to be used as companion diagnostics ... have all been developed independently and utilize different antibody clones, immunohistochemistry (IHC) protocols, scoring algorithms, and cutoffs for PD-L1 positivity,” the investigators noted. “Efficient use of precious tissue and pathology resources will drive a strong desire to minimize testing demand for PD-L1. It is vital to be able to compare PD-L1 diagnostic assays to allow appropriate interpretation of their predictive value with respect to clinical outcome.”
As a “first step towards assay harmonization,” Ratcliffe and colleagues examined correlation among 3 of the Ventana SP263, Dako 22C3, and Dako 28-8 PD-L1 expression assays. For the analysis, they used 500 formalin-fixed, paraffin-embedded archival samples of NSCLC tissue obtained from commercial sources.
Each assay was performed according to instructions on package inserts. The comparative study included a washout period between reads of tissue samples from the same patient to eliminate bias.
The Ventana SP263 served as the reference assay. Negative percentage agreement and positive percentage agreement were calculated for the Dako 22C3 and 28-8 assays relative to Ventana SP263 25% cutoff for PD-L1 expression. The typical within-assay agreement for IHC is >90% overall percentage agreement, the investigators noted.
Clinical and demographic characteristics associated with the tissue samples include relatively even age distribution (<60 years old vs ≥60), 76% male, about 80% stage I-II disease, and 54% nonsquamous histology (43% squamous).
“All 3 assays showed similar patterns of staining,” said Ratcliffe.
The correlation of membrane staining between the Ventana SP263 and Dako 22C3 was 91%. Correlation between the Ventana SP263 and Dako 28-8 was 93.5%. The Dako 28-8 and Dako22C3 had 95% correlation.
Repeat comparisons across different levels of PD-L1 expression yielded similar rates of correlation among the 3 assays, although studies are ongoing to define more clearly the correlation of the Ventana SP263 with lower levels of expression (1-10%).
“There is a high analytical correlation between the 3 different commercially available PD-L1 assays,” Ratcliffe and colleagues concluded. “The assays have closely aligned dynamic ranges, and an overall percentage agreement >90%, which is comparable to typical within assay agreement for IHC, can be achieved across the dynamic range at multiple cutoffs.”
“Currently, direct clinical data supporting a specific diagnostic test should still be considered as the highest standard of proof for diagnostic clinical utility,” they added. “However, our data suggest that the clinical community may consider it appropriate for pathologists to use any of the 3 assays to identify previously defined clinically relevant patient populations.”
Ratcliffe MJ, Sharpe A, Midha A, et al. A comparative study of PD-L1 diagnostic assays and the classification of patients at PD-L1 positive and PD-L1 negative. Presented at the 2016 AACR Annual Meeting; April 16-20; New Orleans, Louisiana. Poster LB-094
<<< View more from the 2016 AACR Annual Meeting