With immunotherapy making inroads into the treatment of non–small cell lung cancer (NSCLC), there is an increasing need for tests for programmed death ligand–1 (PD-L1) that can identify patients for whom such therapy is suitable. But these tests are expensive, and there can be problems with reimbursement, say French researchers, who propose that pathology laboratories produce their own tests.
Pathology laboratories can make PD-L1 tests that can match standard assays, but care needs to be taken to ensure they are validated before clinical use, say the researchers, led by Julien Adam, Department of Pathology, Gustave Roussy Cancer Campus, Villejuif, France.
The researchers compared the performance of more than 25 laboratory-developed tests (LDTs) against standard assays for PD-L1 and found that they achieved concordance in just more than half of cases, Dr Adam reported here at the 17th World Conference on Lung Cancer (WCLC).
Dr Adam said that the results will now form the basis of recommendations for pathology laboratories when developing their own LDTs, so that they can achieve the same level of concordance as standard assays and validate their output.
Discussing the study, Michael Boyer, MD, PhD, chief clinical officer and conjoint chair of medical oncology, from the Chris O'Brien Lifehouse, Camperdown, Australia, said that the study is important because it concerns "how we select our patients for treatment and the mechanisms and methods for testing."
He noted that the development of PD-L1 testing has been "complex" and that the current situation is "messy." Nevertheless, he said: "Knowing the way that our laboratories do the tests and having confidence in the results we get is of great value and great importance.
"To me, it seems, as individual clinicians, we have to understand what our laboratories are doing and, therefore, what the results mean when they come to us," Dr Boyer said. "It also seems to me that the laboratories have to validate their own results."
Commenting on the presentation at the meeting, Shrish Gadgeel, MD, team leader of thoracic oncology, Karmanos Cancer Institute, Wayne State University, Detroit, Michigan, told Medscape Medical News that the role of PD-L1 testing has recently taken on greater significance.
He noted that before the approval of pembrolizumab (Keytruda, Merck), "not every agent was necessarily based on PD-L1 testing, so there was not an added incentive to do testing." Although the approval of pembrolizumab specifies that this drug be used for patients whose tumors have a high expression of PD-L1 (tumor proportion score of 50% or more), other immunotherapies, such as nivolumab (Opdivo, Bristol-Myers Squibb), can be used without such testing. The recent approval of pembrolizumab for first-line use in NSCLC has led experts to recommend that all NSCLC patients be tested for PD-L1 status to determine whether they are eligible for therapy.
Dr Gadgeel said: "Where Dr Adam's data become very important is because it's become challenging in various laboratories in the USA to decide which platform you use and which antibody to use."
Although the approval of pembrolizumab is based on PD-L1 testing with 22C3 antibodies, he pointed out that "in the future, it may not be the same antibody, the same platform, and this harmonization will give us a level of confidence that, if you have yourApproached for comment, H. Jack West, MD, medical director, Thoracic Oncology Program, Swedish Cancer Institute, Seattle, Washington, told Medscape Medical News that the study "corroborates the general conclusion" that the majority of PD-L1 assays provide concordant results, aside from the SP142 test for atezolizumab (Tecentriq, Genentech/Roche).
"However, the current question should be whether and why one would choose to use an assay other than the 22C3 antibody," he continued. "Pembrolizumab is the only agent that is currently restricted by PD-L1 expression, and the 22C3 test has now become a standard of care for newly diagnosed patients with advanced NSCLC."
Dr West added: "This means that the question now needs to be whether any other test provides any significant incremental benefit to lead one to favor it over 22C3. Right now, I think that the clear answer is 'no,' that other PD-L1 tests provide a lateral move at best. At this moment, 22C3 is the de facto standard."
French Study of Tests Produced by Path Labs
At the meeting, Dr Adam began his presentation by noting that primarily two platforms for PD-L1 testing have been used in clinical trials: Dako, which uses the 22C3 and 18-8 assays; and Ventana, which uses the SP142 and SP263 assays.
He said that the harmonization of assays and the establishment of LDTs are "urgently needed" in France, because the two platforms are not available in all pathology laboratories, the assays are expensive, and reimbursement is "insufficient."
Moreover, PD-L1 testing has to be rapidly available for patients in the first-line setting, and performing multiple tests with different assays is not feasible with the small tissue samples that are usually obtained from NSCLC patients.
To examine the performance of the 28-8, 22C3, and SP263 assays across centers and to determine whether LDTs can perform in a manner similar to the official assays, the team selected 41 resected NSCLC tissue specimens known to have various levels of expression of PD-L1.
Seven centers took part in the study. Of those, three used the Dako AS Link 48 platform, two used the Ventana Benchmark Ultra plaform, and the two used the Leica Bond III. The 22C3, 28-8, E1L3N, SP142, and SP263 clones were used for assays, either on the dedicated platform or on an LDT.
This resulted in 35 PD-L1 stainings on 30 protocols, giving a total of 1435 slides across all 41 NSCLC cases. Seven pathologists took part in the study. Each case was scored by a single pathologist who was blinded as to the center, antibody, and platform used. The results were assessed separately for tumor cell and immune cell staining.
As expected, use of the 28-8 and 22C3 assays on the Dako platform and the SP263 assay on the Ventana platform for tumor cell staining resulted in high levels of concordance, at weighted kappa coefficients ranging from 0.71 to 0.89, with at ≥50% tumor cells stained in 95.1% of cases.
One LDT on the Ventana platform was concordant with the standard 28-8 assay (weighted kappa coefficient = 0.80), and two LDTs on the same platform were concordant with the 22C3 assay (weighted kappa coefficient = 0.81 and 0.77).
All five LDTs on the Dako and Leica platforms using the SP263 were found to be concordant with the standard SP263 assay, at weighted kappa coefficients ranging from 0.83 to 0.86.
In addition, two LDTs using the SP142 assay on the Leica platform were concordant with the standard SP263 assay (weighted kappa coefficient = 0.78 and 0.81), and four LDTs on the Dako, Ventana, and Leica platforms using the E1L3N assay were concordant, at weighted kappa coefficients ranging from 0.75 to 0.81.
With respect to immune cell staining, there was poor concordance both with the standard assays and the LDT. Dr Adam speculated that that could be due to a number of factors, including variations in the intensity of staining and the number of cell categories.
He noted that among the 27 LDTs developed on the Dako, Ventana, and Leica platforms, 14 (51.8%) had concordance similar to the standard assays for tumor cell staining, with the 22C3, 28-8, and SP263 assays performing particularly well.
The clone SP263 assay achieved the highest concordance rate across all platforms, at an overall weighted kappa coefficient of 0.81, followed by the E1L3N assay, at 0.78. The lowest concordance was achieved by the SP142 assay, at a weighted kappa coefficient of 0.64 across all platforms on tumor cell staining.
In contrast, low concordance was seen for all immune cell staining assays across all platforms. The best concordance was again observed for the SP263 assay, which had a weighted kappa coefficient here of just 0.64.
Dr Adam concluded that caution is needed as to the use and validation of LDTs and that the findings will be used in France to develop national recommendations for PD-L1 testing in NSCLC.
Speaking at the press conference, Dr Adam told Medscape Medical News that further validation of the results will be sought in additional cohorts and that recommendations on developing LDTs are expected to be published in the second half of 2017.
He said: "We would like people to have clear recommendations on how to perform them but also on how to validate them, because local validation is very important."
Dr Adam continued: "We should look also at the stability of antibodies and what kind of antibodies people use, because E1L3N, which got good results in our study, is right now a research use–only antibody.
"We should have antibodies that are very stable in terms of performance, and this is something that is not completely sure with research use–only antibodies, so they are other issues that are coming."
No funding for the study was reported. The researchers have disclosed no relevant financial relationships.
17th World Conference on Lung Cancer (WCLC). Abstract PL04a.04, p resented December 6, 2016.
own platform...you should be able to use those results clinically.