Recommendations issued on PD-L1, TMB testing

NSCLC and immune checkpoint inhibitor selection

Sherrie Rice

July 2024—In patients with advanced non-small cell lung cancer, a validated PD-L1 immunohistochemistry expression assay should be used, with other targetable genomic biomarker assays where appropriate, for the selection of immune checkpoint inhibitor therapies. And the appropriate validation should be performed on all specimen types and fixatives.

Those are two of the six recommendations of the CAP, Association for Molecular Pathology, International Association for the Study of Lung Cancer, Pulmonary Pathology Society, and LUNGevity Foundation, released in April and now in print in their PD-L1 and tumor mutation burden testing guideline for the selection of such therapies (Sholl LM, et al. Arch Pathol Lab Med. 2024;148[7]:​757–774).

“Despite all the complications we see and how difficult and complex it can be, PD-L1 testing is still a cornerstone of the biomarker testing for non-small cell lung cancer,” said Larissa Furtado, MD, co-chair of the guideline and molecular pathologist at St. Jude Children’s Research Hospital. She went behind the recommendations to explain the rationale for them in a CAP session at the USCAP annual meeting in late March, prior to the guideline’s release.

The four PD-L1 companion diagnostic assays are 22C3 and 28-8 (both Dako/Agilent) and SP142 and SP263 (both Ventana). The anti-PD-1 and anti-PD-L1 immune checkpoint inhibitor combinations approved in the United States for use in advanced and metastatic NSCLC have been approved with a paired PD-L1 IHC assay that was established and validated in randomized trials. “Each companion assay is determined by a specific combination of an antibody clone, an IHC testing platform and reagents, and different scoring systems to assess PD-L1 expression,” Dr. Furtado said.

In addition, a laboratory-developed test may use one of the available anti-PD-L1 clones used for companion diagnostics or another clone that has not been validated in prospective clinical trials, in combination with any available IHC platform.

“For reasons of cost and access,” she said, “PD-L1 IHC antibodies and assays developed outside the scope of randomized, controlled trials have gained widespread use, which leads to confusion on the part of pathologists and clinicians about the best approach for PD-L1 testing to select patients for immune checkpoint inhibitor therapy.”

The other PD-L1-related recommendations suggest using clinically validated IHC assays as intended, when feasible; validating according to the requirements of the lab’s accrediting body if a laboratory-developed test for expression is used; and reporting IHC results using a percentage expression score.

For tumor mutation burden, the groups recommend that clinicians not use TMB alone to select patients with advanced NSCLC for immune checkpoint inhibitors.

The rationale for recommending the use of a regulatory-approved, clinically validated PD-L1 IHC assay

for selecting patients for immune checkpoint inhibitor therapy is based on the evidence from more than 30 studies, including 14 randomized clinical trials, that evaluated the overall survival and response rates following treatment with various immunotherapy agents in tumors with known PD-L1 status. These studies found significant correlation between PD-L1 expression in tumor tissue samples and patient response and survival following immunotherapy with PD-1 and PD-L1 inhibitors given alone or in combination with chemotherapy and/or CTLA-4 inhibitors.

“Although PD-L1 expression is not an absolute predictive biomarker because we see response and benefit in some patients with no visible expression in their tumor sample, and we see lack of benefit in some patients despite high PD-L1 expression, PD-L1 expression can be a useful biomarker to optimize selection for treatment with immune checkpoint inhibitors,” Dr. Furtado said.

In addition to PD-L1 testing, there are other genomic biomarkers for which routine testing is recommended in advanced non-small cell lung cancer, such as EGFR mutations and ALK rearrangement, because several of these biomarkers have FDA-approved therapies or are known to be associated with resistance to therapy. “Knowing the status of EGFR and ALK driver alterations is relevant when you’re considering immunotherapy in patients with advanced non-small cell lung cancer because multiple clinical trials of immune checkpoint inhibitors have shown there is a relative lack of benefit, meaning there is a low response rate, for patients with driver alterations in the EGFR and ALK genes,” Dr. Furtado said.

Concurrent therapy of EGFR TKI and a PD-L1 or PD-1 inhibitor has been associated with immune-related adverse events such as interstitial pneumonitis. Because patients with NSCLC who have targetable genetic alterations derive substantial benefit from appropriate targeted therapy in the first-line setting, genomic biomarker testing in conjunction with PD-L1 IHC testing is indicated to optimize therapy selection.

One of the questions the panel looked at was related to the preanalytic, analytic, and postanalytic considerations for PD-L1 testing,

which can be complicated, Dr. Furtado said, especially when determining what specimen should be tested to provide an accurate result for PD-L1 and tumor mutation burden. The biological heterogeneity of PD-L1 expression can make sample selection challenging.

For instance, she said, it’s not uncommon for primary lung adenocarcinomas to show different tumoral patterns (Fig. 1), “and there can be striking heterogeneity of PD-L1 expression, with less differentiated areas showing little to no PD-L1 expression and higher rated areas showing much more prominence of positivity.”

“It’s possible,” in looking at cases such as shown in Fig. 2, “to consider that the limited biopsy of a larger mass may not accurately represent the PD-L1 status of the whole tumor and that results obtained from small biopsies may differ depending on what region of the tumor is sampled.”

The intra-tumor heterogeneity of PD-L1 expression is not restricted to the different tumoral patterns. Geographic variability in PD-L1 expression can also be seen across different areas of tumor, she said, as demonstrated in the resection specimen in Fig. 3.

How PD-L1 expression from biopsy samples correlates with that of resection specimens is a relevant question, Dr. Furtado said, but the evidence available to answer the question is limited and restricted to retrospective analysis in part because the patients with advanced non-small cell lung cancers who are candidates for immune checkpoint inhibitors will in most cases not undergo resection of the primary tumor or the metastatic lesion.

The studies that compared the performance characteristics of PD-L1 expression in small biopsy samples and resection specimens found high concordance between the two sample types (83 percent to 95 percent), especially at the 50 percent cutoff. At the one percent cutoff, it was 77 percent to 86 percent. In most discordant cases, a lower score is obtained from the small biopsy sample than from the corresponding resection specimen. Despite the intratumoral heterogeneity of PD-L1 expression, “based on these findings, testing of small biopsy specimens is considered to be acceptable,” Dr. Furtado said.

The panel also considered testing of the primary tumor versus metastatic sites but could not make a recommendation because there were no studies that evaluated the response rate to PD-1 or PD-L1 among the checkpoint inhibitors, based on testing specimens from different sites. “However, the available evidence in the literature demonstrates that discordance of expression scores between the primary tumor and synchronous metastatic lesions occurs in about 20 percent to 30 percent of cases,” Dr. Furtado said. “So, when dealing with these cases, a reasonable approach is to test the specimen that has the best technical quality, meaning the specimen that has the highest viable cellularity and best preservation.”

Many patients with NSCLC, particularly those with advanced disease, are diagnosed by fine-needle aspiration or other cytologic techniques, and those specimens may have to be used for PD-L1 testing. The studies that compared PD-L1 expression on tumor cells in the cytology cell blocks from various sources, including fine-needle aspirates and body fluids, with expression in the paired resection specimens showed good concordance, ranging from 67 percent to 94 percent, with an overall concordance of 85 percent.

A small number of studies have evaluated the use of PD-L1 immunocytochemistry testing on alcohol-fixed cytology material. They showed high concordance of PD-L1 expression for the 22C3 and SP263 companion diagnostic assays—97 percent and 89 percent at the cutoffs of one percent and 50 percent (Lozano MD, et al. Cancer Cytopathol. 2019;127[7]:470–480; Noll B, et al. Cancer Cytopathol. 2018;126​[5]:342–352). “In most discordant cases, the score obtained from the aspirate smear was lower than that from the paired resection specimen,” Dr. Furtado said.

While this limited evidence shows there are promising comparisons between cytology and resection specimens, she said, cytology specimens may be employed for PD-L1 testing only if formalin-fixed, paraffin-embedded material is not available and only in laboratories that validated this sample type. Results should be reported with a comment that PD-L1 testing in non-FFPE tissues has not been clinically validated.

The analytical comparisons between PD-L1 IHC assays reveal high concordance

(90 percent or greater) of PD-L1 expression for the 22C3, 28-8, and SP263 companion diagnostic assays, while the SP142 clone generally shows weaker expression compared with the other three assays (Fig. 4). “For this reason,” Dr. Furtado said, “SP142 would not be considered appropriate for the selection of patients for treatment with pembrolizumab, nivolumab/ipilimumab, or even cemiplimab because they’re not interchangeable.”

Despite the overall evidence for the comparable performance for many of the PD-L1 assays, she said, there are not many studies that have compared the predictive ability of the different PD-L1 assays in a well-controlled prospective clinical trial of checkpoint inhibitor therapies.

One such study was a phase three trial of adjuvant atezolizumab versus chemotherapy for advanced non-small cell lung cancer (Herbst RS, et al. N Engl J Med. 2020;383[14]:1328–1339). The patients were enrolled based on the status of PD-L1 testing by SP142, but other antibodies were looked at at the same time. The authors assessed how comparable SP142 was with SP263 and 22C3. “And consistent with prior retrospective analyses, the SP142 assay captures a lower percentage of the patients that have higher-level expression of PD-L1 in the tumor staining cells, while the concordance between SP263 and 22C3 is much higher and overlapping between them,” Dr. Furtado said.

Also important is that when looking at the defined cutoffs for high-level expression of the different antibodies, “there were similar benefits in terms of how the patients responded to the therapy that was administered if you were to select the patients based on different antibodies.”

There are many opportunities, she said, for clinical trials to include this type of study and look more prospectively at how the outcome of these patients could vary based on the application of different antibodies.

For that reason, the expert panel recommends that pathologists use the companion assays that are regulatory approved and clinically validated until larger cohort studies confirm the comparability of the companion diagnostic assays from both technical and clinical standpoints.

Use of LDTs is acceptable following validation to ensure the clinical performance is acceptable.

In reporting the PD-L1 percent expression score, Dr. Furtado said,

“one of the questions is whether it matters if you provide a specific range or at least a certain value of your percent score.” The literature showed that the levels of PD-L1 expression are significantly correlated with patient response and survival following immunotherapy with immune checkpoint inhibitors administered in isolation and in chemo-immunotherapy combinations (Garon EB, et al. N Engl J Med. 2015;372[21]:2018–2028; Sezer A, et al. Lancet. 2021;397[10274]:592–604).

“So studies have demonstrated there is a higher response rate and there are improved outcomes in these patients, particularly in those who have high expression levels of at least or above 50 percent.” Thus, instead of reporting a “low” or “high” score, she said, “it is important to provide a specific level.” Providing an exact percent expression score value may be challenging, though, owing to the subjective nature of visual assessment of PD-L1 expression and the scoring variability among pathologists, particularly on immune cell scoring across different PD-L1 assays. One of the reporting options, she said, is to use a semiquantitative approach in which incremental values and ranges are provided instead of absolute scores—for example: <1 percent, 1–10 percent, 11–20 percent, 21–30 percent.

“This not a formal recommendation; it’s just an option.”

Tumor mutational burden is an FDA-approved test with a tumor-agnostic approval.

But the KEYNOTE-158 clinical trial that led to this tumor-agnostic approval included 102 tumors that were TMB-high, none of which were non-small cell lung cancer, Dr. Furtado said (Marabelle A, et al. J Clin Oncol. 2020;38[1]:1–10). “And the problem is when you look at the way TMB can vary, the range in different tumor types, even within the same tumor type, it raises the question whether the cutoff point for TMB-high that was established in the clinical trial based on a certain assay of greater than 10 mutations per megabase is the number that should be used to select patients with non-small cell lung cancer for therapy with immune checkpoint inhibitors.” The panel recommends that clinicians not use TMB alone to select patients with advanced NSCLC for immune checkpoint inhibitor therapy, owing to the insufficient evidence in this population.

Sherrie Rice is editor of CAP TODAY.