During ASCO 2018, Durm et al reported data from LUN 14-719 (Phase II).40 The objective of the trial was to evaluate the efficacy of consolidation therapy with pembrolizumab 200 mg Q3W for 1 year in patients with unprogressive stage III NSCLC following concomitant chemoradiotherapy. The primary end point was time to metastatic disease or death. A total of 93 patients were included (92 evaluable): 59.8% stage IIIA N2 and 40.2% stage IIIB. The most commonly used chemotherapy regimen was carboplatin–paclitaxel (71.7% of cases). The median number of pembrolizumab cycles was 13.5 (1–19); 84% of patients received at least four cycles and 43.5% of patients achieved overall pembrolizumab injections (one year of treatment). The median time to metastatic disease or death was 22.4 months (95% CI 17.9–NR) and median PFS was 15.4 months (95% CI 11.9–NR), and 16 patients (17.2%) developed grade ≥2 pneumonia (10.8% grade 2% and 5.4% grade 3/4). A death related to radiation pneumonitis was observed. These results are quite consistent with those observed in the PACIFIC trial in terms of both tolerance and efficacy, whereas the beginning of immunotherapy was later (between 28 and 56 days after the end of radiotherapy–chemotherapy at 1–42 days in the PACIFIC trial).40

Data on other combinations are more limited. Results have been reported for the pembrolizumab–entinostat combination. Entinostat is a histone-deacetylase inhibitor with the potential to modulate myeloid-derived suppressor-cell functions, and may synergize with PD1/PDL1 inhibition. The preliminary results of a Phase II trial (57 patients) on entinostat plus pembrolizumab for patients with NSCLC previously treated with ICIs were reported at ASCO 2018.41 Five of the patients included achieved a confirmed partial response (ORR 9%, 95% CI 2.9–19.3). Among the responders, four had PDL1 expression <1% in tissue collected at inclusion. Median DOR was 4.2 months. Twenty (35%) patients experienced grade 3/4–related AEs, and six (10.5%) experienced grade 3/4 IRAEs (three pneumonitides).

Continue Reading

Thirty patients were given vorinostat, another histone-deacetylase inhibitor, and pembrolizumab in a Phase I/II trial. PDL1 expression was ≥1% in 18 of 30 (60%) and ≥50% in eleven of 30 (37%) patients. Six patients were ICI-naïve and 24 ICI-pretreated. The disease-control rate was 67% for the entire study population. Among the ICI-pretreated patients, two achieved a partial response and ten had stable disease, and their PFS was 3.2 months compared to 7.6 months for ICI naïve patients.42Pembrolizumab was evaluated in previously untreated patients with EGFR+, PDL1+ (>1%) NSCLC. Enrollment was stopped due to lack of efficacy after eleven of the 25 planned patients had been treated. Only one patient had an OR (ORR 9%), but repeat analysis of this patient’s tumor definitively proved the original report of an EGFR mutation to be erroneous.


Regulatory agencies8–12,43,44 have now approved three ICIs to treat NSCLC – nivolumab, pembrolizumab, and atezolizumab – which are indicated for second-line or more therapy,6,13 regardless of PDL1-expression rate for nivolumab and atezolizumab, and restricted to >1% expression for pembrolizumab. Two Phase III trials have evaluated nivolumab vs docetaxel, one for advanced squamous-cell NSCLC, the other for nonsquamous-cell NSCLC.8,44 In the latter Phase III trial on nivolumab vs docetaxel for patients whose non-squamous NSCLC had progressed after first-line chemotherapy, OS was prolonged (12.2 vs 9.4 months, respectively) without PFS improvement, but with better ORR for the nivolumab arm (19% vs 12%). Tumor PDL1 expression (Dako antibody clone 28-8) predicted a better response. OS for patients with ≥1% or <1% PDL1 expression, respectively, lasted 17.7 vs 9.0 months or 10.5 vs 10.1 months comparing nivolumab vs second-line chemotherapy for each level.44

For atezolizumab, approval was based on the Phase III OAK trial.45 In that study, atezolizumab reduced the risk of death by 26% compared with docetaxel for patients with advanced NSCLC following the failure of platinum-based chemotherapy. Median OS was prolonged by 4.2 months. A survival benefit with atezolizumab was observed independently of PDL1 status or histology. For the intent-to-treat population, median OS was 13.8 months for atezolizumab vs 9.6 months for the docetaxel arm (HR 0.74, 95% CI 0.63–0.87; P=0.0004). Among the PDL1+ group, median OS was 15.7 months for atezolizumab vs 10.3 months for the control arm (HR 0.74, 95% CI 0.58–0.93; P=0.0102). Among PDL1− patients, median OS was 12.6 months for atezolizumab vs 8.9 months for the docetaxel group (HR 0.75, 95% CI 0.59–0.96; P=0.0205).45

First-line pembrolizumab monotherapy has been approved only for stage IIIB/IV NSCLC with ≥50% PDL1 expression and without EGFR mutation, or ALK or ROS1 (transmembrane tyrosine-kinase receptor) rearrangement.46 Moreover, in January 2017, based on the promising KEYNOTE-021 trial results, the FDA approved first-line pembrolizumab in combination with carboplatin–pemetrexed chemotherapy for patients with nonsquamous NSCLC independently of PDL1 expression.12 In February 2018, the FDA approved durvalumab, a human (IgG1κ) monoclonal antibody that blocks PDL1 interaction with PD1 and CD80, to treat stage III, unresectable, locally advanced NSCLC that had not progressed after concurrent chemoradiotherapy, based on the Phase III PACIFIC trial results showing that durvalumab prolonged PFS by 11.2 months compared to placebo (16.8 vs 5.6 months, HR 0.52, 95% CI 0.42–0.65; P<0.0001). Their respective PFS rates in favor of durvalumab were 55.9% vs 35.3% at 12 months and 44.2% vs 27.0% at 18 months.43


Unlike other ICIs (eg, nivolumab or atezolizumab), pembrolizumab monotherapy depends on tumor-cell PDL1 expression (≥1% for second-line, ≥50% for first line), which is frequent in NSCLC.26,47,48Selecting patients based on their PDL1-expression level obtains better pembrolizumab efficacy, especially for first-line therapy. It is essential to identify biomarkers able to select potential responders to immunotherapy more optimally.

The results of several studies have shown that ICI efficacy was associated with tumor-cell PDL1-expression level.17,44,47,49–52 Indeed, in the KEYNOTE-001 trial,28 a significant association was found between PDL1 expression and pembrolizumab efficacy, with an optimal expression threshold of 50% according to receiver-operating-characteristic curve analysis. Tumors with ≥50% PDL1 expression were defined as strongly positive, those with 1%–49% expression as weakly positive, and <1% expression as negative.28 For the validation cohort (KEYNOTE-001), with 313 patients according to the three aforementioned PDL1 groups, ORRs were 45.2%, 16.5%, and 10.7% and median PFS lasted 6.3, 4.1, and 4.0 months, respectively.28

For nivolumab, in the Phase I trial that included all types of solid tumors, including advanced NSCLC, only PDL1+ (≥5%) tumors responded.53 The results of the Check-Mate 057 trial found better nivolumab efficacy in patients with strongly expressed PDL1.44 The Phase I atezolizumab trial and OAK study also showed an association between PDL1 levels expressed by intratumor-infiltrating immune cells and clinical response.44,54 A meta-analysis of seven studies on ICI-treated NSCLC patients (n=914) found that those with PDL1+ (≥1%) tumors had ORRs significantly higher than patients whose tumors were PDL1− (OR 2.44, 95% CI 1.61–3.68).48

Nonetheless, clinical responses have been observed in patients with PDL1− tumors, showing that IHC-determined PDL1 expression is an imperfect marker predictive of response to ICIs.8–10,44,55 Several things could explain those findings. In the first studies, IHC was not standardized, and several labeling techniques were used with at least four monoclonal antibodies (clones 22C3, 28-8, SP142, and SP263) and developed as companion tests for different PD1 or PDL1 inhibitors. Several groups decided to compare the four available antibodies on different platforms. Hirsch et al showed that three (28-8, 22C3, and SP263) of the four reagents were indeed comparable in terms of sensitivity, specificity, and reproducibility.56 Adam et al found that antibodies 28-8, 22C3, and SP263 had similar analytical performances for labeling tumor cells in seven different centers.57

PDL1 expression is a qualitative variable, which explains the choices of different positivity thresholds depending on the study.10,11,43 According to a recent meta-analysis of anti-PD1/PDL1 therapeutic agents, a 5% threshold seemed to have maximum discriminatory power (OR 2.72, P=0.01).52 In addition, heterogeneous PDL1 expression is possible in a given tumor or between the primary tumor and its metastases.58

Finally, in most studies, PDL1 expression by NSCLC cells was assessed by IHC, without taking into account the immune microenvironment, but no definitive information is available on the role of PDL1+ macrophages in the prediction of response to ICIs.49,50 Moreover, no consensus exists about the dynamic character or not of PDL1 expression over time. Indeed, PDL1 can be induced, notably by IFNγ,18 and tumors not expressing that marker can become PDL1+ because of a chronic inflammatory phenomenon.18 Because modification of PDL1-expression rate has also been observed during NSCLC treatment, interest in rebiopsy of tumors before starting anti-PD1/PDL1 therapy has been raised.59

Other potential biomarkers for the selection of populations sensitive to ICIs have emerged in the literature, eg, the tumor-mutation burden (TMB). Because the most effective ICIs against tumors have elevated levels of somatic mutations, it has been suggested that TMB could play an important role in the response to PD1/PDL1 inhibitors.50 Sequencing of the entire exome of pembrolizumab-treated NSCLC showed that a high TMB was associated with a PFS benefit and higher ORR.60 A study evaluating the role of comprehensive cancer-gene panels (300 genes) to estimate TMB showed that the association between TMB and ICI clinical benefit was also seen when cancer-gene panels were used to estimate TMB.51 Furthermore, the results of several studies have shown that TMB was not associated with PDL1 expression.50 The use of PDL1 expression and TMB might obtain a higher predictive value for response to ICIs than individual use of one or the other of these biomarkers.61


The KEYNOTE-024 trial results showed the superiority of first-line immunotherapy over platinum-based chemotherapy for advanced PDL1 ≥50% NSCLCs.11 Pembrolizumab obtained the most benefit against squamous-cell carcinoma in terms of PFS, with HR of 0.35 (95% CI 0.17–0.71). That was a major breakthrough for these tumors, for which no therapeutic progress had been made for several years. With regard to OS benefit, it has to be considered that only 50% of the chemotherapy-arm patients benefited from crossover to pembrolizumab at disease progression. The safety profile was quite acceptable for a first-line treatment, requiring monitoring and providing information on IRAEs. Despite reservations about PDL1 expression by tumor cells addressed herein, this biomarker enables selection of patients getting the most benefits from first-line pembrolizumab. In all the international recommendations, pembrolizumab is the standard of care for patients with advanced PDL1 ≥50% NSCLC without EGFR mutation or ALK rearrangement and PS 0–1.

For patients whose tumors express <50% PDL1, data on the pembrolizumab–chemotherapy combination seem promising. The results for KEYNOTE-021 cohort G showed doubling of the ORR (55% vs 28%), and the recent findings of the KEYNOTE-189 trial confirmed the superiority of the ICI–chemotherapy combination compared to chemotherapy alone, regardless of PDL1 status.12,47 It should be underlined that the magnitude of the pembrolizumab PFS benefit (alone or with chemotherapy) compared to chemotherapy alone was the same in the three trials: 0.53 in KEYNOTE-021, 0.52 in KEYNOTE-189, and 0.50 in KEYNOTE-024.11,12,36 The safety of the pembrolizumab–pemetrexed–platinum combination was very close to that of chemotherapy alone.36 KEYNOTE-189 trial results should in the near future lead to discussion of that combination as the new standard of care for patients whose NSCLC cells express ≥50% PDL1.

The challenge over the next 2 years is to identify predictive biomarkers of ICI efficacy. PDL1 expression is not optimal for the selection of patients to receive ICIs. Many patients have NSCLC not expressing PDL1, and others have hyperprogression under ICIs, despite having PDL1+ tumors. The search for the best combination of predictive biomarkers (PDL1+, TMB, and/or others) must be a top-priority objective for research teams. Simplification and harmonization of techniques for the reproducibility of biomarker detection are other prominent foci of interest for the near future. Academic societies and regulatory authorities should encourage and launch major projects to harmonize practices and lower the costs of testing for these biomarkers.