A high PD-L1 score may no longer settle the first-line NSCLC question

For more than a decade, the PD-L1 tumor proportion score has been the biomarker that guide first-line treatment in advanced non-small cell lung cancer (NSCLC). A score of 50% or higher has signaled that a patient could skip chemotherapy and start a checkpoint inhibitor alone. A meta-analysis in JAMA Oncology suggests that the biomarker, while still central, is a far blunter instrument than its widespread use implies.

The study published earlier this month and led by Alessandro Di Federico, M.D., from the IRCCS Azienda Ospedaliero-Universitaria di Bologna in Bologna, Italy, and conducted by researchers at Dana-Farber Cancer Institute in Boston and the University of Bologna, pooled 24 phase 3 randomized trials covering 5,546 patients with PD-L1-high NSCLC. Using reconstructed individual patient data, the authors compared PD-(L)1 inhibitor monotherapy or chemoimmunotherapy versus chemotherapy alone in the same biomarker-defined population.

Both approaches beat chemotherapy alone. But the combination, chemoimmunotherapy, pulled ahead. Median overall survival reached 29.2 months with chemoimmunotherapy compared with 19.8 months with a checkpoint inhibitor alone. Median progression-free survival was 11.3 months versus 6.8 months. The advantage held across meta-regression, network meta-analysis and sensitivity analyses were limited to trials with five years of follow-up.

The biomarker story sits underneath those numbers. A high PD-L1 score is supposed to identify the patients most likely to respond to immunotherapy by itself while avoiding toxic effects of chemotherapy. Yet response rates in this group remain modest, and a substantial share of patients progressed early. The authors noted that PD-L1 expression "exhibits spatial and temporal variability," which limits its reliability as a predictive marker. The same tumor can score differently depending on where and when it is sampled, and a single cutoff at 50% lumps together patients with meaningfully different biology.

That heterogeneity has practical consequences for decision-makers. The 50%-or-higher threshold is treated as a defining line in coverage policy and clinical pathways, but the data suggested it does not cleanly separate patients who need chemotherapy from those who can spare it. The researchers could not resolve whether expression in the very high range, 90% or above, behaves differently from the 50% to 89% band, because the underlying trials did not report at that granularity. Histology, smoking status, metastatic sites and co-mutations such as STK11 and KEAP1 likewise could not be factored in.

The result is a biomarker that defines a population without fully predicting its response. Chemotherapy appears to help by reducing tumor burden, boosting antigen presentation and blunting early resistance to immunotherapy, mechanisms that operate regardless of PD-L1 level. In the analysis, fewer patients on the combination progressed within the first three months.

That benefit carried a cost. Chemoimmunotherapy produced significantly more treatment-related adverse events, more grade 3 or higher toxicities, and more treatment discontinuations than monotherapy. “These findings underscore that decisions to intensify treatment should be individualized, carefully balancing the magnitude of efficacy benefit against tolerability, particularly in patients with borderline ECOG [Eastern Cooperative Oncology Group] Performance Status or substantial comorbidities,” wrote Di Federico and his co-authors in the discussion section of the paper.

No completed head-to-head trial has yet tested the two strategies directly. The National Cancer Institute-sponsored INSIGNA trial, which includes a prespecified PD-L1-high subgroup, may offer the most relevant prospective evidence. Newer agents, including the PD-1/VEGF bispecific antibody ivonescimab, are also pushing toward intensification even in biomarker-selected patients.

In light of their findings, Di Federico said that in an ideal study design, researchers would randomize patients with untreated advanced NSCLC and a PD-L1 score of 50% or higher either toi PD-(L)1 inhibitor monotherapy or chemoimmunotherapy, with overall survival as the primary outcome and prespecified subgroup analyses by PD-L1 expression level (50%-89% vs ≥90%), histology, metastatic sites and relevant variants (e.g. STK11, KEAP1), variables that could not be evaluated in their study.