In total, 261 patients with locally advanced or metastatic non-squamous NSCLC treated with platinum-based chemotherapy and an anti-PD(L)1 agent were identified between September 2018 and October 2021. Sixty-six patients were excluded, mainly because no NGS testing had been performed (Supplementary Figure A, see ESM). Overall, 195 patients were included. All of them had received treatment with pembrolizumab, platinum, and pemetrexed.

The demographics and clinical characteristics of the patients at baseline are described in Table 1. Median age was 62 years. One hundred and twenty-four patients (64%) were men. Most patients had a favorable PS of 0 or 1 (87%), were current or former smokers (88%), and had metastatic disease (96%).

Twenty-five patients (12.8%) had received prior treatment for advanced disease. Among these patients, nine had received more than one prior line of treatment. Respectively, 20 patients (10.3%), 12 patients (6.2%), and 7 patients (3.6%) had received prior chemotherapy, targeted therapy, or ICI treatment. Three patients with EGFR mutation (out of 8), one patient with ALK rearrangement (out of 5) and one patient with ROS1 rearrangement (out of 2) received CIT in first line, before any targeted therapy.

PDL1 TPS was available in 183 patients (94%). Sixty-eight patients (35%) had a PDL1 TPS < 1%, 60 patients (31%) had a PDL1 TPS of 1–49%, and 55 patients (28%) had a PDL1 TPS ≥ 50%. MET expression was assessed in 72 patients (37%). Thirty-three patients (46%) had a moderate or high (2+ or 3+) MET expression score.

One hundred and eighteen patients (60.5%) had at least one oncogenic driver alteration, including 80 patients (41%) with KRAS mutations (G12C, n = 38), 8 patients (4.1%) with EGFR mutations (exon19del or Leu858R, n = 5; exon 20 insertion, n = 2; other, n = 1), 5 patients (2.6%) with ALK rearrangements, 2 patients (1%) with ROS1 rearrangements, 1 patient (0.5%) with a RET rearrangement, 11 patients (5.6%) with BRAF mutations (V600E, n = 6; non-V600E, n = 5), 4 patients (2.1%) with MET exon 14 mutations, 2 patients (1%) with MET amplification, and 5 (2.6%) patients with HER2 mutations (Supplementary Figure A, see ESM). TP53 mutations were found in 102 patients (52%) and STK11 mutations in 37 patients (19%). Among the oncogenic drivers, only KRAS (n = 16) and BRAF (n = 1, non-V600E) mutations were associated with SKT11 mutations (Table 1).

Forty-one patients received cisplatin (21%) and the others received carboplatin. Maintenance was conducted for 131 patients (67%). The median number of cycles administered was 8 (IQR 4–13). The main reasons for discontinuing treatment were progressive disease (n = 163, 84%), toxicity (n = 21, 11%), and death (n = 13, 6.7%). Thirty-two patients (16%) were still on treatment at the time of data cutoff.

With a median follow-up of 20.2 months (IQR 15.7–25.2), 151 events were recorded. In the entire cohort, median PFS was 6.4 months (95% CI 5.3–7.3). The proportions of patients who were progression free and alive at 6, 12, and 18 months were 51%, 29%, and 22%, respectively.

Regarding driver subgroups, median PFS (mPFS) was 7.1 months (95% CI 5.4–8.9) for KRAS, 5.5 months (95% CI 2.5–15.3) for EGFR/ALK/ROS1/RET, 12.9 months (95% CI 2.6–not reached [NR]) for BRAF, 1.5 months (95% CI 0.6–NR) for MET, 3.9 months (95% CI 2.6–NR) for HER2, and 5.6 months (95% CI 4.7–7.8) for the no-driver subgroup (Fig. 1, Table 2). After adjustment for age, PS, smoking status, disease stage, and number of previous lines, PFS did not differ between the KRAS, EGFR/ALK/ROS1/RET, BRAF, and no-driver subgroups (p = 0.70). Initial stage (IV-B) was the only variable associated with PFS (p = 0.012).

In univariate analysis, PFS did not differ either between the EGFR/ALK/ROS1/RET subgroup and the others (HR 1.00 [95% CI 0.55–1.80]; p = 0.99), between the KRAS subgroup and the others (HR 0.83 [95% CI 0.60–1.14]; p = 0.24), or between the BRAF subgroup and the others (HR 0.63 [95% CI 0.34–1.16]; p = 0.22), although a trend was observed in favor of KRAS and BRAF patients (Supplementary Figures B–D, see ESM).

In an exploratory analysis, median PFS did not appear to differ between the EGFR patients and those harboring rearrangements (5.6 months [95% CI 0.82–15.35] and 4.7 months [2.6–5.6], respectively). Among the BRAF-mutated patients, median PFS appeared to be higher in case of V600E mutations as compared with non-V600E mutations (20.3 months [95% CI 2.6–NR] vs 8.4 months [0.7–NR]).

Among the patients with PFS ≥ 12 months, two had EGFR mutations (including 1 L858R mutation) (out of eight), two had ALK rearrangements (out of five), four had BRAF V600E mutations (out of 11), 21 had KRAS mutations (ouf of 80), and none had a MET alteration or HER2 mutation.

Regarding the other histo-molecular markers, median PFS was 3.0 months (95% CI 2.5–5.8) for STK11-mutated patients (Fig. 2A), 6.6 months (95% CI 6.0–7.8) for TP53-mutated patients (Supplementary Figure E1, see ESM), 6.9 months (95% CI 4.6–9.9) for patients with MET expression ≥ 2 (Fig. 3A) and 8.6 months (95% CI 6.9–11.0) for patients with PDL1 TPS ≥ 1%. After adjustment for age, PS, smoking status, disease stage, number of previous lines, and STK11, TP53, and PDL1 status, PFS was significantly shorter for patients with STK11 mutations (HR 1.59 [95% CI 1.01–2.51]; p = 0.049), or with an advanced stage (IV-B) (p = 0.009) and significantly longer for patients with PDL1 ≥ 1% (HR 0.62 [95% CI 0.42–0.91]; p = 0.014).

Kaplan-Meier curves of progression-free survival and overall survival according to SK11 mutational status in the overall population (A, B) and in the KRAS sub-group (C, D). Shown are Kaplan–Meier estimates of progression-free survival and overall survival in patients with inactivating STK11 mutations and in patients without inactivating STK11 mutation in the overall population (A, B) and in KRAS-mutated patients (C, D). Tick marks indicate censoring of data. HR hazard ratio, KRASm KRAS mutation, KRASwt KRAS wild-type, mOS median overall survival, mPFS median progression-free survival, STK11m STK11 mutation, STK11wt STK11 wild-type, 95% CI 95% confidence interval

Kaplan-Meier curves of progression-free survival (A) and overall survival (B) according to MET expression score. Shown are Kaplan–Meier estimates of progression-free survival and overall survival in patients harboring MET overexpression (immunohistochemistry staining scores of 2+ or 3+) and in patients without MET overexpression (immunohistochemistry staining scores of 0+ or 1+). Tick marks indicate censoring of data. HR hazard ratio, mOS median overall survival, mPFS median progression-free survival, NR not reached, 95% CI 95% confidence interval

In univariate analysis, the negative impact of STK11 mutations was stronger among KRAS-mutated patients (mPFS of 2.6 months [95% CI 1.4–10.0] for KRAS patients with a SKT11 co-mutation versus 7.5 months [6.5–9.8] for KRAS patients without a STK11 co-mutation; HR 2.38 [95% CI 1.34–4.25]; p = 0.002) (Fig. 2C).

TP53 mutations were not associated with PFS in either the KRAS subgroup (p = 0.30) or the EGFR/ALK/ROS1/RET subgroup (p = 0.88).

Expression of PDL1 ≥ 50% in the EGFR/ALK/ROS1/RET subgroup seemed associated with longer PFS (mPFS 6.4 months [95% CI 3.2–NR] for PDL1 TPS ≥ 50% vs 3.2 months [0.82–NR] for PDL1 TPS < 50%).

Regarding MET expression, which is available for 72 patients (37%), MET overexpression ≥ 2+ was significantly associated with longer PFS in univariate analyses (HR 0.59 [95% CI 0.35–0.99]; p = 0.043) (Fig. 3A).

Median PFS was 5.26 months (95% CI 2.6–9.2) in never-smokers or lighter smokers (≤ 5 pack-years) and 6.47 months (95% CI 5.3–7.9) in ever-smokers (> 5 pack-years); HR 1.18 (95% CI 0.75–1.85); p = 0.30, in univariate analysis (Supplementary Figure F, see ESM).

The ORR was 46% in the whole population, ranging from 17% for the MET-dysregulated subgroup to 55% for the BRAF subgroup (Fig. 4). The ORR was similar for the EGFR/ALK/ROS1/RET, KRAS, and no-driver subgroups (44%, 46%, and 48%, respectively), whereas it was only approximatively 20% in the case of HER2 mutations. Among the patients with BRAF mutations, contrasting response rates were observed, ranging from 83% for BRAF V600E NSCLC to 20% for BRAF non-V600E NSCLC.

Best response to chemotherapy plus immunotherapy according to RECIST 1.1 criteria in each oncogenic driver subgroup

During follow-up, a total of 101 deaths were recorded. The estimated proportions of patients who were alive at 6 months, 12 months, and 18 months were 78%, 62%, and 48%, respectively. Overall, the median OS (mOS) was 16.7 months (95% CI 13.0–20.9).

Regarding the driver subgroups, the median OS was 18.6 months (95% CI 12.16–NR) for KRAS, 13.0 months (95% CI 6.7–20.6) for EGFR/ALK/ROS1/RET, 21.8 months (95% CI 6.3–NR) for BRAF, 15.8 months (95% CI 1.8–NR) for MET, and 15.0 months (95% CI 3.6–NR) for HER2. For patients without any of these oncogenic drivers, mOS was 16.7 months (95% CI 10.0–20.9). After adjustment for prespecified confounders, OS did not differ between the KRAS, EGFR/ALK/ROS1/RET, BRAF, and no-driver subgroups (p = 0.74). Initial stage (IV-B) and age ≥ 70 years old were significantly associated with PFS (p = 0.002 and p = 0.039, respectively).

Regarding other histo-molecular markers, median OS was 9.9 months [95% CI 4.8–13.7] for STK11-mutated patients (Fig. 2B), 16.8 months [95% CI 13.0–21.1] for TP53-mutated patients (Supplementary Figure E2, see ESM), not reached [95% CI 15.0–NR] for patients with MET expression ≥ 2+ (Fig. 3B), and 26.1 months [95% CI 15.9–NR] for patients with PDL1 TPS ≥ 1%. After adjustment for age, PS, smoking status, disease stage, number of previous lines, and STK11, TP53, and PDL1 status, OS was significantly shorter for patients with STK11 mutations (HR 1.88 [95% CI 1.11–3.22]; p = 0.021) and with an advanced stage (IV-B) (p = 0.009) and significantly longer for patients aged < 70 years old (p = 0.039) or with PDL1 TPS ≥ 1% (HR 0.62 [95% CI 0.42–0.91]; p = 0.014).

Like its effect on PFS, the prognostic impact of STK11 was also more pronounced in the KRAS subgroup (mOS of 5.6 months [95% CI 3.2–NR] for KRAS patients with STK11 co-mutations vs 21.1 months [14.2–NR] for KRAS patients without STK11 co-mutations; HR 2.38 [95% CI 1.33–4.25]; p = 0.003) (Fig. 2D). TP53 mutations were not likewise associated with OS in the KRAS subgroup (p = 0.68).

In univariate analyses, OS was significantly longer in patients with MET overexpression ≥ 2+ than those without (HR 0.47 [95% CI 0.24–0.92]; p = 0.028) (Fig. 3B).