Exploratory biomarker analyses were conducted in 190 patients—98 of 229 (43%) receiving nivolumab and 92 of 232 (40%) receiving placebo—with evaluable paired tumour samples collected at screening and blood samples obtained at one or more study time points for whole-exome sequencing (WES). These patients comprised the biomarker-evaluable population. The most frequent reasons for exclusion were failure to meet pathology assessment requirements, unavailable assays and unsuccessful WES results (Extended Data Fig. 1b). Baseline characteristics were generally well balanced between treatment groups and were comparable to those of the overall randomized population (Extended Data Table 1).
ctDNA clearance, minimal residual disease and pathological complete response
In this exploratory biomarker analysis, circulating tumour DNA (ctDNA) was evaluated to measure molecular response during neoadjuvant treatment and minimal residual disease (MRD) during adjuvant treatment. All 190 patients in the biomarker-evaluable population had ctDNA-assessable samples from at least one study time point (Fig. 1a).
Before neoadjuvant treatment began, ctDNA was detectable in 83 of 98 patients (85%) in the nivolumab group and 75 of 92 patients (82%) in the placebo group. No ctDNA was detected in 6 patients (6%) receiving nivolumab and 12 patients (13%) receiving placebo. Baseline characteristics for patients with and without detectable ctDNA are provided in Supplementary Table 1.
At the completion of neoadjuvant treatment, ctDNA was evaluable in 90 of 98 patients (92%) in the nivolumab arm and 78 of 92 patients (85%) in the placebo arm. The largest reduction in ctDNA levels occurred between the start and end of neoadjuvant treatment in both treatment groups (Fig. 2a,b).
Overall, 140 patients had detectable and evaluable ctDNA both before neoadjuvant treatment and at treatment completion: 76 of 98 patients (78%) in the nivolumab group and 64 of 92 patients (70%) in the placebo group. Among these patients, 63 of 76 (83%) in the nivolumab arm and 60 of 64 (94%) in the placebo arm completed all four neoadjuvant treatment cycles (Supplementary Table 2).
ctDNA clearance at the end of neoadjuvant treatment occurred in 50 of 76 patients (66%) receiving nivolumab compared with 24 of 64 patients (38%) receiving placebo (Fig. 2c). Of these patients, 42 of 50 (84%) in the nivolumab group and 22 of 24 (92%) in the placebo group completed all four neoadjuvant treatment cycles. Baseline characteristics were generally comparable across ctDNA clearance subgroups in both treatment arms (Supplementary Table 3). Higher pretreatment ctDNA levels appeared to be associated with more advanced disease stage and greater tumour burden (Supplementary Fig. 1a).
a, Patient-level ctDNA distributions during neoadjuvant and adjuvant treatment in the biomarker-evaluable population. b, ctDNA clearance and MRD status according to ctDNA levels. c, Relationship between ctDNA clearance and pCR status. d, Landmark event-free survival (EFS) from definitive surgery in patients who were MRD-negative before adjuvant C1D1 treatment. e, Baseline characteristics, treatment exposure, clinical outcomes and recurrence patterns in patients who became MRD-positive during adjuvant treatment (n = 13). In a, each dot represents one patient; the centre line of each box represents the median, while the upper and lower borders represent the 75th and 25th percentiles. The whiskers extend 1.5 times the interpercentile range from the upper and lower box limits. Points outside the boxes and whiskers, including minimum and maximum values, are shown as outliers. In a and b, 10−6 represents a ctDNA level of zero. In b and c, ctDNA clearance subgroups were defined according to clearance at the completion of neoadjuvant treatment. In b, MRD subgroups were based on the last available MRD assessment during adjuvant treatment; dashed lines distinguish the neoadjuvant and adjuvant periods. Patients with an unevaluable change in MRD status were MRD-negative before adjuvant treatment or lacked evaluable MRD at one or more additional adjuvant time points. In c, purple text identifies nivolumab-treated patients with ctDNA clearance and pCR. In b and e, patients with unevaluable ctDNA clearance had no detectable baseline ctDNA and/or lacked an evaluable ctDNA result at neoadjuvant treatment completion. In d, the HR and two-sided 95% CI were estimated using an unstratified Cox proportional-hazards model with treatment arm as the sole covariate. In e, each bar represents one patient. CL, clearance; R0, no residual tumour; R1, microscopic residual tumour; R2, macroscopic residual tumour.
Among patients who achieved ctDNA clearance at the end of neoadjuvant treatment, pCR occurred in 25 of 50 patients (50%) in the nivolumab arm and 3 of 24 patients (12%) in the placebo arm. These results corresponded to positive predictive values of 50% and 12%, respectively, with positive likelihood ratios of 2.04 and 2.14 (Supplementary Table 4). Among patients without ctDNA clearance, pCR occurred in none of 25 patients (0%) receiving nivolumab and in 1 of 40 patients (2%) receiving placebo. The corresponding negative predictive values were 100% and 98%, with negative likelihood ratios of 0 and 0.38, respectively.
Among patients with evaluable ctDNA clearance and percentage residual viable tumour (%RVT), the median %RVT among patients with ctDNA clearance was 0% in the nivolumab group (n = 42) and 33% in the placebo group (n = 19). In patients without ctDNA clearance, median %RVT was 50% with nivolumab (n = 17) and 70% with placebo (n = 25) (Supplementary Fig. 1b).
In the nivolumab arm, among patients with ctDNA clearance, 29 (69%) had 0–5% residual viable tumour in the primary tumour (RVT-PT), 9 (21%) had more than 5–80% RVT-PT and 4 (10%) had more than 80% RVT-PT (Supplementary Table 5). In the placebo arm, the corresponding values were 6 (32%), 9 (47%) and 4 (21%). Among patients without ctDNA clearance, 1 patient (6%) in the nivolumab group had 0–5% RVT-PT, 13 (76%) had more than 5–80% and 3 (18%) had more than 80%. In the placebo group, these proportions were 1 (4%), 15 (60%) and 9 (36%), respectively.
In the biomarker-evaluable population, 98 patients—49 of 98 (50%) receiving nivolumab and 49 of 92 (53%) receiving placebo—had evaluable MRD status after surgery and before adjuvant treatment, as well as at least one additional time point during adjuvant treatment. Before adjuvant treatment began, 48 of 49 patients (98%) in the nivolumab arm and 44 of 49 patients (90%) in the placebo arm were MRD-negative. The remaining patients were MRD-positive: 1 of 49 (2%) in the nivolumab group and 5 of 49 (10%) in the placebo group. None of these MRD-positive patients subsequently became MRD-negative during adjuvant treatment.
Among patients who were MRD-negative before adjuvant treatment, 4 of 48 (8%) in the nivolumab arm and 9 of 44 (20%) in the placebo arm became MRD-positive during adjuvant treatment. Baseline characteristics according to MRD status are reported in Supplementary Table 3.
Of the four nivolumab-treated patients who became MRD-positive during adjuvant treatment, one had achieved ctDNA clearance at neoadjuvant treatment completion, one had not achieved ctDNA clearance and two had unevaluable ctDNA clearance status. Of the nine placebo-treated patients who became MRD-positive, one had ctDNA clearance, five had no ctDNA clearance and three had unevaluable ctDNA clearance status.
Among patients who were MRD-positive before adjuvant treatment, including those assessed regardless of whether they underwent surgery, none became MRD-negative during adjuvant treatment. Disease recurrence occurred in 3 of 4 patients (75%) in the nivolumab arm and all 8 patients (100%) in the placebo arm. Figure 1b presents Sankey plots for patients with evaluable ctDNA dynamics, pCR status and recurrence outcomes (nivolumab, n = 46; placebo, n = 44). In this subgroup, 1 patient receiving nivolumab and 5 receiving placebo were MRD-positive after surgery and before adjuvant treatment. Recurrence rates were 22% and 45%, respectively.
Event-free survival according to ctDNA dynamics and pCR status
Among patients who underwent definitive surgery without achieving pCR, the landmark EFS hazard ratio (HR) for nivolumab versus placebo was 0.87 (95% CI, 0.51–1.47). An HR was not calculated for patients with pCR because of the limited sample size (nivolumab, n = 32; placebo, n = 5; Extended Data Fig. 2a).
Consistent with the analysis of all randomized patients, EFS was longer with nivolumab than with placebo in the biomarker-evaluable population. Median EFS was 40.1 months (95% CI, 28.4 to not reached (NR)) with nivolumab versus 15.8 months (95% CI, 10.0–35.1) with placebo, corresponding to an HR of 0.65 (95% CI, 0.43–0.98) (Extended Data Fig. 3).
Among patients with detectable ctDNA before neoadjuvant treatment, the EFS HR for nivolumab versus placebo was 0.58 (95% CI, 0.37–0.92; Extended Data Fig. 2b). An HR was not calculated for patients without detectable baseline ctDNA because of the small sample size (nivolumab, n = 12; placebo, n = 6). The EFS HR was 0.48 (95% CI, 0.22–1.02) among patients with ctDNA clearance before surgery and 0.76 (95% CI, 0.40–1.46) among those without ctDNA clearance (Extended Data Fig. 2c).
In a combined biomarker analysis of ctDNA clearance before surgery and pCR, nivolumab-treated patients with both ctDNA clearance and pCR (n = 25) had longer EFS than patients with ctDNA clearance but no pCR (n = 25; HR, 0.29; 95% CI, 0.10–0.85) and patients without ctDNA clearance or pCR (n = 26; HR, 0.23; 95% CI, 0.08–0.65).
Among nivolumab-treated patients without pCR, the EFS HR for ctDNA clearance versus no clearance was 0.70 (95% CI, 0.31–1.59; Extended Data Fig. 4a). In the placebo arm, comparisons involving patients with both ctDNA clearance and pCR were not performed because only three patients met these criteria. The EFS HR for placebo-treated patients with ctDNA clearance and no pCR (n = 21) versus those without ctDNA clearance and no pCR (n = 39) was 0.77 (95% CI, 0.39–1.54; Extended Data Fig. 4b).
Among patients who were MRD-negative after surgery and before adjuvant treatment, the landmark EFS HR from definitive surgery for nivolumab versus placebo was 0.75 (95% CI, 0.40–1.42; Fig. 2d). An HR was not calculated for patients who were MRD-positive at this time point because of the limited sample size (nivolumab, n = 3; placebo, n = 7). All 13 patients who were initially MRD-negative and subsequently became MRD-positive during adjuvant treatment experienced disease recurrence (Fig. 2e).
Tumour genomic alterations and event-free survival
An additional exploratory analysis examined EFS according to tumour alteration status for KRAS, KEAP1, STK11, SMARCA4, TP53 and CDKN2A. The frequency of selected tumour genomic alterations was generally similar between treatment groups. However, KRAS mutations were numerically less frequent and TP53 mutations numerically more frequent in the nivolumab arm than in the placebo arm (Supplementary Table 6).
KRAS, KEAP1 and STK11 mutations were primarily observed in patients with non-squamous NSCLC, whereas TP53 mutations were common across tumour histologies. CDKN2A mutations were more commonly observed in squamous NSCLC, while CDKN2A copy-number loss occurred at similar frequencies regardless of histology.
Across all 190 biomarker-evaluable patients, the most common individual alteration was a TP53 mutation, identified in 48 patients (25%). The most frequent co-alteration was a TP53 mutation with a CDKN2A alteration, observed in 38 patients (20%) in both treatment groups. Overall, 67 patients (35%) had KRAS, KEAP1 and/or STK11 mutations, including 3 (2%) with KRAS/KEAP1 co-mutations, 7 (4%) with KRAS/STK11 co-mutations and 3 (2%) with co-mutations in all three genes. In addition, 22 patients (12%) had KRAS/TP53 co-mutations, and 31 (16%) had TP53 mutations with KEAP1 and/or STK11 co-mutations (Fig. 3a). Surgical outcomes according to tumour genomic alteration status are reported in Supplementary Table 7.
a, Frequency of KRAS, KEAP1, STK11, TP53 and SMARCA4 tumour mutations and CDKN2A alterations in biomarker-evaluable patients. Tumour genomic alteration status was determined using WES of pretreatment tumour samples collected at screening. CDKN2A alteration included mutation and/or homozygous copy-number loss. b–i, EFS according to tumour genomic status: KRAS mutation (b); KRAS wild type (c); KEAP1 mutation (d); KEAP1 wild type (e); STK11 mutation (f); STK11 wild type (g); TP53 mutation (h); or TP53 wild type (i). Line charts use the same colour scheme as the bar charts. HRs and two-sided 95% CIs were estimated with an unstratified Cox proportional-hazards model using treatment arm as the sole covariate. The 95% CIs for 30-month EFS rates were 10–57% with nivolumab and 18–60% with placebo (b); 51–73% and 30–54% (c); 23–75% and 8–59% (d); 47–70% and 31–55% (e); 10–73% and 9–67% (f); 48–70% and 31–53% (g); 46–70% and 28–54% (h); and 29–74% and 26–62% (i), respectively.
EFS according to KRAS, KEAP1, STK11 and TP53 mutation status is shown in Fig. 3b–i. In the nivolumab group, pCR occurred in 7 patients (47%) with KRAS mutations, 4 (29%) with KEAP1 mutations, 4 (40%) with STK11 mutations and 27 (34%) with TP53 mutations. In the placebo group, pCR occurred in 1 patient (5%) with a KRAS mutation and in no patients with KEAP1 or STK11 mutations. Among patients with TP53 mutations in the placebo group, pCR occurred in 2 patients (3%). EFS according to CDKN2A mutation and/or copy-number loss and SMARCA4 alteration status is shown in Supplementary Fig. 2.
EFS appeared longer with nivolumab than with placebo among patients whose tumours had single or co-occurring alterations in KEAP1, STK11, CDKN2A and/or SMARCA4 (HR, 0.48; 95% CI, 0.28–0.83). Among patients without alterations in these four genes, the EFS HR was 0.90 (95% CI, 0.48–1.69; Extended Data Fig. 5a,b).
The univariate HRs for patients with alterations in at least one of these four genes versus those without alterations were 0.91 (95% CI, 0.50–1.69) with nivolumab and 1.77 (95% CI, 1.01–3.11) with placebo. After adjustment for smoking status, disease stage, tumour histology, tumour PD-L1 expression and tumour mutational burden (TMB), the multivariate HRs were 0.99 (95% CI, 0.53–1.86) with nivolumab and 1.82 (95% CI, 1.01–3.30) with placebo.
EFS also appeared longer with nivolumab than with placebo among patients with TP53 mutations but without KEAP1 or STK11 co-mutations (HR, 0.55; 95% CI, 0.32–0.95; Extended Data Fig. 5c,d). In patients with TP53 mutations and KEAP1 and/or STK11 co-mutations, the EFS HR was 0.44 (95% CI, 0.16–1.24; Extended Data Fig. 5e). Patient characteristics, tumour genomic profiles, treatment status and clinical outcomes for the complete biomarker-evaluable population are provided in Supplementary Fig. 3.
Among the 190 biomarker-evaluable patients, 98 (52% in the nivolumab group and 51% in the placebo group) had baseline TMB below 10 mutations per Mb. The remaining 92 patients—48% receiving nivolumab and 49% receiving placebo—had baseline TMB of at least 10 mutations per Mb. The EFS HR for nivolumab versus placebo was 0.65 (95% CI, 0.39–1.11) in patients with TMB below 10 mutations per Mb and 0.62 (95% CI, 0.32–1.20) in those with TMB of at least 10 mutations per Mb (Supplementary Fig. 4).
Predictive modelling of event-free survival
To identify potential predictors of EFS, investigators trained a random survival forest machine-learning model using data from 80% of biomarker-evaluable patients in both treatment groups. The remaining 20% were used as a test set to assess associations between clinical and genomic variables—including baseline demographic and disease characteristics, pCR, ctDNA clearance and tumour genomic alterations—and EFS.
The model trained on the full biomarker-evaluable dataset achieved a Harrell concordance index of 0.79, indicating a strong correlation between model-derived risk scores and the likelihood of an EFS event. The concordance index was 0.65 in the test dataset.
The factors most strongly associated with a lower risk of EFS events included ctDNA clearance before surgery, non-N2 NSCLC, pCR, squamous tumour histology and nivolumab treatment (Supplementary Fig. 5a). SMARCA4 mutation, CDKN2A alteration and KEAP1 mutation contributed comparatively less to EFS prediction in the model that included patients from both treatment groups.
In a model restricted to patients who received nivolumab, the strongest predictors of reduced EFS risk, in descending order, were pCR, high TMB, high tumour PD-L1 expression, non-N2 NSCLC and ctDNA clearance before surgery (Supplementary Fig. 5b).
EFS risk-score tertiles were calculated in the training population and applied to the test population. Kaplan–Meier curves showed clear separation between predicted high-, intermediate- and low-risk groups across all biomarker-evaluable patients. However, this separation was not observed among biomarker-evaluable patients in the nivolumab arm alone (Supplementary Fig. 5c,d).
Source: www.nature.com