Biomarker-directed targeted therapy plus durvalumab in advanced non-small-cell lung cancer: a phase 2 umbrella trial

Abstract

For patients with non-small-cell lung cancer (NSCLC) tumors without currently targetable molecular alterations, standard-of-care treatment is immunotherapy with anti-PD-(L)1 checkpoint inhibitors, alone or with platinum-doublet therapy. However, not all patients derive durable benefit and resistance to immune checkpoint blockade is common. Understanding mechanisms of resistance-which can include defects in DNA damage response and repair pathways, alterations or functional mutations in STK11/LKB1, alterations in antigen-presentation pathways, and immunosuppressive cellular subsets within the tumor microenvironment-and developing effective therapies to overcome them, remains an unmet need. Here the phase 2 umbrella HUDSON study evaluated rational combination regimens for advanced NSCLC following failure of anti-PD-(L)1-containing immunotherapy and platinum-doublet therapy. A total of 268 patients received durvalumab (anti-PD-L1 monoclonal antibody)-ceralasertib (ATR kinase inhibitor), durvalumab-olaparib (PARP inhibitor), durvalumab-danvatirsen (STAT3 antisense oligonucleotide) or durvalumab-oleclumab (anti-CD73 monoclonal antibody). Greatest clinical benefit was observed with durvalumab-ceralasertib; objective response rate (primary outcome) was 13.9% (11/79) versus 2.6% (5/189) with other regimens, pooled, median progression-free survival (secondary outcome) was 5.8 (80% confidence interval 4.6-7.4) versus 2.7 (1.8-2.8) months, and median overall survival (secondary outcome) was 17.4 (14.1-20.3) versus 9.4 (7.5-10.6) months. Benefit with durvalumab-ceralasertib was consistent across known immunotherapy-refractory subgroups. In ATM-altered patients hypothesized to harbor vulnerability to ATR inhibition, objective response rate was 26.1% (6/23) and median progression-free survival/median overall survival were 8.4/22.8 months. Durvalumab-ceralasertib safety/tolerability profile was manageable. Biomarker analyses suggested that anti-PD-L1/ATR inhibition induced immune changes that reinvigorated antitumor immunity. Durvalumab-ceralasertib is under further investigation in immunotherapy-refractory NSCLC.ClinicalTrials.gov identifier: NCT03334617.

Fig. 1. CONSORT diagram of patient screening and disposition in HUDSON.

Based on tumor molecular profiling, patients were assigned to either biomarker-matched (Group A) or non-matched cohorts (Group B), which included patients with primary and acquired resistance determined by their initial response to prior immunotherapy-containing regimens. Patients were then treated with one of four durvalumab-based combination regimens (Modules 1–4).

Fig. 2. PFS and OS with durvalumab–ceralasertib and with durvalumab plus olaparib, danvatirsen or oleclumab in HUDSON.

a,b, Kaplan–Meier analysis of PFS (a) and OS (b) among all patients who received durvalumab–ceralasertib or who received durvalumab plus olaparib, danvatirsen or oleclumab on HUDSON.

Fig. 3. OS and gene expression profiling in subgroups known to be associated with poor immunotherapy response among patients who received durvalumab plus olaparib, danvatirsen or oleclumab.

a–f, Kaplan–Meier distributions of OS (left) and TIS ssGSEA enrichment score (ES) distributions (violin plots, right) comparing patients with primary or acquired resistance (a), PD-L1-negative or PD-L1-positive status (b), squamous cell carcinoma or adenocarcinoma histology (c), with or without bone/liver metastases (d), high or low TMB (e) and STK11-altered or STK11-wild-type tumors (f). Median OS, 80% CIs and numbers of patients at risk are shown in each Kaplan–Meier plot. Violin plots show data from individual patients (dots), color-coded by site of sample, the median value (text box) and the probability density distribution smoothed using a kernel density estimator (shaded areas). ssGSEA ES distributions were compared using two-sided Wilcoxon rank-sum tests, with no correction for multiplicity of testing; associated P values and the number of tumor samples in each group are reported for each comparison.

Fig. 4. PFS and OS with durvalumab–ceralasertib and with durvalumab plus olaparib, danvatirsen or oleclumab in HUDSON in patient subgroups.

Figure shows medians and 80% CI per Kaplan–Meier methodology.

Fig. 5. Ceralasertib induces systemic immunomodulation that enhances the immune-mediated effects of durvalumab in patients with NSCLC with progression on prior anti-PD-(L)1 treatment.

a, Schema depicting treatment schedule for modules 1 (durvalumab–olaparib), 2 (durvalumab–danvatirsen) and 3 (durvalumab–ceralasertib), showing durvalumab administration (blue arrows), combination agent dosing (continuous colored bars), and pretreatment/on-treatment blood sample collection time points (stars). BID, twice daily. b–g, Effects of ceralasertib and durvalumab treatment on longitudinal blood-derived transcriptomes are depicted in the box plots showing differentially expressed monocyte-associated signatures (b), CD8 T-cell-associated signatures (c), dysfunctional/exhausted CD8 T-cell-associated signatures (d), TNF-α-associated signatures (e), interferon-γ-associated signatures (f) and interferon-α-associated signatures (g). P values are for comparisons with respective cycle 0, day (D) 1 time points. P values for visit dates represent linear mixed model effects, illustrating changes in on-treatment samples. h–q, Effects of ceralasertib, olaparib and danvatirsen with durvalumab on longitudinal blood-derived T cell repertoire: box plots showing productive Simpson T cell clonality with durvalumab–ceralasertib (h); waterfall plots showing changes in clonality on treatment by patient with durvalumab–ceralasertib on cycle 0, day 1 versus cycle 1, day 1 (i) and cycle 1, day 1 versus cycle 1, day 22 (j); box plots showing distributions of T cell clonality with durvalumab–olaparib (k), and waterfall plots showing changes in clonality on treatment by patient with durvalumab–olaparib on cycle 1, day 1 versus cycle 1, day 15 (l); box plots showing distributions of T cell clonality with durvalumab–danvatirsen (m), and waterfall plots showing changes in clonality on treatment per patient with durvalumab–danvatirsen on cycle 1, day 1 versus cycle 1, day 15 (n); box plots of expanded T cell clones (o), newly detected expanded T cell clones (p) and Morisita index for durvalumab plus olaparib, danvatirsen or ceralasertib at indicated visit dates (q). P values for visit dates represent two-sided Wilcoxon paired signed-rank tests between time points or two-sided Wilcoxon signed-rank tests between study modules, with correction for multiplicity of testing (Benjamini–Hochberg procedure), illustrating changes in on-treatment samples. r, Proposed immunomodulatory mechanism of action effects of ceralasertib with durvalumab in the periphery and tumor. For all box plots (b–h, k, m and o–q), box centerlines show medians, box limits show upper and lower quartiles, and whiskers show range.

Fig. 5. Ceralasertib induces systemic immunomodulation that enhances the immune-mediated effects of durvalumab in patients with NSCLC with progression on prior anti-PD-(L)1 treatment.

a, Schema depicting treatment schedule for modules 1 (durvalumab–olaparib), 2 (durvalumab–danvatirsen) and 3 (durvalumab–ceralasertib), showing durvalumab administration (blue arrows), combination agent dosing (continuous colored bars), and pretreatment/on-treatment blood sample collection time points (stars). BID, twice daily. b–g, Effects of ceralasertib and durvalumab treatment on longitudinal blood-derived transcriptomes are depicted in the box plots showing differentially expressed monocyte-associated signatures (b), CD8 T-cell-associated signatures (c), dysfunctional/exhausted CD8 T-cell-associated signatures (d), TNF-α-associated signatures (e), interferon-γ-associated signatures (f) and interferon-α-associated signatures (g). P values are for comparisons with respective cycle 0, day (D) 1 time points. P values for visit dates represent linear mixed model effects, illustrating changes in on-treatment samples. h–q, Effects of ceralasertib, olaparib and danvatirsen with durvalumab on longitudinal blood-derived T cell repertoire: box plots showing productive Simpson T cell clonality with durvalumab–ceralasertib (h); waterfall plots showing changes in clonality on treatment by patient with durvalumab–ceralasertib on cycle 0, day 1 versus cycle 1, day 1 (i) and cycle 1, day 1 versus cycle 1, day 22 (j); box plots showing distributions of T cell clonality with durvalumab–olaparib (k), and waterfall plots showing changes in clonality on treatment by patient with durvalumab–olaparib on cycle 1, day 1 versus cycle 1, day 15 (l); box plots showing distributions of T cell clonality with durvalumab–danvatirsen (m), and waterfall plots showing changes in clonality on treatment per patient with durvalumab–danvatirsen on cycle 1, day 1 versus cycle 1, day 15 (n); box plots of expanded T cell clones (o), newly detected expanded T cell clones (p) and Morisita index for durvalumab plus olaparib, danvatirsen or ceralasertib at indicated visit dates (q). P values for visit dates represent two-sided Wilcoxon paired signed-rank tests between time points or two-sided Wilcoxon signed-rank tests between study modules, with correction for multiplicity of testing (Benjamini–Hochberg procedure), illustrating changes in on-treatment samples. r, Proposed immunomodulatory mechanism of action effects of ceralasertib with durvalumab in the periphery and tumor. For all box plots (b–h, k, m and o–q), box centerlines show medians, box limits show upper and lower quartiles, and whiskers show range.

Extended Data Fig. 1. PFS and OS by treatment cohort.

PFS and OS with a, b, durvalumab-ceralasertib (median OS not yet mature in ATM biomarker-matched cohort), c, d, durvalumab-olaparib, e, f, durvalumab-danvatirsen, g, h, durvalumab-oleclumab.

Extended Data Fig. 2. Gene expression profiling in subgroups with known association with poor immunotherapy response among patients who received durvalumab plus olaparib, danvatirsen or oleclumab.

Gene-wise transcript-per-million (TPM) z-score heatmaps for the 18 genes comprising the tumour inflammation signature (TIS) in tumour samples from patients with a, primary or acquired resistance to prior immune checkpoint blockade therapy (n = 18, n = 31; overall prevalence: n = 81 [43.1%], n = 107 [56.9%]), b, PD-L1-positive or PD-L1-negative tumours (n = 16, n = 18; overall prevalence: n = 73 [58.9%], n = 51 [41.1%]), c, adenocarcinoma or squamous cell histology (n = 38, n = 6; overall prevalence: n = 130 [75.1%], n = 43 [24.9%]), d, with or without bone/liver metastases (n = 23, n = 26; overall prevalence: n = 76 [40.4%], n = 112 [59.6%]), e, high or low TMB (n = 16, n = 18; overall prevalence: n = 60 [41.1%], n = 86 [58.9%]) and f, STK11-altered or STK11-wild-type tumours (n = 5, n = 38; overall prevalence: n = 26 [14.9%], n = 148 [85.1%]). Tumour location and histology are annotated.

Extended Data Fig. 3. PFS and OS with durvalumab-ceralasertib and with durvalumab plus olaparib, danvatirsen or oleclumab in patient subgroups defined by presence or absence of adverse prognostic factors.

Kaplan–Meier analyses of (left) PFS and (right) OS in patients with a, primary resistance, b, acquired resistance, c, PD-L1-negative tumours, d, PD-L1-positive tumours, e, squamous histology, f, non-squamous histology, and g, STK11 mutations.

Extended Data Fig. 4. PFS and OS with durvalumab-ceralasertib in patient subgroups defined by presence or absence of adverse prognostic factors.

Kaplan–Meier analyses of (left) PFS and (right) OS with durvalumab-ceralasertib in patients with a, primary vs acquired resistance, b, PD-L1-negative vs -positive status, and c, squamous cell carcinoma vs non-squamous histology, and in patients d, with or without bone/liver metastases, e, with high or low tumour mutational burden, and according to f, STK11 status and g, KRAS status (variant or wild-type).