Patients with extensive-stage small cell lung cancer whose disease harbors inflamed or YAP1 molecular subtypes may be more likely to derive superior overall survival benefit from durvalumab and chemotherapy vs those with other overexpressed biomarkers.
Patients with extensive-stage small cell lung cancer (ES-SCLC) whose disease harbors Inflamed T-cell or YAP1 molecular subtypes may be more likely to derive a superior overall survival (OS) benefit from treatment with durvalumab (Imfinzi) and chemotherapy compared with those who have other overexpressed biomarkers, according to an exploratory analysis of the phase 3 CASPIAN study (NCT03043872) that was recently presented at the 2022 AACR Annual Meeting.1
An analysis of 2 different RNA sequencing-based SCLC subtyping methods applied within the context of the CASPIAN data demonstrated that patients with these subtypes experienced the longest OS following treatment with the immunotherapy combination compared with patients expressing POU2F3, NEUROD1, and ASCL1.
Overall, the highest median OS were observed in patients with YAP1 or T-cell inflamed signature expression who received durvalumab/EP. In these patients, the median OS were 17.3 months (95% CI, 12.8-NE) and 17.6 months (95% CI, 11.4-NE), respectively.
These data are consistent with previous findings, although further research is warranted to determine the relationship between these biomarkers and immunotherapy response, the study authors noted.
“When the 2 RNA-based subtyping methods were applied to the CASPIAN data, we found that the 2 methods were consistent in differentiating neuroendocrine from non-neuroendocrine subtypes, and showed highest concordance within the non-neuroendocrine subtypes,” lead study author Yashaswi Shrestha, PhD, director of Oncology Translational Medicine and clinical biomarkers at AstraZeneca, said in a presentation of the findings. “Among the 4 subtypes, the inflamed subtype determined by Gay et al and the YAP1 subtype [determined] by Rudin et al showed the longest overall survival in the durvalumab/[chemotherapy] arm, suggesting that [this] subgroup of patients that have an inflamed phenotype are primed to respond to immunotherapy.”
Investigators on the phase 3 CASPIAN study enrolled 805 patients with treatment-naïve ES-SCLC and a World Health Organization performance status of either 0 or 1, a life expectancy of 12 weeks or greater, and measurable disease per RECIST v1.1 criteria. Patients with asymptomatic or treated and stable brain metastases were permitted to enroll.2
The enrolled patients were randomized 1:1:1 to undergo either durvalumab plus etoposide with either carboplatin or cisplatin (EP) for 4 cycles followed by single-agent durvalumab until disease progression, EP for up to 6 cycles followed by optional prophylactic cranial irradiation, or tremelimumab added to durvalumab plus EP for 4 cycles, followed by durvalumab until disease progression.
EP was administered in a dose ranging from 80 to 100 mg/m2 of etoposide along with either area under the curve 5-6 of carboplatin or 75 to 80 mg/m2 of cisplatin. Durvalumab was administered at 1500 mg and tremelimumab was administered at 75 mg.
The coprimary end points of CASPIAN were OS; progression-free survival, overall response rate, safety and tolerability, and patient reported outcomes (PROs) represented secondary end points. Biomarker analysis (tTMB; PD-L1) was an exploratory outcome measure.
Prior findings from CASPIAN revealed that durvalumab/EP significantly improved OS vs EP alone in this patient population, a benefit that was sustained for over 3 years of median follow-up. In the intent-to-treat population, the median OS was 12.9 months (95% CI, 11.3-14.7) with durvalumab/EP and 10.5 months (95% CI, 9.3-11.2) with EP alone (HR, 0.71; 95% CI, 0.60-0.86; P = .0003).
These results established durvalumab/EP as the global standard of care in ES-SCLC.2 The combination is FDA approved for the first-line treatment of patients with ES-SCLC.
Notably, the association between predictive biomarkers and improved outcomes with immune checkpoint inhibitors are still not well understood for patients with SCLC. However, investigators recently identified 4 SCLC molecular subtypes with distinct therapeutic vulnerabilities using 2 different differential gene expression methods.3,4
Investigators in the exploratory analysis therefore evaluated the concordance between the 2 methods and their link to improved OS using RNA sequencing data from the phase 3 CASPIAN study.
Formalin-fixed paraffin-embedded (FFPE) tumor samples previously collected at baseline were used to create whole-transcriptome RNA sequencing data.
Overall, the biomarker-evaluable population (BEP) consisted of 57 out of 268 patients on durvalumab/EP and 47 out of 269 patients on EP monotherapy arm.
For the BEP population, the data cutoff date was March 2021, and the median number of months follow-up was 39.4 months (range, 0.1–47.5). In the BEP population, the median OS was 11.8 months (95% CI, 8.6-14.9) with durvalumab/EP and 9.1 months (95% CI, 7.7-11.3) with EP alone (HR, 0.61; 95% CI, 0.40-0.92).
The Rudin et al. 2019 SCLC subtyping method3 found that YAP1 was commonly expressed in all subtypes, though at a relatively low level. Therefore, a sample was labeled as YAP1 if the YAP1 expression was greater than ASCL1, NEUROD1, and PUO2F3. Among the 4 transcription factors determined to be a subtype of the sample, ASCL1 and NEUROD1 were classified as neuroendocrine while POU2F3 and YAP1 lacked neuroendocrine markers.
The Gay et al. 2021 SCLC subtyping method4 used a 1300 gene signature to arrange the subtypes into 4 hierarchical clusters, each of which were assigned a subtype based on the highest expressed gene of ASCL1, NUEROD1, and POU2F3. The fourth cluster, which had a low expression of all 3 of the gene signatures was designated as the inflamed subtype because of its inflamed gene signature.
The T-cell inflamed signature was identified through assessment of RNA baseline tumor samples from pembrolizumab (Keytruda)-treated patients. The 18-gene T-cell inflamed gene expression signature comprised IFN-γ–responsive genes related to antigen presentation, chemokine expression, cytotoxic activity, and adaptive immune resistance.
The methods revealed high agreement in distinguishing neuroendocrine tumors from non-neuroendocrine subtypes. The highest levels of concordance were identified between non-neuroendocrine subtypes: the prevalence of Rudin YAP1 was 8% and the prevalence of Gay inflamed was 11%.
An OS analysis of the Rudin et al method revealed that, among patients treated with durvalumab/EP, the median OS with ASCL1 markers was 11.5 months, NEUROD1 was 9.5 months, POU2F3 was 4.8 months, and YAP1 was 17.3 months. In patients treated with EP alone, the median OS with ASCL1 markers was 10.7 months, NEUROD1 was 7.1 months, POU2F3 was 6.1 months, and YAP1 was 6.9 months.
An analysis of the Gay et al. method showed that, among patients treated with durvalumab/EP, the median OS with ASCL1 markers was 9.5 months, NEUROD1 was 14.6 months, POU2F3 was 6.8 months, and YAP1 was 17.6 months. In patients treated with EP alone, the median OS with ASCL1 was 8.3 months, NEUROD1 was 10.5 months, POU2F3 was 7.5 months, and YAP1 was 11.3 months.
Specifically, patients with high T-cell inflamed signature had numerically longer OS when receiving durvalumab/EP treatment than with EP alone. Patients in the former arm had a median top quartile OS of 15.8 months (10.4-NE), and a low median OS of 11.3 months (7.3-14.6; HR, 0.64; 95% CI, 0.31-1.32). Patients receiving EP alone had a median top quartile OS of 9.4 months (95% CI, 7.3-15.9), and a low median OS of 8.3 month (95% CI, 6.1-11.4), leading to a 13% reduction in the risk of disease progression or death (HR, 0.87; 95% CI, 0.45-1.66).
“This finding is consistent with previous observations,” Shrestha concluded. “[However], further analysis is necessary for us to gain confidence on the response and resistance biomarkers for immunotherapy and small cell lung cancer.”
Because of the limited sample size of eligible RNA sequencing data, these findings should be considered hypothesis generating, Shrestha noted.