Neo-adjuvant pembrolizumab in stage IV high-grade serous ovarian cancer: the phase II Neo-Pembro trial

Abstract

While immune checkpoint inhibitors (ICIs) have revolutionized cancer treatment, their efficacy in high-grade serous ovarian cancer (HGSOC) remains limited. Some patients, however, achieve lasting responses, emphasizing the need to understand how tumor microenvironment and molecular characteristics influence ICI response. The phase 2 Neo-Pembro study (NCT03126812) included 33 untreated stage IV HGSOC patients, who were scheduled for 6 cycles of carboplatin-paclitaxel and interval cytoreductive surgery. Pembrolizumab (pembro) was added from cycle two and continued for one year. The primary objective was to assess intratumoral immune activation using multiplexed immunofluorescence and immune-related gene expression. Our findings show immune activation, evidenced by an increase in CD3 + , CD8 + , CD8 + /FOXP3+ ratio, TNF-α and interferon-γ signaling. Treatment was well-tolerated. We observed major pathologic responses in 9/33 patients (27%, 95%CI 14-46), with pathologic response strongly associated with immune activation and OS. At a median follow-up of 52.8 months, 8/9 major responders were alive, with 6 patients recurrence-free. In contrast, 4/24 minor responders survived, including one recurrence-free. ctDNA clearance was observed in all major responders and was associated with prolonged PFS and OS. PD-L1 expression and homologous recombination deficiency were predictive of major response and may serve as biomarkers, warranting further exploration. These results suggest major responders may benefit from neo-adjuvant pembro.

Fig. 1. Study flowchart and outcomes including radiologic and pathologic responses, progression-free survival, and overall survival after neoadjuvant pembrolizumab plus chemotherapy.

a Study flowchart. 33 previously untreated stage IV high-grade serous ovarian cancer patients received pembro after one cycle of standard neo-adjuvant carboplatin-paclitaxel. Pembro was added to subsequent chemo cycles and continued as monotherapy up to one year. Surgery was planned after three chemo cycles. ctDNA circulating tumor DNA, PBMC peripheral blood mononuclear cells, PARP Poly (ADP-ribose) polymerase. b Radiological best change (%) in target lesions during neoadjuvant treatment with carboplatin-paclitaxel plus pembrolizumab. Patients whose tumor burden was considered too extensive to achieve complete interval cytoreduction were re-evaluated after four to six cycles of neoadjuvant therapy. c Progression-free survival (PFS); median PFS was 14.2 months (95%CI 11.1-30.2); 3-year PFS was 19% (95%CI 8.8-40). d Overall survival (OS); median OS was 32.0 months (95%CI 21.5-not reached (NR)); 3-year OS was 46% (95%CI 31-68). e Kaplan–Meier of OS for major pathologic responders in the study cohort (green), major pathologic responders in the historical cohort (orange), minor pathologic responders in the study cohort (dark blue), and minor pathologic responders in the historical cohort (light blue). Source data are provided as a Source Data file.

Fig. 2. CONSORT flow diagram.

Flowchart of patients included in the Neo-Pembro trial. HGSOC high-grade serous ovarian cancer, irAE immune-related adverse event, pembro pembrolizumab.

Fig. 3. Comparison of intratumoral immune cell population dynamics in major and minor responders following one cycle of neo-adjuvant chemotherapy monotherapy and subsequent chemotherapy plus pembrolizumab using multiplexed immunofluorescence and immunohistochemistry.

Pre- to post treatment changes of density of immune cell populations in the tumor microenvironment assessed through multiplexed immunofluorescence. Baseline samples available in 9 major responders and 21 minor responders. Post-chemotherapy samples available in 7 major responders and 16 minor responders. Post-pembro plus chemotherapy samples available in 7 major responders and 18 minor responders. Boxplots represent the median, and 25th and 75th percentiles; the whiskers extend from the hinge to the largest value no further than 1.5× IQR from the hinge. Pre- to post pairwise statistical significance was tested using Wilcoxon’s signed-rank test; for differences between major and minor responders, the significance was tested using Wilcoxon’s rank-sum test. All statistical tests were two sided. No adjustments were made for multiple comparisons. Only significant p-values are shown. a CD3 + T cells. b CD3 + CD8 + T cells. c CD3 + FOXP3 + T cells. d CD8 + /FOXP3 + T cell ratio. e CD20 + B cells. f CD68+ macrophages. g CD8+PD-1+ T cells. h CD8+PD-1+ /total CD8+ T cells ratio. i tertiary lymphoid structures (TLS). Source data are provided as a Source Data file.

Fig. 4. Comparison of dynamics of immune-related gene expression and genomic analyses in major and minor responders.

Baseline samples were available in 7 major responders and 21 minor responders. Post-chemotherapy samples were available in 6 major responders and 10 minor responders. Surgery resection samples were available in 4 major responders and 18 minor responders. a–d single sample scoring of four genesets using singscore. (*) indicates p < 0.05 by two-sided paired Wilcoxon test, not corrected for multiple testing. P-values: a baseline vs surgical resection: 4.5 × 10^-5; b baseline vs surgical resection: 1.5 × 10^-5; post-induction chemo vs surgical resection: 0.016. e, f normalized expression level of PD-1 and PD-L1 in each sample. (*) indicates p < 0.05 by two-sided paired Wilcoxon test, not corrected for multiple testing. P-values: e baseline vs surgical resection: 6.7 · 10^-3; post-induction chemo vs surgical resection: 0.031; f baseline vs surgical resection: 5.6 · 10^-3. g, h immune cell type abundance estimate obtained through the average expression of marker genes from Danaher et al.. (*) indicates p < 0.05 by two-sided paired Wilcoxon test, not corrected for multiple testing. P-values: g baseline vs post-induction chemo: 2.0 × 10^-3; baseline vs surgical resection: 0.045. i Mutational load as estimated by the number of somatic variants per million basepairs. Two-sided Wilcoxon test p = 0.075. j Homologous recombination deficiency signature score based on the number of large-scale state transitions in the genomic copy number profile as described in the study of Popova et al.. Only the baseline whole exome sequencing samples were used (major responders: n = 8 and minor responders: n = 20). Horizontal gray lines indicate ploidy-specific cutoffs indicative of homologous recombination deficiency. Points are colored by BRCA1 gene mutational status, where ‘pathogenic’ BRCA1 mutations are clinically established pathogenic variants, and ‘functional damaging’ indicates a BRCA1 variant that has been found to be damaging based on three functional tests, but has not yet been proven to be pathogenic and is clinically a variant of uncertain significance. No BRCA2 mutations were found in this cohort. Source data are provided as a Source Data file.

Fig. 5. PD-L1 expression and association with pathological response and overall survival.

Pre- to post treatment changes of density of immune cell populations in the tumor microenvironment assessed through multiplexed immunofluorescence. Baseline samples available in 9 major responders and 20 minor responders. Post-chemotherapysamples available in 7 major responders and 13 minor responders. Boxplots represent the median, and 25th and 75th percentiles; the whiskers extend from the hinge to the largest value no further than 1.5× IQR from the hinge. Pre- to post pairwise statistical significance was tested using Wilcoxon’s signed-rank test; for differences between major and minor responders, the significance was tested using Wilcoxon’s rank-sum test. All statistical tests were two sided. No adjustments were made for multiple comparisons. Only significant p-values are shown. a Combined positive score. b Immune proportion score (%). c Tumor proportion score (%). d–f representing delta (Δ) expression values in minor pathologic responders (blue) and minor pathologic responders (green), showing post-chemo minus baseline. Source data are provided as a Source Data file.

Fig. 6. Comparison of dynamics of intratumoral immune cell populations and immune-related gene expression presented as delta (Δ) values in major and minor responders.

a–f Dynamics of cell density of immune cell populations in the tumor microenvironment assessed through multiplexed immunofluorescence represented as delta (Δ) expression values in minor pathologic responders (blue) and minor pathologic responders (green), showing post-chemo minus baseline (delta 1; minor, n = 15; major, n = 7), resection minus post-chemo (delta 2; minor, n = 12; major, n = 5), and resection minus baseline (delta 3; minor, n = 18; major, n = 5). Boxplots represent the median, and 25th and 75th percentiles; the whiskers extend from the hinge to the largest value no further than 1.5× IQR from the hinge. The difference between major and minor responders was tested using a two-sided Wilcoxon Rank-sum test. Only significant p-values are shown. a CD3 + T cells. b CD3 + CD8 + T cells. c CD3 + FOXP3 + T cells. d CD8 + /FOXP3 + T cell ratio. e CD20 + B cells. f CD68+ macrophages. Source data are provided as a Source Data file.

Fig. 7. Circulating tumor DNA dynamics during neo-adjuvant treatment.

29 patients were included in the ctDNA analyses. a Relative change in ctDNA level of major responders (green) and minor responders (blue) after one cycle of neo-adjuvant chemotherapy (t₁) and after one subsequent cycle of pembro-chemotherapy (t₂) compared to baseline (t₀). b progression-free survival (PFS) and c overall survival (OS) were similar for patients with a decrease in ctDNA of >30% between t₀ and t₂ (orange) compared to patients without a >30% decrease (light blue). Both PFS (two-sided log-rank p = 0.08) and OS (two-sided log-rank p = 0.35) were considered not significantly different. d Number of cases that display a clearance (light blue) or no clearance (orange) of ctDNA pre-surgery among the minor responders and major responders. Improved e PFS (two-sided log-rank p < 0.001) and f OS (two-sided log-rank p < 0.01) was observed for patients with a ctDNA clearance (light blue) compared to patients without a clearance (orange). Source data are provided as a Source Data file.