Predictors of overall survival in advanced biliary tract cancer in the phase 3 TOPAZ-1 study

INTRODUCTION

TOPAZ-1 (NCT03875235) was the first large, phase 3 study to show that combining durvalumab, an immune checkpoint inhibitor, with gemcitabine and cisplatin (GemCis) could improve overall survival (OS) versus GemCis alone, the prior standard of care (SoC), in advanced biliary tract cancers (aBTCs).¹ At the pre-planned interim analysis of TOPAZ-1 [primary analysis, data cut-off (DCO) August 11, 2021¹], durvalumab plus GemCis showed a significant improvement in OS versus placebo plus GemCis, with no additional toxicity observed. An updated analysis (6-month follow-up; DCO February 25, 2022) further confirmed the favorable benefit-risk profile.² Based on the findings from TOPAZ-1, durvalumab plus GemCis has become a SoC for aBTCs. However, the factors prognostic for OS in aBTCs, or those predictive of survival following immune checkpoint inhibitor treatment, remain poorly defined.³,⁴ Although several patient-related and disease-related factors have been identified as potentially prognostic for survival benefit in aBTCs, none have been definitively confirmed.⁵ This post-hoc analysis aimed to assess patient-related and disease-related factors, and evaluate their potential prognostic or predictive value, in participants with aBTCs treated with durvalumab plus GemCis versus placebo plus GemCis in TOPAZ-1.

METHODS

The design and eligibility criteria of TOPAZ-1 have been described previously¹ and are summarized in the Supplemental Digital Content, http://links.lww.com/HC9/C208. The DCO for this analysis was February 25, 2022. A list of potential prognostic factors for OS was compiled through a literature search of articles published between January 01, 2022, and March 14, 2024, and included pre-specified factors from the original TOPAZ-1 study. These factors were evaluated for their potential to predict OS benefit following treatment with durvalumab plus GemCis versus placebo plus GemCis, in participants of the TOPAZ-1 study.¹ Analyses were based on the full analysis set (FAS), with additional prognostic sensitivity analyses conducted using both the FAS and the complete case set (CCS; excluding participants with ≥1 missing clinical or laboratory measurement). Detailed methods are provided in the Supplemental Digital Content, http://links.lww.com/HC9/C208.

RESULTS

A total of 685 participants were randomized to durvalumab plus GemCis (n = 341) or placebo plus GemCis (n = 344) and comprised the FAS; the CCS comprised 205 and 211 participants, respectively. Baseline demographics, disease characteristics, clinical factors, and laboratory factors were generally balanced across treatment groups and data sets (Supplemental Table S1, http://links.lww.com/HC9/C208).

The OS HR for durvalumab plus GemCis versus placebo plus GemCis, adjusted for stratification factors only, was 0.76 (95% CI: 0.64–0.91)² in the FAS and 0.74 (95% CI: 0.59–0.93) in the CCS.

The following factors were identified as potentially prognostic for OS in the FAS: neutrophil-to-lymphocyte ratio (NLR), cancer antigen 19-9 (CA 19-9), disease status (initially unresectable vs. recurrent), primary tumor location, lactate dehydrogenase, albumin, carcinoembryonic antigen (CEA), and total bilirubin (Supplemental Table S2, http://links.lww.com/HC9/C208). After adjusting for these factors, the OS HR was 0.70 (95% CI: 0.56–0.88). The results of the sensitivity analyses based on the FAS and CCS are reported in Supplemental Table S2, http://links.lww.com/HC9/C208.

In the predictive analysis, OS HRs for durvalumab plus GemCis versus placebo plus GemCis were <1 across all factors assessed, and all interaction p values exceeded 0.05 (Figure 1). Results for the predictive analysis of the TOPAZ-1 stratification factors and pre-specified demographic and disease characteristics have been previously reported.¹

FIGURE 1.

Factors predictive of overall survival benefit with durvalumab plus gemcitabine and cisplatin in TOPAZ-1 in the full analysis set. ^†High expression = tumor area positivity <1%; low expression = tumor area positivity ≥1%. ^‡Cut-offs were selected based on You MS, et al. 2019.^{6 §}The cut-off used to split participants into subgroups was the median value within participants in the TOPAZ-1 study. HRs were estimated by an unstratified Cox proportional hazards model with treatment as the sole covariate. CIs were calculated using a profile likelihood approach. The p-value was derived from an interaction term in an unstratified Cox proportional hazards model with treatment, factor, and their interaction (treatment by factor) as covariates. Abbreviations: CA 19-9, cancer antigen 19-9; CEA, carcinoembryonic antigen; eCCA, extrahepatic cholangiocarcinoma; ECOG PS, Eastern Cooperative Oncology Group performance status; GBC, gallbladder cancer; iCCA, intrahepatic cholangiocarcinoma; LDH, lactate dehydrogenase; NLR, neutrophil (10⁹/L) to lymphocyte (10⁹/L) ratio; OS, overall survival; PD-L1, programmed cell death ligand-1.

DISCUSSION

Baseline factors and OS HRs for durvalumab plus GemCis versus placebo plus GemCis were consistent in the FAS and CCS.

The potential prognostic factors identified in this analysis were generally consistent with prior findings.⁶,⁷,⁸ None of the factors tested were found to be predictive for differential OS benefit following treatment with durvalumab plus GemCis. Previous literature has reported NLR ≥3, CEA ≥5 ng/mL, and CA 19-9 ≥500 U/mL as prognostic factors for OS in patients with gallbladder cancer treated with GemCis.⁶ This aligns with our findings using the Akaike information criterion (AIC) and the FAS. However, in AIC-based sensitivity analyses using the CCS, CEA was not prognostic. Missing data may have influenced these results, as ~10% of values were missing for CEA (9.5%), CA 19-9 (8.8%), and programmed cell death ligand-1 (10.7%), compared with <1.9% for other variables. As the CCS was a subset of the FAS, more data were available in the FAS, allowing for a more robust analysis.

The adjusted OS HRs remained consistent across models in this analysis and with prior analysis using stratification factors only,² supporting the robustness of the treatment effect observed with durvalumab plus GemCis. The adjusted HRs herein provide a more refined estimate of the true effect of durvalumab plus GemCis versus placebo plus GemCis on OS, by considering the confounding effect of certain prognostic factors when interpreting treatment effects, an analysis which aligns with regulatory guidance from the US Food and Drug Administration and European Medicines Agency guidelines.⁹,¹⁰

A strength of this analysis is the application of multiple statistical techniques to the same dataset. The AIC backward selection approach was the primary method for identifying prognostic factors due to its suitability for assessing variable etiologic effects. However, all statistical selection methods are exploratory and have limitations. AIC may be prone to overfitting by including irrelevant variables, especially in larger samples, as it penalizes model complexity less heavily than the Bayesian information criterion. Nonetheless, consistency across sensitivity analyses using AIC and the CCS supports the validity of the findings. A limitation of this analysis is the focus on clinical data rather than translational or biomarker data. Prognostic and predictive factors were selected from the literature and pre-specified in TOPAZ-1, potentially overlooking molecular or biologic markers with prognostic value.

Overall, this post-hoc analysis of the TOPAZ-1 dataset reinforces the strength of the OS benefit associated with durvalumab plus GemCis in participants with aBTCs. While several potentially prognostic factors were identified, some consistent with those highlighted in prior studies, none were predictive of treatment benefit. These results suggest that treatment with durvalumab plus GemCis may offer clinical benefit across all eligible patients with aBTCs, further supporting its role as a SoC treatment.

Supplementary Material

hc9-10-e0868-s001.docx ^{(96.1KB, docx)}

DATA AVAILABILITY STATEMENT

Data underlying the findings described in this manuscript may be obtained in accordance with AstraZeneca’s data sharing policy described at https://www.astrazenecaclinicaltrials.com/our-transparency-commitments/. Data for studies directly listed on Vivli can be requested through Vivli at www.vivli.org. Data for studies not listed on Vivli can be requested through Vivli at https://vivli.org/members/enquiries-about-studies-not-listed-on-the-vivli-platform/. AstraZeneca Vivli member page is also available, outlining further details: https://vivli.org/ourmember/astrazeneca/.

AUTHOR CONTRIBUTIONS

All authors meet all ICMJE criteria for authorship. All authors had access to all the data, were responsible for drafting, reviewing, and editing the manuscript, and had the final responsibility for the decision to submit for publication. Aiwu R. He and Do-Youn Oh had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the results. Access and verification of data: Aiwu R. He, Aleksandra Kuzko, Magdalena Żotkiewicz, Nana Rokutanda, and Do-Youn Oh. Conception and design of the study: Aiwu R. He, Magdalena Żotkiewicz, Nana Rokutanda, and Do-Youn Oh. Data acquisition: Aiwu R. He, Mohamed Bouattour, Vineet G. Gupta, Ludovic Evesque, David B. Zhen, Joon Oh Park, Aumkhae Sookprasert, Alejandro Salvatierra, Gina Vaccaro, Sang Cheul Oh, Sandra Anabel Ostoich, Taroh Satoh, and Do-Youn Oh. Analysis and interpretation of data: Aiwu R. He, Mohamed Bouattour, Vineet G. Gupta, Ludovic Evesque, David B. Zhen, Joon Oh Park, Aumkhae Sookprasert, Alejandro Salvatierra, Gina Vaccaro, Sang Cheul Oh, Sandra Anabel Ostoich, Taroh Satoh, Aleksandra Kuzko, Magdalena Żotkiewicz, Nana Rokutanda, and Do-Youn Oh. Study supervision: Do-Youn Oh.

FUNDING INFORMATION

This study was sponsored by AstraZeneca.

CONFLICTS OF INTEREST

Aiwu R. He reports speaker fees for AstraZeneca, Bristol Myers Squibb, and Eisai, and research grants from AstraZeneca, Genentech, and Merck. Mohamed Bouattour reports consultant fees from AbbVie, AstraZeneca, Bayer, Bristol Myers Squibb, Eisai, Ipsen, MSD, Roche, and Sirtex Medical. Vineet G. Gupta reports speaker fees from AstraZeneca and holds stock in RPG Life Sciences and Zydus Life Sciences. Ludovic Evesque reports travel support from Amgen, AstraZeneca, and Bristol Myers Squibb, consulting fees from AstraZeneca, Merck Serono, MSD, and Takeda, and funding from Servier. David B. Zhen reports advisory fees from Cornerstone Pharmaceuticals, Ipsen, Jazz Pharmaceuticals/Zymeworks, Inc., and QED Therapeutics, and research funding from AstraZeneca, Bayer, Bristol Myers Squibb, Cornerstone Pharmaceuticals, Daiichi Sankyo, Eli Lily, Ipsen, Legend Biotech, Merck, Roche/Genentech, and Seagen. Joon Oh Park reports acting on the advisory board for Adicet Bio, AstraZeneca, Bristol Myers Squibb (Celgene), ImmuneOncia, MediRama, MedPacto, Merck, Merck Serono, and Servier, travel support from Minneamrita Therapeutics LLC, and research support from ABL Bio, Bristol Myers Squibb (Celgene), Eutilex, MedPacto, and Servier. Aumkhae Sookprasert reports acting as an invited speaker for AstraZeneca, Bristol Myers Squibb, Eisai, Jansen, Novartis, Merck, Pfizer, and Roche. Alejandro Salvatierra reports non-financial work as a principal investigator for Fundación Ars Medica. Gina Vaccaro reports advisory board work for AstraZeneca and GlaxoSmithKline, speaker’s bureau for Incyte, and an advisory role for Servier. Sang Cheul Oh declares no conflicts. Sandra Anabel Ostoich reports acting as an invited speaker for AstraZeneca, Bristol Myers Squibb, Eisai, Jansen, Merck, Novartis, Pfizer, and Roche. Taroh Satoh reports acting as an invited speaker for Bristol Myers Squibb, Chugai Pharmaceutical, Daiichi Sankyo, Elli Lilly, Merck-Biopharm, Ono Pharmaceutical, Taiho, Takeda, and Yakult Honsha, advisory board for Bristol Myers Squibb, Daiichi Sankyo, Elli Lilly, Ono Pharmaceutical, Takara-Bio, and Yakult Honsha, and research support from Bristol Myers Squibb, Chugai Pharmaceutical, Daiichi Sankyo, Elli Lilly, Gilead, Hutch-Med, Merck-Biopharm, Ono Pharmaceutical, Parexel, Taiho, and Yakult Honsha. Aleksandra Kuzko, Magdalena Żotkiewicz, and Nana Rokutanda are employees of AstraZeneca. Do-Youn Oh reports consultancy/advisory board for AbbVie, ASLAN, Astellas, Arcus Biosciences, AstraZeneca, Basilea, Bayer, BeiGene, Bristol Myers Squibb (Celgene), Eutilex, Genentech/Roche, Halozyme, Idience, IQVIA, J-Pharma, LG Chem, Merck Serono, Mirati Therapeutics, Moderna, MSD, Novartis, Taiho, Turning Point, Yuhan, and Zymeworks, and research grants from Array, AstraZeneca, BeiGene, Eli Lilly, Handok, MSD, Novartis, and Servier.

ETHICS APPROVAL AND CONSENT TO PARTICIPATE

This study was conducted in accordance with the protocol, consensus ethical principles derived from international guidelines, including the Declaration of Helsinki and Council for International Organizations of Medical Sciences International Ethical Guidelines, applicable International Council for Harmonisation of Technical Requirements for Pharmaceuticals for Human Use Good Clinical Practice Guidelines, and applicable local laws and regulations. The protocol, protocol amendments, informed consent form, investigator brochure, and other relevant documents (eg, advertisements) were approved by an Institutional Review Board/Independent Ethics Committee. Approval reference numbers are shown in Supplemental List S1, http://links.lww.com/HC9/C208​​​​​. Participants or their legal representatives provided written informed consent before participation.