Introduction
Advanced biliary tract cancer (BTC)—encompassing intrahepatic cholangiocarcinoma (iCCA), extrahepatic cholangiocarcinoma (eCCA), and gallbladder cancer (GBC)—has historically carried a poor prognosis when treated with gemcitabine–cisplatin (GC) alone. Two landmark Phase III trials have now demonstrated that augmenting GC with a programmed death-1/programmed death-ligand 1 (PD-1/PD-L1) inhibitor confers a statistically significant overall survival (OS) advantage: TOPAZ-1 with the anti-PD-L1 antibody durvalumab, and KEYNOTE-966 with the anti-PD-1 antibody pembrolizumab 124. Critically, these trials were not conducted as a head-to-head comparison; all cross-trial inferences presented below are therefore indirect and must be interpreted with appropriate caution 15.
Trial Design and Patient Populations
TOPAZ-1 (NCT03875235) randomized 685 previously untreated patients with unresectable or metastatic BTC 1:1 to durvalumab 1,500 mg plus GC (gemcitabine 1,000 mg/m² and cisplatin 25 mg/m², every 3 weeks) for up to 8 cycles, followed by durvalumab maintenance until progression or unacceptable toxicity, versus placebo plus GC. Ampullary carcinoma was explicitly excluded. The trial was conducted at 105 centers across 17 countries 123.
KEYNOTE-966 (NCT04003636) randomized 1,069 patients to pembrolizumab 200 mg every 3 weeks (up to 35 cycles) plus GC versus placebo plus GC. Stratification factors included geographic region (Asia vs. non-Asia), disease stage (locally advanced vs. metastatic), and tumor origin (GBC vs. iCCA vs. eCCA). Eligibility required ECOG performance status 0–1 and measurable disease 146.
Both trials enrolled patients with BTC subtypes (iCCA, eCCA, GBC) and achieved broad geographic representation supporting generalizability across US, EU, and Asian practice settings. KEYNOTE-966 enrolled approximately 56% more patients, providing greater statistical power for subgroup analyses. Follow-up maturity also differs: TOPAZ-1 reports a 3-year OS update (median follow-up 41.3 months), while KEYNOTE-966 provides mature 24-month OS data and a subsequent 3-year follow-up update presented at ASCO 314.
Efficacy Outcomes
Both trials met their primary OS endpoint, establishing chemoimmunotherapy as the new first-line standard of care in advanced BTC.
In TOPAZ-1, the primary publication reported a median OS of 12.8 months with durvalumab plus GC versus 11.5 months with placebo plus GC (hazard ratio [HR] 0.80; 95% CI 0.66–0.97) 2. Updated analyses with greater follow-up reported a median OS of 12.9 versus 11.3 months (HR 0.74; 95% CI 0.63–0.87), with a 36-month OS rate of 14.6% versus 6.9% 13. This 3-year landmark represents a clinically meaningful doubling of the long-term survivor proportion. Among an "extended long-term survivor" (eLTS) subset (alive ≥30 months), 17.0% of patients received durvalumab plus GC versus 8.7% in the placebo arm 3. Progression-free survival (PFS) HR was 0.75 (95% CI 0.63–0.89), and the objective response rate (ORR) was 26.7% versus 18.7%, with a complete response rate of 2.1% versus 0.6% 2.
In KEYNOTE-966, pembrolizumab plus GC yielded a median OS of 12.7 months versus 10.9 months (HR 0.83–0.86 across reporting analyses; 95% CI 0.72–0.95 in the primary publication) 46. The 24-month OS rate was 24.6% versus 19.2%. Median PFS was 6.5 versus 5.6 months (HR 0.85; 95% CI 0.75–0.97), ORR was 28.7% in both arms, and duration of response (DoR) was 8.3 versus 6.9 months 14. In a China-specific subgroup analysis, the OS benefit was numerically more pronounced: median OS 14.1 months versus 9.9 months (HR 0.74), ORR 36.0% versus 28.9%, and DoR 10.2 versus 5.7 months, with a consistent safety profile 12.
Numerically, TOPAZ-1's mature OS hazard ratio (0.74) appears more favorable than KEYNOTE-966's (0.83–0.86). However, the absolute median OS values are nearly identical (~12.7–12.9 months in experimental arms), and differences in sample sizes, patient populations, follow-up maturity, and statistical analyses preclude any conclusion of superiority of one regimen over the other 15.
Safety and Tolerability
The safety profiles of both regimens were broadly consistent with expectations for PD-1/PD-L1 inhibition combined with cytotoxic chemotherapy in a treatment-naïve BTC population, and no new safety signals were identified in either trial's long-term follow-up.
In TOPAZ-1, any-grade adverse events (AEs) occurred in 99.4% of durvalumab-treated patients versus 98.8% with placebo. Grade 3–4 AEs were reported in 75.7% versus 77.8%. Immune-mediated AEs occurred in 12.7% (durvalumab) versus 4.7% (placebo), with grade 3–4 immune-mediated AEs in 2.4% versus 1.5%. Treatment discontinuation due to AEs was 13.0% versus 15.2%, and AE-related deaths occurred in 3.6% versus 4.1%. Serious AEs were reported in 47.3% versus 43.6% 2. The 3-year update confirmed sustained tolerability without new safety concerns 3.
In KEYNOTE-966, grade 3–5 treatment-related AEs (TRAEs) were reported in 71.5% (pembrolizumab plus GC) versus 69.3% (GC alone). Treatment-related deaths occurred in 5.9% (31/529) versus 9.4% (49/534) in the later update. Immune-mediated AEs occurred in 24.3% versus 8.5%, predominantly grade 1–2, and were managed with standard immunotherapy protocols. Quality of life was maintained in the pembrolizumab arm 16.
A critical methodological caveat: TOPAZ-1 reports all-cause grade 3/4 AEs, whereas KEYNOTE-966 reports grade 3–5 TRAEs. These differing definitions limit direct numerical comparison between the two trials and should be considered when counseling patients about relative toxicity profiles 1.
For both regimens, hematologic toxicities from GC (neutropenia, anemia, thrombocytopenia) remain a primary clinical concern requiring routine monitoring. Immune-mediated hepatotoxicity, in particular, warrants vigilance given that patients with BTC may have compromised baseline hepatic reserve; structured monitoring and early corticosteroid use for suspected immune hepatitis are essential components of safe administration 8.
Biomarker-Defined Subgroup Performance
The biomarker landscape in BTC remains an evolving and incompletely validated domain.
PD-L1 expression: In TOPAZ-1, PD-L1 tumor area positivity (TAP) was analyzed; TAP ≥1% was associated with OS HR 0.79 (95% CI 0.61–1.00), while TAP <1% showed HR 0.86 (95% CI 0.60–1.23), suggesting that PD-L1 status does not reliably demarcate responders versus non-responders in this population 2. In the TOPAZ-1 long-term survivor subset, PD-L1 TAP ≥1% was present in 63.6% of durvalumab long-term survivors versus 60.0% in the GC-alone arm, and ORR in long-term survivors was 44.3% versus 33.8%, suggesting a qualitative enrichment for response but not definitive biomarker predictability 11. Across 56 studies encompassing 7,768 BTC patients, pooled PD-L1 positivity is approximately 25–30%, with substantial heterogeneity driven by antibody clone choice (e.g., 5H1 yields ~58% positivity vs. SP142 at ~17%) 151617. This cross-assay variability critically undermines the utility of PD-L1 as a universal treatment-selection biomarker in BTC. A comparative analysis of PD-L1 immunohistochemistry assays confirmed that SP263 tends to yield higher positivity than 22C3 or E1L3N, complicating cross-trial comparisons 16.
MSI-H/dMMR and TMB: Mismatch repair deficiency (dMMR)/microsatellite instability-high (MSI-H) status and high tumor mutational burden (TMB-H) are relatively uncommon in BTC, limiting their practical utility as universal predictive biomarkers for first-line IO-chemotherapy combinations 517. Neither TOPAZ-1 nor KEYNOTE-966 was designed with a biomarker-enriched selection strategy. KEYNOTE-966's primary publication did not report MSI-H or TMB-defined subgroup outcomes.
Anatomical subtypes and viral hepatitis: BTC subtypes (iCCA, eCCA, GBC) harbor distinct molecular landscapes and immune microenvironments, which may modulate IO-chemotherapy benefit heterogeneously. TOPAZ-1 confirmed consistent OS benefit across primary tumor locations 1. KEYNOTE-966 used tumor origin as a stratification factor, and the trial demonstrated overall consistency of benefit 4. For hepatitis B virus (HBV)-positive patients—a significant consideration in Asia—KEYNOTE-966 showed comparable OS benefit in HBV-positive (HR 0.87) and HBV-negative (HR 0.85) subgroups, supporting the applicability of pembrolizumab-based therapy in HBV-endemic regions 15. Real-world data for durvalumab in a cohort of 563 patients (350 receiving durvalumab plus GC) corroborated the trial-level OS signal in routine care, with exploratory findings suggesting that neutrophil-to-lymphocyte ratio ≤3 and locally advanced disease were associated with improved outcomes 7.
Standardized Comparison Table
| Domain | TOPAZ-1 (Durvalumab + GC) | KEYNOTE-966 (Pembrolizumab + GC) |
|---|---|---|
| Phase / Design | Phase III, randomized, double-blind, placebo-controlled | Phase III, randomized, double-blind, placebo-controlled |
| Sample Size | N = 685 (1:1) | N = 1,069 (1:1) |
| BTC Subtypes | iCCA, eCCA, GBC (ampullary excluded) | iCCA, eCCA, GBC |
| Regimen | Durvalumab 1,500 mg + GC Q3W × ≤8 cycles; durvalumab maintenance | Pembrolizumab 200 mg Q3W (≤35 cycles) + GC |
| Median OS (exp. vs ctrl) | 12.9 vs 11.3 months | 12.7 vs 10.9 months |
| OS HR (95% CI) | 0.74 (0.63–0.87)^a | 0.83–0.86 (0.72–0.95 to 0.75–0.98)^b |
| 36-month OS rate | 14.6% vs 6.9% | Not reported at 36 months in retrieved materials |
| 24-month OS rate | Not reported at 24 months in retrieved materials | 24.6% vs 19.2% |
| PFS HR (95% CI) | 0.75 (0.63–0.89) | 0.85 (0.75–0.97) |
| Median PFS | 7.2 vs 5.7 months | 6.5 vs 5.6 months |
| ORR | 26.7% vs 18.7% | 28.7% vs 28.7% |
| DoR (exp. vs ctrl) | 6.4 vs 6.2 months | 8.3 vs 6.9 months |
| Grade 3–4 AEs | 75.7% vs 77.8% (all-cause)^c | 71.5% vs 69.3% (TRAEs, grade 3–5)^c |
| Immune-mediated AEs | 12.7% vs 4.7% | 24.3% vs 8.5% |
| Treatment-related deaths | 3.6% vs 4.1% | 5.9% vs 9.4% |
| Discontinuation due to AEs | 13.0% vs 15.2% | Not specified |
| PD-L1 biomarker | TAP ≥1%: HR 0.79; TAP <1%: HR 0.86 | Not robustly defined in retrieved data |
| HBV subgroup | Not reported in retrieved materials | HR 0.87 (HBV+) vs 0.85 (HBV–) |
| China subgroup (KEYNOTE-966) | N/A | Median OS 14.1 vs 9.9 months; HR 0.74; ORR 36.0% vs 28.9% |
| MSI-H/dMMR / TMB | Not validated as predictive | Not validated as predictive |
| Follow-up maturity | 3-year OS published (median follow-up 41.3 months) | 3-year follow-up update at ASCO (full publication pending) |
Footnotes: AE = adverse event; BTC = biliary tract cancer; ctrl = control; DoR = duration of response; eCCA = extrahepatic cholangiocarcinoma; exp. = experimental; GBC = gallbladder cancer; GC = gemcitabine–cisplatin; HBV = hepatitis B virus; HR = hazard ratio; iCCA = intrahepatic cholangiocarcinoma; MSI-H = microsatellite instability-high; dMMR = deficient mismatch repair; ORR = objective response rate; OS = overall survival; PFS = progression-free survival; Q3W = every 3 weeks; TAP = tumor area positivity; TMB = tumor mutational burden; TRAE = treatment-related adverse event. ^a Updated mature analysis. ^b Primary publication HR 0.83 (95% CI 0.72–0.95); updated analysis HR 0.86 (95% CI 0.75–0.98). ^c Methodological differences in AE reporting between trials preclude direct numerical comparison.
Clinical Interpretation and Treatment Selection
Both durvalumab plus GC and pembrolizumab plus GC have established a superior OS benefit over GC alone and represent evidence-based first-line standards for advanced BTC. Neither regimen has been shown to be definitively superior to the other in the absence of head-to-head trial data 15.
In the absence of validated predictive biomarkers, treatment selection should be individualized based on clinical and operational factors. Maintenance strategy differs meaningfully: TOPAZ-1 employs durvalumab maintenance after ≤8 GC cycles, while pembrolizumab in KEYNOTE-966 is administered for up to 35 cycles, with distinct implications for infusion scheduling, patient preference, and institutional logistics 1. Regional hepatitis context may favor pembrolizumab in HBV-endemic settings, where KEYNOTE-966 data explicitly document consistent benefit across HBV-positive and HBV-negative patients and a favorable China subgroup signal 125. Economic considerations are non-trivial: a cost-effectiveness analysis from a Chinese healthcare system perspective found that durvalumab's addition to GC may not reach conventional cost-effectiveness thresholds without substantial price reductions or patient assistance programs 9. PD-L1 testing, while encouraged, should account for assay-specific variability (clone, threshold), and should not be used as a mandatory selection criterion given inconsistent predictive performance across BTC studies 1617. Immune-mediated hepatotoxicity requires proactive surveillance and structured management algorithms in all patients, particularly given BTC patients' frequent baseline hepatic compromise 8.
In conclusion, both chemoimmunotherapy regimens offer clinically meaningful and durable OS benefits over chemotherapy alone in first-line advanced BTC, with long-term survivor data reinforcing durability 311. The choice between durvalumab- and pembrolizumab-based regimens should be guided by a holistic assessment encompassing regional practice norms, regulatory approvals, patient comorbidity and hepatic function, HBV status, institutional familiarity with biomarker monitoring, and access/pharmacoeconomic considerations—rather than cross-trial efficacy comparisons that the available evidence does not support 157.