Document resource
WHAT IS ALREADY KNOWN ON THIS TOPIC Liposomal irinotecan+oxaliplatin+fluorouracil+leucovorin (NALIRIFOX) improved survival compared with gemcitabine plus nab-paclitaxel (Gem/NabP) in the phase III NAPOLI-3 (NALIRIFOX versus nab-paclitaxel and gemcitabine in treatment-naive patients with metastatic pancreatic ductal adenocarcinoma: a randomised, open-label, phase 3 trial) trial and is now used as a first-line option for metastatic pancreatic ductal adenocarcinoma (PDAC). Modified FOLFIRINOX (mFOLFIRINOX) is also widely used in routine practice, but direct comparative data between NALIRIFOX and mFOLFIRINOX are limited. Real-world evidence comparing NALIRIFOX, mFOLFIRINOX and Gem/NabP remains scarce, particularly in Asian populations.WHAT THIS STUDY ADDS In this multicentre Asian propensity score-matched real-world cohort of advanced PDAC, NALIRIFOX and mFOLFIRINOX were associated with similar overall and progression-free survival. Both regimens were associated with more favourable survival outcomes than Gem/NabP in the matched cohort. Gem/NabP showed the highest objective response rate but the shortest survival. Recorded grade ≥3 toxicities appeared lower with NALIRIFOX, although safety comparisons were limited by retrospective toxicity ascertainment and likely differential reporting.HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY These findings provide comparative real-world data to inform the interpretation of first-line treatment outcomes in advanced PDAC, where head-to-head randomised evidence is lacking. The results support further prospective studies with standardised toxicity reporting, biomarker integration and treatment-sequencing data. In practice, they may help contextualise regimen selection but should be interpreted cautiously given the retrospective design and residual confounding.Introduction Pancreatic ductal adenocarcinoma (PDAC) is an aggressive gastrointestinal malignancy characterised by an asymptomatic early clinical course, late-stage diagnosis and a poor prognosis. The estimated 5-year survival rate remains critically low, particularly for the majority of patients who present with advanced disease. 1 PDAC remains a leading cause of cancer-related mortality globally, posing a substantial public health burden in the Asia-Pacific region and worldwide.2 Given the systemic nature of advanced PDAC, cytotoxic chemotherapy remains the cornerstone of disease management, aiming to prolong survival, control tumour growth and preserve quality of life.Over the past decade, the therapeutic landscape for advanced PDAC has been dominated by two-combination chemotherapy regimens that significantly improved patient outcomes compared with historical single-agent treatments: fluorouracil+leucovorin+irinotecan+oxaliplatin (FOLFIRINOX) and gemcitabine plus nab-paclitaxel (Gem/NabP).3 4 Recognising the substantial toxicity profile of classic FOLFIRINOX, modified FOLFIRINOX (mFOLFIRINOX) regimens—which typically omit the fluorouracil bolus and reduce the irinotecan dose—have been widely adopted to mitigate adverse events while preserving efficacy, making the triplet combination more tolerable in routine real-world practice.5–9 More recently, the liposomal irinotecan+oxaliplatin+fluorouracil+leucovorin (NALIRIFOX) regimen emerged as a promising first-line therapy.10 The landmark phase III NAPOLI-3 (NALIRIFOX versus nab-paclitaxel and gemcitabine in treatment-naive patients with metastatic pancreatic ductal adenocarcinoma (NAPOLI 3): a randomised, open-label, phase 3 trial) trial established NALIRIFOX as a new standard of care, demonstrating statistically significant improvements in overall survival (OS) and progression-free survival (PFS) compared with Gem/NabP in treatment-naïve patients with metastatic PDAC.11 12 Despite these advances, a gap persists in the real-world comparative effectiveness and safety data for NALIRIFOX versus mFOLFIRINOX and Gem/NabP. Patient populations encountered in routine clinical practice exhibit marked heterogeneity in age, performance status, nutritional reserves and comorbidity burden, which frequently do not align with the strict eligibility criteria of randomised controlled trials.10 12 13 Although recent meta-analyses have offered indirect comparisons suggesting comparable efficacy between NALIRIFOX and mFOLFIRINOX,10 13 real-world studies employing robust statistical methods, such as propensity score matching (PSM), are essential to minimise confounding and generate reliable, generalisable insights into regimen performance across diverse clinical settings.14 Currently, evidence from real-world practice remains limited. A recent multicentre European study confirmed that NALIRIFOX shows consistent efficacy and manageable toxicity in unselected patients with advanced PDAC.15 However, no multicentre Asian study has directly compared NALIRIFOX, mFOLFIRINOX and Gem/NabP using PSM. Real-world data are particularly vital in populations where variables such as cachexia prevalence, unique anthropometrics and varying degrees of clinical heterogeneity strongly influence treatment tolerability and sequence planning in ways not fully captured by pivotal trials.Therefore, we conducted the first multicentre Asian propensity score-matched real-world study to compare the efficacy and safety of NALIRIFOX, mFOLFIRINOX and Gem/NabP in treatment-naïve patients with advanced PDAC across 10 major Thai cancer centres.Methods Study design and setting This was a retrospective, multicentre cohort study conducted in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines. 16 Clinical data were systematically collected from 10 major university hospitals and comprehensive cancer centres across Thailand, under the auspices of the Thai Society of Clinical Oncology (TSCO) Gastrointestinal and Hepatobiliary Research Group.Participants The study included all eligible adult patients (aged ≥18 years) with histologically or cytologically confirmed advanced PDAC—defined as either unresectable locally advanced or metastatic disease—who initiated first-line systemic palliative chemotherapy with NALIRIFOX, mFOLFIRINOX or Gem/NabP between January 2021 and July 2024. Key eligibility criteria required an Eastern Cooperative Oncology Group performance status (ECOG PS) of 0–1 and adequate haematological, hepatic and renal organ function prior to cycle 1, day 1. Patients were required to be treatment naïve in the advanced disease setting; however, prior adjuvant chemotherapy was permissible provided it was completed more than 6 months prior to disease recurrence. Patients were excluded if they received an alternative first-line regimen, participated in an interventional clinical trial for first-line therapy or had grossly incomplete data regarding baseline characteristics or follow-up timelines.Because this was a retrospective real-world study, treatment assignment was not randomised. Patients were treated according to institutional standards, clinical guidelines and individual physician judgement. The regimen descriptions reflect common standard dosing protocols across the participating centres, though specific dose modifications, cycle delays and the use of primary or secondary prophylactic granulocyte colony-stimulating factor were at the treating physician’s discretion, based on individual patient tolerability and institutional policies.Data sources and data management Clinical data were abstracted directly from electronic medical records and systematically entered into the Research Electronic Data Capture system following a predefined, standardised data dictionary. 17 18 This harmonised registry framework was maintained by the Excellence Phramongkutklao Hospital Comprehensive Cancer Center and the TSCO governance structure under the registry ID ‘TH PANC version 2.0’. Source documentation included medical oncology clinic notes, multidisciplinary tumour board summaries, structured radiology reports, laboratory data and hospital pharmacy dispensing records. To ensure data integrity, duplicate entries were resolved via manual adjudication by independent clinical reviewers. Comprehensive data completeness checks, range validations and temporal plausibility assessments were performed prior to database lock. The final frozen analytical dataset was generated on 31 December 2024.Endpoints and variables The primary efficacy outcomes evaluated were OS, defined as the time from the initiation of first-line chemotherapy to death from any cause, and PFS, defined as the time from treatment initiation to documented clinical or radiographic disease progression, or death. Secondary outcomes included the objective response rate (ORR) assessed by the treating investigators using the Response Evaluation Criteria in Solid Tumors (RECIST) version 1.1, 19 and the incidence of grade ≥3 treatment-related adverse events graded according to the Common Terminology Criteria for Adverse Events (CTCAE) version 5.0.20 21 Baseline clinical covariates recorded included patient age, biological sex, ECOG PS, primary tumour location (head vs body/tail), disease stage (unresectable locally advanced vs metastatic), metastatic site distribution, baseline serum albumin and baseline carbohydrate antigen 19-9 (CA19-9) tumour marker levels.Sample size and bias control As a retrospective real-world evidence analysis, the overall study size was determined by the total number of eligible consecutive patients identified across all participating sites during the predefined 2021–2024 timeframe (n=513 initially screened). To rigorously minimise inherent treatment selection bias and balance baseline clinical characteristics among the three non-randomised cohorts, we performed PSM. Propensity scores were estimated via multivariable logistic regression. The matching covariates were specifically chosen based on their known prognostic relevance in PDAC and included age, sex, ECOG PS, metastatic burden, primary tumour location and baseline serum albumin.Notably, baseline CA19-9 was intentionally excluded from the PSM model due to incomplete universal availability across historical records and marked statistical skewness between treatment groups. CA19-9 was instead reported descriptively and analytically treated as a residual confounder. Nearest-neighbour matching with a stringent calliper width of 0.2 SD of the logit of the propensity score was applied. Adequate covariate balance postmatching was assessed using standardised mean differences (SMD), with a target SMD of <0.1 indicating successful balance.Missing data handling Missing data for key covariates were proactively addressed through prelock data queries to the participating sites, resulting in less than 5% residual missingness across all primary demographic and clinical variables. Propensity score estimation and subsequent matching were performed using a complete-case analysis approach, with no artificial imputation applied to baseline covariates. For survival analyses, patients who remained alive or without disease progression were censored at the date of their last verified clinical encounter. Sensitivity analyses were also conducted using the unmatched whole cohort with multivariable Cox proportional hazards adjustment, which produced highly concordant effect estimates, thereby supporting the robustness and validity of the primary complete-case matching approach.Statistical analysis A multivariable logistic regression model was used to calculate a propensity score for each patient based on age, sex, ECOG PS, metastatic burden (number and distribution of metastatic sites), tumour location and albumin, following previous PDAC modelling approaches. The complete covariate list is provided in online supplemental table 1. A nearest-neighbour matching algorithm (calliper width 0.2) was then applied. Covariate balance was assessed using SMD <0.1. Efficacy analyses were performed on the matched cohort.SP110.1136/bmjonc-2026-001085.supp1Supplementary data Safety analyses were conducted on the safety population, defined as all matched patients who received at least one full or partial dose of the study treatment and possessed sufficient follow-up documentation to evaluate severe adverse events. Operationally, patients were excluded from the safety denominator if they transferred care to an outside facility immediately following cycle 1, or if they lacked subsequent clinical notes, laboratory results or hospitalisation records necessary to accurately grade CTCAE criteria. Based on this, a small number of patients were excluded from the specific safety denominators (excluded: n=2 for NALIRIFOX, n=8 for mFOLFIRINOX, n=0 for Gem/NabP).Survival probabilities were estimated using the Kaplan-Meier method, and differences between treatment arms were compared using the log-rank test. HRs and their associated 95% CIs were calculated via Cox proportional hazards models. The fundamental proportional hazards assumption was statistically verified using Schoenfeld residuals.22 Subgroup analyses for OS and PFS were systematically performed across prespecified clinical subgroups, including age categories, ECOG status and primary tumour location. All statistical tests were two-sided, with a p value <0.05 considered statistically significant. All analyses were executed using Stata MP V.18.Results Participant flow and characteristics During the study period, a total of 513 patients with advanced PDAC were initially identified across the 10 participating centres. After applying PSM to balance baseline demographics and clinical characteristics, 199 patients were retained for the final comparative analysis. Within this matched cohort, 52 patients received NALIRIFOX, 105 received mFOLFIRINOX and 42 received Gem/NabP (see online supplemental figure 1). The application of PSM successfully achieved SMD of <0.1 for all variables included in the matching model, confirming adequate control of measured baseline confounding (see online supplemental table 1). The matched cohort had a median age of roughly 60 years, with an even distribution of male and female patients, and the vast majority possessed an ECOG PS of 0–1. While the model balanced variables such as tumour location (predominantly pancreatic head) and serum albumin, baseline CA19-9—which was omitted from the PSM algorithm—remained significantly imbalanced postmatching. Specifically, the median baseline CA19-9 was higher at 4327 U/mL in the NALIRIFOX group than at 926 U/mL in the mFOLFIRINOX group (p<0.01). Because CA19-9 is an established prognostic indicator in PDAC, this imbalance represents a source of residual confounding that must be considered when interpreting the efficacy comparisons (table 1).Table 1Baseline characteristics of patients by treatment groups and post-propensity score matchingOverall population(n=513)Post-propensity score matching(n=199)NALIRIFOX(n=52)mFOLFIRINOX(n=301)Gem/NabP(n=160)P valueNALIRIFOX(n=52)mFOLFIRINOX(n=105)Gem/NabP(n=42)P valueAge; year (IQR)62(54, 66)63(53, 69)62(52, 70)0.7062(54, 66)58(49, 66)60(48, 69)0.25Sex0.810.73 Male2616587264822 Female2613673265720ECOG PS<0.01*0.15 0–146101404610140 ≥26200120642Staging0.050.07 Locally advanced unresectable9975792716 Metastasis43204103437826Tumour location0.03*0.11 Head3020590306518 Others229670224024 Body116743112611 Tail112927111413CA19–9, median (IQR) (μ/mL)4327(25, 4329)929(89, 961)1839(92, 1840)<0.01*4327(25, 4329)926(925, 927)1840(1839, 1841)<0.01* CA19–9 <37 (μ/mL)1347250.481373<0.01* CA19–9 ≥37 (μ/mL)37202110379437 Total (n)502491355010140Albumin, median (IQR) (g/dL)3.95(3.4, 4.4)3.7(3.4, 4.1)3.7(3.3, 4.1)0.303.95(3.4, 4.4)3.8(3.3, 4.2)3.7(3.3, 4.4)0.72 Total (n)522961585210442Absolute lymphocyte count, median (IQR) (x10∧9/L)1.686(1.457, 1.818)1.638(1.294, 1.836)1.676(1.438, 1.801)0.681.686(1.457, 1.818)1.665(1.552, 1.759)1.682(1.642, 1.749)0.51 Total (n)523011605210542Baseline CA19-9 was not included in the propensity score model because of missingness and skewness; values are presented descriptively.*Statistically significant; p values derived from χ² or Fisher’s exact test for categorical variables and Kruskal-Wallis test for continuous variables.CA19-9, carbohydrate antigen 19-9; ECOG PS, Eastern Cooperative Oncology Group performance status; Gem/NabP, gemcitabine plus nab-paclitaxel; mFOLFIRINOX, modified FOLFIRINOX; NALIRIFOX, liposomal irinotecan+oxaliplatin+fluorouracil+leucovorin.Efficacy outcomes In the matched population, the median OS from the initiation of first-line therapy was 10.4 months in the NALIRIFOX group, 9.6 months in the mFOLFIRINOX group and 6.0 months in the Gem/NabP group. Survival analysis demonstrated that NALIRIFOX achieved significantly improved OS compared with Gem/NabP (HR 1.68, 95% CI 1.03 to 2.75; p=0.04). However, the comparative difference in OS between NALIRIFOX and mFOLFIRINOX did not reach statistical significance (HR 1.09, 95% CI 0.72 to 1.67; p=0.68), suggesting comparable long-term efficacy. Evaluation of landmark survival rates revealed that the 12-month OS rate was 32.7% for NALIRIFOX, 39.1% for mFOLFIRINOX and 10.1% for Gem/NabP ( figure 1A).Figure 1Kaplan-Meier survival analyses in post-propensity score-matched patients. Panel (A) displays the Kaplan-Meier curve for OS. The median OS was 10.4 months for the NALIRIFOX group, 9.6 months for the mFOLFIRINOX group and 6.0 months for the Gem/NabP group. The difference was statistically significant when comparing NALIRIFOX to Gem/NabP (HR 1.68, 95% CI 1.03 to 2.75; p=0.04) but not when comparing NALIRIFOX to mFOLFIRINOX (HR 1.09, 95% CI 0.72 to 1.67; p=0.68). Panel (B) displays the Kaplan-Meier curve for PFS. The median PFS was 5.0 months for both the NALIRIFOX and mFOLFIRINOX groups, compared with 4.3 months for the Gem/NabP group. The difference was statistically significant between NALIRIFOX and Gem/NabP (HR 1.63, 95% CI 1.05 to 2.50; p=0.03) but not between NALIRIFOX and mFOLFIRINOX (HR 1.02, 95% CI 0.72 to 1.45; p=0.90). Gem/NabP, gemcitabine plus nab-paclitaxel; mFOLFIRINOX, modified FOLFIRINOX; NALIRIFOX, liposomal irinotecan+oxaliplatin+fluorouracil+leucovorin; OS, overall survival; PFS, progression-free survival.PFS outcomes mirrored the OS data. The median PFS was 5.0 months for both the NALIRIFOX and mFOLFIRINOX groups, which was superior to the 4.3 months observed for the Gem/NabP group. Statistical comparisons confirmed that NALIRIFOX achieved significantly longer PFS compared with Gem/NabP (HR 1.63; p=0.03), while differences between NALIRIFOX and mFOLFIRINOX were negligible (figure 1B).Analysis of the ORR via RECIST v1.1 criteria yielded an interesting outcome.19 The ORR was 37.5% for NALIRIFOX, 23.8% for mFOLFIRINOX and highest at 40.5% for the Gem/NabP group (table 2). Subgroup analyses stratifying patients by disease state (unresectable locally advanced vs metastatic disease) demonstrated consistent directionality, reinforcing the primary survival outcomes across the different clinical stages of advanced PDAC (figure 2) (see the online supplemental figures 2 and 3).Figure 2Subgroup analysis of overall survival. Forest plots showing HRs for overall survival in prespecified subgroups. Panel (A) compares NALIRIFOX with mFOLFIRINOX. Panel (B) compares NALIRIFOX with Gem/NabP. Panel (C) compares mFOLFIRINOX with Gem/NabP. The size of the squares is proportional to the number of patients in each subgroup. The horizontal lines represent the 95% CIs. The diamond at the bottom represents the overall HR for all patients. Alb, albumin; ECOG, Eastern Cooperative Oncology Group; Gem/NabP, gemcitabine plus nab-paclitaxel; LA, locally advanced; M1, metastatic; mFOLFIRINOX, modified FOLFIRINOX; NALIRIFOX, liposomal irinotecan+oxaliplatin+fluorouracil+leucovorin.Table 2Survival and response rates in post-propensity score-matched patientsNALIRIFOX(n=52)mFOLFIRINOX(n=105)Gem/NabP(n=42)HR (95% CI)P valueMedian follow-up time, months9.268.775.58Overall survival mOS, months (95% CI)10.4 (9.3 to 13.4)9.6 (7.1 to 11.6)–1.09 (0.72 to 1.67)0.68–6.0 (4.6 to 8.1)1.68 (1.03 to 2.75)0.04* 6-month survival rate, % (95% CI)74.9 (59.9 to 84.9)67.7 (57.7 to 75.8)48.1 (32.1 to 62.4) 12-month survival rate, % (95% CI)32.7 (15.2 to 51.5)39.1 (29.5 to 48.5)10.1 (3.2 to 21.7)Progression-free survival mPFS, months (95% CI)5.0 (2.8 to 6.4)5.0 (3.4 to 6.2)–1.02 (0.72 to 1.45)0.90–4.3 (3.3 to 6.4)1.63 (1.05 to 2.50)0.03* 6-month PFS rate, % (95% CI)38.5 (25.4 to 51.3)43.8 (34.2 to 53.0)40.5 (25.8 to 54.7) 12-month PFS rate, % (95% CI)9.6 (3.5 to 19.4)13.2 (7.5 to 20.6)–Response by RECIST criteria Number of patients with overall response4810542 Overall response rate, n (%)18 (37.5)25 (23.8)17 (40.5) Disease control rate, n (%)36 (75.0)69 (65.7)31 (73.8)Best response Complete response, n (%)0 (0)0 (0)0 (0) Partial response, n (%)18 (37.5)25 (23.8)17 (40.5) Stable disease, n (%)18 (37.5)44 (41.9)14 (33.3) Progressive disease, n (%)12 (25)36 (34.3)11 (26.2)*Statistically significant; p values derived from χ² or Fisher’s exact test for categorical variables and Kruskal-Wallis test for continuous variables.Gem/NabP, gemcitabine plus nab-paclitaxel; mFOLFIRINOX, modified FOLFIRINOX; mOS, median overall survival; mPFS, median progression-free survival; NALIRIFOX, liposomal irinotecan+oxaliplatin+fluorouracil+leucovorin; PFS, progression-free survival; RECIST, Response Evaluation Criteria in Solid Tumors.Safety and tolerability Safety and toxicity evaluations were conducted exclusively on patients possessing sufficient clinical documentation for retrospective adverse event grading (n=50 NALIRIFOX, n=97 mFOLFIRINOX, n=42 Gem/NabP). 20 21 Recorded grade ≥3 toxicities appeared lower with NALIRIFOX, although retrospective capture limits the reliability of cross-regimen comparisons. Specifically, grade ≥3 fatigue was recorded at 5.9% for NALIRIFOX, compared with 47.4% for mFOLFIRINOX and 37.8% for Gem/NabP. Furthermore, no cases of grade ≥3 chemotherapy-induced peripheral neuropathy were captured in the NALIRIFOX group, whereas incidences of 12.5% and 10.0% were recorded in the mFOLFIRINOX and Gem/NabP groups, respectively. It is crucial to note, however, that any-grade fatigue and diarrhoea were documented at nearly 100% and 99%, respectively, for mFOLFIRINOX, strongly suggesting differential reporting bias between the regimens (table 3).Table 3Incidence of grade ≥3 TRAEsmFOLFIRINOXGem/NabPNALIRIFOXP valueAny gradeGrade ≥3Any gradeGrade ≥3Any gradeGrade ≥3Any events, n (%)97 (93.3)97 (98.9)42 (100.0)42 (50.0)50 (96.2)47 (48.5)0.202Non-haematological adverse event Nausea and vomiting, n (%)87 (89.7)16 (18.4)8 (19.0)0 (0)19 (38.0)2 (10.5)<0.005* Diarrhoea, n (%)96 (99.0)1 (1.0)8 (19.0)0 (0)28 (56.0)0 (0)<0.005* Fatigue, n (%)97 (100.0)46 (47.4)37 (88.1)14 (37.8)34 (68.0)2 (5.9)<0.005* CIPN, n (%)96 (99.0)12 (12.5)10 (23.8)1 (10.0)19 (38.0)0 (0)<0.005*Haematological adverse event Anaemia, n (%)82 (84.5)6 (7.3)32 (76.2)20 (62.5)42 (84.0)4 (9.5)0.468 Neutropenia, n (%)95 (97.9)38 (40.0)42 (100.0)12 (28.6)35 (70.0)8 (21.6)<0.005* Infections, n (%)23 (23.7)5 (21.7)6 (14.3)1 (16.7)11 (22.0)4 (36.4)0.452 Thrombocytopenia, n (%)75 (77.3)0 (0)40 (95.2)13 (32.5)26 (52.0)0 (0)<0.005* Thromboembolism, n (%)26 (26.8)5 (19.2)6 (14.3)0 (0)11 (22.0)0 (0)0.268Safety population denominators excluded patients lacking sufficient follow-up documentation for grade ≥3 adverse events (exclusions: n=2 for NALIRIFOX, n=8 for mFOLFIRINOX, n=0 for Gem/NabP).*Statistically significant.CIPN, chemotherapy-induced peripheral neuropathy; Gem/NabP, gemcitabine plus nab-paclitaxel; mFOLFIRINOX, modified FOLFIRINOX; NALIRIFOX, liposomal irinotecan+oxaliplatin+fluorouracil+leucovorin; TRAEs, treatment-related adverse events.Discussion This multicentre, real-world, propensity score-matched cohort study provides valuable insights into the comparative effectiveness and safety of the three dominant first-line systemic regimens used for advanced PDAC. Overall, our findings suggest that NALIRIFOX yields survival outcomes superior to Gem/NabP and broadly comparable to those of the established mFOLFIRINOX regimen in a heterogeneous clinical population.The observed superiority of NALIRIFOX over Gem/NabP in terms of both OS and PFS directly corroborates the primary findings of the randomised NAPOLI-3 trial.12 Validating these trial results in an unselected, real-world Asian cohort is clinically important, as it confirms that the survival benefits of NALIRIFOX are preserved outside the stringent confines of a clinical trial. When directly compared with mFOLFIRINOX, NALIRIFOX demonstrated comparable median OS and PFS. It is worth noting a nuance in the data: despite similar medians, mFOLFIRINOX exhibited a numerically higher 12-month survival rate (39.1%) compared with NALIRIFOX (32.7%). This crossover in the survival curves may reflect variations in the long-term survival tail, differential impacts of censoring at later time points or artefacts of the relatively small sample size in the matched NALIRIFOX arm. Consequently, while the medians indicate equivalence, absolute superiority at later time horizons should not be overinterpreted without larger, longer term prospective validation.10 13 An interesting paradoxical finding in our cohort was that Gem/NabP demonstrated the highest ORR (40.5%), yet yielded the poorest OS and PFS. Because the rigorous PSM algorithm successfully balanced key clinical indicators of patient frailty—such as age, ECOG PS and serum albumin—across the three cohorts, this tension cannot be dismissed by assuming that Gem/NabP was administered exclusively to frailer, less fit patients. Rather, this discordance between early radiological shrinkage and long-term survival durability may reflect unmeasured confounding or fundamental clinical differences not captured in the propensity model. It underscores a vital clinical reality in advanced PDAC: rapid tumour response does not guarantee sustained disease control, and survival-based endpoints remain the most reliable metric for guiding first-line regimen selection.The safety and tolerability comparisons must be interpreted with a high degree of caution. While our retrospective analysis recorded a favourable toxicity profile for NALIRIFOX—5.9% grade ≥3 fatigue vs 47.4% for mFOLFIRINOX—these figures are almost certainly skewed by differential ascertainment and retrospective reporting biases. Established regimens like mFOLFIRINOX, which have been used for years at these cancer centres, benefit from highly standardised documentation practices by nursing and clinical staff. Conversely, adverse events for newer regimens like NALIRIFOX may be undercaptured in routine clinical notes during their early adoption phase. Supporting this hypothesis, prospective data from the NAPOLI-3 trial reported severe fatigue rates for NALIRIFOX that were substantially higher than the 5.9% observed in our retrospective capture.12 23 24 Therefore, while recorded severe toxicities appeared lower with NALIRIFOX, these retrospective safety comparisons remain fundamentally limited.Recent Pan-Asian European Society For Medical Oncology consensus guidelines emphasise that the optimal treatment selection for advanced PDAC must account for both intrinsic therapeutic value and practical real-world clinical feasibility across diverse patient phenotypes.25 Furthermore, emerging data specific to Asian populations increasingly highlight the prognostic relevance of baseline inflammatory and nutritional markers, particularly the neutrophil-to-lymphocyte ratio (NLR) and the platelet-to-lymphocyte ratio.26 A recent meta-analysis encompassing over 20 000 patients reinforced that elevated NLR is powerfully associated with poorer OS in PDAC (multivariable HR 1.79).27 Integrating these inflammatory markers into future real-world studies will be critical to refine risk stratification and improve adjustment for the complex biological heterogeneity that standard clinical covariates currently fail to capture28 .Our study has several important limitations inherent to its retrospective, observational design. Chief among them is the residual baseline imbalance in CA19-9. Because CA19-9 data were skewed and frequently missing, it was excluded from the propensity score model, resulting in a substantially higher median CA19-9 burden in the NALIRIFOX arm. As a recognised prognostic marker, this heavier disease marker burden may have obscured the true comparative efficacy of NALIRIFOX. Additionally, detailed tracking of second-line and subsequent therapies was not systematically available. Differences in the receipt of, access to, or clinical tolerance of active second-line treatments could substantially confound the observed OS outcomes.Conclusion In this multicentre, real-world matched cohort of patients with advanced PDAC, NALIRIFOX and mFOLFIRINOX demonstrated comparable OS and PFS, both of which outperformed Gem/NabP. Recorded grade ≥3 toxicities appeared lower with NALIRIFOX; however, safety comparisons were inherently limited by retrospective toxicity ascertainment and potential reporting biases. While these real-world data confirm the clinical feasibility and efficacy of NALIRIFOX across diverse practice settings, prospective studies with rigorous biomarker integration are necessary to definitively validate its safety profile and address residual confounding factors.