Prognostic value of lactate dehydrogenase-to-albumin ratio in metastatic/recurrent pancreatic cancer treated with first-line gemcitabine plus nab-paclitaxel: a single-center retrospective cohort study

Introduction

The incidence of pancreatic cancer has been increasing worldwide. In the United States, the estimated number of new cases in 2025 was 67,440, accounting for 3.3% of all new cancer cases, and the estimated number of deaths was 51,980, representing 8.4% of all cancer deaths (1). Clinically, a substantial proportion of pancreatic cancer patients are diagnosed at an advanced, unresectable stage, and even when resection is performed, recurrence is not uncommon. The 5-year relative survival rate for patients with distant-stage pancreatic cancer is 3.2%, indicating a highly unfavorable prognosis (1).

In Japan, gemcitabine plus nab-paclitaxel (GnP) has been widely adopted in clinical practice as a first-line chemotherapy for unresectable pancreatic cancer, following the results of the GENERATE trial (2). However, even when treated with the same regimen, clinical outcomes such as treatment efficacy and survival periods vary substantially between patients. This variability is thought to be a result of not only tumor-associated factors such as tumor burden and metastatic sites, but also various patient-associated factors, including systemic conditions such as inflammatory status and nutritional status. Therefore, factors that enable prognosis stratification at the start of treatment are clinically important, and such factors should be as simple as possible. Performance status (PS) is widely recognized as a fundamental clinical indicator that determines chemotherapy eligibility and appropriate treatment intensity, and it is also strongly associated with clinical outcomes (3,4). However, PS has been reported to show inter-rater variability due to its inherent subjectivity and its incorporation of multifactorial elements; therefore, concerns have been raised that PS alone may not adequately capture the heterogeneity of both host and tumor factors (5,6).

Recently, accumulating evidence has shown that indices associated with inflammation and nutrition act as prognostic markers across various cancer types (7-11). Among these, the lactate dehydrogenase-to-albumin ratio (LAR), derived from lactate dehydrogenase (LDH) and albumin, has emerged as an integrated biomarker incorporating LDH, a potential marker of tumor metabolic activity and burden, and albumin, a surrogate marker of host inflammatory and nutritional status. Prior studies have reported its prognostic relevance across multiple malignancies (7). However, evidence specific to pancreatic cancer remains limited, and data focusing on patients treated with first-line GnP are still scarce. Accordingly, it is clinically important to determine whether incorporating LAR, an objective blood-based biomarker, together with PS can improve prognostic stratification in patients with metastatic or recurrent unresectable pancreatic cancer receiving first-line GnP therapy.

Therefore, this study assessed the prognostic value of baseline LAR in this setting and analyzed a simple stratification approach combining LAR and Eastern Cooperative Oncology Group (ECOG) PS. We present this article in accordance with the STROBE reporting checklist (available at https://jgo.amegroups.com/article/view/10.21037/jgo-2026-0223/rc).

Methods

Patients

This single-center retrospective cohort study included consecutive patients with metastatic or recurrent unresectable pancreatic cancer who started GnP as a first-line chemotherapy at Tokyo Medical University Hospital between January 2015 and December 2024. Patients with locally advanced disease were excluded to maintain cohort homogeneity for overall survival (OS) comparisons, as their prognosis differs from that of metastatic or recurrent disease. Owing to the retrospective design, no formal a priori sample size calculation was performed; therefore, all eligible patients during the study period were included. Analyses were conducted as a complete-case analysis, including only patients with no missing data in the prespecified variables. Accordingly, there were no missing values in the covariates included in the prespecified multivariable model. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study protocol was approved by the Institutional Review Board of Tokyo Medical University (No. T2020-0422), and the requirement for individual informed consent was waived because of the retrospective nature of the study.

Data collection

Data were extracted from the electronic medical records at Tokyo Medical University Hospital. Baseline characteristics included age, sex, PS, disease status (metastatic vs. recurrent), and metastatic sites (including liver metastasis and peritoneal dissemination). Serum LDH and albumin levels, absolute neutrophil count, and absolute lymphocyte count were obtained from routine blood tests performed within 7 days before the initiation of GnP treatment. Serum carbohydrate antigen 19-9 (CA19-9) values measured within 28 days before treatment initiation were used.

Variable definitions

In this study, systemic inflammation indices, including LAR and the neutrophil-to-lymphocyte ratio (NLR), were evaluated. Serum albumin was recorded in g/dL in the electronic medical records and converted to g/L for LAR calculation. LAR was calculated as the serum LDH level (U/L) divided by the serum albumin level (g/L). NLR was calculated as the baseline absolute neutrophil count divided by the absolute lymphocyte count. Serum CA19-9 levels were natural log-transformed as ln(CA19-9 + 1) to account for a right-skewed distribution and to accommodate zero values.

Statistical analysis

OS was defined as the time from initiation of GnP to death from any cause; patients without a recorded death were censored at the date of last confirmed contact (last known alive). Follow-up was administratively censored at the cutoff date of 19 November 2025, and patients lost to follow-up were censored at their last confirmed contact. Continuous variables are summarized as the median [interquartile range (IQR)] and categorical variables as the number (%).

Associations with OS were first assessed using univariable Cox proportional hazards models. A prespecified multivariable Cox model was then fitted, including age, sex, ECOG PS, disease status (recurrent vs. metastatic), NLR, LAR, log-transformed CA19-9, liver metastasis, and peritoneal dissemination. For continuous biomarkers [NLR, LAR, and log(CA19-9)], hazard ratios (HRs) were reported per 1-standard deviation (SD) increase. NLR was included a priori as a representative systemic inflammation marker, which has been reported to have prognostic utility across multiple malignancies (12), to adjust for baseline inflammatory status when estimating the association between LAR and OS. The proportional hazards assumption was evaluated using Schoenfeld residuals (global test). To assess potential nonlinearity in the association between LAR and OS, LAR was additionally modeled using restricted cubic splines (RCS) with 4 knots within the multivariable Cox model, with tests for overall and nonlinear effects. For clinical visualization of risk stratification, PS (0–1 vs. 2) and LAR (low vs. high using the cohort median) were combined to define four groups, and OS was compared across groups using Kaplan-Meier curves and the log-rank test. For incremental prognostic assessment, a base clinical model was prespecified including sex, disease status, log-transformed CA19-9, liver metastasis, and peritoneal dissemination. Nested models (base model alone, base + PS, base + LAR, and base + PS + LAR) were compared using Akaike’s information criterion (AIC), Harrell’s concordance index (C-index), and likelihood ratio tests. Additional models incorporating LDH alone or albumin alone into the base model were also evaluated to compare LAR with its individual components. Internal validation of the base + PS + LAR model was performed using bootstrap resampling (1,000 iterations) to estimate optimism and derive the optimism-corrected C-index. All analyses were performed using R version 4.5.2 software (R Foundation for Statistical Computing, Vienna, Austria; https://www.R-project.org), and two-sided P values less than 0.05 were considered to indicate statistically significant differences between groups. Chat GPT (Open AI) was used from January 2026 to March 2026 for limited assistance with R code refinement and preparation of related text. All AI-assisted output was reviewed and approved by the authors. No AI tool was used for data interpretation or for drawing the study conclusions.

Results

The final analytic cohort comprised 205 patients (Figure 1). Of the 448 patients who received GnP at Tokyo Medical University Hospital during the study period, 243 were excluded because GnP was not used as a first-line therapy for unresectable disease (n=185; neoadjuvant, adjuvant, or second-line and subsequent setting), or because the disease was locally advanced and unresectable (n=58). During the follow-up period, 151 deaths occurred. The median follow-up time, estimated using the reverse Kaplan-Meier method, was 34.3 months [95% confidence interval (CI), 31.5–48.2].

Figure 1 Flow chart depicting the patient selection process. GnP, gemcitabine plus nab-paclitaxel.

Patient characteristics are summarized in Table 1. The median age was 67.0 years (IQR, 60.0–73.0 years), and 131 patients (63.9%) were men. ECOG PS was 0 in 120 patients (58.5%), 1 in 67 (32.7%), and 2 in 18 (8.8%). Median OS was 10.9 months (IQR, 7.2–20.3 months). Disease status was metastatic in 96 patients (46.8%) and recurrent in 109 (53.2%). Liver metastasis and peritoneal dissemination were present in 98 (47.8%) and 34 (16.6%) patients, respectively. The median LAR was 4.94 (IQR, 4.19–5.65), the median NLR was 2.80 (IQR, 1.93–4.35), and the median CA19-9 was 139.4 U/mL (IQR, 36.2–887.0 U/mL).

In univariable Cox analyses, older age (per 1-SD increase: HR, 1.20; 95% CI, 1.01–1.41; P=0.04), poorer PS (PS 1 vs. 0: HR, 1.73; 95% CI, 1.21–2.48; P=0.003; PS 2 vs. 0: HR, 3.16; 95% CI, 1.77–5.65; P<0.001), recurrent disease status (vs. metastatic: HR, 0.64; 95% CI, 0.47–0.89; P=0.007), higher LAR (per 1-SD increase: HR, 1.43; 95% CI, 1.23–1.66; P<0.001), higher NLR (per 1-SD increase: HR, 1.31; 95% CI, 1.12–1.53; P<0.001), liver metastasis (HR, 1.64; 95% CI, 1.08–2.48; P=0.02), and peritoneal dissemination (HR, 1.65; 95% CI, 1.09–2.50; P=0.02) were associated with OS (Table 2). Sex and log-transformed CA19-9 were not associated with OS in univariable analyses (Table 2).

In the prespecified multivariable Cox model, PS and LAR remained independently associated with OS (PS 2 vs. 0: HR, 2.38; 95% CI, 1.28–4.42; P=0.006; LAR per 1-SD increase: HR, 1.29; 95% CI, 1.09–1.52; P=0.003) (Table 2). NLR, disease status, liver metastasis, peritoneal dissemination, log(CA19-9), age, and sex were not statistically significant after adjustment. The proportional hazards assumption was met (global test, P=0.48).

When LAR was modeled using RCS (4 knots) within the multivariable Cox model, the overall association with OS was significant (overall P=0.03), with no evidence of nonlinearity (nonlinear P=0.71) (Figure 2).

Figure 2 Restricted cubic spline analysis of the LAR and overall survival. The solid line shows the adjusted HR, with the median LAR used as the reference (HR =1.0). The overall association was significant (overall P=0.03), with no evidence of nonlinearity (nonlinear P=0.71). The model was adjusted for sex, age, ECOG PS, disease status (metastatic vs. recurrence), liver metastasis, peritoneal dissemination, log-transformed CA19-9, and NLR. CA19-9, carbohydrate antigen 19-9; ECOG PS, Eastern Cooperative Oncology Group performance status; HR, hazard ratio; LAR, lactate dehydrogenase-to-albumin ratio; NLR, neutrophil-to-lymphocyte ratio.

Kaplan-Meier curves for the four groups defined by PS (0–1 vs. 2) and LAR (low vs high using the cohort median) demonstrated clear separation in OS (log-rank P<0.001), with the PS 2/high-LAR group showing the least favorable survival and the PS 0–1/low-LAR group showing the most favorable survival (Figure 3).

Figure 3 Kaplan-Meier curves for overall survival according to combined ECOG PS and baseline LAR. Patients were classified into four groups by PS (0–1 vs. 2) and LAR (low vs. high, dichotomized at the cohort median). Overall survival differed significantly among the four groups (log-rank P<0.001). Numbers at risk at 0, 6, 12, 18, and 24 months are shown below the plot. Curves are displayed up to 24 months for visual clarity. ECOG PS, Eastern Cooperative Oncology Group performance status; LAR, lactate dehydrogenase-to-albumin ratio.

In model comparisons, adding PS and LAR to the prespecified base model improved model fit and discrimination. Relative to the base model (AIC, 1,338.3; C-index, 0.607), the base + PS model (AIC, 1,331.2; C-index, 0.643) and the base + LAR model (AIC, 1,326.7; C-index, 0.636) showed more favorable performance. The combined base + PS + LAR model provided the best fit and discrimination among the evaluated models (AIC, 1,320.3; C-index, 0.662). In a nested comparison, adding LAR to the base + PS model further improved model fit (likelihood ratio test, P=0.002). To further examine whether LAR provided prognostic information beyond its individual components, additional models including LDH alone and albumin alone were also assessed; the base + LAR model showed a lower AIC than the base + LDH and base + albumin models (1,326.7 vs. 1,330.4 and 1,336.2, respectively) (Table 3).

Discussion

In this study of patients with metastatic or recurrent unresectable pancreatic cancer treated with first-line GnP, baseline LAR at treatment initiation was associated with OS independently of PS. In the multivariable Cox model, both PS and LAR remained independent prognostic factors. RCS analysis showed no evidence of nonlinearity, suggesting a broadly monotonic increase in risk with increasing LAR, and supporting its use as a continuous risk marker rather than one dependent on an arbitrary cutoff. For clinically meaningful stratification, patients were categorized into four groups by combining PS (0–1 vs. 2) and LAR (low vs. high using the cohort median), and Kaplan-Meier curves demonstrated clear separation in OS across the groups. In model comparisons, adding PS and LAR to the prespecified base model yielded the best fit and discrimination, as reflected by the lowest AIC and the highest C-index. The base + LAR model also showed a lower AIC than the base + LDH and base + albumin models. These findings support the potential value of LAR as a routinely available integrated marker that complements ECOG PS.

This study is characterized by its evaluation of the prognostic significance of LAR in a relatively homogeneous treatment cohort of patients with metastatic or recurrent unresectable pancreatic cancer who received first-line GnP. Recent meta-analytic evidence has supported the prognostic relevance of LAR across multiple malignancies (7). In gastrointestinal malignancies, LAR has also been evaluated in colorectal cancer and biliary tract cancer (13-15). However, most previous studies have relied on cohort-specific cutoff values, and studies examining the continuous association between LAR and survival, as in the present study, remain limited. Furthermore, data are even more limited in patients with metastatic or recurrent unresectable pancreatic cancer treated with first-line GnP. In the present study, LAR was associated with OS in the multivariable Cox analysis, and RCS showed no clear evidence of nonlinearity. These findings suggest that the association between LAR and mortality risk may be broadly monotonic and continuous, rather than dependent solely on a specific threshold. Evaluating LAR as a continuous variable using RCS may reduce information loss due to dichotomisation and better reflect the continuous nature of biological risk (16).

Biologically, LAR may represent an integrated marker that combines tumor-related metabolic activity, hypoxia, necrosis, and tumor burden reflected by LDH (17,18) with host-related inflammatory and nutritional vulnerability reflected by albumin (19,20). Consistent with this interpretation, the base + LAR model showed a lower AIC than the base + LDH and base + albumin models in the present study. This finding suggests that LAR may provide prognostic information beyond that captured by LDH or albumin alone.

The present study also examined the combination of LAR and ECOG PS. ECOG PS was significantly associated with OS in the multivariable Cox analysis, supporting its established prognostic importance. However, PS primarily reflects clinically assessed functional status, including general condition, activities of daily living, and self-care ability (21), and its assessment may be influenced by subjectivity and inter-rater variability (6). Therefore, patients within the same PS category may still differ in tumor-related metabolic activity, systemic inflammation, and nutritional vulnerability. These biological factors are not explicitly assessed by PS. LAR integrates information related to LDH and albumin and may therefore indirectly reflect biological backgrounds that are difficult to capture using PS alone. Thus, the contribution of the present study is not merely the confirmation of LDH, albumin, or PS as individual known prognostic factors, but the evaluation of LAR as a continuous and complementary prognostic marker to PS in a relatively homogeneous pancreatic cancer cohort treated with first-line GnP.

Nevertheless, several limitations of this study should be acknowledged. Owing to its retrospective single-center design, the possibility of selection bias and unmeasured confounding cannot be excluded; therefore, the observed associations do not establish causality. The analyses were restricted to complete cases for the prespecified covariates. Missing data primarily resulted from some laboratory measurements not being obtained as part of routine care at the time, rather than because patients were clinically too ill. Nevertheless, restricting the cohort to complete cases may limit generalizability and could introduce selection bias. Moreover, because OS was used as the outcome, the influence of subsequent therapies beyond first-line treatment could not be fully accounted for, and residual confounding due to temporal changes in treatment strategies and supportive care across the long accrual period cannot be ruled out. In addition, the number of patients with PS 2 was limited, resulting in imprecise estimates for some PS×LAR subgroups. In routine practice, systemic chemotherapy may be withheld in patients with PS 2 because of multiple factors, including impaired organ function, comorbidities, and social issues, and lower-intensity treatment regimens may be preferred. This clinical context may partly explain the small number of PS 2 cases in this cohort.

Conclusions

Baseline LAR was associated with OS independently of ECOG PS in patients receiving first-line GnP. PS combined with LAR may thus enable simple prognostic stratification at treatment initiation, pending future external validation.

Acknowledgments

The authors thank Helena A Popiel, PhD, for English language editing of the manuscript. The authors also acknowledge the use of Chat GPT (Open AI) from January 2026 to March 2026 for English-language refinement, assistance with manuscript wording, figure-formatting support, and limited assistance with revising R code. All AI-assisted output was reviewed and approved by the authors, who take full responsibility for the final content.

Footnote

Reporting Checklist: The authors have completed the STROBE reporting checklist. Available at https://jgo.amegroups.com/article/view/10.21037/jgo-2026-0223/rc

Data Sharing Statement: Available at https://jgo.amegroups.com/article/view/10.21037/jgo-2026-0223/dss

Peer Review File: Available at https://jgo.amegroups.com/article/view/10.21037/jgo-2026-0223/prf

Funding: None.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://jgo.amegroups.com/article/view/10.21037/jgo-2026-0223/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study protocol was approved by the Institutional Review Board of Tokyo Medical University (No. T2020-0422), and the requirement for individual informed consent was waived due to the retrospective nature of the study.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. National Cancer Institute. Cancer Stat Facts: Pancreatic Cancer Bethesda (MD): Surveillance, Epidemiology, and End Results (SEER) Program. Available online: https://seer.cancer.gov/statfacts/html/pancreas.html
2. Ohba A, Ozaka M, Mizusawa J, et al. Modified Fluorouracil, Leucovorin, Irinotecan, and Oxaliplatin or S-1, Irinotecan, and Oxaliplatin Versus Nab-Paclitaxel + Gemcitabine in Metastatic or Recurrent Pancreatic Cancer (GENERATE, JCOG1611): A Randomized, Open-Label, Phase II/III Trial. J Clin Oncol 2025;43:3345-54. [Crossref] [PubMed]
3. Bridgewater J, Lopes A, Wasan H, et al. Prognostic factors for progression-free and overall survival in advanced biliary tract cancer. Ann Oncol 2016;27:134-40. [Crossref] [PubMed]
4. Kawaguchi T, Takada M, Kubo A, et al. Performance status and smoking status are independent favorable prognostic factors for survival in non-small cell lung cancer: a comprehensive analysis of 26,957 patients with NSCLC. J Thorac Oncol 2010;5:620-30. [Crossref] [PubMed]
5. Sørensen JB, Klee M, Palshof T, et al. Performance status assessment in cancer patients. An inter-observer variability study. Br J Cancer 1993;67:773-5.
6. Chow R, Bruera E, Temel JS, et al. Inter-rater reliability in performance status assessment among healthcare professionals: an updated systematic review and meta-analysis. Support Care Cancer 2020;28:2071-8. [Crossref] [PubMed]
7. Chai D, Yang T, Zhang L, et al. Prognostic value of the lactate dehydrogenase to albumin ratio in cancer patients. Front Nutr 2025;12:1610487. [Crossref] [PubMed]
8. Choi N, Kim JH, Chie EK, et al. A meta-analysis of the impact of neutrophil-to-lymphocyte ratio on treatment outcomes after radiotherapy for solid tumors. Medicine (Baltimore) 2019;98:e15369. [Crossref] [PubMed]
9. Long Y, Zhang Y, Ni L, et al. Prognostic value of platelet-to-lymphocyte ratio in neoadjuvant chemotherapy for solid tumors: A PRISMA-compliant meta-analysis. Medicine (Baltimore) 2021;100:e26202. [Crossref] [PubMed]
10. Tong T, Guan Y, Xiong H, et al. A Meta-Analysis of Glasgow Prognostic Score and Modified Glasgow Prognostic Score as Biomarkers for Predicting Survival Outcome in Renal Cell Carcinoma. Front Oncol 2020;10:1541. [Crossref] [PubMed]
11. Ma X, Zou W, Sun Y. Prognostic Value of Pretreatment Controlling Nutritional Status Score for Patients With Pancreatic Cancer: A Meta-Analysis. Front Oncol 2021;11:770894. [Crossref] [PubMed]
12. Cheng H, Long F, Jaiswar M, et al. Prognostic role of the neutrophil-to-lymphocyte ratio in pancreatic cancer: a meta-analysis. Sci Rep 2015;5:11026. [Crossref] [PubMed]
13. Shu XP, Xiang YC, Liu F, et al. Effect of serum lactate dehydrogenase-to-albumin ratio (LAR) on the short-term outcomes and long-term prognosis of colorectal cancer after radical surgery. BMC Cancer 2023;23:915. [Crossref] [PubMed]
14. Wu J, Wu A, Wang S, et al. The value of lactate dehydrogenase to albumin ratio and immune inflammation biomarkers in colorectal cancer. Front Surg 2023;10:1118403. [Crossref] [PubMed]
15. Li G, He T, Geng M, et al. Prognostic significance of the preoperative lactate dehydrogenase-to-albumin ratio in patients undergoing radical resection for hilar cholangiocarcinoma. Front Med (Lausanne) 2025;12:1705110. [Crossref] [PubMed]
16. Altman DG, Royston P. The cost of dichotomising continuous variables. BMJ 2006;332:1080. [Crossref] [PubMed]
17. Van Wilpe S, Koornstra R, Den Brok M, et al. Lactate dehydrogenase: a marker of diminished antitumor immunity. Oncoimmunology 2020;9:1731942. [Crossref] [PubMed]
18. Forkasiewicz A, Dorociak M, Stach K, et al. The usefulness of lactate dehydrogenase measurements in current oncological practice. Cell Mol Biol Lett 2020;25:35. [Crossref] [PubMed]
19. Almasaudi AS, Dolan RD, Edwards CA, et al. Hypoalbuminemia Reflects Nutritional Risk, Body Composition and Systemic Inflammation and Is Independently Associated with Survival in Patients with Colorectal Cancer. Cancers (Basel) 2020;12:1986. [Crossref] [PubMed]
20. Robinson TP, Hamidi T, Counts B, et al. The impact of inflammation and acute phase activation in cancer cachexia. Front Immunol 2023;14:1207746. [Crossref] [PubMed]
21. Neeman E, Gresham G, Ovasapians N, et al. Comparing Physician and Nurse Eastern Cooperative Oncology Group Performance Status (ECOG-PS) Ratings as Predictors of Clinical Outcomes in Patients with Cancer. Oncologist 2019;24:e1460-6. [Crossref] [PubMed]