Predictive value of the prothrombin time-to-albumin ratio for 28- and 90-day mortality in lung cancer patients admitted to the intensive care unit: A retrospective cohort study.

1. Introduction

Lung cancer is a highly aggressive malignancy originating from the bronchial mucosa or pulmonary glands. It remains one of the fastest-growing malignancies in terms of incidence and mortality rates. [ 1 , 2 ] In most countries, the 5-year survival rate following diagnosis is limited to 10% to 20%, [ 3 ] posing a substantial threat to public health and imposing a significant burden on healthcare systems and society. As medical research advances, the identification of effective biomarkers and therapeutic targets has become critical for improving lung cancer outcomes.

Among various prognostic indicators, prothrombin time (PT) and albumin (Alb) have garnered increasing attention due to their essential roles in physiological homeostasis. PT serves as a key marker of coagulation function, while Alb not only maintains plasma colloid osmotic pressure but also plays a crucial role in nutritional status, immune modulation, and systemic homeostasis. Studies investigating the PT/Alb ratio (PAR) suggest that it may serve as a novel prognostic marker. Research has indicated its potential association with liver function and poor long-term prognosis in patients with hepatocellular carcinoma post-hepatectomy. [ 4 ] Additionally, hypoalbuminemia has been strongly correlated with lung cancer prognosis, particularly in non-small cell lung cancer, and is similarly implicated in various other solid tumors. [ 5 – 7 ] A 2023 multicenter case-control study using a public database identified a curvilinear relationship between PT and in-hospital mortality in critically ill patients with cancers. [ 8 ] Clinically, patients with lung cancer frequently exhibit coagulation dysfunction and metabolic disturbances, suggesting a potential interplay between PT, Alb, and disease progression. Although lung cancer prognosis is influenced by multiple factors, no studies have specifically explored the prognostic relevance of PAR in this context. A comprehensive analysis of the relationship between lung cancer and PAR may provide novel insights and facilitate the identification of potential biomarkers for early diagnosis, disease monitoring, and prognosis assessment, ultimately improving clinical outcomes.

While various Alb-based composite biomarkers have been investigated in oncology populations – including the Alb-globulin ratio in cervical cancer, [ 9 ] gastrointestinal cancers, [ 10 ] and non-small cell lung cancer [ 11 ] ; the Alb-to-fibrinogen ratio in multiple myeloma [ 12 ] and hepatocellular carcinoma [ 13 ] ; and the fibrinogen-to-Alb ratio in colorectal cancer [ 14 ] – no study to date has specifically evaluated the prognostic value of the PAR in critically ill patients with lung cancer. Unlike other Alb-based composites, PAR uniquely integrates 2 physiologically distinct domains: coagulation function (reflected by PT) and nutritional/inflammatory status (reflected by Alb). This dual aspect is particularly relevant in the intensive care unit (ICU) setting, where coagulopathy and hypoalbuminemia frequently coexist and independently contribute to poor outcomes. Therefore, PAR may capture synergistic prognostic information that single biomarkers or other composites cannot provide.

The prognosis of lung cancer patients admitted to the ICU differs substantially from that of general oncology cohorts. Recent meta-analyses have demonstrated that critically ill patients with lung cancer have pooled ICU mortality rates of 38% to 46%, substantially higher than expected in non-ICU oncology populations. [ 15 ] ICU patients represent a highly selected subgroup experiencing acute, life-threatening deteriorations due to cancer-related complications, treatment-related adverse events, or intercurrent critical illnesses. [ 15 ] In this context, short-term outcomes are driven more immediately by the severity of acute physiological derangement than by underlying cancer prognosis alone, underscoring the need for ICU-specific prognostic tools.

We hypothesized that a higher PAR would be associated with an increased risk of both 28- and 90-day all-cause mortality in this population. Clinically, an elevated PAR at ICU admission may serve as a simple, readily available bedside indicator of poor physiological status, potentially identifying patients who warrant more intensive monitoring or early intervention. Given that the requirement for mechanical ventilation, vasopressor use, and higher severity scores (Simplified Acute Physiology Score II [SAPS II], Sequential Organ Failure Assessment [SOFA]) are established predictors of mortality in this population, [ 15 ] we further hypothesized that PAR would demonstrate predictive performance comparable to or better than these established but more complex ICU scoring systems. This study aims to leverage data from the Medical Information Mart for Intensive Care IV (MIMIC-IV, version 2.2; MIT Laboratory for Computational Physiology (MIT LCP]) database to investigate the association between PAR and 28- and 90-day mortality in patients with lung cancer.

2. Methods

2.1. Data sources

The study data were sourced from the MIMIC-IV database, [ 16 ] curated and maintained by the MIT Computational Physiology Laboratory. This database encompasses comprehensive clinical information on 73,181 patients admitted to the ICU of Beth Israel Deaconess Medical Center between 2008 and 2019, including demographic details, vital signs, laboratory results, and relevant clinical parameters. The use of the MIMIC-IV database was approved by the Institutional Review Boards of Beth Israel Deaconess Medical Center and the Massachusetts Institute of Technology. As the database is deidentified and publicly available with appropriate data use agreements, the requirement for informed consent was waived. Access to MIMIC-IV requires completion of a qualifying exam and approval from the database administrators. A designated researcher from our team (Fengjuan Jia) successfully obtained authorization to extract data for research purposes (ID: 13274138).

2.2. Data extraction

Patients with a confirmed diagnosis of lung cancer were included in the study. For this study, ICU stays were used as the unit of analysis. Patients with lung cancer were identified using International Classification of Diseases codes recorded in hospital discharge summaries. For patients with multiple ICU admissions during the study period, each admission was treated as an independent observation. For the Cox proportional hazards models, the time scale was days from ICU admission. Patients were administratively censored at 28 and 90 days post-ICU admission for the respective outcome analyses. Mortality data in MIMIC-IV are derived from the Social Security Administration Death Master File and hospital records, providing complete follow-up for vital status within the study period. The inclusion criteria were a definitive diagnosis of lung cancer, age ≥18 years, and availability of complete Alb and PT data. Exclusion criteria were age <18 years and incomplete critical data.

The extracted variables encompassed demographic and hospitalization data: ICU and hospital admission counts, mortality status, date of death, gender, age, race, ICU admission time, and ICU length of stay. Vital signs included heart rate, mean arterial pressure (MAP), respiratory rate, body temperature, and pulse oximetry saturation (SpO 2 ). Laboratory parameters included Hematocrit, hemoglobin (Hb), platelet count, white blood cell count (WBC), lymphocyte count, neutrophil count, Alb, serum creatinine, blood glucose, PT, activated partial thromboplastin time (APTT), international normalized ratio (INR), alanine aminotransferase, alkaline phosphatase, aspartate aminotransferase, total bilirubin, and lactate. Disease severity scores included Logistic Organ Dysfunction System (LODS), Oxford Acute Severity of Illness Score (OASIS), SAPS II, SOFA, Glasgow Coma Scale, and Acute Physiology Score III (APS III). Sepsis status and comorbidities included myocardial infarction, congestive heart failure, peripheral vascular disease, cerebrovascular disease, dementia, chronic pulmonary disease, rheumatic disease, peptic ulcer disease, liver disease, diabetes, paralysis, kidney disease, malignancy, and acquired immune deficiency syndrome (AIDS). All selected data were derived from the first recorded measurements upon ICU admission.

The PAR was calculated as PT (in seconds) divided by serum Alb (in g/dL).

2.3. Study endpoint

The primary endpoint of this study was all-cause mortality at 28 and 90 days post-admission.

2.4. Data processing

Variables with more than 15% missing data were excluded from statistical analysis. No variables in this study exhibited a missing data rate between 5% and 15%. For continuous variables with missing data (<5% for all covariates), mean imputation was employed. We acknowledge that mean imputation may attenuate variance, but given the low proportion of missingness (<5%), the potential for substantial bias is minimal. This approach is consistent with common practice in clinical ICU studies using large databases.

2.5. Data analysis

All statistical analyses and visualizations were conducted using RStudio (version: 2024.09.1 + 394, Posit Software, PBC). Normally distributed variables were expressed as mean ± standard deviation (SD), while non-normally distributed variables were reported as median and interquartile range. Cox univariate and multivariate regression analyses were performed to assess the independent prognostic value of PAR, with hazard ratios (HRs) and 95% confidence intervals (CIs) reported. Variables for multivariable Cox regression models were selected based on univariable Cox regression analyses. Variables with P < .05 in univariable analysis were considered for inclusion in the multivariable models. To assess potential multicollinearity, variance inflation factors were calculated for all covariates in the final models. The proportional hazards assumption for all Cox models was assessed using Schoenfeld residuals. No significant violations were detected for the primary predictor PAR or for the global models (28-day mortality: χ 2 = 0.116, P = .73; 90-day mortality: χ 2 = 2.02, P = .16), confirming that the proportional hazards assumption was met ( Figures S1 and S2 , Supplemental Digital Content). Receiver operating characteristic (ROC) curve analysis was used to compare the predictive performance of PAR, PT, Alb, and established ICU severity scores (LODS, OASIS, SAPS II, SOFA, APS III) [ 17 , 18 ] for 28- and 90-day mortality in patients with lung cancer. For the fixed 28- and 90-day mortality outcomes, standard ROC analysis for binary classification was used to evaluate predictive performance, as there was no censoring within these specific time windows. The optimal PAR cutoff value was determined using the Youden index, stratifying patients into high-PAR and low-PAR groups. Kaplan–Meier survival analysis was applied to estimate event incidence rates across PAR groups, and differences were evaluated using the log-rank test. In the fully adjusted model, each 1-SD increase in PAR (SD ≈ 1.75) was associated with a 34% higher risk of 28- and 90-day mortality (HR = 1.34, 95% CI: 1.24–1.44, P < .001 for both). A P -value < .05 was considered statistically significant.

3. Results

3.1. Baseline characteristics

The MIMIC-IV v2.2 database included 73,181 patients, among whom 2855 were diagnosed with lung cancer and were ≥18 years old. As shown in Figure 1 , among 2855 initially identified patients, 1817 (63.6%) were excluded due to missing Alb and 79 (2.8%) due to missing PT, yielding a final cohort of 959 patients. The high proportion of missing Alb reflects real-world ICU practice, where this laboratory value is not universally obtained. At 28 days, 438 patients had died, while 521 survived. By 90 days, 589 patients had died, and 370 remained alive. The patient selection process, data preprocessing, and analysis workflow are illustrated in Figure 1 . Baseline characteristics for the total cohort, as well as stratifications by 28- and 90-day survival status, are presented in Tables 1 and 2 . PAR, defined as the ratio of PT to Alb, was identified as a significant predictor of 28- and 90-day mortality based on Cox univariate regression analysis, suggesting its potential as a superior prognostic indicator.

3.2. PAR is an independent risk factor for 28- and 90-day mortality

In the analysis of 28-day mortality, univariate Cox regression identified significant differences between survivors and non-survivors in several variables, including first ICU admission, presence of sepsis, heart rate, MAP, respiratory rate, SpO 2 , hematocrit, Hb, WBC, Alb, chloride, creatinine, INR, PT, APTT, and liver disease. Notably, PAR differed significantly between the 2 groups (Table 1 ). The variables included in the multivariable Cox models were selected based on univariable analyses ( P < .05), and no significant multicollinearity was detected (all variance inflation factors were <5). Table 3 presents the results of multivariate Cox regression analysis. In Model I, which adjusted for basic demographic factors (PAR, gender, race, and age), PAR was significantly associated with 28-day mortality (HR = 1.13; 95% CI: 1.10–1.17; P < .001). In Model II, after adjusting for additional covariates, PAR remained an independent predictor (HR = 1.18; 95% CI: 1.13–1.23; P < .001).

For the analysis of 90-day mortality, univariate Cox regression revealed significant differences between survivors and non-survivors in factors such as first ICU admission, presence of sepsis, heart rate, MAP, respiratory rate, body temperature, SpO 2 , hematocrit, Hb, WBC, Alb, chloride, INR, PT, APTT, aspartate aminotransferase, rheumatic disease, and liver disease (Table 2 ). PAR also differed significantly between the groups. Variables with a P -value < .05 were included in the multivariate Cox proportional hazards model. As shown in Table 4 , Model I, which accounted for PAR, gender, race, and age, demonstrated a significant association between PAR and 90-day mortality (HR = 1.13; 95% CI: 1.10–1.17; P < .001). In Model II, after adjusting for confounding variables, PAR remained an independent prognostic factor (HR = 1.18; 95% CI: 1.13–1.23; P < .001).

3.3. ROC curve analysis

To further validate the predictive performance of PAR, ROC curves were generated for PAR, PT, Alb, and commonly used ICU scoring systems to assess 28-day mortality in patients with lung cancer (Figs. 2 and 3 ; Table 5 ). For 28-day mortality, ROC curve analysis revealed that the PAR score yielded an AUC of 0.694 (95% CI: 0.6606–0.7267). According to DeLong tests, PAR demonstrated significantly superior predictive performance compared with PT, OASIS, and SOFA. In contrast, its discriminative ability was statistically comparable to that of LODS, SAPS II, and APS III. Detailed AUC estimates and corresponding pairwise comparison P -values are provided in Table 5 .

Similarly, for 90-day mortality, the PAR score achieved an AUC of 0.703 (95% CI: 0.6689–0.7366). Consistent with the 28-day findings, PAR showed significantly better predictive accuracy than PT, OASIS, and SOFA, while performing similarly to LODS, SAPS II, and APS III (Figs. 4 and 5 ; Table 6 ).

3.4. Clinical outcomes and Kaplan–Meier curves

The optimal PAR threshold was determined using the maximum Youden index derived from the ROC curve. Based on this critical value, patients were stratified into high- and low-PAR groups, as detailed in Tables 7 and 8 . For 28-day mortality, significant associations were observed between PAR classification and first ICU admission status, as well as 28-day mortality, with both showing statistically significant differences. Although ICU length of stay and total hospitalization duration were longer in the high-PAR group than in the low-PAR group, the differences were not statistically significant (Table 7 ). Kaplan–Meier survival analysis further confirmed that 28-day mortality was significantly higher in the high-PAR group compared with the low-PAR group ( P < .001, Fig. 6 A).

For 90-day mortality, significant associations were also identified between PAR classification and first ICU admission status, ICU length of stay, and 90-day mortality, all of which exhibited statistically significant differences. While the high-PAR group had a longer total hospital stay than the low-PAR group, this difference did not reach statistical significance (Table 8 ). Kaplan–Meier survival analysis similarly demonstrated a significantly higher 90-day mortality rate in the high-PAR group ( P < .001, Fig. 6 B).

4. Discussion

Early-stage lung cancer is typically asymptomatic, leading to delayed diagnosis, particularly among critically ill patients admitted to the ICU. Currently, no reliable prognostic model exists to accurately predict outcomes in these patients, underscoring the clinical necessity of identifying robust prognostic indicators. This study utilized data from the MIMIC-IV database to investigate potential biomarkers for the early and rapid assessment of lung cancer prognosis. Univariate and multivariate Cox regression analyses demonstrated that PAR is an independent predictor of mortality. ROC curve analysis showed that PAR had moderate predictive accuracy for 28- and 90-day mortality (AUCs 0.694 and 0.703, respectively). Notably, PAR significantly outperformed OASIS and SOFA at both time points (all P < .05), and showed predictive performance comparable to LODS, SAPS II, and APS III (all P > .05). Although APS III exhibited a numerically higher AUC for 28-day mortality, this difference was not statistically significant ( P = .347). Kaplan–Meier survival analysis confirmed that patients with high PAR had significantly higher mortality rates. Additionally, a comparative analysis of ICU and hospital length of stay between high-PAR and low-PAR groups was conducted. While the high-PAR group exhibited a significantly higher 28-day mortality rate, no statistically significant association was observed between PAR and length of hospital stay or ICU stay. However, a potential correlation cannot be ruled out, necessitating further investigation to elucidate the relationship between PAR and clinical outcomes.

PT serves as a key indicator of coagulation function, with prolonged PT reflecting deficiencies or dysfunctions in coagulation factors. It assesses the extrinsic pathway of blood coagulation by measuring plasma exposure to tissue factor. Prolonged PT has been strongly associated with adverse outcomes across multiple malignancies, including colorectal cancer, [ 19 ] lung cancer, [ 20 ] and hepatocellular carcinoma. [ 21 ] During the COVID-19 pandemic, an analysis of early clinical characteristics in patients with lung cancer with COVID-19 identified prolonged PT and elevated high-sensitivity cardiac troponin I as independent risk factors for in-hospital mortality. [ 22 ] Additionally, in patients with lung cancer experiencing thromboembolic complications, PT was found to be progressively prolonged in the mortality group, with a significant correlation between poor prognosis and PT extension. In NSCLC, coagulation abnormalities are frequently observed, and coagulation parameters have been proposed as prognostic markers for survival. [ 23 ] A meta-analysis further confirmed that patients with lung cancer exhibit significantly prolonged PT compared with healthy controls, with PT extension serving as a hallmark of lung cancer and an indicator of poor prognosis. [ 24 ] The underlying mechanisms by which prolonged PT influences tumor progression may involve coagulation factor depletion, hepatic biosynthetic dysfunction, and coagulation system activation. [ 25 ] In lung cancer, activation of the coagulation-fibrinolytic system is thought to be partially mediated by increased neutrophil elastase release, a process associated with disease stage and primarily implicated in NSCLC. This mechanism contributes to tumor progression through inflammation and coagulation-driven tumor interactions. [ 26 ] Furthermore, tumor-induced dysregulation of the coagulation system may directly influence PT. Procoagulant substances secreted by tumor cells can activate the coagulation cascade, resulting in PT alterations. These disruptions compromise normal blood flow, elevate thrombosis risk, impair vital organ perfusion, and ultimately contribute to increased mortality in patients with lung cancer.

Alb is the most abundant plasma protein, serving as both a key marker of nutritional status and a critical factor in tumor progression. In patients with advanced lung cancer, hypoalbuminemia contributes to tissue edema, impairs cellular oxygenation and nutrient exchange, weakens immune function, and increases susceptibility to complications such as infections, thereby elevating mortality risk. A study by Gul et al, examining the relationship between nutritional status and lung cancer prognosis identified poor nutritional status as a major determinant of poor outcomes in patients with advanced disease. [ 27 ] Similarly, hypoalbuminemia has been linked to significantly worse prognosis in patients with colon and cervical cancers compared with those with normal Alb levels. [ 28 – 30 ] Additionally, Alb has been proposed as an indicator of invasive potential in laryngeal cancer. [ 31 ] A prognostic scoring system incorporating Alb and the neutrophil-to-lymphocyte ratio has demonstrated enhanced stratification capability in lung cancer, providing a potential tool for guiding clinical decisions, particularly in patients with small cell lung cancer receiving immune checkpoint inhibitors. [ 32 ] Beyond its role as a marker of systemic inflammation, [ 33 ] Alb influences chemotherapy efficacy and tumor metabolism through its binding to chemotherapeutic agents and fatty acids. [ 34 ] However, conflicting findings exist. A retrospective study of 378 patients with small cell lung cancer by Tas et al found that neither Alb nor BMI exhibited prognostic significance among chemotherapy-treated patients, though this may be attributed to the study’s relatively small sample size. [ 35 ]

Lung cancer is associated with poor prognosis, with disease stage remaining the primary prognostic indicator in clinical practice. Beyond staging, few reliable biomarkers have been identified. However, emerging evidence suggests that Alb-based composite indices hold prognostic value. For instance, the neutrophil-to-Alb ratio has been shown to provide prognostic insights in patients receiving palliative care for advanced lung cancer. [ 36 ] Similarly, the Alb-to-globulin ratio has been identified as an independent protective factor in patients with advanced NSCLC treated with anlotinib. [ 37 ] A meta-analysis further supports the Alb-to-fibrinogen ratio as a predictor of cancer prognosis. [ 38 ] Studies investigating novel indices that integrate PT and Alb have demonstrated their predictive utility in other diseases. For example, the product of serum Alb and PT has been found to effectively assess the severity of acute pancreatitis, particularly in cases involving organ failure. [ 39 ] Additionally, the PT–international normalized ratio/Alb ratio has been proposed as a marker of liver function and a predictor of poor long-term prognosis following liver resection in patients with hepatocellular carcinoma. [ 4 ]

Currently, no established biomarker or scoring system reliably predicts ICU mortality in patients with lung cancer, and no prior study has evaluated the prognostic significance of a combined PT-to-Alb index in this population. Most research has focused on the independent associations of PT or Alb with disease prognosis. This study, however, sought to assess the prognostic value of PAR in ICU patients with lung cancer, and the findings support its potential utility for early risk stratification. By integrating nutritional status and coagulation function, PAR offers a more comprehensive reflection of physiological dysfunction in critically ill patients with lung cancer. From a clinical perspective, identifying a high PAR at ICU admission may serve as a simple bedside “red flag” indicating poor physiological reserve and increased mortality risk. This could prompt clinicians to conduct more comprehensive evaluations for underlying causes of coagulopathy and malnutrition, intensify monitoring frequency, or consider earlier involvement of multidisciplinary teams. The potential clinical utility of PAR is supported by decision curve analysis ( Figures S3 and S4 , Supplemental Digital Content). Nevertheless, this study has several limitations. Although the MIMIC-IV database provides extensive clinical data, missing data remain a concern, particularly regarding detailed tumor molecular biology information, cancer stage, histological subtypes, and specific anticancer treatments. This limitation hinders a deeper analysis of the relationship between PAR and specific molecular features of lung cancer, preventing exploration of potential mechanistic pathways. A total of 63.6% of patients were excluded from the present study due to missing Alb data. This phenomenon is consistent with real-world clinical practice in the ICU, in which Alb measurement is not routinely performed for all patients. Furthermore, direct comparison between included and excluded patients was not feasible, as the excluded cases lacked the primary exposure variable (PAR) and essential clinical covariates. This represents an inherent limitation of our study. The missing laboratory data were nonrandom in nature, and our study findings are therefore only generalizable to the patient cohort with complete PAR measurements. Although we modeled PAR linearly in the Cox regression, we did not formally explore potential nonlinear relationships, which should be investigated in future studies. Additionally, as a retrospective observational study, despite efforts to control for confounding factors through multivariate analysis, a causal relationship between PAR and lung cancer mortality cannot be definitively established as in a prospective randomized controlled trial. Thus, the findings suggest an association rather than causation, necessitating further prospective validation. Also, although the MIMIC-IV database is a large and well-characterized critical care dataset, it derives from a single-center healthcare system in the United States, which may restrict the generalizability of our findings to other populations and clinical settings. Further multicenter studies are therefore warranted to validate our results. Moreover, the lack of dynamic PAR monitoring precludes an analysis of its temporal trends and their impact on prognosis, representing another limitation of this study. Additionally, ICU stays rather than unique patients were used as the unit of analysis, and approximately 6.8% of admissions were readmissions. A sensitivity analysis limited to first ICU stays or clustering by patient was not performed, which is a limitation. However, the proportion of readmissions was low, and our primary objective was to reflect real-world ICU practice.

In conclusion, PAR is a practical prognostic indicator for lung cancer, showing a moderate association with 28- and 90-day mortality in critically ill patients, and may be useful for early risk stratification. Incorporating PAR into routine clinical assessments may enhance prognostic evaluation, enabling a more comprehensive understanding of patients’ physiological status and facilitating personalized treatment strategies. For patients with elevated PAR, targeted interventions such as enhanced nutritional support, coagulation function monitoring, and therapeutic optimization may contribute to improved outcomes. Future research should focus on prospective randomized controlled trials to establish a causal relationship between PAR and lung cancer mortality. Additionally, large-scale multicenter studies encompassing diverse populations and geographic regions would enhance the generalizability of findings. Investigating the association between PAR and lung cancer molecular biology may also uncover novel therapeutic targets and biomarkers, providing a foundation for precision medicine. Furthermore, the development of dynamic PAR monitoring methods could enable real-time assessment of disease progression, allowing for timely adjustments in treatment strategies.

5. Conclusion

In conclusion, PAR is a simple, inexpensive, and readily available biomarker with moderate predictive value for short-term mortality in critically ill patients with lung cancer. Its performance is comparable to established ICU severity scores, supporting its potential use as a tool for early risk stratification. Decision curve analysis further supports its clinical utility, suggesting that PAR may serve as a practical bedside prognostic tool. Given these promising findings, prospective multicenter validation is warranted to confirm its clinical applicability and to determine whether PAR-guided risk assessment can ultimately improve patient outcomes.

Acknowledgments

We would like to express our thanks to all the investigators and participants.

Author contributions

Conceptualization: Guodong Xia.

Funding acquisition: Tao Xu, Xianying Lei.

Investigation: Fengjuan Jia.

Resources: Fengjuan Jia.

Supervision: Guodong Xia.

Writing – original draft: Fengjuan Jia.

Writing – review & editing: Tao Xu, Lirong Hu, Songlin Wu, Li Liu, Xianying Lei.

Abbreviations: AIDS acquired immune deficiency syndrome Alb albumin APS III Acute Physiology Score III APTT activated partial thromboplastin time CI confidence interval Hb hemoglobin HR hazard ratio ICU intensive care unit INR international normalized ratio LODS Logistic Organ Dysfunction System MAP mean arterial pressure MIMIC-IV the Medical Information Mart for Intensive Care IV OASIS Oxford Acute Severity of Illness Score PAR prothrombin time/albumin ratio PT prothrombin time ROC receiver operating characteristic SAPS II Simplified Acute Physiology Score II SD standard deviation SOFA Sequential Organ Failure Assessment SpO 2 = pulse oximetry saturation WBC white blood cell count

This work was supported by the Sichuan Science and Technology Program (2022YFS0632) and the Luzhou Science and Technology Bureau (2024SYF164). Funding related to 2022YFS0632 was received by Xianying Lei. Funding related to 2024SYF164 was received by Tao Xu.

The data utilized in this study were obtained following ethical approval from the administrators of MIMIC-IV.

The authors have no conflicts of interest to disclose.

The datasets generated and/or analyzed during the current study are not publicly available, but are available from the corresponding author on reasonable request.

Supplemental Digital Content is available in the online version of this article ( http://dx.doi.org/10.1097/MD.0000000000049035 ).

How to cite this article: Jia F, Xia G, Xu T, Hu L, Wu S, Liu L, Lei X. Predictive value of the prothrombin time-to-albumin ratio for 28- and 90-day mortality in lung cancer patients admitted to the intensive care unit: A retrospective cohort study. Medicine 2026;105:22(e49035).

Contributor Information

Fengjuan Jia, Email: jiafengjuan@swmu.edu.cn.

Guodong Xia, Email: 894242130@qq.com.

Tao Xu, Email: daishaoye2239@sina.com.

Lirong Hu, Email: 254884552@qq.com.

Songlin Wu, Email: wslin919@163.com.

Li Liu, Email: niuniudoctor@swmu.edu.cn.

References

[1] Howlader N, Forjaz G, Mooradian MJ, et al. The effect of advances in lung-cancer treatment on population mortality. N Engl J Med. 2020;383:640–9.

[2] Feng RM, Zong YN, Cao SM, Xu RH. Current cancer situation in China: good or bad news from the 2018 Global Cancer Statistics? Cancer Commun (Lond). 2019;39:22.

[3] Sung H, Ferlay J, Siegel RL, et al. Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin. 2021;71:209–49.

[4] Haruki K, Shiba H, Saito N, et al. Risk stratification using a novel liver functional reserve score of combination prothrombin time-international normalized ratio to albumin ratio and albumin in patients with hepatocellular carcinoma. Surgery. 2018;164:404–10.

[5] Miura K, Hamanaka K, Koizumi T, et al. Clinical significance of preoperative serum albumin level for prognosis in surgically resected patients with non-small cell lung cancer: comparative study of normal lung, emphysema, and pulmonary fibrosis. Lung Cancer. 2017;111:88–95.

[6] Yang JR, Xu JY, Chen GC, et al. Post-diagnostic C-reactive protein and albumin predict survival in Chinese patients with non-small cell lung cancer: a prospective cohort study. Sci Rep. 2019;9:8143.

[7] Zhang CL, Gao MQ, Jiang XC, et al. Research progress and value of albumin-related inflammatory markers in the prognosis of non-small cell lung cancer: a review of clinical evidence. Ann Med. 2023;55:1294–307.

[8] Liang JD, Qin ZA, Yang JH, et al. Association between PT, PT-INR, and in-hospital mortality in critically ill patients with tumors: a retrospective cohort study. Front Public Health. 2023;11:1036463.

[9] Wang Z, Lin J, Chen D. A meta-analysis of albumin, globulin, and albumin globulin ratios for predicting prognosis of cervical cancer. Eur J Cancer Prev. 2025;34:337–48.

[10] Delladio W, Barroca C, Roberts WS, Nguyen H. The effectiveness of albumin-to-globulin ratio as a prognostic biomarker in primary gastrointestinal cancer: a systematic review and meta-analysis. Cureus. 2024;16:e61830.

[11] Li Z, Li C, You R, et al. Longitudinal investigation of albumin-to-globulin ratio for human cancers demonstrates benefit in postoperative serial remeasurement. NPJ Precis Oncol. 2025;9:23.

[12] Carlisi M, Presti RL, Pirrera A, Augello AE, Siragusa S, Caimi G. Prognostic potential of albumin to fibrinogen ratio in multiple myeloma patients. Ann Hematol. 2025;104:5881–9.

[13] Lapsiwala B, Gurram P, Vanani S, et al. S1723 prognostic role of fibrinogen-albumin ratio in patients with hepatocellular carcinoma: a systematic review and meta-analysis. Off J Am Coll Gastroenterol. 2024;119:S1250–1.

[14] Shimada M, Goi T, Tagai T, et al. Fibrinogen-albumin ratio: an independent recurrence factor for colorectal cancer stages II and III in a competing risk model. J Anus Rectum Colon. 2025;9:321–9.

[15] Elkhapery A, Taifour H, Niu C, Soubani AO. Prognosis of patients with lung cancer admitted to the intensive care unit: a systematic review and meta-analysis. J Intensive Care Med. 2025:8850666251339451.

[16] Johnson AEW, Bulgarelli L, Shen L, et al. MIMIC-IV, a freely accessible electronic health record dataset. Sci Data. 2023;10:1.

[17] DeLong ER, DeLong DM, Clarke-Pearson DL. Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach. Biometrics. 1988;44:837–45.

[18] Heagerty PJ, Lumley T, Pepe MS. Time-dependent ROC curves for censored survival data and a diagnostic marker. Biometrics. 2000;56:337–44.

[19] Lee S, Huh SJ, Oh SY, et al. Clinical significance of coagulation factors in operable colorectal cancer. Oncol Lett. 2017;13:4669–74.

[20] Tas F, Kilic L, Serilmez M, Keskin S, Sen F, Duranyildiz D. Clinical and prognostic significance of coagulation assays in lung cancer. Respir Med. 2013;107:451–7.

[21] Wang XP, Mao MJ, He ZL, et al. A retrospective discussion of the prognostic value of combining prothrombin time(PT) and fibrinogen(Fbg) in patients with Hepatocellular carcinoma. J Cancer. 2017;8:2079–87.

[22] Nie L, Dai K, Wu J, et al. Clinical characteristics and risk factors for in-hospital mortality of lung cancer patients with COVID-19: a multicenter, retrospective, cohort study. Thorac Cancer. 2021;12:57–65.

[23] Qi Y, Fu J. Research on the coagulation function changes in non small cell lung cancer patients and analysis of their correlation with metastasis and survival. J BUON. 2017;22:462–7.

[24] Bayleyegn B, Adane T, Getawa S, Aynalem M, Kifle ZD. Coagulation parameters in lung cancer patients: a systematic review and meta-analysis. J Clin Lab Anal. 2022;36:e24550.

[25] Zhang K, Gao HF, Mo M, et al. A novel scoring system based on hemostatic parameters predicts the prognosis of patients with advanced pancreatic cancer. Pancreatology. 2019;19:346–51.

[26] Gabazza E, Taguchi O, Yamakami T, Machishi M, Ibata H, Suzuki S. Correlation between increased granulocyte elastase release and activation of blood coagulation in patients with lung cancer. Cancer. 1993;72:2134–40.

[27] Gul B, Metintas S, Ak G, Yilmaz S, Metintas M. The relationship between nutritional status and prognosis in patients with locally advanced and advanced stage lung cancer. Support Care Cancer. 2021;29:3357–65.

[28] Gupta A, Gupta E, Hilsden R, et al. Preoperative malnutrition in patients with colorectal cancer. Can J Surg. 2021;64:E621–9.

[29] Yoshikawa N, Yoshihara M, Tamauchi S, Ikeda Y, Yokoi A, Kajiyama H. Hypoalbuminemia for the prediction of survival in patients with stage IVB cervical cancer. PLoS One. 2022;17:e0273876.

[30] Wang X, Xu J, Zhang H, Qu P. The effect of albumin and hemoglobin levels on the prognosis of early-stage cervical cancer: a prospective, single-center-based cohort study. BMC Womens Health. 2023;23:553.

[31] Shen LF, Wang QY, Yu Q. The systemic immune-inflammation index and albumin as prognostic predictors in laryngeal carcinoma. Nutr Cancer. 2021;73:1916–23.

[32] Kuang Z, Miao J, Zhang X. Serum albumin and derived neutrophil-to-lymphocyte ratio are potential predictive biomarkers for immune checkpoint inhibitors in small cell lung cancer. Front Immunol. 2024;15:1327449.

[33] Soeters PB, Wolfe RR, Shenkin A. Hypoalbuminemia: pathogenesis and clinical significance. JPEN J Parenter Enteral Nutr. 2019;43:181–93.

[34] Yu L, Hua Z, Luo X, Zhao T, Liu Y. Systematic interaction of plasma albumin with the efficacy of chemotherapeutic drugs. Biochim Biophys Acta Rev Cancer. 2022;1877:188655.

[35] Tas F, Ozturk A, Erturk K. Prognostic significance of body mass index and serum albumin as the indicators of nutritional status in small cell lung cancer. Postgrad Med. 2024;136:208–17.

[36] Feng C, Yu H, Lei H, Cao H, Chen M, Liu S. A prognostic model using the neutrophil-albumin ratio and PG-SGA to predict overall survival in advanced palliative lung cancer. BMC Palliat Care. 2022;21:81.

[37] Kut E, Menekse S. Prognostic significance of pretreatment albumin-bilirubin (ALBI) grade and platelet-albumin-bilirubin (PALBI) grade in patients with small cell lung cancer. Sci Rep. 2024;14:1371.

[38] Sun DW, An L, Lv GY. Albumin-fibrinogen ratio and fibrinogen-prealbumin ratio as promising prognostic markers for cancers: an updated meta-analysis. World J Surg Oncol. 2020;18:9.

[39] Li S, Liu Z, Wu H. The product value of serum albumin and prothrombin time activity could be a useful biomarker for severity prediction in AP: an ordinal retrospective study. Pancreatology. 2019;19:230–6.

Associated Data

Supplementary Materials