Dynamic inflammation-based prognostication in acute-on-chronic liver failure: The COSSH-CAR model as a step forward in personalized risk stratification

Abstract

Acute-on-chronic liver failure (ACLF) is a swiftly deteriorating condition characterized by profound systemic inflammation and failure of multiple organ systems, leading to high early mortality. There remains a critical need for more effective biomarkers to facilitate timely and accurate risk assessment. Recent findings by Zhu and Yan demonstrated that evaluating temporal changes in the C-reactive protein to albumin ratio (CAR), especially the 7-day variation, offers superior prediction of 28-day mortality compared with single baseline measurements. By integrating the 7-day variation of CAR with the model for end-stage liver disease sodium score and the grade of hepatic encephalopathy, the Chinese Group on Study of Severe Hepatitis B (COSSH)-CAR model was created, which surpassed traditional prognostic tools such as the Child-Pugh, model for end-stage liver disease, and COSSH-ACLF. This comment highlights the importance of using dynamic biomarker trajectories rather than static values for prognostic evaluation. CAR is biologically compelling because it captures both the inflammatory burden and the patient’s nutritional/physiological reserve. While the COSSH-CAR model is promising and based on routinely obtainable laboratory data, its widespread adoption will depend on validation in larger, diverse, and non-hepatitis B virus-related cohorts. Future work should examine CAR kinetics in prospective and interventional studies and consider how they may support individualized management strategies. Collectively, these observations suggest that the CAR could represent an important addition to current ACLF prognostic frameworks.

Key Words: Acute-on-chronic liver failure; C-reactive protein to albumin ratio (CAR); Chinese Group on Study of Severe Hepatitis B-CAR; Dynamic biomarkers; Systemic inflammation; Hepatic encephalopathy; Risk stratification

Core Tip: The dynamic shift in the C-reactive protein to albumin ratio (CAR) during the initial week provides additional prognostic insight into Chinese Group on Study of Severe Hepatitis B-defined acute-on-chronic liver failure, surpassing the predictive capacity of traditional indices such as the model for end-stage liver disease sodium score and hepatic encephalopathy grade. Embedding 7-day variation of CAR within the Chinese Group on Study of Severe Hepatitis B-CAR score yields a biologically plausible, practical bedside instrument. Nonetheless, confirmation through robust, multicenter validation is essential before widespread clinical application can be recommended.

INTRODUCTION

Acute-on-chronic liver failure (ACLF), which is defined by abrupt organ failure and exceptionally high short-term mortality rates, is considered the most catastrophic manifestation of chronic liver disease decompensation. International guidelines consistently stress its acute and emergent presentation[1,2]. The Asian Pacific Association for the Study of the Liver (APASL) defines ACLF as an acute hepatic insult leading to jaundice and coagulopathy, which is complicated within 4 weeks by ascites and/or hepatic encephalopathy in individuals with preexisting chronic liver disease[3]. Similarly, the European Association for the Study of the Liver (EASL)-Chronic Liver Failure (CLIF) Consortium (CLIF-C) describes ACLF as the “most severe form of acutely decompensated cirrhosis”, characterized by multiorgan failure and profound systemic inflammation[4,5].

Geographic variations in etiology are well recognized. In East Asia, ACLF often develops following hepatitis B virus reactivation or superinfection, whereas in Western populations, it more commonly arises from alcoholic hepatitis or severe infections[6,7]. Irrespective of cause, patients exhibit marked systemic inflammation - reflected in elevated cytokines, leukocytosis, and increased C-reactive protein (CRP) - together with metabolic disturbances that precipitate organ dysfunction[1]. For example, the PREDICT study demonstrated that inflammatory triggers such as infection, alcohol misuse, or bleeding are strongly associated with increased CRP levels and leukocyte counts in ACLF patients. This evidence supports the paradigm that systemic inflammation is central to ACLF pathophysiology and that its biomarkers may hold prognostic importance[8].

Accurate prediction of outcomes in patients with ACLF is clinically critical for patient triage, prioritization of limited resources, and timely evaluation for liver transplantation. While traditional tools such as the Child-Pugh score and model for end-stage liver disease (MELD)/MELD sodium (MELD-Na) score provide useful prognostic guidance, their predictive precision remains limited. More refined, disease-specific scores have since been proposed. Compared with conventional models, the CLIF-C ACLF score, which incorporates multiple failure indices, and the Chinese Group on Study of Severe Hepatitis B (COSSH)-ACLF score, which is tailored for hepatitis B virus (HBV)-related ACLF, both offer improved accuracy[9]. However, these approaches largely rely on baseline or single time-point assessments. Given the inherently dynamic trajectory of ACLF, increasing attention has turned to serial biomarker trends as potential predictors of outcomes[10,11].

In this context, the recent study by Zhu and Yan[9] is notable for evaluating the prognostic role of changes in the CRP to albumin ratio (CAR) during the first week of hospitalization. Their findings highlight the change in CAR from baseline to day 7 (ΔCAR-7) as a robust predictor of 28-day survival. By combining the ΔCAR-7 with the MELD-Na score and hepatic encephalopathy grade, the authors proposed a novel COSSH-CAR model, which outperformed established prognostic systems[12]. In context, conventional ACLF prognostic models such as MELD-Na and CLIF-C ACLF primarily depend on static measures of organ function, whereas basic inflammatory markers such as the neutrophil-lymphocyte ratio (NLR) have been linked to patient mortality. In contrast, the COSSH-CAR model distinctively integrates the dynamic CAR with established hepatic scoring parameters, offering a more responsive and temporally updated assessment of disease severity and prognosis[13].

In this editorial, we critically analyze these results. We outline the study’s methodological innovations, discuss the mechanistic rationale linking CRP-albumin dynamics to disease progression and evaluate implications for both clinical decision-making and future research. We underscore the advantages of the COSSH-CAR model - its reliance on objective laboratory data, dynamic assessment, and strong discriminatory power - while also acknowledging its limitations, including its retrospective single-etiology design. Finally, we contextualize the CAR within the broader landscape of emerging biomarkers and personalized medicine approaches in ACLF, suggesting further validation and integration into clinical practice[12,14].

KEY INSIGHTS AND INNOVATIONS

The principal contribution of Zhu and Yan[9] lies in their emphasis on dynamic monitoring of the CAR rather than relying on a single admission value. Historically, most prognostic models for liver failure have been derived from static laboratory or clinical variables obtained at baseline. By contrast, this study incorporated daily CAR measurements over the first week of hospitalization and quantified the change from baseline. Among these factors, the ΔCAR-7 emerged as the most informative predictor of short-term outcomes. This timepoint appears to capture the early trajectory of the host response to ACLF, distinguishing patients whose inflammation is resolving from those in whom it continues to escalate. As the authors highlight, ACLF evolves rapidly, and an early trend may be more prognostically meaningful than an isolated baseline value. Compared with the baseline CAR or alternative timepoints, the ΔCAR-7 demonstrated superior predictive power for 28-day mortality. These findings align with an emerging view of ACLF as a dynamic disease process; supporting evidence comes from recent work showing that serial changes in bilirubin, creatinine, and other laboratory indices (as in the dynamic prediction [DP]-ACLF) outperform static scores in predicting outcomes. Thus, the integration of the ΔCAR-7 reinforces a growing paradigm in which sequential assessment enhances risk stratification in patients with ACLF[8,12,15].

A second major innovation is the development of the COSSH-CAR model, a prognostic tool specifically designed for HBV-related ACLF. Using multivariate logistic regression, the authors combined three components: (1) ΔCAR-7 score; (2) Baseline MELD-Na score; and (3) Hepatic encephalopathy grade. Each element has established biological and clinical relevance: MELD-Na is a globally recognized index of hepatic function on the basis of bilirubin, creatinine, international normalized ratio (INR), and sodium levels, whereas hepatic encephalopathy reflects both liver and neurological compromise. By incorporating the ΔCAR-7, the model introduces a dynamic inflammatory biomarker into an otherwise static framework[8,16].

The COSSH-CAR score was rigorously validated against patient outcomes and directly compared with established prognostic tools. Notably, it showed significantly greater discriminative ability (higher area under the curve) than the Child-Pugh score, MELD-Na score, and even the COSSH-ACLF score. This is particularly striking given that prior studies have demonstrated that the COSSH-ACLF score itself is highly robust. For example, Tong et al[10] reported a 28-day area under the receiver operating characteristic curve of approximately 0.807. Surpassing this benchmark underscores the incremental value of including dynamic CAR measurements[12].

In summary, Zhu and Yan[10] offers several key innovations: (1) Dynamic CAR monitoring - shows that the ΔCAR-7 is a stronger predictor of mortality than the baseline CAR; (2) Composite modeling – integrating the ΔCAR-7 with the MELD-Na score and hepatic encephalopathy grade into the COSSH-CAR model; (3) Enhanced predictive performance – demonstrated improved discrimination and clinical utility compared with widely used scores; and (4) Etiology-specific focus – addressing the prognostic gap in HBV-predominant ACLF populations through a tailored tool[10,12].

These advances illustrate the value of incorporating serial biomarker data into prognostic frameworks. In addition to similar efforts, such as the DP-ACLF model, this work strengthens the case that longitudinal monitoring captures ACLF trajectories more accurately than static assessments do. Collectively, these insights represent an important step toward more precise and individualized prognostication in patients with ACLF[8].

MECHANISM

The prognostic value of the dynamic CAR can be better understood by considering the underlying pathophysiology of ACLF. This syndrome is characterized by a profound systemic inflammatory response superimposed on advanced liver dysfunction, which collectively drives multiorgan failure. CRP and albumin reflect distinct but complementary aspects of this process.

CRP is a prototypical positive acute-phase protein synthesized by hepatocytes in response to proinflammatory cytokines, particularly interleukin 6 (IL-6). Elevated CRP mirrors the magnitude of systemic inflammation and infection. Moreau et al[12] demonstrated that patients with ACLF exhibit markedly higher CRP levels than other cirrhotic populations do, underscoring the heightened inflammatory milieu. In clinical practice, bacterial infections - commonly precipitants of ACLF - trigger surges in CRP, and persistent or rising levels correlate with complications and poor survival in patients with cirrhosis. Thus, an increasing CRP trajectory during the first week of hospitalization likely reflects uncontrolled inflammation or new infection, both of which portend unfavorable outcomes[1,17].

By contrast, albumin is a negative acute-phase reactant and serves as a marker of hepatic synthetic reserve and nutritional status. Under normal conditions, the serum albumin concentration ranges from 35-50 g/L, but the level decreases in individuals with chronic liver disease and further decreases during acute inflammation because of both reduced hepatic synthesis and redistribution from systemic capillary leakage. Hypoalbuminemia is independently associated with mortality across multiple clinical contexts and is a well-established component of liver disease severity indices such as the Child-Pugh score. In addition to being a biomarker, albumin also has antioxidant, anti-inflammatory, and oncotic functions; therefore, its depletion may directly exacerbate oxidative stress and immune dysregulation[13,18].

By combining these two signals, the CAR amplifies prognostic information: Elevated CRP together with reduced albumin produces disproportionately high values. Across critical illnesses - including sepsis, cancer, and ICU populations - the CAR has consistently emerged as a strong predictor of mortality. For example, Zhou et al[19] reported that higher CAR values were independently associated with in-hospital mortality among patients with sepsis in the ICU. In decompensated cirrhosis, the CAR has likewise been linked to survival outcomes, reinforcing its utility as a marker of the balance between the inflammatory burden and physiological reserve[14,19,20].

Within ACLF, dynamic changes in the CAR likely reflect the patient’s trajectory of recovery vs deterioration. A declining CAR over the first week - driven by resolution of inflammation and stabilization of albumin - signals clinical improvement, whereas a rising CAR suggests escalating inflammation or worsening hepatic function. The ΔCAR-7 therefore captures this temporal shift and provides a surrogate for early treatment response. This rationale aligns with broader evidence in ACLF: Sequential measures such as bilirubin trends or evolving ACLF grades have consistently outperformed single baseline values in predicting survival[1].

From a mechanistic perspective, the two components of CARs map directly onto ACLF biology. CRP is acutely upregulated by common triggers such as infection and hepatotoxic insults, whereas decreased albumin reflects synthetic failure, inflammation-driven redistribution, and nutritional depletion. A rising CAR indicates both an intensifying cytokine storm and a weakening metabolic reserve. Moreover, inflammation itself exacerbates both sides of the ratio: Cytokine-induced vascular leakage decreases the level of albumin, whereas, simultaneous immune activation increases the level of CRP. CRP may even contribute to pathogenesis through complement activation and opsonization[1].

In summary, ΔCAR-7 embodies the dynamic interplay between inflammation and host resilience in ACLF. Its superior prognostic performance over the baseline CAR is intuitive: A static value provides only a snapshot confounded by chronic inflammation and preexisting disease severity, whereas the change over time integrates disease progression, treatment response, and host adaptability. This makes the dynamic CAR a compelling candidate biomarker, bridging mechanistic plausibility with clinical utility[14].

CLINICAL AND RESEARCH IMPLICATIONS

The study by Zhu and Yan[9] carries several important takeaways for clinical practice and research. From a frontline perspective, serial measurement of CRP and albumin in ACLF patients - particularly at admission and again on day 7 - offers valuable prognostic insight. A rising CAR over the first week should raise concern, prompting closer monitoring, escalation of supportive care (e.g., intensive care unit transfer, early vasopressor use), re-evaluation for occult infection, or expedited transplant consideration. Conversely, a stable or declining CAR trend may reassure clinicians that current management is effective. The incorporation of the ΔCAR-7 score into the COSSH-CAR score provides a practical tool - at least within HBV-related ACLF - to integrate this dynamic biomarker into prognostic decision-making. Since CRP and albumin are inexpensive and widely available, routine serial testing adds minimal burden to clinical workflows.

Several caveats merit consideration; different definitions of ACLF exist across the APASL, EASL-CLIF, and COSSH frameworks. Zhu and Yan[9] cohort was evaluated using COSSH criteria (bilirubin ≥ 205 μmol/L and INR ≥ 1.5), which are particularly relevant to HBV-ACLF. Whether the ΔCAR-7 score retains similar predictive performance in patients with non-HBV ACLF or under the EASL-CLIF definitions remains uncertain. For example, Western ACLF cohorts - often driven by alcoholic hepatitis or fungal sepsis - may exhibit distinct CRP/albumin dynamics. Thus, external validation is necessary before broad clinical adoption[21]. Moreover, a persistently elevated CAR despite therapy may signal the need for escalation of care.

Another key point is the model’s discriminatory power. In their analysis, the COSSH-CAR significantly outperformed established tools such as the Child-Pugh, MELD-Na, and COSSH-ACLF. While exact area under the curve values were not reported, prior work has shown that COSSH-ACLF achieves approximately 0.807 for 28-day mortality prediction, surpassing the benchmark representing a meaningful gain, potentially > 0.85. Clinically, this improved discrimination could be transformative. High COSSH-CAR scores might be used to identify patients for early transplant referral, trial enrollment, or costly rescue therapies (e.g., extracorporeal liver support), whereas lower scores might justify more conservative approaches[10,12].

From a research perspective, this study highlights the value of dynamic biomarkers in prognostication. A patient with moderate ACLF and a rising CAR may require urgent therapeutic escalation, whereas a patient with mild disease and a declining CAR could be managed conservatively. Previous studies, such as Yu et al[14] DP-ACLF model, have shown that incorporating trends in bilirubin, creatinine, and other parameters enhances predictive accuracy. Zhu and Yan[9] demonstrated that even a single dynamic index - the ΔCAR - can significantly increase prognostic precision. Future investigations should explore whether the ΔCAR could further improve existing frameworks (e.g., DP-ACLF and CLIF-C ACLF). Moreover, ΔCAR-7 itself could be tested as a surrogate trial endpoint, reflecting the therapeutic response to interventions such as antibiotics, anti-inflammatory agents, or corticosteroids in patients with ACLF[14].

This study also encourages the exploration of other dynamic inflammatory or composite biomarkers. Trends in bilirubin, INR, and creatinine have already shown promise. Ratios such as the neutrophil-percentage-to-albumin ratio or the neutrophil-albumin ratio may provide additional prognostic depth. Comparative studies of the ΔCAR-7 of Δneutrophil-percentage-to-albumin ratio, neutrophil-albumin ratio, or systemic immune-inflammation indices could clarify the most effective approach[8,21].

Finally, these findings reinforce a personalized medicine paradigm in ACLF management. Static scores treat patients as homogeneous, but dynamic biomarkers may allow for risk-adapted treatment strategies. For example, a patient with moderate ACLF and a worsening CAR might require urgent intervention, whereas another patient with milder disease and a falling CAR could be managed more conservatively. Ultimately, a panel of dynamic markers (e.g., CAR, NLR, and lactate) may help stratify patients with ACLF into biologically distinct subgroups with tailored treatment algorithms. The COSSH-CAR model represents an early but promising step toward this individualized approach, aligning with broader efforts in systems medicine and multiomics integration for liver failure[2,22].

TRANSLATIONAL SIGNIFICANCE

For hepatologists and intensivists, the study by Zhu and Yan[9] highlights the value of serial CRP and albumin monitoring in patients with ACLF. Tracking the CAR during the first week, particularly the ΔCAR-7, provides actionable prognostic information. A rising CAR despite therapy should prompt reassessment of infection control, nutritional support, or transfer to intensive care, whereas a falling or stable CAR may indicate effective treatment.

The COSSH-CAR score, which incorporates the ΔCAR-7, MELD-Na, and hepatic encephalopathy grade, can be calculated on day 7 to refine risk stratification. This shifts prognostic assessment from a static baseline model to a dynamic re-evaluation, aligning ACLF care with practices in other critical illnesses where serial scoring is standard (e.g., sepsis). Importantly, CRP and albumin are inexpensive and widely available, making this approach feasible even in resource-limited settings, particularly in HBV-endemic regions of Asia.

The next step is external validation across diverse cohorts, including APASL-defined and EASL-defined ACLF, to test generalizability beyond HBV-related cases. Prospective studies should also assess whether the ΔCAR-7 could predict longer-term outcomes, such as 90-day or 180-day survival, and whether its prognostic value differs between transplant-free survival and overall survival.

CAR dynamics could serve as both a trial stratification factor and an endpoint. Patients with high or increasing CAR values might be prioritized for trials of anti-inflammatory or nutritional interventions, with ΔCAR-7 cells used as an early marker of treatment response. Comparative studies against other emerging tools - such as the DP-ACLF model or machine learning-based prognostic systems - are also warranted. The integration of CAR into advanced algorithms could further increase the predictive accuracy[8]. Mechanistic research is needed to clarify the biological drivers of CAR trends. Linking the ΔCAR to cytokine kinetics, immune cell profiles, or metabolic shifts could validate its role as a surrogate of systemic inflammation and host resilience. Finally, prospective studies are warranted to evaluate whether the ΔCAR-7 could be used to predict extended outcomes, such as 90-day or 180-day survival. Such insights would not only strengthen confidence in CAR-guided care but also help identify therapeutic targets.

LIMITATIONS AND FUTURE RESEARCH NEEDS

While the COSSH-CAR findings are encouraging, they should be interpreted with caution until further validation is possible. The analysis was retrospective, which inherently carries risks of selection bias and unmeasured confounding. Interventions during the first week (e.g., antibiotics, vasopressors, or albumin infusions) may have influenced CAR changes independently of disease severity. Only a prospective study controlling for these variables can confirm that the ΔCAR-7 is an independent prognostic marker. This analysis was conducted within the COSSH-ACLF framework, with a focus on HBV-related ACLF. Its applicability to other etiologies remains uncertain, and external validation in more diverse cohorts is needed. The APASL and EASL-CLIF definitions differ in inclusion criteria, and geographic variation in ACLF etiology (e.g., alcoholic hepatitis or non-alcoholic steatohepatitis) may affect the CAR’s prognostic value. CRP, in particular, may be less informative in non-HBV contexts, highlighting the need for validation in diverse cohorts.

The choice of day 7 for ΔCAR-7 is empirical. While it likely captures an early response to therapy, patients who deteriorate or die before day 7 may not be accurately represented. Earlier (day 3 for ΔCAR) or later (day 14 for ΔCAR) timepoints, as well as continuous monitoring, might provide additional prognostic information. Prospective studies should investigate longitudinal CAR trajectories beyond the first week to account for late deterioration. Therapeutic interventions may affect CAR irrespective of the underlying disease course. During the early phase of management, treatments such as antibiotics, vasopressor therapy, or albumin supplementation can alter CAR trends without necessarily reflecting true changes in disease severity. For example, albumin administration may artificially lower CAR by increasing serum albumin levels, whereas anti-inflammatory drugs, including corticosteroids, can reduce CRP production, leading to decreased CAR values independent of clinical improvement[23,24], Future studies should account for therapy-specific effects to ensure that ΔCAR-7 reflects underlying physiology rather than treatment artifacts. Ultimately, prospective, multicenter validation is essential. Research should also evaluate whether incorporating the ΔCAR-7 into clinical practice improves patient outcomes, potentially through trials comparing dynamic CAR-guided management with standard care. Cost-effectiveness and real-world utility should also be assessed.

BIOMARKERS AND PERSONALIZED CLINICAL STRATEGIES IN PRACTICE

Dynamic CAR is easily derived from standard laboratory tests, making it a practical bedside tool even in resource-limited settings. This finding exemplifies the growing trend toward biomarker-driven, personalized care in hepatology. Like indices such as the NLR, platelet-to-lymphocyte ratio, and gamma-globulins, the CAR is an inexpensive, inflammation-based measure. Its main advantage lies in its simplicity and interpretability, as it relies on only two routine laboratory values with clear physiological significance. Clinicians can easily calculate the CAR at the bedside, unlike more complex composite scores, making it well suited for individualized risk stratification[2,20].

For practical application: (1) Low or decreasing CAR: Patients with a declining CAR may be managed conservatively, with a focus on supportive care and monitoring. Escalation to intensive therapies, such as artificial liver support, could be deferred if the trajectory indicates improvement; and (2) High or increasing CAR: Rising CAR signals escalating risk, prompting aggressive diagnostic and therapeutic action. This may include intensified infection surveillance, broad-spectrum antibiotics, early consideration of renal or respiratory support, and expedited transplant evaluation.

CAR can also function as part of a broader biomarker panel, complementing measures such as lactate clearance, procalcitonin trends, or leukocyte activation markers. Emerging molecular or metabolic biomarkers could further improve prognostication. Elevated levels of proinflammatory cytokines such as IL-6, tumor necrosis factor alpha, and IL-1β and chemokines such as IL-8 are characteristic of ACLF, whereas metabolic indicators - including lactate, ammonia, circulating bile acids, and amino acid metabolites - mirror the extent of organ dysfunction. The incorporation of these molecular and metabolic biomarkers into a comprehensive ACLF “dashboard”, together with the CAR, could substantially improve the precision of prognostic assessment[25]. In the future, an ACLF “dashboard” may integrate CAR as one axis, alongside other dynamic features, to guide both risk assessment and treatment decisions. Machine learning studies have shown that incorporating multiple dynamic parameters improves predictive accuracy, yet CAR retains the advantages of transparency, ease of calculation, and applicability in resource-limited settings[26].

Dynamic CAR trends may also inform therapy selection. For example, a hyperinflammatory trajectory could indicate potential benefits from immune-modulating treatments, such as corticosteroids, in patients with alcoholic hepatitis. Conversely, patients with stable or falling CARs might derive limited benefits. Similarly, CAR dynamics could guide decisions on extracorporeal liver support or early transjugular intrahepatic portosystemic shunt placement. In summary, integrating the CAR into clinical practice supports precision ACLF care, allowing interventions to be matched to each patient’s evolving inflammatory profile.

FUTURE DIRECTIONS

Several promising avenues for future research have emerged from the COSSH-CAR findings. Prospective interventional trials are needed to evaluate whether incorporating CAR dynamics improves patient outcomes. For example, one could compare standard prognostic scoring (MELD-Na, CLIF-C ACLF) with algorithms augmented with ΔCAR-7, measuring endpoints such as 28-day survival or intensive care unit-free days. Alternatively, ΔCAR-7 could serve as an enrollment criterion, targeting patients with increasing CAR trajectories for novel ACLF therapies. Multicenter validation cohorts across Asia, Europe, and the Americas are essential given the regional variability in ACLF etiology. Observational registries, similar to the CANONIC or the APASL ACLF studies, could incorporate serial CAR measurements to facilitate retrospective analyses and benchmarking.

Biomarker research can be expanded by combining the ΔCAR with genomic, transcriptomic, or proteomic markers to refine risk stratification. Investigating whether CAR dynamics correlate with distinct ACLF phenotypes - for example, circulatory vs neurological-dominant failure - could uncover pathophysiological subtypes and guide tailored interventions. Mechanistic studies may evaluate whether interventions that modulate CAR translate to improved outcomes. Anti-inflammatory therapies, such as IL-6 inhibitors, could be assessed for their impact on CRP and CAR trends, whereas nutritional strategies (e.g., high-protein diets or branched-chain amino acid supplementation) might stabilize albumin and reduce CAR. Determining whether these changes improve survival would provide actionable guidance for therapy selection.

Currently, experimental ACLF models do not directly characterize the kinetics of CAR or CRP, and existing interpretations are primarily inferred from general inflammatory research. Comprehensive mechanistic studies are warranted to investigate how targeted interventions (e.g., IL-6 blockade or albumin therapy) modulate CRP/albumin trends, influence disease mechanisms, and ultimately impact clinical outcomes, thereby clarifying the biological significance of CAR dynamics in ACLF.

Furthermore, although Zhu and Yan[9] reported a 90-day mortality rate of 40.5%, the COSSH-CAR model’s predictive validity beyond the 28-day period has not yet been investigated. Future studies should therefore explore whether the ΔCAR-7 or COSSH-CAR can also reliably predict longer-term outcomes, such as 90-day or 180-day survival. This explicitly highlights that extended prognostic performance remains to be established.

Finally, integrating CAR monitoring into electronic health records and clinical decision support systems could enable real-time alerts for increasing CAR, guiding timely interventions. Predictive dashboards incorporating dynamic CAR trends could continuously refine risk thresholds and inform personalized care strategies. Collectively, these approaches envision a dynamic, data-driven framework for ACLF prognosis, with CAR as a central biomarker guiding both risk assessment and therapeutic decisions.

DISCUSSION

The COSSH-CAR model was originally established using a cohort of patients with HBV-related ACLF. Therefore, its applicability and predictive accuracy in patients with ACLF arising from other causes, such as alcohol-associated liver disease or non-alcoholic steatohepatitis, have yet to be fully validated. The work by Zhu and Yan[9] highlights the significant prognostic potential of dynamic inflammatory markers in ACLF. Their findings that the first-week change in the CRP/albumin ratio (ΔCAR-7) strongly predicts short-term mortality and its integration into the COSSH-CAR model underscore the value of serial CAR monitoring. The superior discriminative performance of the COSSH-CAR suggests that incorporating a dynamic CAR could meaningfully enhance clinical decision-making in patients with HBV-related ACLF while also illustrating a broader movement toward dynamic, biomarker-driven prognostication.

Mechanistically, ΔCAR-7 reflects the evolving interplay between systemic inflammation and hepatic/nutritional reserves, providing both physiological insight and practical appeal, as CRP and albumin are routinely available laboratory tests. The evidence supports the notion that a dynamic CAR can complement existing scores, although careful validation, accounting for therapeutic interventions, and evaluation across diverse patient populations remain essential. In the future, integrating the ΔCAR with genomic, transcriptomic, or proteomic markers may enable even more precise, personalized approaches in ACLF care.

In clinical application, an elevated COSSH-CAR score may help identify patients who would benefit from early intensive interventions, such as advanced supportive care or transplant assessment, whereas a marked decline in CAR over time could support a more conservative therapeutic approach. This underscores the potential of the COSSH-CAR model to facilitate personalized treatment strategies in patients with ACLF by aligning clinical decisions with dynamic patient risk profiles.

Since the COSSH-CAR model relies solely on two routinely available laboratory parameters, it can be easily computed at the bedside; however, its application depends on the availability of timely serial CRP and albumin measurements. Embedding ΔCAR-7 into electronic alert systems or clinical decision-support platforms could enable automated risk stratification. Furthermore, incorporating the COSSH-CAR as a variable within machine learning frameworks - along with other clinical and biomarker inputs - may increase the precision of ACLF outcome prediction. Nonetheless, successful integration of such systems will require rigorous validation to account for data incompleteness and variability in therapeutic interventions.

CONCLUSION

In summary, dynamic CAR represents a promising addition to the toolkit of hepatologists. While it is not a definitive solution, it enhances the understanding of individual patient trajectories and can guide early, risk-adapted interventions. Clinicians are encouraged to monitor CAR trends, and researchers are encouraged to rigorously evaluate their prognostic and therapeutic utility. Ultimately, such efforts may advance precision-guided strategies for ACLF and improve outcomes in this high-risk population.