Published online Jan 7, 2027. doi: 10.3748/wjg.123000
Revised: June 29, 2026
Accepted: September 4, 2026
Published online: January 7, 2027
Processing time: 188 Days and 19.2 Hours
Inflammatory activity and nutritional status are closely associated with the progression of primary biliary cholangitis (PBC) and other chronic liver diseases.
To determine the predictive performance of the Glasgow prognostic score (GPS), modified GPS (mGPS), and high-sensitivity mGPS (hs-mGPS) for hepatic decompensation in PBC.
We performed a multicenter retrospective cohort study involving patients diagnosed with compensated PBC at seven hospitals between January 2013 and December 2024. GPS, mGPS, and hs-mGPS were calculated using baseline measurements of C-reactive protein (CRP), high-sensitivity CRP, and albumin levels. The endpoint was the first occurrence of hepatic decompensation during follow-up. Time-dependent receiver operating characteristic curves (ROCs) were used to compare the predictive performance of the GPS, mGPS, hs-mGPS, Paris I, Paris II, Mayo, United Kingdom-PBC (UK-PBC), and GLOBE scores for decompensation events.
This analysis included 762 patients with compensated PBC. The median age was 56 years (interquartile range: 49-66), and 606 patients (79.5%) were female. Kaplan-Meier analysis showed that the 5-year decompensation rates increased progressively across higher score strata for all three indices (P < 0.0001 for all). In fully adjusted multivariate models, GPS (hazard ratio [HR] = 3.605, 95% confidence interval [CI]: 2.783-4.670), mGPS (HR = 3.103, 95%CI: 2.469-3.900), and hs-mGPS (HR = 3.261, 95%CI: 2.549-4.171) independently predicted decompensation risk (P < 0.0001 for all). Time-dependent ROC analysis indicated that the GPS score yielded the highest area under the curve (0.754) for predicting decompensation, followed by hs-mGPS (0.725), mGPS (0.696), GLOBE (0.663), UK-PBC (0.638), Mayo (0.624), Paris I (0.597), and Paris II (0.582), whereas the difference between GPS and hsmGPS was not statistically significant.
GPS, mGPS, and hs-mGPS are independent prognostic markers for decompensation in PBC. The GPS has the strongest predictive accuracy, supporting its potential clinical utility in early risk stratification and management guidance.
Core Tip: This multicenter cohort study evaluated the predictive value of the Glasgow prognostic score (GPS) and its derivatives, modified GPS (mGPS) and high-sensitivity mGPS (hs-mGPS), for hepatic decompensation in primary biliary cholangitis (PBC). GPS integrates C-reactive protein and albumin to reflect systemic inflammation and nutritional status. In 762 patients with compensated PBC, GPS, mGPS, and hs-mGPS were independent predictors of decompensation, with GPS demonstrating the highest predictive accuracy. These findings highlight GPS as a simple, cost-effective tool for early risk stratification and personalized management in PBC.
- Citation: Huang M, Guo F, Bian LG, Wu CH, Zou W, Duan D, Tang Y, Chen SY, Zheng MY, Niu JK, Lei HT. Glasgow prognostic score as a predictor of decompensation in patients with primary biliary cholangitis. World J Gastroenterol 2027; 33(1): 123000
- URL: https://www.wjgnet.com/1007-9327/full/v33/i1/123000.htm
- DOI: https://dx.doi.org/10.3748/wjg.123000
Hepatic decompensation in cirrhosis marks a critical shift toward advanced chronic liver disease, substantially worsening prognosis and sharply increasing mortality risk[1]. Primary biliary cholangitis (PBC) is a chronic autoimmune liver disease characterized by progressive non-suppurative destruction of the interlobular bile ducts. In patients who respond poorly to treatment, the disease can progress to decompensated cirrhosis, a condition that leads to severe complications such as ascites, variceal hemorrhage, and hepatic encephalopathy. These complications significantly impair both quality of life and long-term survival[2]. A growing body of evidence points to systemic inflammation as a key driver of the transition from compensated to decompensated cirrhosis[3]. Systemic inflammation and immune dysfunction are now widely regarded as central factors in the progression of chronic liver disease and the onset of decompensation[4-6].
Current prognostic models for cirrhosis, such as the model for end-stage liver disease (MELD) score, rely mainly on measures of synthetic liver function and overall clinical status[7]. However, these models do not account for the two core pathological features of PBC: Chronic autoimmune-mediated inflammation and cholestasis. As a result, the predictive accuracy of PBC outcomes remains limited.
The Glasgow prognostic score (GPS) and its derivatives, modified GPS (mGPS) and high-sensitivity mGPS (hs-mGPS), were calculated using C-reactive protein (CRP), high-sensitivity CRP (hs-CRP), and albumin levels. This allows them to systematically reflect both the systemic inflammation and nutritional status. Although originally developed for cancer prognosis, where they are well-established predictors of outcomes in various malignancies[8], these scores have since been adopted in hepatology. Nevertheless, existing studies have largely focused on viral hepatitis or alcohol-related cirrhosis[9], and the value of GPS in predicting decompensation risk in PBC remains uncertain.
Therefore, this study evaluated whether baseline GPS, mGPS, and hs-mGPS could predict decompensation events in patients with PBC. We hope that these findings will aid in the early identification of high-risk individuals and help optimize clinical management strategies.
Between January 2013 and December 2024, we initially screened 1092 patients with compensated PBC across seven participating hospitals for this multicenter, retrospective cohort study. The participating hospitals included the Second Affiliated Hospital of Kunming Medical University (n = 524), First Affiliated Hospital of Kunming Medical University (n = 340), Qujing First People’s Hospital (n = 62), Yunnan Second People’s Hospital (n = 58), Qujing Second People’s Hospital (n = 49), Yu-xi People’s Hospital (n = 38), and Dali Prefecture People’s Hospital (n = 21). The median follow-up duration for the entire cohort was 53 months (interquartile range [IQR]: 41-64 months). A flowchart detailing the patient selection process is shown in Figure 1. In our cohort, no patients were lost to follow-up and no deaths occurred during the study period.
The diagnosis of PBC was established based on guidelines from the European Association for the Study of the Liver and the British Society of Gastroenterology/United Kingdom-PBC (UK-PBC), which required the presence of at least two of the following three criteria: (1) Serum alkaline phosphatase (ALP) elevation consistent with cholestatic liver injury; (2) Presence of antimitochondrial antibodies at a titer ≥ 1:40; and (3) Liver biopsy findings compatible with PBC[10,11]. The compensated stage was defined as the absence of decompensation events, including ascites (confirmed by imaging), hepatic encephalopathy (diagnosed based on West Haven criteria), or variceal bleeding (confirmed by endoscopy)[1]. All enrolled patients received standardized ursodeoxycholic acid therapy at a daily dose of 13-15 mg/kg for a minimum of 12 months, with treatment adherence verified by cross-checking medical and pharmacy records[10,11].
Participants were excluded from the analysis if they met any of the following conditions: (1) Coexisting viral hepatitis, alcoholic liver disease, or autoimmune hepatitis-PBC overlap syndrome; (2) Diagnosis of malignancy or active infection, defined by clinical symptoms, leukocytosis, elevated procalcitonin, and/or microbiological evidence; (3) Absence of standard ursodeoxycholic acid therapy; (4) History of corticosteroids, immunosuppressants, or albumin infusion; and (5) Incomplete clinical follow-up data.
Biochemical response per the Paris I criteria was defined as ALP ≤ 3 × upper limit of normal (ULN), aspartate ami
| Score | GPS | mGPS | hs-mGPS |
| 0 | CRP ≤ 10 mg/L and ALB ≥ 35 g/L | CRP ≤ 10 mg/L | hs-CRP ≤ 3 mg/L |
| 1 | CRP > 10 mg/L or ALB < 35 g/L | CRP > 10 mg/L and ALB ≥ 35 g/L | hs-CRP > 3 mg/L and ALB ≥ 35 g/L |
| 2 | CRP > 10 mg/L and ALB < 35 g/L | CRP > 10 mg/L and ALB < 35 g/L | hs-CRP > 3 mg/L and ALB < 35 g/L |
The primary outcome was the onset of hepatic decompensation, which was defined as the first documented episode of ascites, variceal hemorrhage, or hepatic encephalopathy[1]. Time-to-event was measured from the date of PBC diagnosis until the first decompensation event, death, or last clinical follow-up, whichever came first. Patient follow-up adhered to the British Society of Gastroenterology/UK-PBC guidelines[11]. Individuals with low-risk PBC underwent annual serum liver biochemical testing, supplemented by a written risk assessment every 3 years.
Statistical analyses were performed using SPSS version 31.0.1.0 (IBM Corp., Armonk, NY, United States) and R version 4.5.1 (R Foundation for Statistical Computing). Sample size was estimated a priori based on the primary endpoint of hepatic decompensation, and the final cohort of 762 patients met the required sample size. Continuous variables exhibiting skewed distributions are presented as medians (IQR) and were compared using the Mann-Whitney U test. Categorical variables are expressed as n (%) and were compared using the χ2 test or Fisher’s exact test, as appropriate. Survival curves were generated using the Kaplan-Meier method, and the associations of the GPS, mGPS, and hsmGPS with decompensation events were evaluated using univariable and multivariable Cox proportional hazards regression models. Multicollinearity among covariates in the fully adjusted model (model C) was assessed using variance inflation factor. The predictive accuracy of each score for decompensation events during follow-up was assessed using time-dependent receiver operating characteristic (ROC) analysis at the 5-year time point, with the area under the curve (AUC) and 95% confidence intervals (CIs) calculated. Differences in AUC values between the models were compared using DeLong’s test. To verify the robustness of the primary findings, subgroup and sensitivity analyses were conducted, and subgroup analyses were stratified by the MELD score (< 10 vs ≥ 10). In the sensitivity analysis, patients with hyper
A total of 762 patients with compensated PBC were included in the final analysis (Figure 1). The median age was 56 years (IQR: 49-66), 606 (79.5%) were female, and the median body mass index was 22.06 (IQR: 20.02-24.22). As shown in Table 2, compared to the non-decompensated group, patients who developed decompensation exhibited significantly higher levels of total bilirubin, AST, ALP, gamma-glutamyl transferase, CRP, and MELD scores but lower albumin levels (all P < 0.05). GPS, mGPS, and hs-mGPS scores also differed significantly between the groups (P < 0.001).
| Characteristic | Total (n = 762) | Decompensation (n = 113) | Non-decompensation (n = 649) | P value |
| Age (years) | 56 (49-66) | 57 (50-70.5) | 56 (49-65) | 0.112 |
| Sex | 0.973 | |||
| Female | 606 (79.5) | 90 (79.6) | 516 (79.5) | |
| Male | 156 (20.5) | 23 (20.4) | 133 (20.5) | |
| BMI (kg/m2) | 22.06 (20.02-24.22) | 21.50 (19.95-23.90) | 22.06 (20.04-24.38) | 0.222 |
| Hypertension | 0.974 | |||
| Yes | 183 (24) | 27 (23.9) | 156 (24) | |
| No | 579 (76) | 86 (76.1) | 493 (76) | |
| Diabetes mellitus | 0.898 | |||
| Yes | 85 (11.2) | 13 (11.5) | 72 (11.1) | |
| No | 677 (88.8) | 100 (88.5) | 577 (88.9) | |
| Atherosclerosis | 0.990 | |||
| Yes | 47 (6.2) | 7 (6.2) | 40 (6.2) | |
| No | 715 (93.8) | 106 (93.8) | 609 (93.8) | |
| TBIL (μmol/L) | 18.4 (12.4-40.08) | 27.1 (12.65-80.6) | 17.8 (12.35-35) | 0.002 |
| ALT (U/L) | 53.5 (28-106) | 63 (29.1-111.5) | 52 (28-104) | 0.249 |
| AST (U/L) | 54 (33-103.93) | 75 (37.85-119) | 50 (32-98.95) | 0.002 |
| ALP (U/L) | 176.15 (113-320) | 252 (139.75-440.5) | 171 (108.5-285) | < 0.001 |
| GGT (U/L) | 205 (71-415.75) | 253 (104-521.5) | 196 (68-404.5) | 0.011 |
| INR | 1 (0.94-1.09) | 1.01 (0.91-1.13) | 1 (0.94-1.09) | 0.737 |
| GLOB (g/L) | 33.15 (29-37.6) | 32.5 (28.5-39.05) | 33.2 (29-37.5) | 0.829 |
| ALB (g/L) | 37.9 (33.5-42) | 34 (29.5-38.4) | 38.4 (34.2-42.2) | < 0.001 |
| CRP (mg/L) | 2.47 (1.1-7.35) | 10.8 (3.63-12.75) | 1.97 (1.02-5) | < 0.001 |
| MELD score | 3.86 (1.53-7.34) | 5.57 (2.16-9.13) | 3.67 (1.48-6.79) | 0.007 |
| AMA-M2 | 0.877 | |||
| 1 | 528 (69.3) | 79 (69.9) | 449 (69.2) | |
| 0 | 234 (30.7) | 34 (30.1) | 200 (30.8) | |
| SP100 | 0.623 | |||
| 1 | 84 (11.1) | 14 (12.4) | 70 (10.8) | |
| 0 | 676 (88.9) | 99 (87.6) | 577 (89.2) | |
| GP210 | 0.361 | |||
| 1 | 164 (21.5) | 28 (24.8) | 136 (21) | |
| 0 | 598 (78.5) | 85 (75.2) | 513 (79) | |
| Paris I | 0.001 | |||
| Response | 495 (65) | 58 (51.3) | 437 (67.3) | |
| Suboptimal response | 267 (35) | 55 (48.7) | 212 (32.7) | |
| Paris II | 0.011 | |||
| Response | 319 (41.9) | 35 (31) | 284 (43.8) | |
| Suboptimal response | 443 (58.1) | 78 (69) | 365 (56.2) | |
| Mayo | 4 (3-5) | 5 (4-6) | 4 (3-5) | < 0.001 |
| GLOBE | < 0.001 | |||
| > 0.3 | 133 (17.5) | 37 (32.7) | 96 (14.8) | |
| ≤ 0.3 | 629 (82.5) | 76 (67.3) | 553 (85.2) | |
| UK-PBC | 0.57 (-0.32 to 1.98) | 1.49 (0.26-3.10) | 0.46 (-0.40 to 1.77) | < 0.001 |
| GPS | < 0.001 | |||
| 0 | 475 (62.3) | 26 (23.0) | 449 (69.2) | |
| 1 | 198 (26.0) | 49 (43.4) | 149 (23.0) | |
| 2 | 89 (11.7) | 38 (33.6) | 51 (7.8) | |
| mGPS | < 0.001 | |||
| 0 | 616 (80.8) | 51 (45.1) | 565 (87.1) | |
| 1 | 58 (7.6) | 24 (21.3) | 34 (5.2) | |
| 2 | 88 (11.6) | 38 (33.6) | 50 (7.7) | |
| hs-mGPS | < 0.001 | |||
| 0 | 445 (58.4) | 27 (23.9) | 418 (64.4) | |
| 1 | 166 (21.8) | 36 (31.9) | 130 (20.0) | |
| 2 | 151 (19.8) | 50 (44.2) | 101 (15.6) |
Kaplan-Meier analysis demonstrated significantly different cumulative incidences of decompensation across the GPS, mGPS, and hs-mGPS scores (all P < 0.0001; Figure 2). The 5-year decompensation rates for scores of 0, 1, and 2 were as follows: GPS: 6.7%, 29.4%, 60.7%; mGPS: 9.6%, 53.4%, 62.3%; hs-mGPS: 6.8%, 29.6%, 49.8%.
As shown in Table 3, GPS, mGPS, and hs-mGPS were significantly associated with the risk of decompensation in both the univariate and multivariate Cox regression analyses. In the fully adjusted model (model C), compared with a re
| Variable | Univariate analysis | Multivariate analysis | ||||||
| Model A | Model B | Model C | ||||||
| HR (95%CI) | P value | HR (95%CI) | P value | HR (95%CI) | P value | HR (95%CI) | P value | |
| GPS | ||||||||
| 0 | Reference | Reference | Reference | Reference | ||||
| 1 | 5.593 (3.475-9.003) | < 0.001 | 5.805 (3.589-9.389) | < 0.001 | 5.911 (3.647-9.580) | < 0.001 | 5.540 (3.346-9.173) | < 0.001 |
| 2 | 14.289 (8.617-23.694) | < 0.001 | 13.618 (8.133-22.804) | < 0.001 | 13.942 (8.290-23.450) | < 0.001 | 13.573 (7.835-23.514) | < 0.001 |
| P for trend | 3.690 (2.922-4.660) | < 0.001 | 3.595 (2.837-4.556) | < 0.001 | 3.654 (2.875-4.643) | < 0.001 | 3.605 (2.783-4.670) | < 0.001 |
| mGPS | ||||||||
| 0 | Reference | Reference | Reference | Reference | ||||
| 1 | 7.747 (4.755-12.62) | < 0.001 | 8.512 (5.169-14.016) | < 0.001 | 8.537 (5.180-14.070) | < 0.001 | 8.025 (4.805-13.405) | < 0.001 |
| 2 | 10.09 (6.546-15.551) | < 0.001 | 9.219 (5.926-14.340) | < 0.001 | 9.511 (6.082-14.873) | < 0.001 | 8.459 (5.211-13.732) | < 0.001 |
| P for trend | 3.238 (2.644-3.965) | < 0.001 | 3.128 (2.544-3.846) | < 0.001 | 3.188 (2.584-3.932) | < 0.001 | 3.103 (2.469-3.900) | < 0.001 |
| hs-mGPS | ||||||||
| 0 | Reference | Reference | Reference | Reference | ||||
| 1 | 7.161 (4.291-11.952) | < 0.001 | 7.315 (4.349-12.305) | < 0.001 | 7.307 (4.322-12.353) | < 0.001 | 7.343 (4.269-12.630) | < 0.001 |
| 2 | 12.224 (7.523-19.862) | < 0.001 | 11.951 (7.257-19.679) | < 0.001 | 12.187 (7.382-20.120) | < 0.001 | 11.407 (6.669-19.511) | < 0.001 |
| P for trend | 3.283 (2.639-4.084) | < 0.001 | 3.245 (2.591-4.064) | < 0.001 | 3.300 (2.628-4.143) | < 0.001 | 3.261 (2.549-4.171) | < 0.001 |
Furthermore, in the fully adjusted model, each one-point increase in GPS (HR = 3.605, 95%CI: 2.783-4.670), mGPS
As shown in Figure 3, time-dependent ROC analysis demonstrated that all evaluated scores were significant predictors of decompensation. Their predictive performance, expressed as the AUC with 95%CIs, was as follows: GPS, 0.754 (0.703-0.806); mGPS, 0.696 (0.637-0.756); hs-mGPS, 0.725 (0.671-0.779); Paris I, 0.597 (0.541-0.653); Paris II, 0.582 (0.526-0.639); Mayo, 0.624 (0.564-0.685); UK-PBC, 0.638 (0.579-0.696); and GLOBE, 0.663 (0.605-0.720). As shown in Figure 4, the GPS model demonstrated superior predictive ability compared to other prognostic models (all P < 0.05), with the exception of hs-mGPS.
Subgroup analysis was performed by stratifying patients by MELD score (< 10 vs ≥ 10). In the MELD < 10 subgroup (n = 528, decompensation events = 58) and the MELD ≥ 10 subgroup (n = 234, decompensation events = 55), the fully adjusted models demonstrated that GPS (HR = 3.982, 95%CI: 2.768-5.729), mGPS (HR = 3.685, 95%CI: 2.569-5.288), and hs-mGPS (HR = 4.347, 95%CI: 3.069-6.159) were significantly associated with decompensation risk. Similarly, in patients with MELD scores ≥ 10, GPS (HR = 3.521; 95%CI: 2.215-5.597), mGPS (HR = 3.014, 95%CI: 2.113-4.299), and hs-mGPS (HR = 2.207; 95%CI: 1.509-3.226) were also significantly associated with decompensation risk in the fully adjusted model (all P < 0.001; Figure 5 and Table 4).
| Variable | Univariate analysis | Multivariate analysis | ||||||
| Model A | Model B | Model C | ||||||
| HR (95%CI) | P value | HR (95%CI) | P value | HR (95%CI) | P value | HR (95%CI) | P value | |
| MELD ≥ 10 | ||||||||
| GPS | 3.173 (2.142-4.698) | < 0.001 | 3.024 (2.017-4.534) | < 0.001 | 3.072 (2.041-4.623) | < 0.001 | 3.521 (2.215-5.597) | < 0.001 |
| mGPS | 2.683 (2.000-3.599) | < 0.001 | 2.592 (1.909-3.519) | < 0.001 | 2.629 (1.933-3.577) | < 0.001 | 3.014 (2.113-4.299) | < 0.001 |
| hs-mGPS | 2.205 (1.584-3.069) | < 0.001 | 2.115 (1.499-2.984) | < 0.001 | 2.096 (1.485-2.959) | < 0.001 | 2.207 (1.509-3.226) | < 0.001 |
| MELD < 10 | ||||||||
| GPS | 3.598 (2.618-4.943) | < 0.001 | 3.484 (2.519-4.817) | < 0.001 | 3.673 (2.648-5.094) | < 0.001 | 3.982 (2.768-5.729) | < 0.001 |
| mGPS | 3.266 (2.407-4.431) | < 0.001 | 3.101 (2.278-4.221) | < 0.001 | 3.217 (2.345-4.413) | < 0.001 | 3.685 (2.569-5.288) | < 0.001 |
| hs-mGPS | 3.937 (2.909-5.328) | < 0.001 | 3.846 (2.821-5.244) | < 0.001 | 4.132 (2.997-5.697) | < 0.001 | 4.347 (3.069-6.159) | < 0.001 |
To assess the robustness of the findings, sensitivity analysis was performed by excluding patients with hypertension, arteriosclerosis, or diabetes mellitus. In the fully adjusted model (model B), GPS (HR = 4.265, 95%CI: 3.118-5.834), mGPS (HR = 3.400, 95%CI: 2.605-4.437), and hs-mGPS (HR = 3.423, 95%CI: 2.580-4.541) remained significantly associated with decompensation risk (all P < 0.001; Table 5).
| Variable | Univariate analysis | Multivariate analysis | ||||
| Model A | Model B | |||||
| HR (95%CI) | P value | HR (95%CI) | P value | HR (95%CI) | P value | |
| GPS | 4.306 (3.230-5.740) | < 0.001 | 4.341 (3.246-5.805) | < 0.001 | 4.265 (3.118-5.834) | < 0.001 |
| mGPS | 3.602 (2.827-4.589) | < 0.001 | 3.547 (2.776-4.534) | < 0.001 | 3.400 (2.605-4.437) | < 0.001 |
| hs-mGPS | 3.464 (2.676-4.483) | < 0.001 | 3.463 (2.661-4.506) | < 0.001 | 3.423 (2.580-4.541) | < 0.001 |
This multicenter cohort study is the first to systematically evaluate the predictive value of the GPS for decompensation risk in patients with PBC within a large cohort from Southwest China. Our principal finding was that GPS, mGPS, and hs-mGPS were all independent predictors of hepatic decompensation in patients with PBC, with the original GPS demonstrating superior predictive performance. These results provide crucial evidence for the early identification of high-risk patients and may inform the optimization of clinical management strategies. The robustness of these associations was further corroborated by the comprehensive sensitivity and subgroup analyses.
The GPS, initially developed in oncology, integrates albumin and CRP levels to provide a composite reflection of a patient’s nutritional and systemic inflammatory status[8,15] and has been validated as an independent prognostic indicator in various malignancies[16,17]. Our study extends its utility to PBC, revealing that higher GPS, mGPS, and hs-mGPS scores are significantly associated with an elevated risk of hepatic decompensation. This finding aligns with the established literature: Cash et al[15] reported significantly elevated CRP levels in patients with PBC, Mendelian randomization studies have further confirmed CRP as a potential risk factor for PBC[18]. Elevated CRP is also associated with the progression of cirrhosis and increased mortality risk[19] and has prognostic value in acute liver failure with sepsis[20]. By contrast, hypoalbuminemia has been identified as a predictor of poor prognosis in patients[21]. The pathophysiological mechanism is thought to involve a self-perpetuating cycle of persistent inflammatory injury and progressive nutritional depletion. In the context of PBC-specific autoimmune bile duct injury, elevated CRP levels signify a persistent systemic inflammatory state that correlates with regulatory T-cell function and is associated with the activation of pro-inflammatory T helper cell (Th) 1/Th17 cells[22], which in turn correlates with hepatic stellate cell activation and is associated with liver fibrosis progression[23]. Furthermore, inflammatory cytokines may shift hepatic protein synthesis toward acute-phase reactants while suppressing albumin production[24]. Hypoalbuminemia not only directly reflects impaired hepatic synthetic function but is also indicative of a malnourished state, which may compromise the intestinal mucosal barrier and increase the risk of bacterial translocation[22]. These interconnected pathways are collectively associated with immune dysregulation and fibrosis progression, ultimately driving the transition to decompensated disease.
The prognostic strength of the GPS stems from its integrated representation of the “inflammation-nutrition” axis. By integrating both CRP and albumin, it provides a more comprehensive risk assessment than single biomarkers, effectively identifying patients in a high-risk state of concurrent “high inflammation and low nutrition.” Our time-dependent ROC analysis indicated that the predictive performance of GPS (AUC = 0.753) was significantly superior to that of mGPS (AUC = 0.695). This discrepancy arises primarily from the differential weighting of albumin levels. The mGPS assigns points for hypoalbuminemia only in the presence of elevated CRP levels, potentially overlooking patients with significant hypoalbuminemia but normal CRP levels. In chronic liver diseases such as PBC, low albumin levels often result from cholestasis-related malabsorption, chronic decline in synthetic function, and chronic malnutrition[25,26]. The original GPS, by independently scoring both parameters, can more sensitively identify this “low-inflammation, high-malnutrition” phenotype, thereby enabling more comprehensive and precise risk stratification for decompensation. Consequently, the original GPS may yield more accurate prognostic information than mGPS in the clinical management of PBC.
Regarding the hs-mGPS, it uses a lower cut-off for CRP (hs-CRP > 3 mg/L) compared with the conventional CRP threshold (> 10 mg/L) used in the original GPS. This lower threshold is designed to detect subtler systemic inflammation that may be missed by standard CRP assays[27]. In PBC, a chronic autoimmune condition characterized by persistent low-grade inflammation, hs-CRP may theoretically offer advantages in capturing minute inflammatory fluctuations. However, our results showed that hs-mGPS (AUC = 0.725) did not outperform the original GPS (AUC = 0.754), suggesting that the higher conventional CRP threshold (> 10 mg/L) already effectively identifies clinically meaningful inflammation in this population. This may be because PBC-related inflammation, once it reaches a level sufficient to impact prognosis, exceeds the conventional CRP threshold; the added sensitivity of hs-CRP may not translate into additional prognostic gain. These findings support the use of conventional CRP in routine clinical practice, as it is more accessible and cost-effective, particularly in resource-limited settings.
It should also be noted that the UK-PBC and GLOBE scores, with which we compared the GPS, were originally developed to predict liver transplantation or overall mortality rather than decompensation events specifically[12]. Therefore, direct AUC comparisons between these models and the GPS should be interpreted with caution. Nevertheless, the consistently superior performance of GPS in predicting decompensation in our cohort supports its potential utility as a complementary tool for risk stratification in PBC.
From a clinical perspective, our findings provide actionable insights for risk stratification. Based on the observed 5-year decompensation rates and the multivariate HRs, we propose that patients with a baseline GPS ≥ 1 should be considered at elevated risk and may benefit from more intensive surveillance, including shortened follow-up intervals and earlier consideration of second-line therapies such as obeticholic acid or fibrates, particularly when concomitant biochemical non-response is present[10,11]. Patients with GPS = 2, representing concurrent elevated CRP and hypoalbuminemia, warrant the highest level of vigilance and timely multidisciplinary intervention. While these recommendations require prospective validation, they provide a framework for integrating GPS into clinical decision-making for PBC.
The principal strengths of our study include its multicenter design, substantial sample size, and extended follow-up period, which enhance the generalizability of our findings. Moreover, the GPS utilizes routinely available laboratory parameters (CRP and albumin), making it easy to calculate, cost-effective, and readily translatable into clinical practice. However, several limitations of this study must be acknowledged. The retrospective design is susceptible to selection bias, and although we adjusted for numerous known confounders in our multivariable models, residual confounding from unmeasured factors cannot be entirely excluded. Additionally, as a multicenter retrospective study, minor inter-center variations in laboratory assays and follow-up protocols may exist, which represents an inherent limitation of this study. Furthermore, although this study was conducted across seven hospitals, all participating centers are located in Yunnan Province and other regions of Southwest China. This geographic concentration may limit the generalizability of our findings to other ethnic populations or healthcare settings with different demographic and clinical characteristics. External validation in diverse, multi-ethnic cohorts is warranted. Therefore, further prospective studies are required to validate our conclusions.
In conclusion, the GPS and its derivatives are effective tools for predicting the risk of hepatic decompensation in patients with PBC, with the original GPS exhibiting the best predictive performance. This easily calculable score, based on readily available biomarkers, holds significant promise for clinical adoption, facilitating early risk stratification, and paving the way for more personalized management of PBC.
The GPS and its derivatives are effective tools for predicting the risk of hepatic decompensation in patients with PBC, with the original GPS exhibiting the best predictive performance. This easily calculable score, based on readily available biomarkers, holds significant promise for clinical adoption, facilitating early risk stratification, and paving the way for more personalized management of PBC.
| 1. | D'Amico G, Bernardi M, Angeli P. Towards a new definition of decompensated cirrhosis. J Hepatol. 2022;76:202-207. [RCA] [PubMed] [DOI] [Full Text] [Cited by in Crossref: 198] [Cited by in RCA: 181] [Article Influence: 45.3] [Reference Citation Analysis (3)] |
| 2. | Ampuero J, Lucena A, Berenguer M, Hernández-Guerra M, Molina E, Gómez-Camarero J, Valdivia C, Gómez E, Casado M, Álvarez-Navascuez C, Jorquera F, García-Buey L, Díaz-González Á, Morillas R, García-Retortillo M, Sousa JM, Pérez-Medrano I, Simón MÁ, Martínez J, Arenas J, Londoño MC, Olveira A, Fernández-Rodríguez C; ColHai Registry. Predictive factors for decompensating events in patients with cirrhosis with primary biliary cholangitis under different lines of therapy. Hepatology. 2024;80:791-806. [RCA] [PubMed] [DOI] [Full Text] [Cited by in Crossref: 8] [Cited by in RCA: 15] [Article Influence: 7.5] [Reference Citation Analysis (0)] |
| 3. | Sánchez-Aldehuelo R, Villanueva C, Genescà J, García-Pagán JC, Castillo E, Calleja JL, Aracil C, Bañares R, Téllez L, Paule L, Morillas RM, Poca M, Peñas B, Augustin S, Abraldes JG, Alvarado-Tapias E, Bosch J, Albillos A. Progressive systemic inflammation precedes decompensation in compensated cirrhosis. JHEP Rep. 2025;7:101231. [RCA] [PubMed] [DOI] [Full Text] [Cited by in RCA: 10] [Reference Citation Analysis (1)] |
| 4. | Clària J, Arroyo V, Moreau R. Roles of systemic inflammatory and metabolic responses in the pathophysiology of acute-on-chronic liver failure. JHEP Rep. 2023;5:100807. [RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)] [Cited by in Crossref: 47] [Cited by in RCA: 45] [Article Influence: 15.0] [Reference Citation Analysis (2)] |
| 5. | Geng A, Brenig RG, Roux J, Lütge M, Cheng HW, Flint EE, Lussier POG, Meier MA, Pop OT, Künzler-Heule P, Matter MS, Wendon J, McPhail MJW, Soysal S, Semela D, Heim M, Weston CJ, Ludewig B, Bernsmeier C. Circulating monocytes upregulate CD52 and sustain innate immune function in cirrhosis unless acute decompensation emerges. J Hepatol. 2025;83:146-160. [RCA] [PubMed] [DOI] [Full Text] [Cited by in Crossref: 11] [Cited by in RCA: 13] [Article Influence: 13.0] [Reference Citation Analysis (0)] |
| 6. | Gustot T, Stadlbauer V, Laleman W, Alessandria C, Thursz M. Transition to decompensation and acute-on-chronic liver failure: Role of predisposing factors and precipitating events. J Hepatol. 2021;75 Suppl 1:S36-S48. [RCA] [PubMed] [DOI] [Full Text] [Cited by in Crossref: 54] [Cited by in RCA: 50] [Article Influence: 10.0] [Reference Citation Analysis (2)] |
| 7. | Mazumder NR, Fontana RJ. MELD 3.0 in Advanced Chronic Liver Disease. Annu Rev Med. 2024;75:233-245. [RCA] [PubMed] [DOI] [Full Text] [Cited by in Crossref: 13] [Cited by in RCA: 31] [Article Influence: 15.5] [Reference Citation Analysis (0)] |
| 8. | Okimoto S, Kobayashi T, Tashiro H, Kuroda S, Ishiyama K, Ide K, Abe T, Hashimoto M, Iwako H, Hamaoka M, Honmyo N, Yamaguchi M, Ohdan H. Significance of the Glasgow Prognostic Score for patients with colorectal liver metastasis. Int J Surg. 2017;42:209-214. [RCA] [PubMed] [DOI] [Full Text] [Cited by in Crossref: 17] [Cited by in RCA: 16] [Article Influence: 1.8] [Reference Citation Analysis (0)] |
| 9. | Zhou DS, Xu L, Luo YL, He FY, Huang JT, Zhang YJ, Chen MS. Inflammation scores predict survival for hepatitis B virus-related hepatocellular carcinoma patients after transarterial chemoembolization. World J Gastroenterol. 2015;21:5582-5590. [RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)] [Cited by in CrossRef: 19] [Cited by in RCA: 23] [Article Influence: 2.1] [Reference Citation Analysis (0)] |
| 10. | European Association for the Study of the Liver. EASL Clinical Practice Guidelines: The diagnosis and management of patients with primary biliary cholangitis. J Hepatol. 2017;67:145-172. [RCA] [PubMed] [DOI] [Full Text] [Cited by in Crossref: 1222] [Cited by in RCA: 1131] [Article Influence: 125.7] [Reference Citation Analysis (5)] |
| 11. | Hirschfield GM, Dyson JK, Alexander GJM, Chapman MH, Collier J, Hübscher S, Patanwala I, Pereira SP, Thain C, Thorburn D, Tiniakos D, Walmsley M, Webster G, Jones DEJ. The British Society of Gastroenterology/UK-PBC primary biliary cholangitis treatment and management guidelines. Gut. 2018;67:1568-1594. [RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)] [Cited by in Crossref: 294] [Cited by in RCA: 266] [Article Influence: 33.3] [Reference Citation Analysis (3)] |
| 12. | Chen S, Duan W, You H, Jia J. A brief review on prognostic models of primary biliary cholangitis. Hepatol Int. 2017;11:412-418. [RCA] [PubMed] [DOI] [Full Text] [Cited by in Crossref: 12] [Cited by in RCA: 15] [Article Influence: 1.7] [Reference Citation Analysis (0)] |
| 13. | Takamori S, Takada K, Shimokawa M, Matsubara T, Fujishita T, Ito K, Toyozawa R, Yamaguchi M, Okamoto T, Yoneshima Y, Tanaka K, Okamoto I, Tagawa T, Mori M. Clinical utility of pretreatment Glasgow prognostic score in non-small-cell lung cancer patients treated with immune checkpoint inhibitors. Lung Cancer. 2021;152:27-33. [RCA] [PubMed] [DOI] [Full Text] [Cited by in Crossref: 15] [Cited by in RCA: 41] [Article Influence: 6.8] [Reference Citation Analysis (0)] |
| 14. | Di Rosa M, Sabbatinelli J, Giuliani A, Carella M, Magro D, Biscetti L, Soraci L, Spannella F, Fedecostante M, Lenci F, Tortato E, Pimpini L, Burattini M, Cecchini S, Cherubini A, Bonfigli AR, Capalbo M, Procopio AD, Balistreri CR, Olivieri F. Inflammation scores based on C-reactive protein and albumin predict mortality in hospitalized older patients independent of the admission diagnosis. Immun Ageing. 2024;21:67. [RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)] [Cited by in RCA: 15] [Reference Citation Analysis (1)] |
| 15. | Cash WJ, McCance DR, Young IS, McEneny J, Cadden IS, McDougall NI, Callender ME. Primary biliary cirrhosis is associated with oxidative stress and endothelial dysfunction but not increased cardiovascular risk. Hepatol Res. 2010;40:1098-1106. [RCA] [PubMed] [DOI] [Full Text] [Cited by in Crossref: 23] [Cited by in RCA: 23] [Article Influence: 1.4] [Reference Citation Analysis (0)] |
| 16. | Liu R, Wang L, Ye J, Li X, Ma W, Xu X, Yu J, Wang W. Preoperative glasgow prognostic score was an effective prognostic indicator in patients with biliary tract cancer. Front Immunol. 2025;16:1560944. [RCA] [PubMed] [DOI] [Full Text] [Cited by in RCA: 5] [Reference Citation Analysis (0)] |
| 17. | Dolan RD, McMillan DC. The prevalence of cancer associated systemic inflammation: Implications of prognostic studies using the Glasgow Prognostic Score. Crit Rev Oncol Hematol. 2020;150:102962. [RCA] [PubMed] [DOI] [Full Text] [Cited by in Crossref: 14] [Cited by in RCA: 36] [Article Influence: 6.0] [Reference Citation Analysis (0)] |
| 18. | Gao W, Peng C, Wang Z, Li Y, Liu M. Genetic association and causal relationship between multiple modifiable risk factors and autoimmune liver disease: a two-sample mendelian randomization study. J Transl Med. 2024;22:425. [RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)] [Cited by in Crossref: 4] [Cited by in RCA: 11] [Article Influence: 5.5] [Reference Citation Analysis (1)] |
| 19. | Turco L, Garcia-Tsao G, Magnani I, Bianchini M, Costetti M, Caporali C, Colopi S, Simonini E, De Maria N, Banchelli F, Rossi R, Villa E, Schepis F. Cardiopulmonary hemodynamics and C-reactive protein as prognostic indicators in compensated and decompensated cirrhosis. J Hepatol. 2018;68:949-958. [RCA] [PubMed] [DOI] [Full Text] [Cited by in Crossref: 133] [Cited by in RCA: 120] [Article Influence: 15.0] [Reference Citation Analysis (2)] |
| 20. | Wang D, Wang X, Mu J, Kuang Z, Zhang J, Lu X, Wang X, Lin F. Prognostic indicators and outcome in patients with acute liver failure, sepsis and with and without shock: a retrospective cohort study. Ann Med. 2025;57:2438833. [RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)] [Cited by in Crossref: 13] [Cited by in RCA: 10] [Article Influence: 10.0] [Reference Citation Analysis (1)] |
| 21. | Yao TT, Qian JD, Wang GQ. Efficacy of ursodeoxycholic acid combined with prednisolone and immunosuppressant triple therapy in the treatment of refractory primary biliary cholangitis. Med Clin (Barc). 2020;155:165-170. [RCA] [PubMed] [DOI] [Full Text] [Cited by in Crossref: 1] [Cited by in RCA: 7] [Article Influence: 1.2] [Reference Citation Analysis (1)] |
| 22. | Chen R, Sun Y, Hu Y, Tai W. Innate immunity of bile and cholangiocytes in primary biliary cholangitis. Front Immunol. 2025;16:1655287. [RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)] [Cited by in RCA: 4] [Reference Citation Analysis (0)] |
| 23. | Simbrunner B, Villesen IF, Königshofer P, Scheiner B, Bauer D, Paternostro R, Schwabl P, Timelthaler G, Ramazanova D, Wöran K, Stift J, Eigenbauer E, Stättermayer AF, Marculescu R, Pinter M, Møller S, Trauner M, Karsdal M, Leeming DJ, Reiberger T, Mandorfer M. Systemic inflammation is linked to liver fibrogenesis in patients with advanced chronic liver disease. Liver Int. 2022;42:2501-2512. [RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)] [Cited by in Crossref: 5] [Cited by in RCA: 43] [Article Influence: 10.8] [Reference Citation Analysis (0)] |
| 24. | Wiedermann CJ. Hypoalbuminemia as Surrogate and Culprit of Infections. Int J Mol Sci. 2021;22:4496. [RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)] [Cited by in Crossref: 64] [Cited by in RCA: 305] [Article Influence: 61.0] [Reference Citation Analysis (12)] |
| 25. | Jagdish RK, Maras JS, Sarin SK. Albumin in Advanced Liver Diseases: The Good and Bad of a Drug! Hepatology. 2021;74:2848-2862. [RCA] [PubMed] [DOI] [Full Text] [Cited by in Crossref: 16] [Cited by in RCA: 91] [Article Influence: 18.2] [Reference Citation Analysis (0)] |
| 26. | Li Y, Liu B, Li X. High C-reactive protein-to-albumin ratio levels are associated with osteoporosis in patients with primary biliary cholangitis. Front Endocrinol (Lausanne). 2024;15:1415488. [RCA] [PubMed] [DOI] [Full Text] [Reference Citation Analysis (0)] |
| 27. | Wu TH, Tsai YT, Chen KY, Yap WK, Luan CW. Utility of High-Sensitivity Modified Glasgow Prognostic Score in Cancer Prognosis: A Systemic Review and Meta-Analysis. Int J Mol Sci. 2023;24:1318. [RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)] [Cited by in Crossref: 16] [Cited by in RCA: 27] [Article Influence: 9.0] [Reference Citation Analysis (1)] |