Primary biliary cholangitis (PBC), also known as primary biliary cirrhosis, is an immune-mediated, chronic progressive cholestatic liver disease with unclear etiology, likely associated with environmental and genetic factors (1).It is defined by non-purulent damage to the small bile ducts (2).The incidence rate in Europe and Asia is 1.9-40.2/100000 and 4.8-5.6/100000, respectively, with a prevalence in middle-aged and elderly women (3–5). Without effective treatment, patients often progress to liver fibrosis, cirrhosis and liver cancer, eventually leading to death (6).

Ursodeoxycholic acid (UDCA) remains the only primary treatment option for PBC. Research indicates that after 5 years of UDCA therapy, the incidence of liver decompensation in PBC patients is between 3.81% and 4.31%, while the rate of hepatocellular carcinoma development is 1.62%. Prognosis worsens notably once cirrhosis develops, with a 5-year survival rate (without the need for liver transplantation) of 77.1% in patients with compensated cirrhosis, dropping to 35.9% in those with decompensated cirrhosis (7).

We conducted a retrospective analysis of clinical data from 315 PBC patients, focusing on liver-related mortality as the outcome, to explore the factors affecting survival. Utilizing the identified risk factors, nomograms were developed to estimate 5-year OS for these patients.

This study included 315 PBC patients, analyzing death as the endpoint to identify risk factors for mortality. Cox univariate regression analysis showed that Age, Sex, TBIL, and NLR significantly influenced mortality risk in PBC patients, while multivariate Cox regression confirmed Age, TBIL, and NLR as independent predictors of death.

As individuals age, the immune system undergoes progressive decline, characterized by diminished functionality of T and B cells, reduced immune tolerance, and elevated levels of pro-inflammatory cytokines. This state of immunosenescence exacerbates autoimmune responses and promotes disease progression in PBC (10, 11). Additionally, the regenerative capacity of hepatocytes in elderly patients is significantly impaired, leading to diminished liver repair mechanisms and accelerated progression of fibrosis and cirrhosis (12). Furthermore, older patients often present with multiple comorbidities which may worsen the severity of PBC, complicate treatment, and increase the risk of mortality (13).Mei Lu (14) et al. observed 3,488 patients with PBC across different ethnicities, categorizing their ages into ≤40 years,41-50 years,51-60 years,61-70 years and >70 years. The prevalence of PBC was found to increase from 3% to 7.5%.Nikolaos K. Gatselis et al. (15) conducted a retrospective study using a database from a prospective study, involving 482 patients with PBC, and found that advanced age was a poor prognostic factor for liver cirrhosis. Similarly, a study conducted in Hong Kong demonstrated that as age increased, the prevalence of PBC rose, and age was found to be a risk factor for overall mortality (HR 1.03, 95% CI: 1.01–1.04) and non-transplant mortality (HR 1.01, 95% CI: 1.0002–1.02) (16). In our study, ROC curves constructed in the modeling group using Prism indicated a cutoff value of 57.5, with patients over 58 showing an increased risk of mortality. This finding aligns with previous reports (17).

The Neutrophil-to-Lymphocyte Ratio (NLR), calculated as the ratio of absolute neutrophil count to absolute lymphocyte count, serves as an indicator of systemic inflammation. Elevated NLR levels suggest an intensified systemic inflammatory response. In the context of PBC, prolonged inflammation may contribute to liver cell injury and exacerbate fibrotic progression (18). Furthermore, a high NLR may reflect suppressed immune function, particularly a decline in lymphocyte-mediated anti-inflammatory and immune-regulatory activities, thereby accelerating disease progression (19). In advanced stages of PBC, infections are commonly observed, and an elevated NLR may signal either an ongoing infection or a worsening inflammatory response, both of which can negatively impact prognosis (20).A high NLR, resulting from increased neutrophils or decreased lymphocytes, is linked to poor outcomes in several cancers (21–23). In a study on hepatocellular carcinoma, Cheng-Hsiang Lo et al. (24) found that both continuous (p=0.006) and categorical (p=0.003) NLR values significantly correlated with survival. Patients with an NLR ≥ 2.4 exhibited an OS of 38.2%, compared to a 1-year survival rate of 83.6% in those with an NLR < 2.4 (p < 0.001). NLR also independently predicted liver-related toxicity (p = 0.002). In lung cancer patients, a high NLR was strongly linked to worse prognosis, with an HR of 1.798 (95% CI: 1.284-2.518, p=0.001) (25). Elevated NLR (>2.46) was an independent risk for reduced non-transplant survival, and when combined with UK-PBC and GLOBE scores, it enhanced prognostic accuracy (26). Despite its association with adverse outcomes, NLR is less frequently reported in PBC studies. Here, we categorized patients using an NLR cutoff of 2 based on Qian Wei et al. (9), grouping them into NLR >2 and NLR ≤2. Findings indicated that NLR was a significant mortality risk factor in PBC, likely due to increased infection risk during cirrhotic decompensation.

Total bilirubin (TBIL) is identified as a key prognostic marker for survival in PBC patients (27, 28). In PBC patients, the intrahepatic bile ducts sustain damage, which disrupts normal bile secretion processes. This impairment leads to an elevation in TBIL levels, indicating a decline in the liver’s metabolic and detoxification capabilities. Consequently, there is an accumulation of endogenous toxins, which exacerbates oxidative stress, ultimately resulting in further hepatic damage (29).In a multicenter study of 4,199 PBC patients treated with UDCA, Willem J. Lammers et al. (30) found that both age (HR: 1.05, 95% CI: 1.04-1.06) and serum bilirubin above ULN (HR: 2.56, 95% CI: 2.22-2.95) were risk factors for death. In another study by Willem J. Lammers et al. (31), 1118 of 4845 patients with PBC were found to end up with liver transplantation or death. It was also found that bilirubin levels had a strong relationship with clinical outcome. The 10-year survival rate for PBC patients with normal bilirubin levels was 86% compared to 41% for those with elevated level. Numerous guidelines, response criteria, and risk scores also support the stabilization or reduction of bilirubin levels to lower disease progression risk (32–35). Carla F. Murillo Perez et al. (36)found that Age, TBIL, and ALP were better markers of 1-year survival in PBC patients. They created the “ABA” tool that categorized the patients into three groups. Low-risk patients were defined as those over 50 years of age with bilirubin ≤1x ULN and ALP ≤3x ULN. High-risk patients were those younger than 50 years and with bilirubin >1x ULN and ALP >3x ULN, while all others fell into the medium-risk category. Initial 10-year survival rates were 89% for the low-risk group, 77% for the intermediate group, and 59% for the high-risk group, with respective 1-year survival rates of 86%, 76%, and 40%.All these indicate that serum bilirubin is an indicator for predicting the prognosis of PBC. In this study, TBIL >10x ULN (188 μmol/L) was shown to be a notable predictor for mortality.

The infiltration of inflammatory cells into areas of hepatic necrosis is a characteristic feature of cholestatic liver injury (37), likely caused by cell swelling, apoptosis, membrane integrity disruption, and release of cellular components due to cholestasis. Damaged cells release pro-inflammatory mediators via damage-associated molecular patterns (DAMPs), which subsequently stimulate immune cells, including macrophages, neutrophils, and natural killer cells, triggering an inflammatory cascade (38). Among the variables included in this study, TBIL and NLR were most strongly correlated. We also found that the NT index, a newly created variable based on these two indicators, revealed differing prognoses across subgroups: best for the LN-LT group, followed by the LN-HT/HN-LT group, and worst for the HN-HT group. This finding is consistent with research on inflammatory mechanisms in cholestasis.

The UK-PBC (32) and GLOBE scores (30), while useful, are complex and challenging to calculate clinically. In this study, the AUC values of the GLOBE score were lower than nomograms in both the validation and training cohorts. Although the UK-PBC score had a higher AUC than the nomogram in the modeling group, it was lower in the validation group. This indicated that neither the UK-PBC nor GLOBE scores were as accurate as the nomogram for this patient cohort, making them less favorable for clinical use. The “ABA” tools of Carla F. Murillo Perez et al. (36) were only used to predict the 1-year prognosis and were not effective for the 5-year prognosis. We constructed a nomogram with three objective variables, Age, NLR, and TBIL, to assess the 5-year OS of patients, the nomograms achieved an AUC of 0.7251 in the training cohort and 0.7721 in the validation cohort, indicating good consistency. The higher AUC value in the validation group was more related to the smaller population in the validation cohort after randomization. The predictions of the calibration curve for 5-year survival in the training cohort were in general agreement with the actual observations. The same results were obtained in the validation group. Simultaneously, we used DCA to assess the model’s clinical utility in predicting 5-year survival for PBC patients. DCA showed clinical value in the modeling group, but the validation group had a narrower threshold, suggesting larger studies are needed to validate this model. The accuracy of the model was thus considered accurate. Utilizing a nomogram threshold of 116.3, patients were stratified into low- and high-risk categories. In the training cohort, 5-year OS were 93.2% for low-risk patients and 75.9% for high-risk patients, showing a statistically significant difference (p < 0.05). Likewise, in the validation cohort, the 5-year survival rates were 95.6% for the low-risk group and 81.0% for the high-risk group. Both groups showed statistically significant differences. The high-risk group had no LN-LT cases, while the low-risk group had no HN-HT cases. HN-HT was prevalent in 56% of the high-risk group, and LN-LT in 57% of the low-risk group, suggesting better 5-year survival with LN-LT and poorer with HN-HT.

This study further analyzed patient responses at 6 and 12 months. Prior research has indicated that the biochemical response at 6 months holds similar predictive value to the response at 12 months (39). We examined whether differences existed between low- and high-risk groups across different time points and with identical biochemical responses. In the training cohort, among those with non-response at 6 months based on the Barcelona criteria, the 5-year OS was 95% in the low-risk group and 74.3% in the high. Among those with a biochemical response at 6 months, the 5-year OS was 92.6% for low-risk and 76.9% for high-risk patients. At the 12-month mark, non-responsers had a 5-year OS of 93.3% in the low-risk group and 73% in the high-risk group. In contrast, responders was 93.1% in the low and 77.4% in the high-risk group.

In the validation cohort, non-responders at 6 months had 5-year OS rates of 91.3% for low-risk and 65.6% for high-risk patients. For patients with a biochemical response at 6 months, the OS reached 100.0% in the low-risk group and 90.0% in the high-risk group. Furthermore, in the poor responder group at 12 months, the 5-year survival rate was 94.3% for low- and 54.2% for high-risk patients. Conversely, within the 12-month responder group, the survival rate was 96.7% for low-risk and 95.7% for high-risk individuals. Significant differences in OS between low- and high-risk groups were observed for response and non-response cases at most time points, except for the 12-month non-response group, likely due to the smaller sample size after stratification.

In the modeling, validation, and different biochemical response groups, excluding the poor responders at 12 months, the low and high-risk groups for PBC patient mortality within 5 years demonstrated statistically significant differences, indicating that the risk stratification for 5-year mortality is meaningful. This study confirms that a patient’s 5-year survival rate can be assessed based on three key indicators: Age, TBIL, and NLR at the time of diagnosis. To further promote the clinical application of this model, we plan to develop a calculation tool (e.g., a mini-program or calculator) based on the nomogram model. This tool will integrate the three core parameters—Age, NLR, and TBIL—to enable rapid risk scoring and classification, while providing intuitive risk levels and survival rate predictions to facilitate clinical decision-making.

Despite the recognized multifactorial nature of mortality determinants in patients with PBC, ALP has been established as an independent prognostic factor in prior studies (40). Though baseline ALP levels between our training and validation cohorts demonstrated uncomparable distributions (P=0.184), multivariate Cox regression analyses did not retain ALP as an independent mortality predictor. Notably, subgroup comparisons of sex-based ALP reference ranges and serial ALP trajectory analyses were not conducted in this investigation. Given the underrepresentation of male patients (n=52, 17%) in our PBC cohort, the methodological constraints preclude robust sex-stratified interpretation. Future work will expand enrollment to evaluate sex-adjusted ALP threshold effects and validate the generalizability of the dynamic stratification strategy based on ULN of ALP across multiple centers. This study has other limitations, as it is based on a single-center analysis with a relatively small sample size. Given that the patients treated at our center often have more severe conditions, there may be inherent biases. To further elucidate the impact of Age, TBIL, and NLR on mortality in PBC patients, we plan to implement regular model updates to continuously optimize its performance. These updates will involve integrating data from multiple regional medical centers to enhance sample diversity and representation, as well as establishing an independent external validation cohort to further evaluate the model’s predictive accuracy and generalizability.