A novel scoring system as a predictor for delayed pain disappearance after percutaneous balloon compression for trigeminal neuralgia Authors: Tao Wu, Yundong Yang, Shizhong Zhang, Xiaoliang Liu, Guojun Wang, Wenming Lv Published: 2025 ABSTRACT AbstractBackgroundThe objective of this study was to establish a scoring system for predicting the risk of delayed pain disappearance (DD) after percutaneous balloon compression (PBC) for trigeminal neuralgia.MethodsA retrospective analysis was conducted on 282 patients with trigeminal neuralgia who underwent PBC, with 217 in the modeling cohort and 65 in the validation cohort. Patients in the modeling cohort were grouped according DD status (DD and non-DD groups). Logistic regression analysis was used to identify risk factors for DD and obtain corresponding regression coefficients. Based on these regression coefficients, a scoring system was established and validated in both cohorts.ResultsFifty-nine patients in the modeling cohort experienced DD. In the logistic regression analysis, pain involving V2 (P = 0.032), non-pear-shaped balloon (P = 0.029), history of radiofrequency therapy (P = 0.002), diabetes (P = 0.039), disease duration ≥ 8 years (P = 0.004), and herpes simplex virus infection (P < 0.001) were identified as independent risk factors for DD. A scoring system was established, consisting of the six variables above. Each factor was assigned a score, with the total score ranging from 0 to 11. The scoring system demonstrated good discriminative and calibration abilities. Using a cutoff score of 6, internal validation demonstrated satisfactory diagnostic sensitivity, specificity, accuracy, positive predictive value, negative predictive value, as well as strong discriminative and calibration capabilities. Analysis of the validation cohort confirmed the scoring system’s diagnostic sensitivity (68.8%), specificity (79.6%), accuracy (76.9%), positive predictive value (52.4%), and negative predictive value (88.6%), with positive and negative likelihood ratios of 3.09 and 0.39, respectively.ConclusionsThe scoring system we developed for predicting DD after PBC shows strong predictive ability and calibration. This tool is essential for improving postoperative evaluation and averting the need for retreatment.Keywords: Percutaneous balloon compression, Trigeminal neuralgia, Risk factors, Scoring system Introduction Trigeminal neuralgia (TN) is a chronic facial pain disorder characterized by sudden, electric shock-like pain in the area innervated by the trigeminal nerve [1]. Pharmacotherapy is the primary treatment for TN. When medication is ineffective or causes intolerable side effects, surgical intervention becomes a preferable option. Surgical treatments include microvascular decompression (MVD), radiofrequency thermocoagulation (RFT), percutaneous balloon compression (PBC), and stereotactic radiosurgery [2]. PBC involves selective compression of damaged myelinated pain-transmitting nerve fibers by inflating a balloon within Meckel’s cave [3, 4] while preserving the motor branch of the trigeminal nerve [5]. Pain is alleviated by deactivating the trigger mechanism of the trigeminal nerve pain pathway. The reported rates of immediate pain relief after PBC range from approximately 74%–100% [6–8]. Gradual pain disappear after the procedure is referred to as “delayed pain disappearance (DD)” [9], which reportedly occurs in 9.31% to 19% of patients undergoing PBC [10, 11]. After 2–3 min of balloon compression of the trigeminal ganglion, mechanical compression injury occurs in the nerve fibers, leading to changes in the nerve cell membrane potential and the function of ion channels [12]. Stress and inflammatory responses are also initiated. Cytokines such as IL-2, IL-4, IL-10, tumor necrosis factor (TNF), and neurotransmitters such as 5-HT released by nerve cells continue to act in the microenvironment, which may constitute the molecular mechanisms underlying DD [13–15]. In a previous study, the average duration of DD was approximately 9 days [16]. However, our preliminary research found that DD occurs within 30 days of PBC. Similarly, after MVD, Esmaeilzadeh et al. reported a median DD duration of 1 month (range, 4 days to 3 months) [17]; they suggested that right-sided pain and vascular contact at the root entry zone of the nerve may be key factors contributing to the delayed onset of MVD effects. Preliminary study indicated that factors such as a history of RFT, diabetes, and herpes zoster influence the occurrence of DD after PBC [9, 18]. Despite these findings, a comprehensive scoring system to assess the risk of DD after PBC is lacking. This study aimed to identify risk factors associated with DD after PBC and to establish a scoring system to accurately predict it. Materials and methods Patients A total of 282 TN patients who underwent PBC performed by the same neurosurgeon between 2017 and 2023 were included for analysis. Based on the timing of surgery and follow-up, data from 217 early stage patients were selected to establish the scoring system; data from another 65 patients were used for external validation. This study was approved and conducted by the hospital’s ethics committee. (KYLL-202204-034).All participants were familiar with and supported the present study and signed the informed consent. Inclusion and exclusion criteria Inclusion Criteria: (1) TN patients whose medical history, symptoms, physical examination findings, and brain imaging (CT and MRI) conform to the diagnostic guidelines of the Headache Classification Committee of the International Headache Society[19]; (2) Patients who have not responded adequately to oral carbamazepine treatment or have experienced significant adverse effects from this medication; (3) The procedure must be performed by the same chief physician of the surgical team, and all patients must provide informed consent to participate in the study. Exclusion criteria: (1) patients who cannot tolerate general anesthesia or who exhibit significant dysfunction in vital organs (such as the heart, lungs, liver, or kidneys), or those with noticeable neurological impairments; (2) secondary and idiopathic TN patients; and (3) patients with incomplete medical records or insufficient follow-up data. PBC procedure Patients were positioned supine, and the Hartel puncture technique was employed following general anesthesia [20]. A No. 14 needle was inserted into the cheek, 3 cm lateral to the mouth’s angle, under fluoroscopic guidance using a C-arm machine (Ziehm 8000, Germany). The needle was advanced toward the foramen ovale. Subsequently, a 4-F Fogarty catheter connected to a T-joint was carefully advanced into Meckel’s cave along the same path. Gadodiamide (0.4 mL–0.8 mL) was slowly injected to inflate the balloon into a pear shape, maintaining compression for 3–5 min. After inflation, the gadodiamide was withdrawn, and both the catheter and puncture needle were removed. A sterile dressing was applied to the puncture site and compressed for 10 min. Outcome evaluation and data collection After PBC, patients were assessed for pain severity immediately upon awakening from anesthesia and then daily during hospitalization. After discharge, follow-ups were conducted via phone, WeChat, or outpatient visits—monthly in the first 3 months after surgery, and then every 3 months until symptom recurrence or disappearance. If the patient’s pain showed no improvement for more than 6 months after PBC, it was considered ineffective and further treatment strategies were developed. Patients were categorized based on their pain response to PBC as either immediate disappearance, DD, or no improvement. Pain was classified using the Barrow Neurological Institute (BNI) pain intensity score (Table 1). BNI grade I indicates no pain. Immediate disappearance was defined as achieving BNI grade I immediately after awakening from anesthesia. DD referred to patients in whom BNI grade I was gradually achieved within 30 days of the procedure. Patients who did not experience immediate disappearance or DD were categorized as no improvement. Table 1BNI pain intensity scale scoreScoreDefinitionINo pain, no medicationIIOccasional pain, not requiring medicationIIISome pain, adequately controlled with medicationIVSome pain, not adequately controlled with medicationVSevere pain/no pain reliefBNI, Barrow Neurological Institute BNI pain intensity scale score BNI, Barrow Neurological Institute We recorded various factors that could influence DD, including age, gender, disease duration, laterality of pain, previous treatments, comorbidities, and volume and shape of the balloon used during PBC. The study data were collected from January 1, 2018 to June 30, 2023. Statistical analysis Statistical analyses were performed using SPSS software version 22.0 (IBM Corp., Armonk, NY, USA) with two-sided testing and a significance level of α = 0.05. To establish the scoring system, the 217 early stage patients were divided into a DD group (59 cases) and a non-DD group (158 cases) for univariate and multivariate analyses, resulting in a logistic regression equation. The regression coefficients (β values) of the independent variables in the equation were used for scoring. The variable with the smallest absolute β value was taken as the reference, and its risk score was assigned as 1 point. The risk scores of other variables were calculated as β value of the variable/β value of the reference variable, with results rounded to the nearest integer. The Hosmer–Lemeshow (H–L) goodness-of-fit test was used to evaluate the calibration of the scoring system [21]. A p value > 0.05 indicated no statistically significant difference between predicted and observed values, suggesting good calibration [7]. The discriminatory ability of the scoring system was assessed using the area under the receiver operating characteristic (ROC) curve (AUC). An AUC between 0.5 and 0.7 indicated low discrimination, 0.7–0.9 indicated good discrimination, and 0.9–1 indicated high discrimination. The maximum Youden index was calculated from the ROC curve to determine the cutoff value for DD. Diagnostic sensitivity, specificity, accuracy, and other indicators were calculated based on this cutoff value. In the 65 patients of the external validation group, corresponding statistical indicators were utilized to assess the quality of the scoring system, followed by the H–L goodness-of-fit test and ROC curve analysis. Using predetermined thresholds, the diagnostic sensitivity, specificity, accuracy, and other relevant metrics of the scoring system were calculated. Results Patient characteristics Patient characteristics are shown in Table 2. Table 2The demographic characteristics of 282 patientsMGVGGender M9121F12644Age (range: years) 65 (33 ~ 89)64 (32 ~ 88)≥ 6514245 < 657520Disease duration (range:years) 8 (0.04 ~ 50)8 (0.06 ~ 30)Laterality Left side7919Right side13846Division V142V27528V31213V1+2578V1+322V2+3387V1+2+3295Previous treatments MVD598RFT5915PBC32GG20Therapeutic effect of drugsNo response11043Multiple side effects7514Concomitant disease:Hypertension7915Diabetes mellitus5914Coronary heart disease359Cerebral infarction2317Brain atrophy53Herpes zoster11517Chronic bronchitis57MG, modeling group; VG, validation group; M, Male; F, Female; V1, ophthalmic branches; V2, maxillary branches; V3, mandibular branches; MVD, Microvascular decompression, RFT, radio frequency thermocoagulation, PBC, Percutaneous balloon compression, GG, Glycerol gangliolysis The demographic characteristics of 282 patients MG, modeling group; VG, validation group; M, Male; F, Female; V1, ophthalmic branches; V2, maxillary branches; V3, mandibular branches; MVD, Microvascular decompression, RFT, radio frequency thermocoagulation, PBC, Percutaneous balloon compression, GG, Glycerol gangliolysis Risk factors for DD Among the 217 patients used to establish the scoring system (who were grouped as DD or non-DD), univariate and multivariate logistic regression analyses revealed that involvement of V2 (P = 0.032), non-pear-shaped balloon (P = 0.029), history of RFT (P = 0.002), diabetes (P = 0.039), disease duration ≥ 8 years (P = 0.004), and herpes simplex virus (HSV) infection (P < 0.001) were independent risk factors for DD (Table 3). Based on the logistic regression analysis results, we derived the equation: where X1,X2,X3,X4,X5,andX6 represent the following factors: involvement of V2, non-pear-shaped balloon, history of RFT, diabetes, disease duration ≥ 8 years, and HSV infection, respectively. Table 3Univariate and multivariate logistic regression analysis for disappearance of delayed pain in PBCVariableUnivariate analysisMultivariate analysisOR95% CIPβOR95% CIPLowHighLowHighAge(≥ 65 years)0.9750.5141.8500.938Sex: M1.2070.6612.2050.542duration ≥ 8 years2.8231.5275.2200.001*1.6795.3611.68417.0690.004*MVD2.3401.2614.3420.006*− 1.9590.1410.0161.2640.080RFT12.7216.26625.8290.000*3.56535.3243.630342.8070.002*Concomitant diseaseHypertension2.7991.5135.1790.001*0.9552.5990.8318.1290.101Diabetes mellitus4.9322.37910.2250.000*1.4044.0711.07315.4410.039*Herpes zoster5.2760.93915.4910.000*4.43284.1399.622735.7140.000*Coronary heart disease1.3010.5293.1980.568Left side0.7760.4121.4620.434Involved V211.9541.58690.0900.002*2.92618.6511.286270.4340.032*Balloon volume ≤ 0.4 ml1.0700.5851.9560.827Non-"pear-shaped" balloon16.0997.54334.3610.000*1.3013.6731.13911.8430.029*M, male; MVD, microvascular decompression; RFT, radio frequency thermocoagulation; V2, maxillary branches of trigeminal nerve; PBC, percutaneous balloon compression*P < 0.05 was statistically significant Univariate and multivariate logistic regression analysis for disappearance of delayed pain in PBC M, male; MVD, microvascular decompression; RFT, radio frequency thermocoagulation; V2, maxillary branches of trigeminal nerve; PBC, percutaneous balloon compression *P < 0.05 was statistically significant Construction of the scoring system According to the Framingham Heart Study method, the reference values were set as follows: involvement of V2, non-pear-shaped balloon, history of RFT, diabetes, disease duration ≥ 8 years, and HSV infection were assigned a value of 1. Absence of V2 involvement, pear-shaped balloon, no history of RFT, no diabetes, disease duration < 8 years, and no HSV infection were assigned a value of 0. The constant βj was set as the minimum parameter of 1.301. The formula (Wij–WiFRR) × β/βj was applied, and the resulting score was rounded to the nearest integer. Based on the β values corresponding to the above six variables, each variable was assigned a value to establish a risk score, which ranged from 0 to 11 points (Table 4). The predicted probability of the scoring system was 0.014 and the AUC was 0.963 (95% confidence interval [CI] 0.936–0.990; P < 0.05; Fig. 1). AUC > 0.750 indicates good discriminatory ability. Calibration as assessed by the H–L test showed χ2 = 10.525 and P = 0.230, indicating good calibration. The sensitivity and specificity of the scoring system was 84.7% and 94.9%, respectively. The positive and negative predictive values were 86.2% and 94.34%, respectively, resulting in an accuracy of 92.2%. The positive and negative likelihood ratios were 5.36 and 0.06, respectively. Table 4Screening scoring systemVariableReference (Wij)Parameter (β)Risk distance (Wij-WiFRR)×βRisk score (value)Involved V2: no/yes0/12.926/−0/2.9260/2Non-“pear-shaped”balloon: no/yes0/11.301/−0/1.3010/1RFT: no/yes0/13.565/−0/3.5650/3Diabetes mellitus: no/yes0/11.404/−0/1.4040/1Symptom duration: < 8 years/ ≥ 8 years0/11.679/−0/1.6790/1Herpes zoster: no/yes0/14.432/−0/4.4320/3 Screening scoring system ROC curve of predication model Scoring system validation Using the total score as the diagnostic variable, the model yielded an AUC of 0.950 (95% CI 0.922–0.979; Fig. 2A), indicating good discriminative ability. Calibration as assessed by the H–L test showed P = 0.501, suggesting good calibration. The maximum Youden index was calculated based on the ROC curve. Using a cutoff value of 6, risk of DD was categorized as low (6–11 points) or high (0–5 points). DD detection rates in the high- and low-risk groups were 66.6% and 2.3%, respectively (χ2 = 107.895; P < 0.001). The sensitivity, specificity, positive predictive value, negative predictive value, and accuracy of the risk scoring system was 94.9%, 82.3%, 72.6%, 97.7%, and 85.7%, respectively. Calibration as assessed by the H–L test showed χ2 = 6.338 and P = 0.501, indicating good calibration. When applying the scoring system to evaluate DD risk after PBC in the validation group (n = 65), scores ranged from 0 to 11 points. The AUC was 0.805 (95% CI 0.683–0.928; P < 0.05; Fig. 2B), indicating good discriminative ability. Using the same cutoff value to categorize high- and low-risk groups as above, the DD detection rates in these groups were 52.4% and 11.4%, respectively (χ2 = 12.889; P < 0.001). The scoring system showed sensitivity, specificity, positive predictive value, negative predictive value, and accuracy of 68.8%, 79.6%, 52.4%, 88.6%, and 76.9%, respectively. The positive and negative likelihood ratios were 3.09 and 0.39, respectively. ROC curve of internal test and external test. A: ROC curve of internal test; B: ROC curve of external test. Discussion DD is a common phenomenon after treating patients with TN. The reported incidence rate of DD ranges from 2.9% to 40.6%. Most studies have focused on analyzing risk factors for DD in patients undergoing MVD [22–24]; however, significant discrepancies exist among their findings. Although scoring systems have been designed to assist clinicians in identifying TN patients suitable for MVD or Gamma Knife radiosurgery [25, 26], there is currently no scoring system for predicting DD after PBC. We found that involvement of the V2 branch, non-pear-shaped balloon morphology, history of RFT, diabetes mellitus, disease duration ≥ 8 years, and HSV infection are independent risk factors for DD after PBC. In addition, we developed a risk prediction scoring system for post-PBC DD, which quantifies the risk of DD occurrence, aiding clinicians in decision making and preventing premature interventions for TN patients whose pain does not resolve immediately after PBC. Factors influencing DD For patients with TN, the duration of neuralgia affects the degree of nerve myelination as well as the degree of psychological impact on the patient [27]. A longer duration of symptoms is associated with an increased risk of DD after PBC. Broggi et al. and Bederson et al. each reported that patients with a longer disease duration had lower rates of postoperative pain relief and higher recurrence rates [28, 29]. Similarly, Mureb et al. noted that a shorter disease duration predicts better pain relief outcomes after MVD [30]. Our findings are consistent with the above conclusions, as the patients in this study with a disease duration > 8 years had a higher risk of DD after PBC. Patients with TN who have had the condition for a longer duration tend to have more severe demyelination of the trigeminal nerve, which can generate ectopic impulses capable of spontaneous pain, leading to excessive excitation of the afferent component of the nerve [31]. PBC causes mechanical damage, leading to changes in neuronal membrane potential. However, excessive excitation of the afferent nerve component in some neurons may persist for a period of time [32]. Regeneration of myelin by Schwann cells after the procedure may block the transmission of these neural impulses, which takes time, thus, the disappearance of pain is delayed. Thermal injury has also been implicated in the process of myelin sheath damage and regeneration [33]. RFT involves inserting an electrode to generate an electric current, which creates an electric field with thermal energy that induces ion migration, thereby blocking the transmission of pain signals along axons. The thermal coagulation selectively damages unmyelinated Aδ and C nerve fibers responsible for pain conduction while sparing the heat-resistant myelinated Aα and Aβ fibers that mediate touch and motor sensation [34]. Postoperative microscopic examination has revealed varying degrees of damage to the myelin sheath, axons, and connective tissue surrounding Schwann cells. This further induces an increase in c-Fos immunoreactivity and leads to cellular stress. It is hypothesized that such cellular stress, along with the regeneration of connective tissue and nerve fibers after RFT, may cause local adhesions in Meckel’s cavity [35]. These adhesions could prevent the balloon from achieving optimal compression of the trigeminal ganglion during PBC, potentially contributing to DD. Therefore, for patients with a history of RFT, special attention should be paid during PBC to the balloon’s position, pressure, shape, and compression time to mitigate the impact of prior RFT. If a longer duration of symptoms and RFT directly affect the neural fiber structure of the trigeminal ganglion, then diabetes and HSV infection exert their effects by altering the humoral microenvironment surrounding the nerves. Diabetic peripheral neuropathy (DPN) is the most common complication of diabetes. Approximately two-thirds of patients with diabetes experience it [36]. Although involvement of the nerves of the upper and lower extremities is most common, the trigeminal and facial nerves may be involved as well [37]. In this study, all types of diabetes were uniformly regarded as a risk factor—type was not distinguished. Hyperglycemia in patients with diabetes can lead to neuronal inflammation, elevated glycosylated hemoglobin, and mitochondrial dysfunction, leading to sensory nerve hyperalgesia and neuralgia [38]. The conduction failure between A β, A δ myelinated fibers, and unmyelinated C fibers caused by TN-associated demyelination [21] is consistent with the type of nerve damage in peripheral neuropathy involving small fiber A δ and C-class nerve fibers affected by DPN [39], and experimental evidence has shown that DPN can cause Schwann cell apoptosis, myelin damage, and impaired neuronal cellular repair [40]. Electrophysiological experiments have shown that a high glucose environment affects trigeminal nerve function, and that high glucose itself can enhance the sensitivity of patients to pain, making the pain threshold of diabetic patients significantly lower than that of normal people [41, 42]. This is also one of the possible explanations for DD after treatment. Therefore, we believe that the extracellular environment of high glucose in patients with diabetes, the stimulation of related cytokines, and the increased pain sensitivity of diabetic patients may be the reasons for DD after PBC in in patients with TN complicated with diabetes. After infecting a host, the HSV often remains latent in nerve cells; HSV-1 is actually capable of persisting in the trigeminal ganglion for years. HSV reactivation after surgical treatment of TN is widely recognized [43]. Because of balloon compression-induced direct mechanical damage to trigeminal ganglion neurons [44], stress occurs in the surrounding cells, leading to increased secretion of pro-inflammatory cytokines, such as interferon and IL-6. This, in turn, triggers the reactivation of HSV-1 from its latent state. In addition, the decline in the patient’s immune function after treatment impairs the effectiveness of HSV-specific CD8 + T cells residing in the trigeminal ganglion [45]. The reactivated virus is transported along nerve fibers to the cells of the perioral skin and mucous membranes, where it proliferates again, causing a recurrence of HSV-1 and potentially even inducing the onset of herpes zoster encephalitis [46]. Furthermore, animal experiments have demonstrated that HSV-1, after infecting the host and remaining latent in the trigeminal ganglion, can reactivate and increase the animal’s sensitivity to TN while lowering the pain threshold [47]. For patients with a history of HSV infection, the use of antiviral medications such as acyclovir after PBC has been shown to alleviate delayed pain symptoms, which disappear as the infection resolves. TN patients with involvement of the V2 branch are at a higher risk for DD after PBC. This increased risk may stem from the mechanical damage that balloon compression can inflict on myelinated nerve fibers. The procedure can also exert traction and pressure on the trigeminal ganglion, potentially disrupting its blood supply, which can lead to demyelination and reduced pain conduction. Furthermore, the V2 branch possesses a more intricate network of independent and anastomosing vascular connections than the V1 and V3 branches [48]. As a result, patients with V2 involvement may be more prone to experiencing DD. In addition to the above five preoperative factors, the shape of the balloon is the only intraoperative factor that influences the occurrence of DD. During PBC, the pear-shaped balloon is considered the gold standard for surgical success [49]. The presentation of balloon morphology is influenced by factors, such as balloon position, balloon penetration angle, and filling volume. Owing to the anatomy of Meckel’s cave and the surrounding bones, a balloon presenting with a pear-shaped morphology indicates that it has completely filled Meckel’s cave, exhibiting optimal compression of the nerve vessels on the inner wall [50]. Insufficient filling of Meckel’s cave may only serve as a separation of blood vessels and nerves, causing partial nerve compression. Non-pear-shaped balloons indicate poor compression of the trigeminal ganglion and also have a significant impact on postoperative complications [51]. Therefore, we speculate that non-pear-shaped balloons do not fully fill Meckel’s cave and result in trigeminal nerve damage and incomplete obstruction of blood flow. Post-treatment apoptosis of neurons gradually occurs after a period of time. Therefore, patients with a non-pear-shaped balloon presentation during PBC may experience DD or no pain improvement. The scoring system of DD Based on the aforementioned influencing factors, we have developed a scoring system to predict the risk of DD after PBC. The system incorporates the following six variables: involvement of V2, non-pear-shaped balloon, history of RFT, diabetes, disease duration ≥ 8 years, and HSV infection. In both the model construction and validation cohorts, the system effectively identified patients at high risk of DD, with the AUC consistently exceeding 0.750 in both groups, demonstrating its strong ability to distinguish between high- and low-risk populations. The scoring system exhibited good calibration in both the training and validation sets, and accuracy metrics further supported its excellent performance, particularly in terms of negative predictive value, making it suitable for assessing the risk of DD after PBC. For TN patients who do not experience immediate pain relief after PBC surgery, we recommend using this scoring system for risk assessment. Low-risk patients may consider alternative surgical or RFT treatment options, while high-risk patients are advised to undergo a 1-month observation period. If there is no improvement in pain at that time, further treatment should be sought. This study is the first to analyze the risk factors for DD in TN patients after PBC, and to develop and validate a scoring system in an independent population. Using conventional data collection methods, we ensured that the approach is easy to implement, highly practical, and readily applicable in clinical practice. Limitations This study is a single-center, retrospective analysis. A single medical institution has its specific patient population, geographic location, medical standards, specialty characteristics, and admission criteria. In addition, retrospective studies are “backward-looking,” as the data are derived from historical records generated in the past (such as electronic medical records and archives). Therefore, such studies inevitably introduce selection bias, which is an inherent methodological limitation. In the future, we will further validate the findings by implementing clear and broad inclusion and exclusion criteria, enrolling consecutive cases, conducting multivariate analysis, and performing sensitivity analysis. Second, the sample size of this study is relatively small, particularly for external validation, and there is a lack of validation data from other medical centers. As a result, the efficacy derived from this study may not accurately reflect its predictive performance. Furthermore, the predictive model lacks long-term outcome validation, leading to a “silent degradation” of model performance. Therefore, in the future, we will design a continuous monitoring system, establish a regular re-evaluation mechanism for model performance, develop a clear model iteration plan, and design corresponding mitigation measures to ensure the accuracy and stability of the predictive model. Conclusion The scoring system we developed for predicting DD after PBC shows strong predictive ability and calibration. This tool is essential for improving postoperative evaluation and averting the need for retreatment. Acknowledgements We thank Liwen Bianji (Edanz) (https://www.liwenbianji.cn) for editing the English text of a draft of this manuscript. Author contributions Conceptualization: T. W.and Y. Y.; Methodology: Y. Y. and W.L.; Formal analysis and investigation: X.L. and W.L.; Writing—original draft preparation: T. W.; Writing—review and editing: W.L.,S.Z. and G.W.; Supervision: G.W. and W.L. All authors reviewed the manuscript. Funding This work was supported by the China Postdoctoral Science Foundation (No. 2022M711742 to Xiaoliang Liu); and Youth Science Foundation Cultivation and Support Program of Shandong First Medical University (No. 202221-050 to Xiaoliang Liu); and Shandong Provincial Natural Science Foundation (No. ZR2023QH032 and ZR2024MH206 to Xiaoliang Liu) and Tai’an Science and Technology Innovation Development Project (No. 2022NS207 to Shizhong Zhang). Data availability To protect patient privacy, the data sets generated and/or analyzed during the current study are not publicly available. Data sets that hide sensitive patient information are available from the corresponding author on reasonable request. Declarations Ethics approval and consent to participate The protocol was in accordance with the principles of the Declaration of Helsinki and was approved by the local ethics committee. Competing interest The authors declare no competing interests. Footnotes Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Tao Wu and Yundong Yang contributed equally to this work. Contributor Information Xiaoliang Liu, Email: liuxiaoliang1020@163.com. Guojun Wang, Email: wguojun2002@163.com. Wenming Lv, Email: 252838358@qq.com. References Associated Data Data Availability Statement To protect patient privacy, the data sets generated and/or analyzed during the current study are not publicly available. Data sets that hide sensitive patient information are available from the corresponding author on reasonable request.