Exocrine pancreatic insufficiency (EPI) is a common complication of chronic pancreatitis, significantly affecting patients' quality of life. Early identification of patients at risk is essential, particularly in primary care. Aim: To develop a personalized prognostic model for assessing the risk of EPI in middle-aged patients with chronic pancreatitis based on clinical, laboratory, and functional parameters. Methods: The study included 114 patients treated at the Ternopil Primary Health Care Center. Clinical evaluation, laboratory tests (hemoglobin, triglycerides, HbA1c), coprogram scores, abdominal ultrasound, and questionnaires (PEI-Q, GSRS) were analyzed. Multivariate regression identified predictors of reduced fecal elastase levels. Model performance was evaluated using ANOVA, diagnostic accuracy metrics, and ROC curve analysis. Results: Significant predictors included age, hemoglobin, comorbidity index, coprogram score, triglycerides, HbA1c, PEI-Q, and GSRS scores. The final model demonstrated high predictive accuracy (R² = 0.989; p < 0.001), with sensitivity of 95.92%, specificity of 81.25%, and overall accuracy of 93.86%. ROC analysis confirmed strong discriminative ability. Conclusions: The proposed model reliably predicts EPI risk in chronic pancreatitis patients and can support early diagnosis and personalized treatment strategies in primary care settings.
Chronic pancreatitis is a progressive inflammatory disease of the pancreas that leads to structural
damage and gradual loss of organ function [
Traditional methods for detecting EPI have limited sensitivity and are often applied only at
advanced stages of the disease [
The aim of the study was to develop a personalized prognostic model for assessing the risk of exocrine pancreatic insufficiency in patients with chronic pancreatitis, based on clinical, laboratory, and functional parameters. The proposed model is intended to enhance the effectiveness of early detection of exocrine pancreatic dysfunction and support informed decision-making regarding further treatment and prevention in primary care practice.
The study was conducted at the Ternopil Primary Health Care Center. A total of 114 middle-aged patients diagnosed with chronic pancreatitis participated in the study. Written informed consent was obtained from all participants. The study protocol was approved by the local ethics committee.
A comprehensive clinical evaluation included physical examination, laboratory blood tests (hemoglobin, HbA1c, triglycerides), coprological assessment using a scoring system, abdominal ultrasound, and completion of the PEI-Q (to assess exocrine pancreatic insufficiency) and GSRS (to evaluate dyspeptic symptoms) questionnaires.
The key variables incorporated into the regression model for predicting fecal α-elastase levels included: patient age (X₁), hemoglobin level (X₂), comorbidity index (X₃), coprogram score (X₄), triglyceride level (X₅), HbA1c (X₆), PEI-Q score (X₇), and GSRS score (X₈).
Statistical analysis was performed using SPSS software, version 26.0. A multivariate regression analysis was applied to construct the predictive model. Model performance was evaluated through analysis of variance (ANOVA), assessment of operational characteristics (sensitivity, specificity, positive predictive value, negative predictive value, and overall accuracy), as well as receiver operating characteristic (ROC) curve analysis with calculation of the area under the curve (AUC).
A multivariate regression analysis (Table 1) was conducted to identify the most significant predictors of decreased fecal α-elastase levels in middle-aged patients suffering from chronic pancreatitis. The results of the analysis demonstrated no statistically significant correlation between serum albumin levels and fecal α-elastase levels. Consequently, this variable was excluded from the development of the regression model for predicting fecal α-elastase levels in the specified patient cohort.
A subsequent multivariate regression analysis allowed for the identification of the most statistically significant predictors associated with reduced fecal α-elastase levels in middle-aged patients with chronic pancreatitis. Based on the obtained data, regression coefficients were determined to construct a predictive regression model (Table 2).
Beta 522.221 -0.054 -1.749 -0.195 -1.082 -1.513 -12.052 0.045 -61.885 -1.220
Beta 519.815 -0.053 -1.725 -1.116 -1.043
As a result of the conducted multivariate regression analysis, the following predictive model was developed: Y = 519.815 − 0.053X₁ − 1.725X₂ − 1.116X₃ − 1.043X₄ − 1.624X₅ − 11.658X₆ − 62.935X₇ − 1.212X₈ (R = 0.995; R² = 0.989; p < 0.05)
Where:
The results of the ANOVA confirmed a high level of statistical significance (p < 0.001), validating the regression model's predictive accuracy for fecal α-elastase levels in middle-aged patients with chronic pancreatitis. This indicates that the model performs more precisely than using average predictor values alone (Table 3). The analysis of the coefficient of determination (R² = 0.995) further substantiates the high precision of the developed regression model, confirming its adequacy for forecasting fecal α-elastase levels in this specific patient population.
To evaluate the adequacy of the developed predictive model, an analysis of operational characteristics was conducted, including true positives, false negatives, false positives, and true negatives. These indicators were determined by applying logistic regression to the positive and negative outcomes in accordance with the predictive model, using contingency tables for detailed data analysis (Table 4).
The analysis of operational characteristics of the developed regression model for predicting fecal αelastase levels in middle-aged patients with chronic pancreatitis enabled the calculation of aggregated performance indicators (Table 5). A high level of accuracy was found across all investigated operational metrics, confirming the robustness and adequacy of the model for this patient category.
Note: Se – Sensitivity; Sp – Specificity; PPV – Positive Predictive Value; NPV – Negative Predictive Value; Acc – Accuracy.
To assess the predictive value of the developed regression model for forecasting fecal α-elastase levels in middle-aged patients with chronic pancreatitis, a ROC analysis was performed. Based on the generated ROC curves and AUC indicators, a thorough evaluation of model quality was carried out (Figure 1).
ROC analysis demonstrated a high diagnostic value of age in relation to fecal α-elastase levels (AUC = 0.930, p < 0.001) (Figure 1). This indicates a strong association between increasing age and reduced exocrine pancreatic function. These findings highlight the prognostic significance of age as a risk factor for exocrine insufficiency.
Вік
ROC analysis demonstrated an excellent diagnostic value of hemoglobin in relation to fecal αelastase levels (AUC = 0.973, p < 0.001) (Figure 2). This finding indicates a very strong association between hemoglobin concentration and exocrine pancreatic function. These results emphasize the prognostic importance of hemoglobin as a potential marker of exocrine insufficiency. y t i v i t i s n e S
AUC = 0,973
P < 0,001 0 20 40 60 80
100 100-Specificity
AUC = 0,811
P < 0,001 0 20 40 60 80
100 100-Specificity
ROC analysis demonstrated a good diagnostic value of the comorbidity index in relation to fecal αelastase levels (AUC = 0.811, p < 0.001) (Figure 3). This indicates a significant association between higher comorbidity burden and impaired exocrine pancreatic function. These findings suggest the comorbidity index as an important predictor of exocrine insufficiency.
Ідекс коморбідності
ROC analysis revealed a good diagnostic performance of the coprogram result in relation to fecal αelastase levels (AUC = 0.811, p < 0.001) (Figure 4). The data suggest a significant link between abnormal coprogram findings and impaired exocrine pancreatic activity. These outcomes underline the utility of coprogram evaluation as a supplementary marker of exocrine insufficiency.
ROC analysis demonstrated a satisfactory diagnostic value of triglyceride levels in relation to fecal α-elastase (AUC = 0.760, p = 0.012) (Figure 5). This finding indicates a moderate association between elevated triglycerides and impaired exocrine pancreatic function. These results point to triglyceride concentration as a potential auxiliary marker of exocrine insufficiency.
AUC = 0,811
P < 0,001 0 20
40 60 100-Specificity 80
100 ТГ
AUC = 0,760
P = 0,012 0 20 80
100 0 20
AUC = 0,915 P < 0,001 80
100
ROC analysis revealed a strong diagnostic performance of HbA1c in relation to fecal α-elastase levels (AUC = 0.915, p < 0.001) (Figure 6). The results demonstrate a clear association between higher HbA1c values and diminished exocrine pancreatic function. These data emphasize the relevance of
ROC analysis indicated a robust diagnostic value of the PEI-Q score for predicting fecal α-elastase levels (AUC = 0.871, p = 0.002) (Figure 7). The data demonstrate a clear association between higher PEI-Q scores and decreased exocrine pancreatic function. These results underscore the utility of the PEI-Q score as an effective predictor of exocrine insufficiency.
100
80 ty 60 iiit v s n eS 40 20
0 100 80 iiitty 60 v s n e S 40 20 0
HbA1c
AUC = 0,871 P = 0,002 80
100
ROC analysis demonstrated a good diagnostic value of the GSRS (Dyspepsia Syndrome) score in relation to fecal α-elastase levels (AUC = 0.844, p < 0.001) (Figure 8). This indicates a significant association between higher dyspepsia scores and impaired exocrine pancreatic function. These findings support the GSRS (Dyspepsia Syndrome) score as a useful predictor of exocrine insufficiency.
GSRS (синдром диспепсії) 100 80 iiitty 60 v s n eS 40 20 0 0 20
AUC = 0,844 P < 0,001 80
100
Based on the generated ROC curves and AUC indicators, a thorough evaluation of model quality was carried out. The results confirmed a high level of accuracy in prediction across all predictors based on the AUC values (Table 6).
Note: AUC – Area Under the Receiver Operating Characteristic Curve.
The study successfully developed and validated a personalized prognostic model for evaluating the risk of exocrine pancreatic insufficiency in middle-aged patients with chronic pancreatitis, based on clinical, laboratory, and functional indicators. The multivariate regression model demonstrated a high predictive capacity (R² = 0.989; p < 0.001), with statistically significant contributions from variables
(Dyspepsia
rome) such as age, hemoglobin level, comorbidity index, coprogram score, triglyceride level, HbA1c, PEI-Q score, and GSRS score.
The model exhibited strong diagnostic performance, with high sensitivity (95.92%), specificity (81.25%), positive predictive value (96.91%), and overall accuracy (93.86%). ROC curve analysis further confirmed the model’s high diagnostic power across all included predictors.
These findings underscore the clinical value of the developed model as an effective tool for early detection of exocrine pancreatic dysfunction in patients with chronic pancreatitis. Its integration into primary care practice may enhance diagnostic accuracy, facilitate timely therapeutic decisions, and improve prevention strategies tailored to individual patient risk profiles.
[2] Babinets L, Halabitska I. Characteristics of joint pain in patients with primary osteoarthritis and comorbid conditions with exocrine pancreatic insufficiency. Lekarsky Obzor. 2021;70(2):62-4. [3] Redkva OV, Babinets LS, Halabitska IM. Evaluation of parameters of actual typical pathogenetic syndromes in comorbidity of type 2 diabetes mellitus and chronic pancreatitis. Wiadomosci lekarskie (Warsaw, Poland : 1960). 2021;74(10 cz 2):2557-9. [4] Capurso G, Traini M, Piciucchi M, Signoretti M, Arcidiacono PG. Exocrine pancreatic insufficiency: prevalence, diagnosis, and management. Clinical and experimental gastroenterology. 2019;12:129-39. [5] Babinets LS, Halabitska IM, Kotsaba YY, Borovyk IO, Migenko BO, Ryabokon SS, et al. The effect of the proteolisis' system activity for the trophological status of patients with osteoarthrosis and excretory insufficiency of pancreas. Wiadomosci lekarskie (Warsaw, Poland : 1960). 2018;71(2 pt 1):273-6. [6] Mienye ID, Obaido G, Jere N, Mienye E, Aruleba K, Emmanuel ID, et al. A survey of explainable artificial intelligence in healthcare: Concepts, applications, and challenges. Informatics in Medicine Unlocked. 2024;51:101587. [7] Babinets LS, Shaihen OR, Homyn HO, Halabitska IM. Specific aspects of clinical course in case of combination of chronic pancreatitis and concomitant viral hepatitis C. Wiadomosci lekarskie (Warsaw, Poland : 1960). 2019;72(4):595-9. [8] Min M, Patel B, Han S, Bocelli L, Kheder J, Vaze A, et al. Exocrine Pancreatic Insufficiency and Malnutrition in Chronic Pancreatitis: Identification, Treatment, and Consequences. Pancreas. 2018;47(8):1015-8.