Doctors treating amoebic liver abscess face a stubborn problem: the infection can smolder quietly in one patient and turn life-threatening in another, and the usual warning signs do not always separate the two in time. A research team in Iraq has now built and internally validated a multivariable prediction model that combines routine blood chemistry with immune markers to estimate which patients with the disease are most likely to have severe illness. The work, published in the Egyptian Liver Journal, drew on seventy patients treated at two teaching hospitals and fifty healthy comparison subjects.
Amoebic liver abscess is caused by the single-celled parasite Entamoeba histolytica, which colonizes the colon and can reach the liver through the portal vein. The condition is described in the study as the second most common cause of parasite-related mortality worldwide and the most prevalent inflammatory space-occupying lesion of the liver. About ten percent of the global population carries the parasite in the colon, and roughly one in ten of those carriers may develop invasive amoebiasis. In Iraq, where the disease is prevalent, diagnostic difficulty is a recurring theme because the abscess can mimic acute hepatitis, perforated peptic ulcers, or acute cholecystitis.
The stakes of delayed recognition are high. The researchers note that severe disease, including rupture of the abscess, carries a death rate ranging from eighteen to forty-five percent, while mild disease has very low mortality. Early and accurate diagnosis matters because delays in identification and treatment can lead to complications. Known prognostic factors include advanced age, malnutrition, alcohol consumption, diabetes mellitus, and chronic liver disease, and inadequate sanitation, low socioeconomic level, rural living, and compromised host defenses have all been associated with liver abscess.
The study enrolled seventy patients diagnosed with amoebic liver abscess over twelve months, from December 2024 to December 2025, at Al-Ramadi Teaching Hospital and Baghdad Teaching Hospital. Patients were divided into a mild group of forty and a severe group of thirty. Diagnosis rested on clinical features such as fever, right upper quadrant pain, and hepatomegaly, together with abdominal ultrasonography, stool examination, and positive serology for Entamoeba histolytica infection. Because secondary bacterial infection can occur in these patients, the team also cultured aspirated material to check for it. Fifty apparently healthy individuals matched in age to the patients served as controls.
Severity was not left to clinical impression alone. Patients were classified as severe if they met one or more predefined criteria: an abscess at least five centimeters across, multiple abscesses, left lobe involvement, leukocytosis above fifteen thousand cells per cubic millimeter, serum albumin below three grams per deciliter, serum bilirubin above two milligrams per deciliter, pleural effusion, or abscess rupture. Those with none of these features were classed as mild. The researchers then measured a panel of serum markers in all participants: C-reactive protein, alanine aminotransferase, aspartate aminotransferase, alkaline phosphatase, albumin, total bilirubin, lactate dehydrogenase, and the cytokines interleukin-6 and interleukin-10.
The biomarker results showed a clear gradient across the groups. Severe patients had markedly higher levels of CRP, ALT, AST, ALP, total bilirubin, LDH, IL-6, and IL-10 than both mild patients and controls, with differences reaching statistical significance at p below 0.001. In the severe group, mean values were 52.7 for CRP, 92.8 for ALT, 85.3 for AST, 277.2 for ALP, 3.5 for total bilirubin, 421.6 for LDH, 96.3 pg/mL for IL-6, and 88.2 pg/mL for IL-10, compared with milder values in the mild group and the lowest values in controls. Serum albumin moved in the opposite direction, falling significantly in severe patients. Neutrophil-lymphocyte and platelet-lymphocyte ratios rose in the severe group, though they did not differ significantly between mild patients and controls.
To find which markers independently predicted severity, the team first ran univariate logistic regression. Increased CRP, ALT, AST, ALP, total bilirubin, and IL-6 were each significantly associated with severe disease, as was abscess size, while albumin showed a protective association, with lower levels seen in severe cases. NLR, PLR, and IL-10 did not reach statistical significance. A multivariable logistic regression model with bidirectional stepwise selection then identified AST, ALP, abscess size, total bilirubin, and IL-6 as the strongest positive predictors of severity, with CRP, ALT, and LDH contributing less, and albumin negatively associated with severe disease.
The model was converted into a nomogram, a graphical tool that assigns points to each predictor according to its regression coefficient so that a clinician can total them into an individualized probability of severe disease. In the nomogram, AST and ALP contributed the most points, albumin contributed negative points reflecting its protective role, and abscess size and total bilirubin followed. CRP, IL-6, ALT, and LDH added smaller positive contributions. Discrimination analysis, performed exclusively within the patient group, showed excellent separation between mild and severe disease: AST and ALP achieved an area under the curve of 1.000, followed by ALT at 0.998, IL-6 at 0.989, albumin at 0.980, LDH at 0.970, total bilirubin at 0.939, and CRP at 0.885. ROC analysis gave AST, ALP, and ALT sensitivities of 99.2, 99, and 98 percent with specificities of 97, 98.3, and 96.5 percent, while CRP and total bilirubin were less accurate.
Internal validation by bootstrap resampling supported the stability of the regression coefficients, with all biomarkers retaining significant associations with severe disease. Calibration analysis showed good agreement between predicted and observed probabilities, with a non-significant Hosmer-Lemeshow test and calibration points close to the ideal forty-five-degree reference line. Decision curve analysis found that across a range of threshold probabilities the model delivered a higher net benefit than treat-all or treat-none strategies, and clinical impact curves suggested that most patients the model flagged as high risk were true severe cases. The authors emphasize that these results are preliminary and require confirmation through external validation in an independent cohort because of the limited sample size.
The study has limitations the authors state plainly. It was a two-center study vulnerable to selection bias from patient geographic dispersion, center-specific diagnostic and treatment protocols, and variation in underlying illness composition. Data for some important factors had gaps and omissions. The nomogram was internally validated, but the small number of patients meant external validation could not be carried out, and the authors call for large-scale, prospective, multicenter research to confirm the model’s viability, robustness, generalizability, and clinical applicability before routine use. Even so, they argue the model, built largely from tests already available in most hospitals, could be adapted for tertiary hospitals, resource-limited endemic settings, and emergency departments, offering a practical way to identify patients who may need closer monitoring or more aggressive intervention early in the course of the disease.
Subject of Research: A multivariable prediction model using immunological and physiological biomarkers to assess disease severity in amoebic liver abscess
Article Title: Development and validation of a multivariable predictive model using immunological and physiological biomarkers for disease severity in amoebic liver abscess
Article References: Fahad, S. M., & Hanoush, N. H. (2026). Development and validation of a multivariable predictive model using immunological and physiological biomarkers for disease severity in amoebic liver abscess. Egyptian Liver Journal, 16(1), Article 77. https://doi.org/10.1186/s43066-026-00562-z
Image Credits: AI Generated
DOI: 10.1186/s43066-026-00562-z
Keywords: amoebic liver abscess, Entamoeba histolytica, prediction model, nomogram, interleukin-6, C-reactive protein, liver enzymes, ROC analysis, bootstrap validation, Iraq, disease severity, biomarkers
News Source: Ophelia Keating. (October 28, 2026). Biomarker model predicts severe amoebic liver abscess in Iraqi patients. Scienmag.
