Demographic, Clinical, and Laboratory Predictors of Liver Involvement in COVID-19 Patients with Gastrointestinal Symptoms

Study design and setting

We conducted a prospective, cross-sectional observational study at three tertiary referral centers in Al-Anbar Governorate, Iraq: Fallujah Teaching Hospital (450 beds), Ramadi Teaching Hospital (400 beds), and Al-Karma General Hospital (250 beds), collectively serving a catchment population of 1.8 million.

Study population and timeline

Between March 2021 and July 2025, we prospectively enrolled adult patients (≥18 years) with RT-PCR-confirmed SARS-CoV-2 infection (cycle threshold <35) presenting with gastrointestinal symptoms within 7 days of admission. Gastrointestinal symptoms included diarrhea, vomiting, nausea, abdominal pain, or anorexia. All patients underwent comprehensive hepatic panel assessment within 72 hours of admission. Medication histories, though collected, were incompletely documented in medical records.

Exclusion criteria

We excluded patients with pre-existing chronic liver disease, hepatic malignancy, recent hepatotoxic medication use (when documented), pregnancy, incomplete medical records (<80% data completion), or concurrent non-COVID-19 infections.

Diagnostic definitions

Hepatic involvement was defined biochemically as ALT or AST > 2 × upper limit of normal (ULN), with secondary criteria including ALP > 1.5 × ULN, GGT > 2 × ULN, or total bilirubin >1.5 mg/dL.¹⁵ This definition, while standard, cannot distinguish between direct viral hepatotoxicity, medication-induced injury, hypoxic hepatitis, or systemic inflammatory responses. Structural assessment via imaging (ultrasound, elastography) or histopathology was not performed due to resource limitations and safety considerations during pandemic conditions. Severity was stratified as mild (2–5 × ULN), moderate (5–10 × ULN), or severe (>10 × ULN).

Disease severity followed WHO criteria: mild (ambulatory), moderate (pneumonia without hypoxemia), severe (pneumonia with hypoxemia), or critical (ARDS, septic shock, or multi-organ failure).

Sample size calculation

Sample size was calculated using n = Z²p(1-p)/d², where Z = 1.96, p = 0.35 (anticipated prevalence), and d = 0.05, yielding a minimum of 349 participants. Accounting for 10% attrition, we enrolled 384 patients.

Data collection

Standardized case report forms captured demographic (age, sex, BMI), clinical (comorbidities, vaccination status, symptom chronology), and laboratory data. Medication data collection was attempted but proved incomplete due to documentation limitations. Laboratory investigations included complete blood count, comprehensive metabolic panel, hepatic function tests (ALT, AST, ALP, GGT, bilirubin, albumin, prothrombin time), inflammatory markers (CRP, ferritin, IL-6, procalcitonin), coagulation profile (D-dimer, fibrinogen, aPTT), and viral load via cycle threshold values.¹⁴

Statistical analysis

Analyses were performed using SPSS version 26.0 and R version 4.1.0. Continuous variables are presented as mean ± SD or median (IQR); categorical variables as frequencies and percentages. Univariate comparisons used chi-square/Fisher’s exact test for categorical variables and t-tests/Mann-Whitney U tests for continuous variables. Variables with P < 0.10 entered multivariate logistic regression with backward stepwise selection. However, strong correlations between disease severity and other predictors limit interpretation of independent effects. Multicollinearity was assessed using variance inflation factor <5; model fit via Hosmer-Lemeshow test and Nagelkerke R².

ROC curves determined optimal cut-offs using Youden’s index, with calculation of AUC, sensitivity, specificity, PPV, and NPV. A clinical scoring system was developed using β-coefficient-weighted regression coefficients. Internal validation employed bootstrap resampling (1,000 iterations); external validation remains pending. Statistical significance was set at P < 0.05.

Ethical considerations

The study received ethics approval from the Research Ethics Committee of the University of Anbar College of Medicine (Reference: UOA/MED/EC/2021-015) and adhered to the Declaration of Helsinki. Written informed consent was obtained from all participants or legal representatives, with confidentiality maintained through data anonymization.

Study flow and key findings

Of 384 patients screened, 347 were enrolled after excluding 37 (pre-existing liver disease, incomplete data, transfers, concurrent infections). Biochemically-defined hepatic involvement occurred in 168 patients (48.4%). The study encompassed March 2021–July 2025 across three tertiary centers in Al-Anbar, Iraq, covering Delta and Omicron variants. Hepatic involvement was associated with 2.84-fold increased mortality risk (P < 0.001), though causality cannot be inferred ( Figure 1).

Figure 1. Patient flow diagram and hepatic involvement distribution.

Flow diagram illustrating patient screening, enrollment, exclusion criteria application, and final allocation into hepatic involvement and no hepatic involvement groups. Numbers indicate patient counts at each stage.

Patient characteristics and hepatic involvement prevalence

Of 384 patients screened, 347 met inclusion criteria. Biochemical hepatic involvement occurred in 168 patients (48.4%). Mean age was 52.3±16.8 years with male predominance (57.1%, n=198). Baseline characteristics stratified by hepatic involvement are presented in Table 1.

Age demonstrated a dose-dependent association with hepatic involvement, with patients ≥61 years exhibiting 3.02-fold increased odds versus those aged 18–40 years. Male sex and obesity (BMI ≥30 kg/m²) showed associations of 1.78-fold and 2.00-fold increased odds, respectively.

Clinical manifestations and comorbidity profile

Pre-existing comorbidities showed significant associations with hepatic involvement risk. Among diabetic patients, 62.9% developed hepatic involvement versus 40.4% of non-diabetic patients. Clinical characteristics are detailed in Table 2.

Gastrointestinal symptom burden demonstrated a dose-response relationship, with patients presenting ≥5 symptoms exhibiting 2.06-fold increased odds of hepatic involvement versus 1–2 symptoms. Disease severity showed the strongest association, with severe or critical cases demonstrating 10.42-fold increased odds versus mild disease, suggesting hepatic abnormalities may primarily reflect systemic illness severity rather than independent organ-specific injury.

Laboratory biomarker performance

Laboratory investigations revealed distinct patterns differentiating patients with and without biochemically-defined hepatic involvement ( Table 3).

ALT demonstrated superior discriminative ability (AUC = 0.896) for identifying biochemical hepatic dysfunction. Combined assessment of ALT >80 U/L and ferritin >400 ng/mL achieved enhanced diagnostic accuracy (sensitivity 85.7%, specificity 88.3%).

Independent predictors of hepatic involvement

Multivariate logistic regression identified eight variables associated with hepatic involvement after statistical adjustment ( Table 4). However, the predominant association with disease severity (aOR = 10.42) suggests these may represent markers of systemic illness rather than independent hepatic risk factors.

Model performance: Hosmer-Lemeshow P = 0.412; Nagelkerke R² = 0.624; AUC = 0.854 (95% CI: 0.814–0.894). Internal bootstrap validation (1,000 iterations) showed minimal optimism (0.018). External validation required before clinical application. The strong association with disease severity suggests confounding by systemic illness.

Clinical outcomes and prognostic impact

Biochemically-defined hepatic involvement correlated significantly with adverse clinical outcomes and healthcare resource utilization ( Table 5).

Patients with hepatic involvement exhibited prolonged hospitalizations (14.2 vs. 9.6 days), doubled ICU admission rates (48.2% vs. 22.3%), and three-fold increased mortality (19.0% vs. 6.7%; all P< 0.001). These associations likely reflect overall disease severity rather than isolated hepatic dysfunction.

Clinical risk scoring system

Based on β-coefficients from the multivariate model, we developed a preliminary weighted scoring system pending external validation ( Table 6).

Risk Stratification: Low risk (0–2 points, 15% probability), intermediate risk (2.5–3.5 points, 35%), high risk (4–5.5 points, 65%), very high risk (≥6 points, 85%). Internal validation performance: sensitivity 82.1%, specificity 78.8%, PPV 77.5%, NPV 83.2%. External validation in independent cohorts required before clinical implementation.

Temporal patterns and variant analysis

Hepatic involvement rates showed non-significant differences between Delta-predominant (52.3%, March 2021–December 2021) and Omicron-predominant periods (44.8%, January 2022–July 2025; P = 0.168). This analysis did not adjust for vaccination status changes, treatment protocol evolution, or other temporal confounders, limiting interpretability ( Figure 2 and Table 7).

Figure 2. Receiver operating characteristic curves for laboratory biomarkers.

Receiver operating characteristic (ROC) curves demonstrating predictive performance of individual biomarkers (ALT, AST, ferritin, D-dimer, GGT) and combined model for hepatic involvement in COVID-19 patients with gastrointestinal symptoms. Combined model (AUC = 0.924) integrates ALT, ferritin, D-dimer, and clinical predictors. All biomarkers demonstrated significant discriminative performance (P < 0.001). AUC, area under the curve.

Age-related vulnerability and hepatic reserve

Advanced age showed strong association with hepatic involvement (aOR = 2.68 for ≥61 years), consistent with age-related COVID-19 vulnerability.¹⁷ This may reflect immunosenescence, diminished hepatic regenerative capacity, increased comorbidity burden, or simply correlation with disease severity.¹⁸ Our findings extend previous observations¹⁹ while acknowledging inability to establish independent causation.

Sex-based disparities in hepatic injury

Male sex showed 82% increased odds of hepatic involvement (aOR = 1.82), consistent with sex-based outcome disparities in COVID-19.²⁰ Proposed mechanisms include androgen-mediated ACE2 upregulation, sex-specific immune responses, and baseline hepatic vulnerability differences.²¹ However, these associations may reflect general COVID-19 severity patterns rather than organ-specific susceptibility.

Metabolic comorbidities and hepatic injury risk

Diabetes mellitus emerged as the strongest comorbidity predictor (aOR = 2.94), aligning with meta-analyses demonstrating increased COVID-19 severity in diabetic populations.²² While mechanisms may include pre-existing non-alcoholic fatty liver disease, amplified cytokine responses, and hyperglycemia-mediated dysfunction,²³ our study cannot distinguish these from general disease severity effects. Hypertension’s association (aOR = 2.12) similarly may reflect systemic disease burden rather than specific hepatic vulnerability²⁴

Gastrointestinal-hepatic axis in COVID-19

Diarrhea demonstrated association with hepatic involvement (aOR = 1.96), potentially supporting gut-liver axis disruption.²⁵ However, both manifestations may simply reflect greater viral burden or systemic inflammatory response rather than organ-specific pathophysiology.²⁶ Without structural assessment or mechanistic studies, causal relationships remain speculative.

Laboratory biomarkers for risk stratification

ALT >80 U/L demonstrated strong discriminative ability (AUC = 0.896) for detecting biochemical hepatic abnormalities.²⁷ However, aminotransferase elevation’s non-specificity—potentially reflecting viral hepatitis, drug toxicity, hypoxic injury, or systemic inflammation—limits clinical interpretation.

Ferritin elevation (aOR = 3.86 for >400 ng/mL) and D-dimer elevation (aOR = 3.21 for >1000 ng/mL) likely reflect systemic inflammatory and thrombotic processes rather than specific hepatic injury.^28,29 Their associations with hepatic enzyme elevation may indicate shared pathways or confounding by disease severity.

Disease severity and multi-organ involvement

The overwhelming association between hepatic involvement and COVID-19 severity (aOR = 10.42 for severe/critical disease) suggests biochemical abnormalities primarily reflect systemic illness rather than independent hepatic injury.³⁰ Multiple mechanisms may contribute: hypoxic hepatitis from respiratory failure, cytokine-mediated injury, microvascular thrombosis, or medication effects.³¹ This strong association raises questions about whether other identified “predictors” represent independent risk factors or merely additional severity markers.

Clinical utility of the predictive model

The proposed scoring system demonstrates reasonable internal performance (sensitivity 82.1%, specificity 78.8%) but requires external validation before clinical implementation.³² The model’s heavy weighting toward disease severity markers limits its utility for identifying patients with disproportionate hepatic risk. Furthermore, without distinguishing injury mechanisms, targeted interventions remain challenging.

Regional context and healthcare implications

Our findings align with regional studies documenting hepatic manifestations in COVID-19.^9,10,13 The non-significant difference between variant periods (P = 0.168), while suggesting consistent patterns, lacks adjustment for critical confounders including vaccination uptake, treatment protocol changes, and healthcare system adaptations.^11,12

Clinical outcomes and prognostic significance

The three-fold mortality increase associated with biochemical hepatic involvement (19.0% vs. 6.7%, P < 0.001) likely reflects its role as a severity marker rather than independent prognostic factor.¹⁴ Without adjusting for disease severity in outcome analyses, the specific contribution of hepatic dysfunction remains unclear. Resource utilization increases similarly may reflect overall illness burden rather than hepatic-specific complications.

Mechanistic insights and future directions

While our findings suggest multifactorial pathogenesis, the inability to distinguish between direct viral injury, medication toxicity, hypoxic hepatitis, and systemic inflammatory effects represents a fundamental limitation. Future investigations incorporating imaging, histopathology, and comprehensive medication data are essential for mechanistic understanding and targeted intervention development.

Strengths and limitations

Strengths include systematic examination of an important patient subset, comprehensive biochemical assessment, prospective design, and transparent reporting of associations. The multicenter design enhances regional generalizability.

Critical limitations include: (1) exclusive reliance on biochemical definitions without structural assessment via imaging or histopathology; (2) inability to distinguish injury mechanisms (viral, drug-induced, hypoxic, inflammatory); (3) incomplete medication data precluding assessment of drug-induced hepatotoxicity; (4) lack of vaccination status analysis despite its collection; (5) strong confounding by disease severity limiting interpretation of independent predictors; (6) single-region setting limiting generalizability; (7) absence of external validation for the proposed scoring system; and (8) cross-sectional design precluding temporal assessment of biomarker evolution.

Clinical implications and recommendations

Until external validation and mechanistic clarification are achieved, clinical application should remain cautious. Biochemical hepatic monitoring remains appropriate for COVID-19 patients with gastrointestinal symptoms, though interpretation should consider multiple potential etiologies including medication effects and systemic illness severity. The proposed scoring system requires validation in independent cohorts before implementation.

Future research priorities

Essential priorities include: (1) external validation in diverse populations with different healthcare systems; (2) incorporation of imaging (ultrasound, elastography) to characterize structural changes; (3) systematic documentation of medication exposures to assess drug-induced hepatotoxicity; (4) analysis of vaccination impact on hepatic involvement risk; (5) adjustment for disease severity in multivariable models to identify truly independent predictors; (6) longitudinal studies examining temporal evolution and recovery patterns; and (7) mechanistic studies distinguishing between competing injury pathways.