Clinical characteristics and prognostic factors in patients with stool cytomegalovirus positivity

The clinical association between cytomegalovirus (CMV) DNA detection in stool samples and patient outcomes remains underexplored. This study aimed to assess prognostic factors and viral kinetics in patients with positive stool CMV polymerase chain reaction (PCR).

This retrospective cohort study included adult patients with positive stool CMV-PCR results at Taipei Veterans General Hospital (2016-2021). Clinical data, plasma, and stool viral loads (VLs) were analyzed. Receiver operating characteristic (ROC) curves and area under the curve (AUC) evaluated 30-day mortality prediction, with optimal cutoffs maximizing sensitivity and specificity. Kaplan-Meier survival analyses and Cox proportional hazards models identified predictors of 30-day mortality.

A total of 114 patients (mean age: 64.0 years, 64% male) were included. The median stool CMV VL was 629 copies/mL (interquartile range [IQR]: 263-7949). Plasma CMV DNA was detected in 76% with a median VL of 341 copies/mL (IQR: 10-1771). Stool and plasma VLs showed moderate correlation ( ρ = 0.38, p < 0.0001). ROC analysis identified cutoffs for predicting 30-day mortality: stool 9654 copies/mL (AUC = 0.54; sensitivity 42%; specificity 81%) and plasma 1738 copies/mL (AUC = 0.60; sensitivity 47%; specificity 70%). In multivariate Cox analysis, stool CMV VL >9,654 copies/mL (adjusted hazard ratio: [HR] 2.69, 95% confidence interval [CI]: 1.06-6.84; p = 0.04) and plasma CMV VL >1738 copies/mL (adjusted HR: 2.66, 95% CI: 1.14-6.17; p = 0.02) were independent predictors of 30-day mortality. Septic shock and steroid use were also associated with increased mortality, whereas antiviral therapy ≥7 days was independently protective (adjusted HR: 0.26, 95% CI: 0.10-0.64; p = 0.003).

Stool and plasma CMV VLs, antiviral treatment duration, and host factors such as immune status may influence outcomes in patients with intestinal CMV reactivation. Larger studies are needed to validate optimal VL thresholds for risk stratification.