Clinical and laboratory predictors of mpox severity and duration: an Italian multicentre cohort study (mpox-Icona)

Abstract

Background: Severe and prolonged mpox courses have been described during the 2022-2023 outbreak. Identifying predictors of severe evolution is crucial for improving management and therapeutic strategies. We explored the predictors of mpox severity and tested the association between mpox severity and viral load in biological fluids. We also analysed the predictors of disease duration and kinetics of inflammatory markers and described the viral presence and duration of shedding in biological fluids.

Methods: This multicentre historical cohort study included adults diagnosed with laboratory-confirmed mpox diagnosis between May 2022 and September 2023 at 15 Italian centres. Patients were followed up from the day of diagnosis until clinical recovery. Biological fluids (blood, urine, saliva, and oropharyngeal and rectal swabs) were collected from each subgroup during the course of the disease and after healing. The primary outcomes were disease severity (presence of mucosal involvement, extended rash, or need for hospitalisation) and its association with the cycle threshold value (Ct-value, surrogate of viral load) in biological fluids, using standard linear and linear mixed-effect logistic regression models. Among the secondary outcomes, predictors of disease duration were assessed using a linear regression model.

Findings: A total of 541 patients were enrolled, including four (0.74%) women, with a median age of 38 years (IQR 33-44). Among the 235 people living with HIV (PLWH) (43.44%), 22 (4.07%) had a CD4 count lower than 350 cells/μL. Severe mpox was reported in 215 patients (39.74%). No patient died. Multivariable analysis showed that, severe mpox was more likely among Caucasians (OR 1.82; 95% CI 1.14-2.90, p = 0.012) and patients who had an onset of fever (1.95; 1.27-2.99, p = 0.002), lymphadenopathy (2.30; 1.52-3.48, p < 0.001), sore throat (2.14; 1.27-3.59, p = 0.004), and peri-anal lesions (2.91; 1.93-4.37, p < 0.001). There was a significant difference (p = 0.003) between the median Ct-value in the upper respiratory tract for patients presenting with either mild (35.15; IQR 28.77-42.01) or severe infection (31.00; 25.00-42.01). The risk of developing severe disease decreased by approximately 5% per Ct increase (0.95; 0.91-0.98; p = 0.005). The disease lasted longer in the case of proctitis (+4.78 days; 1.95-7.61, p = 0.001), sore throat (+3.12; 0.05-6.20, p = 0.046), extended rash (+3.42; 0.55-6.28, p = 0.020), as well as in PLWH with a low CD4 count (+12.51; 6.79-18.22, p < 0.001).

Interpretation: The identification of predictors of severe or prolonged disease and the direct association MPXV Ct-value in the upper respiratory tract and disease severity could be useful in establishing proper management and early treatment of new mpox cases.

Funding: ICONA Foundation; Italian Ministry of Health "Ricerca Corrente Linea 2", INMI Lazzaro Spallanzani IRCCS.

Keywords: Ct-value; Evolution; MPOXV; Recovery; Severity; mpox.

Fig. 1

Flowchart for patient enrolment and inclusion in the different analyses.

Fig. 2

(A) Kruskal–Wallis test for comparison of median cycle threshold (CT) values for MPXV in the upper respiratory tract during the first week of infection. The analysis includes a subset of 233 patients (97 with severe disease and 136 with mild disease). The analysis shows that patients with severe diseases have lower Ct values (i.e. higher viral load) than those with mild infection (p = 0.003); (B) Mixed logistic regression model to assess the association of disease severity with MPXV Ct-value in the upper respiratory tract. The model includes 233 patients enrolled in eight different centres and shows that the risk of severe infection decreases by 5% per unit-Ct increase (p = 0.005). This provides evidence that lower Ct-values (i.e. higher viral load) is strongly associated with the severity of diseases.

Fig. 3

Decay model according to Kenward and Roger mixed effect model. Black solid lines represent estimates and relative 95% CI. Black dashed lines represent normal upper values. Model parameters are reported both as natural log (additive association) and exponential (multiplicative association). The constant value represents the estimate at the time of symptom onset, and the slope represents an estimate of the decay coefficient over time. A) Exponential decay model for C reactive protein (CRP) between day 0 and 30 since symptoms onset. The model shows that the CRP average is 7.93 mg/dL at the onset of symptoms, eventually dropping by a multiplicative coefficient of 0.84 per day (p < 0.001). B) Exponential decay model for neutrophil-to-lymphocyte ratio (NLR) between day 0 and 30 since symptoms onset. The model shows that the NLR average is 2.72 at symptom onset and eventually drops by a multiplicative coefficient of 0.95 per day since symptom onset (p < 0.001).