Abstract
BACKGROUND: There is a risk of novel mutations of SARS-CoV-2 that may render COVID-19 resistant to most of the therapies, including antiviral drugs and vaccines. The evidence around the application of therapeutic plasma exchange (TPE) for the management of critically ill patients with COVID-19 is still provisional, and further investigations are needed to confirm its eventual beneficial effects. AIMS: To assess the effect of TPE on the risk of mortality in patients with COVID-19-associated pneumonia, using three statistical procedures to rule out any threats to validity. METHODS: We therefore carried out a single-centered retrospective observational non-placebo-controlled trial enrolling 73 inpatients from Baqiyatallah Hospital in Tehran (Iran) with the diagnosis of COVID-19-associated pneumonia confirmed by real-time polymerase chain reaction (RT-qPCR) on nasopharyngeal swabs and high-resolution computerized tomography chest scan. These patients were broken down into two groups: Group 1 (30 patients) receiving standard care (corticosteroids, ceftriaxone, azithromycin, pantoprazole, hydroxychloroquine, lopinavir/ritonavir), and Group 2 (43 patients) receiving the above regimen plus TPE (replacing 2 l of patients' plasma by a solution, 50% of normal plasma, and 50% of albumin at 5%) administered according to various time schedules. The follow-up time was 30 days and all-cause mortality was the endpoint. RESULTS: Deaths were 6 (14%) in Group 2 and 14 (47%) in Group 1. However, different harmful risk factors prevailed among patients not receiving TPE rather than being equally split between the intervention and control group. We used an algorithm of structural equation modeling (of STATA) to summarize a large pool of potential confounders into a single score (called with the descriptive name "severity"). Disease severity was lower (Wilkinson rank-sum test p < 0.001) among patients with COVID-19 undergoing TPE (median: -2.82; range: -5.18; 7.96) as compared to those not receiving TPE (median: -1.35; range: -3.89; 8.84), confirming that treatment assignment involved a selection bias of patients according to the severity of COVID-19 at hospital admission. The adjustment for confounding was carried out using severity as the covariate in Cox regression models. The univariate hazard ratio (HR) of 0.68 (95%CI: 0.26; 1.80; p = 0.441) for TPE turned to 1.19 (95%CI: 0.43; 3.29; p = 0.741) after adjusting for severity. CONCLUSIONS: In this study sample, the lower mortality observed among patients receiving TPE was due to a lower severity of COVID-19 rather than the TPE effects.