The pandemic of the coronavirus disease (COVID-19) started in late 2019. It quickly spread worldwide, notably in Europe [1
It is responsible for severe pneumonia, resulting in a high rate of transfers to intensive care units (ICU) and in-patient mortality of 5% to 32% [2
]. Severity has been related to an exaggerate immune response, the cytokine release syndrome (CRS), mediated by pro inflammatory cytokines, including interleukinIL-6, IL-12 and tumor necrosis factor α, leading to various organ dysfunction including tthe lungs, brain and heart [3
]. Previously, tocilizumab, an antibody targeting IL-6 receptors proved efficient against CRS [4
]. Opposing CRS may decrease further inflammatory pulmonary lesions, i.e., respiratory deterioration requiring mechanical ventilation, transfers to ICU and death [5
While waiting for the results of ongoing trials studying the effects of tocilizumab on COVID-19 pneumonia, starting on 23 March 2020, we administrated off-label tocilizumab to patients with severe COVID-19 pneumonia as a compassionate use.
In the present report, we assessed the effect of tocilizumab on mortality and mechanical ventilation in a cohort of patients hospitalized for severe COVID-19 pneumonia. To mitigate selection bias, we performed a triple analysis, including propensity-score matching, Cox multivariable and inverse probability score weighting analyses, to compare patients who received tocilizumab, to those who did not.
As the main finding of this single-center retrospective study, which focused on 246 patients hospitalized for severe COVID-19 pneumonia, we observed a protective association between treatment by tocilizumab and clinical outcomes, which included deaths and invasive mechanical ventilation, at 28-days of follow-up.
The study cohort was similar to that of previously described COVID-19 patients with a median age of 68 years; 27.6% presented cardiovascular history and 30.1% were obese [7
]. Median delay between first symptoms and treatment was 8 days, corresponding to the delay of CRS onset described in SARS-Cov-2 [8
Attenuating CRS may partly explain the significant decrease in the primary outcome [5
]. Several therapeutic interventions (corticosteroids, interleukin-1 blockade) have been used to mitigate inflammatory organ injury in viral pneumonia [9
]. The recent preliminary results from the RECOVERY trial provides evidence that treatment with dexamethasone reduces mortality in patients with COVID-19 under respiratory support. This study demonstrates the clinical relevance of the strategy based on inflammatory regulation in severe COVID-19 pneumonia [10
Indeed, CRS was related to interleukin accumulation, and a recent randomized trial studying dexamethasone yielded significant benefits [11
], comforting the role of immunomodulation therapeutic strategies [9
]. In our study, clinical improvement observed in patients treated by tocilizumab was akin to that described in two observational studies [12
], the most recent describing 544 patients with severe COVID-19 pneumonia criteria, the beneficial effect of tocilizumab (n = 179 in this group), with an adj.HR of 0.61 (95%CI = 0.40–0.92), regarding the same endpoint as in our study [13
]. In these two studies, similar to currently enrolling randomized controlled trials, tocilizumab dosage was higher: 8 mg/kg (up to 800 mg) in one to three injections. In comparison, in our study, dosage of tocilizumab was 400 mg, injected once. Further confirming our results with the same dosage would improve the availability of this costly biotherapy for which access may become an issue.
The preliminary results of the COVACTA trial, testing the effect of tocilizumab in randomized control trial in a heterogeneous cohort of patients, did not show significance of the primary composite endpoint nor secondary mortality endpoint. However, the proportion of patients under mechanical ventilation or in ICU who were excluded in our observational study, is unknown as of now. Moreover, previous observational study protocols mentioned tocilizumab dosage as high as 800 mg twice [12
], similar to that of ongoing trials with two injections of up to 800 mg each, within a 3-day period [14
]. In comparison, in our study, dosage of tocilizumab was 400 mg, injected once, following previous reports of improved outcomes of chimeric antigen receptor-T-cell-induced CRS with an 8 mg/kg dosage [4
]. Further confirming our results with the same dosage would improve the availability of this costly biotherapy for which access may become an issue [16
Indeed, in our hospital, reasons for injecting only one dose of 400 mg tocilizumab was mainly driven by cost issues and difficulties in procuring this treatment. It was only made available through the extensive work of our Pharmacology department, even more so that the time of administration corresponded to the first peak of the pandemic in Europe and France in particular. Previous pharmacokinetic and pharmacodynamic analyses of tocilizumab treatment yielded an equivalence between weekly subcutaneous injections of a lower dosage, as compared to intravenous injections once per month, to treat rheumatoid arthritis [17
]. Contrary to this chronic disease, COVID-19 may be assimilated to an acute infection and thus, may not require an anti-IL6 effect as prolonged as in rheumatoid arthritis, which may explain how clinical efficacy was obtained so quickly in the present study.
We acknowledge several limitations. First, the single-center nature of this study requires external validation; however, it guarantees homogeneity in the care of all patients, in our non-ICU departments dedicated to treat COVID-19 patients, i.e., observed differences are more likely to be due to tocilizumab. Second, although we aimed to mitigate selection bias using three statistical methods, including propensity-score matching, Cox multivariable and IPSW analyses, residual confounders are plausible [18
]. Due to comorbidities and the lack of beds in the context of the COVID-19 pandemic, a significant proportion of patients were not labeled as having a full-engagement status; hence options were limited regarding the possibilities of being transferred to a critical care medical department, as well as invasive mechanical ventilation. However, these criteria did not alter indications for tocilizumab, as nearly 33% were limited at admission. Furthermore, we acknowledge that limitation of care is more granular than a binary categorical variable such as “not-to-be-resuscitated”, and involves more grades. In this retrospective study, the status of patients was hard to represent accurately, the number one reason for this being that their status evolved through time. Indeed, patients who were at first not to be ventilated, after the first few days, may have changed to be in full engagement, due to signs of improvement. Similarly, patients who were not labeled as limited when they were admitted to medical wards may have been limited during night shifts by the intensivist on duty based on comorbidities or evolution since admission. Accurately representing these variations in a simple model was not feasible, hence, we opted for the most pragmatic approach we had at our disposal: assessing when patients were flagged as full-engagement, as opposed to others. All analyses were adjusted for this criterion in multivariable models. Furthermore, we performed additional sensitivity analyses to further mitigate selection bias; analyses which yielded similar results with significant association between tocilizumab and better survival without mechanical ventilation, even focusing on patients with full therapeutic engagement (n = 155).
Third, because arterial partial pressure of O2
was not available in all patients, we used a SpO2
ratio to assess respiratory dysfunction, a validated marker in acute lung injury [20
]. Fourth, use of non-invasive ventilation and high-flow oxygen support changed during the study period, following evolving guidelines which advocated against doing so during the first weeks of the pandemic to decrease virus aerosol propagation, and were then made more flexible. The relatively low proportion of patients who benefited from these treatments and the fact that there was no difference between the two groups regarding this criterion decreases the chances of it being a bias, although, residual confounding biases may remain. Fourth, there was a mild difference in age between tocilizumab and control groups. However, age was a variable that was accounted for in all multivariable analyses and also in computing the propensity-score. Thus, we did not observe any interaction between age on the efficacy of treatment by tocilizumab. Finally, we did not systematically assay IL-6, which may have proven valuable to identify patients for whom the effect was greater [6
These results point towards a clinical benefit of tocilizumab; however, they may not replace a fully-fledged randomized controlled trial, focusing on dosage adjustment and on patients managed early, so that CRS may be adequately attenuated in patients with COVID-19 evolving towards clinical deterioration.