Anakinra for severe forms of COVID-19 – Authors' reply

We thank Marc Laine and Laurent Bonello, Adil Rashid Khan and colleagues, and Naim Khan for their interest in our Article on treatment of patients with severe forms of COVID-19 with the interleukin (IL)-1 receptor antagonist anakinra.¹

Our colleagues pointed out (as did we in our Article) that the design of our study was not perfect from a statistical point of view because we compared the patients receiving anakinra with a historical control group. Only high-quality randomised trials can avoid confounding factors, but the urgent context of the COVID-19 pandemic means randomised trials are not always appropriate.

Laine and Bonello underline the significant effect of hypercoagulability in patients with severe COVID-19. At the time our study began, the issue of hypercoagulability was not fully understood. Of note, the proportion of patients who were given anticoagulants for a pre-existing disease before admission for COVID-19 was higher in the historical control group compared with the anakinra group (11 [25%] of 44 vs 5 [10%] of 52; p=0·044). The inclusion of this variable in our multivariate analysis did not modify the results. Once the patients were admitted to our hospital, they all received at least a prophylactic treatment with low molecular weight heparin. During the hospital stay, more patients developed a thromboembolic event in the anakinra group (ten [19%] of 52 patients) than in the control group (five [11%] of 44 patients). Therefore, a confounding effect favouring anakinra and related to this kind of complication appears unlikely.

With regards to the questions raised by Adil Rashid Khan and colleagues, it is absolutely true that some patients reached the primary endpoint at day 0, but no patient was on mechanical ventilation at inclusion. When excluding patients who experienced an event at day 0, the hazard ratio (HR) was 0·30 (95% CI 0·14–0·69). When excluding patients with an event on day 0 and day 1, the HR was 0·25 (0·09–0·70). As stated in our Article,¹ the patients in the historical control group had to fulfil the same inclusion and exclusion criteria as those in the anakinra group; however, we should have indicated that the inclusion criteria had to be fulfilled for the first time during hospitalisation, within the 24 h preceding anakinra initiation. Because our patients received standard oxygen therapy in medical wards, we were not able to assess a reliable percentage of inspired oxygen ratio at inclusion. As a result, a partial pressure of arterial oxygen to percentage of inspired oxygen ratio, which could classify the patients as having moderate or severe acute respiratory distress syndrome (ARDS) according to the Berlin definition, could not be calculated. We must underline that none of our patients had non-invasive ventilation or high flow oxygen therapy through nasal cannula at inclusion. Although the figure of 73% for the composite endpoint in the historical control group appears very high, it is in accordance with previously reported rates of mortality and admission to intensive care units for mechanical ventilation, in similar cohorts. In a follow-up study of 339 patients aged 71 years (SD 8) or older, more than 70% were severe or critical, with an incidence of ARDS of 21% and a mortality rate of 19%.² In the UK, a retrospective cohort study indicated an inpatient overall mortality of 38%, increasing to 60% in patients older than 80 years.³ Of note, the mortality rate was 44% in the historical control cohort reported by Cavalli and colleagues;⁴ whereas, in our study the mortality rate was 43%, at the same timepoint (21 days). Finally, the significant difference in the use of hydroxychloroquine between the two groups was taken into account in our study, but did not modify the results after a multivariable analysis. Moreover, the efficacy of hydroxychloroquine in COVID-19 remains a matter of debate, particularly in the most severe cases.

Naim Khan raises the difficult question of the optimal immunomodulation to be chosen to control the cytokine storm implicated in severe cases of COVID-19. Dexamethasone has proved effective in patients with ARDS admitted to an intensive care unit for mechanical ventilation, and it could represent a drug of choice in case of an insufficient response to anakinra.⁵ To overcome the SARS-CoV-2-induced autoinflammatory phenomenon, one could also suggest the combination of anakinra with other biologics targeting different cytokines (including IL-6, TNF) and potentially intracellular Janus kinases. The main unresolved issue about such a proposition is the adverse events, particularly in terms of infection.