A total of 279 unique publications was yielded in this study. According to flow diagram, 251 were excluded after primary and subsequent full-text review (Figure 1). Thus, 28 articles included 991 COVID-19 patients underwent tocilizumab administration were included in this study10,11,19-44. Of the 28 included studies, 7 of which included 377 patients were case-control studies with a control group not using tocilizumab28,29,32,36,39,40,42.
Baseline characteristics of all 28 included studies were listed in table 1. Of the total 991 patients, 769 were male (77.6%) and pooled median age was 62-year-old (range 55-73). The most common comorbidities were hypertension, cardiovascular disease and diabetes mellitus (DM). 324 patients undergone mechanical ventilation was reported in 19 studies. COVID-19 patients usually experience hyper-inflammation evidenced by exaggerated IL-6 and C-reactive protein (CRP). After administration of tocilizumab, most available data showed a dramatic drop of CRP value and serum IL-6 level tended to spike shortly in first and then decrease.
Risk of bias of all studies was assessed by the ROBINS-I tool. Discriminations were primarily found in domain of “bias due to confounding” and “bias due to deviations from intended interventions”. Consequently, the overall judgment of risk of bias for all included studies was moderate (Figure S1).
Impact of tocilizumab treatment on ORR and mortality of COVID-19 patients
Of the 991 COVID-19 patients in all 28 studies, on top of antiviral treatment, additional use of tocilizumab was associated with a pooled ORR of 72% (95% CI, 66-79%; I2 = 72%, P< 0.0001) (Figure 2A). Data were unavailable for mortality in 3 studies, and a pooled all-cause death rate of 16% (95% CI, 11-22%; I2 = 67%, P< 0.0001) was reported in 870 patients when treated with tocilizumab. 17 studies10,11,19,21-23,25-29,31,33,35-37,40 comprising 562 patients were defined as severe COVID-19 infection, and the pooled ORR of treatment with tocilizumab was 78% (95% CI, 70-85%; I2 = 65%, P< 0.0001), while pooled ORR in 64 non-severe infection was 93% (95% CI, 85-98%; I2 = 0%, P=0.66) (Figure S2).
Of the 7 case-control studies, 377 patients had been assigned to tocilizumab therapy and 380 to standard-care treatment, the pooled results demonstrated a significantly lower risk of all-cause death at follow-up in patients with tocilizumab (RR: 0.61, 95% CI 0.42–0.90; P=0.012; Figure 3A). Moreover, Figure 3B illustrated that 231 patients received tocilizumab in experimental group also had a higher ORR than 269 patients in the control group, although the RR was statistically insignificance [1.10 (95% CI 0.96-1.27; P=0.175)].
A subgroup analysis of 56 COVID-19 patients with solid organ transplantation indicated that the pooled ORR was 53% (95% CI, 26-78%; I2 = 67%, P=0.030) and the pooled all-death rate was 28% (95% CI, 12-68%; I2 = 73%, P=0.010) when compared with 935 non-transplant patients infected with COVID-19 (pooled ORR of 75%, 95% CI 69-81; I2 = 68%, P< 0.0001 and pooled all-death rate of 15%, 95% CI 10-21%; I2 = 69%, P< 0.0001). Overall, these results presented a poorer clinical outcome in solid organ transplant recipients than non-transplant patients when administrated tocilizumab (Figure S3 and S4).
Impact of tocilizumab treatment on duration of hospitalization, ICU stay and mechanical ventilation
At the point of study completion, median LOH and LOI was 14 days (range 11-34) and 8 days (range 5-11) in patients receiving tocilizumab, which was specified in 15 studies10,20-23,28,29,31,32,35,36,38-41 and 8 studies10,23,27,29,32,33,36,42 , separately. Moreover, the median duration of mechanical ventilation was 7 days (range 5-10) after treatment of tocilizumab which was documented in 9 articles10,11,21-23,27,28,40,41 (Table 2).
Optimal timing and safety in use of tocilizumab
The timing of tocilizumab administration was clarified in 8 studies.10,22-25,27,36,40 With this data, the second-order polynomial equation was given as following: y = p1*x2 + p2*x + p3. (p1 = 0.004099 (-0.006127, 0.01433), p2 = -0.05933 (-0.257, 0.1384), p3 = 0.3474 (-0.5564, 1.251). The significance of the model was inferred by P value (P<0.001), and the goodness of fit was accessed by R2 which was 0.667 (Figure S5). Thus, we deduced, by this polynomial equation, that the optimal timing of tocilizumab administration was the 7.15 day from the symptom onset and with a lowest all-cause death rate of 13.11% (Figure S6).
Nearly all included studies confirmed the safety of tocilizumab for treatment of COVID-19 patients no matter in the context of ICU or general ward. Although the occurrence was very low, the most commonly reported adverse events were the elevation of hepatic enzymes of at least 3-times above the normal level, pancytopenia and cutaneous rash. (Table 3) Occasionally, bacteremia caused by neutropenia was reported by individual medical centers.
Publication bias and sensitivity analysis
The funnel plots were used to assess the evidence of bias towards studies. According to the Figure 4, there was no conclusive evidence of publication bias of ORR and mortality in any studies (P=0.611 and P=0.875). In the sensitivity analysis, the overall estimates remained consistent across the analysis of ORR and all-cause death rate in all included studies, which suggests our results were credible. (Figure S7)