The study selection and exclusion process has been documented using the PRISMA flow diagram (Figure 1). Out of total 362 records screened, 13 observational studies [7-11, 20-27] were included in qualitative (systematic) review as well as in quantitative analysis. Non-comparative intervention studies (n=10) were not included in meta-analysis.
Study characteristics and outcomes data reported in various observational studies included in this systematic review are included in Table 1. Studies done by Garcia et al, Kewan et al, Klopfenstein et al, Quartuccio et al, Ramaswamy et al, Sanz et al and Wadud et al had significantly more severe disease patients (higher level of inflammation, systemic disease and organ damage) in tocilizumab group that can affect the primary outcomes like mortality, whereas all other studies[7, 8, 20, 21, 25, 26] had similar disease severity in both treatment groups. Four studies [8, 20-22] were done in mild to moderate disease whereas nine studies[7, 9-11, 23-27] were done in moderate to severe disease group of patients (Table 1).
Risk of bias (ROB) within the studies
The overall risk of bias was assessed as moderate except for three studies, Campochiaro et al has been assessed as low risk of bias and Klopfenstein et al and Quartuccio et al evaluated as having serious risk of bias. Overall ROB is similar to bias due to confounding in individual studies. Andrews et al, Somers et al and Quartuccio et al had moderate bias in selection of participants. Andrews et al had moderate bias with missing data. (Figure 2), (Figure 1S – Weighted Summary plot). Overall ROB for observational studies was assessed as moderate, as majority studies were assessed as having moderate ROB. In addition, adjusted mortality rates were presented, which were less likely to be affected by confounding factors as these factors were adjusted during analysis. Hence ROB assessed for GRADE was regarded as having no serious issues.
For unadjusted mortality rate, pooled analysis revealed no difference with tocilizumab treatment in comparison to control group [OR = 0.90 (95% CI = 0.74 - 1.11), p=0.33; I2=81%] (Figure 3a). This data was extracted from twelve observational studies (OS) with a total of 819 and 1931 patients in tocilizumab and control group, respectively. Sensitivity analysis with exclusion of study done by Sanz et al and Capra et al resulted in I2 = 14%, with significant decrease in overall mortality odds with tocilizumab [OR = 0.65 (95% CI = 0.50 – 0.84), p=0.001; I2=14%]. Subgroup analysis revealed that tocilizumab significantly decrease mortality in mild to moderate patients [OR = 0.40 (95% CI = 0.23 – 0.69), p=0.001; I2=76] but no difference in severe patients [OR = 1.05 (95% CI = 0.84 – 1.31), p=0.68; I2=81%].
For adjusted mortality, pooled analysis showed decrease in adjusted hazard ratio (HR) with tocilizumab versus control group [HR = 0.54 (95% CI = 0.44-0.68), p<0.00001; I2=74%; 6 OS] (Figure 3b). Sensitivity analysis was not performed as all studies showed decrease mortality with tocilizumab and hence heterogeneity was ignored.
Progression of Disease – ICU admission or Need for ventilation (Intubation)
Pooled analysis revealed statistically significant decrease in number of individuals with progression of disease with tocilizumab treatment in comparison to control group [OR = 0.34 (95% CI = 0.21 – 0.55), p<0.0001; I2=76%] (Figure 4). This data was extracted from five OS with a total of 212 patients in tocilizumab group and 266 in control group. Sensitivity analysis was not performed as heterogeneity is because of difference in characteristics of patients included in studies of review. Therefore, GRADE quality of evidence is downgraded for increased heterogeneity.
Pooled analysis revealed that there was no difference in odds of hospital discharge in tocilizumab group as compared to standard of care [OR = 1.35 (95% CI = 0.84 – 2.18), p=0.22; I2=22%] (Figure 2S). Data extracted from 4 OS with 157 and 175 patients in two groups respectively.
No statistically significant difference in the rate of clinical improvement was observed between two treatment groups [OR = 1.02 (95% CI = 0.61 – 1.69), p=0.95; I2=84%] (Figure 3S). Results derived from 4 OS with 164 and 148 patients in tocilizumab and control group, respectively.
There was a significant increase in odds of secondary bacteraemia in individuals treated with tocilizumab versus control [OR = 2.75 (95% CI = 1.86 – 4.06), p< 0.00001; I2=84%] (Figure 5a). This data was extracted from 3 OS with a total of 254 and 558 patients in tocilizumab and control groups, respectively.
Secondary pneumonia or superinfections
There was a significant increase in odds of secondary pneumonia or superinfections in individuals treated with tocilizumab versus control [OR = 2.93 (95% CI = 1.92 – 4.46), p< 0.00001; I2=77%] (Figure 5b). This data was extracted from 4 OS with a total of 282 and 581 patients in tocilizumab and control groups, respectively.
Publication bias was low, the funnel plot for twelve studies appears to be asymmetrical around the intervention effect estimate for mortality rate (Figure 4S). Regression test for funnel plot asymmetry showed the value of t = -1.3752 and p-value of 0.1991, indicating low publication bias. Egger’s regression test for progression of disease (t = 0.1022, p-value = 0.9251), hospital discharge (t = -0.8767, p-value =0.4731) and clinical improvement (t = -2.4392, p-value =0.1349) showed low publication bias. Hence, overall publication bias was concluded as low for systematic review.
GRADE analysis of the primary and secondary outcomes (Table 2)
The GRADE recommendation for the primary outcome adjusted HR for mortality was ‘HIGH’ evidence quality as there were no serious issues in the ROB analysis, inconsistency, indirectness with lack of imprecision. ROB is regarded as low as all the confounders were adjusted during analysis. As all the studies showed mortality benefit with tocilizumab with adjusted HR, heterogeneity was ignored. However, for non-adjusted mortality rate, recommendation is low, as there is high heterogeneity along with high ROB. In addition, there was imprecision of results along with plausible confounding suggesting spurious effect. The GRADE recommendation for progression of disease was graded as ‘Moderate’ evidence quality because of moderate ROB and high heterogeneity. The recommendation for hospital discharge and clinical improvement were graded as “Low” and “Very Low” respectively. The GRADE recommendation for secondary bacteraemia and pneumonia or superinfections were “High” and “Moderate” evidence quality respectively, as there was presence of serious issues with ROB and high heterogeneity. The GRADE recommendation is shown in table 2.