At the end of the search, 2775 articles were retrieved from the reviewed databases, of which 1050 were duplicates, and were excluded from the study. After screening by title, 1442 articles were removed out of 1725 ones, and 283 studies were evaluated based on their abstracts. Then, 182 articles were entered into the screening stage based on their full texts, of which 110 ones were excluded from the study due to different outcomes, 35 due to different communities, 5 due to non-English language, 10 due to unavailability, and 5 due to different methods. Finally, 17 clinical trial studies were included (Fig. 1).

The search outputs and study selection

Among all selected articles, 15 studies were randomized clinical trials, and 2 ones were non-randomized clinical trials. Also, 6 studies were published in 2020, and the rest in 2021. In the final 11 selected studies, the comparison group received a placebo while in the rest of the studies, they received another drug, such as dexamethasone, favipiravir, or sarilumab. The mean age and BMI in all studies was 61.4 years, and 27.7 kg/m2, respectively (Table 1). In seven studies, the tocilizumab consumption increased the risk of mortality, and the control group of two studies received only the usual treatment of COVID-19 without additional drugs, or placebo [22, 23]. The control group of three studies received placebo in addition to the usual treatment of COVID-19 [24,25,26]. Another study compared the effects of dexamethasone, and tocilizumab [27] while dexamethasone was prescribed at a dose of 4 mg/kg/day, and was more effective than tocilizumab in preventing death due to the COVID-19 disease. Finally, the last study examined the effect of the usual dose of tocilizumab on the first group receiving doses of 200, and 120 mg, and the second group receiving doses of 80, and 40 mg. It is likely that the higher mortality in the first group than the second one is due to sicker patients, and the prescription of higher doses of the drug for them [28]. Among the series of articles related to reduction in mortality, the study by Capra et al. showed the most significant rate [29] (Table 1). Among the ten articles in which the use of tocilizumab reduced mortality, in one study, the control group received favipiravir [30], in another study, the two separate control groups received sarilumab, and placebo [31], in one study, they received methylprednisolone [32], and in five studies, the control groups received the usual treatment of COVID-19 [33,34,35,36]. In two studies, grouping was performed based on the response [37], and disease severity [38] (Table 1).

After combining the results of these studies, the cumulative risk of death in COVID-19 patients taking tocilizumab was 0.93 compared to patients who had not taken the drug (RR: 0.93; 95% CI: 0.86, 1.00; I2: 72.39%) (Fig. 2). The results of publication bias using the Egger test and funnel plot showed publication bias did not occur in this analysis (B: − 1.08; standard Error: 0.608; P-value: 0.075). The funnel plot is shown in Fig. 3. The percentage of heterogeneity in this analysis was 72.39% which was less than 75%, and indicates the presence of heterogeneity between studies, but its rate is acceptable (Fig. 3). In addition to these findings, meta-regression results are shown in Fig. 3. This analysis was performed to evaluate the effect of patients’ body mass index, and age on the association between the use of tocilizumab, and the rate of death due to COVID-19. The results showed with increasing age, the effect of this drug on reducing death due to COVID-19 in patients decreased (B: − 0.028; Standard Error: 0.046; P-value: 0.557; 95% CI: − 0.128, 0.079), and with increasing the body mass index, the effect of tocilizumab on reducing death caused by COVID-19 in patients increased (B: 0.115; Standard Error: 0.082; P-value: 0.258; 95% CI: − 0.147, 0.377), but the association between both variables (the age, and body mass index) was not statistically significant in meta-regression analysis (Fig. 3).

The effect of tocilizumab (Actemra) on the occurrence of death in patients with COVID-19

The funnel plot (for assessing heterogeneity) and L’Abbe plot (for assessing heterogeneity)

In addition to death, the outcomes of the need for a ventilator, and hospitalization in ICU were also examined, the results of which are shown in Fig. 4. The results showed tocilizumab consumption increased hospitalization in ICU by 4% which was not statistically significant (RR: 1.04; 95% CI: 0.90, 1.20; I square: 0.00%) (Fig. 4). The results of the present meta-analysis also showed the use of this drug reduced the need for a ventilator by 2%, but this association was also not statistically significant (RR: 0.98; 95% CI: 0.90, 1.08; I square: 26.87%) (Fig. 4). The results of publication bias analysis using the Egger test, and funnel plot showed publication bias did not occur in any of these associations (B: -0.36; Standard Error: 0.400; P-value: 0.376) (B: 0.13; Standard Error: 0.657; P-value: 0.840). The funnel plot is shown in Fig. 4. In addition, meta-regression was performed to evaluate the effect of COVID-19 patients’ body mass index, and age on the association between the use of tocilizumab, hospitalization in ICU, and the need for ventilator. The results showed with increasing age, and the body mass index, increased the effect of tocilizumab on the need for a ventilator in patients with COVID-19 ((B: 0.001; standard Error: 0.027; P-value: 0.950; 95% CI: − 0.059, 0.067), and (B: 0.006; standard Error: 0.116; P-value: 0.960; 95% CI: − 0.36, 0.37)). Also, these variables lead to decreased the effect of tocilizumab on hospitalization of patients with COVID-19 in ICU ((B: − 0.028; standard Error: 0.025; P-value: 0.425; 95% CI: -0.105, 0.057), and (B: − 0.040; standard Error: 0.037; P-value: 0.447; 95% CI: -0.52, 0.43)).

The effect of tocilizumab (Actemra) on the occurrence of ICU admission and need to ventilator in patients with COVID-19 (forest and funnel plot)

In this meta-analysis, the association between underlying diseases, severe forms of the COVID-19 disease, and the use of tocilizumab was investigated, and the results have been shown in Table 2. The results showed hypertension, cardiovascular diseases, asthma, malignancy, chronic pulmonary disorders, chronic liver disorders, and obesity could make patients with COVID-19 more prone to its severe forms, and taking tocilizumab. Among the mentioned diseases, the chronic liver disorder could have a greater effect, but this association was not statistically significant (RR: 1.43; 95% CI: 0.42, 4.86; I2: 0.00%) (Table 2). Only hypertension was significantly associated with the severe form of the COVID-19 disease in patients, and tocilizumab consumption (RR: 1.03; 95% CI: 1.00, 1.12; I2:: 28.08%) (Table 2). In addition, myocardial infraction, diabetes mellitus, COPD, and chronic kidney disorders reduced the risk of developing severe COVID-19 forms, and the use of tocilizumab according to the combination of preliminary study results (Table 2).

The qualitative assessment of articles based on the Cochrane checklist revealed that the initially selected studies exhibited a low risk of bias. Notably, the least bias was observed in the areas of allocation concealment and random sequence generation, indicating minimal risk of selection bias (see Fig. 5). Conversely, the most prevalent bias was associated with the blinding of outcome assessment, indicative of detection bias, with several initial studies manifesting challenges in this domain (see Fig. 5).

Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies

Within the selected studies, the study conducted by Stone et al. exhibited the least bias, whereas the study by Pomponio et al. demonstrated the highest level of bias (see Fig. 6). These findings underscore the importance of considering specific domains of bias in the individual studies, allowing for a nuanced understanding of the methodological rigor and limitations across the body of evidence.

Risk of bias summary: review authors' judgements about each risk of bias item for each included study