Letter to the editor: “Clinical significance of micronutrient supplements in patients with coronavirus 2019 disease: A comprehensive systematic review and meta-analysis"

Dear Editor,

We would like to share ideas on publication “Clinical significance of micronutrient supplements in patients with coronavirus 2019 disease: A comprehensive systematic review and meta-analysis [1]”. Beran et al. identified 26 studies involving 5633 COVID-19 patients, comparing outcomes from the use of three individual micronutrient supplements with standard-of-care (SOC) practices. Although the authors discussed about zinc deficiency and the risk of complications and mortality, they concluded that supplementation with this micronutrient was not statistically associated with the primary outcome of reduced mortality from COVID-19.

Although important results have been presented, Beran et al. interpreted their data relying on significance tests to make inferences. Thus, the clinical impact of variables under investigation, that is, the genuine clinical significance of the main findings was lost during the discussion. For example, the RR for the mortality outcome associated with zinc supplementation groups was 0.79 (95%CI 0.60 to 1.03) and 0.75 (0.49–1.13) compared to the SOC group. Regarding confidence intervals (CIs), treatment results suggest compatibility for a wide range of effects, from a 51% reduction to a 13% increase in mortality rates. In this scenario, what are the possible implications for risk patients affected by SARS-CoV-2 (1st) It is possible that the intervention is, in fact, beneficial; however, based on results of the significance test, supplementation would not be recommended (2nd) It may be that intravenous zinc supplementation (up to 13% mortality range) is not a recommended treatment for all patients; therefore, it is not prudent to state that “there are no associated effects”.

There is an old recommendation that statistical significance tests should not constitute the platform for inferences [[2], [3], [4], [5]] because they lead to misunderstandings such as that by Beran et al. Similar failure occurs when CIs are used to judge whether the null (or unit) value is in or out of the range, but the researcher must prioritize the analysis for the importance of the ES and the accuracy of plausible clinical values that patients should expect from the treatment [3,5]. Although low power also manifests itself as wide CIs, the analysis of the statistical power is crucial for the feasibility of prospective meta-analyses.

Thus, power curves under the fixed-random effects model as a function of the number of studies were generated based on estimated parameters (ES, number of studies, sample size and heterogeneity) by the meta-analysis that included 5 studies comparing groups supplemented with zinc or SOC practices for mortality outcome. Examination of curves (Fig. 1 ) suggests that at least 8 studies are needed to detect differences with reasonable power in future meta-analyses. Therefore, the inclusion of a priori power analysis in protocols is also a recommended procedure.

Fig. 1.

Power curves as a function of the number of studies. Points along the curve indicate the estimated power according to specified parameters. The dotted line indicates 80% power. Analysis was performed using the R software (version 4.1.1).

Grants and funding

None.

Contributions of each author

CC - 1a. Substantial contributions to study conception and design.

1 b and 1c. Substantial contributions to acquisition of data.

Declaration of competing interest

The authors declare no potential conflict of interest.