Six sets of extracted data pertained to experimental groups that did not represent the effect of interest. This affected meta-analytic findings for credibility assessments of misinformation and real information, and intentions to share real information.

The authors estimated the SE from the standardized effect sizes (d) and sample sizes by condition (n) [3]. Where only unstandardized effects were reported, n was also used to calculate d. As such, the d and their SEs were functions of n. Importantly, n, d, and SE constitute all the information used in the meta-analysis.

For 3 studies reporting unstandardized effects, authors made implausible, notably incorrect estimates of n. This affected findings for credibility assessments of misinformation and real information. Multiple other errors regarding n values were also noted.

Lu et al [1] include a moderation analysis by intervention type: active versus passive. In active inoculation, participants contribute to the refutation process [4,5], but Lu et al [1] erroneously classified interventions by medium rather than participant contributions to the refutation process. Incorrect classification occurred for at least 9 high-powered studies.

These represent the most important findings of our analysis, which was constrained to appraisals of the supplementary materials, the primary studies, and two email exchanges with a representative author. We request this letter’s publication to raise awareness of these concerns. We further ask that all issues detailed in the full report be adequately addressed.