Dear editor,

We read with interest the systematic review and meta-analysis on adherence to oral zinc supplementation for acute diarrhoeal disease in children under five [Reference Pradhan1]. The topic is important for translating guideline recommendations into real-world practice, and the authors should be commended for focusing on adherence rather than efficacy alone. Their prospective registration, broad search strategy, and explicit use of a structured risk of bias tool are notable strengths that align with contemporary guidance for systematic reviews [Reference Page2].

Several methodological issues, however, warrant consideration when interpreting the pooled adherence estimates. First, although the review is framed as including observational studies, the eligibility criteria and final sample also incorporate at least one prospective open-label interventional study. Mixing observational and interventional designs is not inherently problematic, yet it should be clearly acknowledged and justified, since different designs may have distinct patterns of adherence, measurement error, and risk of bias [Reference Higgins3].

Second, the application of the NHLBI tool for observational cohort and cross-sectional studies is described as yielding a 0–6 summary score, with higher scores interpreted as a higher risk of bias. The original developers caution against creating simple summary scores and against reversing the intuitive direction of such scores because different items are not psychometrically equivalent and do not contribute equally to bias [Reference Sanderson, Tatt and Higgins4]. Without specifying which items were selected, how they were coded, and why higher values denote worse quality, readers may struggle to understand how study-level bias influenced the synthesis. This is particularly relevant because one included study is interventional, for which an observational tool may not be fully appropriate.

Third, the meta-analytic pooling of adherence proportions occurs in the presence of extremely high between-study heterogeneity (I² values close to 100%) and clear evidence of publication bias. When outcome definitions, settings, and measurement methods differ substantially across studies, a single pooled prevalence may have limited meaning [Reference Higgins3]. A stronger approach would be to emphasise the range and pattern of adherence across contexts, supplemented by exploratory subgroup or sensitivity analyses based on setting, regimen duration, and adherence definition, rather than relying on a precise summary estimate.

Finally, the review treats adherence as a homogeneous construct, although the included studies define it variably as completion of a 10- or 14-day course, or as achieving a minimum number of days based on caregiver report. Methodological literature on outcome harmonisation stresses that conceptually distinct definitions should be handled with caution in quantitative synthesis, or at least subjected to careful stratification [Reference Page2, Reference Higgins3]. Explicitly mapping and grouping these definitions would help readers understand whether low adherence reflects shorter duration, missed doses, early discontinuation after symptom resolution, or other behavioural patterns.

While this review addresses a clinically important question and is methodologically transparent in several respects, the mixing of designs, non-standard risk of bias scoring, very high heterogeneity, and variable outcome definitions suggest that the pooled adherence estimates should be interpreted cautiously. We hope these comments assist readers and editors in appraising the evidence and inform future work on adherence to zinc and other supportive therapies in childhood diarrhoea.