FIGURE 3.

Forest plot of the effect of small-quantity LNSs on prevalence of iron deficiency (ferritin <12 µg/L). Ferritin concentrations were adjusted for inflammation (i.e., C-reactive protein and/or α-1-acid glycoprotein concentrations, as available), using a regression correction approach adapted from the Biomarkers Reflecting Inflammation and Nutritional Determinants of Anemia (BRINDA) project (29). Individual study estimates were generated from log-binomial regression controlling for baseline measure when available and with clustered observations using robust SEs for cluster-randomized trials. Pooled estimates were generated using inverse-variance weighting in both fixed- and random-effects models. LNS, lipid-based nutrient supplement; PR, prevalence ratio.