vis/OneDrive/Desktop/Job%20Applications%20for%20BU/Fink%20et%20al%20--%20clinical%20trial%20growth%20moni... Q Update ng Zambia.pdf 5 8 100% ... for the study. d of the study. "(OL'0 = d) sdnora ble 3 for further information). group. On average, across the 3 groups, the household. A total of 126 community y 2015. For these meetings, all children ed at baseline, and 1% was receiving AZ, no case of severe acute malnutrition s encountered at baseline. Almost 90% of Id (median age: 13 mo) at baseline, with a mean WAZ of -0.7, a mean weight-for- , and a mean MUAC of 14.8 cm. Despite re made across the 3 rounds of meetings 1 child/household was in the study; in mmunity were invited. A total of 3341 from 1 household were in the study, 2 aseline demographic and anthropometric the caregiver refused measurement, and e HBGM group immediately after com- ly significant differences were found in ross the 42 CBGM+NS villages between id not have valid anthropometric assess- th charts were installed by study staff in were not recorded correctly on the mobile end in September 2015. Of the 512 end- ber 2014. We reassessed 512 children es the overall study design and participant nrolled in the study across 127 villages in survey. In the HBGM group, in most 4-2948 and by ERES Converge in Zambia 014-June-011. Consent for study partici- 47 of 569 targeted children (96.1%) and om the head of each household. Given the of Human Research Administration under children's stunted growth status at base- nts for multiple testing (multiplicity) were mo period with initially aligned outcome ified by age and formally tested for effect ed risk of the study, no trial monitoring lly interacted models. Given the limited -2 showed children in the "red zone"), such heterogeneous treatment effects, we on about children not reaching age-specific ved by the Harvard TH Ghan School of the low likelihood of finding large group rventions primarily focused on providing inses were likely for parents with children growth faltering seems to occur empire effects also seemed likely for children To test this empirically, we performed ention group with mean outcomes in the Huber (17). All models compare mean FINK ET AL TABLE 2 Intervention impact estimates Mean difference Odds ratio HAZ NTER-NDA WAZ Food diversity Stunted growth Underweight Unadjusted HBGM 0.153 (-0.101, 0.407) 0.018 (-0.116, 0.153) 0.178" (-0.021, 0.377) 0.159 (-0.543, 0.862) 0.794 (0.531, 1.189) 0.61 1 (0.300, 1.243) CBGM+NS -0.105 (-0.340, 0.131) -0.050 (-0.182, 0.083) 0.018 (-0.178, 0.215) -0.339 (-0.976, 0.297) 1.143 (0.769, 1.699) 0.804 (0.441, 1.467) Control Ref. Ref. Ref. Ref. Ref Ref. Clusters, n 125 12 126 126 125 125 497 504 107 512 497 Observations, n Adjusted HBGM 0.127 (-0.107, 0.361) -0.017 (-0.133, 0.098) 0.183 (0.037, 0.328) 0.219 (-0.500, 0.938) 0.734 (0.407, 1.324) 0.725 (0.235, 2 234) CBGM+NS -0.152 (-0.341, 0.036) -0.118* (-0.230, -0.006) -0.066 (-0.189. 0.056) -0.216 (-0.889, 0.457) 1.294 (0.689. 2.432) 1.091 (0.423, 2.816) Ref. Ref. Ref. Ref. Ref. Ref. Control 125 12 126 125 125 Clusters, n 481 493 Observations, n 482 486 482 Results are from multivariate linear regression models. Coefficients displayed in columns 2-5 represent mean differences (95% CIs). Coefficients displayed in columns 6 and 7 represent ORs (95% CIs). All adjusted models include controls for age (months); sex; twin status; height, weight, and weight for height at baseline; health at baseline; caregiver education; household composition; and household wealth quintile. We used cluster-robust SEs to account for within-cluster correlation. We used f tests to test hypotheses. Stunted growth was defined as HAZ