Page 55 - 2019_08-Haematologica-web

Rapid growth and hepcidin suppression in infancy
the hierarchy of factors associating with hepcidin and fer- ritin concentrations at five and 12 months. In these mod- els, higher weight gain since birth remained a significant independent predictor of reduced hepcidin at five months of age in both cohorts (Table 3), and at 12 months in VPM (Table 4), irrespective of birthweight; however, weight change did not associate with ferritin concentration in these analyses. At 12 months of age, hepcidin was posi- tively associated with CRP and season and inversely asso- ciated with sTfR, while at both time points and in each cohort we found significant inverse associations of ferritin with sTfR (Tables 3 and 4).
Together, these analyses not only highlight associations of hepcidin with markers of erythropoietic drive, inflam- mation and season that align with previous observations
in older African pre-school children,21 but also reveal con- sistent evidence that prior weight gain predicts hepcidin (although not ferritin) at key time points across these infant populations, irrespective of birthweight and espe- cially in earlier infancy.
Factors predicting changes in hepcidin and ferritin over time within infants: longitudinal analysis
We then utilized the longitudinal data, fitting fixed effects models to examine how changes over time in time- variant parameters (sTfR, CRP, plasma iron, infections, season, weaning status and growth) predicted changes over time in hepcidin and ferritin within infants. Fixed effects models eliminate confounding caused by any time- invariant characteristic, measured or unmeasured, and
A
B
Figure 2. The relative influence of time-variant factors on changes in hepcidin and ferritin over time during the first year of life in Gambian infants: standardized Forest plots summarizing fixed effect models. Plots depict how a change of one standard deviation of an explanatory variable over time induces a change in standard deviation of outcome variables hepcidin (left panels) and ferritin (right panels) over time within a child in (A) the VPM cohort and (B) the VA cohort. Outcome variables are modeled simultaneously as “seemingly unrelated variables” to account for hepcidin/ferritin correlation, and data represent the pooled analysis of 100 datasets in which any missing data were imputed by multiple imputation, combined using Rubin’s combination rules, as described in detail in the Online Supplementary Methods. Weaning was defined as the first recorded occurrence of any type of feeding other than exclusive breastfeeding. sTfR: soluble transferrin receptor; CRP: C- reactive Protein; Hb: hemoglobin. The wet season was classified as July-October. Plots depict standardized coefficients with 95% Confidence Intervals. *P<0.05; **P<0.01; ***P<0.001. Interactions of weight change with birthweight group [above or below World Health Organization weight-for-age Z-score (WAZ) of -0.5, i.e. close to the median observed across these two cohorts] and sex were modeled. #P<0.05; ##P<0.01; ###P<0.001; respectively, for differences in the coefficients for weight change with respect to higher birthweight females as the reference group. Unstandardized and standardized coefficients are given in Table 5; equivalent analysis based on the dataset prior to multiple imputation is given in Online Supplementary Table S5. MH: higher birthweight males; ML: lower birthweight males; FH: higher birthweight females; FL: lower birthweight females; VPM: Vaccination and Paediatric Microbiome study; VA: a vitamin A supplementation randomized con- trolled trial.22
haematologica | 2019; 104(8)
1547