137 Body composition measurement methods in preterm infants 7 Risk of bias (quality) assessment Two reviewers (D.F.J.Y. andD.d.J.) primarily assessed bias using the Critical Appraisal Skills Program (CASP) checklist. 15 In case of any discrepancies between the two reviewers, the two reviewers discussed and asked the expert opinion of the two other reviewers until an agreement was made. The synthesis was based on the final decision made under agreement of all reviewers. The quality of individual studies was assessed with CASP checklists. In addition the Oxford Centre for Evidencebased Medicine’s Levels of Evidence was used to grade the level of evidence of each manuscript.16 Strategy for data synthesis A narrative synthesis was primarily done by two researchers (D.F.J.Y. and M.M.v.W.) and was reviewed by D.d.J., J.C.F.K. and H.N.L. before finalization. Results Out of 1884 identified records, 48 full-text articles were assessed for eligibility and 19 were included in this synthesis. (Figure 1) Nine studies (n=1539) reported about the predictive value or validity of body proportionality measures. Five studies (n = 319) investigated the validity of bioelectrical impedance analysis (BIA), three studies (n=90) investigated the validity of SFT, two studies (n=24) investigated the validity of ADP, one study (n=63) investigated the predictive value of ultrasound and one study (n=15) investigated the validity of MRI. There were no human studies which reported about the validation of DXA and isotope dilution studies in preterm infants. (Table 3) Body composition measurements were performed at various postnatal ages, ranging from 24 hours postpartum to 4 months corrected age. Body proportionality measures Table3showsourfindingsonthepredictivevalueandvalidityofbodyproportionality measures. Weight and length indices had a moderate to good predictive value for fat-free mass (in grams). On the contrary, the predictive value of weight and length indices for fat mass was poor to moderate and fat mass percentage was poorly predicted by weight/length indices. 17-21 Larcadeetal. andSimonetal. assessedpredictiveequationswithclinical parameters, such as caloric and macronutrient intake, and z-scores for weight, length and head circumference.22,23 They found that fat-free mass (g) could be predicted by the amount of human milk feeding, respiratory support, antenatal corticosteroid use, growth parameters and sex. A newly modeled equation by Larcade et al showed good agreement for fat free mass (g). However, Larcade and colleagues did not assess fat mass percentage, while Simon and colleagues could only explain 24% of the variance in fat mass percentage with their predictive model. 22,23 (see table 3)
RkJQdWJsaXNoZXIy MTk4NDMw