96 Chapter 5 Statistical analysis Continuous variables were presented as mean and standard deviations (SD). Dichotomous variables were presented as count and percentages. Mixed models were used to predict the change in IGF-I and the SDS of weight, length and head circumference for every individual in the previously mentioned time frames . Partial correlations, controlling for gestational age at birth and postmenstrual age at the time of body composition measurements, were reported for associations between the (change in) SD scores for weight, length and head circumference and fat mass, and fat free mass. Likewise, partial correlations were reported for IGF-I and the aforementioned outcomes. Due to the non-normal distribution of fat (free) mass percentage the association between the (change in) SD scores for weight, length, and head circumference and fat (free) mass percentage was explored using linear regression modeling. Gestational age at birth and PMA at the time of body composition measurement were entered in the regression model as covariates. Potential confounders were evaluated with stepwise (backward) regression analysis. All statistical analyses were done using IBM® SPSS® Statistics 24 for Windows (IBM Corp., Armonk, NY, USA). Two-sided statistical significance was assumed at p-values less than 0.05. Results Eighty-seven infants were included in the primary analysis for growth and IGF-I. Sixty-five infantswere assessed at termequivalent age of whom58 completed body composition measurement. (Figure 1) The sample size for IGF-I measurements per postmenstrual age is shown in table 1. Baseline characteristics are shown in table 2.
RkJQdWJsaXNoZXIy MTk4NDMw