Bibian van der Voorn

133 EARLY LIFE GROWTH ON CORTISOL AFTER PREMATURITY preterm birth with cortisol levels that were either higher at bedtime 33 or throughout the day 34 . Since we did not include a term control group, the results of these studies cannot be compared to ours. Moreover, we collected only 2 salivary samples post- awakening, and may therefore have missed the peak cortisol concentration. In addition, a single collection day may not be sufficient to demonstrate differences at age 8 year 35 . It is increasingly recognized that preterm birth constitutes a major risk factor for cardiometabolic disease 36 . However, in studies providing long-term follow-up of preterm populations, no differences in insulin resistance or blood pressure were demonstrated between subjects with and without intrauterine growth restriction 37,38 . Possibly, the subjects within these cohorts were too young to demonstrate such differences. This may also partly explain the lack of association in our sample at age 8 y. Life-long follow-up of preterm populations is warranted, since it is conceivable that alterations in HPA axis activity in preterm infants during early life are involved in pathways leading to cardiometabolic disease and neurodevelopmental impairments. The main strength of this study is the use of a well-described birth cohort, in which extensive information on early growth was available. In addition, serial measurement of serum cortisol in early infancy as well as in childhood, and measurement of salivary cortisol at age 8 y were performed. There are several limitations. First, this study was a post-hoc analysis of a nutritional RCT, in which a term born control group was not included. Second, attrition at age 8 y limited our sample size. We therefore carefully assessed the possibility of attrition bias by comparing participants with non-participants and excluded subjects. Since there were no differences, analyzing with GEE gave us the opportunity to use all available data in longitudinal analyses, while accounting for missing data 20 . With this approach, we followed the suggested reporting requirements for addressing attrition as described by Fewtrell 21 . Third, subgroup analyses were performed in relatively few subjects. Fourth, the exact timing of cortisol sampling was not the same for all subjects and the early morning would have had our preference. Since this was not possible, we assessed fasting blood levels at term age, 3 and 6 mo. corrected age, as a second best option 39 . However, since there was no difference in mean sampling time between AGA and SGA subjects we considered this to have no influence on our results. In addition, despite our effort to standardize our salivary cortisol collection protocol by giving clear instructions and reporting time of sampling, it was not optimal, and the LLOQ limited the sensitivity of values in the lower ranges (i.e., ≤1 nmol/L). Considering the recently published guidelines of Stalder et al., the small