The analysis of heart rate fluctuations, or heart rate variability (HRV), may be applied to explore children\'s neurodevelopment. However, previous studies have reported poor reliability (repeatability) of HRV measures in children at rest and during light exercise. Whether the reliability can be improved by controlling variables such as physical activity, breathing rate and tidal volume, or by selecting non-conventional techniques for analysing the data remains as an open question. We evaluated the short-term repeatability of RR-interval data from medicated children with congenital hypothyroidism (CH). The alpha(1) exponents, obtained by detrended fluctuation analysis (DFA), from the data of 21 children collected at two different sessions were compared. Elapsed days between sessions were 59 +/- 33, and data were obtained during 10 min, trying to restrict the children\'s activity while being seated. We found statistical agreement between the means of alpha(1) exponents for each session (p = 0.94) and no bias with a low-coefficient variation (9.1%); an intraclass correlation coefficient ri = 0.48 ([0.14 0.72], 95% confidence interval) was also estimated. These findings, which were compared with results obtained by conventional time and frequency techniques, indicate the existence of agreement between the alpha(1) exponents obtained at each session, thereby providing support concerning the repeatability of HRV data as analysed by DFA in children with congenital hypothyroidism. Of particular interest was also the agreement found by using the central frequency of the high-frequency band and the parameter pNN20, both showing better or similar ri than alpha(1) (0.77 [0.57 0.89] and 0.51 [0.17 0.74], respectively), yet considerably better repeatability than other conventional time and frequency parameters.