Estimation of biological reference intervals of the hemogram in children and young people aged 2 to 18 years using empirical percentile, bootstrap, Harrell & Davis, and Horn’s robust methods

Abstract

Introduction:  Clinical  laboratory  results  should  be interpreted according to biological intervals obtained from reference individuals. Intra- and inter-variability of biological magnitudes and of other factors such as nutrition and geographical origin, play an important role. These values are usually calculated with statisti-cal tools that are not tested for statistical assumptions, or without considering the required sample size for validity.

Methods: Percentile, bootstrap, Harrell & Davis’, and Horn’s robust methods were used in this study to estimate biological reference intervals of 20 parame-ters of the hemogram from 842 young people aged between 2 and 18 years. The classification of pediatric populations by age and sex recommended by Soldin et al. (2003) was considered. Also, the intervals proposed by such authors were compared with those contained in this study for leukocytes, red blood cells, hemoglo-bin, and platelets in order to determine if there were substantial changes in the populations studied, and to establish which of the methods evaluated would provide an interval for the clinician to determine the hematological status of the population studied.

Results and conclusions: In most cases, bootstrap and Horn’s robust methods provided wider intervals than  those  obtained  with  empirical  percentile  and Harrell & Davis’ methods. The lower limit calculated with the robust estimator was far below the median value, and the bootstrap method produced the higher upper limit. Marked differences were observed af-ter comparing each interval with those proposed by Soldin  et  al.,  even  when  intra-individual  variability was low, as in the case of hemoglobin. These results confirmed the need for each laboratory to estimate its own biological intervals using standardized protocols