According to a study published in the Journal of Clinical Endocrinology & Metabolism, the best algorithm for the early detection of abnormal growth conditions in children (such as Turner syndrome, growth hormone deficiency, and celiac disease) is the Grote clinical decision rule. In addition, the World Health Organization (WHO) growth charts yielded more sensitive values—albeit less specific—when used to evaluate algorithm performance.
Current algorithms used to monitor childhood growth worldwide have provided contradictory indications about their ability to diagnose abnormal growth conditions, and their respective performances have never been analyzed head-to-head. The researchers in this external validation study compared the 7 existing algorithms used to define abnormal growth in children as well as study algorithm performance based on the use of the WHO vs national growth charts.
Researchers extracted growth data on 341 children, including those with Turner syndrome (n=112), growth hormone deficiency (n=163), or celiac disease (n=66), as well as on 3406 healthy children. They applied all 7 algorithms to the data retrospectively using a case-referent approach: each algorithm was analyzed for specificity (percentage of referents considered to have normal growth), sensitivity (percentage of cases considered to have abnormal growth prior to diagnosis), and the theoretical reduction in time to diagnosis (the difference between the actual age at diagnosis vs the age at which the algorithm showed abnormal growth). Growth-monitoring algorithms with specificity greater than 98% were compared for sensitivity using the McNemar test for pairs. Researchers also used this test to compare algorithm performance and the theoretical reduction in time to diagnosis based on which growth chart was used: WHO or national growth charts.
The results showed that specificity values were higher with WHO vs national growth charts (47.4% to 100% vs 44.8% to 100%); however, sensitivity values using WHO growth charts were lower (0% to 85.7% vs 0% to 87.5%). Among the 2 algorithms with specificity over 98%, the Grote clinical decision rule was found to be more sensitive than the Coventry consensus for children with Turner syndrome (40.2% vs 8%; P<.0001), growth hormone deficiency (44.2% vs 6.1%; P <.0001), and celiac disease (4.6% vs 0%; P <.25). For patients with growth hormone deficiency, the Grote clinical decision was also more sensitive when using the WHO growth chart vs national growth charts (54% vs 44.25; P <.0001). Similarly, the Grote clinical decision rule offered a greater theoretical reduction in time to diagnosis using WHO vs national growth charts in children with growth hormone deficiency (0.3 vs 0 years; P <.0001) and Turner syndrome (0.9 vs 0 years); however, the overall short reduction in time to diagnosis is a potentially limiting factor of this algorithm regardless of which growth chart was used.
Limitations of the study included a small sample size consisting of cases obtained from a single region in France, indicating potential recruitment bias. The algorithm performance was evaluated against only 3 conditions associated with abnormal growth; however, growth patterns for other conditions may differ, affecting the performance of the algorithms to monitor growth in children.
The Grote clinical decision rule had the best algorithm performance for the early detection of Turner syndrome, growth hormone deficiency, and celiac disease based on specificity and sensitivity values. However, this algorithm showed limited potential for the theoretical reduction in time to diagnosis. While sensitivity was higher using the WHO growth chart, a loss in specificity was observed, and further studies are needed to establish algorithms that improve sensitivity without sacrificing specificity.
This study was funded by the Ministry of Health, Paris-South University, and more. Please see reference for full list of authors’ disclosures.
Scherdel P, Matczak S, Léger J, et al. Algorithms to define abnormal growth in children: external validation and head-to-head comparison [published online August 20, 2018]. J Clin Endocrinol Metab. doi: 10.1210/jc.2018-00723.