ESPE Abstracts (2015) 84 P-3-1028

Comparison of the Performance of Algorithms Proposed to Standardize Growth Monitoring

Pauline Scherdela,b, Soraya Matczakc, Raja Braunerd, Jean-François Salaüne, Barbara Heudea & Martin Chalumeaub,c


aINSERM U1153, Equipe de recherche sur les Origines précoces de la santé et du développement de l’enfant (ORCHAD), Centre de Recherche Epidémiologie et Biostatistique Sorbonne Paris Cité (CRESS), University, Paris, France; bINSERM U1153, Equipe de recherche en Epidémiologie Obstétricale, Périnatale et Pédiatrique (EPOPé), Centre de Recherche Epidémiologie et Biostatistique Sorbonne Paris Cité (CRESS), Université Paris De, Paris, France; cDepartment of General Pediatrics, Necker Children’s Hospital, AP-HP, Paris, France; dUnité d’Endocrinologie Pédiatrique, Fondation Ophtalmologique Adolphe de Rothschild, Université Paris Descartes, Paris, France; eAssociation Française de Pédiatrie Ambulatoire, Gradignan, France


Background: In industrialised countries, the main goal of growth monitoring (GM) of apparently healthy children is the early detection of severe underlying conditions. However, empirical evidence suggests globally poor performances of GM, with important diagnosis delays for priority target conditions and many unnecessary referrals for diagnostic work-up for disease-free children.

Objective and hypotheses: To evaluate and to compare the performance for early detection of GH deficiency (GHD) of seven algorithms proposed in the literature to standardize GM: the WHO criterion, the Coventry consensus, the Dutch consensus, GHRS criteria, the Grote clinical decision rule and Saari’s clinical decision rules.

Method: In a case-referent study, we applied the seven algorithms on growth data of 33 children with GHD related to pituitary-stalk interruption syndrome (cases), and 2 200 apparently healthy children followed longitudinally from birth (referents). The sensitivities, specificities and theoretical improvement in time to diagnosis of these rules using French growth charts were calculated, and highly specific (>98%) rules were compared.

Results: Sensitivities and specificities varied from 15 to 100% and 48 to 100% respectively. Among the two algorithms with a specificity >98%, the Grote clinical decision rule had a higher sensitivity (67% vs 15%, P<0.0001) and offered a better theoretical improvement in time to diagnosis (4 months vs 0 months, P<0.0001) than the Coventry consensus respectively.

Conclusion: Among algorithms proposed for GM, the Grote clinical decision (Grote 2008) had the best performance for early detection of GHD, using French growth charts. Its performance on other target conditions of GM and using WHO growth charts must be evaluated.

Article tools

My recent searches

No recent searches.