A Bayesian Approach to Diagnose Growth Hormone Deficiency in Children: Insulin-Like Growth Factor Type 1 Is Valuable for Screening and IGF-Binding Protein Type 3 for Confirmation.

2020 
BACKGROUND The utility of insulin-like growth factor type 1 (IGF-1) is well established in the diagnosis of growth hormone deficiency (GHD), whereas IGF-binding protein type 3 (IGFBP-3) has a more controversial role. Most studies evaluated the value of these peptides by assessing their sensitivity and specificity but not considering the low prevalence of GHD among short children (<2%). OBJECTIVE To evaluate the utility of basal IGF-1 and IGFBP-3 values in the GHD diagnosis process with a Bayesian approach, based on pre- and post-test probability. METHODS We determined ROC curves, sensitivity, specificity, and positive and negative predictive values for IGF-1 and IGFBP-3 obtained from patients with GHD (n = 48) and GH-sufficient children (n = 175). The data were also analyzed by classifying the children into early childhood and late childhood (girls and boys younger and older than 8 and 9 years, respectively). RESULTS The area under the curve (AUC) of the receiver operating characteristic curve of IGF-1-SDS (standard deviation score) was greater than that of IGFBP-3-SDS (AUC 0.886 and 0.786, respectively, p = 0.001). In early childhood, the AUC of IGFBP-3-SDS was significantly improved (0.866) and similar to IGF-1-SDS (0.898). IGF-1-SDS, in comparison to IGFBP-3-SDS, had a greater sensitivity (92 vs. 45.8%, respectively), lower specificity (69 vs. 93.8%, respectively), and lower positive predictive value (5.7 vs. 13.1%, respectively), with similar negative predictive values. CONCLUSION IGF-1-SDS is a useful screening tool in the diagnosis of GHD. Although IGFBP-3-SDS lacks sensitivity, its high specificity supports the role to confirm GHD in short children, especially in early childhood. This strategy could simplify and reduce the necessity of a second laborious and expensive GH stimulation test to confirm the diagnosis of GHD.
    • Correction
    • Source
    • Cite
    • Save
    • Machine Reading By IdeaReader
    19
    References
    2
    Citations
    NaN
    KQI
    []