Editor,—We read with interest the paper by Mitchell and colleagues1 and wish to add our own observations on this subject. In 1996 the Regional Endocrine Laboratory started to provide a service for the measurement of insulin-like growth factor binding protein (IGFBP-3) following early reports that this was a good marker of growth hormone secretion. We then undertook a retrospective audit of the measurement of serum insulin-like growth factor (IGF-1) and IGFBP-3 as predictive markers of growth hormone deficiency (GHD) in children undergoing growth hormone stimulation tests (glucagon and insulin tolerance tests). Between October 1996 and January 1998, 93 children had simultaneous measurements of IGF-1 and 78 children had measurements of IGFBP-3. We defined GHD as a peak growth hormone level of < 20 mU/litre and complete GHD as a peak < 10 mU/litre in response to a stimulation test.

The results for IGF-1 and IGFBP-3 were compared to reference ranges for age available in the laboratory and classified as low or normal. The reference range for IGF-1 was constructed by the laboratory using their own assay and that for IGFBP-3 being supplied by the manufacturers of the kit (Nichols Institute, San Juan Capistrano, California, USA). We calculated their sensitivity and specificity as predictors of GHD using the two different cut off levels and the likelihood ratio—that is, the likelihood that the result would be seen in someone with as opposed to someone without GHD (table 1).

Eight children had both a low IGF-1 and IGFBP-3, which produced a sensitivity of 22.2% and specificity of 90.4%, with a likelihood ratio of 2.3 in predicting GHD. Therefore the combination of a low IGF-1 and low IGFBP-3 would be highly suggestive of GHD, but a significant number of children with GHD will have normal values for either of these two markers.

Thus it can be seen that a single measurement of IGFBP-3 performed no better than IGF-1 as a marker of growth hormone secretion despite previous claims. Neither marker had a high likelihood ratio and would therefore not be good as a single predictive test. Although we realise that some of the normal IGFBP-3 results could have resulted from the presence of IGFBP-3 protease activity interfering with the assay in children with radiation induced GHD this is not likely to alter our findings significantly.

Thus we agree with Mitchell et al and other authors2 that IGFBP-3 measurements are not good predictive markers of growth hormone secretion and do not replace the need for careful clinical evaluation and growth hormone stimulation tests in short, slowly growing children.