Large databases facilitate the examination of those factors, which may determine response to treatment, potentially to develop prediction models. We have used KIMS (Pfizer's international metabolic database of GH-treated adults), to examine parameters that might predict clinical response to GH therapy in hypopituitary adults.
Patients/methods: Patients with non-functioning pituitary adenoma on GH titrated to maintain serum IGF-I between minus 2 and 2 SDS, and with appropriate replacement of other hormone deficiencies were selected from the database (n-326). Potential response-predictors (gender, age at GH start, GH peak at diagnosis, type of tumour treatment, GH starting dose, duration of GHD, baseline: IGF-I SDS, LDL-cholesterol, and QoL-AGHDA) were correlated against changes in waist circumference, waist/hip ratio (WHR), LDL-cholesterol and QoL-AGHDA score after 1 year of treatment. Potential effects of regression to the mean were explored.
Results: Significant inverse correlation was found between QoL at baseline vs change in QoL (r= minus 0.54) and between LDL cholesterol vs change in LDL cholesterol (r= minus 0.44). The correlation between gender vs. change in QoL was also statistically significant (r= 0.17 p < 0.01)
Conclusions: Males are more responsive to GH than females although the significance is low consistent with inter-individual variation. Baseline LDL-cholesterol (n=326) and QoL-AGHDA score (n=217) predict LDL and QoL responses respectively to GH treatment. Thus patients with worse QoL (higher QoL-AGHDA score) and/or higher LDL-cholesterol values improve most. The effect of regression to the mean accounted for some of the change in LDL-C. However, the limited ability to predict treatment response, as shown by small R values suggests that all GH-deficient patients should be considered for GH therapy and continue according to individual clinical response.
22 - 24 Mar 2004
British Endocrine Societies