Previous studies looking at patients with various etiologies of short stature either at single institutions or from large databases such as the Kabi International Growth Study (KIGS) or the ANSWER Program have attempted to define pre-Rx factors that best predict short-term GH responsiveness [1–4]. Goals of these studies have been to inform the selection of patients most likely to benefit from GH therapy, to predict the magnitude of the increase in HV during therapy, and to help pick the optimal dose of GH to produce the desired increase in HV. Younger age has been found to predict better responsiveness to GH, as well as lower peak GH levels in response to provocative testing, taller parents, and a higher BMI SD score. In patients with IGHD, these predictors make clinically relevant “sense” since they define the characteristics of this group, and previous studies have shown that the most severely GH-deficient children tend to be the most responsive to GH [1, 3, 4].
An earlier report using the NCGS database showed that the increase in height SDS during the first year of GH therapy was significantly greater in patients with peak GH < 3 ng/mL (+1.14) than in those with peak GH of 3 to 7 ng/mL (+0.81) and 7 to 10 ng/mL (+0.72) . In addition, children with severe GH deficiency tend to have increased body fat stores, as noted in a study that found a high correlation between baseline leptin and first-year change in height SDS (r = 0.49; P <0.0001) in a sample of 150 Swedish children with a range of peak GH values . Thus it is not surprising that the best responders in our IGHD group have a significantly higher BMI SDS than the average responders, although the difference in mean BMI SDS between the average and poor responders was relatively small. In the ISS group, BMI was also positively associated with response. Parental height clearly also contributes to responsiveness, as short children of average or tall parents have greater genetic potential and usually have a greater height deficit to recoup during the first 2 years of therapy. For IGHD patients, the bigger the difference between the mid-parental target height SDS and the patient’s baseline height SDS, the greater the HV was likely to be during both the first and second years of GH treatment, but for ISS patients, the effect of this difference on HV was much smaller during the first year and non-significant for the second year.
While our data confirm some of the predictors already known, our approach to defining GH responsiveness in patients with IGHD and ISS differs from other approaches in several respects. We were able to define the best and the poorest responders not based on arbitrary criteria, such as a certain increase in HV or in height SDS, factors that depend to some extent on the age of the patients, but based on age, sex, and diagnosis-specific HV curves for both the first and second years of GH therapy from large numbers of similarly treated patients. These curves, developed by Bakker et al. [5, 6] were derived from a large sample of GH-treated patients with IGHD and ISS contained within the same NCGS database as was used for our analyses. We also wanted to examine both the best and the poorest responders using the same baseline criteria to see whether the factors associated with poor response deviated from those found in the average responders to the same degree but in the opposite direction as the best responders. The fact that there was a much greater separation between the best and average responders compared with the average and poor responders suggests that the best responders make up a more distinct subset of patients than do the poor responders. Poor response to GH may stem from factors we have not assessed, such as undefined genetic differences in skeletal responsiveness and poor compliance with treatment. While we do not have any data on compliance, such issues with GH treatment are not uncommon [10, 11]. Age and dose were eliminated as predictors of response for both IGHD and ISS in our analysis because they were similar between the comparison groups. Of the pre-Rx variables that differed between the patients, we found that the best responders in both the IGHD and ISS groups had lower pre-Rx HVs and higher BMI, with differences seen only during the first year of therapy; second-year responsiveness to GH was not related to pre-Rx HV. Lower peak stimulated GH and taller parents predicted second-year responsiveness in the patients with IGHD; there were no strong predictors of second-year responsiveness in the ISS group.
For comparison, in an analysis of 169 Swedish prepubertal children with a mean age of 8.3 years and a wide range of stimulated GH levels, all treated with 0.1 U/kg/day, Kristrom et al. reported that the maximum GH after arginine-insulin, age at start of treatment, mid-parental height SDS, pre-Rx height SDS and HV, and the difference between pre-Rx height SDS and mid-parental height SDS (diff SDS) all showed significant correlations (r = 0.22 to 0.43) with the increase in height SDS over 2 years of therapy . By stepwise linear regression, diff SDS and log GHmax were the strongest predictors. Using the KIGS database for patients with only IGHD, Geffner and Dunger  reported that degree of GH deficiency, age, height – mid-parental height SDS, and weight SDS (which correlated strongly with height SDS) were, in that order, the variables most predictive of HV during the first year of therapy. For children with ISS, Ranke et al.  reported that by multivariate linear regression, the four variables associated with first-year response to GH were age, GH dose, height – mid-parental height SDS, and weight SDS. In both examples, if you eliminate age and dose, the degree of GH deficiency remains the most consistent variable.
For IGHD, it is clear that the best response in the first and second years was seen in the patients with the greatest difference between starting height and the target height. More recently, Lee et al., using the ANSWER study database involving 698 children with IGHD, found that the change in HV at 4 months, baseline age, baseline height SDS, and baseline BMI SDS were, in that order, the best predictors of the increase in height SDS at 1 and 2 years after starting GH; they did not report predictors of response for their smaller group of 123 children with ISS . We found more predictors of excellent response to GH therapy in the IGHD group than in the ISS group, and several of these predictors for IGHD but not for ISS had significant effects into the second year.
One unique aspect of this study was our ability to see how the classification of IGHD and ISS responses to rhGH, which were found during the first year of therapy, changed when these groups were followed into the second year. One might have predicted that the best and poorest responders during the first year would largely remain in the same categories during the second year, but this was not the case. While a large proportion (42%) of patients with IGHD who were best responders in the first year remained in this group during the second year, and relatively few (4%) became poor responders, a greater proportion (58%) of the IGHD first-year poor responders became average responders during the second year, with only 37% remaining poor responders during the second year. In the ISS group, the comparable second-year numbers were 53% moving from poor to average responders and 36% remaining poor responders. A potential weakness of the analysis is that our results may have been biased by the possibility that some children who responded poorly during year 1 may have been taken off treatment by either the physician or by the family; they would not show up in our analysis because we only looked at patients completing 2 years of therapy. Our drop out analysis confirms that bias since those who did drop out tended to have a somewhat poorer height velocity, albeit only about 5% on average less than those who remained in the study for 2 years. Nevertheless, most of those who grew poorly in the first year but who continued treatment grew satisfactorily or well during the second year.
There are important implications of these findings. First, as suggested above, poor responders may be more similar to average responders than to best responders. Second, as Tanner et al. demonstrated some 46 years ago , in general, the correlation of height gain in prepubertal children from year to year is only about 0.3 and this may underlie growth responsiveness to hormonal treatment as well. Compliance may also be a factor as could changes in the psychosocial environment . Thus, these findings suggest that a suboptimal first-year growth response does not always predict a continued poor response in the second year. The treating physician may consider continuing therapy into the second year, with appropriate attention to factors such as compliance and psychosocial factors, and the knowledge that there may be episodic changes in HV that are inherent to the growth process and not obscured by GH treatment.