0021-972X/99/$03.00/0 The Journal of Clinical Endocrinology & Metabolism Copyright © 1999 by The Endocrine Society
Vol. 84, No. 4 Printed in U.S.A.
Editorial: A Mathematical Model for Predicting Growth Response to Growth Hormone Replacement Therapy—A Useful Clinical Tool or an Intellectual Exercise? The end of this century marks a 40-year experience with GH replacement therapy, the first 25 years with human pituitary extracts (hpGH) and the subsequent 15 years with biosynthetic recombinant human GH (rhGH). The clinical research experience with hpGH was hampered by the inherent limitation in supply, with frequent interruptions in treatment, inadequate dosages, and outcomes that were often disappointing (1, 2). Nonetheless, early investigators were able to document correlation of growth response with hpGH dose (3–5), and that younger age, greater height for age, and more retarded bone age dictated a better response (4). Clinical trials with rhGH began in the U.S. in 1981, and the predictive factors for growth response of 79 patients were reported in 1986 (6). In this mixed population with idiopathic and organic GH deficiency (GHD), on a standard GH dose of 0.1 mg kg body weight injected three times per week, 48% of the variability in first-year growth response could be accounted for in a model that included pretreatment somatomedin-C concentration, age, body mass, and maternal height. Subsequent analysis from the National Cooperative Growth Study (NCGS), a post-marketing ongoing survey, of 523 children with idiopathic and 109 children with organic GHD, produced a model in which 40% of the variability of growth response in the idiopathic group was explained by age, log maximum GH response to stimulation, weight adjusted for height, dosing schedule, dose, and mean parental height. Those with organic GHD grew less well; for them three variables, pretreatment growth rate, log maximum GH response, and age, predicted only 20% of their growth response variability (7). Analysis of the other large post-marketing database, the European-based Kabi International growth study (KIGS), revealed first-year treatment response in 472 prepubertal children with idiopathic GHD to be a function of age, height sds minus mean parental height sds, GH injection frequency and dose per kilogram, birth weight sds, and weight for height at the time of treatment. These factors accounted for 56% of the variability in response (8). Achermann et al. (9) have recently confirmed the influence of birth weight sds on response to GH replacement in 16 GHD children. In this issue of JCEM, Ranke et al. (10) (see page 1174) report a further tapping of the KIGS database. In an effort to develop a tighter predictive model, they have used stricter inclusion criteria, confining their study to those idiopathic GHD patients who were receiving six or seven GH injections Received February 4, 1999. Accepted February 11, 1999. Address correspondence to: Arlan L. Rosenbloom, M.D., Children’s Medical Services Center, 1701 SW 16th Avenue, Gainesville, Florida 32608. E-mail:
[email protected].
per week and who had height assessment between 11 and 13 months of treatment. Despite these stricter criteria, the modestly larger population in the present study (n 5 593) had only 45% of variability in growth response attributable to the variables analyzed in the previous study. It is likely that the lowered predictability, despite stricter inclusion criteria, resulted from the exclusion of frequency of injections, which was no longer a variable in the current cohort. When the log of the serum GH response to provocative testing was added to the model, 61% of response variability was explained, and this factor was the most influential. This is not surprising, because the definition of GHD was highly sensitive, but not very specific. There is abundant data to indicate that upwards of 50% of patients thus diagnosed will no longer have GHD with retesting a few years later or in adulthood (11). This study confirms that GHD patients respond more dramatically to rhGH than do questionably GHD short children. This model was validated, albeit with considerable variability, in several other population groups. As has been noted in other studies, subsequent years of growth response to GH replacement are primarily predicted by the first-year growth velocity. The finding of a positive correlation between birth weight and response to GH replacement in the earlier KIGS study (8), confirmed in the present report (10) is unexplained. Birth weight and length are usually normal in idiopathic GHD (12), including severe congenital deficiency such as that due to GH gene deletion (13). This would indicate that GH is not critical for fetal growth. The suggestion that the better GH treatment response with greater birth weight might reflect greater GH sensitivity in heavier infants (9, 10) is inconsistent not only with the lack of influence of GH on birth weight, but with the fact that GH insensitivity or deficiency typically results in increased body fat and weight for height (12–14). The finding of a positive influence of weight for height at the time of treatment raises the question of whether these two variables are linked. In their small population, Achermann et al. (9) found no relation between birth weight sds and body mass index sds at the time of treatment. In these analyses, birth weight may be more a reflection of birth size than of adiposity, because lengths are not recorded. Small for gestational age (SGA) babies may be at increased risk for postnatal GHD. Balsamo et al. (15) have analyzed the growth responses to three years of GH replacement therapy in 16 GHD children who were SGA compared with 16 who were appropriate weight for gestational age. Although both groups had substantial and similar catch-up growth in the first-year, the SGA group had a more rapid decline in velocity in the subsequent years. The authors concluded that a constitutional component of statural deficiency in the SGA
1172
ROSENBLOOM
children prevailed over their hormonal deficiency during treatment with GH. Such an explanation might also apply to the findings with the KIGS population. In the Balsamo et al. (15) study, SGA was defined as length below the 10th percentile. With a median birth weight SDS of 20.6, a mean of 20.5, and a range of 22 to 13.6, the KIGS population likely includes a substantial number of patients who were SGA at birth. The positive correlation of growth response to GH with weight sds at the time of treatment as an independent variable (i.e. not the result of a relatively higher dose for lean body mass or surface area in the heavier child) has also not been explained. Increased weight for height may reflect the severity of GHD (which would positively affect growth response) independently of serum GH response to provocative testing. The authors are enthusiastic about the ability of this model and others to be developed for other conditions that are being treated with GH, “. . . to assist the clinician in the individualization of the patient’s treatment from the start to the end of GH therapy.” These models are to be made available through user-friendly computer programs. It is difficult to imagine how this promise can be fulfilled by a model that explains, on average, only 60% of response variability the first year and considerably less thereafter. Despite the availability of comparably predictive models based on large populations for some time, there has not been an effort to apply these models to individual patient monitoring. Although much is written about problems with the diagnosis and definition of GHD (11), there does not seem to be any concern among pediatric endocrinologists about the lack of a reliable predictive model for GH response. There is good reason for this. The growth specialist bases clinical judgments on auxologic data (pretreatment growth velocity, current stature and weight, and measured parental heights), and on severity of GH deficiency, duration and etiology of GHD, associated deficiencies, osseous maturation, and initial response to therapy. Because the goal of treatment is for the child to catch up to and continue growing on the normal curve in a track appropriate for genetic endowment, the above information and a population appropriate growth curve in a chart or on a computer growth program are the relevant tools. The individual physician treating an individual patient has more specific data at hand than were used in the development of the model, such as pretreatment growth velocity, bone age, and IGF-1 or IGFBP-3 measurement. Putting less data than the physician has about the patient into a model will neither provide more information to use for decision-making or counseling nor suggest how to evaluate inadequate growth responses, which do not require a complex model to be recognized. General reliability in predicting average responses in other groups is not relevant to the question of usefulness or accuracy in applying the model to individual patients.
1173
In answer to the question posed in the title of this commentary, therefore, although the clinical usefulness of the model described by Ranke et al. (9) is dubious, their analysis remains useful. Despite the lack of standardization for the data collection and the use of a variety of clinical laboratories, analyses of predictive variables for growth response in large groups of patients such as the KIGS and NCGS databases can provide clues about the natural history of growth response to rhGH in GHD and other disorders. An example is identifying the need for changing dosage requirements with age and development stage. The results of these analyses can inform clinical decision making and can provoke questions to address in prospective studies. I thank Janet H. Silverstein for her review and advice. Arlan L. Rosenbloom Children’s Medical Services Center Gainesville, Florida 32608 References 1. Burns EC, Tanner JM, Preece MA, Cameron N. 1981 Final height and pubertal development in 55 children with idiopathic growth hormone deficiency, treated for between 2 and 15 years with human growth hormone. Eur J Pediatr. 137:155–164. 2. Coste J, Letrait M, Carel JC, et al. 1997 Long-term results of growth hormone treatment in France in children of short stature: population, register based study. Brit Med J. 315:708 –713. 3. Aceto T, Frasier DS, Hayles AB, et al. 1973 Collaborative study of the effects of human growth hormone in growth hormone deficiency: III. First eighteen months of therapy. In: Raiti S. ed. Advances in human growth hormone research. Washington, DC: DHEW Publication No. (NIH); 74-612; 695–714. 4. Tanner JM, Whitehouse RH, Hughes PCR, Vince FP. 1971 Effect of human growth hormone treatment for 1 to 7 years on growth of 100 children, with growth hormone deficiency, low birth weight, inherited smallness, Turner’s syndrome, and other complaints. Arch Dis Child. 46:745–782. 5. Frasier SD, Costin G, Lippe BM, Aceto T, Bunger PF. 1981 A dose response curve for human growth hormone. J Clin Endocrinol Metab. 53:1213–1217. 6. Sherman BM, Frane J, Johanson AJ, Kaplan SL. 1988 Predictors of response in treatment with methionyl human growth hormone. In: Underwood LE, ed. Human growth hormone: progress and challenges. New York and Basel: Marcel Dekker; 131–142. 7. Blethen SL, Compton P, Lippe BM, Rosenfeld RG, August GP, Johanson A. 1993 Factors predicting the response to growth hormone (GH) therapy in prepubertal children with GH deficiency. J Clin Endocrinol Metab. 76:574 –579. 8. Ranke MB, Lindberg A, Guilbaud O. 1994 Prediction of growth in response to treatment with growth hormone. In: Ranke MB, Gunnarsson R, eds. Progress in growth hormone therapy–5 years of KIGS. Mannheim: J & J Verlag; 97–111. 9. Achermann JC, Hamdani K, Hindmarsh PC, Brook CGD. 1998 Birth weight influences the initial response to growth hormone treatment in growth hormone-insufficient children. Pediatrics 102:342–345. 10. Ranke MB, Lindberg A, Chatelain P, et al. 1999 Derivation and validation of a mathematical model for predicting the response to exogenous recombinant human growth hormone (GH) in pubertal children with idiopathic GH deficiency. J Clin Endocrinol Metab. 84:1174 –1183. 11. Preece MA. 1997 Making a rational diagnosis of growth-hormone deficiency. J Pediatr. 131:S61– 64. 12. Goodman HG, Grumbach MH, Kaplan SL. 1968 Growth and growth hormone. II. A comparison of isolated growth hormone deficiency and multiple pituitary hormone deficiencies in 35 patients with idiopathic hypopituitary dwarfism. N Engl J Med. 278:57– 68. 13. Rivarola MA, Phillips JA III, Migeon CJ, Heinrich JJ, Hjelle BJ. 1984 Phenotypic heterogeneity in familial isolated growth hormone deficiency type I-A. J Clin Endocrinol Metab. 59:34 – 40. 14. Rosenbloom AL, Guevara-Aguirre J, Rosenfeld RG, Pollock BH. 1994 Growth in growth hormone insensitivity. Trends Endocrinol Metab. 5:296 –303. 15. Balsamo A, Tassoni P, Cassio A, et al. 1995 Response to growth hormone therapy in patients with growth hormone deficiency who at birth were small or appropriate in size for gestational age. J Pediatr. 126:474 – 477.
