Division of Medical Sciences (J.A., M.C.S., P.M.S.), University of Birmingham, Queen Elizabeth Hospital, Birmingham B15 2TH, United Kingdom; Department of Postgraduate Medicine (R.N.C.), University of Keele, Hartshill, Stoke-on-Trent, ST4 7QB, United Kingdom; and Department of Diabetes and Endocrinology (A.S.B.), Birmingham Heartlands and Solihull National Health Service Trust, B9 5SS, United Kingdom
Address correspondence to: Prof. P. M. Stewart, Division of Medical Sciences, Queen Elizabeth Hospital, Edgbaston, Birmingham, B15 2TH, United Kingdom. E-mail: p.m.stewart{at}bham.ac.uk.
To the editor:
The paper by Holdaway et al. (1) demonstrates once again the importance of reducing GH hypersecretion to restore mortality rates to those of the background population in patients with acromegaly. As reported in previous studies, including our own from the West Midlands region of the United Kingdom (2, 3), serum GH at last follow-up was the most significant predictor of mortality, with mortality rates reduced to normal by achieving GH concentrations less than 12 µg/liter. However, the conclusion that IGF-I normalization reduces mortality to expected levels is at odds both with our recent data and with the data presented.
The authors acknowledge that the ability to predict the effect of IGF-I on mortality is weakened by the use of different IGF-I assays over the study period. More importantly, there were only 32 patient events (deaths) in the subset of patients for whom IGF-I data were available, reducing the power of analysis even further. This aside, data presented in the paper (see Table 3) appear to show that the percentage of patients with normalized IGF-I values was no different in the survivors or deceased cohort (54 vs. 53%). Furthermore, no statistically significant difference in IGF-I SD scores was observed between these two groups (1.8 ± 3.7 vs. 3 ± 3.1), making it extremely unlikely in our view that IGF-I would have had a significant effect on mortality. This contrasts with the expected changes in GH values that were consistently higher in the deceased group; GH (µg/liter) at last follow-up, 27.8 ± 101 vs. 6.6 ± 21 in survivors and 65% having GH greater than 2 µg/liter at last follow-up vs. 33% of survivors.
Subsequent multivariate analysis of predictors of mortality indicated that serum IGF-I was not an independent predictor of mortality when both GH and IGF-I measurements were included in the analysis. When GH was omitted, serum IGF-I became a significant predictor of mortality. This is not surprising, because there is a well-recognized correlation between GH and IGF-I, although this relationship is linear only up to GH values of approximately 12.5 µg/liter; beyond this point, IGF-I levels plateau, resulting in a curvilinear relationship (4).
In our much larger series of 419 patients with acromegaly, IGF-I data were available in 360 patients (61 deceased) (3). There was no significant increase in mortality in the subgroup of patients with raised posttreatment serum IGF-I levels compared with those with normal posttreatment IGF-I levels. Internal comparison of these groups, controlled for age and sex, revealed a rate ratio of 1.20 (0.712.02; P = 0.50).
The introduction of GH receptor antagonists to treat patients with acromegaly has placed emphasis on using IGF-I rather than GH as the "assay of choice" in assessing underlying disease activity. However, the discordant GH/IGF-I results in treated acromegalic patients (5, 6) together with the complexity of the relationship between GH and IGF-I concentrations in patients with GH deficiency (7) cast doubt on the wisdom of such an approach. Certainly, based on our data [and we would argue those of Holdaway et al. (1)] analyzing mortality as a robust evidence-based outcome, such a move would be premature.
Received March 19, 2004.
References