To the Editor:
We recently discovered that an error had been made in handling the data in our study on the use of recombinant human growth hormone (GH) to reduce visceral adipose tissue (VAT) in HIV-associated adipose redistribution syndrome (HARS) that appeared in the Journal of Acquired Immune Deficiency Syndromes (2004).1 This error resulted in reporting incorrect data for baseline and change values for VAT and abdominal subcutaneous adipose tissue (SAT), as described below. Other study outcomes and overall conclusions were not affected.
In the double-blind trial,1 HIV-infected patients with abdominal adiposity were randomized to receive either placebo (PL) or 4 mg GH, dosed daily (DD) or on alternate days (AD), for weeks 1-12. From week 12 to 24, half of the patients who received 4 mg DD were rerandomized to receive placebo (the DD-PL group), and half to receive GH 4 mg AD (the DD-AD group); the other groups were reassigned to receive either GH 4 mg AD (AD-AD group) or GH 4 mg DD (the PL-DD group).
Cross-sectional areas of VAT and abdominal SAT were measured by single-slice abdominal computed tomography (CT) scans at the level of the L4-5 intervertebral disk space at baseline, week 12, and week 24. Radiologists working at a central site determined values of parameters from the scans. The radiologists were unaware of the study treatment and sequence in which scans were obtained. The central reading site provided a listing of the data to statisticians at a separate site that listed all parameters assessed by the scans.
The errors in the article resulted from the fact that incorrect parameters for VAT and abdominal SAT were inadvertently chosen from the listing that was provided by the central reading site and used in the analyses. The CT scan parameters that should have been selected were the muscle-free, organ-free, bone-free measures of VAT and abdominal SAT. The CT scan parameters that were inadvertently chosen included not only the area for fat in the region where VAT and SAT were imaged but also total pixels that included muscle, organ, and bone in those regions. Correct baseline data by sex are shown in Table 1 with the data that were previously reported (in Table 2 of the original article1) also shown in the table.
The erroneously published data for changes in VAT and SAT by treatment group at 12 weeks (as in Table 3 of the original article1), and the correct data for these changes, are shown in Table 2.
In the article,1 it was stated that, from baseline to week 12, VAT decreased significantly in the DD group (−8.6%, P < 0.001 as compared to PL) but did not decrease significantly in the AD group (−4.2%, P = 0.052 as compared to PL). Upon reanalysis, using the correct 12-week data, significant (P < 0.001) reductions in VAT from baseline were observed in both the DD and AD groups (−18.8% and −16.8%, respectively) as compared to the PL group, which exhibited no significant change in VAT from baseline.
The article1 also reported that in patients who first received GH daily for 12 weeks and then received PL during weeks 12-24 (the DD-PL group), VAT did not significantly reaccumulate during weeks 12-24, although trunk fat by DXA significantly reaccumulated during that period. On reanalysis using the correct VAT data, it was found that VAT did significantly reaccumulate during weeks 12-24 (by 17.7 cm2, P = 0.012) in the DD-PL group. The finding that VAT did reaccumulate after stopping GH in this study is consistent with the findings of Engelson et al in a previously reported trial.2
In the DD-PL patients, VAT at week 24 (156.4 cm2) was not significantly different (P = 0.273) from their mean VAT at baseline (164.5 cm2). However, significant (P ≤ 0.001) losses of VAT from baseline were observed at week 24 in the other three treatment groups (PL-DD, AD-AD, and DD-AD), as shown in Figure 1 (which is the corrected version of Fig. 3 in the original article1).
At week 24, the corrected overall percentage reductions in VAT from baseline were −9.2%, −10.7%, −1.6%, and −21.4% in the PL-DD, AD-AD, DD-PL, and DD-AD groups, respectively, rather than −8.1%, −5.4%, −5.3%, and −9.5%, as previously reported.1
On reanalysis of the changes in area of abdominal SAT on CT scan from baseline to week 24, the DD-AD group still showed a significant (P < 0.05) change in the mean value of this variable (but to 132.8 cm2 at week 24 from 151.3 cm2 at baseline, P = 0.0019, rather than to 152.3 cm2 at week 24 from a baseline of 173.5 cm2, as previously reported1). There were still no significant reductions in mean abdominal SAT in other groups from baseline to week 24, and reanalysis revealed no significant differences in the percentage change in SAT from baseline to week 24 in any treatment group (−2.5%, +0.6%, +3.3%, and +9.4% in AD-AD, PL-DD, or DD-PL, respectively, with P values ranging from 0.099 to 0.566). Previously, at week 24, a significant (P = 0.01) percentage decrease in SAT of −5.9% from baseline had been observed in the DD-AD group, while no significant percentage SAT changes from baseline were seen in the other groups (−1.8%, −0.2%, and +2.7% in AD-AD, PL-DD, or DD-PL, respectively) at week 24.1
In addition, the units for IGF-I in the paper should have been expressed in ng/mL, not ng/dL, and the footnote to the original Table 2 referring to “mean (SD)” should have stated “mean (SEM).”
In summary, the original article1 overestimated the baseline VAT and SAT. On the other hand, it underestimated percent reductions in VAT and SAT on GH therapy. It also incorrectly stated that VAT and SAT reduction were statistically significant only in the group treated with GH 4 mg/day, not in the AD group, and that VAT did not significantly reaccumulate in subjects after GH treatment ended. The revised analyses show that mean VAT and SAT for subjects at baseline were less than originally reported, that reductions in VAT and SAT were significant at 12 weeks on both GH regimens (AD and DD), and that VAT (but not SAT) reaccumulated significantly after cessation of active treatment. IGF-I should have been expressed in ng/dL. Please note that the main conclusions of the original report remain the same, namely, that a 12-week course of GH dosed at 4 mg/day significantly reduced VAT, trunk fat, dorsocervical fat, non-HDL cholesterol, and related cardiovascular risk parameters in HIV patients with excess abdominal fat and that AD therapy had smaller and fewer significant effects on most of these parameters.
Donald P. Kotler, MD*
Norma Muurahainen, MD, PhD†
Carl Grunfeld, MD, PhD‡
Christine Wanke, MD§
Melanie Thompson, MD‖
Michael Saag, MD¶
Gregg Simons, PhD#
Joseph M. Gertner, MRCP**
*Division of GI Immunology St. Luke's Roosevelt Hospital Center, New York, NY
†Serono, Inc. Rockland, MA
‡UCSF VA San Francisco, CA
§Tufts-NEMC Boston, MA
‖AIDS Research Consortium Atlanta, GA
¶UAB Birmingham, AL
#Quality Clinical Research Hall, MA
**The GI Company Framingham, MA