Letters to the Editor

Compare trial: Updated data

To the Editor:

In 2006, our article ‘‘A randomised trial to compare thepharmacokinetic, pharmacodynamic and antiviral effectsof peginterferon alfa-2b and peginterferon alfa-2a in pa-tients with chronic hepatitis C (COMPARE),” whichwas an 8-week comparison of peginterferon alfa-2a andpeginterferon alfa-2b, was published in the Journal of

Hepatology [1].In COMPARE and other Schering-Plough-supported

trials, Schering-Plough used a proprietary quantitativeTaqMan PCR assay to quantify hepatitis C virus (HCV)-RNA levels. In early 2007, Schering-Plough discoveredthat, in some instances, this assay yielded incorrect HCV-RNA levels. The COMPARE study was affected by theassay inaccuracies, and certain samples from patientsenrolled in these studies required reanalysis. Samples iden-tified as having been measured incorrectly have since beenreanalyzed, and we can confirm that the overall conclusionsreported in our published article remain unchanged.

Although the overall conclusions of this publicationremain unchanged, some of the data were adjusted.Briefly, all values presented in Table 3 in the article wereadjusted slightly; however, the adjustments did not resultin any change in the statistical significance betweenpeginterferon alfa-2a and peginterferon alfa-2b. Thereanalysis data indicate that the average decrease inHCV-RNA and the mean maximum decrease in HCV-RNA are significantly greater among patients receivingpeginterferon alfa-2b than among those receiving pegin-terferon alfa-2a at week 1 (P < .05 for both comparisons)and week 4 (P < .0001 for both comparisons). In addi-tion, at week 4, the number of patients receiving peginter-feron alfa-2a is reduced from 18 to 17, and the numberreceiving peginterferon alfa-2b is reduced from 18 to 16.

Peginterferon alfa-2a versus peginterferon alfa-2b datadid not change, but correlation coefficients and probabili-ties for the week 8 carry-forward response (W8cfresp) datafor cyclin D, IP10, ISG15, ISG54, perforin, PKR, signaltransducer and activator of transcription (STAT)-1,STAT-2, and 2050OAS were affected (Table 4 in the article).For each of these transcripts, the reanalysis indicated a re-duced Spearman correlation coefficient compared with the

original data. Similar changes were noted for the W8cfrespdata in Table 5 in the article, resulting in a reduced correla-tion coefficient for principle component (Prin) 1 and Prin2,an increased correlation coefficient for Prin3, and a nega-tive correlation at Prin4 and Prin5 compared with the ori-ginal data. The reanalysis indicated that the differences ingene transcript levels between responders and non-responders for inducible protein 10 (IP10) and RNA-dependent protein kinase (PKR) (maximum fold increasefrom baseline; Cmax) and IP10, PKR, and interferon-stim-ulated gene (ISG) 54 (area under the curve; AUC) re-mained statistically significant (P < .05) (Table 2 in thearticle). The differences in ISG15, 2050 oligoadenylate syn-thetase (OAS), and ISG54 (Cmax) and ISG15 and 2050

OAS (AUC) between responders and non-responders thatwere statistically significant in the original analysis were nolonger significant (Table 2 in the article).

All the data presented in Figs. 4A and B in the articlewere slightly affected. For peginterferon alfa-2a, the dif-ference in mean AUC of IP10 between responders andnon-responders remained statistically significant; how-ever, for peginterferon alfa-2b, the difference in meanAUC for ISG54 between responders and non-respondersfailed to reach statistical significance using the reanalysisdata (Fig. 4A). In the reanalysis, a patient previously con-sidered a responder was reclassified as a non-responder: inthe reanalysis, 9 of 14 patients were responders and 5 werenon-responders (Fig. 4B in the article). Data presented inFig. 5 in the article were slightly affected; however, themean maximal log10 change in HCV-RNA from baselineremained significantly greater among patients receivingpeginterferon alfa-2b than among those receiving pegin-terferon alfa-2a at week 1 and week 4 (P < .05) for bothcomparisons. Finally, the data presented in Fig. 6 in thearticle were also slightly affected; however, the rate of de-cline in HCV-RNA during the 8 weeks of treatment re-mained significantly greater among patients receivingpeginterferon alfa-2b than among patients receivingpeginterferon alfa-2a.

As stated, the adjustments noted above did notchange the overall conclusions reported in our publishedarticle. However, we believe that full disclosure of the

www.elsevier.com/locate/jhep

Journal of Hepatology 49 (2008) 288–294

assay inaccuracies, the subsequent reanalysis, andresults of the reanalysis were imperative to communicateto the medical community.

Appendix A. Supplementary data

Supplementary data associated with this article canbe found, in the online version, at doi:10.1016/j.jhep.2008.05.009.

Reference

[1] Silva M, Poo J, Wagner F, Jackson M, Cutler D, Grace M, et al. Arandomised trial to compare the pharmacokinetic, pharmacody-

namic, and antiviral effects of peginterferon alfa-2b and peginter-feron alfa-2a in patients with chronic hepatitis C (COMPARE). JHepatol 2006;45:204–213.

Marcelo O. SilvaHead, Liver and Liver Transplant Unit,

Hospital Universitario Austral,A.J.D. Peron 1500,

Pilar (B1629),

Buenos Aires, Argentina

Tel.: +54 2322 482884E-mail address: msilva@cas.austral.edu.ar

doi:10.1016/j.jhep.2008.05.009

Spanish reports of hepatotoxicity associated with Herbalife� products

To the Editor:

We have read with interest the articles recently pub-lished in this journal by Elinav et al. [1] and Schoepferet al. [2], as well as the editorial by Stickel [3], reportingand discussing two series of cases from Israel and Swit-zerland of hepatotoxicity associated with the consump-tion of Herbalife� slimming products. Stickel [3] raisesthe question of why products which are distributed inat least 60 countries only seem to induce hepatotoxicityin two.

In Spain, as perhaps in other countries, Herbalife�

distributes its products through door-to-door salesmenand the internet. Some products are registered as dieteticsupplements, but others are not registered at all. In2005, three cases of hepatotoxicity (two hepatitis andone of increased liver enzymes) were reported to thePharmacovigilance Centre of Asturias, an area in thenorth of Spain. Another case of increased liver enzymeswas reported in 2006. All the four cases occurred in wo-men between 47 and 57 years old, and all came from thesame hospital. Three of these were published last Febru-ary in a Spanish medical journal [4].

In the two cases of hepatitis the patients were sisters.One of whom developed severe liver damage. After tak-ing Herbalife� products for 1 year and losing 20 kg inweight, she was referred to hospital suffering from wide-spread pruritus, jaundice, fatigue and abdominal strain.Biochemical tests showed liver transaminases over1000 U/l and 12 mg/dl of total bilirubin. Test antibodiesto hepatitis were negative. Abdominal ultrasoundshowed a cholelithiasis that was surgically removed.Although she was recommended to cease taking Herba-life� products, she continued to do so for another 2years and developed chronic liver disease with recurrentexacerbations. Liver biopsy showed acute hepatitisgrade 4 on chronic liver disease stage 3. She was diag-

nosed as having idiopathic or toxic autoimmune hepati-tis, and is currently being treated with corticosteroids.Her sister took Herbalife� products for six monthsand developed diarrhea, hyperbilirubinemia, jaundiceand pruritus. Biochemical tests showed transaminasesover 1000 U/l and 26.7 mg/dl of total bilirubin. Viralantibodies were negative. Subacute cholestatic hepatitiswas diagnosed and the withdrawal of Herbalife� prod-ucts was recommended. Some months later, the patienthad totally recovered. The two sisters took the followingHerbalife� products: formulas 1, 2, 3 and 4, RoseOx,Herbalifeline, Guarana, Classic Aromatized Tea andHerbal Aloe.

The patients in the other two cases of reported hepa-totoxicity were also taking Herbalife� products over aperiod of 1 month in one case and 3 years in the other,and developed increased transaminases which were de-tected on routine analysis. After the withdrawal of Her-balife� products, the enzyme levels returned to normal.

To explain the mechanism involved in the hepatotox-icity induced by Herbalife� products, Elinav et al. [1]have suggested genetic susceptibility to develop an im-mune-mediated liver toxicity associated with one ormore constituents. The family ties between the two casesof hepatitis reported in Spain and the diagnosis of auto-immune hepatitis in one case support this observation.We agree with Stickel [3] when he explains that it is ex-tremely difficult to identify the crucial compounds whenconsumers of Herbalife� are taking several products.Although toxic or bacteriological contamination ofsome batches cannot be dismissed, we believe that somecomponents of Herbalife� products could be the hepa-toxicity inducers. Firstly, as Schoepfer et al. [2] havesuggested, the green tea contained in the Classic Aroma-tized Tea and also in other Herbalife� products. In thelast 5 years, evidence on the ability of green tea to induce

Letters to the Editor / Journal of Hepatology 49 (2008) 288–294 289