icon-folder.gif   Conference Reports for NATAP  
  International HIV and Hepatitis Drug Resistance Workshop
June 8-12, 2010,
Dubrovnik Croatia
Back grey_arrow_rt.gif
Debate Follows Data on Genotypic Predictions of HIV Tropism
  International HIV and Hepatitis Drug Resistance Workshop, June 8-12, 2010, Dubrovnik, Croatia
Mark Mascolini
A comparison of genotypic results on HIV-1 use of CCR5 or CXCR4 receptors found nearly 90% agreement among nine independent European Union (EU) labs and a reference center in Cologne [1]. The results add to evidence that genotyping may be a reliable way to determine whether a CCR5 antagonist like maraviroc will work in individual patients. But several attendees at this Resistance Workshop expressed concern about how convincing this and other coreceptor genotyping studies have been.
The nine EU labs used the geno2pheno coreceptor tool to predict coreceptor use in viral samples of 166 patients randomly selected from the labs. Technicians analyzed the samples with geno2pheno false-positive rates set at 5% and 10%. EU guidelines recommend a false-positive rate cutoff of 5% to ensure safe use of maraviroc. The investigators also sent 96 of the 166 samples for phenotypic Trofile testing.
The labs successfully analyzed 118 of 166 samples (71%), 94 of them HIV-1 subtype B and the other 24 non-B subtypes, including 11 C subtypes. With a geno2pheno false-positive rate setting of 10%, the nine labs called 87 of 118 sequences R5 and 31 X4. The Cologne reference center also predicted R5 use in 87 of 118 sequences and X4 use in 31. Overall agreement between the nine labs and the Cologne center was 104 of 118 (88%). Results proved concordant for 83 of 94 subtype B samples (88%) and 21 of 24 non-Bs (87.5%).
With a false-positive rate setting of 5%, the nine EU labs called 94 of 118 sequences R5 and 24 X4. The Cologne reference center classified 98 of 118 as R5 and 20 as X4. Overall concordance was again 88% (104 of 118 sequences). Coreceptor use results were concordant in 81 of 94 subtype B samples (86%) and 23 of 24 non-B samples (96%).
Of the 96 viral isolates sent for Trofile phenotyping, 66 (69%) corresponded to isolates that both the reference lab and the individual labs could predict. Fifty-seven of these 66 samples (86%) were samples that got the same coreceptor call in Cologne and the nine labs. Trofile results were the same as genotyping results in 48 of 57 samples (84%) with geno2pheno at the 5% false-positive rate and in 46 of 57 samples (81%) with geno2pheno at the 10% rate.
All told, 48 of 66 samples (73%) yielded the same result with Trofile, in Cologne, and in the individual EU labs with geno2pheno at the 5% false-positive rate, as were 46 of 66 (70%) at the 10% false-positive rate.
The investigators noted that low viral loads did not explain genotypic discrepancies between labs, so they suggested that differences in primers and procedures could explain the variation. They stressed that geno2pheno interprets poor-quality sequences as CXCR4 use to avoid false prediction of CCR5 use and inappropriate prescription of maraviroc.
In a roundtable after this presentation and others on genotyping for coreceptor use, more than one Resistance Workshop attendee voiced doubts about the strength of data presented so far. The basic problem, voiced by Douglas Richman (University of California, San Diego) and others, is that 12% discordance between labs using the same genotyping system does not inspire confidence, especially when added to the 29% of viral samples that could not be analyzed at all.
1. Sierra S, Balduin M, Zazzi M, et al. Pairwise comparison of genotypic tropism prediction between 1 reference centre and 10 European laboratories. International HIV and Hepatitis Drug Resistance Workshop. June 8-12, 2010. Dubrovnik. Croatia. Abstract 14.