HIV-1 resistance testing has become the standard-of-care for managing treatment-experienced patients and is increasingly recommended for treatment-naïve patients. The two standard methods for resistance testing are genotyping, which is generally used in treatment-naïve patients and where relatively simple resistance patterns are expected, and phenotyping, which is more commonly used in patients with complex patterns of resistance. Genotyping can be more sensitive to minority variants, which may impact subsequent virologic response, while phenotyping measures susceptibility directly and thus does not require interpretation of complex genotypic determinants.

A third approach is the hybrid VirtualPhenotype™-LM (VPT-LM) bio-informatics engine, which calculates a fold change (FC) from a genotype based on analyses of paired genotype-phenotype data for >59,000 clinical samples. Resistance is then interpreted in the virco®TYPE HIV-1 report, based on, for most drugs, clinical cutoffs (CCOs) that are calculated using virological response data from patients receiving the drugs. The virco®TYPE HIV-1 report thus combines the advantages of phenotyping and genotyping, to provide clinicians with reliable predictions of clinical susceptibility - that is, the proportion of activity remaining for each drug.

Two recent reports have demonstrated particular strengths of the virco®TYPE HIV-1system. A report from Margriet van Houtte and colleagues* in the October 2009 Journal of Medical Virology highlighted the strong correlations between FC predictions made using VPT-LM with FCs measured in an independent laboratory phenotyping system. A second report, from Bart Winters and colleagues,** published online August 3, 2009 in the Journal of Virological Methods, highlighted recent enhancements to the virco®TYPE HIV-1 report including CCOs based on a much larger clinical data set and an important change in the handling of mixed viral sequences by VPT-LM that improves sensitivity to minority variants. 

The report by van Houtte et al assessed concordance between FC values calculated using VPT-LM and FC values measured in the PhenoSense™ (PS) phenotypic assay for 287 to 902 records (depending on the drug) from the Stanford HIV database. FC values for the assays were closely correlated, with a mean correlation coefficient of 0.90 using single PS measurements and 0.94 when VPT-LM results were compared to the mean of repeated PS measurements. These results were comparable to the corresponding correlation coefficients of 0.87 and 0.95, respectively, for measurements obtained in the Antivirogram® phenotyping assay on which VPT-LM is based. Importantly, the version of VPT-LM used in these analyses addressed mixtures by summing partial values for each of the mutations detected at a given position, not unlike a phenotypic assay measuring the average impact of the various mutants across a mixed viral population.

In the report from Winters et al, CCOs for the virco®TYPE HIV-1 report were updated to include analyses of clinical response data for 6550 patients with 2299 records reserved for validation. New data were added from 2 large cohort studies as well as the TITAN, POWER, and DUET clinical studies, yielding a total of approximately triple the number of records previously available. The updated CCOs were generally close to the previous values, with marginally higher cutoffs for some NRTIs. The new CCOs were typically equal to or marginally better than the previous cutoffs in predicting response, and resistance calls for clinical samples remained relatively stable between CCO versions. Notably, CCOs were derived for several additional agents, and are now available for all protease inhibitors and for all RT inhibitors except efavirenz, nevirapine and emtricitabine.

Winters and colleagues also described a new "Worst-Case Scenario" (WCS) approach to calculating resistance when mixtures are detected in a genotype. Previously, and in the report from van Houtte et al described above, when 2 amino acids were observed at a position, the resistance value was defined as the sum of 1/2 of the weight factor for each residue; similarly for mixtures with 3 (1/3 each) or 4 (1/4 each) residues. WCS, in contrast, incorporates the full value of only the residue that is associated with the lowest susceptibility to a given drug. This new approach departs from creating the most direct correlation with a measured phenotype and instead conservatively assumes the selection of the more resistant quasispecies from the test sample during future therapy. It thus focuses on predicting the susceptibility to the drug in a future treatment regimen, rather than making the most precise assessment of susceptibility in the actual test sample. In the report from Winters and colleagues, WCS generally resulted in relatively modest increases in FC values, reflecting the fact that only positions with mixtures were impacted.

Assessing the simultaneous impact of the updated CCOs and the WCS approach, Winters et al found that in either the development or validation datasets, odds ratios for predicting virologic response for a given CCO category using the new virco®TYPE HIV-1 version were generally higher than or equivalent to those based on the previous version. The impact on resistance calls was also assessed for >50,000 recent clinical samples received from December 1, 2006 through December 1, 2008. Changes between fully sensitive and resistant were uncommon, (0% and <0.15% for any drug), while the incidence of changes to or from intermediate susceptibility was low to moderate for most drugs. Changes for >2% of samples to a more sensitive class were seen for nelfinavir (10.3%), indinavir/r (5.6%), and saquinavir/r (3.3%). Shifts for >2% of samples to a more resistant category occurred for abacavir (8.7%), stavudine (6.2%), lamivudine (5.0%) and zidovudine (3.3%).

These data highlight several important points: First, high concordance between the vircoTYPE®_HIV-1 calculated fold change values and the results reported by a conventional measured phenotype can be obtained using the VPT-LM bio-informatics engine. Second, the cutoffs for resistance, partial resistance or sensitivity in the virco®TYPE HIV-1 report are now based on virological response data for all protease and most RT inhibitors, and these CCOs are based on robust analysis and validation datasets. Finally, a new “Worst-Case Scenario” approach has been introduced for assessing mixed sequences that uses to full advantage the sensitivity of genotypic testing when making predictions of future response. The virco®TYPE HIV-1 report thus continues to use the latest clinical research to provide clinicians with reliable, up-to-date assessments of antiretroviral resistance and sensitive predictions of residual response to drugs as used in future treatment regimens.
 

* Margriet Van Houtte,1 Gaston Picchio,2* Koen Van Der Borght1, Theresa Pattery,1, Pierre Lecocq,1 and Lee T. Bacheler3

**Bart Winters1, Elke Van Craenenbroeck1, Koen Van der Borght1, Pierre Lecocq1, Jorge Villacian1, Lee Bacheler3

1Virco BVBA, Mechelen, Belgium
2Virco Lab, Inc. Raritan, NJ, USA
3VircoLab, Inc., 101 Glenview Place, Chapel Hill, NC 27514, USA

DISCLAIMER: The Van Houtte study reviewed in this clinical update did not use a worst-case analysis of viral mixtures when comparing measured and predicted FC values