Variety of interpretation systems for human immunodeficiency virus type 1 genotyping: confirmatory information or additional confusion?

Sequence quality analysis tool for HIV type 1 protease and reverse transcriptase

Access to antiretroviral therapy is increasing globally and drug resistance evolution is anticipated. Currently, protease (PR) and reverse transcriptase (RT) sequence generation is increasing, including the use of in-house sequencing assays, and quality assessment prior to sequence analysis is essential. We created a computational HIV PR/RT Sequence Quality Analysis Tool (SQUAT) that runs in the R statistical environment. Sequence quality thresholds are calculated from a large dataset (46,802 PR and 44,432 RT sequences) from the published literature ( http://hivdb.Stanford.edu ). Nucleic acid sequences are read into SQUAT, identified, aligned, and translated. Nucleic acid sequences are flagged if with >five 1-2-base insertions; >one 3-base insertion; >one deletion; >six PR or >18 RT ambiguous bases; >three consecutive PR or >four RT nucleic acid mutations; >zero stop codons; >three PR or >six RT ambiguous amino acids; >three consecutive PR or >four RT amino acid mutations; >zero unique amino acids; or <0.5% or >15% genetic distance from another submitted sequence. Thresholds are user modifiable. SQUAT output includes a summary report with detailed comments for troubleshooting of flagged sequences, histograms of pairwise genetic distances, neighbor joining phylogenetic trees, and aligned nucleic and amino acid sequences. SQUAT is a stand-alone, free, web-independent tool to ensure use of high-quality HIV PR/RT sequences in interpretation and reporting of drug resistance, while increasing awareness and expertise and facilitating troubleshooting of potentially problematic sequences.

Description of two commercially available assays for genotyping of HIV-1

HIV-1 resistance testing is one important part in the diagnostics of antiretroviral treatment and is commonly done by genotyping. Currently, two systems are commercially available and, despite being far from easy to use, these have achieved a high degree of sophistication. Modifications of standard kit protocols might be necessary based on the clinical situation. Although resistance reports based on decision rules are a part of both systems, considerable knowledge and skills are nevertheless required by the user to establish useful clinical data out of detected resistance patterns. Both systems described here have their advantages and disadvantages; a decision for one or the other system needs to be based on individual requirements. The future might lie in so-called 'next-generation sequencing' systems based on pyrosequencing, which enable a high throughput and the detection of minor variants of less than 1%.

The development of an expert system to predict virological response to HIV therapy as part of an online treatment support tool

OBJECTIVE

The optimum selection and sequencing of combination antiretroviral therapy to maintain viral suppression can be challenging. The HIV Resistance Response Database Initiative has pioneered the development of computational models that predict the virological response to drug combinations. Here we describe the development and testing of random forest models to power an online treatment selection tool.

METHODS

Five thousand, seven hundred and fifty-two treatment change episodes were selected to train a committee of 10 models to predict the probability of virological response to a new regimen. The input variables were antiretroviral treatment history, baseline CD4 cell count, viral load and genotype, drugs in the new regimen, time from treatment change to follow-up and follow-up viral load values. The models were assessed during cross-validation and with an independent set of 50 treatment change episodes by plotting receiver-operator characteristic curves and their performance compared with genotypic sensitivity scores from rules-based genotype interpretation systems.

RESULTS

The models achieved an area under the curve during cross-validation of 0.77-0.87 (mean = 0.82), accuracy of 72-81% (mean = 77%), sensitivity of 62-80% (mean = 67%) and specificity of 75-89% (mean = 81%). When tested with the 50 test cases, the area under the curve was 0.70-0.88, accuracy 64-82%, sensitivity 62-80% and specificity 68-95%. The genotypic sensitivity scores achieved an area under the curve of 0.51-0.52, overall accuracy of 54-56%, sensitivity of 43-64% and specificity of 41-73%.

CONCLUSION

The models achieved a consistent, high level of accuracy in predicting treatment responses, which was markedly superior to that of genotypic sensitivity scores. The models are being used to power an experimental system now available via the Internet.

Clinical evaluation of the potential utility of computational modeling as an HIV treatment selection tool by physicians with considerable HIV experience

The HIV Resistance Response Database Initiative (RDI), which comprises a small research team in the United Kingdom and collaborating clinical centers in more than 15 countries, has used antiretroviral treatment and response data from thousands of patients around the world to develop computational models that are highly predictive of virologic response. The potential utility of such models as a tool for assisting treatment selection was assessed in two clinical pilot studies: a prospective study in Canada and Italy, which was terminated early because of the availability of new drugs not covered by the system, and a retrospective study in the United States. For these studies, a Web-based user interface was constructed to provide access to the models. Participating physicians entered baseline data for cases of treatment failure and then registered their treatment intention. They then received a report listing the five alternative regimens that the models predicted would be most effective plus their own selection, ranked in order of predicted virologic response. The physicians then entered their final treatment decision. Twenty-three physicians entered 114 cases (75 unique cases with 39 entered twice by different physicians). Overall, 33% of treatment decisions were changed following review of the report. The final treatment decisions and the best of the RDI alternatives were predicted to produce greater virologic responses and involve fewer drugs than the original selections. Most physicians found the system easy to use and understand. All but one indicated they would use the system if it were available, particularly for highly treatment-experienced cases with challenging resistance profiles. Despite limitations, the first clinical evaluation of this approach by physicians with substantial HIV-experience suggests that it has the potential to deliver clinical and economic benefits.

Transmitted drug resistance in nonsubtype B HIV-1 infection

HIV-1 nonsubtype B variants account for the majority of HIV infections worldwide. Drug resistance in individuals who have never undergone antiretroviral therapy can lead to early failure and limited treatment options and, therefore, is an important concern. Evaluation of reported transmitted drug resistance (TDR) is challenging owing to varying definitions and study designs, and is further complicated by HIV-1 subtype diversity. In this article, we discuss the importance of various mutation lists for TDR definition, summarize TDR in nonsubtype B HIV-1 and highlight TDR reporting and interpreting challenges in the context of HIV-1 diversity. When examined carefully, TDR in HIV-1 non-B protease and reverse transcriptase is still relatively low in most regions. Whether it will increase with time and therapy access, as observed in subtype-B-predominant regions, remains to be determined.

A comparison of three computational modelling methods for the prediction of virological response to combination HIV therapy

OBJECTIVE

HIV treatment failure is commonly associated with drug resistance and the selection of a new regimen is often guided by genotypic resistance testing. The interpretation of complex genotypic data poses a major challenge. We have developed artificial neural network (ANN) models that predict virological response to therapy from HIV genotype and other clinical information. Here we compare the accuracy of ANN with alternative modelling methodologies, random forests (RF) and support vector machines (SVM).

METHODS

Data from 1204 treatment change episodes (TCEs) were identified from the HIV Resistance Response Database Initiative (RDI) database and partitioned at random into a training set of 1154 and a test set of 50. The training set was then partitioned using an L-cross (L=10 in this study) validation scheme for training individual computational models. Seventy six input variables were used for training the models: 55 baseline genotype mutations; the 14 potential drugs in the new treatment regimen; four treatment history variables; baseline viral load; CD4 count and time to follow-up viral load. The output variable was follow-up viral load. Performance was evaluated in terms of the correlations and absolute differences between the individual models' predictions and the actual DeltaVL values.

RESULTS

The correlations (r(2)) between predicted and actual DeltaVL varied from 0.318 to 0.546 for ANN, 0.590 to 0.751 for RF and 0.300 to 0.720 for SVM. The mean absolute differences varied from 0.677 to 0.903 for ANN, 0.494 to 0.644 for RF and 0.500 to 0.790 for SVM. ANN models were significantly inferior to RF and SVM models. The predictions of the ANN, RF and SVM committees all correlated highly significantly with the actual DeltaVL of the independent test TCEs, producing r(2) values of 0.689, 0.707 and 0.620, respectively. The mean absolute differences were 0.543, 0.600 and 0.607log(10)copies/ml for ANN, RF and SVM, respectively. There were no statistically significant differences between the three committees. Combining the committees' outputs improved correlations between predicted and actual virological responses. The combination of all three committees gave a correlation of r(2)=0.728. The mean absolute differences followed a similar pattern.

CONCLUSIONS

RF and SVM models can produce predictions of virological response to HIV treatment that are comparable in accuracy to a committee of ANN models. Combining the predictions of different models improves their accuracy somewhat. This approach has potential as a future clinical tool and a combination of ANN and RF models is being taken forward for clinical evaluation.

HIV type 1 drug resistance in adults receiving highly active antiretroviral therapy in Abidjan, Côte d'Ivoire

As antiretroviral therapy continues to scale-up in developing countries, there is concern that high levels of HIV drug resistance to antiretroviral drugs will occur. Here we describe rates of emergence of HIV-1 drug resistance and factors associated with their occurrence among adults who received antiretroviral therapy (ART) for >1 year through the Côte d'Ivoire national drug access program from 1998 to 2003. To detect genotypic drug resistance, we sequenced all 1- and 2-year specimens with detectable HIV RNA viral load. To assess factors associated with emerging drug resistance, we used log normal regression with interval censoring, including covariates in the model for self-reported drug adherence, CD4 cell count, and HIV viral load at therapy initiation, and observed changes in these measures, type of prescribed ART drugs, diagnoses of opportunistic illness, and demographic characteristics. An estimated 14.2% [95% confidence limits (CL) 11.7, 16.9] and 26.6% (95% CL 22.7, 30.8) of patients developed primary drug-resistant mutations within 1 year and 2 years after initiation of therapy, respectively. Factors associated with drug resistance included drug nonadherence, partial or lack of viral suppression, higher viral load or lower CD4 at initiation of therapy, and initiation of ART with what is now considered substandard dual combination therapy. Our results demonstrate the need to strengthen adherence and continuity in treatment programs in order to avoid interruption of ART drugs. Treatment programs should pay attention to indicators of emerging drug resistance: incomplete or lesser decreases in viral load or increases in CD4 cell counts following initiation of therapy, and the occurrence of AIDS opportunistic illnesses.

Molecular assays for monitoring HIV infection and antiretroviral therapy

Infection with HIV results in lifelong persistence of the virus in the body of infected persons, independent of antiretroviral treatment. Therefore, efficient and meaningful therapy monitoring has been developed since its introduction in the 1980s. Whereas, primarily, the measurement of the CD4 cell count was the most important clinical marker of disease progression, nowadays the estimation of plasma viral load with molecular methods plays a major role as a marker of therapy success. To optimize therapy changes in patients failing on antiretroviral therapy regimen, HIV-1 genotyping has been introduced and is now widely accepted as an additional diagnostic tool. Due to this increase in diagnostic parameters, clinicians and virologists have to cope with many different methods. This review should give a brief overview of the current commercially available assays for detection and quantification of HIV, as well as for HIV-1 genotypic resistance testing. Quantitative reverse transcriptase PCR, real-time PCR, nucleic acid sequence-based amplification and the branched DNA system are described in detail, and the advantages and disadvantages are discussed. In addition, two commercially available HIV-1 genotyping assays are compared. However, a general recommendation to favor one system over the other cannot be given, because the final decision of which system to use should be decided on the individual requirements.

RB-finder: an improved distance-based sliding window method to detect recombination breakpoints

Recombination detection is important before inferring phylogenetic relationships. This will eventually lead to a better understanding of pathogen evolution, more accurate genotyping, and advancements in vaccine development. In this paper, we introduce RB-Finder, a fast and accurate distance-based window method to detect recombination in a multiple sequence alignment. Our method introduces a more informative distance measure and a novel weighting strategy to reduce the window size sensitivity problem and hence improve the accuracy of breakpoint detection. Furthermore, our method is faster than existing phylogeny-based methods since we do not need to construct and compare complex phylogenetic trees. When compared with the current best method Pruned-PDM, our method is a few hundred times more efficient. Experimental evaluation of RB-Finder using synthetic and biological datasets showed that our method is more accurate than existing phylogeny-based methods. We also show how our method has potential use in other related applications such as genotyping.

A study of seven rule-based algorithms for the interpretation of HIV-1 genotypic resistance data in Thailand

Since the free therapy program was started by the Thai government, the number of patients infected by HIV-1 with access to antiretroviral drugs has increased. The selection of effective interpretation algorithms for antiretroviral drug resistance has become even more important for clinical management. In this retrospective study, the level of agreement was evaluated in 721 antiretroviral-therapy failing HIV-1 subjects. Regarding genetic diversity, about 89% was recognized as non-B variants (CRF01_AE). The level of complete concordant interpretation score in all seven algorithms was recognized in non-nucleoside reverse transcriptase inhibitors (NNRTIs) and protease inhibitors (PIs) (67%), but not in nucleoside reverse transcriptase inhibitors (NRTIs) (52%). Over 10% of the major discordance score with TRUGENE was revealed in didanosine (Agence Nationale de Recherches sur le SIDA[ANRS]; Detroit Medical Centre [DMC]), abacavir (ANRS; Centre Hospitalier de Luxembourg [CHL]), and also with delavirdine, indinavir and amprenavir (Grupo de Aconselhamento Virológico [GAV]). A good to excellent agreement range of kappa scores was detected for most antiretroviral drugs. However, poor agreement with the TRUGENE system (k<0.40) was seen in the ANRS system with didanosine, abacavir and lopinavir; GAV system in indinavir and amprenavir; and DMC system in ritonavir. These might be an option for resource limited countries when selecting the use of a low cost or free algorithm interpretation, which has excellent agreement as the U.S. Food and Drug Administration (FDA)-approved TRUGENE commercial system.

The Development of Artificial Neural Networks to Predict Virological response to Combination HIV Therapy

Introduction

When used in combination, antiretroviral drugs are highly effective for suppressing HIV replication. Nevertheless, treatment failure commonly occurs and is generally associated with viral drug resistance. The choice of an alternative regimen may be guided by a drug-resistance test. However, interpretation of resistance from genotypic data poses a major challenge.

Methods

As an alternative to current interpretation systems, we have developed artificial neural network (ANN) models to predict virological response to combination therapy from HIV genotype and other clinical information.

Results

ANN models trained with genotype, baseline viral load and time to follow-up viral load (1,154 treatment change episodes from multiple clinics), produced predictions of virological response that were highly significantly correlated with actual responses (r2=0.53; P<0.00001) using independent test data from clinics that contributed training data. Augmented models, trained with the additional variables of baseline CD4+ T-cell count and four treatment history variables, were more accurate, explaining 69% of the variance in virological response. Models trained with the full input dataset, but only those data involving highly active antiretroviral therapy (three or more full-dose antiretroviral drugs in combination), performed at an intermediate level, explaining 61% of the variance. The augmented models performed less well when tested with data from unfamiliar clinics that had not contributed data to the training dataset, explaining 46% of the variance in response.

Conclusion

These data indicate that ANN models can be quite accurate predictors of virological response to HIV therapy even for patients from unfamiliar clinics. ANN models therefore warrant further development as a potential tool to aid treatment selection.

Impact of HIV-1 pol diversity on drug resistance and its clinical implications

PURPOSE OF REVIEW

HIV knowledge is based on subtype B, common in resource-rich settings, whereas globally non-B subtypes predominate. Inter-subtype pol diversity encompasses multiple genotypic differences among HIV variants, the consequence of which is unknown. This review summarizes publications from the past year relevant to the impact of HIV diversity on drug resistance evolution and its potential clinical implications.

RECENT FINDINGS

The benefit of antiretroviral therapy in non-B infected patients is ongoing, though subtype heterogeneity in rates of disease progression is observed. Pol inter-subtype diversity is high, and known subtype B drug resistance mutations occur in non-B subtypes. New mutations and subtype-specific mutation rates are identified, however, unexplained drug susceptibilities are seen, and additional insight is offered on structural pathogenic mechanisms of resistance in non-B subtypes. These differences may affect genotypic interpretation and our ability to apply drug resistance to patient care.

SUMMARY

Current evidence suggests good treatment response and comparable drug resistance evolution in HIV-1 B and non-B infected patients, with increasingly emerging differences. Impact of inter-subtype diversity on drug susceptibility and on evolution of drug resistance should continue to be a major research focus to increase our understanding and ability to improve global patient care.

The AmpliChip CYP450 genotyping test: Integrating a new clinical tool

The AmpliChip CYP450 Test, which analyzes patient genotypes for cytochrome P450 (CYP) genes CYP2D6 and CYP2C19, is a major step toward introducing personalized prescribing into the clinical environment. Interest in adverse drug reactions (ADRs), the genetic revolution, and pharmacogenetics have converged with the introduction of this tool, which is anticipated to be the first of a new wave of such tools to follow over the next 5-10 years. The AmpliChip CYP450 Test is based on microarray technology, which combines hybridization in precise locations on a glass microarray and a fluorescent labeling system. It classifies individuals into two CYP2C19 phenotypes (extensive metabolizers [EMs] and poor metabolizers [PMs]) by testing three alleles, and into four CYP2D6 phenotypes (ultrarapid metabolizers [UMs], EMs, intermediate metabolizers [IMs], and PMs) by testing 27 alleles, including seven duplications. CYP2D6 is a metabolic enzyme with four activity levels (or phenotypes): UMs with unusually high activity; normal subjects, known as EMs; IMs with low activity; and PMs with no CYP2D6 activity (7% of Caucasians and 1-3% in other ethnic groups). Levels of evidence for the association between CYP2D6 PMs and ADRs are relatively reasonable and include systematic reviews of case-control studies of some typical antipsychotics and tricyclic antidepressants (TCAs). Evidence for other phenotypes is considerably more limited. The CYP2D6 PM phenotype may be associated with risperidone ADRs and discontinuation due to ADRs. Venlafaxine, aripiprazole, duloxetine, and atomoxetine are newer drugs metabolized by CYP2D6 but studies of the clinical relevance of CYP2D6 genotypes are needed. Non-psychiatric drugs metabolized by CYP2D6 include metoprolol, tamoxifen, and codeine-like drugs. CYP2C19 PMs (3-4% of Caucasians and African Americans, and 14-21% of Asians) may require dose adjustment for some TCAs, moclobemide, and citalopram. Other drugs metabolized by CYP2C19 are diazepam and omeprazole. The future of pharmacogenetics depends on the ability to overcome serious obstacles, including the difficulties of conducting and publishing studies in light of resistance from grant agencies, pharmaceutical companies, and some scientific reviewers. Assuming more studies are published, pharmacogenetic clinical applications may be compromised by economic factors and the lack of physician education. The combination of a US FDA-approved test, such as the AmpliChip CYP450 Test, and an FDA definition of CYP2D6 as a 'valid biomarker' makes CYP2D6 genotyping a prime candidate to be the first successful pharmacogenetic test in the clinical environment. One can use microarray technology to test for hundreds of single nucleotide polymorphisms (SNPs) but, taking into account the difficulties for single gene approaches such as CYP2D6, it is unlikely that very complex pharmacogenetic approaches will reach the clinical market in the next 5-10 years.

Genotypic drug resistance interpretation algorithms display high levels of discordance when applied to non-B strains from HIV-1 naive and treated patients

Genotypic drug resistance interpretation algorithms have been developed on patients infected with HIV-1 subtype B to interpret complex patterns of mutations. As non-B strains are characterised by the natural presence of several resistance-related mutations, we examined to what extent this might result in interalgorithm discordances in naive and treated patients. We compared the prediction by three algorithms (ANRS, Stanford and Rega) of drug susceptibilities to diverse HIV-1 strains from 272 naive and 156 treated patients. In naive patients, higher levels of interalgorithm discordance were observed for predictions of protease inhibitor (0.60-39%) than for predictions of reverse transcriptase inhibitor susceptibility (0-4%). The main reason for discordant protease inhibitor interpretation was the presence of resistance mutations that were natural protease polymorphisms. In contrast, in the treated patients, more interalgorithm discordances were observed for predictions of reverse transcriptase inhibitor (5-48%) than protease inhibitor susceptibilities (10-31%). Discordances were related to disagreement between the intermediate and susceptible scores, the intermediate and resistant scores and the interpretations of complex mutation patterns, related to cross-resistance and antagonistic interactions.

Discordances between interpretation algorithms for genotypic resistance to protease and reverse transcriptase inhibitors of human immunodeficiency virus are subtype dependent

The major limitation of drug resistance genotyping for human immunodeficiency virus remains the interpretation of the results. We evaluated the concordance in predicting therapy response between four different interpretation algorithms (Rega 6.3, HIVDB-08/04, ANRS [07/04], and VGI 8.0). Sequences were gathered through a worldwide effort to establish a database of non-B subtype sequences, and demographic and clinical information about the patients was gathered. The most concordant results were found for nonnucleoside reverse transcriptase (RT) inhibitors (93%), followed by protease inhibitors (84%) and nucleoside RT inhibitor (NRTIs) (76%). For therapy-naive patients, for nelfinavir, especially for subtypes C and G, the discordances were driven mainly by the protease (PRO) mutational pattern 82I/V + 63P + 36I/V for subtype C and 82I + 63P + 36I + 20I for subtype G. Subtype F displayed more discordances for ritonavir in untreated patients due to the combined presence of PRO 20R and 10I/V. In therapy-experienced patients, subtype G displayed a lot of discordances for saquinavir and indinavir due to mutational patterns involving PRO 90 M and 82I. Subtype F had more discordance for nelfinavir attributable to the presence of PRO 88S and 82A + 54V. For the NRTIs lamivudine and emtricitabine, CRF01_AE had more discordances than subtype B due to the presence of RT mutational patterns 65R + 115 M and 118I + 215Y, respectively. Overall, the different algorithms agreed well on the level of resistance scored, but some of the discordances could be attributed to specific (subtype-dependent) combinations of mutations. It is not yet known whether therapy response is subtype dependent, but the advice given to clinicians based on a genotypic interpretation algorithm differs according to the subtype.

DR_SEQAN: a PC/Windows-based software to evaluate drug resistance using human immunodeficiency virus type 1 genotypes

Background

Genotypic assays based on DNA sequencing of part or the whole reverse transcriptase (RT)- and protease (PR)-coding regions of the human immunodeficiency virus type 1 (HIV-1) genome have become part of the routine clinical management of HIV-infected individuals. However, the results are difficult to interpret due to complex interactions between mutations found in viral genes.

Results

DR_SEQAN is a tool to analyze RT and PR sequences. The program output includes a list containing all of the amino acid changes found in the query sequence in comparison with the sequence of a wild-type HIV-1 strain. Translation of codons containing nucleotide mixtures can result in potential ambiguities or heterogeneities in the amino acid sequence. The program identifies all possible combinations of 2 or 3 amino acids that derive from translation of triplets containing nucleotide mixtures. In addition, when ambiguities affect codons relevant for drug resistance, DR_SEQAN allows the user to select the appropriate mutation to be considered by the program's drug resistance interpretation algorithm. Resistance is predicted using a rule-based algorithm, whose efficiency and accuracy has been tested with a large set of drug susceptibility data. Drug resistance predictions given by DR_SEQAN were consistent with phenotypic data and coherent with predictions provided by other publicly available algorithms. In addition, the program output provides two tables showing published drug susceptibility data and references for mutations and combinations of mutations found in the analyzed sequence. These data are retrieved from an integrated relational database, implemented in Microsoft Access, which includes two sets of non-redundant core tables (one for combinations of mutations in the PR and the other for combinations in the RT).

Conclusion

DR_SEQAN is an easy to use off-line application that provides expert advice on HIV genotypic resistance interpretation. It is coded in Visual Basic for use in PC/Windows-based platforms. The program is freely available under the General Public License. The program (including the integrated database), documentation and a sample sequence can be downloaded from

A web-based genotyping resource for viral sequences

The Genotyping tool at the National Center for Biotechnology Information is a web-based program that identifies the genotype (or subtype) of recombinant or non-recombinant viral nucleotide sequences. It works by using BLAST to compare a query sequence to a set of reference sequences for known genotypes. Predefined reference genotypes exist for three major viral pathogens: human immunodeficiency virus 1 (HIV-1), hepatitis C virus (HCV) and hepatitis B virus (HBV). User-defined reference sequences can be used at the same time. The query sequence is broken into segments for comparison to the reference so that the mosaic organization of recombinant sequences could be revealed. The results are displayed graphically using color-coded genotypes. Therefore, the genotype(s) of any portion of the query can quickly be determined. The Genotyping tool can be found at: http://www.ncbi.nih.gov/projects/genotyping/formpage.cgi.

Updated European Recommendations for the Clinical Use of HIV Drug Resistance Testing

In most European countries, HIV drug resistance testing has become a routine clinical tool. However, its practical implementation in a clinical context is demanding. The European HIV Drug Resistance Panel was established to make recommendations to clinicians and virologists on this topic and to propose quality control measures. The panel recommends resistance testing for the following indications: i) drug-naive patients with acute or recent infection; ii) therapy failure, including suboptimal treatment response, when treatment change is considered; iii) pregnant HIV-1-infected women and paediatric patients with detectable viral load when treatment initiation or change is considered; and iv) genotype source patient when post-exposure prophylaxis is considered. In addition, for drug-naive patients with chronic infection in whom treatment is to be started, the panel suggests that resistance testing should be strongly considered and recommends testing the earliest sample for drug resistance if suspicion of resistance is high or prevalence of resistance in this population exceeds 10%. The panel does not favour genotyping over phenotype, however it is anticipated that genotyping will be used more often because of its greater accessibility, lower cost and faster turnaround time. For the interpretation of resistance data, clinically validated systems should be used to the greatest extent possible. It is mandatory that laboratories performing HIV resistance tests take regular part in quality assurance programs. Similarly, it is necessary that HIV clinicians and virologists take part in continuous education and meet regularly to discuss problematic clinical cases. Indeed, resistance test results should be used in the context of all other clinically relevant information for predicting therapy response. The panel also encourages the timely collection of epidemiological information to estimate the impact of transmission of resistant HIV and the prevalence of HIV-1 non-B subtypes in the different European countries.