Future research and collaboration: the "SINERGIE" project on HCV (South Italian Network for Rational Guidelines and International Epidemiology)

The SINERGIE (South Italian Network for Rational Guidelines and International Epidemiology) project is intended to set up a collaborative network comprising virologists, clinicians and public health officials dealing with patients affected by HCV disease in the Calabria Region. A prospective observational data-base of HCV infection will be developed and used for studies on HCV natural history, response to treatment, pharmaco-economics, disease complications, and HCV epidemiology (including phylogenetic analysis). With this approach, we aim at improving the identification and care of patients, focusing on upcoming research questions. The final objective is to assist in improving care delivery and inform Public Health Authorities on how to optimize resource allocation in this area.

HIV-1 subtype F1 epidemiological networks among Italian heterosexual males are associated with introduction events from South America

About 40% of the Italian HIV-1 epidemic due to non-B variants is sustained by F1 clade, which circulates at high prevalence in South America and Eastern Europe. Aim of this study was to define clade F1 origin, population dynamics and epidemiological networks through phylogenetic approaches. We analyzed pol sequences of 343 patients carrying F1 subtype stored in the ARCA database from 1998 to 2009. Citizenship of patients was as follows: 72.6% Italians, 9.3% South Americans and 7.3% Rumanians. Heterosexuals, Homo-bisexuals, Intravenous Drug Users accounted for 58.1%, 24.0% and 8.8% of patients, respectively. Phylogenetic analysis indicated that 70% of sequences clustered in 27 transmission networks. Two distinct groups were identified; the first clade, encompassing 56 sequences, included all Rumanian patients. The second group involved the remaining clusters and included 10 South American Homo-bisexuals in 9 distinct clusters. Heterosexual modality of infection was significantly associated with the probability to be detected in transmission networks. Heterosexuals were prevalent either among Italians (67.2%) or Rumanians (50%); by contrast, Homo-bisexuals accounted for 71.4% of South Americans. Among patients with resistant strains the proportion of clustering sequences was 57.1%, involving 14 clusters (51.8%). Resistance in clusters tended to be higher in South Americans (28.6%) compared to Italian (17.7%) and Rumanian patients (14.3%). A striking proportion of epidemiological networks could be identified in heterosexuals carrying F1 subtype residing in Italy. Italian Heterosexual males predominated within epidemiological clusters while foreign patients were mainly Heterosexual Rumanians, both males and females, and South American Homo-bisexuals. Tree topology suggested that F1 variant from South America gave rise to the Italian F1 epidemic through multiple introduction events. The contact tracing also revealed an unexpected burden of resistance in epidemiological clusters underlying the need of public interventions to limit the spread of non-B subtypes and transmitted drug resistance.

Standardized representation, visualization and searchable repository of antiretroviral treatment-change episodes

BACKGROUND

To identify the determinants of successful antiretroviral (ARV) therapy, researchers study the virological responses to treatment-change episodes (TCEs) accompanied by baseline plasma HIV-1 RNA levels, CD4+ T lymphocyte counts, and genotypic resistance data. Such studies, however, often differ in their inclusion and virological response criteria making direct comparisons of study results problematic. Moreover, the absence of a standard method for representing the data comprising a TCE makes it difficult to apply uniform criteria in the analysis of published studies of TCEs.

RESULTS

To facilitate data sharing for TCE analyses, we developed an XML (Extensible Markup Language) Schema that represents the temporal relationship between plasma HIV-1 RNA levels, CD4 counts and genotypic drug resistance data surrounding an ARV treatment change. To demonstrate the adaptability of the TCE XML Schema to different clinical environments, we collaborate with four clinics to create a public repository of about 1,500 TCEs. Despite the nascent state of this TCE XML Repository, we were able to perform an analysis that generated a novel hypothesis pertaining to the optimal use of second-line therapies in resource-limited settings. We also developed an online program (TCE Finder) for searching the TCE XML Repository and another program (TCE Viewer) for generating a graphical depiction of a TCE from a TCE XML Schema document.

CONCLUSIONS

The TCE Suite of applications - the XML Schema, Viewer, Finder, and Repository - addresses several major needs in the analysis of the predictors of virological response to ARV therapy. The TCE XML Schema and Viewer facilitate sharing data comprising a TCE. The TCE Repository, the only publicly available collection of TCEs, and the TCE Finder can be used for testing the predictive value of genotypic resistance interpretation systems and potentially for generating and testing novel hypotheses pertaining to the optimal use of salvage ARV therapy.

Identification and structural characterization of novel genetic elements in the HIV-1 V3 loop regulating coreceptor usage

BACKGROUND

The interaction between HIV-1 gp120 and CCR5 N terminus is critical for R5-virus entry and affects CCR5 antagonists' activity. Knowledge of how different genetic signatures of gp120 V3 domain effect the strength of this interaction is limited.

METHODS

HIV-1 coreceptor usage was assessed in 251 patients using enhanced-sensitivity Trofile assay and V3 sequencing plus tropism prediction by Geno2pheno algorithm. Bayesian partitional model and recursive model selection have been used to define V3 genetic determinants correlated with different coreceptor usage. Gp120 interaction with CCR5 N terminus was evaluated by docking-analysis/molecular-dynamic simulations starting from the model described previously.

RESULTS

Selected V3 genetic determinants (beyond known aminoacidic positions) significantly correlate with CCR5- or CXCR4-usage, and modulate gp120 affinity for CCR5 N terminus. This is the case for N5Y and N7K, absent in CCR5-using viruses and present in 4.5% and 6% of CXCR4-using viruses, respectively, and A19V, occurring in 2.6% of CCR5-using viruses and 22.0% of CXCR4-using viruses (P=10(-2) to 10(-7)). Their presence determines a decreased affinity for CCR5 N terminus even stronger than that observed in the presence of the well-known mutation S11R (N5Y: -6.60 Kcal/mol; N7K: -5.40 Kcal/mol; A19V: -5.60 Kcal/mol; S11R: -6.70 Kcal/mol; WT: -6.90 Kcal/mol). N7K significantly increases the distance between V3 position 7 and sulphotyrosine at CCR5 position 14 (crucial for binding to gp120; from 4.22 Å to 8.30 Å), thus abrogating the interaction between these two important residues.

CONCLUSIONS

Key determinants for tropism within the V3 sequence, confirmed by structure- and by phenotypic-tropism, have been identified. This information can be used for a finer tuning of potential efficacy of CCR5-antagonists in clinical practice, and to provide molecular implications for design of new entry inhibitors.

Near full-length sequence analysis of HIV type 1 BF recombinants from Italy

Recombination between HIV-1 subtypes B and F has generated several circulating and unique recombinant forms, particularly in Latin American areas. In Italy, subtype B is highly prevalent while subtype F is the most common pure non-B subtype. To investigate the recombination pattern in Italian BF recombinant viruses, we characterized full-length sequences derived from 15 adult patients, mostly Italian and infected by the heterosexual route. One of the BF mosaics was a CRF29, three sequences clustered with low bootstrap values with CRF39, CRF40, and CRF42. With the exception of the CRF29-like sequence, the other recombination patterns were unique, but two possible clusters were identified. Analysis of the gp120 V3 domain suggested a possible link with subtype F from Eastern Europe rather than from Latin America, favoring the hypothesis of local recombination between clade B and F viruses over that of import of BF recombinants from Latin America. HIV-1 subtypes B and F appear prone to generation of unique recombinants in Italy, warranting epidemiological surveillance and investigation of a possible clinical significance.

Efficacy of antiretroviral therapy switch in HIV-infected patients: a 10-year analysis of the EuResist Cohort

INTRODUCTION

Highly active antiretroviral therapy (HAART) has been shown to be effective in many recent trials. However, there is limited data on time trends of HAART efficacy after treatment change.

METHODS

Data from different European cohorts were compiled within the EuResist Project. The efficacy of HAART defined by suppression of viral replication at 24 weeks after therapy switch was analyzed considering previous treatment modifications from 1999 to 2008.

RESULTS

Altogether, 12,323 treatment change episodes in 7,342 patients were included in the analysis. In 1999, HAART after treatment switch was effective in 38.0% of the patients who had previously undergone 1-5 therapies. This figure rose to 85.0% in 2008. In patients with more than 5 previous therapies, efficacy rose from 23.9 to 76.2% in the same time period. In patients with detectable viral load at therapy switch, the efficacy rose from 23.3 to 66.7% with 1-5 previous treatments and from 14.4 to 55.6% with more than 5 previous treatments.

CONCLUSION

The results of this large cohort show that the outcome of HAART switch has improved considerably over the last years. This result was particularly observed in the context after viral rebound. Thus, changing HAART is no longer associated with a high risk of treatment failure.

Naturally occurring hepatitis C virus (HCV) NS3/4A protease inhibitor resistance-related mutations in HCV genotype 1-infected subjects in Italy

OBJECTIVES

To assess the prevalence of hepatitis C virus (HCV) NS3/4A protease inhibitor (PI) resistance mutations in HCV genotype 1-infected PI-naive individuals in Italy.

PATIENTS AND METHODS

One hundred and twelve patients infected with HCV genotype 1a or 1b (based on Versant HCV Genotype 2.0 or 5'UTR/core sequencing) and never treated with any HCV PI were evaluated. The whole NS3 region was analysed by population sequencing and mutations related to resistance to linear and macrocyclic PIs were recorded.

RESULTS

Forty-six HCV-monoinfected and 66 HCV/HIV-coinfected subjects were studied. Complete NS3 sequence information was obtained for 109 (97.3%) samples: 67 subtype 1a and 42 subtype 1b. Subtype assignment by NS3 sequencing was concordant in 100.0% and 83.9% of cases with the original 5'UTR sequencing and Versant result, respectively. At least one mutation related to PI resistance was detected in 21 (19.3%) isolates. However, 11 of these had only Q80K, expected to confer resistance to one investigational macrocyclic compound, and were detected only in subtype 1a. Boceprevir and telaprevir resistance-related mutations were detected in 10 (9.2%) isolates and included V36L, T54S and V55A. Only one isolate harboured two mutations (V36L and T54S). There was no association between HCV PI resistance and HIV coinfection or exposure to HIV PIs.

CONCLUSIONS

A minority of untreated HCV genotype 1 patients in Italy harbour a virus population carrying HCV PI resistance-related mutations. The clinical implications of this finding warrant further analysis.

Performance of genotypic tropism testing on proviral DNA in clinical practice: results from the DIVA study group

OBJECTIVE

The DIVA study is aimed at setting up a standardized genotypic tropism-testing on proviral-DNA for the routine clinical diagnostic-laboratory.

METHODS

Twelve local centres and 5 reference centres (previously cross-validated) were identified. For inter-center validation-procedure, 60 peripheral-blood mononuclear cells (PBMCs) aliquots from 45 HAART-treated patients were randomly chosen for population V3 sequencing on proviral-DNA at local HIV centre and at reference-laboratory. Viral tropism was predicted by Geno2Pheno algorithm (False Positive Rate [FPR] = 20%) as proposed by the European-Guidelines. Quantification of total HIV-1 DNA was based on a method described by Viard (2004).

RESULTS

Quantification of HIV-1 DNA was available for 35/45 (77.8%) samples, and gave a median value of 598 (IQR:252- 1,203) copies/10 PBMCs. A total of 56/60 (93.3%) samples were successfully amplified by both the reference and the local virological centers. The overall concordance of tropism prediction between local and reference centers was 54/56 (96.4%). Results of tropism prediction by local centers were: 33/54 (61.1%) R5 and 21/54 (38.9%) X4/DM.

CONCLUSION

There was high concordance in the genotypic tropism prediction based on proviral DNA among different virological centers throughout Italy. Our results are in line with other European studies, and support the use of genotypic tropism testing on proviral DNA in patients with suppressed plasma HIV-1 RNA candidate to CCR5-antagonist treatment.

Human immunodeficiency virus-1 B and non-B subtypes with the same drug resistance pattern respond similarly to antiretroviral therapy

We analysed the 12-week virological response to protease inhibitor (PI) or non-nucleoside reverse transcriptase inhibitor (NNRTI) therapy in 1108 patients carrying B or non-B human immunodeficiency virus (HIV)-1 subtypes with matched resistance mutation patterns. Response rates were not significantly different for non-B and B subtypes stratified for treatment status (51.5% vs. 41.5% in naïve patients; 46.7% vs. 38.7% in experienced patients) or regimens (46.9% vs. 39.7% with PI; 56.7% vs. 40% with NNRTI). No difference in response was detected in patients harbouring B and non-B subtypes with any resistance profile. Further studies are advisable to fully test this approach on larger datasets.

HIV-1 mutational pathways under multidrug therapy

BACKGROUND

Genotype-derived drug resistance profiles are a valuable asset in HIV-1 therapy decisions. Therapy decisions could be further improved, both in terms of predicting length of current therapy success and in preserving followup therapy options, through better knowledge of mutational pathways- here defined as specific locations on the viral genome which, when mutant, alter the risk that additional specific mutations arise. We limit the search to locations in the reverse transcriptase region of the HIV-1 genome which host resistance mutations to nucleoside (NRTI) and non-nucleoside (NNRTI) reverse transcriptase inhibitors (as listed in the 2008 International AIDS Society report), or which were mutant at therapy start in 5% or more of the therapies studied.

METHODS

A Cox proportional hazards model was fit to each location with the hazard of a mutation at that location during therapy proportional to the presence/absence of mutations at the remaining locations at therapy start. A pathway from preexisting to occurring mutation was indicated if the covariate was both selected as important via smoothly clipped absolute deviation (a form of regularized regression) and had a small p-value. The Cox model also allowed controlling for non-genetic parameters and potential nuisance factors such as viral resistance and number of previous therapies. Results were based on 1981 therapies given to 1495 distinct patients drawn from the EuResist database.

RESULTS

The strongest influence on the hazard of developing NRTI resistance was having more than four previous therapies, not any one existing resistance mutation. Known NRTI resistance pathways were shown, and previously speculated inhibition between the thymidine analog pathways was evidenced. Evidence was found for a number of specific pathways between NRTI and NNRTI resistance sites. A number of common mutations were shown to increase the hazard of developing both NRTI and NNRTI resistance. Viral resistance to the therapy compounds did not materially effect the hazard of mutation in our model.

CONCLUSIONS

The accuracy of therapy outcome prediction tools may be increased by including the number of previous treatments, and by considering locations in the HIV genome which increase the hazard of developing resistance mutations.

A prognostic model for estimating the time to virologic failure in HIV-1 infected patients undergoing a new combination antiretroviral therapy regimen

Background

HIV-1 genotypic susceptibility scores (GSSs) were proven to be significant prognostic factors of fixed time-point virologic outcomes after combination antiretroviral therapy (cART) switch/initiation. However, their relative-hazard for the time to virologic failure has not been thoroughly investigated, and an expert system that is able to predict how long a new cART regimen will remain effective has never been designed.

Methods

We analyzed patients of the Italian ARCA cohort starting a new cART from 1999 onwards either after virologic failure or as treatment-naïve. The time to virologic failure was the endpoint, from the 90^th day after treatment start, defined as the first HIV-1 RNA > 400 copies/ml, censoring at last available HIV-1 RNA before treatment discontinuation. We assessed the relative hazard/importance of GSSs according to distinct interpretation systems (Rega, ANRS and HIVdb) and other covariates by means of Cox regression and random survival forests (RSF). Prediction models were validated via the bootstrap and c-index measure.

Results

The dataset included 2337 regimens from 2182 patients, of which 733 were previously treatment-naïve. We observed 1067 virologic failures over 2820 persons-years. Multivariable analysis revealed that low GSSs of cART were independently associated with the hazard of a virologic failure, along with several other covariates. Evaluation of predictive performance yielded a modest ability of the Cox regression to predict the virologic endpoint (c-index≈0.70), while RSF showed a better performance (c-index≈0.73, p < 0.0001 vs. Cox regression). Variable importance according to RSF was concordant with the Cox hazards.

Conclusions

GSSs of cART and several other covariates were investigated using linear and non-linear survival analysis. RSF models are a promising approach for the development of a reliable system that predicts time to virologic failure better than Cox regression. Such models might represent a significant improvement over the current methods for monitoring and optimization of cART.

Low rate of virological failure and maintenance of susceptibility to HIV-1 protease inhibitors with first-line lopinavir/ritonavir-based antiretroviral treatment in clinical practice

Protease inhibitor (PI)-resistant HIV-1 has hardly ever been detected at failed boosted PI-based first-line antiretroviral regimens in clinical trials. However, this phenomenon has not been investigated in clinical practice. To address this gap, data from patients starting a first-line lopinavir/ritonavir (LPV/rtv)-based therapy with available baseline HIV-1 RNA load, a viral genotype and follow-up viral load after 3 and 6 months of treatment were extracted from the Italian Antiretroviral Resistance Cohort Analysis (ARCA) observational database. Based on survival analysis, 39 (7.1%) and 43 (7.8%) of the 548 examined patient cases had an HIV-1 RNA >500 and >50 copies/ml, respectively, after 6 months of treatment. Cox proportional hazard models detected baseline HIV-1 RNA (RH 1.79, 95%CI 1.10-2.92 per 1-log(10) increase, P=0.02) and resistance to the nucleoside backbone (RH 1.04, 95%CI 1.02-1.06 per 10-point increase using the Stanford HIVdb algorithm, P<0.001) as independent predictors of HIV-1 RNA at >500 copies/ml, but not at the >50 copies/ml cutoff criteria. Higher baseline viral load, older patient age, heterosexual route of infection and use of tenofovir/emtricitabine were predictors of failure at month 3 using the 50-copy and/or 500-copy threshold. Resistance to LPV/rtv did not occur or increase in any of the available 36 follow-up HIV-1 genotypes. Resistance to the nucleoside backbone (M184V) developed in four cases. Despite the likely differences in patient population and adherence, both the low rate of virological failure and the lack of development of LPV/rtv resistance documented in clinical trials are thus confirmed in clinical practice.

No pol mutation is associated independently with the lack of immune recovery in patients infected with HIV and failing antiretroviral therapy

An investigation was undertaken to determine whether specific pol mutations hinder long-term immune recovery regardless of virological response. In total, 826 patients with >50 HIV RNA copies/ml, who underwent genotypic resistance testing between 1 January 2000 and 31 December 2003 after >3 years of antiretroviral treatment, and were followed up for >3 years after genotypic resistance testing, were analyzed retrospectively. The outcome of the study was the lack of immune recovery after >3 years of follow-up, defined as a slope by linear regression <0. The viremia detectability ratio was defined as the number of HIV RNA values of >50 copies/ml divided by the number of HIV RNA measurements during follow-up. Logistic regression was used for univariable and multivariable analysis. Median (Q1, Q3) values at baseline were the following: age 40 (37, 45) years, years on antiretroviral therapy 4.45 (3.65, 5.47), HIV RNA 3.91 (3.39, 4.53) log(10) copies/ml, CD4+ T-cell 358 (211, 524)/µl. After 3.13 years of follow-up, 375 patients (45.4%) showed a lack of immune recovery. The risk of lack of immune recovery increased independently with increasing baseline CD4+ counts (OR=1.104 per 50-cell increase, 95% CI=1.069-1.142, P<0.0001), increasing viremia detectability ratio during follow-up (OR=1.145 per 0.1-unit increase, 95% CI=1.093-1.202, P<0.0001), and with earlier calendar years of resistance testing (overall effect: P=0.0007). In conclusion, no pol mutation is associated independently with the lack of immune recovery.

HIV-1 drug-resistance patterns among patients on failing treatment in a large number of European countries

BACKGROUND

Information about patterns of HIV-1 drug resistance among treatment-exposed patients is crucial for the development of novel effective drugs. Currently no system exists that monitors patterns of resistance in patients failing therapy.

METHODS

The study included 1,988 HIV-1 sequences from patients experiencing therapy failure collected between 2000 and 2004 in 15 European countries. Genotypic resistance was interpreted using the ANRS algorithm. Phenotypic resistance was predicted using the Virco geno- to phenotype system.

RESULTS

80.7% of the sequences included at least one drug-resistance mutation. Mutations were found for NRTIs (73.5%), NNRTIs (48.5%), and protease inhibitors (35.8%). Ninety percent of sequences with genotypic resistance harbored M184V, M41L, K103N, D67N, and/or T215Y. Among NRTIs, resistance was most frequently predicted for lamivudine. About half of all sequences had reduced susceptibility for NNRTIs. Resistance to most boosted protease inhibitors was found in < 25%. No sequence had resistance to all currently available drugs.

CONCLUSION

Levels of resistance among patients with therapy failure were high. The patterns of resistance reflect resistance to drugs available for a longer time. Fully suppressive regimens can be designed even for the most mutated HIV because boosted protease inhibitors have remained active against most circulating viruses and new drug classes have become available.

Antiretroviral therapy optimisation without genotype resistance testing: a perspective on treatment history based models

Background

Although genotypic resistance testing (GRT) is recommended to guide combination antiretroviral therapy (cART), funding and/or facilities to perform GRT may not be available in low to middle income countries. Since treatment history (TH) impacts response to subsequent therapy, we investigated a set of statistical learning models to optimise cART in the absence of GRT information.

Methods and Findings

The EuResist database was used to extract 8-week and 24-week treatment change episodes (TCE) with GRT and additional clinical, demographic and TH information. Random Forest (RF) classification was used to predict 8- and 24-week success, defined as undetectable HIV-1 RNA, comparing nested models including (i) GRT+TH and (ii) TH without GRT, using multiple cross-validation and area under the receiver operating characteristic curve (AUC). Virological success was achieved in 68.2% and 68.0% of TCE at 8- and 24-weeks (n = 2,831 and 2,579), respectively. RF (i) and (ii) showed comparable performances, with an average (st.dev.) AUC 0.77 (0.031) vs. 0.757 (0.035) at 8-weeks, 0.834 (0.027) vs. 0.821 (0.025) at 24-weeks. Sensitivity analyses, carried out on a data subset that included antiretroviral regimens commonly used in low to middle income countries, confirmed our findings. Training on subtype B and validation on non-B isolates resulted in a decline of performance for models (i) and (ii).

Conclusions

Treatment history-based RF prediction models are comparable to GRT-based for classification of virological outcome. These results may be relevant for therapy optimisation in areas where availability of GRT is limited. Further investigations are required in order to account for different demographics, subtypes and different therapy switching strategies.

Prediction of response to antiretroviral therapy by human experts and by the EuResist data-driven expert system (the EVE study)

OBJECTIVES

The EuResist expert system is a novel data-driven online system for computing the probability of 8-week success for any given pair of HIV-1 genotype and combination antiretroviral therapy regimen plus optional patient information. The objective of this study was to compare the EuResist system vs. human experts (EVE) for the ability to predict response to treatment.

METHODS

The EuResist system was compared with 10 HIV-1 drug resistance experts for the ability to predict 8-week response to 25 treatment cases derived from the EuResist database validation data set. All current and past patient data were made available to simulate clinical practice. The experts were asked to provide a qualitative and quantitative estimate of the probability of treatment success.

RESULTS

There were 15 treatment successes and 10 treatment failures. In the classification task, the number of mislabelled cases was six for EuResist and 6-13 for the human experts [mean±standard deviation (SD) 9.1±1.9]. The accuracy of EuResist was higher than the average for the experts (0.76 vs. 0.64, respectively). The quantitative estimates computed by EuResist were significantly correlated (Pearson r=0.695, P<0.0001) with the mean quantitative estimates provided by the experts. However, the agreement among experts was only moderate (for the classification task, inter-rater κ=0.355; for the quantitative estimation, mean±SD coefficient of variation=55.9±22.4%).

CONCLUSIONS

With this limited data set, the EuResist engine performed comparably to or better than human experts. The system warrants further investigation as a treatment-decision support tool in clinical practice.

Comparative determination of HIV-1 co-receptor tropism by Enhanced Sensitivity Trofile, gp120 V3-loop RNA and DNA genotyping

Background

Trofile^® is the prospectively validated HIV-1 tropism assay. Its use is limited by high costs, long turn-around time, and inability to test patients with very low or undetectable viremia. We aimed at assessing the efficiency of population genotypic assays based on gp120 V3-loop sequencing for the determination of tropism in plasma viral RNA and in whole-blood viral DNA. Contemporary and follow-up plasma and whole-blood samples from patients undergoing tropism testing via the enhanced sensitivity Trofile^® (ESTA) were collected. Clinical and clonal geno2pheno_[coreceptor] (G2P) models at 10% and at optimised 5.7% false positive rate cutoff were evaluated using viral DNA and RNA samples, compared against each other and ESTA, using Cohen's kappa, phylogenetic analysis, and area under the receiver operating characteristic (AUROC).

Results

Both clinical and clonal G2P (with different false positive rates) showed good performances in predicting the ESTA outcome (for V3 RNA-based clinical G2P at 10% false positive rate AUROC = 0.83, sensitivity = 90%, specificity = 75%). The rate of agreement between DNA- and RNA-based clinical G2P was fair (kappa = 0.74, p < 0.0001), and DNA-based clinical G2P accurately predicted the plasma ESTA (AUROC = 0.86). Significant differences in the viral populations were detected when comparing inter/intra patient diversity of viral DNA with RNA sequences.

Conclusions

Plasma HIV RNA or whole-blood HIV DNA V3-loop sequencing interpreted with clinical G2P is cheap and can be a good surrogate for ESTA. Although there may be differences among viral RNA and DNA populations in the same host, DNA-based G2P may be used as an indication of viral tropism in patients with undetectable plasma viremia.

Changing patterns in HIV-1 non-B clade prevalence and diversity in Italy over three decades

BACKGROUND

HIV-1 non-B subtypes have recently entered Western Europe following immigration from other regions. The distribution of non-B clades and their association with demographic factors, over the entire course of the HIV-1 epidemic, have not been fully investigated in Italy.

METHODS

We carried out a phylogenetic analysis of HIV-1 pol sequences derived from 3670 patients followed at 50 Italian clinical centres over nearly three decades.

RESULTS

Overall, 417 patients (11.4%) carried non-B subtypes. The prevalence of non-B strains increased from 2.6% in 1980-1992 to 18.9% in 1993-2008 (P<0.0001) in a subset of 2479 subjects with a known year of diagnosis. A multivariate analysis on a subset of 1364 patients for whom relevant demographic data were available indicated that African ethnicity, heterosexual route of infection and year of diagnosis were independently associated with non-B HIV-1 infection (P ≤ 0.0001). All pure subtypes, except for clade K, and seven circulating recombinant forms were detected, accounting for 56.6 and 34.1% of the non-B infections, respectively. The F1 subtype was the most prevalent non-B clade among Europeans and was acquired heterosexually in half of this patient population. Unique recombinant forms accounted for 9.4% of the non-B sequences and showed a B/F1 recombination pattern in one-third of cases.

CONCLUSIONS

The circulation of non-B clades has significantly increased in Italy in association with demographic changes. Spread of the F1 subtype and B/F recombinants appears to predominate, which may result in a redistribution of the relative proportions of the different strains, and this could lead to overlapping epidemics. Thus, the HIV-1 landscape in Italy may in future be distinct from that of the rest of Europe.

Prevalence of mutations and determinants of genotypic resistance to etravirine (TMC125) in a large Italian resistance database (ARCA)

OBJECTIVES

To evaluate whether etravirine (TMC125) might be effective in patients failing therapy with current nonnucleoside reverse transcriptase inhibitors (NNRTIs), we analysed the prevalence of TMC125 mutations and the possible determinants of genotypic resistance to this drug among sequences reported to a large database in Italy [Antiretroviral Resistance Cohort Analysis (ARCA)].

METHODS

We analysed the prevalence of TMC125 resistance-associated mutations (RAMs) and the TMC125 weighted genotypic score (WGS) together with the determinants of genotypic resistance. A total of 5011 sequences from 2955 patients failing NNRTI therapy were evaluated.

RESULTS

Among the sequences in ARCA, 68% had at least one and 9.8% at least three TMC125 RAMs, whereas 31% had a WGS>2. Frequent RAMs were Y181C, G190A, K101E and A98G, whereas V179F, Y181V and G190S appeared in <5% of sequences. Multivariate analysis revealed a higher risk of developing at least three TMC125 RAMs associated with both nevirapine and efavirenz exposure, whereas CD4 counts > or = 200 cells/microL retained their protective effect. An increased risk of WGS>2 was linked to higher HIV RNA values (maximum risk at >5 log(10) copies/mL) and nevirapine exposure; CD4 counts > or = 200 cells/microL were protective.

CONCLUSIONS

The prevalence of TMC125 resistance mutations in the ARCA cohort was 68%. The DUET studies showed that at least three TMC125-associated mutations were required to impair the efficacy of the drug and Y181C/V, V179F and G190S had the greatest effect on response. The prevalence of these mutations among the patients examined in our study was low. However, WGS>2 was found for one-third of our sequences. Previous nevirapine exposure was associated with an increased risk of having WGS>2 (adjusted odds ratio 1.76).

Only slight impact of predicted replicative capacity for therapy response prediction

Background

Replication capacity (RC) of specific HIV isolates is occasionally blamed for unexpected treatment responses. However, the role of viral RC in response to antiretroviral therapy is not yet fully understood.

Materials and Methods

We developed a method for predicting RC from genotype using support vector machines (SVMs) trained on about 300 genotype-RC pairs. Next, we studied the impact of predicted viral RC (pRC) on the change of viral load (VL) and CD4⁺ T-cell count (CD4) during the course of therapy on about 3,000 treatment change episodes (TCEs) extracted from the EuResist integrated database. Specifically, linear regression models using either treatment activity scores (TAS), the drug combination, or pRC or any combination of these covariates were trained to predict change in VL and CD4, respectively.

Results

The SVM models achieved a Spearman correlation (ρ) of 0.54 between measured RC and pRC. The prediction of change in VL (CD4) was best at 180 (360) days, reaching a correlation of ρ = 0.45 (ρ = 0.27). In general, pRC was inversely correlated to drug resistance at treatment start (on average ρ = −0.38). Inclusion of pRC in the linear regression models significantly improved prediction of virological response to treatment based either on the drug combination or on the TAS (t-test; p-values range from 0.0247 to 4 10⁻⁶) but not for the model using both TAS and drug combination. For predicting the change in CD4 the improvement derived from inclusion of pRC was not significant.

Conclusion

Viral RC could be predicted from genotype with moderate accuracy and could slightly improve prediction of virological treatment response. However, the observed improvement could simply be a consequence of the significant correlation between pRC and drug resistance.

Detection of residual human immunodeficiency virus type 1 reverse transcriptase K103N minority species in plasma RNA and peripheral blood mononuclear cell DNA following discontinuation of non-nucleoside therapy

Non-nucleoside reverse transcriptase inhibitor (NNRTI) therapy failed in 30 patients with the typical human immunodeficiency virus type 1 reverse transcriptase K103N mutation, detected using standard genotyping. Following discontinuation of NNRTI therapy for a median of 55.9 weeks and a decrease of K103N mutant species to undetectable levels in plasma RNA, minority K103N species remained detectable, by allele-specific PCR, for longer periods of time and at higher frequency, in peripheral blood mononuclear cell (PBMC) DNA than in plasma RNA (76.7% and 46.7% of samples with residual K103N species detected at median frequencies of 18.0% and 3.8%, respectively). Analysis of PBMC DNA should be considered when searching for residual K103N mutant species in patients previously exposed to NNRTIs.

Effectiveness of antiretroviral regimens containing abacavir with tenofovir in treatment-experienced patients: predictors of virological response and drug resistance evolution in a multi-cohort study

BACKGROUND

In treatment-naïve patients, a combination antiretroviral therapy (cART) containing tenofovir (TDF) and abacavir (ABC) with lamivudine leads to unacceptably high virological failure rates with frequent selection of reverse transcriptase mutations M184V and K65R. We explored the efficacy of at least 16 weeks of ABC + TDF-containing cART regimens in 307 antiretroviral-experienced HIV-1-infected individuals included in observational databases.

METHODS

Virological failure was defined as an HIV RNA > 400 copies/ml after at least 16 weeks of treatment. Patients had received a median of three prior cART regimens. Of these, 76% concomitantly received a potent or high genetic barrier regimen (with at least one protease inhibitor [PI]) or non-nucleoside reverse transcriptase inhibitor or thymidine analogue) while a third non-thymidine nucleoside analogue was used in the remaining patients.

RESULTS

The 1-year estimated probability of virological failure was 34% in 165 patients with HIV RNA > 400 copies/ ml at ABC + TDF regimen initiation. Independent predictors of virological failure were the absence of a potent or high genetic barrier cART, the higher number of cART regimens experienced, and the use of a new drug class. In the subset of 136 patients for whom there were genotypic resistance test results prior to ABC + TDF initiation, the virological failure (1-year estimated probability 46%) was independently predicted by the higher baseline viral load, the concomitant use of boosted PI, and the presence of reverse transcriptase mutation M41L. In 142 patients starting ABC + TDF therapy with HIV RNA pound < or =400 copies/ml, virological failure (1-year estimated probability 17%) was associated only with the transmission category. In a small subset of subjects for whom there were an available paired baseline and follow-up genotype (n = 28), the prevalence of most nucleoside analogue reverse transcriptase inhibitor resistance mutations decreased, suggesting a possible low adherence to treatment. No selection of K65R was detected.

CONCLUSION

The virological response to ABC + TDF-containing regimens in this moderately-to-heavily treatment experienced cohort was good. Higher viral load and the presence of M41L at baseline were associated with worse virological responses, while the concomitant prescription of drugs enhancing the genetic barrier of the regimen conveyed a reduced risk of virological failure. The Appendix provides the names of other members of the MASTER cohort.

Rules-based HIV-1 genotypic resistance interpretation systems predict 8 week and 24 week virological antiretroviral treatment outcome and benefit from drug potency weighting

OBJECTIVES

To test retrospectively the ability of four freely available rules-based expert systems to predict short- and medium-term virological outcome following an antiretroviral treatment switch in pre-treated HIV-1 patients.

METHODS

The HIV-1 genotype interpretation systems (GISs) HIVdb, ANRS, Rega and AntiRetroScan were tested for their accuracy in predicting response to highly active antiretroviral therapy using 8 week (n = 765) and 24 week (n = 634) follow-up standardized treatment change episodes extracted from the Italian Antiretroviral Resistance Cohort Analysis (ARCA) database. A genotypic sensitivity score (GSS) was derived for each genotype-treatment pair for the different GISs and tested as a predictor of virological treatment outcome by univariable and multivariable logistic regression as well as by receiver operating characteristic curve analysis. The two systems implementing drug potency weights (AntiRetroScan and Rega) were evaluated with and without this correction factor.

RESULTS

All four GSSs were strong predictors of virological treatment outcome at both 8 and 24 weeks after adjusting for baseline viro-immunological parameters and previous drug exposure (odds ratios ranging from 2.04 to 2.43 per 1 unit GSS increase; P < 0.001 for all the systems). The accuracy of AntiRetroScan and Rega was significantly increased by drug potency weighting with respect to the unweighted versions (P <or= 0.001). HIVdb and ANRS also increased their performance with the same drug potency weighting adopted by AntiRetroScan and Rega, respectively (P < 0.001 for both analyses).

CONCLUSIONS

Currently available GISs are valuable tools for assisting antiretroviral treatment choices. Drug potency weighting can increase the accuracy of all systems.

Prevalence of transmitted HIV-1 drug resistance in HIV-1-infected patients in Italy: evolution over 12 years and predictors

OBJECTIVES

Transmitted HIV-1 drug resistance (TDR) can reduce the efficacy of first-line antiretroviral therapy.

PATIENTS AND METHODS

A retrospective analysis was performed to assess the prevalence and correlates of TDR in Italy over time. TDR was defined as the presence of at least one of the mutations present in the surveillance drug resistance mutation (SDRM) list.

RESULTS

Among 1690 antiretroviral therapy-naive patients, the most frequent HIV subtypes were B (78.8%), CRF02_AG (5.6%) and C (3.6%). Overall, TDR was 15%. TDR was 17.3% in subtype B and 7.0% in non-B carriers (P < 0.001). TDR showed a slight, although not significant, decline (from 16.3% in 1996-2001 to 13.4% in 2006-07, P = 0.15); TDR declined for nucleoside reverse transcriptase inhibitors (from 13.1% to 8.2%, P = 0.003) but remained stable for protease inhibitors (from 3.7% to 2.5%, P = 0.12) and non-nucleoside reverse transcriptase inhibitors (from 3.7% to 5.8%). TDR to any drug was stable in B subtype and showed a decline trend in non-B. In multivariable analysis, F1 subtype or any non-B subtype, compared with B subtype, and higher HIV RNA were independent predictors of reduced odds of TDR.

CONCLUSIONS

Prevalence of TDR to nucleoside reverse transcriptase inhibitors seems to have declined in Italy over time. Increased prevalence of non-B subtypes partially justifies this phenomenon.

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.

Tracing the HIV-1 subtype B mobility in Europe: a phylogeographic approach

BACKGROUND

The prevalence and the origin of HIV-1 subtype B, the most prevalent circulating clade among the long-term residents in Europe, have been studied extensively. However the spatial diffusion of the epidemic from the perspective of the virus has not previously been traced.

RESULTS

In the current study we inferred the migration history of HIV-1 subtype B by way of a phylogeography of viral sequences sampled from 16 European countries and Israel. Migration events were inferred from viral phylogenies by character reconstruction using parsimony. With regard to the spatial dispersal of the HIV subtype B sequences across viral phylogenies, in most of the countries in Europe the epidemic was introduced by multiple sources and subsequently spread within local networks. Poland provides an exception where most of the infections were the result of a single point introduction. According to the significant migratory pathways, we show that there are considerable differences across Europe. Specifically, Greece, Portugal, Serbia and Spain, provide sources shedding HIV-1; Austria, Belgium and Luxembourg, on the other hand, are migratory targets, while for Denmark, Germany, Italy, Israel, Norway, the Netherlands, Sweden, Switzerland and the UK we inferred significant bidirectional migration. For Poland no significant migratory pathways were inferred.

CONCLUSION

Subtype B phylogeographies provide a new insight about the geographical distribution of viral lineages, as well as the significant pathways of virus dispersal across Europe, suggesting that intervention strategies should also address tourists, travellers and migrants.

Predicting the response to combination antiretroviral therapy: retrospective validation of geno2pheno-THEO on a large clinical database

BACKGROUND

Expert-based genotypic interpretation systems are standard methods for guiding treatment selection for patients infected with human immunodeficiency virus type 1. We previously introduced the software pipeline geno2pheno-THEO (g2p-THEO), which on the basis of viral sequence predicts the response to treatment with a combination of antiretroviral compounds by applying methods from statistical learning and the estimated potential of the virus to escape from drug pressure.

METHODS

We retrospectively validated the statistical model used by g2p-THEO in approximately 7600 independent treatment-sequence pairs extracted from the EuResist integrated database, ranging from 1990 to 2007. Results were compared with the 3 most widely used expert-based interpretation systems: Stanford HIVdb, ANRS, and Rega.

RESULTS

The difference in receiver operating characteristic curves between g2p-THEO and expert-based approaches was significant (P < .001; paired Wilcoxon test). Indeed, at 80% specificity, g2p-THEO found 16.2%-19.8% more successful regimens than did the expert-based approaches. The increased performance of g2p-THEO was confirmed in a 2001-2007 data set from which most obsolete therapies had been removed.

CONCLUSION

Finding drug combinations that increase the chances of therapeutic success is the main reason for using decision support systems. The present analysis of a large data set derived from clinical practice demonstrates that g2p-THEO solves this task significantly better than state-of-the-art expert-based systems. The tool is available at http://www.geno2pheno.org.

Investigation of expert rule bases, logistic regression, and non-linear machine learning techniques for predicting response to antiretroviral treatment

Background

The extreme flexibility of the HIV type-1 (HIV-1) genome makes it challenging to build the ideal antiretroviral treatment regimen. Interpretation of HIV-1 genotypic drug resistance is evolving from rule-based systems guided by expert opinion to data-driven engines developed through machine learning methods.

Methods

The aim of the study was to investigate linear and non-linear statistical learning models for classifying short-term virological outcome of antiretroviral treatment. To optimize the model, different feature selection methods were considered. Robust extra-sample error estimation and different loss functions were used to assess model performance. The results were compared with widely used rule-based genotypic interpretation systems (Stanford HIVdb, Rega and ANRS).

Results

A set of 3,143 treatment change episodes were extracted from the EuResist database. The dataset included patient demographics, treatment history and viral genotypes. A logistic regression model using high order interaction variables performed better than rule-based genotypic interpretation systems (accuracy 75.63% versus 71.74–73.89%, area under the receiver operating characteristic curve [AUC] 0.76 versus 0.68–0.70) and was equivalent to a random forest model (accuracy 76.16%, AUC 0.77). However, when rule-based genotypic interpretation systems were coupled with additional patient attributes, and the combination was provided as input to the logistic regression model, the performance increased significantly, becoming comparable to the fully data-driven methods.

Conclusions

Patient-derived supplementary features significantly improved the accuracy of the prediction of response to treatment, both with rule-based and data-driven interpretation systems. Fully data-driven models derived from large-scale data sources show promise as antiretroviral treatment decision support tools.

Evolution and predictors of HIV type-1 drug resistance in patients failing combination antiretroviral therapy in Italy

Background

This study aimed to examine the evolution of genotypic drug resistance prevalence in treatment-failing patients in the multicentre, Italian, Antiretroviral Resistance Cohort Analysis (ARCA).

Methods

Patients with a drug resistance genotype test performed between 1999 and 2006 at failure of a combination antiretroviral therapy and with complete treatment history were selected. The prevalence of resistance was measured overall, per calendar year, per drug class and per treatment line at failure.

Results

The overall resistance prevalence was 81%. Resistance to nucleoside reverse transcriptase inhibitors (NRTIs) declined after 2002 (68% in 2006; χ2 for trend P=0.004); resistance to non-NRTIs (NNRTIs) stabilized after 2004; and resistance to protease inhibitors (PIs) declined after 2001 (43% in 2006; P=0.004). In first-line failures, NRTI resistance decreased after 2002 ( P=0.006), NNRTI resistance decreased after 2003 ( P=0.001) and PI resistance decreased after 2001 ( P<0.001). Independent predictors of resistance to any class were HIV type-1 transmission by heterosexual contacts as compared with injecting drug use, a higher number of experienced regimens, prior history of suboptimal therapy, higher viral load and CD4+ T-cell counts, more recent calendar year and viral subtype B carriage, whereas the use of PI-based versus NNRTI-based regimens at failure was associated with a reduced risk of resistance. There was an increase of type-1 thymidine analogue and of protease mutations L33F, I47A/V, I50V and I54L/M, whereas L90M decreased over calendar years.

Conclusions

During more recent years, emerging drug resistance has decreased, particularly in first-line failures. The prevalence continues to be high in multiregimen-failing patients.

Robust supervised and unsupervised statistical learning for HIV type 1 coreceptor usage analysis

Human immunodeficiency virus type 1 (HIV-1) isolates differ in their use of coreceptors to enter target cells. This has important implications for both viral pathogenicity and susceptibility to entry inhibitors, recently approved or under development. Predicting HIV-1 coreceptor usage on the basis of sequence information is a challenging task, due to the high variability of the envelope. The associations of the whole HIV-1 envelope genetic features (subtype, mutations, insertions-deletions, physicochemical properties) and clinical markers (viral RNA load, CD8(+), CD4(+) T cell counts) with viral tropism were investigated, using a set of 2896 (659 after filter, 593 patients) sequence-tropism pairs available at the Los Alamos HIV database. Bootstrapped hierarchical clustering was used to assess mutational covariation. Univariate and multivariate analysis was performed to assess the relative importance of different features. Different machine learning (logistic regression, support vector machines, decision trees, rule bases, instance based reasoning) and feature selection (filter and embedded) methods, along with loss functions (accuracy, AUC of ROC curves, sensitivity, specificity, f-measure), were applied and compared for the classification of X4 variants. Extra-sample error estimation was assessed via multiple cross-validation and adjustments for multiple testing. A high-performing, compact, and interpretable logistic regression model was derived to infer HIV-1 coreceptor tropism for a given patient [accuracy = 92.76 (SD 3.07); AUC = 0.93 (SD 0.04)].

Advantages of predicted phenotypes and statistical learning models in inferring virological response to antiretroviral therapy from HIV genotype

Background

Inferring response to antiretroviral therapy from the viral genotype alone is challenging. The utility of an intermediate step of predicting in vitro drug susceptibility is currently controversial. Here, we provide a retrospective comparison of approaches using either genotype or predicted phenotypes alone, or in combination.

Methods

Treatment change episodes were extracted from two large databases from the USA (Stanford-California) and Europe (EuResistDB) comprising data from 6,706 and 13,811 patients, respectively. Response to antiretroviral treatment was dichotomized according to two definitions. Using the viral sequence and the treatment regimen as input, three expert algorithms (ANRS, Rega and HIVdb) were used to generate genotype-based encodings and VircoTYPE™ 4.0 (Virco BVBA, Mechelen, Belgium) was used to generate a predicted phenotype-based encoding. Single drug classifications were combined into a treatment score via simple summation and statistical learning using random forests. Classification performance was studied on Stanford- California data using cross-validation and, in addition, on the independent EuResistDB data.

Results

In all experiments, predicted phenotype was among the most sensitive approaches. Combining single drug classifications by statistical learning was significantly superior to unweighted summation ( P<2.2x10-16). Classification performance could be increased further by combining predicted phenotypes and expert encodings but not by combinations of expert encodings alone. These results were confirmed on an independent test set comprising data solely from EuResistDB.

Conclusions

This study demonstrates consistent performance advantages in utilizing predicted phenotype in most scenarios over methods based on genotype alone in inferring virological response. Moreover, all approaches under study benefit significantly from statistical learning for merging single drug classifications into treatment scores.

Recombination analysis and structure prediction show correlation between breakpoint clusters and RNA hairpins in the pol gene of human immunodeficiency virus type 1 unique recombinant forms

Recombination is recognized as a primary force in human immunodeficiency virus type 1 (HIV-1) evolution, increasing viral diversity through reshuffling of genomic portions. The strand-switching activity of reverse transcriptase is required to complete HIV-1 replication and can occur randomly throughout the genome, leading to viral recombination. Some recombination hotspots have been identified and found to correlate with RNA structure or sequence features. The aim of this study was to evaluate the presence of recombination hotspots in the pol gene of HIV-1 and to assess their correlation with the underlying RNA structure. Analysis of the recombination pattern and breakpoint distribution in a group of unique recombinant forms (URFs) detected two recombination hotspots in the pol region. Two stable and conserved hairpins were consistently predicted corresponding to the identified hotspots using six different RNA-folding algorithms on the URF parental strains. These findings suggest that such hairpins may play a role in the higher recombination rates detected at these positions.

Evolution and predictors of HIV type-1 drug resistance in patients failing combination antiretroviral therapy in Italy

BACKGROUND

This study aimed to examine the evolution of genotypic drug resistance prevalence in treatment-failing patients in the multicentre, Italian, Antiretroviral Resistance Cohort Analysis (ARCA).

METHODS

Patients with a drug resistance genotype test performed between 1999 and 2006 at failure of a combination antiretroviral therapy and with complete treatment history were selected. The prevalence of resistance was measured overall, per calendar year, per drug class and per treatment line at failure.

RESULTS

The overall resistance prevalence was 81%. Resistance to nucleoside reverse transcriptase inhibitors (NRTIs) declined after 2002 (68% in 2006; chi(2) for trend P=0.004); resistance to non-NRTIs (NNRTIs) stabilized after 2004; and resistance to protease inhibitors (PIs) declined after 2001 (43% in 2006; P=0.004). In first-line failures, NRTI resistance decreased after 2002 (P=0.006), NNRTI resistance decreased after 2003 (P=0.001) and PI resistance decreased after 2001 (P<0.001). Independent predictors of resistance to any class were HIV type-1 transmission by heterosexual contacts as compared with injecting drug use, a higher number of experienced regimens, prior history of suboptimal therapy, higher viral load and CD4+ T-cell counts, more recent calendar year and viral subtype B carriage, whereas the use of PI-based versus NNRTI-based regimens at failure was associated with a reduced risk of resistance. There was an increase of type-1 thymidine analogue and of protease mutations L33F, I47A/V, I50V and I54L/M, whereas L90M decreased over calendar years.

CONCLUSIONS

During more recent years, emerging drug resistance has decreased, particularly in first-line failures. The prevalence continues to be high in multiregimen-failing patients.

Investigation of expert rule bases, logistic regression, and non-linear machine learning techniques for predicting response to antiretroviral treatment

BACKGROUND

The extreme flexibility of the HIV type-1 (HIV-1) genome makes it challenging to build the ideal antiretroviral treatment regimen. Interpretation of HIV-1 genotypic drug resistance is evolving from rule-based systems guided by expert opinion to data-driven engines developed through machine learning methods.

METHODS

The aim of the study was to investigate linear and non-linear statistical learning models for classifying short-term virological outcome of antiretroviral treatment. To optimize the model, different feature selection methods were considered. Robust extra-sample error estimation and different loss functions were used to assess model performance. The results were compared with widely used rule-based genotypic interpretation systems (Stanford HIVdb, Rega and ANRS).

RESULTS

A set of 3,143 treatment change episodes were extracted from the EuResist database. The dataset included patient demographics, treatment history and viral genotypes. A logistic regression model using high order interaction variables performed better than rule-based genotypic interpretation systems (accuracy 75.63% versus 71.74-73.89%, area under the receiver operating characteristic curve [AUC] 0.76 versus 0.68-0.70) and was equivalent to a random forest model (accuracy 76.16%, AUC 0.77). However, when rule-based genotypic interpretation systems were coupled with additional patient attributes, and the combination was provided as input to the logistic regression model, the performance increased significantly, becoming comparable to the fully data-driven methods.

CONCLUSIONS

Patient-derived supplementary features significantly improved the accuracy of the prediction of response to treatment, both with rule-based and data-driven interpretation systems. Fully data-driven models derived from large-scale data sources show promise as antiretroviral treatment decision support tools.

Advantages of predicted phenotypes and statistical learning models in inferring virological response to antiretroviral therapy from HIV genotype

BACKGROUND

Inferring response to antiretroviral therapy from the viral genotype alone is challenging. The utility of an intermediate step of predicting in vitro drug susceptibility is currently controversial. Here, we provide a retrospective comparison of approaches using either genotype or predicted phenotypes alone, or in combination.

METHODS

Treatment change episodes were extracted from two large databases from the USA (Stanford-California) and Europe (EuResistDB) comprising data from 6,706 and 13,811 patients, respectively. Response to antiretroviral treatment was dichotomized according to two definitions. Using the viral sequence and the treatment regimen as input, three expert algorithms (ANRS, Rega and HIVdb) were used to generate genotype-based encodings and VircoTYPE() 4.0 (Virco BVBA, Mechelen, Belgium) was used to generate a predicted -phenotype-based encoding. Single drug classifications were combined into a treatment score via simple summation and statistical learning using random forests. Classification performance was studied on Stanford-California data using cross-validation and, in addition, on the independent EuResistDB data.

RESULTS

In all experiments, predicted phenotype was among the most sensitive approaches. Combining single drug classifications by statistical learning was significantly superior to unweighted summation (P<2.2x10(-16)). Classification performance could be increased further by combining predicted phenotypes and expert encodings but not by combinations of expert encodings alone. These results were confirmed on an independent test set comprising data solely from EuResistDB.

CONCLUSIONS

This study demonstrates consistent performance advantages in utilizing predicted phenotype in most scenarios over methods based on genotype alone in inferring virological response. Moreover, all approaches under study benefit significantly from statistical learning for merging single drug classifications into treatment scores.

Stochastic modelling of genotypic drug-resistance for human immunodeficiency virus towards long-term combination therapy optimization

MOTIVATION

Several mathematical models have been investigated for the description of viral dynamics in the human body: HIV-1 infection is a particular and interesting scenario, because the virus attacks cells of the immune system that have a role in the antibody production and its high mutation rate permits to escape both the immune response and, in some cases, the drug pressure. The viral genetic evolution is intrinsically a stochastic process, eventually driven by the drug pressure, dependent on the drug combinations and concentration: in this article the viral genotypic drug resistance onset is the main focus addressed. The theoretical basis is the modelling of HIV-1 population dynamics as a predator-prey system of differential equations with a time-dependent therapy efficacy term, while the viral genome mutation evolution follows a Poisson distribution. The instant probabilities of drug resistance are estimated by means of functions trained from in vitro phenotypes, with a roulette-wheel-based mechanisms of resistant selection. Simulations have been designed for treatments made of one and two drugs as well as for combination antiretroviral therapies. The effect of limited adherence to therapy was also analyzed. Sequential treatment change episodes were also exploited with the aim to evaluate optimal synoptic treatment scenarios.

RESULTS

The stochastic predator-prey modelling usefully predicted long-term virologic outcomes of evolved HIV-1 strains for selected antiretroviral therapy combinations. For a set of widely used combination therapies, results were consistent with findings reported in literature and with estimates coming from analysis on a large retrospective data base (EuResist).