2018 update to the HIV-TRePS system: the development of new computational models to predict HIV treatment outcomes, with or without a genotype, with enhanced usability for low-income settings.

Objectives: Optimizing antiretroviral drug combination on an individual basis can be challenging, particularly in settings with limited access to drugs and genotypic resistance testing. Here we describe our latest computational models to predict treatment responses, with or without a genotype, and compare their predictive accuracy with that of genotyping. Methods: Random forest models were trained to predict the probability of virological response to a new therapy introduced following virological failure using up to 50 000 treatment change episodes (TCEs) without a genotype and 18 000 TCEs including genotypes. Independent data sets were used to evaluate the models. This study tested the effects on model accuracy of relaxing the baseline data timing windows, the use of a new filter to exclude probable non-adherent cases and the addition of maraviroc, tipranavir and elvitegravir to the system. Results: The no-genotype models achieved area under the receiver operator characteristic curve (AUC) values of 0.82 and 0.81 using the standard and relaxed baseline data windows, respectively. The genotype models achieved AUC values of 0.86 with the new non-adherence filter and 0.84 without. Both sets of models were significantly more accurate than genotyping with rules-based interpretation, which achieved AUC values of only 0.55-0.63, and were marginally more accurate than previous models. The models were able to identify alternative regimens that were predicted to be effective for the vast majority of cases in which the new regimen prescribed in the clinic failed. Conclusions: These latest global models predict treatment responses accurately even without a genotype and have the potential to help optimize therapy, particularly in resource-limited settings.

Phylogenetic evidence for underreporting of male-to-male sex among human immunodeficiency virus-infected donors in the Netherlands and Flanders.

BACKGROUND: Separate transmission networks for human immunodeficiency virus (HIV) coexist. Molecular typing of viral genomes can provide insight in HIV transmission routes in donors for whom risk behavior-based donor selection failed. STUDY DESIGN AND METHODS: This study includes all HIV-infected Dutch and Flemish donors in the period 2005 to 2014 (n = 55). Part of the HIV polymerase (pol) gene was amplified, sequenced, and compared with more than 10,000 HIV strains obtained from HIV-infected Dutch and Flemish patients. The most likely transmission route was determined based on HIV phylogeny and the donor's self-reported risk behavior during the exit interview. RESULTS: HIV-infected donors were predominantly male (69%), were repeat donors (73%), were born in the Netherlands or Belgium (95%), and harbored HIV Subtype B (68%). Seventy-five percent of HIV-infected male donors were part of robust phylogenetic clusters linked to male-to-male sex, while only 24% of HIV-infected male donors reported male-to-male sex during posttest counseling. Sex between men and women accounted for 13% of HIV infections in male donors and 93% of HIV infections in female donors based on phylogenetic analysis. Only 40% of HIV-infected female donors had HIV Subtype B; 65% of female donors reported a foreign partner and indeed HIV sequences interspersed with sequences from HIV-endemic areas abroad, in particular sub-Saharan Africa. CONCLUSION: HIV typing helps to understand HIV transmission routes in donor populations. We found substantial underreporting of male-to-male sex among HIV-infected male donors. Donor education on HIV risk factors and the danger of window-period donations and a donor environment that encourages frank disclosure of sexual behavior will contribute to a decrease of HIV-infected donors.

An update to the HIV-TRePS system: the development and evaluation of new global and local computational models to predict HIV treatment outcomes, with or without a genotype.

OBJECTIVES: Optimizing antiretroviral drug combination on an individual basis in resource-limited settings is challenging because of the limited availability of drugs and genotypic resistance testing. Here, we describe our latest computational models to predict treatment responses, with or without a genotype, and compare the potential utility of global and local models as a treatment tool for South Africa. METHODS: Global random forest models were trained to predict the probability of virological response to therapy following virological failure using 29 574 treatment change episodes (TCEs) without a genotype, 3179 of which were from South Africa and were used to develop local models. In addition, 15 130 TCEs including genotypes were used to develop another set of models. The 'no-genotype' models were tested with an independent global test set (n = 1700) plus a subset from South Africa (n = 222). The genotype models were tested with 750 independent cases. RESULTS: The global no-genotype models achieved area under the receiver-operating characteristic curve (AUC) values of 0.82 and 0.79 with the global and South African tests sets, respectively, and the South African models achieved AUCs of 0.70 and 0.79. The genotype models achieved an AUC of 0.84. The global no-genotype models identified more alternative, locally available regimens that were predicted to be effective for cases that failed their new regimen in the South African clinics than the local models. Both sets of models were significantly more accurate predictors of outcomes than genotyping with rules-based interpretation. CONCLUSIONS: These latest global models predict treatment responses accurately even without a genotype, out-performed the local South African models and have the potential to help optimize therapy, particularly in resource-limited settings.

An update to the HIV-TRePS system: the development of new computational models that do not require a genotype to predict HIV treatment outcomes.

OBJECTIVES: The optimal individualized selection of antiretroviral drugs in resource-limited settings is challenging because of the limited availability of drugs and genotyping. Here we describe the development of the latest computational models to predict the response to combination antiretroviral therapy without a genotype, for potential use in such settings. METHODS: Random forest models were trained to predict the probability of a virological response to therapy (<50 copies HIV RNA/mL) following virological failure using the following data from 22,567 treatment-change episodes including 1090 from southern Africa: baseline viral load and CD4 cell count, treatment history, drugs in the new regimen, time to follow-up and follow-up viral load. The models were assessed during cross-validation and with an independent global test set of 1000 cases including 100 from southern Africa. The models' accuracy [area under the receiver-operating characteristic curve (AUC)] was evaluated and compared with genotyping using rules-based interpretation systems for those cases with genotypes available. RESULTS: The models achieved AUCs of 0.79-0.84 (mean 0.82) during cross-validation, 0.80 with the global test set and 0.78 with the southern African subset. The AUCs were significantly lower (0.56-0.57) for genotyping. CONCLUSIONS: The models predicted virological response to HIV therapy without a genotype as accurately as previous models that included a genotype. They were accurate for cases from southern Africa and significantly more accurate than genotyping. These models will be accessible via the online treatment support tool HIV-TRePS and have the potential to help optimize antiretroviral therapy in resource-limited settings where genotyping is not generally available.

Monitoring Recently Acquired HIV Infections in Amsterdam, The Netherlands: The Attribution of Test Locations.

Background: Surveillance of recent HIV infections (RHI) using an avidity assay has been implemented at Dutch sexual health centres (SHC) since 2014, but data on RHI diagnosed at other test locations is lacking. Setting: Implementation of the avidity assay in HIV treatment clinics for the purpose of studying RHI among HIV patients tested at different test locations. Methods: We retrospectively tested leftover specimens from newly diagnosed HIV patients in care in 2013-2015 in Amsterdam. Avidity Index (AI) values ≤0.80 indicated recent infection (acquired ≤6 months prior to diagnosis), and AI > 0.80 indicated established infection (acquired >6 months prior to diagnosis). An algorithm for RHI was applied to correct for false recency. Recency based on this algorithm was compared with recency based on epidemiological data only. Multivariable logistic regression analysis was used to identify factors associated with RHI among men who have sex with men (MSM). Results: We tested 447 specimens with avidity; 72% from MSM. Proportions of RHI were 20% among MSM and 10% among heterosexuals. SHC showed highest proportions of RHI (27%), followed by GPs (15%), hospitals (5%), and other/unknown locations (11%) (p < 0.001). Test location was the only factor associated with RHI among MSM. A higher proportion of RHI was found based on epidemiological data compared to avidity testing combined with the RHI algorithm. Conclusion: SHC identify more RHI infections compared to other test locations, as they serve high-risk populations and offer frequent HIV testing. Using avidity-testing for surveillance purposes may help targeting prevention programs, but the assay lacks robustness and its added value may decline with improved, repeat HIV testing and data collection.

Predicting Virological Response to HIV Treatment Over Time: A Tool for Settings With Different Definitions of Virological Response.

OBJECTIVE: Definitions of virological response vary from <50 up to 1000 copies of HIV-RNA/mL. Our previous models estimate the probability of HIV drug combinations reducing the viral load to <50 copies/mL, with no indication of whether higher thresholds of response may be achieved. Here, we describe the development of models that predict absolute viral load over time. METHODS: Two sets of random forest models were developed using 50,270 treatment change episodes from more than 20 countries. The models estimated viral load at different time points following the introduction of a new regimen from variables including baseline viral load, CD4 count, and treatment history. One set also used genotypes in their predictions. Independent data sets were used for evaluation. RESULTS: Both models achieved highly significant correlations between predicted and actual viral load changes (r = 0.67-0.68, mean absolute error of 0.73-0.74 log10 copies/mL). The models produced curves of virological response over time. Using failure definitions of <100, 400, or 1000 copies/mL, but not 50 copies/mL, both models were able to identify alternative regimens they predicted to be effective for the majority of cases where the new regimen prescribed in the clinic failed. CONCLUSIONS: These models could be useful for selecting the optimum combination therapy for patients requiring a change in therapy in settings using any definition of virological response. They also give an idea of the likely response curve over time. Given that genotypes are not required, these models could be a useful addition to the HIV-TRePS system for those in resource-limited settings.

Limited overlap between phylogenetic HIV and hepatitis C virus clusters illustrates the dynamic sexual network structure of Dutch HIV-infected MSM.

OBJECTIVE: MSM are at increased risk for infection with HIV-1 and hepatitis C virus (HCV). Is HIV/HCV coinfection confined to specific HIV transmission networks? DESIGN AND METHODS: A HIV phylogenetic tree was constructed for 5038 HIV-1 subtype B polymerase (pol) sequences obtained from MSM in the AIDS therapy evaluation in the Netherlands cohort. We investigated the existence of HIV clusters with increased HCV prevalence, the HIV phylogenetic density (i.e. the number of potential HIV transmission partners) of HIV/HCV-coinfected MSM compared with HIV-infected MSM without HCV, and the overlap in HIV and HCV phylogenies using HCV nonstructural protein 5B sequences from 183 HIV-infected MSM with acute HCV infection. RESULTS: Five hundred and sixty-three of 5038 (11.2%) HIV-infected MSM tested HCV positive. Phylogenetic analysis revealed 93 large HIV clusters (≥10 MSM), 370 small HIV clusters (2-9 MSM), and 867 singletons with a median HCV prevalence of 11.5, 11.6, and 9.3%, respectively. We identified six large HIV clusters with elevated HCV prevalence (range 23.5-46.2%). Median HIV phylogenetic densities for MSM with HCV (3, interquartile range 1-7) and without HCV (3, interquartile range 1-8) were similar. HCV phylogeny showed 12 MSM-specific HCV clusters (clustersize: 2-39 HCV sequences); 12.7% of HCV infections were part of the same HIV and HCV cluster. CONCLUSION: We observed few HIV clusters with elevated HCV prevalence, no increase in the HIV phylogenetic density of HIV/HCV-coinfected MSM compared to HIV-infected MSM without HCV, and limited overlap between HIV and HCV phylogenies among HIV/HCV-coinfected MSM. Our data do not support the existence of MSM-specific sexual networks that fuel both the HIV and HCV epidemic.

The genotypic inhibitory quotient and the (cumulative) number of mutations predict the response to lopinavir therapy.

For 95 protease inhibitor-experienced HIV-1-infected patients, the genotypic inhibitory quotient (GIQ; trough level/number of mutations) was calculated for lopinavir. Three different sets of mutations showed equal predictive value. However, the use of cumulative numbers of mutations for calculation of the GIQ showed significantly better association with the virological response. Furthermore, the predictive value of the GIQ was no different from that of the number of mutations alone.

Comparison of the risks of atherosclerotic events versus death from other causes associated with antiretroviral use.

BACKGROUND: Studies considering the risk of atherosclerotic disease (AtD) associated with the use of HAART have reported inconsistent results. METHODS: Data on antiretroviral therapy (ART) use, risk factors for cardiovascular disease (CVD), AtD and death from other causes in 18 603 HIV-infected patients from two established cohorts were evaluated. The relative hazards of AtD and death from other causes were calculated using a proportional hazards competing risks framework. The impact of protease inhibitor (PI)-containing, non-nucleoside reverse transcriptase inhibitor (NNRTI)-containing or PI + NNRTI-containing regimens on these outcomes were compared to nucleoside reverse transcriptase inhibitor (NRTI)-only regimens or stopping therapy, adjusting for known CVD risk factors. RESULTS: In 77 480 person-years of follow-up (median duration 3.49 years) there were 318 AtD events including 92 myocardial infarctions and 2044 deaths. Older age, hypertension, diabetes mellitus, having smoked and HIV disease stage were significantly associated with increased risk of AtD. PI- and NNRTI-containing regimens significantly reduced the joint risk of either AtD or death from other causes compared to NRTI-only or stopping therapy [hazard ratio (HR) for PI-containing ART, 0.76, 95% confidence interval (CI), 0.73-0.78, P< 0.001; NNRTI-containing ART, 0.69, 95% CI, 0.65-0.74; P< 0.001). PI-containing ART was associated with a borderline significant increased risk of myocardial infarction (cause-specific HR for PI-containing ART 1.19, 95% CI, 1.01-1.40, P = 0.04) but not with increased risk of AtD compared to NRTI-only regimens or stopping therapy (cause-specific HR for PI-containing ART, 1.03, 95% CI, 0.95-1.13, P = 0.44). CONCLUSIONS: Overall benefits of PI- and NNRTI-based ART in reducing mortality significantly outweigh any risks of AtD in the "short-term" follow-up of this study. Traditional cardiac risk factors play an important role in determining AtD risk status.

Mortality and progression to AIDS after starting highly active antiretroviral therapy.

OBJECTIVES: To examine survival and progression to AIDS among HIV-infected patients after starting highly active antiretroviral therapy (HAART). METHODS: The study population consisted of 3724 patients from the ATHENA observational cohort who initiated HAART. We considered progression to either an AIDS-defining disease or death, distinguishing HIV-related and non-related (including therapy-related) deaths. A time-dependent multivariate hazards model was fitted to the patient data and 5-year survival probabilities under various therapy scenarios estimated. RESULTS: A total of 459 patients developed AIDS and 346 died during 12 503 person-years of follow-up. HIV-related mortality decreased from 3.8 to 0.7 per 100 person-years between 1996 and 2000 whereas non-HIV-related mortality did not change (0.4 and 0.9, respectively, P = 0.25). For asymptomatic and symptomatic therapy naive patients younger than 50 years with CD4 counts above 10 x 10(6) and 150 x 10(6) cells/l, respectively, predicted 5-year survival probabilities were above 90% when HAART was used continuously. This limit was 450 x 10(6) cells/l when HAART was used during 20 weeks in each 24 week-period of follow-up, and 110 x 10(6) cells/l when patients delayed initiation of HAART for 1 year after becoming eligible for treatment. CONCLUSIONS: Survival probabilities were high among HIV-infected patients initiating HAART at an early stage of infection. The best therapy strategy is therefore to start HAART at this stage of infection. However, deferring HAART in patients with high CD4 cell counts may be clinically more appropriate given toxicity and adherence problems. The lack of any change in non-HIV-related mortality suggests that toxicity has not yet become a major risk factor for death.

Single nucleotide polymorphism in gene encoding transcription factor Prep1 is associated with HIV-1-associated dementia.

BACKGROUND: Infection with HIV-1 may result in severe cognitive and motor impairment, referred to as HIV-1-associated dementia (HAD). While its prevalence has dropped significantly in the era of combination antiretroviral therapy, milder neurocognitive disorders persist with a high prevalence. To identify additional therapeutic targets for treating HIV-associated neurocognitive disorders, several candidate gene polymorphisms have been evaluated, but few have been replicated across multiple studies. METHODS: We here tested 7 candidate gene polymorphisms for association with HAD in a case-control study consisting of 86 HAD cases and 246 non-HAD AIDS patients as controls. Since infected monocytes and macrophages are thought to play an important role in the infection of the brain, 5 recently identified single nucleotide polymorphisms (SNPs) affecting HIV-1 replication in macrophages in vitro were also tested. RESULTS: The CCR5 wt/Δ32 genotype was only associated with HAD in individuals who developed AIDS prior to 1991, in agreement with the observed fading effect of this genotype on viral load set point. A significant difference in genotype distribution among all cases and controls irrespective of year of AIDS diagnosis was found only for a SNP in candidate gene PREP1 (p = 1.2 × 10(-5)). Prep1 has recently been identified as a transcription factor preferentially binding the -2,518 G allele in the promoter of the gene encoding MCP-1, a protein with a well established role in the etiology of HAD. CONCLUSION: These results support previous findings suggesting an important role for MCP-1 in the onset of HIV-1-associated neurocognitive disorders.

National estimate of HIV prevalence in the Netherlands: comparison and applicability of different estimation tools.

OBJECTIVES: to determine limitations and strengths of three methodologies developed to estimate HIV prevalence and the number of people living with HIV/AIDS (PLWHA). METHODS: the UNAIDS/WHO Workbook method; the Multiparameter Evidence Synthesis (MPES) adopted by the Health Protection Agency; and the UNAIDS/WHO Estimation and Projection Package (EPP) and Spectrum method were used and their applicability and feasibility were assessed. All methods estimate the number infected in mutually exclusive risk groups among 15-70-year-olds. RESULTS: using data from the Netherlands, the Workbook method estimated 23 969 PLWHA as of January 2008. MPES estimated 21 444 PLWHA, with a 95% credible interval (CrI) of 17 204-28 694. Adult HIV prevalence was estimated at 0.2% (95% CrI 0.15-0.24%) and 40% (95% CrI 25-55%) were undiagnosed. Spectrum applied gender-specific mortality, resulting in a projected estimate of 19 115 PLWHA. CONCLUSION: although outcomes differed between the methods, they broadly concurred. An advantage of MPES is that the proportion diagnosed can be estimated by risk group, which is important for policy guidance. However, before MPES can be used on a larger scale, it should be made more easily applicable. If the aim is not only to obtain annual estimates, but also short-term projections, then EPP and Spectrum are more suitable. Research into developing and refining analytical tools, which make use of all available information, is recommended, especially HIV diagnosed cases, as this information is becoming routinely collected in most countries with concentrated HIV epidemics.

Rising HIV-1 viral load set point at a population level coincides with a fading impact of host genetic factors on HIV-1 control.

OBJECTIVE: Heterozygosity for a 32 base pair deletion in the CCR5 gene (CCR5wt/Δ32) and the minor alleles of a single-nucleotide polymorphism in the HCP5 gene (rs2395029) and in the HLA-C gene region (-35HLA-C; rs9264942) has been associated with a lower viral load set point. Recent studies have shown that over calendar time, viral load set point has significantly increased at a population level. Here we studied whether this increase coincides with a fading impact of above-mentioned host genetic markers on HIV-1 control. METHODS: We compared the association between viral load set point and HCP5 rs2395029, -35HLA-C rs9264942, and the CCR5wt/Δ32 genotype in HIV-1-infected individuals in the Netherlands who had seroconverted between 1982 and 2002 (pre-2003 seroconverters, n = 459) or between 2003 and 2009 (post-2003 seroconverters, n = 231). RESULTS: Viral load set point in post-2003 seroconverters was significantly higher than in pre-2003 seroconverters (P = 4.5 × 10(-5)). The minor alleles for HCP5 rs2395029, -35HLA-C rs9264942 and CCR5wt/Δ32 had a similar prevalence in both groups and were all individually associated with a significantly lower viral load set point in pre-2003 seroconverters. In post-2003 seroconverters, this association was no longer observed for HCP5 rs2395029 and CCR5wt/Δ32. The association between viral load set point and HCP5 rs2395029 had significantly changed over time, whereas the change in impact of the CCR5wt/Δ32 genotype over calendar time was not independent from the other markers under study. CONCLUSION: The increased viral load set point at a population level coincides with a lost impact of certain host genetic factors on HIV-1 control.

HIV transmission patterns among The Netherlands, Suriname, and The Netherlands Antilles: a molecular epidemiological study.

We aimed to study patterns of HIV transmission among Suriname, The Netherlands Antilles, and The Netherlands. Fragments of env, gag, and pol genes of 55 HIV-infected Surinamese, Antillean, and Dutch heterosexuals living in The Netherlands and 72 HIV-infected heterosexuals living in Suriname and the Antilles were amplified and sequenced. We included 145 pol sequences of HIV-infected Surinamese, Antillean, and Dutch heterosexuals living in The Netherlands from an observational cohort. All sequences were phylogenetically analyzed by neighbor-joining. Additionally, HIV-1 mobility among ethnic groups was estimated. A phylogenetic tree of all pol sequences showed two Surinamese and three Antillean clusters of related strains, but no clustering between ethnic groups. Clusters included sequences of individuals living in Suriname and the Antilles as well as those who have migrated to The Netherlands. Similar clustering patterns were observed in env and gag. Analysis of HIV mobility among ethnic groups showed significantly lower migration between groups than expected under the hypothesis of panmixis, apart from higher HIV migration between Antilleans in The Netherlands and all other groups. Our study shows that HIV transmission mainly occurs within the ethnic group. This suggests that cultural factors could have a larger impact on HIV mobility than geographic distance.

Effect of transmitted drug resistance on virological and immunological response to initial combination antiretroviral therapy for HIV (EuroCoord-CHAIN joint project): a European multicohort study.

BACKGROUND: The effect of transmitted drug resistance (TDR) on first-line combination antiretroviral therapy (cART) for HIV-1 needs further study to inform choice of optimum drug regimens. We investigated the effect of TDR on outcome in the first year of cART within a large European collaboration. METHODS: HIV-infected patients of any age were included if they started cART (at least three antiretroviral drugs) for the first time after Jan 1, 1998, and were antiretroviral naive and had at least one sample for a genotypic test taken before the start of cART. We used the WHO drug resistance list and the Stanford algorithm to classify patients into three resistance categories: no TDR, at least one mutation and fully-active cART, or at least one mutation and resistant to at least one prescribed drug. Virological failure was defined as time to the first of two consecutive viral load measurements over 500 copies per mL after 6 months of therapy. FINDINGS: Of 10,056 patients from 25 cohorts, 9102 (90·5%) had HIV without TDR, 475 (4·7%) had at least one mutation but received fully-active cART, and 479 (4·8%) had at least one mutation and resistance to at least one drug. Cumulative Kaplan-Meier estimates for virological failure at 12 months were 4·2% (95% CI 3·8-4·7) for patients in the no TDR group, 4·7% (2·9-7·5) for those in the TDR and fully-active cART group, and 15·1% (11·9-19·0) for those in the TDR and resistant group (log-rank p<0·0001). The hazard ratio for the difference in virological failure between patients with TDR and resistance to at least one drug and those without TDR was 3·13 (95% CI 2·33-4·20, p<0·0001). The hazard ratio for the difference between patients with TDR receiving fully-active cART and patients without TDR was 1·47 (95% CI 0·19-2·38, p=0·12). In stratified analysis, the hazard ratio for the risk of virological failure in patients with TDR who received fully-active cART that included a non-nucleoside reverse transcriptase inhibitor (NNRTI) compared with those without TDR was 2·0 (95% CI 0·9-4·7, p=0·093). INTERPRETATION: These findings confirm present treatment guidelines for HIV, which state that the initial treatment choice should be based on resistance testing in treatment-naive patients. FUNDING: European Community's Seventh Framework Programme FP7/2007-2013 and Gilead.

Life expectancy of recently diagnosed asymptomatic HIV-infected patients approaches that of uninfected individuals.

OBJECTIVE: To compare life expectancies between recently diagnosed HIV-infected patients and age and sex-matched uninfected individuals from the general population. DESIGN: : National observational HIV cohort in the Netherlands. METHODS: Four thousand, six hundred and twelve patients diagnosed with HIV between 1998 and 2007 and still antiretroviral therapy-naive as of 24 weeks after diagnosis were selected. Progression to death compared to the age and sex-matched general population was studied with a multivariate hazards model in 4174 (90.5%) patients without AIDS events at 24 weeks. Life expectancy and number of life years lost were calculated using the predicted survival distribution. RESULTS: During 17 580 person-years of follow-up since 24 weeks after diagnosis [median follow-up 3.3 years, interquartile range (IQR) 1.6-5.8], 118 deaths occurred, yielding a mortality rate of 6.7 [95% confidence interval (CI) 5.5-8.0] per 1000 person-years. Median CD4 cell counts at 24 weeks were 480 cells/microl (IQR 360-650). According to the model, the median number of years lived from age 25 was 52.7 (IQR 44.2-59.3; general population 53.1) for men and 57.8 (49.2-63.7; 58.1) for women without CDC-B event. The number of life years lost varied between 0.4 if diagnosed with HIV at age 25 and 1.4 if diagnosed at age 55; for patients with a CDC-B event this range was 1.8-8.0 years. CONCLUSION: The life expectancy of asymptomatic HIV-infected patients who are still treatment-naive and have not experienced a CDC-B or C event at 24 weeks after diagnosis approaches that of non-infected individuals. However, follow-up time is short compared to the expected number of years lived.

CD4 cell counts of 800 cells/mm3 or greater after 7 years of highly active antiretroviral therapy are feasible in most patients starting with 350 cells/mm3 or greater.

OBJECTIVE: CD4 cell count changes in therapy-naive patients were investigated during 7 years of highly active antiretroviral therapy (HAART) in an observational cohort. METHODS: Three endpoints were studied: (1) time to >or=800 CD4 cells/mm in 5299 therapy-naive patients starting HAART, (2) CD4 cell count changes during 7 years of uninterrupted HAART in a subset of 544 patients, and (3) reaching a plateau in CD4 cell restoration after 5 years of HAART in 366 virologically suppressed patients. RESULTS: Among patients with <50, 50 to 200, 200 to 350, 350 to 500, and >or=500 CD4 cells/mm at baseline, respectively, 20%, 26%, 46%, 73%, and 87% reached >or=800 CD4 cells/mm within 7 years of starting HAART. Periods with HIV RNA levels >500 copies/mL and age >or=50 years were associated with lesser increases in CD4 cell counts between 6 months and 7 years. Having reached >or=800 CD4 cells/mm at 5 years, age >or=50 years, and >or=1 HIV RNA measurement >1000 copies/mL between 5 and 7 years were associated with a plateau in CD4 cell restoration. CONCLUSIONS: Restoration to CD4 cell counts >or=800 cells/mm is feasible within 7 years of HAART in most HIV-infected patients starting with >or=350 cells/mm and achieving sufficient suppression of viral replication. Particularly in patients >or=50 years of age, it may be beneficial to start earlier than current guidelines recommend.