Simon Collins, HIV i-Base

Results from a new online resource developed to predict treatment outcomes for settings without access to genotypic resistance tests were presented in a poster at the conference. [1]

The system was developed by training computer models to predict virological response to therapy using data from approximately 15,000 treatment changes drawn from over 15 countries. The models use CD4, viral load, treatment history and the drugs in the new regimen in making their predictions and can generate predictions of response at selected time points out to 48 weeks for all available combinations or for a selected combination. The system includes the option to select drugs that are available in each country and to exclude drugs that are contraindicated.

The accuracy of the models was assessed with an independent test set of 800 cases. Two further test sets from Romania (n=39) and South Africa (n=56) were also reported together with subset of 57 cases from the 800 test set that had genotypes available.

The mean area under the curve and overall accuracy were 0.77 and 71% with the 800 test dataset (with similar results during cross validation). The comparable results were 0.68 and 67% for the Romanian and 0.69 and 68% for the South African test sets respectively. When the 57 case test set was used to compare the performance of the models with and without genotype information the results were 0.77 and 74% using the genotype, compared to 0.76 and 68% for the ‘no-genotype’ models.

The models are now available via the RDI’s online treatment selection tool HIV-TRePS. Importantly, the resource includes the option to include, with permission, anonymised information on treatment decisions and outcomes to be collected to help further development of the system. [2]

The resource has been developed by researchers at RDI who were involved in much of the original pioneering work into HIV drug resistance technology and more recently have been developing prediction tools to interpret genotype results using computer-developed neural networks.

Future reports on how this resource is used in practice will be important given the extremely restricted access to resistance testing in most resource-limited countries and that this is unlikely to change in the near future.

References

1. Larder BA et al. Predicting response to antiretroviral therapy without a genotype: a treatment tool for resource-limited settings. 16th IAS Conference on HIV Pathogenesis, Treatment and Prevention, 17–20 July 2011, Rome. Poster MOPE146.
2. The resource can be accessed free online after one-time free registration.

http://www.hivrdi.org

A poster by Chalwarith and colleagues at CROI presented the results from a retrospective comparison of two Thai cohorts to determine if there were differences in virological failure and resistance mutation rates in once-yearly versus twice-yearly viral load monitoring. [1]

Table 1: Mutations in a group of patients in the two cohorts. From Chalwarith et al.

The study compared 424 patients who received a viral load test annually via the Universal Health Coverage Programme to 154 patients who received two viral loads a year via the Social Security Health Programme. All patients were on stable HAART at the Chiang Mai University Hospital. There were no significant differences in measured baseline characteristics. Men comprised 46% of the sample. The mean age was 40 years. Nearly 98% of patients were on two NRTIs plus an NNRTI. Median CD4 count was 60 cells/mm3 (IQR: 30-138 cells/mm3).

The investigators found no significant differences in incidence of virological failure (defined as viral load >1000 copies/mL) between the two cohorts. The rate in the twice-yearly group was 5.3% (8 patients; 3.08 per 100,000 person days) versus 8% (34 patients; 4.32/100,000 patient days) in the once-yearly group. The hazard ratio for the once-yearly group was 1.37 (95%CI: 0.63-2.95; p=0.428). Neither did sex, age or baseline CD4 count predict virological failure. However, adherence, measured as total doses taken over total prescribed, was predictive (HR: 0.01; 95%CI: 0.01-0.03; p<0.001).

Patterns of NRTI mutations also did not differ between the two groups. Table 1 shows the mutations that were detected in a subset of patients.

comments

The authors conclude that once-yearly viral load monitoring in resource-limited settings is justified. This is an observational comparison and there may be other differences between the cohorts that could easily confound the comparison. Also, numbers are small so the lack of difference could be due to a type II error. So the conclusion the authors draw may be too strong based on those data. However, following the findings of DART, this study may add to the argument that (as a number of experts suggest) once-yearly viral loads are a good compromise, where feasible, between the need for minimising patient risk and maximising the number of people on treatment.

Ideally large studies are needed before we can be comfortable that we know what the differences in outcome from different monitoring strategies are.

Ref: Chalwarith R et al. Frequency of plasma HIV RNA monitoring: Impact on outcome of antiretroviral therapy. 17th CROI. 16-19 February 2010. Poster abstract 500.

http://www.retroconference.org/2010/Abstracts/37132.htm

Making switching decisions based on clinical monitoring, as happens in many resource-limited settings, has led to concern about the potential widespread transmission of drug resistant virus in populations where antiretroviral options are limited.

A poster at CROI, authored by Andrew Phillips and colleagues showed results from a computer simulation model designed to predict transmission of drug resistance according to the monitoring strategy used as the basis for guiding switches to second line treatment.

The investigators modelled a scenario using a stochastic simulation model of a high prevalence heterosexual epidemic beginning in 1985 and introducing treatment in 2003 for people with WHO stage 4/<200 cells/mm3 CD4. The first line regimen used in the model was d4T+3TC+NVP and second line AZT+ddI+LPV/r.

This was used to predict the proportion of new infections with transmitted drug resistant virus from 2010 to 2020 according to the timing of introduction of 6 monthly viral load monitoring (based on threshold of 500 copies/mL) to guide switching from first to second line.

In 2010, it was assumed that 20% people with HIV were diagnosed and 12% on ART (44% coverage).  The authors made an optimistic assumption that the diagnosis rate would increase after 2010 and treatment started at CD4 <350 cells/mm3.

The model predicted that the levels of transmitted resistance from introducing viral load in 2010 or 2015 would be similar after 2015. The predicted proportion of newly infected people with transmitted drug resistance in 2020 was 5.4% if viral load monitoring were introduced in 2010, 6.1% if introduced in 2015, and 12.4% if clinical WHO Stage 4 monitoring were used throughout.

When a viral load threshold of 5000 copies/mL, instead of 500 copies/mL, for six monthly monitoring was used, the predicted proportion increased to 6.0% in 2020.

Using viral load monitoring only once every 3 years, with the first at year 1, predicted a value of 7.2%, while use of a single viral load measurement at 1 year (and no subsequent measures) a value of 8.5%.

Looking at the death rate in the HIV-positive population the values predicted were: 2.7 (2.4 for people on treatment) 2.8 (2.7) and 3.1 (3.3) per 100 person years, with use of viral load monitoring from 2010, 2015 and use of WHO 4 clinical monitoring throughout the period, respectively. (2.9 [2.7] per 100 person years with WHO 3 or 4 clinical monitoring).

But, if the assumed increase in diagnosis and coverage since 2010 did not take place, the death rate was predicted to be 4.7 per 100 person years, even with use of viral load.

The authors wrote: “There is a long term need for introduction of some form of cheap, practical, and sustainable viral load monitoring in resource limited settings which can be used in rural as well as urban settings. These tests do not need to be able to do more than distinguish those with viral load levels of above and below some low threshold such as 500 copies/mL.”

They added, “Our results also indicate that even very infrequent (eg 3 yearly) testing, is likely to provide significant benefit in reducing resistance transmission”.

Ref: Phillips A et al. Predicted effect on transmission of HIV-1 resistance of timing of implementation of viral load monitoring to determine switches from first- to second-line antiretroviral regimens in resource-limited settings. 17th CROI. 16-19 February 2010. Poster abstract 596.

http://www.retroconference.org/2010/Abstracts/38818.htm

First line regimens in resource limited settings (RLS) – as currently recommended by WHO – are usually two nucleosides, 3TC plus a thymidine analogue (TA) either d4T or AZT, and one NNRTI.

Most programmes have limited access to virological monitoring and genotype resistance testing. Because of this most treatment switches are based on clinical or immunological failure.

A considerable number of patients are expected to receive failing TA containing regimens for extended periods before switching to second line. Since the nucleoside drugs in second line regimens may be compromised by presence of TAMS there is concern over the consequences of accumulation of TAMS before switching.

Alessandro Cozzi-Lepri and investigators for the EuroSIDA Study Group used European cohort data to estimate the rate and predictors of accumulation of TA mutations (TAMS) in patients who continue to receive failing regimens. In an article published in the 1 September 2009 issue of the Journal of Infectious Diseases they report lower than anticipated accumulation of TAMs in patients experiencing virological failure.

The investigators analysed data from patients in the EuroSIDA study who experienced virological failure (defined as first viral load >=500 copies/mL after >=6 months), with >= 2 genotype resistance tests (GRTs) while receiving the same TA-containing regimen, with a viral load of >500 copies at both. The time of the first genotype test results in a pair was defined as t0, the date of the very first genotype used in the analysis as baseline-t0.

In this analysis, the majority (87%) of genotype results were obtained retrospectively from stored samples.

The rate of TAM accumulation was calculated as the number of TAMs detected at t1 that were not present at t0 divided by the interval between t0 and t1. The investigators used a multivariate Poisson regression model to identify independent predictors of TAM accumulation.

They also simulated a scenario in which all patients studied were switched to a WHO recommended second line nucleosides (eg AZT+ddI or ABC+ddI) after the extended period on failing TA-containing HAART. This was used to estimate the decrease in susceptibility of subsequent regimens due the accumulation of TAMs.

The study population of 339 patients provided 603 pairs of GRTs. At t0 their median age was 39 years and 14% were female. Of this group 67% had one pair of GRTs; 18% had two; 6% had 3 and 9% more than three pairs of GRTs. Their median viral load was 4.11 log copies/mL and CD4 244 cells/mm³. They were very treatment experienced, 53% had failed 1-3 drugs before baseline t-0 and the remainder 4 or more drugs; 35% had failed an NNRTI and 72% a PI.

During the interval t0-t1 (median 6 months, range 1-89 months) the investigators reported the patients having very stable viral loads (mean absolute change +0.03 log copies/mL, 95% CI -0.3 to +0.09, p=0.29) and CD4 counts (mean absolute change -5.74 cells/mm³, 95% CI -2.52 to +14, p=0.17).

Over t0-t1 all patients were receiving either AZT or d4T, which they received for a median of 9 and 15 months duration respectively from virological failure to t1. Twenty-nine percent received an AZT-containing regimen (176 pairs) and 71% a d4T-containing regimen (427 pairs). Besides the TA, the majority (70%) of patients were receiving 3TC at t0. Other frequently used nucleosides were ddI (25%) and abacavir (18%). The most common NNRTIs were NVP (34%), EFV (18%), but some patients were also receiving PIs, NFV (19%), IDV (26%) and LPV (9%). The investigators noted frequent switching in the drugs besides the TA between t0 and t1. In 478 (79%) patients, more than 1 drug used at t0 was no longer used at t1.

At t0, 90% of the study population had at least one TAM and a median of 3 (range 0-6). Of these 81% had TAM profile 1 (TAM1) – 41L, 210W and 215F mutations, and 62% TAM profile 2 (TAM2) – 67N, 70R and 219EQ; 65% had 41L and 68% 215Y TAM1 mutations and 52% 67N TAM2 mutations.

At t1 93% had at least one new TAM. The investigators noted that he rate of accumulation of TAM1 mutations was twice as fast as that of TAM2.

Between t0 and t1, 126 additional TAMs were accumulated during 548 patient years of follow up (PYFU), which the investigators estimated to give a rate of 0.23 per year (95% CI 0.20-0.27) or, in other words, 1 in 4.3 years (95% CI 3.7-5.0).

The rate was faster (0.3 per year) in the subset (330 pairs) with 0-3 TAMs at t0 and was slower, with a rate of 0.11 per year in the patients who already had 4-5 TAMs at baseline (245 pairs).

Using the Rega IS and the ANRS systems the investigators predicted the response to subsequent WHO recommended nucleoside pairs. Both systems appeared to show that regimens containing tenofovir (particularly with 3TC) were likely to have the greatest activity at t0 and the least reduction in activity t0-t1. These predictions however depend on the accuracy of current expert knowledge regarding which mutations may reduce susceptibility to tenofovir.

When they looked at predictors of TAM accumulation, they found that also greater susceptibility to non thymidine analogues in the failing regimen was associated with faster accumulation of TAMs (50% faster per additional active drug, RR 1.5 [95% CI,1.05-2.14], p=0.02).

Other predictive factors were acquisition of HIV through heterosexual contact (vs homosexual almost 2-fold difference in rate RR1.89 [95%CI 1.01-3.57] p=0.05) and TAM2 profiles at t0 (vs TAM1, 87% faster, RR 1.87 [95% CI 1.06-3.27], p=0.03). NNRTI+PI or PI based regimens at t0 were associated with slower accumulation of TAMs (RR 0.32 [95% CI, 0.12-0.84], p=0.02).

The investigators concluded that their data suggest, “In patients who continue to receive TA-based, virologically failing regimens, the rate of accumulation of TAMS is relatively slow, on average, though the higher the initial predicted activity of the regimen, the faster the rate at which TAMs accumulate. Nucleoside pairs including tenofovir, although expensive, seem more likely to be active against viruses harbouring TAMs and also to experience the highest drop of activity in the face of TAM accumulation. Additional research in this area is needed to inform programme planning in RLS.”

COMMENT

That two distinct pathways of TAMs can emerge under pressure of TA-containing HAART that is not fully suppressive is well described. TAM 1 has been associated with high-level resistance to AZT and most other NRTIs, including tenofovir and abacavir and TAM2 with lower levels of resistance to TDF and other NRTIs.

The finding that the rate of emergence of TAMs was slower than expected in this estimation by Cozzi-Lepri and colleagues is reassuring for programmes with limited access to monitoring and, alongside DART results, will make a big contribution to ongoing discussions about “What to measure?” “How often?” and “What are the consequences?”

The authors note that only 9% of their patients had non-B subtypes and that 24% were receiving WHO recommended first line regimens, which could limit the extent to which their results might be generalisable to patients in RLS. However, they suggest that the similarities between their estimation and that observed in RLS may make this bias negligible. They also were not able to establish an explanation why patients in EuroSIDA were left on failing regimens from these data, and so could not rule out selection bias.

While the average rate accumulation of TAMs is relatively slow and suggests a public health approach would be good, there still needs to be work on identifying why some patients do fail fast. Is it a function of the virus? The drug selection? Genes? What monitoring is needed to catch the small percentage of patients that don’t respond?

While the average rate accumulation of TAMs is relatively slow and suggests a public health approach would be good, there still needs to be work on identifying why some patients fail more rapidly and what monitoring is needed to catch the small percentage of patients that don’t respond?

One of the main predictors of faster accumulation suggested by this analysis (and others) was a function of the virus and drug selection. For example, the greater the amount of resistance already accumulated at the time of failure the slower the rate of accumulation of additional mutations.

And, as the authors stress “all possible efforts should be continued to increase the availability of drug options in RLS.”

Ref:
Cozzi-Lepri A et al. Rate of accumulation of thymidine analogue mutations in patients continuing to receive virologically failing regimens containing zidovudine or stavudine: implications for antiretroviral therapy programs in resource limited settings. J Infectious Dis 200: 687-97, 1 September 2009.