This chapter focussed on the analysis of the virological and CD4 count response to HAART in the two cohorts.
In both cohorts by 48 weeks of therapy, >80% of patients had achieved a virai load ^500 copies/mL. The initial rate of virological response appeared to be slower in I.Co.N.A. compared to UNIF however this is likely to be explained by the much higher frequency of viral load monitoring in UNIF. Thus, for example, the proportion of patients with viral load s500 copies/mL by week 4 in I.Co.N.A is expected to be low simply because very few patients would have a viral load measured so early after HAART initiation. The estimated proportion of patients who achieved a viral load ^50 copies/mL by 48 weeks of therapy in UNIF was, however, much higher than that calculated for the I.Co.N.A. patients. The reason for this difference is less clear considering that the analysis of I.Co.N.A patients has been performed using only viral load measurements available after January 1, 1998. However, it is worth noting that I.Co.N.A. is a multi-centre study so that ultra-sensitive assays could have been introduced at later time points in some of the centres, and this differential “time of use" of ultra-sensitive assays may have affected the Kaplan-Meier estimate of the rate of patients with viral load suppressed s50 copies/mL at the different time points. Indeed, a consistent proportion of viral loads collected during the follow-up (n=881 of 2,390 total viral load measurements, 36.9%) had been measured using assays with cut-offs greater than 50 copies/mL even in patients who started HAART after January 1, 1998 (data not shown).
A summary of the most important predictors identified in the two cohorts is provided in Table 3.5.1. Patients who started HAART when they were antiretroviral-naïve are those who have the best chance of experiencing a good virological response to therapy, both in terms of initial suppression of viral load and of maintaining this suppression. In these analyses, around 90% of antiretroviral-naïve patients achieved a viral load below 500 copies/mL (and about 80% below 50 copies/mL) by two years of treatment with HAART. Furthermore, oniy 30% of these patients experienced a viral rebound above 500 copies/mL (about 50% above 50 copies/mL) by three years from the date of initial suppression. This finding can be explained by the fact that patients who started HAART after being previously exposed to mono-dual therapy are likely to harbour viruses that are less susceptible to antiretroviral therapy or this may simply reflect poorer adherence to therapy in this group of patients. As mentioned before in this thesis, both drug resistance and poor adherence are well established as key
factors affecting the chance of initial virological response as well as the chance of viral
rebound'="'=^"°^'^^.
Some patients who started HAART soon after the introduction of protease inhibitors suffered from receiving sub-optimal combinations (typically including saquinavir-HG as the only PI), that is known to have low bio-availability^®'^’^®®. Despite the smaller number of patients in UNIF starting HAART with saquinavir HG as the only PI the results were consistent in the analyses of both cohorts emphasizing that patients who started these sub-optimal HAART combinations had a lower chance of achieving and maintaining virological success. This result has also been shown by a number of other analyses of observational cohort studies 174-176,181,191.193
Among the I.Co.N.A patients, regimens containing three NRTIs (typically including zidovudine, lamivudine and abacavir) appeared to be associated with a better response in the short term compared to the standard Pl-containing HAART. This is an interesting result considering that OT analyses could favour regimens that are associated with a higher rate of toxicity and consequently worse adherence (see chapter 2). However, the long-term consequences of being exposed to a regimen which to date, especially in patients with high viral loads, has not been convincingly shown to be as virologically potent as a PI- or NNRTI-containing regimens needs to be taken into account. Indeed, these same data show that the chance of virological rebound could be much higher in these patients than that observed in patients starting two NRTI and a PI. The number of patients in this analysis was small so confirmatory analyses on larger numbers and longer follow-up are needed. Furthermore, results should be interpreted cautiously since patients have not been randomised.
Another strong factor associated with the chance of achieving a suppressed viral load by a given time is the pre-therapy levels of viral load, which, in antiretroviral-naïve patients, can be taken as the natural viral load set-point^^'^^®°. The analysis performed in this chapter showed that the viral load pre-therapy is associated with both the time to achieve viral success and the absolute probability of being virologically successful. However, the analysis of UNIF cohort has shown that, in previously treated patients stating HAART, in the long-term, viral load may lose its prognostic value in favour of the CD4 count. Patients with better immunological conditions at the time of starting therapy appeared to have a greater chance of maintaining virological success, irrespective of their initial viral load. However, this result is still controversial, has not been confirmed by analyses of other cohorts^^®’^^’^^®’^®^’^®^ and was previously found only in one cohort of pre-treated patients^®®. It is conceivable that the significance of baseline CD4 count is different in groups of antiretroviral-naïve and pre-treated patients because it partly indicates the impact of, and perhaps therefore adherence to.
past therapies as well as the natural underlying state of the patient’s infection. Further, in pre-treated patients, the number of new drugs started at the time of HAART was also found to be a strong predictor of the virological response.
In addition to the factors analysed here, there are obviously other factors that are known to affect the virological response to HAART. These are typically adherence, plasma drug concentration levels, and drug-resistance. Adherence data are not yet routinely collected in the cohort studies analysed in this thesis, and only limited measurements of adherence and plasma drug concentrations are available in a group of patients enrolled in I.Co.N.A. Interestingly, in I.Co.N.A., injecting drug users had a poorer virological response than patients who acquired HIV through other modalities of transmission after adjusting for all the other factors considered. It is conceivable to explain this result with poorer adherence to therapy in this group of patients. Indeed, mode of HIV transmission is often used as a “proxy” of adherence in analyses of cohort studies. The association between drug resistance and virological response to therapy is addressed in chapter 7 of this thesis.
Table 3.5.1 Summary of most important predictors in the two cohorts Cohort
UNIF I.CO.N.A
Outcomes
Lower chance of initial IDUs IDUs
virological success Higher viral load High viral load
Pre-treated SQV as only PI
Less new drugs started PI-HAARTvs. 3NRTI SQV as only PI
Higher risk of IDUs Younger age
virological rebound Lower CD4 count SQV as only PI
Pre-treated 3 NRTI vs. PI-HAART
SQV as only PI NNRTI vs. PI
Lower chance of CD4 Higher viral load Lower VL suppression
3.5.2 CD4 count response to HAART
After the initiation of HAART and over a period of follow-up of up to 108 weeks, the CD4 count increased, on average, by 4 to 7 cells/[xl per month. Assuming a linear increase, this means that it may take about 28 months for a patient receiving HAART to achieve an increase of at least 200 cells/fxi. This estimate is roughly in agreement with the plot of the raw CD4 changes observed in UNIF (Figure 3.3.8) and in I.Co.N.A.
(Figure 3.4.5).
There is agreement at this point in time that the variability in CD4 count increase is mostly explained by the degree of virological suppression on HAART. Thus, from this analysis, the larger rise in CD4 count in antiretroviral-naïve patients and in pre-treated patients starting at least three new drugs, was completely explained by the greater decrease in viral load experienced by these patients.
Co-infection with HCV, age, and the CD4 nadir before therapy initiation have often been indicated as main predictors of CD4 count recovery on HAART®°®’®^®‘®^®. However, whether HCV infection is directly responsible for the lack of CD4 regeneration or is simply a proxy for poor adherence needs further evaluation®^®’^^^. In the analysis of the I.Co.N.A database, co-infection with HCV did not seem to provide prognostic information over and above the mean viral load suppression over follow-up, mode of HIV transmission and CD4 count pre-therapy. In the UNIF database the information regarding the HCV sero-status was not available (see Table 1.10.3).
In these analyses, age retained a marginally significant association with the CD4 count increase, after adjusting for the average virological suppression over follow-up. Specifically, CD4 count reconstitution appeared to be slower in older patients but this effect was rather small and confined to the comparison between very old and very young patients. However, the risk of AIDS is affected by age independently of CD4 count and viral load so further data are needed to evaluate whether age should be taken into account when deciding the time of initiation of HAART (see chapter 5 for full discussion of this issue).
CHAPTER 4: USE OF A TIME TO THERAPY FAILURE' APPROACH TO