When should antiretroviral therapy be started for HIV infection? Interpreting the evidence from observational studies
AIDS 2003; 17(13):1863-1869
Andrew N. Phillips a,b; Alessandro Cozzi Lepri a,b; Fiona Lampe a,b; Margaret Johnson a; Caroline A. Sabin a,b
From the aRoyal Free Centre for HIV Medicine and the bDepartment of Primary Care and Population Sciences, Royal Free and University College
Medical School, UCL, Rowland Hill St, London NW3 2PF, UK.
"...There is little point in starting ART in patients who are not ready, unless their short-term risk of disease makes it essential to start... toxicities can affect quality of life, and this will be a major consideration for many people when deciding whether to initiate ART at a given time. Our view has been that it is important first to address whether immediate or deferred ART is likely to delay AIDS and death most...
"...(the authors conclude) In a person with CD4 cell count 350, immediate ART is likely to result in a lower risk of AIDS or death than deferred ART for the initial few years. However, the absolute risk in these initial years is relatively low, so the longer-term view is important. For this, it is important to consider the risk of virological failure when starting, and the benefits in this respect of deferring ART may not be as great now as they were in 1996/1997. Even so, it seems likely that those deferring ART who survive to initiate ART eventually will experience somewhat lower long-term risk of AIDS and death than if they had started ART immediately".
The point during the course of HIV infection at which antiretroviral therapy (ART) should be initiated remains uncertain. This is reflected in the extent of recent changes in recommendations for treatment guidelines. After a shift in the mid to late 1990s towards earlier initiation of ART - with the most aggressive guidelines suggesting initiation of ART in the overwhelming majority of people diagnosed with HIV - the more recent trend has been for later starting, with a delay in initiation until the CD4 cell count has fallen to 200 ? 106 cells/l in the most conservative versions. The time difference between starting ART according to the most aggressive and the most conservative criteria is on average over 5 years, and for some people could mean a potential ART delay of over 10 years; consequently, this issue is of major relevance.
In the absence of any ongoing randomized trial to address the issue directly, it has been necessary to assemble various pieces of evidence from prospective studies to estimate which is likely to prove the best strategy. Interpretation of the relevance and contribution of each study analysis is, however, not straightforward. In this article, we review the key findings and provide a framework for their use in deciding in a person with CD4 cell count about 350 whether immediate or deferred ART is likely to most delay AIDS and death. This CD4 cell count level is chosen because it is an area of most uncertainty for ART initiation. Although the key ultimate endpoints of importance are AIDS and
death, the effectiveness of current ART is such that relatively few people will develop them within a 5 year time span. Since we have not followed people on triple ART regimens for longer time periods than this, we have also, therefore, to consider an intermediate endpoint - virological failure - which, as outlined below, is a key determinant of who is more likely to develop AIDS or die over much longer time periods.
We consider in the following section the risk of AIDS and death before starting ART in those who defer. In the subsequent two sections we then focus on the comparison of the risk of AIDS and death according to whether ART is started immediately or deferred. We separately consider the short-term perspective, for which we have direct estimates of risk of AIDS and death, and the longer-term perspective, for which we must use evidence on risk of virological
failure. Finally, we briefly discuss some other issues for consideration and draw some conclusions.
Assessing the AIDS risk for those deferring antiretroviral therapy
The possibility of considering deferral when treatment is known to be effective at all CD4 cell counts results from the knowledge that there are CD4 cell levels at which deferral of ART appears to be safe. However, any delay in ART
carries with it a risk of development of an AIDS disease. It is important to be able to monitor this risk closely. It is well known that the CD4 cell count and viral load are key predictors of the risk and both are typically monitored every
3-6 months in people under clinical care for HIV infection. This suggests that this time period of 3-6 months is the most relevant one over which to assess AIDS risk and to guide clinicians and patients in deciding whether/when to
initiate ART. However, until now, few such data have been available. We have recently used information from a large collaborative project that pooled data from several seroconverter studies to estimate the 6-month risk of AIDS according to the viral load, CD4 cell count and age . Clinicians and patients can use such estimates to decide whether to continue to delay ART, at least for a further 3-6 months when the next measurement of CD4 cell count and viral load are obtained, or if the risk has reached a threshold level that is considered too high. This threshold level of risk – which conceptually is the same as that for which initiation of prophylaxis for Pneumocystis carinii pneumonia also
recommended - has generally been considered to occur at a CD4 cell count of 200-250; however, while the CD4 cell count is the principal consideration, the risk also depends appreciably on the viral load and age. Since viral load does not tend to change markedly, in log terms, in most untreated individuals over time, the short-term risk of AIDS and death largely changes with decreases in the CD4 cell count. For this reason, the issue of when to start ART comes down to a great extent to the question of how low the CD4 cell count should be allowed to decline before ART is initiated. Given this, it is relevant to know the probability that a given CD4 cell count, to which therapy is to be deferred, will be reached with the next few months to years. Recent estimates of this from the Multicenter AIDS Cohort Study have been published.
AIDS risk: the short-term perspective
The answer to the question of whether ART should be started immediately in a person with CD4 cell count may depend on how far into the future one wishes to consider. Figure 1 illustrates the risk of AIDS over time for two such notional groups: one of which starts ART immediately and the other defers. The lines represent the short-term (e.g., 3 month) risk at each time point. The initial AIDS risk is low in both groups (about 0.8% risk in 3 months in a 35-year-old person with a CD4 cell count of 350 and viral load 30 000 copies/ml according to our recently produced estimates). Studies of HIV natural history tell us that this risk of AIDS will tend to rise without initiation of ART, the rate of rise depending on the rate of decline in CD4 cell count and the level of viral load. If
ART is started immediately, then the risk of AIDS (and death) falls markedly, even in those with high CD4 cell counts (although the absolute risk reduction in those starting at high CD4 cell counts is low) and risk remains low for as long as large groups of such people have been followed: at present up to about 5 years. This has most clearly been shown in the recently reported analysis from the ART Cohort Collaboration. For those people with baseline CD4 cell count > 350, the rate of AIDS or death dropped by more than half between the first few months after the start of ART and 6 months later.
Consequently, until ART is started, the risk of AIDS is almost certainly greater in those who defer ART than in those who start immediately. The absolute magnitude of this difference will depend on the initial AIDS risk, being higher in
those at higher risk. Even after initiation of ART in those who initially defer, the risk of AIDS in this group is unlikely to fall, in a time span of a few years, below that in those who started ART immediately i.e., the line in Fig. 1, for the
deferred group initially remains above that for the immediate group even after ART is started because the former started ART at a lower baseline CD4 cell count. The recently reported results from the ART Cohort Collaboration compared AIDS rates according to the baseline CD4 cell count group. The higher AIDS rate in those initiating ART at a lower CD4 cell count remained after a follow-up of 3 years (Fig. 2).
Therefore, existing available evidence strongly suggests that the risk of AIDS and death over the subsequent 3 years is lower, albeit by only a small amount if the CD4 cell count is high, if ART is initiated immediately rather than deferred. So, if the dominant concern is to avoid AIDS and death over the next 3 years, then the better strategy for the person with CD4 cell count 350 is to start ART immediately. A recent analysis from the Swiss HIV Cohort, in which
people starting ART were matched with people who appeared similar except that they were not initiating ART, reached this same conclusion. This was the case even when follow-up experience in those who did not start ART was right
censored when the CD4 cell count declined to < 200 if ART had not been initiated by this point. It is important to emphasize that the absolute reduction in risk when therapy is initiated is much greater in those with lower
CD4 cell count, so while there is a clear reduction even in those with CD4 cell count > 350, the absolute risk was very low even before this reduction (Fig. 2).
So why are the above findings not considered to provide conclusive evidence that ART should be initiated immediately, rather than deferred, in people with relatively high CD4 cell count? The reason is that ART is now so effective that it
actually provides hope of avoiding AIDS for an entire natural lifespan. It is this perspective - rather than just 3 years - that many people will take. Even if untreated, the cumulative risk of AIDS over 3 years in a person with a viral load of 10 000-29 900 copies/ml and CD4 cell count of 200-350 is < 10%. Over 5 years, the cumulative risk is around 15%. The great majority of people with a CD4 cell count of 350 will remain AIDS-free 5 years later whether they initiate ART immediately or defer until their short-term risk has become unacceptably high. Therefore, the more important question is where the risk line (in Fig. 1) will be from 3 years onwards (perhaps up to 50 years or more), rather than in the next 3 years. This is considered in the next section.
Before this, other analyses assessing short-term AIDS risk have been presented as part of the 'when to start' debate and these should be put in context. Some studies have presented analyses showing the risk of AIDS and/or death after
starting highly active antiretroviral therapy (HAART) according to the CD4 cell count and, in some cases, the viral load at the pre-HAART baseline. Such data are useful for assessing prognosis in people starting ART but they do not directly address the question of exactly what CD4 cell count should be chosen for the start of ART as they do not account for the risk of AIDS and death before starting ART if it is delayed. Further, higher risk of AIDS after starting
ART in those with lower CD4 cell count at initiation cannot be used to conclude that the effect of ART is less in those with lower CD4 cell count at initiation, because the AIDS risk during ART reflects the AIDS prognosis at the time of
initiation as well as the effect of ART. Cole et al. have attempted to address this lead-time bias issue by jointly modelling the three processes operating in a person who defers ART: the probability of getting AIDS or dying before
starting ART, the probability of starting ART, and the probability of developing AIDS and/or dying after starting ART. This essentially formalizes the approach we have used above in our interpretation of data from studies of HIV natural
history and people on HAART (illustrated in Fig. 1, before 5 years). Based on data from the Multicenter AIDS Cohort Study and the Women's Interagency HIV Study, they concluded that deferring ART until the CD4 cell count was < 200 resulted in a greater 3-year AIDS risk (although adjustment for lead-time bias indicated that the effect of deferral to < 200 was of substantially lesser magnitude than was suggested by the crude analysis of time since ART initiation referred to above) but that deferral of ART in those with CD4 cell count 350-500 until < 350 was reached (but not < 200) did not result in any greater AIDS risk (although confidence intervals for the relative hazards were relatively wide). Therefore, these findings are consistent with the points made
The longer term perspective: minimizing risk of virological failure and resistance development
Since we have only around 5 years of experience of triple ART regimens, we can only indirectly evaluate the risk of AIDS diseases and death over the longer term, by comparing intermediate endpoints that we believe relate most closely
to the long-term risk of AIDS and death. For some years, use has been made of viral load outcomes to assess and compare the effects of antiretroviral drugs. It is well established that the effect of ART on viral load is strongly
correlated with the effect on the CD4 cell count and risk of AIDS, and it has become widely accepted that the goal of ART is to achieve and maintain a viral load < 50 copies/ml. So long as such a level of viral suppression is maintained in a person with a CD4 cell count > 200, the risk of AIDS remains very low. The need for viral suppression as low as 50 copies/ml is to minimize the risk of emergence of viral strains resistant to drugs. Virologic failure - failure of viral load to become, or to stay, at a low level on ART - is often associated with development of drug resistance. Since resistance to one drug in a class tends to also reduce susceptibility of virus to other drugs in that class, virological failure and development of drug resistance is a serious problem with long-term consequences for maintaining viral suppression. It seems reasonable, and conservative, to assume that in the future development of new drugs to which the virus of most infected people is sensitive will be limited. In those who achieve viral load < 50 copies/ml within 24 weeks of the start of ART, the risk of virological failure is very low; this suggests that prevention of development of AIDS for many years to come - perhaps even over a natural lifespan - depends largely on avoiding resistance within the first few months after initiation of ART. Resistance cannot always be detected directly because such tests lack sensitivity. Effectively then, the emphasis is on avoiding virological failure on ART: resistance will probably be present in most such patients, whether it is detected in genotypic or phenotypic resistance tests or not.
Returning to Fig. 1, the fear is that the risk of AIDS in those on ART will eventually increase again at some point because of the development of resistance to the available ART regimens. So which group - the immediate or the deferred- is likely to achieve the longer duration of reduced AIDS risk? If this duration (which we know already is at least 5 years) is the same in both groups then, as indicated in Fig. 1, the increase in AIDS risk will occur earlier in the immediate group (because they started ART earlier). Therefore, the risk lines will have crossed, with that for the deferred group moving from being the higher to the lower. It is difficult to conceive of any plausible set of
circumstances in which the risk lines would then re-cross at some point in the future, so the deferred group would probably experience lower AIDS risk from that point on. A similar outcome would result if the duration of reduced
AIDS risk were greater in those who defer. However, if the duration of reduced AIDS risk is shorter in the group who defer, then the difference in AIDS risk after the risk lines cross will be smaller or, indeed, the risk lines may never cross at all.
As discussed above, the principal means we currently have of trying to examine this issue is to assess whether the risk of virological failure is lower if ART is started immediately or if it is deferred. There are reasons for believing that either could be the case. Consider, on the one hand, a person with a CD4 cell count of 350 in 2003 and who might have a CD4 cell count of 200 in 2008 if ART is deferred. Immediate ART may carry the lower risk of virological failure if it is the case that the risk of virological failure is lower in people initiating ART at higher CD4 cell counts and/or lower viral loads. We studied almost 3000 people from three large cohort studies who initiated ART after 1996 and found no evidence that the risk of virological failure was lower in those starting ART at CD4 cell counts > 350 compared with those initiating ART at CD4 cell counts of 200-350. Those with higher baseline viral loads took a longer time to achieve initial viral suppression than those with lower baseline viral load, but
the probability of achieving viral load suppression by 32 weeks was similar, as was the subsequent rate of viral rebound. Other authors have studied this issue but not on individual patient data. Consider, on the other hand, this same person with a CD4 cell count of 350 in 2003 who defers ART until 2008; the risk of virological failure may be reduced because of advances in ART (becoming less toxic, easier to take and more potent) and adherence management that may have occurred during this period. Of course, it is impossible to answer this definitively because we do not know what treatment improvements will occur in the next few years. However, it is informative to look at the changes in early virological failure rates seen since the introduction of HAART in 1996. This is best achieved by looking at complete, open clinic cohorts with no patient selection. Figure 3 shows data from an open clinic on the percentage
with virological failure by 1 year from starting ART (with at least three drugs). Failure is defined as having a viral load > 400 copies/ml measured between 24 weeks and 1 year, followed by another value > 400 copies/ml. Those without a
viral load available between weeks 24 and 48 are excluded. There has been a decline in the percentage with virological failure between 1996 and 1998, but there has been little change since then, with the value levelling out at approximately 15%. More extensive data are required on this issue, including for rates of virological failure over more than 1 year. However, this preliminary examination suggests that - in so far as future AIDS risk be determined by 1-year rate of virological failure - the durability of reduction in AIDS risk will be no shorter, and perhaps longer, in those who defer ART. As argued above, this indicates that the AIDS risk will be lower over the longer term in the deferrers.
Before concluding, some other issues require consideration. First, the degree to which the patient feels ready to start ART is probably the most important determinant of virological response. There is little point in starting ART in patients who are not ready, unless their short-term risk of disease makes it essential to start. Second, while the data shown take account of drug toxicity in so far as it directly (owing to fatal adverse events) or indirectly (owing to an effect on adherence) affects risk of virological failure and death, there is also the fact that toxicities can affect quality of life, and this will be a major consideration for many people when deciding whether to initiate ART at a given time. Our view has been that it is important first to address whether immediate or deferred ART is likely to delay AIDS and death most. Given this information, patients can individually then factor in the quality-of-life issues as they relate to them personally before making their final decision. However, this is not easy as there are insufficient data on quality of life in most trials and cohorts. In addition, treatment interruption is relatively common (although by no means universal). Those who initiated ART with a higher CD4 cell count and lower viral load will probably be able to interrupt safely for longer, if at some point they do interrupt ART. This is another important consideration when deciding whether to start ART at a given time.
In a person with CD4 cell count 350, immediate ART is likely to result in a lower risk of AIDS or death than deferred ART for the initial few years. However, the absolute risk in these initial years is relatively low, so the longer-term view is important. For this, it is important to consider the risk of virological failure when starting, and the benefits in this respect of deferring ART may not be as great now as they were in 1996/1997. Even so, it seems likely that those deferring ART who survive to initiate ART eventually will experience somewhat lower long-term risk of AIDS and death than if they had started ART immediately.