BHIVA draft adult antiretroviral treatment guidelines: Evidence and trial design

1.4 Basing recommendations on evidence

In producing this document the committee used an evidence-based medicine approach. In reality, it is recognised that if only the most reliable form of clinical evidence is taken into account (i.e. the result of at least one or more randomised controlled trial with clinical endpoints) it would be impossible to formulate these guidelines, as many important aspects of clinical practice remain to be formally evaluated and very few trials with clinical endpoints are ongoing or planned. Results from clinical trials with viral load and CD4 count changes as endpoints had to be used as they are the only source of evidence in many instances. However, most such trials have been performed in order to obtain drug approval and the designs are not well suited to addressing questions of how to use the drugs in clinical practice. The most important drawbacks of such trials in this regard are the short duration and the lack of follow-up data on patients who switch therapy from the original regimen. In most cases the only available data on long term outcomes are from routine clinic cohorts.

While such cohorts have the advantage of being representative of routine clinic populations, the lack of randomization to different regimens means that comparisons between the outcomes of different regimens are highly susceptible to bias (Pocock, Phillips). Expert opinion forms an important part of all consensus guidelines; however, this is the least valuable and robust form of evidence.

1.5 Use of evidence published as abstracts

In writing these updated guidelines, it has been recognised that there is often a considerable time lag between initial presentation of important data (whether oral or in abstract/poster format) and full publication. Consequently, there is a danger in relying on data that have not been subjected to formal peer review and published in full. We have therefore avoided citing any research findings which appeared only in abstract format more than 3 years ago (i.e. before mid-1998).

1.6 Implications for research

Unless guidelines are interpreted and applied sensibly and with caution valuable research initiatives that might improve standards of care would be stifled. It would be wrong to suggest that certain clinical controlled trials would be unethical if they did not conform to the guidelines, especially when these guidelines were based mainly upon opinions rather than more reliable evidence. The National Health Service executive has stated that clinical guidelines cannot be used to mandate, authorise or outlaw treatment options [Hurwitz].

1.7 Use of surrogate marker data

CD4 cell counts and plasma viral load are used as markers of the biological activity of antiretroviral therapy in Phase I and II trials. Reduction in viral load leads to a rise in peripheral blood CD4 count, with greater rises being seen in those with greater and more sustained viral suppression (Staszewski).

Changes in these markers in response to therapy has been found to be strongly associated with clinical response (Pozniak, FDA July 1997, Grabar, Delta Coordinating Committee and Virology Group 1999) and CD4 counts measured in people on antiretroviral therapy have been shown to be associated with risk of AIDS-defining diseases which is no higher than the risk expected at that CD4 count in untreated individuals (ref Mocroft, Miller, Weverling, Ledergerber). The CD4 count is a better indicator of the immediate risk of AIDS-defining diseases than the viral load in those on antiretroviral therapy (Lundgren ICAAC).

Favourable responses to therapy, i.e. a decline in plasma HIV-1 RNA and increases in CD4 cell counts, have led to accelerated licensing of antiretroviral agents on the grounds that it is impracticable to wait years for large clinical endpoint trials to be completed before drugs are approved [Pozniak,FDA July 1997,FDA 2000].

Drugs are given full approval on the basis of trials lasting 48 weeks and, in some countries, accelerated approval based on data to 16 weeks.

Most clinicians would agree that a drug-licensing policy based on surrogate markers is reasonable and humane. However, it is important not to forget that CD4 count and viral load responses do not precisely reflect the clinical outcome to be expected (ie they are not perfect surrogates of the clinical response - Fleming, De Gruttola 1, Babiker). This is because the drugs have other effects with clinical consequences besides those reflected in viral load and CD4 count changes. The short length of trials leading to drug approval obviously means that at the time a drug is licensed nothing is known about long term consequences of use of the drug.

1.8 Issues concerning design and analysis of clinical trials

1.8.1. Trial Designs

Most trials of antiretroviral drugs are performed by pharmaceutical companies as part of efforts to obtain licensing approval and the designs are often not well suited to derive information on how to use the drugs in clinical practice. Besides the short duration of follow-up, the key limitation is the lack of data on outcomes in people who change from the original randomized regimen. The results are therefore only clearly interpretable so long as a high proportion of participants remain on the original allocated regimens. Clinical questions about which drugs to start with or switch to require longer-term trials which continue despite changes to the original treatment. From a clinical viewpoint it makes no sense to ignore what happens to patients after a given specific regimen has been stopped. Besides, use of a given drug can affect outcomes long after it has been stopped (eg due to selection for virus resistant to drugs not yet encountered, overlapping toxicities). However, interpretation of such trials is not straightforward and of course account must be taken of which drugs were used subsequent to the original regimen in each arm. Planned or ongoing trials which adopt such an approach include Initio, CPCRA's FIRST, ACTG 384 and Optima.

Study design may have an important influence on the rate of discontinuation of trial drugs. An open trial design may result in higher levels of discontinuation from what is perceived to be the least effective regimen, while a double-blind, placebo-controlled study may reduce adherence in all groups because of the large pill burden. It is also important to recognise that controlled clinical trials provide an optimal treatment setting and results in clinical practice are usually not as good.

1.8.2. Methods of Analysis

Several methods have been used to analyze viral load and CD4 count responses, including the change from baseline at a given time and the time-weighted change from baseline (Area Under the Curve (AUC)).

For virological response, however, the most common approach relates to whether the viral load is below some low level, usually 50 copies /mL, which is approximately the lower limit of quantification of most viral load assays in routine use . The proportion of people with viral load

One reason for the choice of this outcome measure is that some studies have indicated that if this low level is not achieved then subsequent viral load rebound is more likely (Kempf, Raboud). However, when comparing treatment regimens, differences between the treatment groups in viral load (with levels higher than 50 copies/ml) are highly predictive of subsequent differences in clinical outcome (Brun-Vezinet, Boucher, Loveday et al 1997), and so the restriction of the comparison to the proportions with viral load £ 50 copies/ml would not utilise other information contained in viral load measurement. A related approach which is suited to assessing the response to an initial regimen is to calculate the time to virological failure.

Virological failure is typically defined by failure to achieve viral suppression or viral load rebound after achieving

1.8.3 'Intention to treat' and 'on treatment' analyses

Randomization in a trial is to ensure balance in prognosis between the treatment arms at baseline. If outcomes cannot be assessed for some patients then this can disturb this balance and can create bias in the comparison between the treatment arms. The principal of analysis by "intention to treat" is to include outcomes for all randomized patients in the analysis in order to avoid risk of such bias. However, in order to do this clearly it is necessary to continue to collect data on all patients, even if they switch from the original regimen. But this is rarely done, so the "intention to treat" principal is operationalized by imputing values for those patients who have "dropped out" of the trial. When the outcome is the proportion of people with viral load 50 copies/mL all patients who have earlier switched therapy or have the viral load value missing for any reason (missing=failure approach) (Gazzard). Such an approach implicitly equates failure of a regimen due to inadequate potency with inability to tolerate a regimen due to pill burden, inconvenience and/or adverse effects, although the implications of these two outcomes are likely to be different. The approach is often referred to as "conservative", due to the fact that it gives the minimum proportion

So-called "on treatment" analyses only consider outcomes in those on the original allocated treatment. In the context of the proportion of people with viral load 50 copies /mL will reflect the speed with which clinicians make the decision to switch therapy in response to the first viral load value(s) > 50 copies/mL. It is difficult to see how it provides a useful means to compare the efficacy of different regimens. Within the context of a time to virological failure analysis the "on treatment" analysis (in which follow-up on patients switching drugs is right-censored – by definition this occurs before there is evidence of viral load failure and so should not cause bias ) may be more revealing.

In situations where there is a high proportion (say > 25%) of patients who do not have a viral load value at a given time point, interpretation is inherently difficult and no analytical approach is entirely satisfactory.

1.8.4 Equivalence

Large numbers of patients are usually required to show equivalence between regimens (i.e. to demonstrate no or small difference in response between treatments), and many surrogate marker studies are under powered to demonstrate this. Stating that studies have shown no significance difference between the treatment arms is very different from saying that the arms show equivalence. Graphical representations of CD4 cell count rises or viral load declines in response to therapy that appear to be overlapping, may hide differences in efficacy between drugs. The confidence interval for the difference in outcome between treatment arms should be examined carefully in such studies. Lack of adherence to allocated regimens is even more of an issue in equivalence trials since the intention-to-treat analysis would tend to dilute the difference in outcome between the treatment groups. Unless discontinuations and treatment changes during the trial reflect what would happen in clinical practice, the results from the intention-to-treat analysis would be biased towards equivalence.

1.8.5 Cross study comparisons — presentation of data

It is tempting to compare results of individual drug combinations assessed in different trials. Such comparisons are fraught with difficulties because of differences in entry criteria (particularly with respect to viral load and CD4 cell counts), methods of analysis (e.g. intention to treat versus on treatment), degrees of adherence and sensitivities of viral load assays.

1.9. Adverse event reporting

Many previously unsuspected side effects of antiretroviral therapy have been reported only after drug licensing. It is vital that any possible adverse events be reported rapidly by prescribers in order that recognition of adverse events occurs swiftly. A blue-card scheme, organised by the Medicines Control Agency, the Committee for Safety of Medicines (CSM) and the MRC, operates in the UK for reporting adverse events relating to the treatment of HIV.