Any discussion about "Salvage Therapy" is facilitated by the understanding that antiretroviral treatment decisions are made in any patient over the dimension of time and in the context of a variety of clinical circumstances rather than as a singular event in a patient who is part of a homogenous patient population. A review of the ICAAC "Interactive Session on Retroviral Therapy" is helpful, then, in providing the context in which decisions regarding changes in treatment are being made in 1998. In this interesting symposium comprised of both experienced and inexperienced HIV providers, case examples guided discussion about the choice of antiviral interventions through the use of a device at each attendee's chair that enabled the participant to give a single answer to multiple choice questions that followed each case example. For instance, the first case scenario presented to the audience a 32 year old man newly diagnosed with HIV. Laboratory information about this man included a CD4 count of 560 and a viral load of 23,000. Based upon this information, the audience was asked to choose one "best" answer to the question "what would be your recommendation for treatment?" Of the choices available, 17% of the audience would advise this man to continue careful monitoring with no therapy at this time, 49% would advise the use of two nucleoside analogues (NA) and a protease inhibitor (PI), while 28% would recommend two NAs and a non-nucleoside reverse transcriptase inhibitor (NNRTI), and yet 3% would use three NAs. As the session progressed, this man's clinical information and his viral load and/or T cell data would change, prompting the audience to respond differently with their treatment recommendations. Clearly the topic of "Salvage Therapy" cannot be discussed without close attention to an individual's antiviral history, information which has a profound impact on the outcome of any antiviral intervention that follows the first one. The issues around switching therapy really involve getting answers to three important questions: when to change, with what, and towards what goal?

As demonstrated by the audience during the interactive symposium, there is diverse opinion on when changes in therapy are important. Given a scenario that described a patient whose three drug regimen (stavudine/didanosine/indinavir) achieved an undetectable (below 400 copies/ml) viral load for 15 months, followed by detectable levels of 1200 on two occasions, the audience was divided about whether or not to change treatment. In this example, 74% of the audience would elect to continue present therapy with careful monitoring, while 26% thought a change was necessary. Then, when asked at which level of viral load would a change in therapy be appropriate, 36% chose a level of 5000 copies/ml, 31% chose a level of 10,000 copies, 17% chose a level of 20,000 copies, and 15% said they would not rely upon viral load testing to make decision. As a follow up point in this case where the clinical situation demonstrated a stable viral load, but a falling CD4 count, a query discovered that a decline to 200 cells/mm3 would prompt almost half of our audience to recommend a treatment change. There are several reasons for such diversity of opinion about when changes in therapy are needed. In the patient receiving antiretroviral therapy, knowing that persistent viral replication (even though it may result in a low but measurable viral load) selects for viral strains that accumulate resistance mutations in their genome with subsequent treatment failure urges many to change therapy at the first indication that antiviral control is being lost (rising viral load). However, both the limitation in antiretroviral therapeutic options and the clinical and immunological stability of any given patient will conspire to give any HIV treater pause before recommending any treatment change. Discussion about the utility of genotypic and phenotypic laboratory tests in guiding this decision making process revealed that the audience was divided fairly evenly: 30% would use a phenotypic test, 35% a genotypic test, and 29% thought guessing was just as good.

Dr. Jonathon Shapiro from Tel Hashomer Hospital in Tel Aviv, Israel led a thoughtful discussion about when switches in therapy are appropriate, and towards what goal these therapeutic changes could realistically aim. Characterizing a person's antiretroviral (ARV) experience will determine whether viral load "undetectability" is likely and realistic (as it is in treatment naïve and at times in mildly experienced persons) or whether a reduction in viral burden is more achievable (as it may be in those persons with lots of antiretroviral experience). The following table summarizes this thinking:

Switch Early Good therapeutic options Current regimen challenging to patient Immune status deteriorating

Delay Switch Few therapeutic options Current regimen well tolerated; Immunological state stable/improving

In a return to the discussion about the use of resistance testing (genotypic or phenotypic) as an aid in the decision making process, several clinical situations were identified in which such testing with today's technology might be useful:

1. To exclude the possibility of resistant virus in a person newly infected with HIV

2. To assist in determining the cause of therapeutic failure in a given person
   1. Compliance
   2. Drug absorption problems
   3. Drug metabolism problems
   4. Resistance to antiretrovirals

3. Determine most appropriate drug regimen in a patient failing the current regimen

4. Though a controversial use of this technology, to assist in a decision to discontinue ARV treatment in persons with advanced disease, who demonstrate multidrug resistant strains and who also are experiencing many drug toxicities. (Caution here, however, as several small studies presented here and in Geneva demonstrating the use of "megatherapy" with nine or more drugs in highly ARV-experienced persons have shown at least short-lived antiretroviral responses in 30-40%)

The development of resistance testing as a tool that is useful in making careful treatment decisions must demonstrate that the test can predict which drugs will no longer be of clinical benefit as well as identifying which drugs will be of benefit now, and show that the use of these tests will improve clinical outcomes more than what can be accomplished by using current techniques (using a compilation of information such as patient antiviral history, T-cell count, viral burden, etc.).

With the context (in which treatment-change decisions are made) now framed, Dr. Deeks began his discussion of issues around salvage therapy by referring to the DHHS Guidelines as a work in progress, despite the current broad recommendation that such regimens contain two NAs and two PIs. Given the large numbers of HIV-infected persons receiving treatment with Indinavir, he cited a study presented in Geneva that demonstrated that use of Ritonavir with Saquinavir as a "salvage regimen" was not a good choice in the setting of Indinavir failure. He went on to describe a prospective, randomized open label trial in 20 patients all demonstrating resistance to Indinavir with viral load values > 2500 copies after 24 weeks. These individuals received a combination regimen containing either Nelfinavir/Saquinavir/Abacavir/NA or Nelfinavir/Saquinavir/Abacavir/Nevirapine. As a result, the participants who received the first regimen showed limited viral load reduction with rebound to baseline at 24 weeks of study, Those who received the second regimen demonstrated levels of viral load that were undetectable (<50 copies/ml) at 24 weeks in 5 of nine patients in this group, demonstrating (we think) the importance of multiple drugs and multiple classes of drugs in a setting of antiviral failure to a PI containing regimen.

However, when taking a protease-inhibitor containing regimen as initial therapy, not all viral load rebound means the development of antiviral resistance. Citing the observations from ACTG 343 (using Indinavir) and ACTG 347 (using Amprenivir), Dr. Deeks presented the "Predator/Prey Theory" to explain the emergence of wild type virus in persons with rebound viral load who were receiving Indinavir (or in the case of ACTG 347, Amprenavir). In ACTG 343, most patients exhibiting viral rebound in the Indinavir-containing arm of the study grew virus that demonstrated a wild type genotype in the protease gene. That is to say, that virus grown in these persons failed to demonstrate any mutations in the protease gene that would render HIV protease resistant to Indinavir. Furthermore, those individuals that had the highest CD4 rises were at highest risk for viral rebound on Indinavir: for every 100 cell CD4 rise, there was a 30% greater risk of relapse. This led to the "adding fuel to the fire" appellation that tries to theorize why this is happening. As the theory goes, a vigorous rise in CD4 cells provides more "cell targets" for HIV. During the initial months of therapy, selected strains of wild type virus that are somehow less responsive to the inhibitory effects of Indinavir will gain a selective growth advantage and "prey-upon" the rich cellular milieu that has arisen as a result of an early and brisk immune response to antiviral treatment. That this phenomenon has been similarly observed in the ACTG study using an Amprenavir containing arm (wild type virus showing no amprenavir mutations in persons demonstrating viral rebound while taking an amprenavir-containing regimen) suggests something unusual is happening. Moreover, this scenario suggests that drug intensification may be needed, rather than switching and does not suggest what many would have surmised to explain this viral rebound -- that the patient was poorly adherent to therapy. This is all new and not understood, thereby making treatment decisions in this situation more complicated (though perhaps providing an indication for good genotypic analysis of the protease gene in this clinical setting).

As a quick overview of new agents and their potential in the setting of "treatment failure" (however that is defined) the following points were made.

1. Abacavir is a very potent guanosine nucleoside analogue in treatment naïve individuals, and is associated with mutations in the RT gene at positions 184, 65, 74, and 115. Many of the mutations occur in persons taking AZT and 3TC, but many mutations are need before resistance develops to Abacavir. As a salvage treatment, there are limited controlled data available.

2. Adefovir is a nucleotide analogue that inhibits reverse transcriptase. As such, resistance to this drug develops when mutations occur at codon 65 and 70 on th RT gene. Salvage study 408 showed that in persons with the 184 mutation (common to 3TC) study participants could achieve a 0.97 log reduction in viral load. The addition of AZT resistant mutations to 184 diminished the response to adefovir: with a 0.58 log decrease in viral load when low level AZT resistance was added to 184, and a 0.44 log decrease in the presence of high level AZT resistance.

3. Efavirenz is the newest NNRTI to be approved demonstrating equal efficacy to to Indinavir/AZT/3TC, when used in combination with AZT and 3TC . Its utility in a treatment-failure setting seems unlikely when other NNRTIs have been used (nevirapine and delavirdine), because of broad cross resistance to all drugs in the NNRTI class that occurs as a result of mutations at the 103 postion of RT (the K103N mutation).

4. Combination therapy with Efavirenz/Adefovir/Nelfinavir/Abacavir in persons failed by Indinavir therapy, is being evaluated in an ongoing ACTG 372B trial.

5. Hydroxyurea has stimulated interest in the treatment failure setting, particularly because it has a different mechanism of action that seems to enhance the antiviral effect of several NAs (DDI in particular) in in vitro studies. Moreover, HU can restore virus sensitivity to DDI when resistance has developed as a consequence of previous DDI use. By inhibiting the cell in the G1 cell cycle, Hydroxyurea (HU) changes the enzymatic pathway by which the building blocks of genetic material (dideoxynucleosides) are prepared (by a process called phosphorylation) for DNA chain formation (crucial for productive HIV infection into the host genome of the infected cell).

Even though HU has no demonstrable antiretroviral activity in and of itself, several small clinical trials have demonstrated its ability to diminish viral load further when added to DDI or DDI/D4T regimens. Most clinicians avoid its use with AZT because of the combined effect of both drugs on bone marrow function: neutropenia and anemia are common when used together. Neutropenia may be a problem when HU is used without any other marrow-suppressive medications, and usually patients with a neutrophil count under 1500 before commencement of HU are at higher risk. Another concern in many trials evaluating HU is the blunting of the CD4 cell rise which usually occurs as a consequence of a declining viral load in the presence of effective antiretroviral therapy. However, its ease of use (twice daily), relatively good tolerance and different mechanism of action have made popular the addition of HU to a regimen containing DDI in the treatment failure setting.

ICAAC abstract I-202 presented one experience (from UCLA and USC) where charts were retrospectively reviewed in individuals taking HU + DDI with other NAs following virologic failure. 20 men and three women were included in the analysis. On average, the group experienced a decline in plasma viral load of 0.7log by 28 weeks, but also a fall of 34 CD4 cells. 3 of 23 (13%) had > 1 log viral load decline and 4/23 (17%) demonstrated undetectability (<200 copies). Those that were able to achieve undetectable levels had lower mean viral loads before the DDI/HU regimen began (13,654 copies) as opposed to those that remained undetectable (330, 265 copies; p=0.028). Prior DDI experience did not influence the response to DDI + HU therapy.