It might seem easy to define treatment failure and devise strategies to recover from it. But Keith Henry, M.D., who heads a clinic in St. Paul, Minnesota, for persons with drug-resistant HIV, observes, "Prescribing salvage therapy is at the level of anecdotal medicine. The field cries out for large databases to follow how people respond."
Despite the meager available information, the Public Health Services (PHS) branch of the U.S. Department of Health and Human Services (HHS) has published recommendations for such rescue or salvage therapies as part of its overall guidelines for treating HIV (see "HHS Guidelines for the Use of Antiretroviral Agents in HIV Infected Adults," revised December 1, 1998, pages 8-13 and pages 35-7). The PHS's advice is to switch therapy in those who experience only minor decreases in HIV levels when starting a new regimen or whose HIV rebounds to detectable levels on an established regimen. Although noting that "there are extremely limited data to suggest that any of these alternative regimens will be effective," the document states, "[the new regimen] should ideally involve complete replacement of the regimen with different drugs to which the patient is naïve and to which cross-resistance is not anticipated." The PHS further advises that "it may be prudent to delay changing therapy in anticipation of the availability of newer and more potent agents."
Many small trials using approved drugs have recently accumulated some preliminary data beyond the anecdotes. They confirm that some success is possible without experimental antiviral agents. Mostly too new to be published in medical journals, these trials were reported at the five major biomedical conferences concentrating on HIV held in the past half-year. These include June's 2nd International Workshop on HIV Drug Resistance and Treatment Strategies, July's 12th World AIDS Conference, September's 38th Interscience Conference on Antimicrobial Agents and Chemotherapy (ICAAC) and November's 4th International Congress on Drug Therapy in HIV Infection (held biennially in Glasgow and referred to as the "Glasgow conference"). Finally, the 6th Conference on Retroviruses and Opportunistic Infections, which took place in Chicago as this Treatment Issues was going to press, added further observations on salvage therapies and their appropriate use. All these reports are enlarging the "database" that Dr. Henry and the experts who write and revise the PHS guidelines need to put salvage therapy strategies on a sounder footing.
All Is Not Lost
Because CD4 counts often remain high in persons who experience viral rebound while continuing regimens that had been successful, researchers disagree over the proper time for giving up on current therapy and switching to a rescue regimen. On the one hand, changing drugs soon would curtail the accretion of resistance mutations in HIV as it replicated under the old regimen and preserve treatment options from the hobbling effects of cross-resistance. On the other hand, it might be better not to rush through all available drugs, exposing HIV to a series of not-quite-suppressive therapies in rapid order, but to maintain patients on a given regimen as long as they are immunologically or clinically stable, even if viral loads are significant.
One of the original bases for the go-slow approach was observations by Steven Deeks, M.D., at San Francisco General Hospital, of patients there who had experienced lack of complete viral suppression or HIV rebound on protease inhibitor-containing regimens (5th Retrovirus Conference, abstract 419). Failure was indicated by viral loads repeatedly above 500 copies/ml. Although their HIV was not completely suppressed, these individuals remained immunologically and clinically stable for at least a year (as measured by CD4 count and the appearance of new opportunistic conditions). After another year's follow-up, differences have emerged between those responding and not responding to treatment (6th Retrovirus Conference, abstract 494). After 96 weeks, median increases in CD4 were 180 cells/mm3 among virologic successes and 117 cells/mm3 among virologic failures.
People with incomplete viral suppression did have sustained CD4 count increases over their pre-therapy levels, but by the second year those increases were significantly less than for those with HIV levels that had fallen below the level of quantification. The extent of CD4 increase at all time points was less in persons with higher viral loads. Those with viral loads over 10,000 copies/ml had a median week 72 CD4 count of only 69 cells/mm3 over pre-therapy counts.
The reason why persons with only temporary success on antiviral therapy continue to do better than before may be related to a highly intriguing observation made by the San Francisco researchers. Using their specially created assay, they found that the average half-life of CD4 cells in untreated people with HIV is relatively short, 27 days compared to 77 days in individuals who have brought their viral loads down below 500 copies/ml (abstract LB2). Four persons with partial or no response to therapy had CD4 cells with an intermediate average half-life of 43 days. Therapy must be providing some partial protection even when the result is not obvious in terms of viral load.
One obvious possibility is that even seemingly failed therapies still keep viral loads down to some extent. Another is that the longer CD4 cell survival time in the nonresponders is due to HIV drug-resistance mutations that make the virus less prone to kill cells directly or indirectly. The San Francisco researchers reported that protease inhibitor-resistant HIV lost 50% to 80% of its ability to replicate in human thymus tissue grafted into mouse HIV infection models (6th Retrovirus Conference, abstract 4). A thymus less affected by HIV would be able to produce more new CD4 cells to replace those lost in the course of HIV infection.
Researchers at the Hôpital Bichat-Claude Bernard in Paris, who were following 21 people experiencing viral rebound while on protease inhibitors, had somewhat similar findings (6th Retrovirus Conference, abstract 331). They reported that protease-resistant HIV with drug-resistant protease enzymes became increasingly less "fit" as the number of protease mutations increased. In this case fitness, which was defined as the ability to infect fresh laboratory cell cultures, declined by an average of 27.5% in the observed patients. Decreasing viral fitness correlated with a higher CD4 count in the persons examined. Loss of fitness was not related to viral load, supposedly because the increased CD4 count ultimately allowed for more infected cells — a new host-virus equilibrium was created.
Resistance Assays Help if You Pay Attention
Resistance assays potentially can help doctors select which drugs are still active in a particular patient. A salvage therapy presentation at the 2nd Resistance Workshop illustrated the insights that may be gleaned by utilizing these assays (abstract 53). Nick Hellman, of ViroLogic in South San Francisco, described how his company's phenotypic resistance assay helped explain the response of 18 persons to a four-drug combination during a salvage therapy study at San Francisco General Hospital. Eight volunteers who had failed to respond to an indinavir-containing combination were treated with abacavir/nevirapine/nelfinavir/Fortovase while another ten received abacavir plus a second nucleoside analog and the nelfinavir-Fortovase protease inhibitor combination. (The volunteers had never received nevirapine or any other NNRTI.)
Notably, four of the volunteers turned out to have indinavir-sensitive HIV. No one had HIV sensitive to all four drugs in their regimen, but those with HIV still sensitive to two or three of the drugs in the combination experienced a profound reduction in viral load that reached a median of 2.5 logs (99.7%). The volunteers with HIV sensitive to zero or one drug had an immediate viral load drop that reached a median of 1.3 logs (95%) by the second week. This was as good as the first group had achieved at that point, but the second group's viral load rapidly returned to the baseline value.
The British Columbia Centre for Excellence in HIV/AIDS checked the predictive power of phenotypic resistance assays in 84 patients receiving ritonavir/saquinavir (ICAAC, abstract I-78). Inability to respond to combination regimens containing these two protease inhibitors was highly correlated with phenotypic resistance to saquinavir, especially, but also to ritonavir to some extent. No person with HIV resistant to both attained viral loads below 500 copies/ml, the limit of detection with the viral load assay in use. But the patients whose tests showed sensitivity to both drugs did not always respond completely. Only about half the patients with prior protease inhibitors and test results indicating sensitivity to both protease inhibitors achieved viral loads below 500. A study of the relation between genotype and response to salvage therapy (also a ritonavir/saquinavir combination) was conducted by Stanford University in 51 persons with a long history of treatment failure that included at least one protease inhibitor (2nd Drug Resistance Workshop, abstract 54). Nineteen (37%) of the study participants attained plasma viral loads below 500 copies/ml, while 18 (35%) had no virologic response at all. Failure was predicted by any combination of three protease mutations at specific points on the enzyme (amino acids 30, 46, 54, 82, 84, and 90).
In November French researchers presented very early results from the "VIRADAPT" trial, the first prospective data on the value of genotyping in selecting salvage regimens (Glasgow, abstract OP7.1). The 108 participants were divided into two groups before their new therapy was selected. In the first group, doctors altered their patients' regimen on the basis of personal experience and accepted standards of care. In the second, the doctors received the results of the genotypic resistance assays. Of the 47 already followed for six months, 39% on the genotyping arm had viral loads below 400 copies/ml, whereas only 9.5% in the standard-of-care arm had achieved this level. It should be pointed out, though, that the participants randomized to the standard-of-care arm turned out to have longer, more complicated treatment histories and more resistance-related mutations in their HIV than the genotyped group.
The Need for Real World Experience
The Community Program for Clinical Research on AIDS (CPCRA) trial 046 has yielded the most recent results on the benefits of genotypic resistance assays for selecting therapy (6th Retrovirus Conference, abstract LB8). This pilot study enrolled 153 patients whose viral loads had rebounded after more than 16 weeks on a three-drug combination and analyzed their HIV for resistance-conferring genetic mutations. Half the enrollees and their doctors received the results of these assays, whereas the other half were kept in the dark. The results after 12 weeks of treatment were somewhat underwhelming: Only 29% of the genotype group and 17% of the no-genotype group had viral loads below 400. Only 23% of the group whose results were withheld achieved this level of viral suppression. This outcome might underestimate the assay's worth: Individual doctors were allowed to alter the genotype-based regimens advised by the investigators. Such second-guessing occurred more than half the time.
Maverick doctors are just one example of how the real world differs from theory. Here's another: There is a growing mystery regarding the nature of indinavir failure, as several trials have reported that most people experiencing failure to control HIV while taking indinavir-containing triple combinations do not show evidence of genotypic (or phenotypic) resistance to indinavir (6th Retrovirus Conference, abstract 492). Resistance-conferring mutations to one of the other drugs in the combination (a nucleoside analog such as 3TC or an NNRTI like efavirenz) are the only relevant findings yielded by genotypic assays.
An analysis by Diane Havlir, M.D., of the University of California San Diego found this situation among participants in ACTG 343, a maintenance therapy trial in which treatment-naïve volunteers first received AZT/3TC/indinavir for six months. If their HIV was below 400 copies/ml at that point, they were randomized to either continue the triple combination or switch to either AZT/3TC or indinavir alone. (Continuing on triple therapy quickly proved superior.) Those whose HIV rebounded on or failed to respond to AZT/3TC/indinavir usually (but not always) had normal indinavir blood levels (6th Retrovirus Conference, abstract LB12). Dr. Havlir's explanation is that 3TC-resistance is the easiest for HIV to acquire. In the context of increased CD4 cell populations that provide more targets for HIV, loss of a single drug's suppressive effect may be enough to allow HIV to escape. If such is the case, it would not be necessary to change all the drugs in the combination, just the one that now proves ineffective in resistance assays. (Alternatively, a fourth drug could be added to the combination, either to reverse or ensure against viral rebound.)
Suspicions remain, though, that inferior indinavir blood concentrations are really to blame. For the average person, indinavir levels reach a problematic level, around the minimum needed for adequate viral suppression, at their lowest point between doses. People with poor absorption or rapid indinavir excretion will go into the danger zone. Indinavir's complicated dosing requirements (three times a day on a nearly empty stomach) and fears of indinavir kidney stones may also discourage strict adherence. The French Trilge trial, a maintenance therapy trial similar to ACTG 343, in fact found evidence that volunteers failing on indinavir did have subnormal indinavir concentrations in their blood and were not taking the prescribed number of pills (6th Retrovirus Conference, abstract 493).
Regardless of how this mystery is resolved, resistance assays have yet to prove their worth under real world conditions, in which lab quality varies considerably. The ability of commercial medical laboratories to detect mutated HIV subpopulations will be crucial. Minority strains of HIV making up as much as 25% of a person's entire HIV easily escape detection. The tests may deem the HIV in a patient susceptible to a particular drug, but if resistance exists in a small portion of the HIV present, that minority will rapidly become the majority upon exposure to the drug, rendering it useless.
Lots of Little Trials
The most ambitious salvage therapy trial to date is CNA2007, which enrolled 101 persons failing on their current regimens and gave them a triple combination that included the new drugs efavirenz (a nonnucleoside reverse transcriptase inhibitor or NNRTI), abacavir (a nucleoside analog) and amprenavir (a protease inhibitor). (For a fuller description of this trial, see the preceding article on abacavir.)
At week 16, only about one-fourth of the participants had viral loads below 400 copies/ml. Response depended heavily on the baseline drug resistance of individual participants' HIV, as measured by both phenotypic and genotypic assays (6th Retrovirus Conference, abstract 133). When the trial started, HIV was analyzed from 65 trial participants who remained on all three drugs. Of those baseline HIV isolates, 55% were susceptible to amprenavir, 42% to abacavir and 75% to efavirenz. Four of nine volunteers (44.4%) whose HIV was sensitive to two of the drugs experienced a viral load decrease of ten-fold (1 log) or a reduction to below 400 copies/ml. For those with HIV sensitive to all three drugs, the response rate was 46%. Analogous with the indinavir studies mentioned above, viral rebound by week 16 was mainly associated with resistance to efavirenz, not to amprenavir.
These same issues loom even larger in "mega-HAART" regimens, a sort of "kitchen-sink" approach to salvage therapy in which a maximum number of antiviral agents are administered concurrently. The hope is that enough of these compounds are at least partially effective and that in total, they will yield a significant suppression of viral load. A group of Frankfurt physicians have been administering six to eight drugs, including at least three nucleoside analogs, one or two NNRTIs and two to three protease inhibitors, to patients with a history of multiple viral breakthroughs on various regimens (6th Retrovirus Conference, abstract 130). The response in 37 patients has been promising, with 70% testing below 500 copies/ml on at least one viral load test. Once again, response could be predicted through HIV drug-resistance assays, with the ability to find three or four drugs still active against an individual's HIV crucial to the regimen's success. But it was difficult for most patients to continue on such a complicated regimen. Many had to switch to a simpler maintenance therapy, in which case their viral loads generally increased.
The British Columbia Centre for Excellence in HIV/AIDS reported on a similar mega-HAART study this fall (Glasgow conference, abstract P1). Patients were treated with an individualized regimen containing five, six or more drugs (termed "Multi-Drug Rescue Therapy" or MDRT). Nineteen of the 55 patients (34%) achieved a viral load below 500 copies/ml at some point. In the two studies, viral load decreases correlated inversely with the extent of phenotypic or genotypic resistance. Still, there were responders in both Vancouver and Frankfurt who had multidrug resistance. Last winter, a group of doctors in St. Louis and St. Paul reported on using d4T/3TC/ritonavir/saquinavir to rescue 24 persons with viral loads over 5,000 in two nelfinavir trials (see 5th Retrovirus Conference, abstract 510). The combination was relatively successful overall with 17 of the 24 participants (71%) having viral loads below 500 copies/ml assay at week 24. Now, with a median of 61 weeks of follow-up, 58% of the study participants still have viral loads below the 500 copies/ml cutoff (6th Retrovirus Conference, abstract 392). The results were highly dependent on the viral load at the time of the switch to ritonavir/saquinavir. Ten of the 12 (83%) participants with initial viral loads below 30,000 copies/ml achieved viral loads under 500 copies/ml at week 48, compared to 4 of 12 (25%) with higher initial viral loads.
A number of similar small studies have since taken a further look at the role of double protease inhibitor combinations to rescue treatment failure. Double protease inhibitors often require more mutations in HIV before the virus can escape. If there is no HIV population in a patient containing all these mutations together, one protease inhibitor covers the virus that is resistant to the other. Metabolic interactions between the two frequently raise their blood concentrations, too, making for a more potent attack.
More recent reports have given the ritonavir/saquinavir combination mixed reviews when used in therapies following other protease inhibitors, particularly indinavir. In one French study of 67 persons previously on indinavir-containing combinations (ICAAC, abstract I-194), only 7% had viral loads below 500 copies/ml at month 6.
Participants in this study were the hardest cases imaginable. They had had either little initial response to the indinavir combination or had viral loads that returned to baseline after the initial response, plus they had previously received every available nucleoside analog. Some have suggested that doubling the saquinavir dose in this combination to 800 mg twice a day would have been more effective in such an extreme group. (See Bernard Clotet's presentation, "New Directions in Protease Inhibitor Therapy," at the Hoffmann-La Roche-sponsored Satellite Symposium 3 held in conjunction with the 4th International Congress on HIV Drug Therapy.) In contrast, a retrospective chart review of 17 patients at Johns Hopkins who had failed regimens containing indinavir (13 persons) or nelfinavir (4 persons) found that 71% (12) achieved viral loads persistently below 400 copies/ml on ritonavir/saquinavir (12th World AIDS Conference, abstract 12330). This enhanced performance is probably due to the fact that the study participants had a relatively low viral load at the time of switching therapy (and, one can infer, fewer resistance mutations). The median viral load when switching to ritonavir/saquinavir was only 13,500 copies/ml.
The use of nelfinavir as the single protease inhibitor in salvage combinations for people failing other protease inhibitors did not work out very well in three observational studies (12th World AIDS Conference, abstracts 12281, 12285, 41215), probably due to the presence of prior protease mutations conferring cross-resistance to nelfinavir. In the best and largest (56 participants) of these uncontrolled, open-label studies, only 36% achieved viral loads under 500 copies/ml.
On the other hand, the nelfinavir/saquinavir combination may be useful, at least temporarily. One representative study from Germany used a four-drug, twice-a-day (bid) regimen consisting of two nucleoside analogs plus nelfinavir (1,250 mg bid) and saquinavir (1,000 mg bid) in 25 patients who had failed at least one protease inhibitor-containing triple combination (12th World AIDS Conference, abstract 12303). The median baseline viral load was a relatively low 20,000. The proportion of patients with plasma RNA levels below 500 was 60% at week 8, but only 45% at week 24. Likewise, another small German study of nelfinavir/saquinavir (plus d4T) rescue observed its cohort's viral loads drop over the first four weeks from a median of 149,000 to 45,000 copies/ml, but then the median viral load returned almost to baseline by week 12 (12th World AIDS Conference, abstract 22375).
Trying Something Completely Different
Double protease inhibitor combinations involve taking at least four, and usually more, different drugs (including two complementary nucleoside analogs), which makes for an onerous dosing schedule and risks introducing multiple drug-related toxicities. One of the original salvage therapy studies used only the nonnucleoside reverse transcriptase inhibitor nevirapine in combination with 3TC and indinavir to create a relatively successful and nontoxic combination for rescuing failed nucleoside analog regimens (one person of the 22 study participants had received ritonavir and two had prior loviride). (See M. Harris et al. Journal of Infectious Disease, June 1998, pages 1514-20; 5th Retrovirus Conference, Feb. 1-5, 1998, abstract 429a; and 12th World AIDS Conference, abstract 22337). A number of other studies have tried using efavirenz or nevirapine plus a protease inhibitor in persons with long histories on nucleoside analogs but no experience with NNRTIs or protease inhibitors. Among the first was a study by Gail Skowron, of Brown University, using d4T/nevirapine/nelfinavir in which 16 of 19 (84%) volunteers had a viral load below 400 by week 9 (5th Retrovirus Conference, abstract 350).
A number of preliminary studies, some no more than chart reviews, have tried using nevirapine plus a protease inhibitor and one or two nucleoside analogs in similar populations. Typical is one London hospital study (ICAAC, abstract I-195) in which doctors prescribed nelfinavir plus nevirapine and nucleoside analogs to 19 patients who had previous protease inhibitors but no NNRTIs and viral loads greater than 10,000 copies/ml. In 17 of the 19, the nucleoside analogs were merely continued or recycled from past therapies, and the response was not great. Only 5 of the 19 (26%) had viral loads below 400 copies/ml at 24 weeks. A 195-person study, ACTG 364, presented at the 6th Retrovirus Conference found that the nucleoside analogs plus efavirenz was markedly better than two nucleoside analogs plus nelfinavir (abstract 489). The result was unexpected since the trial population had no previous exposure to either protease inhibitors or NNRTIs. It may stem from better adherence to efavirenz's once-daily dosing schedule and side effects. Nelfinavir is a twice-a-day drug and causes a high frequency of diarrhea. (For more details, see chart on page 7).
Then there was the 101-person CNA2007 protocol using the efavirenz/abacavir/amprenavir combination described above. It was conducted in a cohort that was heavily experienced in both nucleoside analogs and protease inhibitors. The triple regimen of experimental drugs only performed respectably in the volunteers without prior NNRTIs. Another newcomer that has been tried in salvage therapy is the nucleotide analog adefovir from Gilead Sciences. The company's protocol 408 (ICAAC, abstracts I-84 and I-108; 4th International Congress on HIV Drug Therapy, abstract OP5.3) added adefovir onto current therapy in 429 volunteers with a mean baseline viral load of 72,000 copies/ml and a mean baseline CD4 count of 352 cells/mm3. The volunteers were allowed to switch background therapy at will. About 17% of the 182 persons receiving adefovir for 24 weeks had viral loads below 500 copies/ml, compared to 5% in the placebo arm.
According to a Gilead genotypic substudy, the reverse transcriptase codon 184 mutation that makes HIV resistant to 3TC paradoxically increases the virus's sensitivity to adefovir. Conversely, adefovir has no effect on HIV with the mutations that confer high-level AZT resistance. But how many people have HIV that has been exposed to 3TC and not AZT? About 15% of the 142 persons in the protocol 408 substudy had HIV resistant to 3TC but not AZT. These people's viral loads declined by an average of 0.94 log (89%) at 24 weeks. Fifty-three percent had high-level resistance to both AZT and 3TC, and exhibited a modest viral load drop of 0.5 log (70%) at 24 weeks. (Worse yet for adefovir, 32% of protocol 408's participants who had been on adefovir for 48 or more weeks developed signs of serious renal dysfunction and had to stop the drug.)
Another possible innovative salvage strategy involves hydroxyurea, which purportedly circumvents the drug-resistant HIV issue by constricting the production of natural nucleosides for assembly into DNA and thus increases the potency of analogs to DNA's natural nucleoside building blocks. The effect of adenosine analogs such as ddI and adefovir are especially strengthened by hydroxyurea, to the extent that they can be active against HIV that would normally resist them. The British Columbia Centre included hydroxyurea in its MDRT combination mentioned above. At the 12th World AIDS Conference, Steve Miles of UCLA astounded one audience with an account of how 18 heavily pretreated patients fared on a d4T/3TC/hydroxyurea combination. Despite having HIV with d4T and 3TC resistance mutations, the 18 had a median peak viral load reduction of 1.7 log (98%) by week 8. The average length of time with at least a 0.7 log (80%) viral load reduction was 17 weeks. This regimen seemed to elicit a strong but temporary response and frequently caused severe bone marrow suppression.
An increasing number of trials are including hydroxyurea. One, presented by Stanford University researchers at the 6th Retrovirus Conference (abstract 135), included both ddI and adefovir as well as hydroxyurea. The ten volunteers, who also received efavirenz, were naïve to NNRTIs, but had taken an average of 2.5 protease inhibitors and 4 nucleoside analogs in the past. To enter the trial, they had to have HIV containing a set of mutations thought to provide broad cross-resistance against nucleoside analogs (based on AZT- and 3TC-induced mutations) and protease inhibitors (see Treatment Issues, June 1998, pages 1-2). Nonetheless, the average viral load when starting the salvage regimen was a moderate 22,000. Only three of the volunteers had 2 log (99%) reductions in viral load, one of whom sustained that response through week 20. Slowly or rapidly, viral loads in the rest returned to baseline values after an initial response. The two volunteers so far who continued on the regimen for more than 20 weeks had to discontinue it because of adefovir's well-known renal toxicities. (One of these was the single sustained responder.)
A second hydroxyurea-containing regimen was much more successful (6th Retrovirus Conference, abstract 400). Thirty-eight volunteers, who had failed to suppress their HIV during protease inhibitor therapy, received hydroxyurea, efavirenz, two new protease inhibitors (mainly ritonavir and indinavir) and two nucleoside analogs (mainly ddI and 3TC). From week 8 on, 80% of the group had viral loads below 400 copies/ml. Follow-up was for 24 weeks. The source of this mega-HAART regimen's success is unclear since so many different drugs comprise it, and hydroxyurea's influence, if any, remains obscure.
Looking to the Future
This last trial exhibits the weakness of many salvage regimen studies: too few people taking too many drugs. It is impossible to evaluate the role of each. At best they prove the principle that rescuing people with long histories of treatment failure is still possible. Even when successful, though, these regimens are so toxic and cumbersome that they do not represent practical options for most people. More streamlined regimens are needed, which may be possible with some of the new drugs in development that present easier dosing schedules and at least partially escape cross-resistance with older drugs of the same class. Among the most advanced of the promising new agents are the protease inhibitors ABT-378 from Abbott Laboratories and tipranavir from Pharmacia & Upjohn. Amprenavir, a protease inhibitor from Glaxo that is on the verge of FDA approval, also has some promise in this area. Then there is Trimeris's virus-cell fusion inhibitor T-20, which works by a completely different mechanism than other available antiviral drugs. These four are reviewed in the first article of this issue.