Along with their new legitimacy as a potential treatment strategy, "drug holidays" are now known as structured treatment interruptions (STIs), and they are being studied in two very different contexts. First, some researchers have been exploring the use of STIs in a salvage population as a way of shifting viral population from resistant to wild type, perhaps eliminating the former (see Treatment Issues
, "Going Wild (Type) with Structured Treatment Interruptions
," September/October 1999). But, second, STIs are also being studied in people who have had undetectable, or very low, viral loads for an extended period. In this population, the rationale for the STI is to promote an immune response to HIV: an STI might allow endogenous HIV antigens to stimulate cellular immune responses. This could lead to partial immune control of the virus, allowing people to take drugs intermittently, or, in some cases, it could even result in complete immune control and the discontinuation of HAART. Such results are especially appealing because there is almost no chance that HAART can eliminate HIV from the body, because drug regimens are complicated, toxic, and difficult to adhere to, and because long-term side effects, such as lipodystrophy, are becoming more common and worrisome.
Nevertheless, STIs also carry significant risk to the patient -- without proven benefit -- so they are not currently recommended. Unfortunately, many patients are not aware of the theory behind STIs and are simply stopping therapy, which leads to an increase in viral load and possible loss of CD4 cells in an uncontrolled setting. Another potentially serious risk is the development of drug resistance. At this point, such negative outcomes are much more likely than the handful of documented success stories.
From Berlin to Bethesda and Brussels
The use of STIs to boost the immune system was initially suggested by the now famous "Berlin patient," who started therapy (ddI, hydroxyurea and indinavir) before seroconversion but stopped treatment three times since then. The first time, viral load increased. The second and third times, however, viral load remained undetectable (the second interruption was due to a bout of hepatitis A). Remarkably, since the third interruption started in December of 1996, therapy has not been reinitiated and viral load is still undetectable. Although HIV antibodies haven't been detected, he has CD4 cells that respond to HIV's p24 protein and cytotoxic T-lymphocytes (CTLs) that are primed for HIV-infected cells.
Somewhat similar results were obtained in a study at New York's Aaron Diamond Research Center. Thirty-eight men with primary HIV infection were put on AZT/3TC
plus either ritonavir
, indinavir, or ritonavir/indinavir. Eleven later discontinued therapy, two of whom had especially interesting results. Both stopped and restarted therapy several times, eventually stopping for two years. During that time, one's viral load ranged between 50 and 500 copies and the other's ranged between 50 and 2,000 copies. In addition, they had a strong population of anti-HIV CTLs.
In both cases, therapy was initiated during primary infection and treatment was discontinued several times. It is not known what, if any, role these factors played, but there is some reason to think they might be important. Several studies suggest that after the first interruption in treatment, the immune system can manage to get more control over HIV -- for some patients, viral loads rise more slowly and stabilize at lower levels. And many think that treatment during primary infection protects the immune system, which could mean that those treated early would have a better chance of controlling the virus without HAART. Some researchers are also using therapeutic vaccines, hydroxyurea, or IL-2 to help activate an immune response.
Recently, there have been more conference reports of patients who discontinue therapy but have had no or little rebound in viral load. First, there was NIAID's NoHRT study (ICAAC, abstract 689). Dr. Davey reported on 18 patients who started HAART during chronic infection. The median CD4 count was 921, median time that viral load had been below 500 copies was 108 weeks, and the median time of a protease inhibitor-based HAART regimen was 137 weeks. Twelve (67%) had been previously treated with IL-2. One of the 18 patients had no detectable HIV DNA in peripheral blood mononuclear cells (PBMCs) and seven had undetectable levels of latently-infected resting CD4 lymphocytes (five of these seven had taken IL-2); of ten that had lumbar punctures, all had a cerebral spinal fluid viral load below 50 copies; and three of five who had lymph node biopsies had fewer than 50 copies per ten million cells.
Despite the promising characteristics of this group, all had rebounds when they discontinued HAART and IL-2. Viral loads increased approximately 0.2 log per day and surpassed 50 copies in a mean of 11 days and 500 copies in 18 days. In addition, HIV was detectable in resting CD4 cells of all patients. No difference was noted in the rates of virologic rebounds between those who had and had not received IL-2 therapy. However, there was one notable result. The so-called "Bethesda patient" only had a slight rebound in viral load that ranged between 50 and 500 copies for six months. Unlike earlier cases, the Bethesda patient did not start treatment during primary infection.
Not to be outdone, a team from Brussels presented an STI study in which three patients discontinued therapy without a rebound in viral load for 4 to 24 months (ECCAT, abstract 413). An initial group of seven patients started a variety of treatments during primary infection and continued for 12 to 36 months. Researchers did not find any regimen-specific differences between the four who had rebounds in viral load and the three who did not. However, they did note that the rebounders, in contrast with those who did not have rebounds, had increased markers of activated CD8 cells prior to the STI.
HAART Initiated During Primary Infection
At the Infectious Disease Society of America's conference in Philadelphia last November, Rosenberg presented results of a study designed to test whether initiating HAART early in primary infection would preserve an HIV-specific immune response (abstract 725). A total of 25 patients started HAART before seroconversion when no HIV antibodies were detected but there were high viral loads (median 13.3 million copies). Mean CD4 count was 418 cells.
At baseline, patients were given a test to measure CD4 lymphoproliferative responses (LPRs), which determine the degree to which a patient's CD4 cells respond to HIV antigens. Results are given in terms of a stimulation index (SI). If the SI is below ten, it indicates an inadequate immune response. Initially three of the 25 (12%) had an SI above ten. HAART was initiated and all patients' viral loads were suppressed below the limit of detection (limit not given). After two months of drug therapy, 16 of 22 patients (73%) had an SI greater than ten (mean of 42) and two had an SI greater than 100. At month six, the mean SI increased to 44, where it remained through month twelve. At month twelve, thirteen of fifteen patients (87%) had an SI greater than ten. (The other two patients had drug-resistant HIV and responded to therapy slowly.)
After one year of successful HAART, two of the patients consented to at least two STIs. The first interrupted therapy for a few weeks. Viral load peaked at 17,000 copies after about three weeks and the SI was over 100. Therapy was successfully restarted for four months, at which point the patient developed hepatitis A and had a second (unplanned) treatment interruption of about four weeks. After two weeks, viral load rebounded to 37,000 copies, but 72 hours later it fell to 6,000 and, after another 48 hours, to 400. Upon recovery from the hepatitis A infection, the viral load rose to 7,000 copies, HAART was reinitiated, and viral load became undetectable again. During this second STI, the SI was over 800, one of the highest SIs recorded in an HIV-positive person. A third STI lasted about three months. Generally, viral load remained below 5,000 copies, although there were two occasions on which it rose to 12,000 (the second time, the patient had gonorrhea of the throat) before falling again. With a viral load of 3,600, the patient restarted therapy and viral load became undetectable.
The second patient underwent two STIs. During the first, which lasted about three weeks, viral load rebounded to 120,000; however, the SI went from 20 at baseline to 130. Viral load became undetectable once HAART was restarted. The second time, the patient's viral load remained below 5,000 copies for four months. At that point, therapy was initiated again and viral load became undetectable.
This year's Retrovirus Conference included several other STI studies in which therapy was started during primary infection. From a group of fifteen patients, Altfeld reported on seven who underwent an STI (abstract 357). Mean baseline viral load for the entire group was 13.8 million copies, but it became undetectable after initiating HAART. Viral loads remained undetectable for at least one year, when the present study began. Ten of the fifteen also showed low levels of CTL responses against one to four epitopes (HIV protein sequences).
All seven who discontinued therapy had a rebound in viral load within one to eight weeks, as well as an increase in CTL responses. There was both an increase in the number of HIV epitopes recognized (median two; range zero to three) and a greater magnitude of response to epitopes that had been previously recognized. In addition, two patients who had no CTL response prior to the STI developed one afterward. Two of the patients had a second STI, which resulted in a greater increase in cytotoxic T-lymphocyte (CTL) response.
Zala presented results of an STI involving patients who initiated d4T, ddI, and nevirapine therapy, with or without hydroxyurea (abstract 558). Of eighteen patients (half of whom were taking hydroxyurea) who had completed their 48-week follow-up, 83% had viral loads below 500 copies and 67% were below 50 copies. From this group, eight underwent an STI after a median of 57 weeks (range 53 to 75 weeks) of therapy. While five of the eight were receiving hydroxyurea, researchers did not indicate whether any of the eight had undetectable viral loads prior to the STI. Viral loads rebounded to over 5,000 copies in all eight within 21 days of the STI. After peaking (peak viral loads were not reported), viral loads dropped by a median of 0.69 log (range 0.48 to 1.05). No significant differences were found between those who did and did not take hydroxyurea. Of interest, one patient who was not taking hydroxyurea discontinued therapy due to peripheral neuropathy at week 20, and his viral load remained below 50 copies after 46 weeks without therapy.
At a Retrovirus late-breaker, Jin presented the first STI study that involved the use of an experimental, therapeutic vaccine (LB 12). Four patients who began treatment with AZT/3TC/indinavir between seven and 100 days (median 60 days) after infection were enrolled. All achieved an undetectable viral load (<50 copies) after therapy, and this was sustained for 2.5 years. At that point, they began a course of an experimental vaccine that combined recombinant ALVAC 1452 and gp160. One week after their final vaccination and a median of 1162 days (3.2 years) of HAART, the patients discontinued therapy.
Two patients had a rapid rebound in viral load, which became detectable at days thirteen and 23 and had a doubling time of 1.4 days. However, the other two did not have a detectable viral load until days 68 and 85, and doubling times were 4.5 and 3.2 days. In addition, the two with a delayed viral rebound had significant increases in CTLs to more than one viral antigen, whereas one of the normal rebounders had no measurable CTL response to the vaccines and the other only showed a response to the gag protein. Finally, after four and eight months off therapy, the slower rebounders had viral loads of 5,600 and 330 copies, respectively; at four months off therapy, the normal rebounders' viral loads were 3,500 and over 50,000 copies.
In these four studies, patients undergoing an STI consistently had increases in viral load, but several patients also had either a spontaneous drop (Zala, Rosenberg) or lower viral loads with subsequent STIs (Rosenberg). There were also indications of a cellular immune response during the STIs in three studies -- increased stimulation index (Rosenberg), greater CTL response (Altfeld, Jin), or a broader and more pronounced response to epitopes (Altfeld). And although hydroxyurea did not seem to have an effect on outcome, Jin's research indicates that a therapeutic vaccine might enhance a cellular immune response.
HAART Initiated During Chronic Infection
Returning to the IDSA conference, Haslett presented retrospective, cross-sectional data that imply patients who do not start HAART until chronic infection can also gain immune control over HIV during treatment interruption (abstract 15). Like Rosenberg (discussed above), Haslett compared the LPR SI of eleven patients who had stopped therapy for at least one week to eleven patients who had never stopped therapy. The "interrupters" had a median SI of 45, which was significantly higher than that of non-interrupters (median SI=3; p<0.05).
Again, several presentations at the Retrovirus Conference involved STIs in people who started HAART during chronic infection, several of which were updates of reports given at earlier conferences. In a study presented by Garcia (LB 11), 10 patients who were treated with d4T, 3TC, and ritonavir or indinavir underwent three STIs. They had been on HAART for 52 weeks and had a viral load below 20 copies for at least 32 weeks before the first STI. Prior to each STI, all patients also had a viral load below 20 copies and seven were below five copies (one was lost to follow-up after the first STI). The STIs were spaced six months apart, and after each STI patients restarted the same drug therapy.
After each STI, viral load rebounded with a doubling time of 2.23, 3.38, and 3.25 days for the first, second, and third stops, respectively. During the second treatment interruption, four patients initially had viral rebounds to pre-HAART levels, but levels then spontaneously dropped by 0.8 to 2.09 log. The four also showed HIV-specific CTL responses and strong CD4 cell responses to HIV antigens. Similarly, four patients had viral loads that were 0.6 to 1.5 log lower than pretreatment and showed HIV-specific cellular immune responses during the third STI.
A study by Ruiz also showed a slower doubling time, from a mean of 1.6 to 2.2 days, between 12 patients' first and second STIs (abstract 354). While there was no change in the percentage of CD4 and CD8 cells during the STIs, viral load rebound did lead to a significant rise in CD38 expression (an indication of CD8 cell activation). There was no noticeable difference in outcomes between the five who did and the seven who did not take IL-2 before the study.
In another IL-2 study, Smith examined nine patients on HAART and daily low-dose IL-2 therapy (abstract 355). (Typically, IL-2 is given in higher, cyclical doses.) HAART was discontinued, but patients remained on IL-2. Viral loads became detectable in all nine after 19 days and rapidly increased for the next two weeks (2.0 day doubling time) to almost 350,000 (±230,000) copies. However, over the next four weeks, viral load steadily decreased, stabilizing at a low of approximately 26,000 (±8,000) copies, a level that was significantly lower than the pre-HAART mean viral load (about 70,000 ±20,500, p<0.01). Just after viral load peaked, CD4 counts dropped by 24% compared to baseline. With viral rebound, there was also an increase in CD8 cells to about 200% of baseline, which remained high while viral load dropped. Without a control group, however, it is difficult to know the impact of this approach to IL-2 therapy.
Papasavvas compared CD4 and CD8 cellular immune responses in five patients on HAART who underwent an STI to five untreated controls (abstract 353). In contrast with the control group, the STI group had significant increases in CD4 cell responses against p24 and gp160, which were followed by significant increases in interferon-gamma-secreting CD8 T-cell responses against HIV envelope antigens. After a median STI of 46 days, three of the five patients resumed HAART and maintained or further increased their cell-mediated HIV responses. The other two remained off therapy with viral loads below 1,080 copies and continued high cell-mediated responses.
Lori presented a case-controlled study (abstract 352) comparing eight patients on HAART with nine on ddI and hydroxyurea (the PANDA cohort). In contrast with the HAART patients, the PANDA cohort had low but detectable viral loads and a strong HIV-specific cellular immune response. Groups were similar in terms of time on treatment (more than two years for both), CD4 counts (HAART=549±175, PANDA=495±127), and CD8 counts (HAART=874±203, PANDA = 805±335). However, seven of eight on HAART had a viral load below 50 copies while only one of nine in the PANDA cohort was below 50 copies.
Both groups stopped treatment for up to eight weeks. Failure after the STI was defined as a rebound in viral load to greater than 10,000 copies or CD4 count decrease to below 200 cells. Five of the HAART patients failed by week six (one at week two and four at week six) and restarted therapy, and the average viral load rebounded from 97 to 16,863 copies (2.25 log). In addition, HAART patients' CD4 counts dropped an average of 208 cells and CD8 counts increased by an average of 323 cells (CD4/CD8 ratios dropped significantly, to below 0.3). In contrast, none of the PANDA patients failed during the STI; viral load increased from 843 to 1,596 copies (0.27 log) but fell to pre-STI levels by week eight; and CD4 and CD8 counts, as well as the CD4/CD8 ratios, remained stable. Interpreting the results is difficult, however, since the PANDA group, unlike the HAART group, also had an eight-week STI before this study.
The largest prospective STI trial so far is the Swiss Intermittent Treatment Trial (SITT), which plans to enroll 120 patients. The trial design involves four cycles of two weeks off therapy followed by eight weeks of therapy. At week 40, treatment is discontinued indefinitely unless viral load rises above 5,000 copies. At the Retrovirus Conference, preliminary data were available (abstract 458). Median pre-HAART CD4 count was 398 cells and median viral load was 36,000 copies. Before beginning the study, patients had been on HAART for a median of 24 months and the median time with a viral load below 50 copies was 21 months. Of the 96 patients to undergo the first STI, 30% showed no increase in viral load. Fifty-four and 23 people went through the second and third STIs, respectively, but there was no evidence that viral rebounds were less frequent or smaller. Since this trial is much larger than the others, future data could prove especially interesting.
Don't Try This at Home
So far, STI trials have been small, and the number of people who have maintained an undetectable viral load without therapy is even smaller. Some context was provided by Sherer's retrospective examination of thirteen people who had viral loads below 500 copies after stopping therapy for a median of 2.5 years (Retrovirus, abstract 351). All of their antiretroviral regimens consisted of only one or two nucleoside analogs (median three months, range 1 to 54); two had received intermittent therapy, and two were treated during primary infection. Three of ten were heterozygous for the CCR-5 deletion, and two of twelve had clade A HIV. Pre-therapy viral loads were available for seven of the thirteen, and they ranged from below 500 to 9,000. They conclude that isolated cases of individuals with low or undetectable viral loads after stopping therapy do not necessarily show exceptional therapeutic effect, especially if they were treated during primary infection before a viral set point had been achieved.
Another reason for caution comes from Price's comparison of the viral load rebound in cerebrospinal fluid (CSF) to that in peripheral blood in four patients who stopped treatment (Retrovirus, abstract 306). Two patients had virtually no change in CSF/plasma viral load ratios (0.020 and 0.007). The other two, however, had a plasma viral load increase of 0.5-1.0 log and a CSF viral load increase of more than two to three log. As a result, the CSF/plasma ratio went from 0.01 to 0.61 in one patient and from 0.001 to 0.448 in the other. There was also an increase in white blood cell counts, from three to 31/mm3 in one and from zero to 46/mm3 in the other. While there were no neurological symptoms, the results suggest potential risks of STIs.
STIs also pose a risk of drug resistance. This is especially true with patients who are taking non-nucleosides (NNRTIs), such as nevirapine or efavirenz, since these drugs have long half-life. If a person discontinues all drugs simultaneously, nucleoside analogs and protease inhibitors will leave the body more quickly than the NNRTIs; and HIV only needs to develop one mutation in order to become resistant to all available NNRTIs. In the trials discussed above, the development of resistance has not been a problem. However, these have been small studies in very controlled environments, so the potential for resistance to develop is not really known. Moreover, it's possible that each STI allows latent viral reservoirs to be reseeded with new virus. This could increase the amount of viral DNA in the latently infected cells, making future eradication more difficult.
For the time being, then, the role of STIs in controlling HIV remains unclear, and they should never be undertaken without medical supervision. Determining whether or not they are useful -- and, if so, for whom -- requires more clinical studies. And while STIs have a number of attractions, they are a gamble. Those who have managed to suppress their viral loads for an extended period of time, especially if they have tolerable side effects, might be best off continuing with their current regimen or trying other options, such as a therapeutic vaccine.
STIs with Virologic Failure
In a late-breaker, Dr. Steve Deeks presented results from a prospective study of STIs in patients who were on protease inhibitor-based therapy and who had a viral load over 2,500 copies for at least twelve months (LB-10). Eighteen patients were randomized to discontinue therapy. They had about three years of protease inhibitor therapy and a detectable viral load for about 31 months (range, 28 to 33 months). Upon discontinuing therapy, mean CD4 count was 245 cells (range, 104 to 307), mean viral load was just below 40,000 copies (range, 10,000 to 100,000), and the median decrease in protease inhibitor susceptibility was 56-fold (range, 23 to 79). Viral load, CD4 levels and phenotypic resistance were measured weekly for the first 12 weeks, at which point therapy was restarted and tests were done every four weeks.
At 12 weeks, the median CD4 decrease was 94 cells (28 to 120) and the median viral load increase was 0.82 log (0.34-0.92). In 16 patients, the virus shifted to a protease susceptible (<2.5 fold decrease in susceptibility) phenotype -- or wild-type virus -- after two to 15 weeks (mean 8.5 weeks). Slower reversion to wild-type virus was associated with more protease inhibitor resistant mutations at baseline and with a smaller viral load reduction on therapy as compared to pretherapy viral load. Nucleoside analog resistance persisted in seven patients, but typically at a much lower level than at baseline. Replicative fitness (the ability of the virus to reproduce) increased from a median of 22.3% to 67.1% (p=0.004). Deeks noted that prior to the reversion to a wild-type virus, there was a slow decline in CD4 cell counts and a slow rise in viral load; however, after the shift CD4 counts dropped and viral load increased dramatically. He suggested that the former could be associated with residual antiviral activity and the latter with an increase in replicative fitness.
In four of eight patients who reverted to wild type, PBMC cultures (taken 12 to 36 weeks after discontinuing therapy) showed resistant virus that was identical to that at baseline. That is, while the virus in their blood had switched to protease inhibitor susceptible, resistant virus could still be found in PBMCs, which suggests that wild-type outgrew resistant virus but did not eliminate it. Moreover, it is possible that the other patients also still had resistant virus, even though tests could not detect it.
Researchers also looked at the fractional replacement rate of CD4 cells. Prior to interrupting therapy, the rate in those patients on "failing" therapy was 0.011/day, which was significantly lower than the rate of 0.032/day observed in patients who were not on therapy. This suggests that despite having a detectable viral load, those patients receiving drug treatment had a longer CD4 half-life than patients receiving no treatment. In addition, during STIs, there was a reduction in either CD4 half-life or CD4 production.
While Deeks did not draw any conclusions about STIs as a treatment option in this population, he did find evidence of continued antiretroviral therapy benefit even in cases where patients have resistant virus, which might be because the resistant virus is less fit than wild-type. This implies that, at least for patients with few or no treatment options, continuing drug therapy when viral load is detectable is more beneficial than discontinuing therapy altogether.
Back to the GMHC Treatment Issues Spring 2000 contents page.