Portal de Periódicos da CAPES

Supervised interruptions of antiretroviral therapy

2002; Lippincott Williams & Wilkins; Volume: 16; Linguagem: Inglês

10.1097/00002030-200216004-00022

1473-5571

Steven G. Deeks, Bernard Hirschel,

HIV/AIDS drug development and treatment

Introduction The current recommended approach to the treatment of HIV infection is to initiate at least three antiretroviral drugs, and to continue combination therapy indefinitely as long as viremia remains undetectable [1–3]. Most treatment naïve patients who initiate and adhere to standard regimens achieve and maintain undetectable levels of plasma viremia. As a result, CD4+ T cell counts increase and – in many patients — normalize after several years of treatment. This standard approach has resulted in dramatic declines in HIV-related morbidity and mortality [4,5]. Despite clear success, there is a growing appreciation that there are significant limitations with continuous administration of HAART. HIV does not appear to be curable, at least with currently available treatment options. Durable control of viral replication therefore requires the continuous use of multi-drug regimens; any attempt to simplify therapy and use a less potent regimen typically results in rapid virologic rebound [6–8]. Thus, if the goal of therapy is to prevent viral replication (and viral evolution), patients are required to remain adherent to complicated regimens indefinitely. This may be feasible for some patients, particularly as regimens become easier to take and less toxic. However, current regimens are often complex, poorly tolerated and expensive [9]. Long-term toxicity of therapy is now evident, including lipoatrophy, insulin resistance, hyperlipidemia, peripheral neuropathy, and accelerated liver disease [10]. For many patients, the immediate risk of therapy may outweigh the immediate benefit of therapy. Based in large part on these concerns, and the significant economic costs associated with treatment, ‘structured’ or ‘supervised’ treatment interruptions (STI) are now being considered in a variety of settings. There are many reasons to interrupt therapy in a controlled manner. The rationale for such interruptions is complex and often overlapping: 1. ‘Autovaccination’: The anti-HIV immune response tends to decrease during continuous antiretroviral treatment, perhaps because of the removal of antigenic stimulation [11]. Controlled treatment interruptions may result in expansion of HIV-specific immune responses. 2. ‘Intermittent therapy’ to reduce costs and drug toxicity: The vast majority of HIV infected patients live in resource poor countries [12]. Intermittent therapy may reduce costs, thus allowing more people to be treated. Intermittent therapy may also reduce toxicities and prevent ‘pill fatigue’. 3. ‘Salvage therapy’: A growing proportion of patients have limited therapeutic options due to the presence of highly resistant HIV. Interrupting therapy in a controlled manner may allow the emergence of drug-susceptible virus, thus allowing a more durable response to subsequent ‘salvage’ therapy. These are very diverse clinical situations. We shall examine each of them in turn, citing studies presented at meetings or published through September 2002. Preference is given to studies that have been published in the peer-reviewed literature. However, abstract presentations of unpublished data are also considered. Supervised treatment interruptions to stimulate anti-HIV immune responses The immune response against HIV There is substantial evidence supporting a protective role of HIV-specific cellular immunity in both treated and untreated HIV disease. As HIV-specific immune response emerges during seroconversion, plasma HIV RNA levels decrease significantly, eventually reaching a steady-state level that is unique to each patient and which predicts subsequent rates of disease progression (the viral load ‘set-point’) [13,14]. This reduction of viremia is temporally associated with the emergence of an HIV-specific cytotoxic T lymphocyte (CTL) response [15–17]. CTL escape mutants typically emerge during this time, indicating that immunologic pressure was sufficient to cause viral evolution [18–20]. Potent HIV specific T cell responses are often found in untreated patients who maintain low-level viremia for years (‘long-term non-progressors’) but not in patients with progressive disease [21–28]. Finally, and perhaps most convincingly, when CD8+ T cells are depleted in SIV infected monkeys, viremia increases. As CD8+ T cells reaccumulate, viremia decreases [29,30]. Collectively, these observations indicate that immunologic control of viral replication is the rule rather than the exception. It is therefore reasonable to assume that strategies aimed at enhancing this immune response could result in lower viral load set-points and an improved prognosis. The frequency of HIV-specific CD4+ and CD8+ T cells generally declines during HAART, presumably as a consequence of reduced antigenic exposure [11,24,31–35]. When HAART is interrupted after prolonged periods of viral suppression, the numbers of HIV-specific T cells generally increase [35–37]. Transient episodes of detectable viremia (viral ‘blips’) during treatment are associated with enhanced HIV-specific T cell responses, presumably because the level of residual viral replication is sufficient to maintain high numbers of memory T cells [37–39]. Ortiz and colleagues observed that patients with incomplete adherence had episodes of intermittent viremia, and that these episodes predicted both greater HIV-specific immune responses and more durable control of viral replication when treatment was subsequently discontinued [37]. These observations suggest that the magnitude of virus-specific T cell immunity is determined in part by the level of antigenic exposure, and that the preservation of large numbers of effector cells requires ongoing or intermittent antigenic exposure. There exists a complex and poorly understood dynamic relationship between the degree of virus exposure necessary to stimulate HIV-specific immunity (as a consequence of antigenic exposure) and the degree of virus exposure necessary to deplete HIV-specific CD4+ T cells (as a consequence of the cytopathic effects of viral replication) [40,41]. Thus, although interrupting therapy in patients with completely suppressed viral replication should result in expansion of HIV-specific T cells, it remains unclear how much virus is too much, or how often the immune system should be stimulated (i.e., how many interruptions are necessary). This has practical limitations in the design of STI studies. It is not clear if treatment should be interrupted for a fixed, brief period (e.g., 2 weeks as in the SSITT study described below) or interrupted until plasma HIV RNA levels increase to above a fixed maximum amount (e.g., greater than 50 000 copies RNA/mL). It is likely that if an optimal balance between the positive and negative consequences of viral replication exists, it will vary from patient to patient and be difficult to predict. In summary, three related observations provide a strong conceptual basis for treatment interruptions where the goal is to expand HIV-specific cellular immunity. First, most accept the central thesis that immunologic control of HIV commonly occurs in some clinical situations. Thus it is reasonable to assume that enhanced HIV-specific immune responses will result in a lower viral load set-point and an improved prognosis. Second, early initiation of HAART results in preservation of HIV-specific CD4+ T cell responsiveness (as measured by the lymphoproliferative response). Third, interruption of therapy expands the number of HIV-specific T cells, as defined by cytokine expression and/or T cell proliferation. Treatment interruptions after aggressive therapy of primary infection The response to STI may be greatest in those patients who initiate therapy very early during HIV disease. HAART-mediated viral suppression prevents viral evolution, thus reducing the risk that immunologic escape mutations will emerge. Early administration of HAART also prevents the irreversible loss of antigen-specific CD4+ T cells [31,42–44]. The importance of starting HAART as early as possible was perhaps most clearly shown in a series of experiments involving SIV infected macaques. When animals were treated very early, repeated treatment interruptions resulted in durable control of viral replication (in the absence of therapy) [45–47]. In contrast, when treatment was started during chronic infection, repeated treatment interruptions failed to control viral replication, and disease progressed rapidly [48]. Similar data have been generated in clinical studies involving HIV-infected adults. HIV-specific CD4+ T cell function (as measured by lymphoproliferation) is often preserved when HAART is initiated during primary infection [23,32,49]. In contrast, HIV-specific CD4+ lymphroproliferative responses decline when treatment is delayed [49]. Based in part on these concepts, Rosenberg and colleagues initiated a series of clinical studies involving recently infected patients. Most patients in this cohort initiated therapy prior to or at the time of seroconversion. Early treatment was associated with preserved HIV-specific CD4+ T cell lymphoproliferative responses; however, HIV-specific CTL responses were of low magnitude and limited breadth [44]. After achieving durable viral suppression on HAART, patients underwent a series of treatment interruptions, with therapy resumed if plasma HIV RNA levels increased to greater than 50 000 copies/mL once or greater than 5000 copies/mL during three consecutive visits. For most patients, each subsequent treatment interruption was associated with enhanced HIV-specific CTL responses and a lower off-therapy viral load ‘set-point’. Careful prospective immunologic studies revealed that sequential treatment interruptions were associated with increasing HIV-specific T cell responses [49,50], although it is not clear that enhanced antigen-specific cellular immune responses were the sole mediator of virologic control in these patients [51,52]. Despite these impressive results most other studies have failed to yield convincing data that treatment interruption during early HIV infection results in enhanced immunologic control of HIV. Miro and colleagues initiated HAART in 12 recently infected patients. In contrast to the patients reported by Rosenberg, these patients were generally treated after seroconversion and had limited evidence of HIV-specific CD4+ T cell proliferative responses. Sequential two-month treatment interruptions failed to convincingly show enhanced immunologic control of viral replication [53]. Markowitz and colleagues initiated HAART in a well characterized cohort of recently infected adults. Sixteen patients eventually discontinued therapy (11 having received a therapeutic HIV vaccine during HAART). All patients experienced viral rebound within weeks of their interruption. After achieving a mean peak level of viremia of 4.4 log10 copies RNA/mL, there was a spontaneous reduction of plasma HIV RNA levels. However, the steady level of viremia had a distribution similar to that observed in a cohort of untreated patients, suggesting that early and aggressive HAART did not result in enhanced control of viral replication. There was no consistent relationship between the immunologic response and the level of viremia, nor was there any evidence that vaccination affected the virologic outcome off HAART [54]. Several factors may have accounted for the failure of virologic control in this study, including (i) the initiation of HAART after sero-conversion, which presumably resulted in irreversible loss of virus-specific T cells and/or the selection of escape mutants, (ii) the use of long-term HAART, thus resulting in a time-dependent loss of immune response, and/or (iii) the use of a single interruption [54]: Theoretically, a single interruption may not be sufficient since predator/prey dynamics favor the virus over the host in most situations [37,55,56]. Indeed, an initial brief interruption may be required simply to expand existing HIV specific T cell responses to high levels, thus preventing rapid rebound during subsequent interruptions. Despite inconsistent clinical data, these studies provide a strong rationale for pursuing immune-based strategies aimed at controlling viral replication. Such strategies may require the use of therapeutic vaccines rather than treatment interruptions as a means to stimulate virus-specific immunity [46], although initial clinical studies in this area have been disappointing [57,58]. Treatment interruptions in chronically infected patients responding to HAART The vast majority of patients initiate HAART during chronic infection. How efficacious are treatment interruptions in these patients? Treatment interruptions in chronically infected patients are associated with an expanded HIV-specific T cell response in most studies [38,39,59–61]. Recent data indicate that this increase in HIV-specific CD8+ T cells is primarily due to an expansion of pre-existing epitopespecific T cells rather than the generation of T cells recognizing novel epitopes, and that much of the measured increase reflects a redistribution of T cells from lymphoid tissue to peripheral blood [35]. When treatment is interrupted or discontinued after months of HAART, viremia commonly rebounds back toward the pre-treatment set-point [55,56,62,63]. Thus, although a single interruption results in higher numbers of HIV specific T cells, it is clear that these cells act to reestablish the pre-therapy viral load set-point, and that repeated interruptions will be necessary to achieve greater immunologic control of viral replication. The effectiveness of repeated treatment interruptions in establishing a reduced viral load set-point off therapy has been addressed in several studies, most of which were small and uncontrolled. The results have been mixed. Garcia and colleagues, for example, performed a series of three treatment interruptions in 10 patients responding to HAART. Although virologic rebound was observed in all patients, the doubling times increased significantly between the first and third treatment interruptions. Patients who generated a stronger HIV-specific immune response generally had a lower viral load set-point at the end of the study. Ruiz and colleagues also observed longer viral load doubling times and enhanced HIV specific T cell responses in a small cohort of patients undergoing sequential treatment interruptions [64]. In contrast, Ortiz and colleagues performed a series of one month treatment interruptions in chronically infected patients responding to HAART and found no evidence of enhanced virologic control (despite clear evidence that STI resulted in a greater number of HIV-specific T cells) [65]. The Swiss-Spanish-Intermittent-Treatment-Trial The largest treatment interruption study completed to date is the Swiss-Spanish-Intermittent-Treatment-Trial (SSITT). This study enrolled 133 patients, all of whom were treated with HAART for at least 9 months prior to study entry and all of whom had suppressed viremia below 50 copies/ml during the preceding 6 months (median duration of undetectable viremia at study entry was 21 months). Antiretroviral treatment was interrupted for two weeks and then restarted and continued for eight weeks (Fig 1). After four such cycles of two-week treatment interruptions, therapy was discontinued and resumed at week 52 if viremia exceeded 5000 copies RNA/ml. The endpoints measured were the proportion of patients with a viremia below 5000 copies RNA/ml after 12 and 56 weeks of antiretroviral treatment.Fig. 1. Swiss-Spanish-Intermittent-Treatment-Trial (SSITT) Treatment Protocol.: The SSITT study enrolled 133 subjects with durable viral suppression on HAART (plasma HIV RNA levels < 50 copies/mL at week 0). Shaded areas reflect periods of treatment with HAART. Treatment was interrupted for two weeks and then resumed for 8 weeks. After four on/off treatment cycles, therapy was discontinued. The primary endpoint was measured at week 52.After repeated short treatment interruptions, about one in six patients appeared to stabilize his or her viremia at a level < 5000 copies/ml. Low pre-HAART levels of plasma viremia, a small number of lymphocytes with HIV provirus DNA at study entry, and the absence of rebound during the short treatment interruptions between week 0 and 40 appeared to predict durable partial control of viral replication after treatment discontinuation. During more than 500 treatment interruptions, there were only two instances of acute retroviral syndrome, a potentially serious reaction which is likely more common in patients who remain off therapy for more that two weeks [66–68]. What about the HIV-specific cellular immune response? Results from the study revealed a consistent increase in CD8+ T cells recognizing HIV peptides between study entry and week 39. However, this increase did not predict the level of viremia after treatment discontinuation; indeed, patients with higher rebounds tend to have more, not less HIV-specific CD8+ T cells. This result is largely inconsistent with the ‘autovaccination hypothesis’ for patients who started their HAART during chronic HIV infection. In summary, the data regarding STI in chronically infected adults has been disappointing as it pertains to the hypothesis that re-exposure of the immune system to autologous virus will result in the establishment of a low off-treatment ‘set-point’. Multiple factors may account for such failure. In particular, it is likely that the dominant virus population prior to HAART had CTL escape mutations, and that re-exposure to virus resulted in expansion of ineffective cells (original antigenic sin), or the expansion of cells that are only partially effective [35]. The irreversible loss of some immunologic effector cells (particularly HIV-specific CD4+ T cells) may also be a factor. In support of this hypothesis, several groups have reported functional and proliferative defects in CD4+ T cells during uncontrolled viral replication [69–71]. Use of immunosuppressive therapy in supervised treatment interruptions There is an accumulating body of evidence indicating that HIV-induced proliferation and activation of CD4+ T cells (particularly naïve cells) drives disease progression [72,73]. This raises the possibility that drugs aimed at suppressing T cell turnover or activation may have therapeutic benefit. Two recent pilot studies addressed the role of immunosuppresion prior to and during treatment interruption. Each study showed intriguing trends suggesting either greater virologic control and/or greater HIV-specific cellular immune responses in HAART-treated patients receiving hydroxyurea or mycophenolate mofetil [74,75]. It is noteworthy that hydroxyurea was critical component of the suppressive HAART regimen in several of the more successful STI protocols, including the ‘Berlin patient’ and most of the studies directed by Lori and Lisziewicz [36,45,48,76]. These studies provide indirect evidence suggesting that HIV-mediated T cell activation might preclude the generation of effective T cell immunity before and perhaps during treatment interruption. Supervised treatment interruptions to decrease drug exposure (intermittent therapy) Perhaps the two most compelling reasons to interrupt therapy in a controlled manner are to reduce drug toxicity and reduce drug costs. In our experience, most patients are interested in STI as a means to limit toxicity or to provide a break from the demands of taking medications. It is reasonable to assume that the shortterm toxicity of therapy will rapidly resolve once drugs are discontinued. However, it is not yet clear whether many of the long-term toxicities associated with antiretroviral therapy resolve during a treatment interruption [77]. There are two broadly defined approaches to intermittent therapy: short-cycle interruptions of days to weeks and long-cycle interruptions of months to years (Fig. 2). ‘Short-cycle’ treatment interruptions (e.g., one week on, one week off) are based on the observation that virologic rebound to greater than 50 copies RNA/mL is often delayed for a few weeks after therapy is interrupted [56], suggesting the possibility that resuming therapy soon after it is interrupted might prevent the negative consequence of virologic rebound. Dybul and colleagues performed a pilot ‘proof-of-concept’ study of 10 individuals who underwent repeated cycles of one week on HAART (ritonavir, indinavir, stavudine and lamivudine) followed by one week off HAART. Plasma HIV RNA levels remained below 50 copies/mL and CD4+ T cell counts remained stable in all patients adhering to the one week on/one week off schedule. Levels of serum cholesterol and triglycerides improved, but insulin resistance and anthropomorphic measurements did not [78]. Repeated short-term interruptions were not associated with enhanced HIV-specific immune responses; however, ‘autovaccination’ is not a rationale for the short cycle intermittent therapy approach. Two of ten patients were unable to rigidly adhere to the one week on/one week off schedule and consequently experienced virologic rebound.Fig. 2. Intermittent Therapy with a goal of reducing drug exposure.: Solid lines reflect plasma HIV RNA levels and dash lines reflect CD4+ T cell counts. Shaded areas reflect periods of antiretroviral treatment. In both examples HAART is initiated and maintained for 48 weeks. (i) Panel A: In the ‘short-cycle’ intermittent treatment approach HAART is interrupted and resumed before significant levels of viral replication occur. Although a one week-on/one-week off approach is shown, other approaches are now being studied (including five days on, two days off). (ii) Panel B: In the ‘long-cycle’ intermittent treatment approach, HAART is interrupted and resumed once peripheral CD4+ T cell counts decrease to a minimal level (e.g., 350 CD4+ T cells/mm3). Therapy is interrupted once CD4+ T cell counts increase to above a specified levels (e.g., 500 CD4+ T cells/mm3).One of the limitations of this study was the use of ritonavir/indinavir, a regimen that may not be readily available in resource poor regions where intermittent therapy is likely to be most needed. Recognizing this, Dybul recently initiated a similar pilot study using a non-nucleoside reverse transcriptase inhibitor (NNRTI)-based regimen [79]. The risk of interrupting an NNRTI may be greater than the risk of interrupting a protease inhibitor, because the ‘genetic barrier’ to NNRTI resistance is low [80,81]. Eight patients responding to HAART were placed on a repeated treatment schedule of one week of efavirenz, didanosine and lamivudine (all administered once daily) followed by one week off treatment. Plasma viremia remained undetectable during a mean follow-up of 31 weeks. Future studies will look at very short interruptions, including five days on followed by two days off. Although these data provide some support for shortcycle approach, significant concerns persist, including the ability of the average patient to ‘adhere’ to a complex week by week schedule, and the possibility that gradual viral evolution is occurring in the presence of lowlevels of drugs during the treatment interruption periods. Large randomized studies evaluating this approach are ongoing. The second approach involves ‘long-cycle’ treatment interruptions (Fig. 2). For example, some protocols are evaluating treatment schedules based simply on the calendar (e.g., two months on, two months off). Alternatively, the decision to interrupt and resume HAART might be guided by the CD4+ T cell count (e.g., interrupt therapy with CD4+ T cell counts increase to greater than 500 cells/mm3 and resume therapy when counts decline to below 250 cells/mm3). Preliminary data regarding the potential risks and benefits of long-cycle intermittent therapy can be found in observational studies of patients who discontinue therapy in a clinical setting. Hatano and colleagues evaluated 14 patients who interrupted therapy after prolonged HAART-mediated viral suppression. Plasma HIV RNA levels rebounded back toward the pre-HAART viral load set-point [55]. Tebas and colleagues retrospectively identified 72 patients who achieved durable viral suppression and subsequently discontinued therapy. CD4+ T cell counts declined by an average of 16 cells/mm3/month. The slope of the CD4+ T cell count decay was inversely correlated with the treatment-mediated CD4+ T cell count increase on therapy (i.e., those who gained a large number of cells on HAART were at the highest risk for experiencing rapid loss of cells in the absence of HAART) [63]. Gallant and colleagues observed similar results in 101 patients who interrupted therapy. There was a striking correlation between the pre-HAART and post-STI viral loads, suggesting that the off-therapy viral load set-point was not altered by prolonged treatment with HAART. CD4+ T cell counts declined toward the pre-HAART CD4+ T cell count nadir. Those patients whose nadir CD4+ T cell count was low (< 200 cells/mm3) generally had to resume therapy soon after treatment was interrupted [82]. Patients in these and other observational studies generally did well during their interruption, were able to remain off HAART for prolonged periods, and responded well to HAART once it was resumed [83]. The use of intermittent therapy is now being studied in several large randomized clinical studies. The ‘Strategies of Management of Anti-Retroviral Therapy’ (SMART) study compares continued viral suppression with intermittent therapy based on CD4+ T cell counts: stopping HAART when the CD4+ T cell count exceeds 350 cells/mm3, and resuming HAART when the CD4+ T cell count declines to less than 250 cells/mm3 or patients develop clinical complications. The Staccato trial compares three approaches: (i) continuous therapy, (ii) one week on, one week off, and (iii) CD4+ T cell count guided interruptions. Results from these studies are not expected for several years. The ‘Developing Antiretroviral Therapy’ (DART) study will randomize approximately 3000 patients in Uganda and Zimbabwe to continuous versus intermittent therapy (three month on/three months off); a second randomization will compare the role of aggressive laboratory monitoring versus clinical monitoring only. Supervised treatment interruption in patients with drug-resistant HIV In many patients — perhaps most — HAART is not able to durably suppress viral replication to below the level of detection (plasma HIV RNA levels below 50 copies/mL). Reasons for this high virologic failure rate are complex, and include: the sequential use of monotherapy or other suboptimal regimens, nonadherence, and the use of drugs with limited antiviral potency. As a consequence of ongoing viral replication in the presence of drug, mutations associated with drug-resistance are selected and maintained [84]. High-level multi-drug resistant variants typically emerge after prolonged treatment failure, making a complete virologic response to subsequent therapy unlikely [85]. Although many treated patients derive sustained immunologic and clinical benefit from HAART despite incomplete viral suppression, this benefit is unlikely to be permanent in most patients [86]. Novel therapeutic approaches for patients with drug-resistant HIV are needed. Mutations associated with drug-resistance often reduce the replicative capacity, defined as the ability of the virus to replicate in vitro in the absence of drug. The virus is considered less ‘fit’ than the wild-type variant (fitness refers to the ability of two species to compete in a defined environment; the resistant variant is more fit than the wild-type variant in the presence of drug but is often less fit in the absence of drug) [87]. When patients with multi-drug resistance discontinue HAART, the wild-type virus is able to outcompete the resistant virus; within a few weeks to months of an interruption, the dominant plasma virus population shifts from drug-resistant to drug-susceptible (Fig. 3).Fig. 3. Schematic of viral dynamics after interruption of therapy in patients with drug-resistant HIV.: Lines indicate plasma HIV RNA levels before, during and after interruption of HAART (solid line: total HIV RNA level; long-dash line: drug-resistant HIV RNA level, short-dashed line: wild-type HIV RNA level). Shaded areas reflect periods of antiretroviral treatment (treatment is interrupted at week 0 and resumed at week 16). Plasma HIV RNA levels increase immediately after interruption of treatment, perhaps reflecting the removal of treatment exerting some degree of residual antiviral activity against the drug-resistant variant [88]. Archived wild-type HIV RNA (replicating at low levels or stochastically emerging from latent infected cells) begins to replicate after treatment interruption. Due to greater replicative fitness in the absence of therapy, wild-type HIV RNA rapidly outcompetes drug-resistant HIV, and becomes the dominant virus population after approximately 8 weeks [88]. Resumption of aggressive therapy results in rapid declines of the wild-type variant. Theoretically, the ability of the residual drug-resistant virus to escape HAART is reduced given its reduced level at the time treatment is resumed (i.e., the level of antiviral potency necessary to durably suppress the drug-resistant variant may be lower at week 16 than at week 0 given the reduced total viral burden of the resistant variant).The proportion of patients who experience a shift in viral phenotype during STI varies from study to study [88–92]. The likelihood that a shift will occur likely depends on duration of HAART (patients treated for very long periods may have few cells containing replication competent wild-type HIV and thus be less likely to experience a shift) and on the type of mutations present (drug-resistant mutations from protease inhibitor-treated patients may be less stable than mutations from NNRTI-treated patients given the predicted impact these mutations have on viral fitness). In one study, patients with a greater treatment-mediated CD4+ T cell increase were more likely to experience a rapid shift in drug-resistance than patients with a less robust treatment-mediated CD4+ T cell gain [88]. This greater increase in CD4+ T cell counts may be a surrogate for a poorly fit virus [93]; thus, it is not surprising that those patients with the greatest CD4 T cell gain despite virologic failure experience the most rapid loss of drug-resistance when therapy is interrupted. The shift in viral phenotype occurs as a consequence of the outgrowth of a pre-existing and presumably archived wild-type variant; it does not appear to occur as a consequence of ‘back-mutations’ [94]. This shift in virus population from drug-resistant to drug-susceptible is temporally associated with rapid increases in plasma viremia and rapid decreases in CD4+ T cell counts, indicating that the wild-type variant replicates more efficiently in the absence of drug than the drug-resistant variant [88,92]. Although the drug-resistant variant wanes to levels that are no longer detectable using conventional assays, its clear that resistant HIV persists at very low levels and may re-emerge under selective pressure [88,94–97]. Could this shift in drug-phenotype from resistant to susceptible allow a patient to respond in a more durable manner to subsequent salvage therapy, as first suggested by Miller and colleagues? [92] Although most studies suggest that the resistant variant persists at low levels after prolonged treatment interruptions [88,94–97], it remains possible that the ability of partially effective regimen to control a given resistant virus population will be enhanced if that virus population is allowed to wane to a very small population size. In an ongoing study based at San Francisco General Hospital, most patients who interrupted therapy experienced a shift in virus phenotype. Patients who subsequently initiated a new regimen containing no drugs to which their pre-STI virus was susceptible generally did poorly after therapy was resumed. Patients who initiated a regimen containing only two drugs which were fully active against the pre-STI virus generally did well; however, such patients may have done well without first undergoing the substantial risks of an STI. However, all patients in this study who resumed a regimen containing only one drug to which their pre-STI virus was fully sensitive achieved durable viral suppression after resumption of therapy. Historically, such patients generally do not respond well to salvage therapy [98–101].Table 1: Known and expected sides effects of antiretroviral drugsThese studies, and others, provided the theoretical basis necessary to support large randomized clinical trials of STI in the setting of drug-resistant HIV. Three randomized studies addressing this hypothesis were recently completed. Katlama randomized 68 patients with highly resistant HIV and advanced HIV disease to one of two treatment arms: immediate aggressive salvage therapy (7 to 8 drugs) versus an 8-week treatment interruption followed by a similar salvage regimen. The group who underwent a treatment interruption had a remarkably improved virologic response to therapy at week 24 when compared to those who did not undergo a treatment interruption (mean decrease in plasma HIV RNA levels from study entry through week 24 of HAART was -1.08 and -0.39 in the STI vs. immediate HAART groups, respectively; P = 0.01). There results are puzzling, however, as only a minority of patients experienced a complete loss of drug-resistance mutations during the short treatment interruption. Ruiz randomized 46 patients with moderate levels of drug-resistance to either immediate salvage therapy versus a 3-month STI followed by the same fixed regimen. Again, only approximately 50% of the patients in the STI group had a significant shift in their dominant virus population. Patients did well in this study, with no trend suggesting a benefit in patients who underwent an STI. Finally, Lawrence and colleagues from the CPCRA enrolled 274 patients into their long-term, multi-center clinical endpoint study (CPCRA 064]. Participants with genotypically proven multi-drug resistant virus [2–3 class failure) were randomized to immediate ‘optimized’ therapy vs. a 16 week STI followed by optimized therapy. This study was closed to accrual prematurely by the Data and Safety Monitoring Board. A lower than expected clinical event rate was seen and the preliminary data favored the control (immediate therapy) arm, making it unlikely that the study would demonstrate an advantage with STI. Notably, more than one-third of the patients in the STI group experienced a clinical event or lost more than 50% of their CD4+ T cell counts during the four month treatment interruption, underscoring the risky nature of this approach. In summary, the hypothesis that an STI might be beneficial in patients with multi-drug resistant HIV remains viable. However, the risk of disease progression and irreversible immunologic damage is perhaps greatest when therapy is interrupted in patients with limited therapeutic options. Thus, the patients who are most desperate for a novel treatment strategy such as STI are unfortunately the patients who are most likely to derive harm from an STI. Risk of treatment interruptions Although the benefits of STI remain controversial and poorly defined, there is general consensus that STI can be very risky [1,2]. Interrupting an effective regimen is invariably associated with rapid increases in plasma HIV RNA levels and rapid decreases in peripheral CD4+ T cell counts. A small proportion of patients develop an ‘acute retroviral syndrome’, with case reports indicating that such reactions may be quite severe [66–68]. Plasma HIV RNA levels increase in tissue sanctuaries during an STI, particularly within cerebral spinal fluid (CSF) [102]. Such rapid increases in CSF viral replication are likely associated with some injury to the CNS [103]. Rapid increases in viral load within genital fluids have not been well described, but presumably occur given the consistent relationship between plasma and genital fluid viral loads. It is therefore likely that the dramatic increases in viral load immediately after a treatment interruption will result in increased risk of HIV-transmission [104]. Some have argued that STI should be avoided due to its public health implications [105]. Among patients responding well to HAART (plasma HIV RNA level < 50 copies/mL), there is a risk that treatment interruption will select for drug-resistant HIV [106,107]. Theoretically, suboptimal drug levels during the early treatment interruption period may allow virus to replicate in the presence of drug, thus leading to the selective emergence of drug resistance mutations. However, given the observation that the level of viral replication is very low during the first few days of an STI, it remains unclear how an interruption could select for a resistance mutation [78]. Indeed, it is possible that the resistant variant developed during treatment (or was present prior to the first HAART regimen and remained archived during treatment), only to become detectable once treatment was interrupted. Recent observations that resistance can emerge even as plasma HIV RNA levels remain less than 50 copies/mL provide some support for this hypothesis [108,109]. Thus, the reports of resistance emerging during early treatment interruptions do not necessary indicate that the interruption was the cause of the resistance evolution. In the SSITT study, after more than 500 treatment interruptions and 150 patient-years of follow-up, there was only a single case of drug resistance that required the initiation of a ‘salvage’; regimen. Although the hypothesis that interrupting therapy causes drug-resistance remains viable, the level of risk remains undefined and it is not yet clear whether the risk of resistance during STI is greater than the risk of resistance during continuous treatment. Most patients who interrupt therapy experience a decline in CD4+ T cells counts. This decline may be quite rapid in some patients. In the SSITT trial, there was a median CD4+ T cell loss of 14 cells/mL/week during the first twelve weeks of the final treatment discontinuation (Fig. 1). If the treatment interruption continued for more than 12 weeks, the rate of the CD4+ T cell decline decreased to 1.7 cells/mL/week. In some studies, the rate at which CD4+ T cell counts decline is highly correlated with the treatment-mediated CD4+ T cell count increase [63,82]. In other words, those who experienced significant CD4+ T cell gains on HAART are likely to experience rapid CD+ T cell declines in the absence of HAART. Why CD4+ T cell counts decline so rapidly is not yet known, although it is likely to be due to virus-mediated destruction of cells and redistribution of cells from the periphery to lymphoid tissue. Once therapy is resumed after an STI, CD4+ T cells counts generally increase. In the SSITT trial, the median CD4+ T cell count after 6 months of re-treatment was identical to the CD4 count before STI. Although regaining cells lost during a treatment interruption is commonly observed once therapy is resumed, it remains unclear whether immunologic function is fully regained. It is noteworthy in this regard that recent studies suggest that the CD4+ T cell count nadir is a stronger predictor of T cell function than the current CD4+ T cell count on therapy [110]. This observation argues that the process of losing and regaining cells during periods of intermittent therapy may result in permanent immunodeficiency. Finally, there remains the difficult to quantify psychological risk of interrupting therapy. Some investigators have observed that qualify of life improved significantly during an STI, and that patients had subsequent difficulties adjusting to the demands of treatment once treatment was resumed [111]. In one study performed at UCSF, two patients who had been stable on long-term therapy interrupted therapy in a controlled manner. Both patients were unable to subsequently re-commit to continuous treatment and eventually died from HIV-related causes [88]. Conclusion After a period of therapeutic maximalism (‘hit hard, hit early’) [112], there has been a dramatic shift toward a more conservative treatment approach, including the use of treatment interruptions. To what degree this rapid philosophical shift is justified remains controversial [113]. Studies of treatment interruption have proliferated, but few have yet appeared in the peer-reviewed literature, and none have the size and follow-up needed for definitive conclusions. However, until therapy becomes highly effective, easy to administer, safe, well tolerated and widely affordable, some if not most patient populations will likely derive some benefit from the use of controlled treatment interruptions. Defining which patients will benefit and how to derive that benefit is not possible given the current literature. Treatment interruption as a long-term strategy should therefore be limited to research settings when possible. Acknowledgements SD acknowledges support from the NIH (RO1 AI052745), the UCSF Center for AIDS Research (P30 MH59037), and the University-wide AIDS Research Program (CC99-SF-001, ID01-SF-049). BH acknowledges support by the Swiss HIV Cohort Study, supported by the Swiss National Science Foundation (Grant no 3345–062041).