January 29, 2003
When an investigational drug causes a 0.4 log drop in viral load, it is evidence of biological activity. When the first large auto-vaccination study produces the same result, it signals the end of an entire field of HIV research.
October 2002 saw the first publication of results from the Swiss-Spanish Intermittent Treatment Trial (SSITT), the largest study of structured treatment interruptions (STIs) in chronic HIV infection conducted to date. The design of SSITT was based on the "autovaccination hypothesis" -- the idea that short interruptions of HAART might augment the immune response to HIV by exposing the immune system to brief, time-limited bursts of viral replication. Evidence supporting this hypothesis has been reported from studies in acute HIV infection, but results so far in chronic infection have been mixed, at best, and highly controversial. If the media stories that accompanied the release of the SSITT data (published in J. Virology and PNAS) are to be believed, the trial was an abject and definitive failure: "Drug Holidays Not Beneficial for Fighting HIV" quoth the notoriously credulous Reuters Health. A commentary accompanying the PNAS paper, authored by Ume Abbas and John Mellors, was headlined with a lengthy dismissal: "Interruption of antiretroviral therapy to augment immune control of chronic HIV-1 infection: Risk without reward" (Abbas 2002).
But not everyone believes that SSITT represents the final word on autovaccination in chronic infection. A number of other researchers involved in STI studies have pointed out potential shortcomings in the study design, and some even see reason to be encouraged by the data despite these criticisms. The trial did produce a statistically significant result: when pre-treatment viral loads were compared with the viral load setpoint off therapy at the end of the study, there was an average decline of around 0.4 logs (from an average of 19,320 copies pre-HAART to 7,396 copies off therapy at the end of SSITT). While meager and -- as the SSITT authors argue -- likely transient, this change in viral load is surely deserving of closer scrutiny. Instead, the researchers choose to downplay the outcome despite the statistical significance. Abbas and Mellors go a step further, arguing that the investigators may have included viral load data from participants who had restarted HAART in this analysis and that, even if the viral load result is accurate, the risks associated with STIs in chronic infection -- such as the potential development of drug resistance -- are too great to justify further research. Given that any possibility of enhancing immune control of HIV replication in chronic infection could potentially have huge implications for the future of treatment (potentially reducing both the costs and toxicities of long term HAART, for example) it seems important to question whether the SSITT data really support these conclusions.
The first broad overview of the SSITT data was provided by principal investigator Bernard Hirschel at the 2002 Retrovirus Conference (available via webcast at: www.retroconference.org//2002/frame.htm). The trial enrolled 133 chronically infected individuals with CD4 counts over 300 and viral loads less than 50 copies on HAART, with a history of at least six months suppression to less than 500 copies prior to study entry. All participants underwent a two week STI followed by eight weeks of re-treatment, and this cycle was repeated four times prior to an open-ended treatment interruption at week 40. Anyone whose viral load was not re-suppressed to less than 50 copies during the eight week re-treatment periods was automatically excluded from the protocol. At the end of the study, HAART was reinitiated if symptoms of acute HIV infection occurred, or if the viral load measured over 50,000 copies on three occasions, over 100,000 copies twice or over 500,000 once. Hirschel reported that 43 of the original 133 participants were withdrawn from the study prior to week 40, mainly due to viral loads failing to decline below 50 copies within the eight week retreatment periods. An additional 23 withdrew between weeks 40-52, in this case primarily because of viral loads above the thresholds for reinitiating HAART. (It is somewhat unclear to what extent these individuals are included in the published papers -- three are part of the J. Virology substudy, and the authors explain how the reinitiation of HAART is handled in their analyses, but the PNAS article is vague on this point.)
Of the 67 evaluable participants discussed in Hirschel's presentation, 23 maintained viral loads below an arbitrary pre-defined "responder" threshold of 5,000 copies at week 52, but 7 had viral loads in this range before ever starting HAART. The other 44 had viral loads above this threshold, but did not meet the protocol-mandated criteria for restarting therapy. During extended follow-up out to week 96, 4/23 "responders" and 12/44 "non-responders" eventually met these criteria and restarted HAART.
The newly published SSITT studies represent a more detailed exploration of the data described by Hirschel, focusing on the immunological and virological outcomes. The individuals included in these analyses are not broken down into "responder" and "non-responder" categories, but are evaluated as a group. In PNAS, the research team -- led by immunologist Annette Oxenius from the Nuffield Department of Clinical Medicine, John Radcliffe Hospital, Oxford, UK -- reports on the 97 participants in the trial that were recruited in Switzerland (Oxenius, 2002a). In the Journal of Virology, the same team focuses on a subset of 13 individuals for whom detailed data on CD8 T cell responses and HIV viral load were available (Oxenius, 2002b). The titles of the papers serve as pithy summaries of the conclusions: "Stimulation of HIV-specific cellular immunity by structured treatment interruption fails to enhance viral control in chronic HIV infection" (PNAS) and "HIV-Specific CD8 T-Cell Responses Do Not Predict Viral Growth and Clearance Rates during Structured Intermittent Antiretroviral Therapy" (J. Virology).
Taking the larger PNAS study first, the logical presumption from the title is that the participants' viral load remained unchanged compared to their pre-therapy baseline. However, as stated earlier, it did not: "A comparison off plateau viral to pretreatment viral load showed that plateau viral load was significantly lower than pretreatment viral load (P = 0.005, paired t test); however this difference was small (mean log10 pretreatment viral load = 4.2826 and mean log10 plateau viral load 3.896)." Although it is unclear in PNAS whether the reinitiation of therapy was accounted for in this viral load analysis, the J. Virology substudy echoes the finding (an average 0.4log drop in viral load, in this case from a mean of 4.6 to 4.2logs) and clearly states that the data from individuals who restarted HAART was based on two or three samples taken just prior to beginning therapy. It is therefore a little surprising when the researchers go on to say: "This result is comparable to studies of single treatment interruptions where the ensuing plasma viral load was similar to pretreatment viral load (citing Davey, 1999)."
A potential explanation for the downplaying of the viral load outcome may lie in the immunology results. The investigators assessed HIV-specific T cell responses using an ELISpot assay, which counts how many T cells are able to produce the cytokine interferon-gamma when stimulated with peptides derived from HIV. Assessment of the magnitude of the HIV-specific CD8 T cell response using this technique revealed "heterogeneous (mixed) patterns," with some individuals showing increases from the first STI onwards while others experienced no change until the final open ended STI, which led to an increase in the number of HIV-specific CD8 T cells in the majority (but not all) of the participants. Most importantly, no correlation between the magnitude of these responses and viral load levels off therapy could be found. At first blush, this finding implies that HIV-specific CD8 T cells were playing little or no role in controlling HIV replication. However, another potential explanation is that the ELISpot assay being used to measure the CD8 T cell response is not accurately capturing the cells that are exerting an anti-HIV effect. While still controversial, two major lines of evidence support this possibility: 1) CD8 T cells can continue to make the cytokine interferon-gamma even when other important functions have been compromised, and 2) the peptides used in this study to assess whether CD8 T cells were targeting HIV were derived from laboratory isolates of the virus, and these peptides may differ significantly in structure when compared to those derived from each individual's HIV (known as autologous virus). These concerns are discussed further in the "Implications and Future Questions" section, below.
Evaluation of HIV-specific CD4 T cell responses (using the same assay) was restricted to a subset of 30 individuals, at fewer timepoints (baseline, week 39 and once during the final off-therapy period). The magnitude of these responses increased in all 30 participants between weeks 0 and 39 of the study, but their fate during the final cessation of therapy differed depending on the pre-HAART baseline viral load. In people whose baseline viral loads were below the median of 30,976 copies, the average number of HIV-specific CD4 T cells continued to increase at the end of the study. In all but one of those above this median pre-HAART viral load level, the average HIV-specific CD4 T cell response declined during the open-ended STI compared to week 39, while remaining above the week 0 value. The paper does not explicitly state whether an attempt was made to correlate HIV-specific CD4 T cells with the eventual viral load outcome, although such correlations have not been reported in other studies that employed the ELISpot assay. Given that a significant body of recent data suggests that CD4 T cells play a key role in long term control of many viral infections, including HIV, the absence of a comprehensive analysis of these responses in SSITT participants represents another shortcoming of the study.
In the intensive substudy reported in J. Virology, regular blood sampling allowed the researchers to calculate viral "growth rates" and "clearance rates" during the off- and on-therapy periods, respectively. The growth rate was defined as the speed at which viral load increased after each treatment interruption, and it was found that the rate of increase slowed over the course of successive STIs, a potentially encouraging finding. But again, no correlation could be found between the change in viral load growth rate and the HIV-specific CD8 T cell response as measured by ELISpot. The clearance rate (the speed of viral load decline upon HAART resumption) did not change over the course of the study.
The J. Virology paper also takes a closer look at changes in the breadth of the CD8 T cell response as a result of repeated STIs. For each individual, an average of 18 different epitopes (small slices of viral proteins) from HIV was used to assess how many parts of the virus were being targeted by CD8 T cells. A median of three new epitopes were targeted by week 52 of the trial, although again the development of these new responses could not be correlated with the eventual viral load setpoint. It may nevertheless be encouraging that responses to previously unrecognized epitopes were observed, as this typically requires activation of naïve T cells and their conversion to a memory T cell phenotype. Naïve T cell levels decline over the course of HIV infection, and it has previously been unclear whether there is potential to mount responses to new viral epitopes in chronic infection.
The J. Virology paper also includes complete pre-therapy and post-SSITT viral load data for all 13 individuals included in the substudy (see table). It can be observed that 10/13 experienced a decline in viral load, from an average of 160,877 copies to 26,948 copies (a drop of 133,929 copies or about 0.8 logs). In contrast, three participants had an increase in viral load from an average of 23,191 copies to 242,710 copies (an increase of 219,519 copies or around 1 log). Based on these data, it would be of interest to know the proportion of individuals experiencing a decline versus an increase in viral load in the SSITT cohort as a whole, and whether factors could be identified which would help predict these divergent outcomes.
A major limitation of SSITT is that there was no control arm, preventing a clean analysis of the effects of repeated STIs on viral load setpoint as compared with continuous HAART followed by a single interruption. At the Retrovirus conference, Hirschel stated that such a control arm would not have passed muster with his Institutional Review Board when the study was designed in 1999 (although he did not explain why multiple interruptions were considered more ethical than a single interruption at that time). Additional criticisms of the study design have been voiced by other researchers, particularly regarding the fixed eight week retreatment periods, which for some individuals is insufficient time to control viral load to less than 50 copies. As mentioned by Hirschel at the conclusion of his talk, Luis Montaner from the Wistar Institute in Philadelphia has picked up on these points and used them to inform the design of a 42-person STI trial in chronic infection that is due to be completed at the end of this year. The study includes a continuous therapy control arm and employs a series of STIs of 2 weeks, 4 weeks and 6 weeks duration. Both arms then undergo an open-ended interruption at the end of the trial. The timing of each STI is individualized such that no participant interrupts HAART until viral load has been successfully re-suppressed to less than 50 copies.
Although the failure to identify an immunological correlate may partly explain why the researchers shied away from making too much of the observed change in viral load setpoint, this leaves some critical questions unanswered, particularly in terms of whether the responses measured by ELISpot truly reflect functional HIV-specific immunity. A number of recent studies have suggested that this may not be the case, demonstrating differences between ELISpot results and those obtained using techniques that measure HIV-specific T cell proliferation (the ability of T cells to copy themselves in response to the virus, which may an important function). A team of researchers led by Mark Connors at the NIH has recently shown clear differences in the proliferative responses of HIV-specific CD8 T cells in long-term non-progressors (LTNPs) compared to both untreated and HAART-treated individuals with progressive disease (Migueles 2002). These differences were not apparent when the HIV-specific CD8 T cell response was assessed based on interferon-gamma production (the criteria used in ELISpot assay). British immunologist Francis Gotch has reported a similar disconnect between proliferative capacity and interferon-gamma production when HIV-specific CD4 T cells from LTNPs and progressors are compared (Wilson 2002).
The SSITT studies also used selected epitopes from laboratory HIV isolates to measure the magnitude of the virus-specific T cell response, and new data suggests that, due to viral mutation, these epitopes may not match those of the viruses present in each individual (Lee 2002). This problem is most dramatically demonstrated by the case of a SSITT participant infected with HIV from subtype A, in whom no CD8 T cell response could be detected because the peptides used in the ELISpot assay were derived from a subtype B virus. It remains possible that additional functional assessments and/or assays employing autologous HIV isolates from each participant might uncover a closer correlation between HIV-specific immune responses and control of viral load. The role of antibodies in controlling viral load may also deserve further investigation, as no data addressing this question were reported by the SSITT team. The few analyses of antibody responses in STI studies conducted to date have hinted at some improvement in neutralization capacity (a measure of the ability of HIV-specific antibodies to inhibit viral replication), but the information is very limited (Montefiori 2001).
The SSITT researchers did uncover novel evidence that all HIV-specific T cells -- at least as measured by ELISpot -- may not be created equal. As mentioned in PNAS, and recently described more extensively in the journal AIDS (Oxenius 2002c), the magnitude of the HIV-specific T cell response detectable by ELISpot when participants were on HAART (and viral load was suppressed) was inversely correlated with the pre-HAART baseline viral load. In other words, people with higher viral loads prior to starting HAART had fewer detectable HIV-specific T cells when on therapy, and vice versa. This result held true for both HIV-specific CD4 T cells and HIV-specific CD8 T cells.
The implication of these data is that in the setting of untreated infection, when HIV-specific T cells are constantly being activated by ongoing HIV replication, the ELISpot assay captures a mixed population of T cells, only some of which are truly exerting an anti-HIV effect. When viral replication is suppressed by HAART, a subset of the HIV-specific T cells that were measurable prior to treatment appear to persist in a de-activated or "resting" state, although they can still make interferon-gamma and thus are detectable using the ELISpot technique. The statistically significant correlation between the numbers of these resting HIV-specific T cells detected after individuals start HAART and the pre-treatment viral load suggests that these T cells represent the population that was mounting an effective anti-HIV response prior to the initiation of therapy. This finding also suggests that another approach to assessing the effects of STIs on functional HIV-specific T cell immunity might be to look for changes in the size of this resting memory population during the periods when study participants are on HAART.
The overarching theme that emerges from these data is one that has been sounded since the very earliest days of HIV research: the immunological correlates of control of HIV replication are opaque and need to be clarified. Recent basic immunology research in animal models has shed new light on the development and maintenance of virus-specific T cell and antibody responses, but efforts to translate these findings to the setting of human HIV infection are in their infancy. In light of this fact, it seems important to question whether the "small" reduction in viral load setpoint seen in SSITT is really a basis for ending all research evaluating the immunological and virological effects of STIs in chronic infection, as Abbas and Mellors argue in their commentary. When monotherapy with nucleoside analog antiretrovirals was shown to produce a small and transient drop in viral load, these drugs were not dismissed, but rather improved by their eventual inclusion in HAART regimens which produced more lasting effects. Similarly, it can be argued that SSITT represents a first attempt at enhancing control of HIV replication through the use of STIs, and the results -- by showing that viral load setpoint can indeed by reduced -- set a standard for future STI studies in chronic infection to improve upon. However, neither the SSITT authors nor Abbas and Mellors appear very willing to countenance such a possibility.
In part, the negative interpretation of the SSITT results is driven by concerns regarding the risks associated with the use of STIs in chronic infection. Abbas and Mellors cite several potential risks in their commentary, including: symptomatic acute retroviral syndrome, reseeding of viral reservoirs, CD4 T cell decline, increased risk of transmission, poor subsequent adherence to continuous therapy and the potential risk of drug resistance due to declining drug levels during the few days immediately after stopping HAART.
Bernard Hirschel has reported that two cases of acute retroviral syndrome were observed during SSITT, although both successfully resolved upon resumption of HAART. In terms of CD4 T cell decline, average counts dropped during the open ended interruption at the end of SSITT (from an average of 792 at week 40 to 550 at week 96), but Hirschel showed that they were quickly restored in re-treated participants (increasing from 574 to 690 after 16 weeks of HAART). The reseeding of viral reservoirs is an oft-cited but currently somewhat nebulous concern given the difficulty of accessing and accurately measuring such reservoirs. One analysis of the peripheral blood mononuclear cell (PBMC) HIV reservoir over the course of successive STIs has been conducted by Cecile Tremblay from Bruce Walker's group at Massachussetts General Hospital in Boston. Tremblay found no increase in the viral reservoir measured as infectious units per million PBMCs, but it's important to note that this study involved individuals with acute infection and may not be generalizable.
The possibility of an increased risk of HIV transmission during STIs is well supported by data demonstrating a link between viral load levels and the probability of transmission. Indeed, one suspected case of HIV transmission to a seronegative partner has been reported by SSITT investigators (Bernasconi 2001), and this was associated with a viral load increase to 130,000 copies during the first 2 week STI. However, it remains a matter of debate whether this risk is best addressed by ceasing STI studies or ensuring that appropriate information and counseling is provided to individuals considering participation in an STI trial.
At the International Workshop in Montreal in March 2002, Catherine Fagard reported the resistance data from SSITT. Ten participants showed evidence of resistance mutations in reverse transcriptase, and in nine of these cases it was either 184I or 184V (which reduce susceptibility to 3TC). One study participant had a mutation associated with resistance to the protease inhibitor nelfinavir, and this was the only case that required a change in regimen in order to re-suppress viral load to undetectable levels. At the same workshop, Mark Dybul noted that the risk of resistance appears significantly greater for the NNRTI drugs with long half-lives, particularly efavirenz (Sustiva), and notably the use of NNRTIs was not permitted in SSITT. Dybul also found that interrupting HAART before viral load was controlled to less than 50 copies was significantly associated with the development of resistance over the course of subsequent STIs. While these data emphasize the need for care when designing STI protocols, not all researchers agree that the specter of resistance should preclude further trials, particularly given the risk of resistance from continuous therapy is well-described and the relative risk of STIs compared to continuous HAART is as yet unknown.
One extremely valuable perspective on the risks of STIs -- and one that speaks to the potential for conducting STI studies in the future -- comes from the SSITT participants themselves. In a survey reported at the 2002 International AIDS Conference in Barcelona (Le Braz, 2002), study investigators quizzed enrollees about their experience and found that the majority "viewed their experience positively." Only 2% stated they wouldn't participate in a similar STI trial, compared to 72% that said they definitely would and 25% who answered that they'd consider it. In terms of any problems taking HAART after interruptions, 73% reported that that it was no different, 10% said it felt easier while 17% found it more difficult. Furthermore, fewer side effects were reported upon restarting HAART than when therapy was first initiated.
Clearly, the failure of the SSITT protocol to produce any obvious and dramatic immunological or virological benefit threatens to cast a pall over future STI research, at least in terms of autovaccination. Efforts to secure funding and Institutional Review Board (IRB) approval for such studies may well be affected, undermining investigators -- such as Luis Montaner -- attempting to pursue answers to the outstanding questions described in this article. The political aspect of STI research may also be playing a role in the interpretation of SSITT. As Cal Cohen from the Community Research Initiative of New England has observed, there is no incentive for pharmaceutical companies (or, for that matter, the many researchers they fund) to support strategies that aim to allow people with HIV to safely go without drug treatment for extended periods. To ensure that future STI research is guided by the science, not prejudice or politics, the treatment activist community will need to closely monitor developments in the field.