Methods

Study designACTG 5068 was a prospective, randomized, partially double-blind study (figure 1). Subjects receiving their first potent antiretroviral therapy (ART) regimen who had plasma HIV-1 RNA levels <400 copies/mL and CD4⁺ T lymphocyte counts >400 cells/mm³ for >6 months and who had HIV-1 RNA levels <50 copies/mL at study entry were randomized to 1 of 4 arms. Subjects randomized to arms A and C received continuous ART for 44 weeks. Subjects randomized to arms B and D underwent 2 cycles of treatment interruption (the first lasting 4–6 weeks, depending on the time to viral load rebound, and the second lasting 4 weeks) followed by 16-week periods of treatment resumption. Immunization with ALVAC-HIV vCP1452 (arms C and D) or placebo (arms A and B) was performed in 3 sets of 3 weekly injections, beginning 4 weeks before treatment interruption for arms B and D and at comparable time points for arms A and C

After the intervention period, all subjects had an analytical treatment interruption (ATI) that lasted 12–20 weeks, with subjects continuing to not receive therapy for at least 8 weeks after the first quantifiable plasma HIV-1 RNA level >50 copies/mL. Subjects whose viral load reached a new steady state, defined as a change in HIV-1 RNA level of <0.2 log₁₀ copies/mL per week over 3 consecutive weeks before week 12 could end the ATI at that point but were encouraged to complete the 12–20-week ATI. Subjects with steady-state plasma viral loads <10,000 copies/mL whose absolute CD4⁺ T lymphocyte counts or percentages remained >50% of baseline levels were encouraged to continue the treatment interruption for 48 weeks

The protocol was approved by local institutional review boards. Written informed consent was obtained from potential subjects. Human experimentation guidelines of the US Department of Health and Human Services and those of the authors’ institutions were followed in the conduct of the clinical research

EvaluationsSubjects were evaluated by history, physical examination, and laboratory safety monitoring at protocol-defined intervals. During treatment interruptions, CD4⁺ T lymphocyte count and viral load were obtained weekly

Plasma HIV-1 RNA levels were measured in real time at a single ACTG laboratory (Johns Hopkins University) by use of the Amplicor HIV-1 Monitor Test (version 1.5; Roche Molecular Systems). Lymphocyte proliferation assays were performed on fresh blood specimens in accordance with the standard ACTG protocol [13]. Stimulants included pokeweed mitogen (0.1 μg/mL; Sigma), Casta antigen (10 μg/mL; Greer), tetanus toxoid (1U/mL; Wyeth-Lederle), IRC HIV antigen (5 μg/mL p24; Immune Response), recombinant p24, and a control baculovirus preparation (5 μg/mL; Protein Sciences). The stimulation index (the geometric mean counts per minute for each stimulant divided by the geometric mean counts per minute for medium alone) was used for analysis of the response. Cryopreserved lymphocytes were obtained for future additional measurements of CD4⁺ and CD8⁺ T cell responses to HIV antigens

Subject safetySubjects in arms B and D (STI arms) did not undergo their second STI if, by the time the second STI was to begin, they had not experienced resuppression of plasma HIV-1 RNA level to <200 copies/mL while receiving therapy. Any subject with a plasma HIV-1 RNA level not resuppressed to <1000 copies/mL or with a CD4⁺ T lymphocyte count and percentage <70% of baseline values at the scheduled start of the ATI continued to receive therapy. HIV drug-resistance monitoring was performed on the last available plasma sample from these subjects that had an HIV-1 RNA level >500 copies/mL. Sequencing of HIV-1 pol (codons 41–237 of the reverse transcriptase) was performed by standard bulk viral population resistance testing by use of the US Food and Drug Administration–approved TruGene HIV-1 Sequencing Kit and OpenGene DNA Sequence Analysis System (GeneObjects, version 3.2; Bayer Nucleic Acid Diagnostics) [14, 15] at a single ACTG laboratory (University of Alabama, Birmingham) (sequences are available from GenBank [accession numbers DQ497400–DQ497408]). Phylogenetic trees of the aligned, manually edited sequences were constructed using PHYLIP and Treeview, to verify that each patient’s viral sequence was unique [16, 17]

Statistical analysisVirological end points included the end-of-ATI viral load (geometric mean of the values at the last 2 weeks of the ATI), steady-state viral load (geometric mean of 3 consecutive weekly viral load values in which the slope of the RNA versus time curve for all 3 values and the slope of the last 2 values were both between −0.2 and 0.2 log₁₀ copies/mL per week), time from the start of ATI to viral load rebound of >50 copies/mL, the rate of initial rise of viral load during the ATI (calculated from the series of consecutive viral load measurements in which the slope for each consecutive pair of values was >0.2 log₁₀ copies/mL per week), the peak viral load in the ATI, and the postpeak low viral load. Immunologic end points included HIV-1–specific and antigen/mitogen lymphoproliferative responses

The primary end point of the study was the end-of-ATI viral load. The primary prespecified analysis was a comparison of the primary end point between the control arm (no vaccine and no STIs) and each of the 3 other arms (STIs only, vaccine only, and vaccine and STIs). The sample size was determined on the basis of providing the study 80% power to detect—with 2-sided, 5% significance level, exploratory&amp;rank-based tests—a shift in the distribution of the primary end point of 0.62 log₁₀ copies/mL in a single arm versus another single arm. A secondary prespecified analysis was to compare the combined 2 STI arms with the combined 2 non-STI arms and also to compare the combined 2 vaccine arms with the combined 2 non-vaccine arms

Analyses were exploratory at the nominal, 2-sided, 5% level of significance, with no adjustment for multiple testing. Time-to-event end point distributions were estimated with the Kaplan-Meier method and were compared using the log-rank test. Other continuous end points were compared using the 2-sided Wilcoxon&amp;rank sum test. Proportions were compared using Fisher’s exact test. Correlations were estimated and tested with Spearman’s&amp;rank correlation. In intent-to-treat analyses, the worst&amp;rank in the&amp;rank-based tests was assigned to subjects without an observed end point; as-treated analyses were based only on subjects with an observed end point

Results

Study populationBetween May 2001 and October 2003, 97 patients were enrolled; 83 (86%) had an end-of-ATI plasma HIV-1 RNA set point determined. Twenty-four patients were randomized to arm A, 24 to arm B, 25 to arm C, and 24 to arm D. The patients in the 4 groups had similar baseline characteristics, except for race/ethnicity (table 1). Fewer than 20% of the subjects enrolled in arm A were nonwhite, whereas 50% of the subjects in arms B, C, and D were nonwhite. The median CD4⁺ T lymphocyte count was 730 cells/mm³. At study entry, 56 subjects were receiving a nonnucleoside reverse-transcriptase inhibitor (NNRTI), and 45 subjects were receiving a protease inhibitor. Eighty-five subjects were receiving lamivudine

Viral kinetics during the ATIAfter treatment interruption, a rise in plasma HIV-1 RNA level occurred, usually within 2–4 weeks (figure 2). The initial rate of rise in viremia fit a first-order exponential growth curve. Viremia peaked several weeks after initial detection, then fell to a lower level, and tended to stabilize at an equilibrium (set point) viral load

We analyzed the combination of the 2 STI arms (B and D) versus the 2 arms that did not have STI (A and C), in a factorial comparison, and found a consistent pattern of statistically significant differences in the kinetics of HIV-1 RNA rebound during the ATI in the combined STI arms compared with the arms that did not have STI (table 2). In analysis of subjects who had the observed evaluations at the end of the ATI period (n=83), those in the STI arms showed a longer doubling time in viral rebound, a lower peak level of viremia, a lower postpeak low, a lower steady-state level of viremia, and a lower median end-of-ATI level of viremia than did those in the non-STI arms. Although there were multiple differences between these groups in the kinetics of viral rebound once initiated, there was no difference in the time to initial detection of viremia (>50 copies/mL) during the ATI. Finally, and perhaps most importantly, significantly more patients who underwent cycles of STI achieved a stable viral load <1000 copies/mL than did subjects who did not undergo this intervention. As a measure of comparability of the arms, there was no difference between the viral loads at week 4 during the first STI in the STI groups and the viral loads at week 4 during the ATI in the non-STI groups (P=.34). By use of the immunization schedule described above, ALVAC-HIV vCP1452 had no apparent effects on viral load measures during the ATI

Of the 14 subjects who did not have observed end-of-ATI plasma HIV-1 RNA levels, 3 were in arm A, 3 were in arm B, 3 were in arm C, and 5 were in arm D. Four subjects withdrew from the study prematurely because they could not adhere to the study visit schedule before the ATI, 4 moved or were lost to follow-up before the ATI, 1 underwent interferon-α treatment for hepatitis C before the ATI, 1 restarted ART prematurely during the first STI, 1 decided against undergoing the ATI, and 2 (1 each in arms B [STI] and C [non-STI]) did not undergo the ATI because their viral load was >1000 copies/mL at the time they were scheduled to start the ATI. Only 1 subject withdrew from the study during the ATI “read-out” period

Intent-to-treat analyses, which assigned the worst&amp;rank to subjects who did not have the required HIV-1 RNA evaluation, gave results similar to those of the analyses of observed data only. The number of subjects who had an end-of-ATI plasma HIV-1 RNA level <1000 copies/mL was greater in the STI arms than in the non-STI arms (10/48 [21%] vs. 3/49 [6%]; P=.04). There was a nonsignificant trend toward a lower median end-of-ATI viral load in the STI arms than in the non-STI arms (HIV-1 RNA level, 4.44 vs. 4.55 log₁₀ copies/mL; P=.12). Although the end-of-ATI viral loads in the control arm were not statistically significantly different from those in any other arm, possibly because of the relatively small sample size, there was a trend toward a lower median value in arm B (STI only) than in arm A (no STI and no vaccine) (HIV-1 RNA level, 4.38 vs. 4.70 log₁₀ copies/mL; P=.14)

There were significant correlations among many of the viral load end points. The steady-state viral load was significantly and positively correlated with the end-of-ATI viral load (P<.0001) (figure 3), the peak viral load (P<.0001), and the rate of the initial rise (P=.025). The time to a viral load >50 HIV-1 RNA copies/mL did not correlate with any other virological measure. The 12-week duration of the ATI period was further validated by comparing the end-of-ATI viral load of the evaluable study population with the steady-state viral load in the 72 subjects who had both end points observed. The median of the absolute difference (end-of-ATI viral load minus steady-state viral load) was only 0.21 log₁₀ HIV-1 RNA copies/mL. Thus, 12 weeks of treatment interruption was probably enough for most subjects to reach new viral load equilibriums

Those subjects who achieved low end-of-ATI viral loads maintained low viral loads while continuing to not receive ART for extended periods of time after the planned ATI (figure 4). Of the 13 subjects with an end-of-ATI plasma HIV-1 RNA level <1000 copies/mL, 7 maintained this level at 48 weeks of not receiving ART; 2 others maintained this level at their last evaluations at weeks 28 and 42. Four subjects had HIV-1 RNA levels between 1000 and 2000 copies/mL at the week 48 evaluation

Immunologic dataConsistent with the viral kinetics and lower viral load set point we observed during the ATI, there was a smaller decrease in CD4⁺ T lymphocyte count during this period in the STI arms than in the non-STI arms (−180 vs. −280 cells/mm³; P=.017). Overall, there was no difference in the median change in CD4⁺ T lymphocyte counts from study entry to the end of the ATI between the STI arms and the non-STI arms (−194 vs. −223 cells/mm³; P=.41). There was no difference between the STI arms and the non-STI arms in the distributions of CD4⁺ T lymphocyte counts immediately before the ATI (P=.83)

An in vitro analysis of the lymphoproliferative response to HIV antigens was performed at multiple points throughout the study (figure 5A). The results show modest but not statistically significant increases in the median response to 2 different HIV protein antigens in the arms given intervention (arms B, C, and D) but not in the placebo arm (arm A). There were no statistically significant differences across the arms in the median changes in responses in individual subjects to the 2 different HIV protein antigens. There was a significant negative correlation between the lymphocyte proliferation assay (LPA) responses to the 2 HIV antigens just before the ATI (after immunization interventions) and the end-of-ATI viral load, although the magnitude of the correlation between higher LPA response and lower viral load was modest (figure 5B)

Safety and HIV drug-resistance dataThere were no deaths, opportunistic infections, or other AIDS-defining events. An acute retroviral syndrome occurred in 1 subject (arm C) during the ATI

Four subjects never achieved a viral load persistently <50 copies/mL after resuming ART following an STI. Three of the 4 subjects had successful amplification and sequencing of virus from plasma specimens obtained in the post-STI period. All 3 were receiving NNRTI-containing regimens before their STIs and had mutations consistent with NNRTI resistance. Two of the 3 patients were receiving lamivudine and were found to have the M184V mutation. Thus, 3 of the 48 subjects undergoing STI, 26 of whom were receiving NNRTIs, displayed HIV resistance against drugs contained in their pre-STI regimens in bulk plasma viral populations

One non-STI subject (arm C) did not have a viral load of <1000 copies/mL when the ATI was scheduled to begin. There was no evidence of HIV drug resistance in the plasma of this subject

Discussion

In this partially blind, randomized, controlled study, we found evidence that delivering pulses of exposure to autologous HIV-1 RNA antigens as a result of cycles of brief structured ATIs could enhance host control of viral replication during a subsequent prolonged ATI, whereas immunization with ALVAC-HIV vCP1452 could not. More patients randomized to receive STIs ended the ATI with plasma HIV-1 RNA levels <1000 copies/mL than did those randomized to receive continuous therapy, whether by intent-to-treat analysis or by analysis of observed data only. Other aspects of the viral kinetic curve during the ATI were favorably altered by this treatment strategy, including the initial rate of rise in plasma HIV-RNA level, the peak level achieved, and the new steady-state equilibrium level at which the HIV-1 RNA level settled. In addition, there was a relative preservation of CD4⁺ T lymphocytes during the ATI in those who underwent STIs

Although a limitation of the study is that we do not have reliable access to the pre-ART HIV-1 RNA set point values for comparison, the randomized design would make a substantial imbalance in pretherapy viral load across the treatment arms unlikely. In addition, 10 of the 13 subjects with end-of-ATI viral loads <1000 copies/mL achieved viral loads >9000 copies/mL at some point during their treatment interruptions, indicating that these were unlikely to be subjects with chronic low-level viremia when not receiving ART before study entry. Importantly, the low end-of-ATI level of viremia in these subjects was stable for the 48 weeks of follow-up. The durability of the effect suggests that a small fraction of patients could achieve sustained viral suppression with this treatment strategy, potentially sparing them the need to receive antiretroviral chemotherapy for a considerable period of time. The 15%–18% response rate, compared with the control group, is modest, but, hopefully, this response could be improved in future studies with refinements of the STI approach and additional immune manipulations

The previous negative reports about STIs were based on uncontrolled observations [18, 19] or on study designs that stopped the ATIs before the new viral load equilibrium had been established, either by design or in accordance with safety stopping rules [18–20]. Our study was successful in allowing all but 1 of the subjects who started ATI to complete it, and it demonstrated that premature discontinuation of the ATI would interfere with the ability to interpret the strength of the immune response induced. A common end point in previous studies was the time to the viral load rising above a certain level [18, 20]. Our data show that there is no correlation between the time to initial detection of viremia (HIV-1 RNA level >50 copies/mL) and the ultimate stable viral load steady state established during the ATI. Because antigen drives an immune response, until enough virus is generated, the time to initial detection of viremia in a laboratory assay may depend more on the kinetics of viral replication and the availability of target cells than on the immune response

The administration of ALVAC-HIV vCP1452, an attenuated canarypox virus loaded with consensus sequences of HIV genes, did not have any effects on measures of viral replication during the ATI. Nor did it demonstrate synergy with exposure to autologous viral antigens, as we had hypothesized. The potency of the vaccine product is critical in testing this hypothesis. Unfortunately, subsequent to the design of this study, the potency of ALVAC-HIV vCP1452 has been disappointing in both prophylactic vaccine studies [9–12] and other therapeutic trials [21–26]. Immunization of HIV-infected subjects with ALVAC-HIV vCP1452 vaccine alone [23, 24] and together with either recombinant gp160 protein [21, 22], Remune vaccine [23], or Lipo-6T vaccine plus interleukin-2 [25] elicited only modest HIV-specific cellular immune responses and, in most studies, demonstrated no clear evidence of control of viremia after ART interruption. It should be noted, however, that, in one study, vaccination was associated with a longer duration of not needing to receive ART [25, 27], and ACTG 5024 did detect a modest level of enhanced viral control after ALVAC-HIV vCP1452 immunization by administration of a single injection every 8 weeks, rather than the 3 cycles of 3 weekly injections administered in this study [26]

We did not find that STIs induced significantly stronger HIV-specific CD4⁺ T LPA responses, and these responses showed only a modest negative correlation with viremia. Other STI studies have not shown a correlation between laboratory measures of HIV-specific cellular immune activity and levels of viremia [28, 29], whereas some studies of therapeutic immunization have shown such a correlation [24, 27]. A negative correlation between LPA response and chronic viral load when not receiving ART has also been described in subjects without therapeutic immunization [30]. It remains unclear to what extent the variability in these results is due to the technical limitations of the assays in the context of multicenter clinical trials, variability among different patient cohorts, or variability of the interventions examined. It is also conceivable that other immunologic functions not assessed by the assays used play a more important role in controlling virus replication and would give more consistent results

With regard to safety, of the 48 subjects in the STI arms of our study, 4 did not experience full resuppression of their virus during the retreatment periods. Mutations associated with resistance to lamivudine and/or NNRTIs were seen in 3 of these 4 subjects. Therefore, STI strategies include a modest risk of resistance, and caution should be exercised. Other trials have shown that the most common drugs to which virus will develop resistance after treatment interruption are lamivudine and NNRTIs, drugs with long half-lives [31–33]. Although detectable viremia does not occur until a median of 2–3 weeks after treatment interruption, the use of agents that have shorter half-lives and higher barriers to resistance than those of lamivudine and NNRTIs may limit the resistance risk in STI studies

In conclusion, this study showed evidence that STIs as a therapeutic immunization strategy could enhance host control of viral replication. By use of this immunization schedule, ALVAC-HIV vCP1452 did not affect HIV-1 RNA measures during an ATI

AIDS Clinical Trials Group 5068 Members

Other investigators at participating sites were P. Tebas and K. Maffei, University of Pennsylvania, Philadelphia; H. Sousa and K. Tashima, Brown University, Providence, Rhode Island; C. Gonzalez and M. Vasquez, New York University, New York; G. McComsey and B. Gripshover, Case Western Reserve University, Cleveland, Ohio; R. Johnson and D. O’Connor, Indiana University, Indianapolis; C. Hurley and R. Corales, University of Rochester, Rochester, New York; A. Nesbit and J. Kuruc, University of North Carolina, Chapel Hill; M. Downing and J. Lawrence, University of California, San Francisco; K. Upton and D. Green, University of Alabama, Birmingham; D. Garmon, Beth Israel Medical Center, New York, New York; S. Koletar and K. Watson, Ohio State University, Columbus; R. Mitsuyasu and S. Chafey, University of California, Los Angeles; D. Ortega, Mount Sinai School of Medicine, New York, New York; B. Royer and N. Conley, University of Washington, Seattle; and M. Rodriguez and D. Demarco, Washington University, St. Louis, Missouri