Antiviral Therapy 12:189 193 Effect of transient antiretroviral treatment during acute HIV infection: comparison of the Quest trial results with CASCADE natural history study Fiona C Lampe 1 *, Kholoud Porter 2, John Kaldor 3, Matthew Law 3, Sabine Kinloch-de Loes 1 and Andrew N Phillips 1 on behalf of the CASCADE Collaboration 1 Royal Free Centre for HIV Medicine, Royal Free and University College Medical School, London, UK 2 MRC Clinical Trials Unit, London, UK 3 National Centre in HIV Epidemiology & Clinical Research, University of New South Wales, Sydney, Australia *Corresponding author: Tel: +44 20 7830 2239; Fax: +44 20 7794 1224; E-mail: f.lampe@pcps.ucl.ac.uk Objective: The benefit of transient combination antiretroviral treatment (CART) during acute HIV infection is uncertain. We used the seroconverter database CASCADE to provide a historical comparison for the Quest trial, in which 79 subjects with acute HIV infection received CART for an average of 2.6 years, and 17.7% (95% confidence interval [CI]: 10.9 27.6) fulfilled the primary endpoint of VL 1,000 copies/ml 24 weeks after CART discontinuation. Methods: We estimated the prevalence of VL 1,000 copies/ml three years after seroconversion and prior to any ART among 385 sexually infected subjects in CASCADE who seroconverted between 1988 and 1996. We conducted a pre-specified comparison with the recently published Quest results, and considered potential biases. Results and discussion: The prevalence of VL 1,000 copies/ml at year three in CASCADE was 10.1% (95% CI: 7.5 13.5) (absolute difference compared to 17.7% in Quest: 7.6%; 95% CI: -0.1 17.8; P=0.053). In CASCADE, VL 1,000 copies/ml was less common among homosexual and heterosexual men compared with women (8.5%, 7.3% and 17.6% respectively) and in subjects with symptomatic infection compared with those without (6.2% and 12.6%, respectively). As Quest had a much greater proportion of symptomatic subjects than CASCADE, any true difference in VL might be underestimated. Therefore this comparison suggests that transient CART in acute infection might result in a modest increase in the probability of low VL subsequently. However, several factors mitigate this conclusion. First, this historical comparison might be subject to other unmeasured confounders. Second, a comparison of median VL at the same time point was not significant (4.02 copies/ml and 4.20 copies/ml in Quest and CASCADE, respectively; P=0.55). Finally, additional analysis suggested that the observed difference in low VL at year three would be consistent with an immediate effect of CART only a delay of usual VL decline without additional benefit. Conclusions: A small but significant proportion of seroconverters have low VL without ART. Transient CART in acute infection might increase the probability of low VL after treatment discontinuation, but such an effect is likely to be modest, and might represent a delay of natural history rather than a long-term therapeutic benefit. Introduction Intervention with combination antiretroviral treatment (CART) during acute HIV infection might provide the best hope of altering the course of disease, via preservation of HIV-specific immune function, which might aid viral control when treatment is stopped [1,2]. Randomized trials to evaluate any such benefit have been difficult to undertake because there is a theoretical basis for early treatment. Therefore, current information regarding the impact of early CART on subsequent off-treatment viral load (VL) comes predominately from case series and trials in which there are no untreated control groups [2]. The largest of these studies is the recently published Quest trial, in which 79 subjects started CART during acute infection and continued treatment for an average of 2.6 years, and 14/79 (17.7%) fulfilled the primary endpoint of plasma VL 1,000 copies/ml 24 weeks after treatment discontinuation and approximately 3 years after CART initiation [3]. We used data from the seroconverter study collaboration CASCADE [4] in order to 2007 International Medical Press 1359-6535 189
FC Lampe et al. provide a comparison group for the Quest results. We estimated the prevalence of plasma VL 1000 copies/ml 3 years after seroconversion among untreated HIV-infected individuals, and assessed the directions of likely biases in comparison of this prevalence with the results from Quest. Although clearly a suboptimal approach compared with a concurrent randomized control group, a historical control is currently the only means available of evaluating the impact of CART in acute infection, and is necessary in order to interpret the results from Quest and other primary HIV infection (PHI) studies. Materials and methods Summary of Quest study methods and results The methods and results of Quest have been reported by the Quest Study Group [3,5]. Overall, 158 subjects, recruited from ten countries during 1998 2002, initiated triple or quadruple CART during confirmed acute HIV infection ( 3 bands on western blot) at a median of 18 days (range: -8 111) from the start of symptoms. Seventy-nine subjects (91.1% male; 72.2% homosexual; median age: 37 years) who remained under follow-up and on treatment for 72 weeks entered a randomized phase of the study, in which they received either therapeutic vaccine or vaccine placebo in addition to CART, and discontinued CART at a median of 2.6 years (range: 1.9 5.8) after treatment initiation. Overall 14/79 (17.7%) subjects fulfilled the primary endpoint of VL 1,000 copies/ml 24 weeks after CART discontinuation (subjects with missing VL [n=6], those who did not discontinue antiretroviral therapy (ART) [n=1] and those who restarted ART [n=7] were included and counted as failures). The median VL at this time was 4.02 log copies/ml (using a last observation carried forward approach for subjects with missing VL and those who restarted treatment after at least 4 weeks off ART) [3]. The proportion of successes did not differ significantly between the randomized arms (vaccine vs no vaccine). The primary endpoint was assessed at 24 weeks after ART discontinuation as this was considered sufficient time for a stable viral load set-point to be established. There was no untreated control group for assessing whether early CART in itself had been beneficial, and therefore a comparison of the proportion of subjects fulfilling the primary endpoint with the corresponding prevalence from natural history cohorts was planned from the inception of Quest. Analysis of CASCADE The CASCADE database includes data on 9,133 seroconverters from 22 cohorts in Europe, Australia and Canada, pooled in July 2003 [4]. We selected sexually infected subjects (n=5745) who seroconverted between January 1988 and December 1996 at age 18 years (n=1374 of 5745) and had 1 VL measurement from year one to year four after seroconversion (n=649), which was prior to starting ART (n=385). Date of seroconversion was defined as the midpoint of an interval of no more than 6 months between the last negative and first positive HIV test, the date of laboratory evidence of seroconversion, or the date of documented seroconversion illness. The time window after seroconversion was chosen to correspond approximately with the measurement of the Quest primary endpoint, and also because VL appears relatively stable during this period [6 11]. Year of seroconversion was restricted because those remaining untreated during the CART era might represent a selected group. We ascertained the prevalence of plasma VL 1,000 copies/ml among the 385 selected subjects using the pre-art VL nearest to year three after seroconversion. We also conducted sensitivity analyses A, including only subjects with a pre-art VL from year two to year four, and B, including all subjects with a pre-art VL from seroconversion to year four (as some subjects might have started ART soon after seroconversion because of high VL or low CD4 + T-cell count, and their exclusion could bias the results). In both cases the pre-art VL measurement nearest to year three was used. In order to investigate the effect of some potential biases in comparison of CASCADE with Quest, we examined the proportion with VL 1,000 copies/ml in CASCADE according to calendar time of VL measurement, frequency of VL measurement, presence of seroconversion symptoms, precision of estimated seroconversion date (using the negative to positive test interval), and demographic factors (risk group and age). All VL assays used had a lower limit of 1,000 copies/ml. Comparisons of proportions were performed using the χ 2 test and confidence intervals for single proportions and differences in proportions were calculated using the Wilson method in Confidence Interval Analysis (available from http://www.bmjbookshop.com). Logistic regression was used to assess the independent effect of various factors on the probability of low VL. We also compared the median VL at year three between the two studies, using the Mann Whitney test. Results Of the 385 subjects selected from CASCADE, 270 (70.1%), 41 (10.7%) and 74 (19.2%) were homosexual men, heterosexual men and heterosexual women, respectively. Median (range) age at seroconversion was 30.7 (18.0 63.5) years, and 146 (37.9%) 190 2007 International Medical Press
Effect of transient CART in acute HIV infection subjects had a documented seroconversion illness. Median (range) date of seroconversion was October 1993 (January 1988 December 1996), and median (range) date of year three VL was September 1996 (March 1989 October 1999), measured at a median (range) of 2.9 (1.0 4.0) years after seroconversion. Median (interquartile range) plasma VL and CD4 + T- cell count at year three were 4.20 (3.58 4.70) log copies/ml and 440 (312 577) cells/ml, respectively. The distribution of year-three VL was as follows: 1,000 copies/ml, 10.1% (n=39); 1,001 5,000 copies/ml, 17.9% (n=69); 5,001 10,000 copies/ml, 12.2% (n=47); 10,000-100,000, 46.5% (n=179); 100,000 copies/ml, 13.3% (n=51). A comparison of the prevalence of VL 1,000 copies/ml at year three in CASCADE (10.1%; 95% CI: 7.5 13.5) with the 17.7% (95% CI: 10.9 27.6) of successes for the primary endpoint in QUEST (VL 1,000 copies/ml 24 weeks after stopping ART; missing and/or restart ART=failure) gives an absolute difference of 7.6% (95% CI: -0.1 17.8; P=0.053, χ 2 test). However, the median VL at 24 weeks after ART in Quest was not significantly different from the median VL in CASCADE at year three (4.02 and 4.20 copies/ml respectively; P=0.51, Mann Whitney test). Table 1 shows the prevalence of year-three VL 1,000 copies/ml in CASCADE according to demographic and other factors in unadjusted and adjusted analyses. Low VL was less common among men than women, and less common among people with a seroconversion illness than those without symptoms. Low VL also tended to be less common among subjects with a short test interval, but there was no evidence of an effect due to age at seroconversion, date of VL measure, or frequency of VL measurement. Sensitivity analyses (see methods) yielded similar results for the prevalence of VL 1,000 copies/ml: 11.5% (95% CI: 8.4 15.6) for sensitivity analysis A (including 304 subjects with a pre- ART VL from year two to year four) and 9.5% (95% CI: 7.1 12.5) for sensitivity analysis B (including 454 subjects with a pre-art VL from seroconversion to year four), giving P=0.14, and P=0.029 respectively for comparison with Quest. The associations of low VL with demographic and other characteristics were also similar in these sensitivity analyses. Discussion Several studies have examined VL natural history from seroconversion [6 12], but we wished to tailor our analysis to a specific comparison with the Quest primary endpoint. Among CASCADE subjects who fulfilled our inclusion criteria, the prevalence of VL 1,000 copies/ml 3 years after seroconversion was approximately 10%, with evidence of variation according to some demographic and seroconversion Table 1. Association of demographic and other factors with VL 1,000 copies/ml at year three post seroconversion % with Unadjusted OR Adjusted OR* VL 1,000 copies/ml (95% CI) of (95% CI) of n (%) (95% CI) at year three VL 1,000 copies/ml VL 1,000 copies/ml Total 385 (100) 10.1 (7.5 13.5) Demographic group P=0.083 P=0.067 Heterosexual woman 74 (19.2) 17.6 (10.6 27.8) 1 1 Heterosexual man 41 (10.7) 7.3 (2.5 19.4) 0.37 (0.10 1.39) 0.36 (0.09 1.37) Homosexual man 270 (70.1) 8.5 (5.7 12.5) 0.44 (0.21 0.91) 0.41 (0.19 0.87) Age at seroconversion P=0.74 P=0.93 <30 years 178 (46.2) 10.7 (6.9 16.1) 1 1 30 years 207 (53.8) 9.7 (6.3 14.5) 0.90 (0.46 1.74) 0.97 (0.49 1.92) Seroconversion illness reported P=0.037 P=0.035 No 239 (62.1) 12.6 (8.9 17.4) 1 1 Yes 146 (37.9) 6.2 (3.3 11.3) 0.46 (0.21 0.99) 0.45 (0.20 0.98) Last negative to first positive test P=0.15 P=0.16 3 months 247 (64.2) 11.7 (8.3 16.4) 1 1 <3 months 138 (35.8) 7.3 (4.0 12.8) 0.59 (0.28 1.24) 0.59 (0.27 1.27) Date of year-three VL P=0.43 P=0.34 <1997 224 (58.2) 11.2 (7.7 16.0) 1 1 1997 161 (41.8) 8.7 (5.3 14.1) 0.76 (0.38 1.51) 0.71 (0.35 1.45) Frequency of VL measurement P=0.86 P=0.86 <3 measurements years one to four 143 (37.1) 10.5 (6.5 16.6) 1 1 3 measurements years one to four 242 (62.9) 9.9 (6.8 14.3) 0.94 (0.48 1.86) 0.94 (0.47 1.89) *Adjusted model includes all factors in table. CI, confidence interval; OR, odds ratios from logistic regression analysis; VL, viral load. Antiviral Therapy 12:2 191
FC Lampe et al. characteristics. Demographic factors have previously been shown to be associated with VL level after seroconversion and subsequent VL trajectories [7 11]. In this and other studies [7 11], low VL was less common among men compared with women. We did not find an association of age or risk group (independent of gender) with low VL at year three, although other analyses have suggested that VL levels tend to be higher among older subjects and among homosexual compared with heterosexual men [9,10]. We also hypothesized that low VL would be less common among those with symptomatic infection [12,13] (assessed by symptoms or the proxy measure of a short test interval) and that low VL would be more common during the early calendar period due to possible underestimation of VL when ascertained retrospectively from frozen samples. In these data, subjects with symptomatic infection were less likely to have low VL at year three, but there was no effect due to calendar period. Subjects included in this analysis were required to have a VL measurement within a specified period. This might have resulted in selection bias if VL was more likely to be measured (or retrospectively ascertained) among those with poor prognosis. However, we found no association between frequency of VL measurement and low VL at year three. Bias might also occur if subjects with poor prognosis were more likely to be excluded because they had started ART early. However, a sensitivity analysis extending the pre-art VL measurement window yielded similar results. This comparison between Quest and CASCADE suggests a small difference between the studies, of borderline statistical significance, in terms of the probability of low VL at year three. However it is a nonrandomized, historical comparison that might be biased by known and unknown confounders. We have attempted to ascertain the directions of bias for several measurable factors. Quest had a similar proportion of men to CASCADE (91% vs 88%); but subjects were older (median age: 35 vs 31 years) and were much more likely to have symptomatic infection (90% vs 38%) [5]. Male gender and acute symptoms both reduce the likelihood of low VL after seroconversion. If there is an effect due to older age, it is likely to be in this same direction. Therefore the overall proportion of subjects with VL 1,000 copies/ml in this CASCADE analysis might be greater than the proportion expected among Quest patients. Consequently, a comparison between the 17.7% of successes in Quest and the 10.1% prevalence of VL 1,000 copies/ml in CASCADE might be conservative, tending to underestimate any true difference. Therefore, if 24 weeks allowed sufficient time for a stable off-art VL to be reached in Quest, these results would provide some indication that transient CART started during acute infection might increase the probability of a low off-treatment VL 3 years later. However, when considering the median VL at year three, rather than the pre-specified endpoint, little difference was found between the two studies. Taken together, these results suggest that any beneficial effect of early CART is likely to be modest. Most previous reports of transient CART strategies in PHI have been based on very small series of patients [2]. There have been claims for an association between early CART and subsequent viraemia control, but these have been difficult to evaluate, as a proportion of HIVinfected subjects have naturally low VL. PHI cohorts have varied in terms of how rapidly after infection CART was started, and the type and length of treatment, and comparisons with historical controls have not given consistent results [1,11,14 16]. Two early studies (one using a treatment interruption strategy) based on eight and nine PHI subjects, suggested that early treatment was associated with suppression of VL following treatment cessation [14,15]. However, three subsequent studies based on 16, 37 and 58 PHI patients suggested that early treatment conferred no significant benefit when compared with historical controls [1,11,16]. In the largest of these, the PRIMO study, 58 patients who started CART during acute or early infection and stopped treatment after sustained virological suppression were compared with untreated historical controls from SEROCO [11]. There was no significant difference in predicted VLs at year three (3.95 and 4.11 log copies/ml for treated and untreated groups, respectively; difference: -0.16 [95% CI: -0.58 0.25] adjusted for age and sex). Recent results from the PRIMSTOP trial of structured treatment interruptions in acutely infected patients showed 6 of 26 participants (23%) with viral load <1,000 copies/ml 24 weeks after the end of treatment [17], which appears consistent with Quest. In the present analysis, as in the SEROCO analysis, the treated and untreated cohorts were compared at the same time point after seroconversion. However, any reduction in viral load in the treated compared with the untreated cohort could be due to the time on CART delaying the increase in VL that would have occurred naturally over the period of 3 years. A much more desirable effect would be that the 2.6 years of treatment from acute infection actually slowed the subsequent natural history, resulting in a lower VL after 6 months off CART than would be expected after an equivalent 6 months of natural history. In order to address this question we conducted an additional analysis in which we performed a simple adjustment to the CASCADE VL measurements in order to account for the fact that the Quest endpoint was measured after only 0.5 years off CART. Using an estimate of natural history VL increase of 0.089 log copies/ml per year [10], we adjusted CASCADE log VLs by subtracting (0.089 [years from seroconversion to VL 192 2007 International Medical Press
Effect of transient CART in acute HIV infection measurement-0.5]). This adjustment resulted in a prevalence of VL 1,000 copies/ml of 14.3% (95% CI: 11.1 18.1) and a non-significant difference compared to Quest (difference 3.4%; 95% CI: -4.4 13.8; P=0.43). In conclusion, a small but significant proportion of HIV-infected individuals have low VL in the absence of ART, and this must be taken into account when interpreting uncontrolled studies of treatment in PHI. This analysis suggests that the proportion of subjects with low VL approximately 3 years after infection in Quest might be higher than would be expected at the same time point after infection among untreated subjects, but the difference appears modest, and is less apparent when comparing the median VL between the two studies. It is also important to appreciate that a difference of this magnitude might represent a delay of natural history rather than a genuine therapeutic benefit of treatment in acute infection. At present there are no results from randomized controlled trials regarding the effect of CART in PHI. These data suggest that such trials are justified and their results necessary in order to inform clinical practice regarding treatment of acute HIV infection. Acknowledgements This analysis was planned following a Quest Core Group discussion. Quest Core Group: Sabine Kinlochde Loes, David Cooper, Luc Perrin, Bruno Hoen, Anders Sonnerborg, Christos Tsoukas, Jan Andersson, Brigitte Autran, Fiona Lampe and Andrew Phillips. Funding and conflict of interest CASCADE is funded through a grant from the European Union [QLK2-2000-01431] and has received additional funding from GlaxoSmithKline. There are no other conflicts of interest. Additional files The additional file CASCADE Steering Committee, Coordinating Centre and Collaborators can be accessed via the Volume 12 Issue 2 contents page for Antiviral Therapy, which can be found at www.intmedpress.com (by clicking on Antiviral Therapy then Journal PDFs ). References 1. Fidler S, Oxenius A, Brady M, et al. Virological and immunological effects of short-course antiretroviral therapy in primary HIV infection. AIDS 2002; 16:2049 2054. Accepted for publication 13 September 2006 2. Smith DE, Walker BD, Cooper DA, Rosenberg ES, Kaldor JM. Is antiretroviral treatment of primary infection clinically justified on the basis of current evidence? AIDS 2004; 18:709 718. 3 Kinloch-de Loes S, Hoen B, Smith DE, et al. Impact of therapeutic immunisation on HIV-1 viremia after discontinuation of antiretroviral therapy initiated during acute infection. J Infect Dis 2005; 192:607 617. 4 CASCADE Collaboration. Changes in the uptake of antiretroviral therapy and survival in people with known duration of HIV infection in Europe: results from CASCADE. HIV Med 2000; 1:224 231. 5 Kinloch S, Cooper D, Lampe F, et al. The Quest cohort: treatment of primary HIV infection with quadruple CART: week 48 preliminary results [abstract 520]. In Program and abstracts of the 10th Conference on Retroviruses and Opportunistic Infections (Boston, MA, USA). Alexandria, VA, USA: Foundation for Retrovirology and Human Health, 2003:240. 6 Hubert J, Burgard M, Dussaix E, et al. Natural history of serum HIV-1 RNA levels in 330 patients with a known date of infection. AIDS 2000; 14:123 131. 7 Lyles CM, Dorrucci M, Vlahov S, et al. Longitudinal human immunodeficiency virus type 1 load in the Italian Seroconversion Study: correlates and temporal trends of virus load. J Infect Dis 1999; 180:1018 1024. 8 Sterling TR, Lyles CM, Vlahov D, Astemborski J, Margolick JB, Quinn TC. Sex differences in longitudinal Human Immunodeficiency Virus type 1 RNA levels among seroconverters. J Infect Dis 1999; 180:666 672. 9 Touloumi G, Pantazis N, Babiker AG, et al. Differences in HIV-RNA levels before the initiation of antiretroviral therapy among 1864 individuals with known HIV-1 seroconversion dates. AIDS 2004; 18:1697 1705. 10 Pantazis N, Touloumi G, Walker AS, Babiker AG. Bivariate modelling of longitudinal measurements of two human immunodeficiency type 1 disease progression markers in the presence of informative drop-outs. Appl Statist 2005; 54:405 423. 11 Desquilbet L, Goujard C, Rouzioux C, et al. Does transient CART during primary infection lower the virological set-point? AIDS 2004; 18:2361 2369. 12 Madec Y, Boufassa F, Porter K, Meyer L, CASCADE Collaboration. Spontaneous control of viral load and CD4 cell count progression among HIV-1 seroconverters. AIDS 2005; 19:2001 2007. 13 Vanhems P, Lambert J, Cooper DA, et al. Severity and prognosis of acute Human Immunodeficiency Virus type 1 illness: a dose-response relationship. Clin Infect Dis 1998; 26:323 329. 14 Rosenberg ES, Altfeld M, Poon SH, et al. Immune control of HIV-1 after early treatment of acute infection. Nature 2000; 407:523 526. 15 Girard PM, Schneider V, Dehee A, et al. Treatment interruption after one year of triple nucleoside analogue therapy for primary HIV infection. AIDS 2001; 15:275 277. 16 Markowitz M, Jin X, Hurley A, et al. Discontinuation of antiretroviral therapy commenced early during the course of human immunodeficiency virus type 1 infection, with or without adjunctive vaccination. J Infect Dis 2002; 186:634 643. 17 Hoen B, Fournier I, Lacabaratz C, et al. PRIMSTOP Study Group. Structured treatment interruptions in primary HIV- 1 infection - The ANRS 100 PRIMSTOP Trial. JAIDS 2005; 40:307 316. Antiviral Therapy 12:2 193