Bridging the Gap: Conference Record [Abstract book, International Conference on AIDS (12th: 1998: Geneva, Switzerland)]

12th World AIDS Conference

Abstracts 42200-42204 813

S42200 Detection in untreated patients of HIV-1 strains with a
ZDV-resistance mutation at codon 215, in different
proportions, and consequence on therapeutic response
Mary-Anne Trabaud. Laboratoire d'Hygiene, Faculte Medecine, 8 Avenue
Rockefeller, 69008 Lyon, France
Objective: To determine whether the search for the presence of strains with drugresistance mutation before treatment could be useful for therapeutic decision.
Methods: Two methods of codon 215 mutation detection were compared: the
selective PCR (Lardier et al.), and the differential hybridization (Eastman et al.,
Chiran). The later method evaluating the proportion of mutant strains has been
used after amplification of proviral as well as plasma viral HIV RNA.
Results: Mutation at codon 215 was found in 9 over 63 patients, with a proportion of mutant strains of 93 to 700% in 6 and of 15 to 30% in 3 selective PCR
detected mutant strains when they represent the nearly total population, but not
when there is a mixture of predominantly wild type species. Comparison of the
results from proviral DNA and viral RNA showed good concordance qualitatively
as well as quantitatively. Viral load and proportion of mutant strains are followed
during subsequent therapy, and compared in patients of the 3 groups (primary
strains consisting of wild type, mixture, or mutant strains)
35*/42201 Placebo control trials of short-course antiretroviral
regimens to reduce mother-to-child HIV
transmission are essential to establish standard of
care in Africa
JamesG. Kahn', E. Marcelle2. 'Inst. for Health for Stud. UCSF Box 0936 San
Francisco, CA; 2lnst. for Health Policy Studies USCF San Francisco, CA, USA
Objectives: Trials in Africa of feasible short-course antiretroviral (ARV) drugs to
prevent mother-to-child HIV transmission have been criticized for using placebocontrol designs instead of equivalence comparisons to AZT/ACTG 076. We
compared the value of these 2 designs for determining local standard of care.
Design/Methods: We review how statistical, design, and implementation issues affect trial validity, generalizability, and rapidity in African settings. We
estimate potential HIV infections prevented with faster trials.
Results: Validity: An equivalence design requires a comparison with known
efficacy. However, the efficacy of 076 cannot be confidently extrapolated to Africa
due to differences in risk for HIV transmission that might affect efficacy, including
immune status, HIV disease state, and breastfeeding practices. Natural history
studies cannot serve as historical untreated controls due to the wide variability
of observed transmission rates (25%-48%). If 076 were found to be better than
short-course ARV, it would be unclear if the alternative were better than nothing.
A placebo trial provides direct information on efficacy. Generalizability: An equivalence design has limited generalizability because the long 076 regimen requires
women to enroll long before they usually seek prenatal care. These women may
have atypical health practices, so that results may not apply well to the general
population of pregnant women. Placebo trials enroll at the time when women usually seek prenatal care. Rapidity of completion: An 076 equivalence trial is more
time-consuming than a placebo trial, needing to recruit women by 30 weeks gestation (when 15-20% of women seek prenatal care) rather than by 34 weeks (when
about 75% seek care) and with a 50% larger sample size. Study completion could
take months to more than a year longer. A fast trial may save lives by speeding
implementation of a perinatal ARV program. Africa has 200,000 mother-to-child
HIV transmissions per year. Once implemented, a program reaching 50% of
care sites, with 75% of women appearing by 34 weeks gestation, and 50% ARV
efficacy would prevent 37,500 mother-to-child HIV transmissions per year.
Conclusion: We believe that using placebo controls is essential to reliably
establish an appropriate local standard of care. A placebo control trial of ARV
is far more likely than an equivalence trial to produce results that are valid and
generalizable for the African countries where these trials are being conducted.
It also requires less time and therefore supports earlier implementation of ARV
programs, which may prevent nearly 40,000 mother-to-child HIV transmissions
per year.
42202 HIV research and women: What influences attrition?
Kathleen Weber1, M. Cohen1, K. Riester3, S. Holman4, N. Hessol5,
Y. Barranday6, S. Back7. 1Cook County Hospital; 1900 W. Polk St., CCSN,
Chicago; 3New England Research Institutes, Watertown, MA; 4Suny Health
Science Center, Brooklyn, NY; 5 University of California, San Francisco, CA;
6 University of Southern California, Los Angeles, CA; 7Montefiore Medical
Center, Bronx NY, USA
Background: To describe factors associated with loss to follow-up (f/u) in the
Women's Interagency HIV Study (WIHS) during the first two years (visits 1-4).
Methods: The WIHS is a US longitudinal study at 6 sites to investigate HIV
disease in women with and at risk for HIV infection through semiannual visits.
Data from 2,412 living women (1,852 HIV+, 560 HIV-) were available. Using a

univariate model, variables including HIV serostatus, demographics, illness stage,
and psychosocial/behavioral factors were analyzed by Fisher's exact and Mantel
extension tests. A participant was categorized as lost to f/u if she missed --2
consecutive visits without subsequent f/u or was disenrolled.
Results: Of 2,412 participants, 348 (14%) were lost to f/u. HIV women had a
higher rate of loss to f/u compared to HIV+ women (21% vs. 13%, p <.001). Loss
to f/u was higher for women who had no prior history of participation in research
and used crack, cocaine, heroin, or injection drugs during the 6 months prior to

enrollment (p -.001 for both). Overall, women who had a primary provider at
baseline had a lower rate of loss to f/u (13.4% vs. 19.8%, p =.001). There was
a strong association between loss to f/u and high plasma viral load (>100,000
copies/ml) at both baseline and last available visit (p --.001). HIV+ women with
CD4 - 200 at baseline had a higher loss to f/u rate (p =.04). There was a marginal
association between loss to f/u and self report of Class C conditions; a health
care visit 2 months prior to baseline; and ethnicity (p =.06) with blacks having
the lowest loss to f/u. There was also differential loss to f/u among consortia sites
(p -.001). After adjusting for HIV serostatus, all variables remained associated
with loss to f/u except primary provider (p =.12).
Conclusions: Although retention rates after 2 years in the WIHS cohort remain
high (86%), there are identifiable behaviors and characteristics that may predict
attrition. These include chemical dependency, advanced disease, no primary
provider, and no prior research experience. Interventions and retention strategies
for participants with these characteristics may improve longitudinal f/u in studies
of women with HIV infection.
S42203 The Italian quality control study for CD4 evaluation in HIV
disease
Franco Pandolfi1, C. Alario1, E. Girardi2, L. Rava'2, G. Ippolito2, A. Kunkl3,
F. Aiuti1. 'Allergy & Clinical Immunology, Sapienza University Viale Dell
'Universita' 38 Rome; 2Spallanzani Hospital Rome; 3 University of Genova,
Genova, Italy
Objectives: To establish a national network for a quality control in evaluating
CD4 cell counts in Italian centers involved in HIV care.
Design: 68 centers were divided according to their geographic location into 8
groups (each with a reference center, RC) and tested 4 times/year (Tests A-D).
Methods: Tests A and B required the analysis of 3 coded whole blood samples
(2 identical) from 2 informed HIV+ patients, one with CD4 counts/mm3 - 200 and
one with -300 sent by each RC to affiliated labs. Test C was to assess the set
up performance and stability of flow cytometers over a 10 days period with the
FITC Combo Kit, which includes a QCWindows FITC reference standard and a
mixture of FITC quantitative and blank standards. In addition, 9 cellular samples
containing mixtures of CD4+ and CD8+ T cell clones at predetermined % of CD4
cells, were also used. Test D consisted in the evaluation of a blood sample from a
HIV patient sent directly from the coordinating center in Rome to all participating
labs. The medians of the determinations performed by the labs involved in each
test were taken as the "true" values. Unsatisfactory performances for % of CD4
were identified as a CD4 with residual values -~5% and with deviates --2. For
absolute numbers of CD4 cells, an unsatisfactory performance was defined as
CD4 counts with residual -~100 CD4 cells/mm3 and with deviates --~2.
Results: Most of the laboratories provided reliable results. Some labs failed
to provide satisfactory results for % (6.25% for test A and 6.17% for test B) or
absolute numbers (16.25% test A and 12.34% test B). Only 3.7% of the labs
gave unsatisfactory results in both tests A and B. Test C showed that 19% labs
had performance parameters outside the acceptable range and their instruments
required calibration, while 17% showed unacceptable results in the analysis of
T-cell clones. In Test D, one lab gave unsatisfactory percentages and 4 labs gave
unreliable absolute numbers.
Conclusions: Most Italian labs provide reliable results in evaluating the numbers of CD4 cells in HIV-1' samples, but the importance of running a quality
control program is highlighted by our experience with those centers providing
data that may lead to incorrect classification of patients. Since determinations of
percentages appear to be more reproducible, we suggest that monitoring of patients under combination therapy should be done not only with absolute numbers
of CD4 cells, but also with their percentages.
42204 Different analyses give highly variable estimates of HIV-1
RNA undetectability and log reduction in clinical trials
Andrew Hill, R. Demasi, M. Kuhn. Avanti Steering Committee; IMA,
Glaxowellcome R & D, Stockley Park Oxeridge, Middlesex, Ubii IBT, London;
Raleigh, UK
Introduction: Efficacy of treatments is being evaluated in terms of the loglo reduction in HIV-1 RNA and the percentage of patients below the limits of detection.
However it is not clear whether these measures can be reliably compared across
clinical trials.
Methods: Different methods of calculating the percentage of patients HIV-1
RNA undetectable and log10 reduction in HIV-1 RNA were made for two trials
of triple combination treatment with AZT/3TC and either indinavir (AVANTI-2, n =
100) or nelfinavir (AVANTI-3, n = 100). HIV-1 RNA was analysed by the Roche
Amplicor assay.
Results: Intent to Treat analysis led to much lower estimates of % undetectable.
The percent of patients undetectable was lower for the ultra-direct assay (-20
copies/ml) than for the standard assay (-400 copies/ml). Reductions in HIV RNA
to under 500 but over 20 copies/mi led to virological failure. Patients with low
baseline RNA levels showed higher rates of undetectability. Percent undetectable

peaked at week 28 and then declined. Estimates of "percent undetectable" at
week 52 on AZT/3TC/indinavir ranged from 78% (As Treated analysis, -500
endpoint) to 46% (Intent to Treat analysis, <20 endpoint). The new "censoring"
methods, used for ACTG 306 and ACTG 320, led to lower estimates of log
reduction in RNA than uncensored analysis. Additionally, different measures of
log reduction (mean, median, maximal) led to different estimates of treatment
effect.