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

802 Abstracts 42146-42150

12th World AIDS Conference

S42146 Formal analysis of the time dependent predictive value of
HIV load and host genetic factor on disease progression
rate
Avidan U. Neumann1, L. Deutsch1, L. Zhang2, L. Kostrikis2, D.D. Ho2,
S. Wolinsky3. 1Bar-llan University, Life Sci. Dept., Ramat-Gan 52900;
Ramat-gan, Istrael; 2ADARC, New York, NY; 3NWU, IL, USA
Background: Viral load has been shown to be a good predictive marker for disease progression rate both at about a year after HIV infection in sero-convertors
and at time of entry for sero-prevalent cohorts. However, it is not yet clear how
does the predictive value of viral load changes in later stages of infection. Various
mutations in co-receptor genes (e.g. CCR5 and CCR2) were also found to be
correlated with disease progression. However, the relation between viral load
and these mutations in the combined effect was not studied. Here, we study the
multi-factorial effect of viral load at different times of infection and of the above
mutations on disease progression.
Methods: 117 sero-convertors from the MACS Chicago cohort were analyzed
longitudinally and cross sectionally. Viral load was available for the full follow-up
period of 80 of them in addition to CD4 and CD8 counts. In addition all patients
had genetic data and viral phenotype data at a few points during progression.
Survival analysis was done on viral load at both random points and every year
from sero-conversion using the quartile levels at 1 year post infection. Bi-variate
survival analysis was also conducted on the CCR5/CCR2 mutations and viral
load at same time.
Results: Viral load is a good predictive marker even when random time points
were taken. However, viral load from specific periods after sero-conversion show
some deviation in predicting survival relative to the one year post infection prediction. Mutations in CCR5 (CCR2) are not significant in patients who begin with
high viral loads.
Conclusion: Viral load is a good predictor even if time from infection is not
known at the time of measurement. However, viral load does effect differently
survival when measured at different times, pointing to the fact that viral load
changes are important for progression. Mutations in CCR5 (CCR2) that are
normally slowing progression do not protect patients that do not control viremia
already at sero-conversion.
|42147, Rectal sno-strips and biopsies: A simple method
to assess rectal viral load and lymphoid
architecture
Stephen Tabet1, C.L. Celum2, C. Surawicz2, S. Horton3, H. Brar2, D. Dondero4,
R. Haggit2. 1901 Boren Ave-Suite 1300 Seattle, WA 98104; 2University of
Washington, Seattle WA; 3Fred Hutchinson Cancer Research Center, Seattle
WA; 4 Viral & Ricket Tsial Disease Laboratory, Berkeley CA, USA
Objectives: First, to assess if rectal biopsies offer a relatively noninvasive way of
obtaining samples for assessing lymphoid architecture at different stages of HIV
infection and response to antiretroviral therapy. Second, to assess if discordance
between HIV viral load (VL) in plasma and rectal samples is determined by
inflammation or other local factors.
Methods: 10 HIV+ men were followed for 8 wks in a HIVNET phase I study
of nonoxynol-9 (N-9). Plasma and rectal sno-strips were obtained at baseline, 2,
5, and 6 wks of N-9 use. VL was assayed with Amplicor RT-PCR on plasma and
rectal sno-strips, using a dilution factor of 10 /~ for sno-strips. Rectal mucosal
biopsies were obtained through an anoscope with endoscopic findings at baseline,
5, and 6 weeks of N-9 use. Biopsy specimens were fixed, routinely processed,
embedded in paraffin and step serial sectioned at 5 plm. Ribbons of step serial
sections were stained with hematoxylin and eosin and examined by a pathologist
who had no knowledge regarding the clinical findings. Semi-quantitative HSV
PCR and HPV hybrid capture were performed on anoscopically-obtained swabs
and perianal swabs, respectively.
Results: None of the 10 men who underwent rectal biopsies had complications.
Three of 33 (9%) samples had > 0.5 log rectal VL greater than plasma VL. Five
of the 6 men with discordant HIV VL in plasma and rectal samples were on
antiretroviral therapy; 4 on highly active antiretroviral therapy. Five rectal biopsy
samples had surface and crypt inflammation with two specimens having a > 1 log
higher VL than plasma VL. Fifty-three of 56 biopsy specimens had intact lymphoid
follicles with a median number of two follicles per biopsy specimen.
Conclusion: Rectal biopsies provide a relatively simple method of obtaining
lymphoid follicles to assess architecture. This study did not find an association
between HIV shedding and local rectal factors (e.g., inflammation and HSV or
HPV load), but larger studies are needed to further assess this finding. The higher
rectal HIV VL in several samples has public health implications since anal sex is
still the primary route of HIV transmission among gay and bisexual men.
S42148 Cervicovaginal HIV-1 RNA levels and plasma viral load
Susan Cu Uvin1, A.M. Caliendo2, T.P. Flanigan1, K.H. Mayer3,
C.C.J. Carpenter2, R. Russo-Patterson1, J. Allegra2. 1The Miriam Hospital 164

Summit Avenue Providence Phode Island 02960; 2Massachusetts General
Hospital Boston MA; 3Memorial Hospital Pawtucket RI, USA
Objective: To assess the relationship of cervicovaginal lavage (CVL) HIV-1 RNA
level to plasma HIV-1 RNA level in women living with HIV.
Methods: 172 paired CVL and plasma specimens were analyzed. Quantitation
of HIV-1 RNA from CVL and plasma specimens was performed using nucleic acid
sequence-based amplification assay (NASBA).

Results: Plasma HIV-1 RNA was detectable in 120/172 (70%) of samples ranging from 330-4,000,000 copies/ml. 63/172 (37%) of CVL samples had detectable
HIV-1 RNA ranging from 180-440,000 copies/ml. CD4 cell counts ranged from
2-1,862/mm3.
HIV-1 RNA Detection in Paired Blood Plasma and CVL Specimens
CVL HIV-1 RNA(+) CVL HIV-1 RNA(-)
Plasma HIV-1 RNA(+) 35% (61/172) 34% (59/172) Plasma
HIV-1 RNA(-) 1% (2/172) 29% (50/172)
Among 61 samples with detectable HIV-1 RNA in both plasma and CVL, 45 had
higher levels in plasma and 16 had higher levels in CVL. For these 16 samples,
CVL HIV-1 RNA was 1.1 to 52-fold higher than plasma HIV-1 RNA. 96% of women
with plasma viral load below the limits of detection also had undetectable CVL
viral load.
Conclusions: CVL HIV-1 RNA was detected across the entire range of measurable plasma HIV-1 RNA. CVL HIV-1 RNA levels may be higher than plasma
levels. There is no absolute plasma HIV-1 RNA below which HIV-1 genital shedding does not occur. These data suggest possible discordance between plasma
and CVL HIV viral load.
42149 Biases in the assessment of viral load reduction in HIV
Sclinical trials
lan Marschner', R.A. Betensky', V. Degruttola1, D.R. Kuritzkes2. 1Harvard
School of Public Health 677 Huntington Ave Boston MA 02115; 2University of
Colorado Denver CO, USA
Background: Clinical trial endpoints based on magnitude of reduction in HIV-1
RNA levels provide an important complement to endpoints based on percent
of patients achieving virologic suppression. However, interpretation of magnitude of reduction can be biased by measurement limitations of virologic assays,
particularly lower and upper limits of quantitation.
Methods: Using data from the US AIDS Clinical Trials Group (ACTG), widely
used crude methods of analyzing HIV-1 RNA reductions were compared with
methods that take account of censoring of HIV-1 RNA measurements. Such
methods include Kaplan-Meier and censored regression analyses. Data from
ACTG study 306, which made use of the bDNA assay to compare double combinations of ZDV, 3TC, d4T and ddl, highlight the methodology and comparisons.
Results: Standard crude methods of analysis consistently underestimated
treatment effects. Although absolute HIV-1 RNA reductions were moderate (on
the order of 0.5-1.5 loglo), treatment differences were underestimated by up
to 0.5 loglo when crude methods were used. In some cases the bias induced
by crude methods masked statistically significant differences between treatment
arms. Adjustment for baseline HIV-1 RNA levels was necessary, and convenient parametric analyses performed as well as more complex non-parametric
analyses.
Conclusions: Widely used crude analyses of magnitude of HIV-1 RNA reduction can substantially underestimate treatment differences. Failure to account for
censoring of HIV-1 RNA measurements can lead to the masking of significant
treatment differences. To obtain complete information about virologic response
to treatment, appropriately analyzed magnitude of change endpoints should be
assessed in conjunction with percent suppression endpoints.
42150 | Management of antiretroviral therapy for HIV-1 infection:
Modeling when to change therapy
Lawrence Wein. MIT E53-343 Cambridge, MA 02142, USA
Objective: To evaluate different strategies for monitoring HIV-infected patients on
antiretroviral therapy - how frequently to measure plasma HIV RNA and perform
genotype tests, and when to change therapy.
Design and Setting: A Monte Carlo simulation model is developed which
tracks many individuals' viral loads and absence/presence of opportunistic infection during the course of therapy. The model uses clinical and virological data
and incorporates noisy, left-censored viral load measurements, imperfect genotype tests and statistical variation in patient parameters. This computer modeling
evaluates strategies intended for use outside controlled trials in clinical practice.
Main Outcome Measures: The objective of the modeling is to minimize the delay in detection of viral rebound while maintaining a low (prespecified) probability
of switching therapy before viral rebound occurs.
Results: The 1997 recommendations of the International AIDS Society-USA
Panel and all drug-switching policies that are based solely on viral load rebound
involve lengthy delays in detection of rebound, particularly when patients are
drug-naive and the detection limit of the viral load assay is 500 HIV RNA log
copies/mL. An assay for plasma HIV RNA with a detection limit of 20 HIV RNA
copies/mL decreases delay of detection of viral rebound substantially, but genotyping achieves only minor marginal reductions in detection delay. A policy based
on switching at a predetermined time after initiation of therapy if viral load has not
yet rebounded (proactive switching) leads to the smallest detection delays and

lowest costs of any of the strategies tested.
Conclusion: The modeling indicates that policies for switching antiretroviral
therapies after viral load rebounds will likely be ineffective at preventing emergence of drug-resistant HIV-1 for individuals who experience drug failure. A
proactive policy, which periodically changes combination regimens prior to detecting evidence of drug failure, may be the only monitoring approach that can
significantly reduce the occurrence of resistance.