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

938 Abstracts 43470-43473

12th World AIDS Conference

category with a reduction of 25% from the figure predicted for 1996 and of around
40% from the one predicted for 1997.
Methods: A simulation model has been developed to investigate possible explanations for the above decline such as treatment effects, or decrease in HIV
incidence in recent years. Starting from incubation times and times of infection
estimated from a Bayesian back-calculation, the model generates the entire progression of each infected individual including times of HIV diagnosis, starting
mono, or dual, or multiple therapy, and time of AIDS onset.
Results: The decline in AIDS incidence observed amongst homosexual men,
can be explained in terms of a very effective therapy suddenly introduced both
at early and late stages of the disease. The magnitude of the decline and its
sustainability clearly depends on the assumptions about treatment efficacy and
duration, and on the level of uptake of the new therapies. Under reasonable assumptions for treatment efficacy and uptake, an efficacy of limited duration would
result in the number of AIDS cases rising again in the near future. Alternatively, a
further decline in HIV incidence in the 1990's would not account for the recent fall
in AIDS incidence. The prevalence of HIV infected persons on treatment is likely
to increase 3 fold between 1996 and 1999.
Conclusions: The fall in AIDS incidence in 1996/97 is probably due to highly
effective combination therapy, assuming a substantial therapy uptake at various
stages of disease. There is a need to plan for a large increase in treatment
demand, especially at the early stages of the disease. It is unlikely that a change
in HIV incidence during the 1990s would have affected current AIDS incidence so
that a further fall in the incidence of HIV infection remains the key medium term
public health target.
43470 1Methods of measuring sexual networks of youth to model
the social and geographic spread of HIV and STD
Benny Jose, S.R. Friedman, P.L. Flom, M. Sandoval, D.C. Des Jarlais,
A. Neaigus. N.D.R.I. 2 World Trade Center, 16th FL, New York, NY 10048, USA
Objectives: Modelers often stress the need for community-level data on sexual
behaviors and networks, but probability surveys to collect such data are rare.
We present methods to sample youth, collect data on sexual behaviors and
networks, and measure the social, geographic and behavioral heterogeneity of
sexual networks.
Methods: In an ongoing multi-stage probability survey of 18-24 year old youth
in Bushwick, a low income section of New York City, all households in randomly
selected face-blocks (the primary sampling units or PSU) are screened for ageeligible youth. A randomly selected youth in each household is tested for HIV
and STDs, and interviewed about sexual behaviors and networks. Questions
asked about the three people with whom subjects most frequent had sex in the
past 12 months include, for each, sociodemographic background, where they
live, relationship type and duration, sexual frequency, and condom use within the
relationship.
Results: Of 986 dwelling units in 19 PSU, 198 age-eligible households were
identified. 165 (83%) youth were interviewed. 151 (92%) of them provided blood
and urine samples. Prevalence of infections: HIV 0%; syphilis 0%; gonorrhea 2%;
chlamydia 9%. 146 (89%) subjects reported having sex in the prior 12 months;
of these, 127 (87%) had 2 or more sexual partners. Among the sexually active,
81% engaged in unprotected sex; 113 provided data on 192 sexual network
partnerships. Unprotected sex occurred in 71% (130/184) of relationships. 27%
of all subjects reported concurrent partnerships. Only 8% of subjects had sexual
contacts outside of their own race/ethnicity. Of all sexual network partnerships,
55% were with Bushwick residents, 37% lived elsewhere in the city, and only 8%
lived outside of New York City.
Conclusions: Data on sexual behaviors, networks, and HIV/STD infections can
be collected from representative samples of youth. Unprotected sex, multiple partnerships, and concurrency are highly prevalent among youth in this low income,
New York City community. Their sexual networks tend to be geographically-limited and racially/ethnically homogeneous. Community-level samples can generate
reasonably accurate network parameters that may improve the predictive value
of HIV/STD transmission models.
|43471 The effect of combination therapy on long-term HIV
disease progression: A population model for Scotland
Tamiza Parpia1, G.M. Raab1, G. Allardice2, D.J. Goldberg2, J.J. McMenamin2.
1Napier University 10 Colinton Road, Edinburgh, Ehio 5DT, Scotland; 2Scieh
Ruchill Hospital, Glasgow Scotland, UK
Objectives: To investigate the effect of combination therapy on long term forecasts
of the immunological state of the Scottish HIV population and the consequences
for policies on resource allocation
Methods: Data collated from surveillance schemes coordinated by the Scottish
Centre for Infection and Environmental Health (SCIEH) involve repeated immunological and clinical measurements on each patient. The forecasting process
involves three stages, the core of which is based on a random slopes model that is
commonly used to analyse such longitudinal data: 1) The model of immunological

(CD4) decline was developed using all patients alive at 30 June 1997, allowing
for the effect of combination therapy which became available in Scotland during
late 1996/early 1997; 2) The model in stage 1 was then used to predict patients
future CD4 values according to different scenarios for the possible long term
effects of treatment on their immunological status; 3) Death rates and rates of
new patients entering the monitoring scheme, estimated from earlier data, were

incorporated into the model. This process enables estimation of the future CD4
profile for patients in Scotland and hence the resources required for their care.
Results: The process was applied to 1657 patients alive at 30 June 1997.
Combination therapy was found to dramatically improve the immunological profile
of the Scottish HIV population. This effect varied by patients' CD4 values at
commencement of combination therapy - over a six month period of observation,
the highest average increase of 2.5% (CD4 count as a percentage of total lymphocytes) or 42 cells/mm3 was seen in patients at a late stage of HIV disease
(i.e. CD4% of <10% or CD4 count <130 cells/mm3).
Conclusion: Recent availability and uptake of combination therapy has had a
substantial impact on the immunological progression of the Scottish HIV population. In the absence of information on the long term effects of these therapies,
forecasting resources required to care for the HIV population must include a
wide range of options to guide policy decisions. We have therefore presented
forecasts according to different scenarios of how these treatments will affect the
population's future immunological state.
S43472 Estimation of date of HIV-seroconversion for prevalent
cases based on CD4 numbers at entry, with application to
natural history studies
Ronald Geskus, L. Lam, R. Van Leeuwen, R.A. Coutinho. 1 Nieuwe Archtergracht
100, 1018 WT, Municipal Health Service, Amsterdam, The Netherlands
Background: Most natural history studies on AIDS are based on seroconverter
cohorts. In the Amsterdam cohort study on HIV infection and AIDS among homosexual men, 31.4% was HIV-positive at entry between October 1984 and April
1985. For each prevalent case at entry, a distribution of time since seroconversion can be estimated by comparing his CD4 level with more than twelve years
of follow-up information on CD4 development from the seroconverters. In this
way, prevalent cases can be incorporated in natural history studies, such as the
influence of early seroconversion on the AIDS-free time.
Methods: For each prevalent case, random dates of seroconversion are drawn
via a two-step empirical procedure. First a random seroconverter is drawn, after
which a random time is chosen from all his time points after seroconversion with
comparable CD4 level as the prevalent case at entry. Combining the seroconversion estimates of the prevalent cases with the seroconverters yields an estimate
of the cohort-wide seroconversion distribution. In an application of this method,
the influence of date of seroconversion on the AIDS-free time is investigated by
imputation of the drawn dates of seroconversion, using a dichotomous analysis
with October 1984 as cut point.
Results: The peak in the seroconversion density occurred in the second half of
1984, just before the start of the cohort study. In the dichotomous analysis, persons who seroconverted before October 1984 have a better AIDS-free survival,
though not significantly at the 5% confidence level (log-rank test).
Conclusions: In the Amsterdam cohort study on HIV infection and AIDS, seroconversion rate was highest during the second half of 1984. In a dichotomous
analysis, persons who seroconverted before October 1984 do not have their
AIDS-free survival significantly different from later seroconverters.
43473 Modeling the dyanmics in plasma viral load and CD4+
lymphocyte count as a response to treatment changes
Elizabeth Brown1, S. MaWhinney1, B. Young1, S. Johnson', D. Kuritzkes1,
K. Kafadar2, R.H. Jones1. 1 U of Colorado Health Sciences Center; 2400 E 13th
Avenue #6 Denver, CO; 2University of Colorado, Denver, CO, USA
Objectives: To develop statistical models to analyze the impact of treatment
changes based on viral genotype.
Design: The sequences of HIV-1 protease and reverse transcriptase were
collected from thirty-four HIV-1 infected patients who had failed current anti-viral
treatment which included a protease inhibitor. Longitudinal data were also collected for patients' treatment, plasma viral load (VL) and CD4+ lymphocyte count
(CD4) histories.
Methods: Each patient follows a different treatment regimen, with VL and CD4
taken at different points in time. Plots of VL and CD4 over time with treatment
history for each patient are important tools in understanding the relationship between these factors. Given the biological mechanism of antiretroviral treatment,
we expected to see an inflection point in the VL and CD4 curves shortly after
treatment change. Assuming the treatment was effective and resistance followed,
a local minimum would occur in the viral load curve. Traditional methods for
analyzing VL and CD4 cannot be used with this complex data set. We develop
new methods which generate bivariate smoothing splines based on state-space
equations to model changes in CD4 and VL.
Results: Preliminary graphs give strong evidence in support of our hypothesis
of points of inflection occurring soon after changes in drug treatment. These
hypotheses are supported in plots from many patients with different treatment
histories and follow-up times.
Conclusions: We are generating smoothing splines using state space equations to model CD4, VL and their rate of change. These models are able to include

the effect of treatment changes on VL and CD4 as well as being able to model
data collected in clinical trials as well as in a real world clinical setting.