Comparison of models for analyzing two-group, cross-sectional data with a Gaussian outcome subject to a detection limit.

A potential difficulty in the analysis of biomarker data occurs when data are subject to a detection limit. This detection limit is often defined as the point at which the true values cannot be measured reliably. Multiple, regression-type models designed to analyze such data exist. Studies have compared the bias among such models, but few have compared their statistical power. This simulation study provides a comparison of approaches for analyzing two-group, cross-sectional data with a Gaussian-distributed outcome by exploring statistical power and effect size confidence interval coverage of four models able to be implemented in standard software. We found using a Tobit model fit by maximum likelihood provides the best power and coverage. An example using human immunodeficiency virus type 1 ribonucleic acid data is used to illustrate the inferential differences in these models.

An Evaluation of Statistical Methods for Analyzing Follow-Up Gaussian Laboratory Data with a Lower Quantification Limit.

Laboratory data with a lower quantification limit (censored data) are sometimes analyzed by replacing non-quantifiable values with a single value equal to or less than the quantification limit, yielding possibly biased point estimates and variance estimates that are too small. Motivated by a three-period, three-treatment crossover study of a candidate vaginal microbicide in human immunodeficiency virus (HIV)-infected women, we consider four analysis methods for censored Gaussian data with a single follow-up measurement: nonparametric methods, mixed models, mixture models, and dichotomous measures of a treatment effect. We apply these methods to the crossover study data and use simulation to evaluate the statistical properties of these methods in analyzing the treatment effect in a two-treatment parallel-arm or crossover study with censored Gaussian data. Our simulated data and our mixed and mixture models consider treated follow-up data with the same variance as the baseline data or with an inflated variance. Mixed models have the correct type I error, the best power, the least biased Gaussian parameter treatment-effect estimates, and appropriate confidence interval coverage for these estimates. A crossover study analysis with a period effect can greatly increase the required study sample size. For both designs and both variance assumptions, published sample-size estimation methods do not yield a good estimate of the sample size to obtain the stated power.

Statistical methods for the analysis of time-location sampling data.

Time-location sampling (TLS) is useful for collecting information on a hard-to-reach population (such as men who have sex with men [MSM]) by sampling locations where persons of interest can be found, and then sampling those who attend. These studies have typically been analyzed as a simple random sample (SRS) from the population of interest. If this population is the source population, as we assume here, such an analysis is likely to be biased, because it ignores possible associations between outcomes of interest and frequency of attendance at the locations sampled, and is likely to underestimate the uncertainty in the estimates, as a result of ignoring both the clustering within locations and the variation in the probability of sampling among members of the population who attend sampling locations. We propose that TLS data be analyzed as a two-stage sample survey using a simple weighting procedure based on the inverse of the approximate probability that a person was sampled and using sample survey analysis software to estimate the standard errors of estimates (to account for the effects of clustering within the first stage [locations] and variation in the weights). We use data from the Young Men's Survey Phase II, a study of MSM, to show that, compared with an analysis assuming a SRS, weighting can affect point prevalence estimates and estimates of associations and that weighting and clustering can substantially increase estimates of standard errors. We describe data on location attendance that would yield improved estimates of weights. We comment on the advantages and disadvantages of TLS and respondent-driven sampling.

High-risk human papillomavirus reactivation in human immunodeficiency virus-infected women: risk factors for cervical viral shedding.

OBJECTIVE: To evaluate the presence of and estimate risk factors for reactivation of latent high-risk human papillomavirus (HPV) cervical infection in human immunodeficiency virus (HIV)-infected and HIV-uninfected women. METHODS: Data from 898 women in the HIV Epidemiology Research Study (HERS) were used to evaluate cervical HPV latency and reactivation. Prior exposure to HPV types (16, 18, 31, 35, and 45) was determined by serologic testing at enrollment, and cervical shedding of HPV was detected by polymerase chain reaction at 6-month intervals. Human papillomavirus cervical shedding and sexual history were used to estimate rates of reactivation and recurrence. Repeated measures survival analysis was used to estimate hazard ratios and 95% confidence intervals for reactivation and recurrence. Rates of total HPV shedding (recurrence and reactivation) during follow-up were assessed by HIV status and rate ratios were calculated. RESULTS: Reactivation of latent HPV infections was observed in HIV-infected women, but few reactivation events were identified in HIV-uninfected women. Factors consistently associated with reactivation in HIV-infected women included CD4 count less than 200/mm and age younger than 35 years. Women infected with HIV had 1.8 to 8.2 times higher rates of viral shedding (reactivation plus recurrence) compared with HIV-uninfected women. CONCLUSION: Women with a history of cervical HPV infection may be at risk of reactivation of latent viral infection even in the absence of sexual activity, and this risk is higher in women with HIV infection. LEVEL OF EVIDENCE: II.

HIV genital shedding and safety of Carraguard use by HIV-infected women: a crossover trial in Thailand.

OBJECTIVE: To evaluate the safety, including impact on genital HIV RNA shedding, of Carraguard vaginal gel in HIV-infected women. DESIGN: This is a randomized, controlled, crossover study of Carraguard in HIV-infected women in Thailand. METHODS: Each woman (CD4 cell count 51-500 cells/microl and not on antiretroviral therapy) used each treatment (Carraguard, methylcellulose placebo, and no-product) once daily for 7 days during each 1-month period (3-week wash-out). Women were randomized to one of the six possible treatment sequences. Safety assessments were conducted at baseline (pregel), 15 min postgel, day 7, and day 14, and included HIV RNA measurements in cervicovaginal lavage (CVL) specimens. RESULTS: Sixty women were enrolled, and 99% of scheduled study visits were completed. At baseline, median age (34 years), CD4 lymphocyte count (296 cells/microl), plasma HIV viral load (4.6 log10 copies/ml), CVL HIV viral load (3.1 log10 total copies per CVL), and sexual behaviors were similar among randomization groups. HIV viral load, leukocyte and hemoglobin levels, and epithelial cell counts in CVLs were lower 15 min after application of Carraguard or placebo compared with no product; CVL HIV viral load was still lower at day 7 but returned to baseline by day 14. Carraguard use was not associated with prevalent or incident genital findings or abnormal vaginal flora. CONCLUSION: Carraguard appears to be well tolerated for once-daily vaginal use by HIV-infected women. The observed reduction in CVL HIV viral load in the gel months may be clinically relevant but could have resulted from interference with sample collection by study gels.

Estimating HIV incidence in the United States from HIV/AIDS surveillance data and biomarker HIV test results.

The development of an human immunodeficiency virus (HIV) test that detects recent infection has enabled the U.S. Centers for Disease Control and Prevention (CDC) to estimate annual HIV incidence (number of new infections per year, not per person at risk) in the United States from data on new HIV and acquired immunodeficiency syndrome (AIDS) diagnoses reported to HIV/AIDS surveillance. We developed statistical procedures to estimate the probability that an infected person will be detected as recently infected, accounting for individuals choosing whether and how frequently to seek HIV testing, variation of testing frequency, the reporting of test results only for infected persons, and infected persons who never had an HIV-negative test. The incidence estimate is the number of persons detected as recently infected divided by the estimated probability of detection. We used simulation to show that, under the assumptions we make, our procedures have acceptable bias and correct confidence interval coverage. Because data on the biomarker for recent infection or on testing history were missing for many persons, we used multiple imputation to apply our models to surveillance data. CDC has used these procedures to estimate HIV incidence in the United States.

Estimation of HIV incidence in the United States.

CONTEXT: Incidence of human immunodeficiency virus (HIV) in the United States has not been directly measured. New assays that differentiate recent vs long-standing HIV infections allow improved estimation of HIV incidence. OBJECTIVE: To estimate HIV incidence in the United States. DESIGN, SETTING, AND PATIENTS: Remnant diagnostic serum specimens from patients 13 years or older and newly diagnosed with HIV during 2006 in 22 states were tested with the BED HIV-1 capture enzyme immunoassay to classify infections as recent or long-standing. Information on HIV cases was reported to the Centers for Disease Control and Prevention through June 2007. Incidence of HIV in the 22 states during 2006 was estimated using a statistical approach with adjustment for testing frequency and extrapolated to the United States. Results were corroborated with back-calculation of HIV incidence for 1977-2006 based on HIV diagnoses from 40 states and AIDS incidence from 50 states and the District of Columbia. MAIN OUTCOME MEASURE: Estimated HIV incidence. RESULTS: An estimated 39,400 persons were diagnosed with HIV in 2006 in the 22 states. Of 6864 diagnostic specimens tested using the BED assay, 2133 (31%) were classified as recent infections. Based on extrapolations from these data, the estimated number of new infections for the United States in 2006 was 56,300 (95% confidence interval [CI], 48,200-64,500); the estimated incidence rate was 22.8 per 100,000 population (95% CI, 19.5-26.1). Forty-five percent of infections were among black individuals and 53% among men who have sex with men. The back-calculation (n = 1.230 million HIV/AIDS cases reported by the end of 2006) yielded an estimate of 55,400 (95% CI, 50,000-60,800) new infections per year for 2003-2006 and indicated that HIV incidence increased in the mid-1990s, then slightly declined after 1999 and has been stable thereafter. CONCLUSIONS: This study provides the first direct estimates of HIV incidence in the United States using laboratory technologies previously implemented only in clinic-based settings. New HIV infections in the United States remain concentrated among men who have sex with men and among black individuals.

Differentiating normal from abnormal rates of genital epithelial findings in vaginal microbicide trials.

BACKGROUND: Candidate vaginal microbicides could cause genital irritation, which in turn could facilitate HIV transmission instead of preventing it. While genital epithelial findings are documented in a standardized manner in most microbicide trials, little is known about background rates and predictors for many types of genital findings. STUDY DESIGN: A secondary analysis was conducted using data from a Phase II expanded safety study of the candidate microbicide Carraguard gel (Population Council, NY, USA) in Thailand. Genital findings were identified by visual inspection of the cervix, vaginal walls and external genitalia during pelvic exams prior to gel use (screening and enrollment) and during gel use (at 2 weeks and Months 1-12). Women were interviewed about potential risk factors for genital findings at every visit and tested routinely for sexually transmitted and vaginal infections. RESULTS: A total of 258 genital findings were identified in 152 woman-years of follow-up. Genital findings were positively associated with older age, increased parity, self-report of genital symptoms, positive HSV-2 serology, bacterial vaginosis by Nugent scoring and the presence of a genital finding at baseline. Furthermore, vaginal findings were positively associated with vaginal practices and yeast infections. Genital findings were negatively associated with use of hormonal contraception, inconsistently associated with frequency of sex and applicator use, and not associated with condom use. CONCLUSIONS: Several factors that are common in women of reproductive age account for the background rate of genital epithelial findings in this population.

Safety and acceptability of the candidate microbicide Carraguard in Thai Women: findings from a Phase II Clinical Trial.

OBJECTIVE: To determine the safety and acceptability of vaginal application of Carraguard, a carrageenan-derived candidate microbicide gel. DESIGN: A randomized, placebo-controlled, triple-blinded clinical trial was conducted in Chiang Rai, northern Thailand. METHODS: Women were asked to insert one applicator of study gel vaginally at least three times per week (with or without sex) and to use gel with condoms every time they had sex. Safety was assessed by visual inspection of the vagina and cervix, changes in vaginal flora and self-reported symptoms at day 14, month 1 and then monthly for up to 1 year. Acceptability was assessed through reported use of the gel, return of used and unused applicators, and quarterly interviews. RESULTS: One hundred sixty-five women were randomized: 83 to Carraguard and 82 to the placebo (methylcellulose gel) group. Study gel use was similarly high in both groups throughout the trial with an average of four applicators per week. Carraguard use was not associated with abnormal genital clinical findings, abnormal vaginal flora, Pap smear abnormalities or other abnormal clinical signs or symptoms. Adverse events were mostly mild, not attributed to gel use, and similarly distributed between groups. Participants in both groups reported high acceptability. CONCLUSIONS: Carraguard can safely be used an average of four times per week with or without sex and is acceptable to Thai women. A Phase III efficacy trial of Carraguard is warranted and is currently ongoing in South Africa.

Estimating the distribution of a renewal process from times at which events from an independent process are detected.

The analysis of length-biased data has been mostly limited to the interarrival interval of a renewal process covering a specific time point. Motivated by a surveillance problem, we consider a more general situation where this time point is random and related to a specific event, for example, status change or onset of a disease. We also consider the problem when additional information is available on whether the event intervals (interarrival intervals covering the random event) end within or after a random time period (which we call a window period) following the random event. Under the assumptions that the occurrence rate of the random event is low and the renewal process is independent of the random event, we provide formulae for the estimation of the distribution of interarrival times based on the observed event intervals. Procedures for testing the required assumptions are also furnished. We apply our results to human immunodeficiency virus (HIV) test data from public test sites in Seattle, Washington, where the random event is HIV infection and the window period is from the onset of HIV infection to the time at which a less sensitive HIV test becomes positive. Results show that the estimator of the intertest interval length distribution from event intervals ending within the window period is less biased than the estimator from all event intervals; the latter estimator is affected by right truncation. Finally, we discuss possible applications to estimating HIV incidence and analyzing length-biased samples with right or left truncated data.

A two-stage sampling method for clinical surveillance of individuals in care for HIV infection in the United States.

OBJECTIVES: The goals of this study were two-fold: (1) to describe methods for drawing a population-based sample of individuals in care for HIV infection and (2) to compare data from the sample with data from existing surveillance systems that describe care for HIV. METHODS: The authors implemented a two-stage sampling method, using local HIV/AIDS surveillance data as a sampling frame of HIV care providers in three states. At selected providers, medical records of a random sample of patients were abstracted. RESULTS: The medical records of a number of patients, ranging from 253 to 374 individuals per state, were abstracted. The demographics of sampled individuals and of individuals reported to the local HIV/AIDS surveillance program were similar; however, differences existed in the proportion of individuals receiving HIV care consistent with treatment guidelines between the sample and a contemporary facility-based supplemental surveillance project. The median design effect for outcomes collected in the sample was 1.8 (range=0.5-29.6). CONCLUSIONS: This survey method is feasible for collecting population-based data on patients in care for HIV. Sample size and some design elements should be changed in future studies to increase precision of estimates and usefulness of data for local planning and evaluation.

HIV-1 viral load and other risk factors for mother-to-child transmission of HIV-1 in a breast-feeding population in Cote d'Ivoire.

Short-course antiretroviral regimens have been evaluated to reduce mother-to-child transmission of HIV in resource-limited settings. This report from Abidjan, Cote d'Ivoire, examines the risk factors for HIV transmission by 1 and 24 months among breast-feeding women. Eligible HIV-1-seropositive pregnant women enrolled in this randomized double-blind clinical trial were randomly assigned to receive either oral zidovudine (ZDV) (n = 126) prophylaxis or placebo (n = 124). Maternal prophylaxis began at 36 weeks of gestation (300 mg ZDV twice daily antepartum and 300 mg every 3 hours intrapartum); there was no neonatal prophylaxis component. The cumulative risk of transmission in the treatment group was 11.9% and 22.1% by 1 and 24 months, respectively. In adjusted analyses, viral load at enrollment was the strongest predictor of transmission (per log increment: odds ratio [OR] = 4.8, 95% confidence interval [CI]: 2.5-9.5 at 1 month; OR = 5.7; 95% CI: 3.1-10.8 at 24 months). Overall, ZDV prophylaxis was not significantly protective for infection at 1 or 24 months. Comparing ZDV with placebo following dichotomization of viral load (<50,000 vs. > or =50,000 copies/mL) at enrollment, however, there was a significant effect of ZDV seen only among those women with a low viral load at enrollment. The substantial risk of transmission despite ZDV prophylaxis, particularly among those with higher viral loads, underscores the need to find more effective regimens appropriate for use in resource-limited settings.

When to begin highly active antiretroviral therapy? Evidence supporting initiation of therapy at CD4+ lymphocyte counts <350 cells/microL.

We assessed the risk of acquired immunodeficiency syndrome (AIDS)-related opportunistic illness or death among persons first prescribed highly active antiretroviral therapy (HAART) in January 1996 or later in the Centers for Disease Control and Prevention's Adult and Adolescent HIV Spectrum of Disease Project. Patients were included if they were naive to antiretroviral drugs and had no history of AIDS-related opportunistic illness. Risk was assessed as a function of CD4+ lymphocyte count and human immunodeficiency virus load at the time of initiation of HAART in a Cox proportional hazards model. Hazard ratios for AIDS or death were 6.3, 3.5, and 1.7 for persons with baseline CD4+ cell counts of 0-49, 50-199, and 200-349 cells/microL, respectively, compared with the referent (CD4+ cell count > or =500 cells/microL). HAART should not be deferred until the CD4+ cell count reaches <200 cells/microL. The increased hazard associated with CD4+ cell counts of 200-349 cells/microL was modest but supports initiation of HAART at CD4+ cell counts <350 cells/microL, particularly in patients with high virus loads.

Estimating the efficacy of interventions to prevent mother-to-child transmission of human immunodeficiency virus in breastfeeding populations: comparing statistical methods.

Postnatal transmission of human immunodeficiency virus infection through breastfeeding complicates evaluating the efficacy of interventions aimed to reduce mother-to-child transmission risk. Results from trials in Africa evaluating either peripartum antiretroviral therapy or refraining from breastfeeding show an estimated long-term efficacy at 15-24 months of age between 25 and 50 percent. Differences in statistical methods, duration of follow-up, and age at weaning hinder direct comparison between trials. The authors recently outlined theoretically preferred statistical methods for evaluating interventions aimed to reduce risk of mother-to-child transmission of human immunodeficiency virus. When multiple test results and/or supplementary information is available, the more sophisticated methods account for the fact that exact age at infection is unknown, that risk for infection ends at weaning, or that censoring due to death may be informative. The authors apply these methods to four scenarios, using data from four randomized trials carried out in Africa between 1995 and 2000. The authors' findings suggest that, to estimate the cumulative proportion infected at age 6 weeks, a standard Kaplan-Meier approach is likely to give valid results. For estimation of this proportion at age 18 months, more sophisticated methods, such as the extension of the Kaplan-Meier procedure to interval-censored data and competing risks, would be preferred.

Postnatal transmission of HIV-1 after a maternal short-course zidovudine peripartum regimen in West Africa.

BACKGROUND: To assess the postnatal transmission (PT) risk of HIV-1 after a maternal short-course zidovudine regimen in a breastfeeding population. METHODS: Data were pooled from two trials: ANRS 049a DITRAME (Abidjan, Côte d'Ivoire and Bobo-Dioulasso, Burkina-Faso) and RETROCI (Abidjan). Consenting HIV-1 seropositive women were randomized at 36-38 weeks' gestation between September 1995 and February 1998, to receive oral zidovudine or placebo: one tablet twice daily until delivery, and in DITRAME only, for 7 more days. A PT case was infection in a child with a negative HIV-1 PCR at age >/= 30 days who later became infected as defined by a positive HIV-1 PCR, or if aged >/= 15 months, a positive HIV serology. Cumulative risks (CR) of PT were computed using a competing risk approach with weaning as a competing event. FINDINGS: At age 24 months, CR for PT were similar in the zidovudine (9.8%, n = 254) and placebo groups (9.1%, n = 225). In a multivariate model of PT risk factors, the treatment effect was not significant, maternal CD4 cell count < 500 x 10(6)/l at entry tripled the hazard compared to women with CD4 cell counts >/= 500 x 10(6)/l [hazard ratio (HR), 3.14; 95% confidence interval (CI), 1.31-7.49] as well as an increased maternal plasma viral load at entry (HR, 2.65 for 1 log(10) increase; CI, 1.75-4.00). INTERPRETATION: PT occurred at a similar rate between arms and therefore reduced the long-term overall efficacy of this peripartum zidovudine regimen at age 24 months. The higher risk of PT among women with low CD4 cell count emphasizes the importance of identifying interventions to prevent PT for these women.

Kaposi's sarcoma in Uganda: risk factors for human herpesvirus 8 infection among blood donors.

Human herpesvirus 8 (HHV-8) is etiologically linked to Kaposi's sarcoma, a common cancer in Uganda. The authors assessed HHV-8 seroprevalence, risk factors for infection, and HHV-8 assays in a cross-sectional study of Ugandan blood donors. Of 3,736 specimens, the authors selected 203 reactive for HIV, hepatitis B surface antigen (HBsAg), or syphilis, and, randomly, 203 nonreactive specimens. For HHV-8 testing, the authors used two peptide-based enzyme-linked immunosorbent assays (EIAs), ORFK8.1 and ORF65, and an immunofluorescence assay (IFA). Specimens reactive in at least two assays or on IFA alone were considered HHV-8-seropositive. Prevalence estimates were weighted to account for the sampling scheme. Overall HHV-8 seroprevalence was 40%. HHV-8 seroprevalence was higher among HBsAg-positive donors (53%) than HBsAg-negative donors (39%; p =.02) and higher among HIV-positive donors (63%) than HIV-negative donors (39%; p <.001). HHV-8 seroreactivity showed no trend with age. Kappa values for assay concordances were 0.68 (ORFK8.1 EIA and IFA), 0.37 (ORF65 EIA and K8.1 EIA), and 0.29 (ORF65 EIA and IFA). The association between HHV-8 and HBsAg positivity and the lack of association between HHV-8 and age point to primarily nonsexual HHV-8 transmission during childhood. The association with HIV indicates sexual transmission may also occur. The role of ORF65 EIA in testing specimens from Africa warrants further evaluation.

Trends in AIDS incidence and survival among racial/ethnic minority men who have sex with men, United States, 1990-1999.

OBJECTIVES: We describe trends in AIDS incidence, survival, and deaths among racial/ethnic minority men who have sex with men (MSM). METHODS: We examined AIDS surveillance data for men diagnosed with AIDS from 1990 through 1999, survival trends from 1993 through 1997, and trends in AIDS incidence and deaths from 1996 to 1999, when highly active antiretroviral therapy (HAART) was introduced. RESULTS: The percentage of racial/ethnic minority MSM with AIDS increased from 33% of 26,930 men in 1990 to 54% of 17,162 men in 1999. From 1996 through 1998, declines in AIDS incidence were smallest among black MSM (25%, from 66.2 to 49.5 per 100,000) and Hispanic MSM (29%, from 39.3 to 27.8), compared with white MSM (41%, from 17.9 to 10.5). Declines in deaths of MSM with AIDS were also smallest among black MSM (53%, from 39.7 to 18.6 deaths per 100,000) and Hispanic MSM (61%, 21.6 to 8.4), compared with white MSM (63%, 12.3 to 4.5). Survival improved each year for all racial/ethnic groups but was poorest for black MSM in all years. CONCLUSIONS: Since the introduction of HAART, a combination of factors that include relatively higher infection rates in more recent years and differences in survival following AIDS diagnosis contribute to observed differences in trends in AIDS incidence and deaths among racial/ethnic minority MSM. Increased development of culturally sensitive HIV prevention services, and improved access to testing and care early in the course of disease are needed to further reduce HIV-related morbidity in racial/ethnic minority MSM.

HIV incidence in the United States, 1978-1999.

CONTEXT: HIV incidence measurements, which reflect recent or current transmission, are valuable for monitoring the epidemic and evaluating prevention programs. OBJECTIVES: To summarize HIV incidence patterns and trends in U.S. population groups. DATA SOURCES: Publications in English from 1980 through mid-2000. STUDY SELECTION AND STATISTICAL METHODS: We searched the literature for reports of HIV incidence in the United States. Locally weighted scatterplot smoothing was used to generate smooth curves to estimate trends in incidence. Spearman rank correlation was used to estimate the correlation coefficient between prevalence and incidence. DATA SYNTHESIS: In 74 eligible reports, HIV incidence varied widely (0.002-19.8 per 100 person-years [py]) depending on risk group. Among men who have sex with men (MSM), HIV incidence peaked in the early 1980s (5-20/100 py) and then declined but remained high during the 1990s (2-4/100 py). Among injection drug users (IDUs), incidence decreased since the mid-1980s but differed by geographic area; in the 1990s, incidence remained high in the East (1-3/100 py) but was lower in the West (<0.5/100 py). Throughout the late 1980s and 1990s, incidence was low and stable in broader populations (blood donors: <0.01/100 py; military personnel: 0.01-0.07/100 py). The correlation between HIV incidence and prevalence was strong in populations with a prevalence less than 1% (r = 0.94, p<.0001), moderate in populations with a prevalence from 1% to less than 10% (r = 0.57, p<.0001), and weak in populations with a prevalence at least 10% (r = 0.23, p=.09). CONCLUSIONS: HIV prevention in the United States should continue to focus on MSM and IDUs. HIV incidence measurements should be considered for monitoring HIV transmission in MSM, IDUs, and other populations in which seroprevalence is high.

Twenty-four month efficacy of a maternal short-course zidovudine regimen to prevent mother-to-child transmission of HIV-1 in West Africa.

OBJECTIVE: To assess the 24 month efficacy of a maternal short-course zidovudine regimen to prevent mother-to-child transmission (MTCT) of HIV-1 in a breastfeeding population in West Africa. METHODS: Data were pooled from two clinical trials: DITRAME-ANRS049a conducted in Abidjan, Côte d'Ivoire and Bobo-Dioulasso, Burkina-Faso and RETRO-CI, conducted in Abidjan. Between September 1995 and February 1998, consenting HIV-1-seropositive women were randomly assigned to receive zidovudine (300 mg) or placebo: one tablet twice daily from 36-38 weeks' gestation until delivery, then in DITRAME only, for 7 more days. Paediatric HIV-1 infection was defined as a positive HIV-1 polymerase chain reaction, or if aged > or =15 months, a positive HIV-1 serology. Cumulative risks (CR) of infection were estimated using a competing risk approach with weaning as a competing event. RESULTS: Among 662 live-born children, 641 had at least one HIV-1 test. All but 12 children were breastfed. At 24 months, overall CR of MTCT were 0.225 in the zidovudine and 0.302 in the placebo group, a 26% significant reduction. Among children born to women with CD4 cell counts < 500/ml at enrollment, CR of MTCT were similar, 0.396 in the zidovudine and 0.413 in the placebo group. Among children born to women with CD4 cell counts > or =500/ml, CR of MTCT were 0.091 in the zidovudine and 0.220 in the placebo group, a significant 59% reduction. CONCLUSION: A maternal short-course zidovudine regimen reduces MTCT of HIV-1 at age 24 months, despite prolonged breastfeeding. However, efficacy was observed only among women with CD4 cell counts > or =500/ml. New interventions should be considered to prevent MTCT, especially for African women with advanced HIV-1 immunodeficiency.