Correlation coefficient estimation involving a left censored laboratory assay variable.

When assessing a correlation between two exposure or biological marker variables, one sometimes encounters the problem of indeterminate values for one of the variables due to an assay detection limit. In this event, investigators often report correlation coefficients computed after removing the pairs involving non-detectable values, or after substituting some small constant for those values. These ad hoc practices can lead to bias in both point and confidence interval estimates of the true correlation coefficient. To address this issue, we consider two parametric techniques for estimating the correlation in the presence of left censoring for one of the variables. The first is a maximum likelihood approach, and the second is an adaptation of multiple imputation motivated primarily by potential benefits in confidence interval coverage. Both of the estimators studied reduce to the standard Pearson's correlation coefficient in the event of no censoring, and hence are valid in cases where this measure would be appropriate for the complete data. We assess these approaches empirically and contrast them with ad hoc methods for estimating the correlation between cervicovaginal human immunodeficiency virus (HIV) viral load measurements and CD4+ lymphocyte counts from HIV positive women enrolled in a clinical trial conducted in Bangkok, Thailand.

Safety of late in utero exposure to zidovudine in infants born to human immunodeficiency virus-infected mothers: Bangkok. Bangkok Collaborative Perinatal HIV Transmission Study Group.

BACKGROUND: Short-course zidovudine (ZDV) given in the late antenatal period can reduce mother-infant human immunodeficiency virus (HIV) transmission by one half. Because this intervention is being implemented in developing countries, evidence of its safety is needed. METHODS: In a randomized, double-blinded, placebo-controlled trial in Bangkok, HIV-infected pregnant women received either ZDV (300 mg twice daily from 36 weeks' gestation until labor, then every 3 hours until delivery) or an identical placebo regimen. Infants were evaluated at birth and at 1, 2, 4, 6, 9, 12, 15, and 18 months of age. Growth, clinical events, and hematologic and immunologic measurements were compared between treatment groups. RESULTS: Of the 395 children born (196 in ZDV group and 199 in placebo group), 330 were uninfected, 55 were infected, and 10 had indeterminate infection status. Overall, 319 children (81%) completed 18 months of follow-up, and 14 (4%) died before 18 months of age. Among uninfected children, the mean hematocrit was lower in the ZDV group at birth (49.1% vs 51.5%) but not at later ages; mean weight, height, head circumference, and CD4(+) and CD8(+) T lymphocyte counts were similar in both groups at all ages. Five uninfected children in the ZDV group but only one in the placebo group had a febrile convulsion. No other signs suggestive of mitochondrial dysfunction and no tumors were observed. Among infected children, an estimated 62% in the ZDV group and 77% in the placebo group survived free of Centers for Disease Control and Prevention class C disease during the 18-month follow-up. CONCLUSIONS: No significant adverse events were associated with short-course ZDV during 18 months of follow-up in this population.

Correlating two viral load assays with known detection limits.

A timely objective common to many HIV studies involves assessing the correlation between two different measures of viral load obtained from each of a sample of patients. This correlation has scientific utility in a number of contexts, including those aimed at a comparison of competing assays for quantifying virus and those aimed at determining the level of association between viral loads in two different reservoirs using the same assay. A complication for the analyst seeking valid point and interval estimates of such a correlation is the fact that both variables may be subject to left censoring due to values below assay detection limits. We address this problem using a bivariate normal likelihood that accounts for left censoring of two variables that may have different detection limits. We provide simulation results to evaluate sampling properties of the resulting correlation estimator and compare it with ad hoc estimators in the presence of nondetects. In an effort to obtain improved confidence interval properties relative to the Wald approach, we evaluate and compare profile likelihood-based intervals. We apply the methods to HIV viral load data on women and infants from a trial in Bangkok, Thailand, and we discuss an extension of the original model to accommodate interval censoring arising due to the study design.

Early diagnosis of HIV-1-infected infants in Thailand using RNA and DNA PCR assays sensitive to non-B subtypes.

OBJECTIVES: To evaluate the sensitivity and specificity of RNA and DNA polymerase chain reaction (PCR) for early diagnosis of perinatal HIV-1 infection and to investigate early viral dynamics in infected infants. DESIGN: A cohort study of 395 non-breastfed infants born to HIV-infected mothers in a randomized clinical trial of short-course antenatal zidovudine. METHODS: Infant venous blood specimens collected at birth, 2 months, and 6 months of age were tested by qualitative DNA and quantitative RNA PCR (Roche Amplicor). To determine sensitivity and specificity of DNA and RNA PCR, results were compared with later DNA PCR results and to antibody results at 18 months. The HIV-1 subtype of the mother's infection was determined by peptide serotyping. RESULTS: In the study, 92% of mothers were infected with subtype E. DNA PCR sensitivity was 38% (20 of 53) at birth, and 100% at 2 months (53 of 53) and 6 months (47 of 47). RNA PCR sensitivity was 47% (25 of 53) at birth and 100% (53 of 53) at 2 months. All samples that tested DNA-positive tested RNA-positive. Specificity was 100% for both DNA and RNA testing at all timepoints. For infected infants, the median viral load of RNA-positive specimens was 407,000 copies/ml (5.6 log10) at birth, 3, 700,000 copies/ml (6.6 log10) at 2 months, and 1,700,000 copies/ml (6.2 log10) at 6 months. Infant RNA levels at 2 and 6 months did not differ by maternal zidovudine exposure, or RNA level at birth. CONCLUSION: This RNA PCR assay performed well for diagnosing perinatal HIV subtype E infection, detecting nearly half of infected infants at birth, and 100% at 2 and 6 months, with 100% specificity. Infected infant viral RNA levels were very high at 2 and 6 months, and were unaffected by maternal zidovudine treatment.

Evaluating a new strategy for prophylaxis to prevent Pneumocystis carinii pneumonia in HIV-exposed infants in Thailand. Bangkok Collaborative Perinatal HIV Transmission Study Group.

OBJECTIVE: To evaluate a strategy for prophylaxis against Pneumocystis carinii pneumonia (PCP) for infants in Thailand. METHODS: HIV-infected women were offered trimethoprim-sulfamethoxazole for PCP prophylaxis for their children at 1-2 months of age. When the children reached 6 months of age, investigators simulated a decision to continue or stop prophylaxis on the basis of clinical criteria, and compared their decisions with results of polymerase chain reaction (PCR) testing for HIV. We calculated the proportions of children who received and completed prophylaxis, and compared the rates of pneumonia and death from pneumonia with rates from an earlier prospective cohort. RESULTS: Of 395 eligible infants, 383 (97%) started prophylaxis. By 6 months of age, 10 (2.6%) were lost to follow-up, three (0.8%) were non-adherent, seven (2%) had stopped because of adverse events, four (1%) had died, and 359 (94%) still received prophylaxis. At 6 months of age, 30 (70%) of 43 HIV-infected children and 16 (5%) of 316 uninfected children met the clinical criteria to continue prophylaxis. The incidence of pneumonia at 1 to 6 months of age was 22% (15/68) in the earlier cohort, and 13% (6/46) in the recent cohort [relative risk (RR) 0.6, 95% confidence interval (CI) 0.3-1.4; P= 0.22]; mortality rates were 9% and 4%, respectively (RR 0.5; 95% CI 0.1-2.3; P = 0.47). CONCLUSION: This PCP prophylaxis strategy appeared to be acceptable and safe, may have reduced morbidity and mortality from pneumonia, and should be considered in developing countries where early laboratory diagnosis of perinatal HIV infection is unavailable.

Lack of association between human immunodeficiency virus type 1 antibody in cervicovaginal lavage fluid and plasma and perinatal transmission, in Thailand.

To determine the association between human immunodeficiency virus type 1 (HIV)-specific antibody and RNA levels in cervicovaginal lavage (CVL) samples and plasma, zidovudine treatment, and perinatal transmission, HIV subtype E gp160-specific IgG and IgA were serially measured in a subset of 74 HIV-infected women in a placebo-controlled trial of zidovudine, beginning at 36 weeks of gestation. HIV IgG was detected in 100% of plasma and 97% of CVL samples; HIV IgA was consistently detected in 62% of plasma and 31% of CVL samples. Antibody titers in CVL samples correlated better with the RNA level in CVL samples than with plasma antibody titers. Zidovudine did not affect antibody titers. Perinatal HIV transmission was not associated with antibody in CVL samples or plasma. HIV-specific antibody is present in the cervicovaginal canal of HIV-infected pregnant women; its correlation with the RNA level in CVL fluid suggests local antibody production. However, there was no evidence that these antibodies protected against perinatal HIV transmission.

Short-course antenatal zidovudine reduces both cervicovaginal human immunodeficiency virus type 1 RNA levels and risk of perinatal transmission. Bangkok Collaborative Perinatal HIV Transmission Study Group.

Human immunodeficiency virus (HIV) levels in cervicovaginal lavage (CVL) and plasma samples were evaluated in relation to perinatal transmission in a randomized placebo-controlled trial of brief antenatal zidovudine treatment. Samples were collected at 38 weeks' gestation from 310 women and more frequently from a subset of 74 women. At 38 weeks, after a 2-week treatment period, CVL HIV-1 was quantifiable in 23% and 52% of samples in the zidovudine and placebo groups, respectively (P<.001). The perinatal transmission rate was 28.7% among women with quantifiable CVL HIV-1 and high plasma virus levels (>10,000 copies/mL) and 1% among women without quantifiable CVL HIV-1 and with low plasma virus levels (P<.001). A 1-log increase in plasma HIV-1 increased the transmission odds 1.8 and 6.1 times (95% confidence interval, 0.9-3.5 vs. 2.4-15.4) for women with and without quantifiable CVL HIV-1, respectively (P=.03). CVL HIV-1 is an independent risk factor for perinatal HIV-1 transmission.

Short-course zidovudine for perinatal HIV-1 transmission in Bangkok, Thailand: a randomised controlled trial. Bangkok Collaborative Perinatal HIV Transmission Study Group.

BACKGROUND: Many developing countries have not implemented the AIDS Clinical Trials Group 076 zidovudine regimen for prevention of perinatal HIV-1 transmission because of its complexity and cost. We investigated the safety and efficacy of short-course oral zidovudine administered during late pregnancy and labour. METHODS: In a randomised, double-blind, placebo-controlled trial, HIV-1-infected pregnant women at two Bangkok hospitals were randomly assigned placebo or one zidovudine 300 mg tablet twice daily from 36 weeks' gestation and every 3 h from onset of labour until delivery. Mothers were given infant formula and asked not to breastfeed. The main endpoint was babies' HIV-1-infection status, tested with HIV-1-DNA PCR at birth, 2 months, and 6 months. We measured maternal plasma viral concentrations by RNA PCR. FINDINGS: Between May, 1996, and December, 1997, 397 women were randomised; 393 gave birth to 395 live-born babies. Median duration of antenatal treatment was 25 days, and median number of doses during labour was three. 99% of women took at least 90% of scheduled antenatal doses. Adverse events were similar in the study groups. Of 392 babies with at least one PCR test, 55 tested positive: 18 in the zidovudine group and 37 in the placebo group. The estimated transmission risks were 9.4% (95% CI 5.2-13.5) on zidovudine and 18.9% (13.2-24.2) on placebo (p=0.006; efficacy 50.1% [15.4-70.6]). Between enrolment and delivery, women in the zidovudine group had a mean decrease in viral load of 0.56 log. About 80% of the treatment effect was explained by lowered maternal viral concentrations at delivery. INTERPRETATION: A short course of twice-daily oral zidovudine was safe and well tolerated and, in the absence of breastfeeding, can lessen the risk for mother-to-child HIV-1 transmission by half. This regimen could prevent many HIV-1 infections during late pregnancy and labour in less-developed countries unable to implement the full 076 regimen.

Chlamydial and gonococcal cervicitis in HIV-seropositive and HIV-seronegative pregnant women in Bangkok: prevalence, risk factors, and relation to perinatal HIV transmission.

OBJECTIVES: To determine the prevalence and risk factors associated with cervicitis caused by Chlamydia trachomatis and Neisseria gonorrhoeae in human immunodeficiency virus (HIV) type 1-seropositive and HIV-seronegative pregnant women in Bangkok, and the relation to perinatal HIV transmission. METHODS: As part of a multicenter perinatal HIV transmission study in an antenatal population with 2% HIV seroprevalence, endocervical swabs obtained at mid-pregnancy from a consecutive sample of 222 HIV-seropositive and 219 HIV-seronegative pregnant women at two large hospitals in Bangkok were tested for the presence of C. trachomatis and N. gonorrhoeae by DNA hybridization probe (Gen-Probe). Clinical risk factors and DNA probe results were analyzed in relation to the women's and newborns' HIV infection status. RESULTS: The prevalence of C. trachomatis was 16.2% in HIV-seropositive pregnant women and 9.1% in HIV-seronegative pregnant women (P = 0.03). The prevalence of N. gonorrhoeae was 2.7% in HIV-seropositive pregnant women and 1.4% in HIV-seronegative pregnant women (P = 0.5). The overall population prevalence estimate was 9.2% for C. trachomatis and 1.4% for N. gonorrhoeae. Women with gonococcal infection were more likely to be positive for C. trachomatis (RR(MH) = 5.2, P < 0.01). Young age (<21 years) and primigravid status were associated with C. trachomatis infection among HIV-seropositive women; history of multiple sex partners (>1) were associated with C. trachomatis infection among HIV-seronegative women. For HIV-seropositive women, primigravida status also was associated with C. trachomatis infection. The perinatal HIV transmission rates were similar for those with and without C. trachomatis (24.1% and 23.2%, P = 0.9) and among those with and without N. gonorrhoeae (20% and 23.5%, P = 1.0). CONCLUSIONS: Among pregnant women in Bangkok, C. trachomatis infection was considerably more common than N. gonorrhoeae infection and was associated with HIV infection, young age and first pregnancy (HIV-seropositive women), and multiple partners (HIV-seronegative women). Our data do not suggest an association between perinatal HIV transmission and maternal C. trachomatis or N. gonorrhoeae infection identified and treated during pregnancy. The high prevalence of C. trachomatis found using a test not readily available in Thailand emphasizes the need for improved, inexpensive ways to screen for and diagnose these sexually transmitted infections in developing countries.

PIP: Numerous studies have suggested that Chlamydia trachomatis and Neisseria gonorrhoeae facilitate heterosexual HIV transmission; the impact of these sexually transmitted diseases (STDs) on perinatal HIV transmission is unknown, however, due to the expense of routine screening for STDs during pregnancy in developing countries. As part of a multicenter perinatal HIV transmission study, 222 HIV-positive and 219 HIV-negative women presenting for prenatal care at 2 hospitals in Bangkok, Thailand, during 1993-94 were enrolled. At mid-pregnancy, endocervical swabs were obtained and tested for the presence of C. trachomatis and N. gonorrhoeae by DNA hybridization probe. There were 36 cases (16.2%) of C. trachomatis infection among HIV-positive women and 20 cases (9.1%) among HIV-negative women. There were 6 cases (2.7%) of N. gonorrhoeae among HIV-positive women and 3 cases (1.4%) among HIV-negative women. Based on an estimated antenatal HIV seroprevalence of 2%, these findings imply a general antenatal prevalence of 9.2% for C. trachomatis and 1.4% for N. gonorrhoeae. Women with gonococcal infection were more likely (relative risk, 5.2) to be positive for C. trachomatis as well. C. trachomatis infection among HIV-infected pregnant women was associated with age under 21 years and primigravidity. The overall perinatal HIV transmission rate was 24.2%, with no significant difference according to STD infection status. However, since all women diagnosed with STDs received treatment by the mid-third trimester of pregnancy, it remains possible that untreated STDs facilitate perinatal HIV transmission. The high prevalence of C. trachomatis detected in this study through use of a test not readily available in Thailand emphasizes the need for inexpensive, reliable methods to screen for STDs among pregnant women in developing countries.