Treatment with zidovudine (ZDV) has become routine for most HIV-infected pregnant women to prevent vertical transmission of HIV-1 (1). The Pediatric AIDS Clinical Trial Group Protocol (PACTG) 076 data provided convincing evidence of the effectiveness of ZDV in reducing risk of vertical transmission (2). Since the announcement of this trial's findings, use of ZDV in pregnancy has increased markedly (3). The U.S. Food and Drug Administration (FDA) continues to place ZDV therapy in pregnancy in risk class C (4), indicating that the drug may have teratogenic or embryocidal effects but that available data are insufficient. Here, we present a population-based assessment of prenatal ZDV exposure and congenital anomaly risk.
METHODS
Study Population
The study population was ascertained from 3037 liveborn deliveries between January 1, 1993 and September 31, 1996 by HIV-infected women enrolled in New York State (NYS) Medicaid (a federally mandated medical assistance program for the indigent). Maternal HIV status was determined using a Medicaid claims-based algorithm found to be 97% sensitive and 93% specific in a previous validation substudy (3). Medicaid claims provided data on diagnoses (up to five per inpatient and two per outpatient claim), procedures, payment rates, and prescription drugs. Longitudinal claims files were created for all mothers for up to 3 years before and 1 year after delivery. For the children, longitudinal claims files were created from delivery through 2 years of age. Vital statistics offered demographic, education, parity, gestational age at delivery, birth weight, and self-reported substance use during pregnancy data.
Only the most recent delivery was included in analyses for the 278 women (9.2%) with more than one delivery in the study period. For the 48 twin births in the study period, 1 of 2 cotwins was randomly excluded. Without these minor exclusions, substantially more complex statistical methods would have been needed. Of the 2759 remaining after these exclusions, 104 (8.2%) were excluded because we could not link them to vital statistics. Next, 723 children (27.2%) were excluded because we found no record of any service billed to Medicaid for the child's care. This left 1932 observations in the study population. One possible reason for the complete absence of Medicaid data on the child is the inability to link mother and child Medicaid identifiers. However, at over 70%, our rate of matching mother to child in Medicaid files is near the upper end of the range expected (40%-80%) based on studies in other states (5). Other possible contributing factors to the complete absence of childrens' claims are neonatal death and immediate foster-care placement.
For analyses comparing cohort anomaly rates with population rates from the NYS Congenital Malformations Registry (CMR), 15 cases with missing race data were excluded. For within-cohort comparisons, we excluded individuals (n = 56) with missing data on any study variable (the full set is described later in this paper). In addition, 26 people whose race was coded as something other than white, black, or Hispanic were also dropped. This left 1891 observations in what we refer to as the analytic cohort.
Study Variables
Major congenital malformations were ascertained from the International Classification of Diseases, 9th revision, Clinical Modification (ICD-9-CM) codes on Medicaid inpatient, clinic, and physician claims. Our first major anomaly definition was designed for comparisons with the CMR. Because CMR reporting is hospital based and includes children through age 2 years, only inpatient claims for services provided up to two years of age were used in case-finding. The presence of an ICD-9-CM code for any condition considered by the CMR to be a major malformation qualified a child as a case. Most conditions within the congenital anomalies ICD-9 code range 740 through 759 are considered major malformations (examples include 741.9, spina bifida without hydrocephaly, 746.6, congenital mitral valve insufficiency, and 754.5, varus deformities of feet). However, several codes within this range are for minor congenital anomalies and, consequently, do not by themselves define a case (examples include 744.1, accessory auricle, 747.5, single umbilical artery, and 755.0, polydactyly). In addition, there are some conditions reported to the CMR outside this range that are considered major anomalies by the registry but that are not plausibly associated with ZDV exposure (examples include 760.71, fetal alcohol syndrome and 090, congenital syphilis). For the sake of comparability, these were retained in the case's first definition. A listing of all specific codes included under the first definition can be obtained from the corresponding author.
We used a second case definition approach for the within-cohort analyses. Clinic and physician claims from the first 2 years of life, as well as hospital claims contributed to case finding. For this definition only the major malformation codes within the standard range for congenital anomalies (ICD-9-CM codes 740-759) were considered. In addition to excluding the minor malformation codes listed by the CMR, additional codes from within this range that were believed to increase the likelihood of false positives were dropped from the case definition. The excluded codes were mostly for unspecified anomalies of various organ systems (e.g., 746.9 unspecified anomaly of heart) and subsets of codes for other specified anomalies of various organ systems (e.g., 753.8, other specified anomalies of male genital organs). Guidance on exclusions came from the previous experience of the New York State CMR (Charlotte Druschel, personal communication). The codes included in this second definition can also be obtained from the corresponding author.
A final modification to the second major anomaly definition also was intended to reduce the potential for false positive events. When an eligible anomaly code appeared only on a given child's outpatient claims, that child's full claims history was reviewed by this study's teratologist, Anderson, who was blinded to the mothers' ZDV status. Of 56 claims histories reviewed, 18 anomaly cases initially identified were reclassified as nonevents because it was likely they were acquired after birth.
Although presence of any major anomaly was our principal outcome, we also examined anomalies by the following organ system categories: central nervous and cardiovascular systems, oral clefts, respiratory, digestive, and genitourinary systems. Children with multiple anomalies were allowed to contribute to more than one major malformation category. Because, as we later discuss, there were indications of excess in the cardiovascular and central nervous system categories, we scrutinized further the codes included under these and explored the impact of some additional code exclusions.
Receipt of ZDV was identified as any Medicaid pharmacy claim for filled prescription(s). Exposure was also categorized by trimester of first prescription. Trimester cutpoints were based on back-calculation from gestational age reported on birth certificates (gestational age came from physicians' estimates in 96% of the deliveries and mothers' self-reported last menstrual period in the remainder). When prepregnancy claims were available, we considered any carry-over days from ZDV prescriptions before the first trimester in our assessment of exposure.
Mother's race was coded as white, black, or other, corresponding to the categories used by the CMR, for comparisons with statewide prevalence data. In internal comparisons, we coded race as black, white, or Hispanic. Categories specified for analysis of this and other variables are shown in Table 1. A separate category of missing education data was created given that >5% (i.e., 6.7%) of the sample. Any illicit drug use during pregnancy was identified using a validated approach (6) that searches claims during pregnancy for diagnoses and procedures related to opioid, cocaine, amphetamine, hallucinogen, other psychostimulant, or unspecified illicit drug abuse as well as considering any maternal substance abuse codes in vital statistics records. Length of follow-up was based on the time from delivery to the last available claim before the child's third birthday. A dichotomous indicator of the length of follow-up (<1 versus ≥1 years) was also included.
Analysis Methods
For the comparison against the major congenital anomaly rate in NYS, a standardized morbidity ratio (SMR) was calculated comparing observed with expected case counts adjusting for race, gender, and place of residence. We used CMR data for 1994, roughly the calendar time midpoint of study deliveries. Standard errors for SMRs were estimated using the variance approximation described by Kahn and Sempos (7).
To evaluate the effect of ZDV on major anomalies within the study cohort, odds ratios (ORs) were estimated from multivariable logistic regression models adjusting for all available maternal demographic and clinical characteristics. We explored interactions between the ZDV exposure variable and each covariate in a series of separate adjusted models but intereactions were either not evaluable because of sparse data or did not change our main conclusions. Therefore, interactions were not included. Working from the models including all covariates and no interactions, we dropped individual covariates if their removal from fully adjusted models did not change any estimated ZDV adjusted ORs by ≥3%.
We repeated analyses of ZDV use and major anomaly risk in four subsets created by applying different exclusion criteria to the study cohort.
Subset 1: We excluded deliveries to mothers receiving care during the prenatal period at sites participating in clinical trials of prenatal and perinatal ZDV (n = 505) as well as women with any claim(s) for non-ZDV antiretrovirals (n = 30).
Subset 2: We excluded mothers who were not Medicaid eligible throughout their entire pregnancy (n = 693).
Subset 3: We excluded deliveries in which, although there was a claim for the child at some point in her or his first 2 years of life, there was no child's claim at delivery.
Fully restricted subset: We excluded deliveries meeting any of the previously noted criteria.
For comparisons of the ORs of anomalies in specific major malformation categories in which sample-size in given categories is limited, we estimated only unadjusted ORs (based on the full study cohort) and present findings only for the four anomaly categories with more than ten events. These categories comprised four anatomic systems: central nervous, cardiovascular, gastrointestinal, and musculoskeletal. Standard errors and confidence limits for these ORs were calculated by the exact method of Mehta et al. (8).
RESULTS
Table 1 shows distributions of characteristics in the full sample, the analytic cohort without any missing values, and the fully-restricted subset. As expected, the prevalence of major anomalies was consistently higher based on definition 1 (broader surveillance definition) than definition 2 (more restrictive definition from within cohort analyses). Receipt of ZDV was similar in the full population and in the analytic cohort. However, in the fully restricted subset, the prevalence of ZDV was nearly 40% compared with 30% in the full population and analytic cohorts. Distributions of other characteristics across the cohorts are generally comparable. The fully restricted subset included somewhat fewer younger and more older mothers.
The prevalence of major anomalies in the full cohort based on definition 1 was significantly higher than that observed in the general New York State population. As shown in Table 2, the SMR adjusted for race, gender, and location suggests that the risk of a major anomaly in the study cohort was 2.79 times greater than the general population (95% confidence interval [CI], 2.36-3.17). In all strata defined by race, gender, and location, the number of deliveries observed with major anomalies exceeds the number expected except for the small group of black, non-New York City, male babies.
Table 3 displays the unadjusted and adjusted ORs for any major anomaly associated with any ZDV exposure and with exposure by trimester of first use, using no ZDV exposure as the reference group. Principal findings remained consistent across cohorts.
First, there are moderately increased odds (adjusted ORs between 1.5 and 2.0) of any major anomaly associated with any ZDV exposure. However, the CI on these estimates is also consistent with much weaker associations and, in two instances, includes 1.0. In analyses of ZDV exposure stratified by trimester of first prescription, the ZDV effect is consistently strongest for the third and weakest for the first. CIs on first trimester effects are also wide and included estimates on either side of the null. Third trimester adjusted ORs are near 2.0 (1.84 in the analytic cohort and 2.13 in the fully restricted cohort) but have broad CIs the lower limits of which included, or approached, 1.0.
Table 4 presents the crude ORs for any occurrence of each of the four categories of major anomalies with more than ten occurrences. Crude ORs for any ZDV exposure are elevated for both cardiovascular and central nervous system anomalies (2.24 CI, 1.19-4.21; 4.35 CI, 1.30-16.66, respectively). For central nervous system anomalies, first trimester exposure had the highest OR, at 7.98, but, with only nine occurrences with initial exposure here, the estimate is imprecise. Of the 27 cases involving anomalies of the central nervous system, 10 were for diagnosis of microcephalus (ICD-9-CM code 742.1). Because the frequency of this code and the possibility of error in this diagnosis, we reestimated the OR for central nervous system anomalies excluding the microcephalus ICD-9-CM code from the case definition and still found a significantly greater (p = .007) proportion of births meeting the revised case definition in ZDV exposed than in unexposed. With respect to cardiovascular disease anomalies, 4 of 57 claims were for septal closure anomaly NOS [not otherwise specified] (ICD-9-CM code 745.9). Although this was one of the point nine codes retained in the original case definition, we reestimated the cardiovascular anomaly ORs excluding this code from the case-finding algorithm. As a result, the OR for any AZT exposure moved down to 2.09 (95% CI, 1.16-3.80) but the pattern of risk estimates by trimester of first exposure was unchanged and did not show an increased risk in the first trimester.
DISCUSSION
The protection against vertical transmission conferred by maternal ZDV use is indisputable (2) but the lack of data on potential adverse effects of this therapy is still a concern. The population-based cohort of babies delivered by HIV-infected mothers we studied here had higher rates of major congenital anomalies than did all babies delivered in NYS. The reasons for this might include higher prevalence of other anomaly risk factors among mothers with HIV, effects of HIV infection itself, or, perhaps, ZDV use. To investigate this third hypothesis, we compared anomaly rates of subgroups defined by ZDV exposure history within the cohort of HIV-infected mothers. Babies whose mothers had ZDV exposure during pregnancy had a greater incidence of major malformations than those whose mothers did not. Even so, the risk was not significantly higher for deliveries to mothers with first trimester exposure. This trimester is the period of greatest susceptibility to teratogens because it is when germ layer formation and organogenesis occur (9). When we analyzed the risk of specific categories of anomalies by organ system, we did find an increased unadjusted risk of central nervous system anomalies in the first trimester. However, given that we had only four such events, this association must be interpreted cautiously.
The few published animal studies on the potential teratogenic effects of ZDV have not reported an increase in prevalence of congenital anomalies (10-13). One study in mice did find reduced litter size and greater fetal resorption with maternal ZDV exposure 6 weeks before mating (14). ZDV is, however, known to pass transplacentally (15,16) to the fetus and has been shown to be genotoxic to both mouse (17) and human fetuses (18).
The latest published epidemiologic data from the PACTG Protocol 076 trial cohort showed no increase in major structural anomalies between groups exposed to ZDV and placebo (19). However, this study was designed with the intention of avoiding first trimester exposure. Further, the trial's sample size (n = 211 and n = 213 in the ZDV and placebo arms, respectively) was probably underpowered to detect clinically meaningful differences in prevalence of the uncommon outcome of birth defects. The only other case series of ZDV-exposed pregnancies reported in the literature is even smaller than the PACTG 076 exposed cohort, consisting of 104 deliveries (88 livebirths) with only 8 births with anomalies observed (20).
The Antiretroviral Pregnancy Registry (APR), originally established in 1989 as the Zidovudine in Pregnancy Registry, was begun by the pharmaceutical industry, in conjunction with the U.S. Centers for Disease Control and Prevention, as a means of gathering sufficient data to measure incidence of structural anomalies in antiretroviral-exposed pregnancies. Despite its being an international registry, this effort had prospectively registered and closed only 392 ZDV monotherapy-exposed pregnancies by the end of 1997. Further, 20% of these were closed due to loss-to-follow-up, leaving only 317 pregnancies with data available for analyses (21). Published reports using registry data have not shown a statistically significant increase in the prevalence of major congenital anomalies among children of prospectively-followed mothers exposed to ZDV in any trimester compared to an external standard derived from the Metropolitan Atlanta Birth Defects Registry (22-24).
The major strength of our data on congenital anomaly experience is that they came from a comparably large and relatively unselected cohort. However, as has been pointed out previously (5,25,26), administrative data have several limitations. In a validation study of central nervous system birth defects in Medicaid data, Grisso et al. (5) found serious overreporting of central nervous system anomalies resulting from miscoding of the cesarean section procedure code, 74.1, as ICD-9 diagnosis code 741, that for spina bifida. Even after excluding deliveries with likely miscoding for cesarean section, these authors still reported poor (30%) positive predictive value for Medicaid claims-based CNS anomaly detection. Although the findings of Grisso et al. raise concerns about using Medicaid data to study CNS malformation, in our data, there was only one birth with a 741 code and that child had several other claims with codes indicating a major anomaly of the central nervous system. Further, concern over poor positive predictive value should be mitigated considering that when the prevalence of an outcome is ≥5%, as is the situation with major anomalies, poor positive predictive values (e.g., 50%) can be observed even in the face of high (e.g., 95%) sensitivities and specificities. As long as the sensitivity and specificity of anomaly case-finding is high, any information bias manifested through nondifferential misclassification is likely to be small.
Information biases may still, however, introduce differential misclassification. Two strategies were employed to guard against this. However, neither offers complete protection against this bias. First, analyses were repeated on subsets of the analytic cohort that we believed were less susceptible to misclassification of exposure and outcome data. For example, Medicaid claims data would not have captured ZDV prescribed at sites participating in clinical trials to prevent vertical transmission. Women in these trials may also possess fewer risk factors for deliveries with congenital anomalies because study participants tend to have better health status in general. Misclassification of trial participants as unexposed in the Medicaid data could artifactually lower the anomaly risk estimate in the reference group and consequently lead to an overestimate of the ZDV effect. Consequently, one of our data subsets excluded women known to receive services at any site participating in these clinical trials. Principal findings continued to obtain in this and all other subsets, including the fully restricted one.
The second approach undertaken to guard against information bias was to focus on major anomalies as the outcome of interest and to exclude from the case definition potentially overreported ICD-9-CM codes. We suspected that nonspecific and minor anomalies might be more likely to be reported in Medicaid claims for ZDV-exposed cases who likely were receiving closer medical follow-up. Consequently, the case definition used protects against this form of surveillance bias.
At this point, the totality of epidemiologic information on anomalies following prenatal ZDV exposure is inconclusive. Our data add to limited data available from clinical trials, case series, and registry populations. Close follow-up of clinical trial populations found no increased risk but effect estimates are imprecise. We see an overall excess of anomalies in women with ZDV exposure but, reassuringly, no overall increased risk with first trimester exposure. In fact, risk appeared highest in the third trimester. The Antiretroviral Pregnancy Registry data also show a similar pattern but thus far includes only a small number of deliveries with anomalies (n = 10, as of December 31, 1997) (21).
In gauging the impact of potential adverse effects of ZDV, it is essential to compare them with HIV infection prevented. If we assume that the increased relative odds of major congenital anomalies associated with any ZDV seen in this investigation (1.55) is true, based on the 1994 statewide congenital anomaly rate of 3.4% and the ZDV efficacy estimates from the 076 trial (reduction in vertical transmission risk from 23.5 to 8.3), for every additional major anomaly that could be attributable to ZDV use, approximately 10 vertically transmitted HIV infections would be prevented. Given this 10-fold difference plus the certainty with which the ZDV benefit in preventing vertical transmission compares against the uncertainty regarding teratogenic effects, there seems little doubt current guidelines regarding ZDV use in pregnancy represent best public health practice.
However, as management of pediatric HIV continues to improve, the implicit weights used to compare a ZDV attributable major anomaly against multiple prevented cases of vertically transmitted HIV may change. Questions of the tradeoffs between possible adverse consequences of prenatal antiretroviral therapy and prevention of vertical transmission are potentially may become more complicated and more frequent. Although some early data indicate that prenatal ZDV use limited to late pregnancy or intrapartum periods may prove an effective alternative to earlier prenatal exposure (27-30), it will be a matter of some time before studies are replicated and practice guidelines changed. Recognizing this and the fact that increasing numbers of women are receiving combination antiretroviral therapy in pregnancy, the National Institutes of Health sponsored a workshop on the detection of potential adverse effects following perinatal exposure to antiretrovirals (31). Recommendations included expanding model programs linking data from statewide congenital malformations registries to databases from HIV surveillance systems in states with mandatory HIV reporting. Although this is a complex task, involving intricate data linkage and confidentiality protection issues, it may add an important dimension to our understanding of this possible consequence of ZDV (and other antiretroviral) use during pregnancy. Of course, anything learned about adverse consequences of in utero exposure to ZDV must be balanced against the considerable morbidity avoided and the lives saved by the prevention of vertical transmission of HIV.
Acknowledgments:
This research was supported by grant R01 DA07904-05 from the National Institute on Drug Abuse grant to B. J. Turner. We would like to thank Charlotte M. Druschel, Medical Director of the New York State Congenital Malformations Registry, for her valuable assistance.
REFERENCES
1. Centers for Disease Control. Public health task force recommendations for the use of antiretroviral drugs in pregnant women infected with HIV-1 for maternal health and for reducing perinatal HIV-1 transmission in the United States. MMWR Morb Mortal Wkly Rep 1998;47:1-30.
2. Conner EM, Sperling RS, Gelber R, et al. Reduction of maternal-infant transmission of human immunodeficiency virus type 1 with zidovudine treatment. N Engl J Med 1994 331:1173-80.
3. Turner BJ, Newschaffer CJ, Zhang D, Fanning T, Hauck WW. Translating clinical trial results into practice: the effect of an AIDS clinical trial on prescribed antiretroviral therapy for HIV-infected pregnant women. Ann Intern Med 1999;130:979-86.
4. Physicians desk reference, 53rd ed. Oradell, NJ: Medical Economics Company, 1999:1106.
5. Grisso JA, Carson JL, Feldman HI, Cosmatos I, Shaw M, Strom B. Epidemiological pitfalls using Medicaid data in reproductive health research. J Maternal Fetal Med 1997; 6:230-6.
6. Fanning TR, Turner BJ, Cosler LE, et al. Quality of Medicaid data for HIV/AIDS research: examination of a statewide database. AIDS Publ Pol J 1995; 10:39-47.
7. Kahn HA, Sempos CT. Statistical methods in epidemiology. New York: Oxford University Press, 1989.
8. Mehta CR, Patel NR, Gray R. Computing an exact confidence interval for the common odds ratio in several 2 × 2 contingency tables. J Am Stat Assoc 1985; 78:969-73.
9. Levi PE. Principles and mechanisms of teratogenesis. In: Kolb VM, ed. Teratogens, 2nd completely revised ed. Amsterdam, The Netherlands: Elsevier Science, 1993:1-19.
10. Ayers K. Preclinical toxicity of zidovudine. Am J Med 1988; 85:186-8.
11. Greene JA, Ayers KM, deMiranda P, Tucker Jr. WE Postnatal survival of wistar rats following oral dosage with zidovudine on gestational day 10. Fundam Appl Toxicol 1990; 15:201-6.
12. Ha JC, Nosbisch C, Conrad SH, Ruppenthal GC, Sacket GP, Abkowitz J, Unadkat JD. Fetal and infant toxicity of zidovudine (azidothymidine) in Macaca nemestrina: preliminary observations. J Aquir Immune Defic Syndr 1994; 7:154-7.
13. Ha JC, Nosbisch C, Abkowitz JL, et al. Fetal, infant and maternal toxicity of zidovudine (azidothymidine) administered through pregnancy in Macaca nemestrina. J Aquir Immune Defic Syndr Hum Retrovirol 1998; 18:27-38.
14. Toltzis P, Marx CM, Kleinman N, Levine EM, Schmidt EV. zidovudine-associated embryonic toxicity in mice. J Infect Dis 1991;163:1212-18.
15. Bawdon RE, Soghi S, Dax J. The transfer of anti-human immunodeficiency virus nucleoside compounds by the term human placenta. Am J Obstet Gynecol 1992; 167:1570-4.
16. Olivero OA, Pirila R, Vahakangas K, Pirer MC. Placental transfer and incorporation of AZT into placental DNA in a human placental perfusion model. Proc Am Assoc Cancer Res 1998; 39:631.
17. Olivero OA, Anderson LM, Diwan BA, et al. Transplacental effects of 3`-azido-2`,3`-dideoxythymidine (AZT): tumorigenicity in mice and genotoxicity in mice and monkeys. Reports. J Natl Cancer Inst 1997; 89:1602-8.
18. Olivero OA, Shearer GM, Chougnet C, et al. Incorporation of zidovudine into leukocyte DNA from HIV-1 positive adults and pregnant women, and cord blood from infants exposed in utero. AIDS 1999; 13:919-25.
19. Sperling RS, Shapiro DE, McSherry GD, Britto P, Cunningham BE, Culnane M, et al. Safety of the maternal-infant zidovudine regimen utilized in the Pediatric AIDS Clinical Trial Group 076 Study. AIDS 1998; 12:1805-13.
20. Kumar RM, Hughes PF, Khurranna A. Zidovudine use in pregnancy: a report on 104 cases and the occurrence of birth defects. J Aquir Immune Defic Syndr 1994; 7:1034-9.
21. Glaxo Wellcome, Bristol-Myers Squibb Company, F. Hoffmann-La Roche Ltd. Antiretroviral Pregnancy Registry, interim report, 1 January 1989 through 31 December 1997. Wilmington, NC: Pharmaresearch Corporation, 1998.
22. Eldridge R, Tilson H, Davis G, Hurn B. Pregnancy outcomes following systemic prenatal acyclovir exposure June 1, 1984-June 30, 1993. MMWR Morb Mortal Wkly Rep 1993; 42:806-9.
23. White A, Andrews E, Eldridge R, et al. Birth outcomes following zidovudine therapy in pregnant women. Epidemiologic Notes and Reports. MMWR Morb Mortal Wkly Rep 1994; 43:409-16.
24. White A, Eldridge R, Andrews E. Birth outcomes following zidovudine. Acta Paediatr 1997; 421:86-8.
25. Bright RA, Aborn J, Everitt D. Medicaid data as a resource for epidemiologic studies: strengths and limitations. J Clin Epidemiol 1989; 42:937-45.
26. Carson JL, Strom BS. Medicaid Databases. In: Strom BL, ed. Pharmacoepidemiology. New York: J Wiley and Sons, 1994.
27. Wade NA, Birkhead GS, Warren BL, et al. Abbreviated regimens of zidovudine prophylaxis and perinatal transmission of the human immunodeficiency virus. N Engl J Med 1998; 339:1409-14.
28. Shaffer N, Chuachoowong R, Mock PA, et al. Short-course zidovudine for perinatal HIV-1 transmission in Bangkok, Thailand: a randomised controlled trial. Bangkok collaborative perinatal HIV transmission study group. Lancet 1999; 353:773-80.
29. Wiktor SZ, Ekpini E, Karon JM, et al. Short-course oral zidovudine for prevention of mother-to-child transmission of HIV-1 in Abidjan, Cote d'Ivoire: a randomised trial. Lancet 1999; 353:781-5.
30. Dabis F, Msellati P, Meda N, et al. 6-month efficacy, tolerance, and acceptability of a short regimen of oral zidovudine to reduce vertical transmission of HIV in breastfed children in Cote d'Ivoire and Burkina Faso: a double-blind placebo-controlled multicentre trial. DITRAME study group. Diminution de la Transmission Mere-enfant. Lancet 1999; 353:786-92.
31. Report from the workshop on detection of potential toxicities following perinatal exposure to antiretrovirals. Sponsored by National Institutes of Health, Centers for Disease Control and Prevention, Food and Drug Administration, Health Resources and Services Administration, Workshop held at Bethesda, MD, 1999.
© 2000 Lippincott Williams & Wilkins, Inc.