Despite known interventions to prevent mother-to-child transmission of HIV (PMTCT), an estimated 430,000 children worldwide became newly infected with HIV in 2008.1 Combination antiretroviral therapy (ART) commenced during pregnancy dramatically reduces vertical transmission.2-4 However, because of heavy disease burdens and overstretched health care systems, availability of such interventions is limited in most resource-constrained settings. Global criteria to start ART in pregnancy, though, are expanding. In 2009, the World Health Organization (WHO) recommended that pregnant women with a CD4+ cell count ≤350 cells per microliter start ART5 because they represent the majority of women who transmit HIV to their infants and in whom maternal and infant morbidity and mortality occur.6-10 Since 2008, Zambia has used this CD4 criteria for ART eligibility in pregnancy.11
Access to CD4 testing in many parts of the world remains a formidable challenge. In 2008, WHO estimated that only 12% of HIV-infected pregnant women were assessed for ART eligibility through clinical staging or CD4 screening.12 Even when the test is purportedly available, results can take weeks-even months-to return,13 which may shorten the interval between initiation of ART and delivery. Furthermore, opportunities to impact maternal health may be delayed or missed. Algorithms that predict the likelihood of an HIV-infected pregnant woman having a CD4+ cell count ≤350 cells per microliter could be an attractive alternative, especially if inexpensive and easy to implement. In this report, we evaluate maternal predictors of CD4+ cell count ≤350 cells per microliter.
In Zambia, PMTCT services have been offered routinely during antenatal care since 2002 and have been described elsewhere.14 Opt-out HIV testing is practiced nationwide. Point-of-care hemoglobin and syphilis screening are standard, as is reflex CD4 testing. WHO clinical staging in pregnancy generally occurs after CD4 testing. In Lusaka, a city of 2 million people, 25 public sector sites use the Zambia Electronic Perinatal Record System (ZEPRS), a networked patient-level electronic medical record system that captures information regarding antenatal care and PMTCT.
We conducted a retrospective study of all HIV-infected women enrolling into antenatal care in government clinics in Lusaka from May 1, 2007, to June 30, 2009, through ZEPRS. We included only women newly diagnosed with HIV, thus removing potential biases that might arise from those already on ART. We excluded details of repeat pregnancies, defined as subsequent pregnancies entered into ZEPRS for one mother. We examined factors that could potentially predict CD4+ cell count ≤350 cells per microliter in pregnancy. Maternal demographic, historical, and clinical data are routinely captured at the first antenatal visit. History of stillbirth and preterm birth were ascertained by self-report. Gestational age was estimated by last menstrual period and corroborated or changed according to abdominal palpation of uterine size.
Using multivariate logistic regression, we calculated the adjusted odds ratios (AORs) with corresponding 95% confidence intervals for the associations between each potential factor and CD4+ cell count ≤350 cells per microliter. To ensure the practicality of such a model in the clinical setting, we restricted the potential factors to all categorical variables with at least a 33% increase or decrease in odds and all continuous variables. The misclassification rate, calculated as the sum of the false-positive rate and the false-negative rate, was used as an overall measure of the discriminatory abilities of the predictive models.
We calculated the misclassification rate for all possible combinations of the selected factors. For combinations with continuous variables, we iteratively dichotomized the patients across a wide range of values and calculated separate misclassification rates. We defined the best performing predictor for CD4+ cell count ≤350 cells per microliter as the model that minimized the misclassification rate. We evaluated each model for sensitivity, specificity, positive predictive value, and negative predictive value.
Analyses were performed using SAS version 9.1 (SAS Institute, Cary, NC) and R software version 2.4.1 (http://www.r-project.org). Use of these routinely collected clinical data was approved by the ethical review committees at the University of Zambia (Lusaka, Zambia) and University of Alabama at Birmingham (Birmingham, AL).
Between May 1, 2007, and June 30, 2009, 138,884 pregnancies were registered in ZEPRS. For 823 women with multiple recorded pregnancies, only data from the first pregnancy were included. Of the 133,238 (95.9%) women who had an HIV test, 28,976 (21.7%) tested HIV positive. The 3018 (10.4%) women who knew their positive HIV status before were excluded from the analysis.
Twenty thousand two hundred thirty-three (77.9%) of the 25,958 newly diagnosed HIV-infected pregnant women had a documented CD4 result; of these, 9876 (48.8%) women had a CD4+ cell count ≤350 cells per microliter. Compared to women without a CD4 result available, women with CD4 results were younger (median age of 26 years vs. 27 years; P < 0.01), had a lower median hemoglobin measurement (11.1 vs. 11.6 g/dL; P < 0.01), had lower levels of education (40.5% with primary education vs. 36.2%; P < 0.01), had higher rates of syphilis (22.2% vs. 18.6%; P < 0.01), and were less likely to have a reported history of tuberculosis (TB) (2.8% vs. 3.7%; P < 0.01) and a previous death on an infant before 1 year of age (5.4% vs. 7.0%; P < 0.01). There were no differences with regard to gestational age at first antenatal visit, weight, malaria during the current pregnancy, marital status, parity, and previous child death before 5 years of age (data not shown).
Predictors of CD4+ Cell Count ≤350 Cells Per Microliter in Pregnancy
In multivariable analysis (Table 1), the following categorical variables were significantly associated with CD4+ cell count ≤350 cells per microliter: history of TB [AOR: 1.48, 95% confidence interval (CI): 1.20 to 1.84] and previous death of an infant before 1 year of age (AOR: 1.58, 95% CI: 1.36 to 1.82). Additional associated factors included primary/no education (AOR: 0.86, 95% CI: 0.80 to 0.92), parity > 2 (AOR: 0.88, 95% CI: 0.80 to 0.96), and previous child death before 5 years of age (AOR: 1.31, 95% CI: 1.15 to 1.50). Continuous variables included in the analysis were age at first antenatal visit (years), baseline hemoglobin (grams/deciliter), gestational age at first antenatal visit (weeks), and baseline weight (kilograms). The best performing predictive models based on a single variable are listed in Table 2.
We then constructed predictive models based on combinations of variables (Table 2). A model that utilized only self-reported characteristics (age, history of TB, and previous death of an infant before 1 year of age) had a misclassification rate of 43.8%. Overall, the best performing model to predict CD4+ cell count ≤350 cells per microliter in pregnancy was based on at least 3 of the following: age ≥ 28 years, hemoglobin ≤ 9.8 mg/dL, gestational age ≤ 30 weeks at first antenatal visit, weight ≤ 64 kg, history of TB, or previous death of an infant before 1 year of age. This model had a sensitivity of 45.7%, specificity of 70.7%, positive predictive value of 59.4%, negative predictive value of 58.2%, and misclassification rate of 41.4%. In comparison, a simple model that assumed all HIV-infected pregnant women had CD4+ cell count ≤350 cells per microliter approximated this misclassification rate at 51.7%, with a sensitivity of 100% and specificity of 0%. According to the best performing model, 5066 of 13,599 pregnant women (37.3%) were classified as in need of treatment. The median CD4+ cell count of these women was significantly lower than the median CD4+ cell count of women not classified as requiring treatment [306 (interquartile range: 192-447) vs. 392 (interquartile range: 269-532); P < 0.01].
As a secondary analysis, we performed a multivariate logistic regression analysis using the 6 predictors of the best performing model (including age, hemoglobin, weight, and estimated gestational age at first antenatal visit as continuous variables) to determine predicted probabilities of having a CD4+ cell count ≤350 cells per microliter. We classified a pregnant woman as having a CD4+ cell count ≤350 cells per microliter if the predicted probability was greater than 50%. The misclassification rate from this model was 42.2%.
We found that the following maternal characteristics were associated with CD4+ cell count ≤350 cells per microliter: older age, low hemoglobin, earlier gestational age at first antenatal visit, low weight, history of TB, and previous infant death. While 37.3% of the pregnant women who met criteria of our best performing model not only were in need of ART but also had a significantly lower median CD4+ cell count, the misclassification rate was 41.4%. Thus, a validation analysis was not conducted to see if the predictive models were generalizable. The reduction by 10% in misclassification reflects a 20% relative improvement in correctly predicting that a woman with these maternal characteristics would have CD4+ cell count ≤350 cells per microliter. However, overall performance of the models remained poor. Several variables, such as history of TB, were specific but had low positive predictive values given the low self-reported prevalence of TB in our population. Given that 50% of HIV-infected pregnant women are likely to qualify for ART based on a CD4+ cell count ≤350 cells per microliter,13,15-17 it is highly unlikely that a model will be able to eliminate misclassification dramatically.
To our knowledge, this analysis is unique because it focuses on simple predictors of low CD4+ cell count in pregnant HIV-infected women. Whereas one study examined predictors of mortality in HIV-infected pregnant women,18 another built a predictor model for ART initiation in pregnant women based on self-reported symptoms.19 Our analysis focuses on characteristics that are demographic and historical and thus could be easy to implement in resource-limited settings where nontrained workers sometimes provide care. We acknowledge several limitations of this analysis. This cohort comprises an urban population and may not be generalizable to rural settings.20 We were also limited to factors that are already captured in antenatal care and thus did not include WHO clinical staging for HIV disease. However, other studies have shown that pregnant women often present at WHO stage 1 or 2.17,21,22
Implementation of the 2009 WHO PMTCT guidelines will likely result in half of all HIV-infected pregnant women becoming eligible for ART based on CD4 criteria. This percentage may increase by an additional 6%-20% after WHO clinical staging is performed.9,10,17 Given the present situation of limited access to CD4 testing in many resource-constrained settings and the efficacy of ART in pregnancy,3,23 policy makers could consider initiating ART in all HIV-infected pregnant women who do not have access to CD4+ cell counts.24 However, such a policy would be costly and would present issues of side effects of lifelong ART and multidrug resistance.25 Another alternative is scale-up of point-of-care CD4 testing,26 and efforts should continue to develop the technology so that it remains applicable to resource-constrained settings.
Ensuring timely initiation of ART among eligible pregnant women remains a major public health priority and addresses WHO Millennium Development Goals to reduce maternal mortality and achieve universal access to HIV treatment. Unfortunately, we demonstrate poor performance in predicting CD4+ cell count ≤350 cells per microliter among HIV-infected pregnant women in Lusaka. Our analysis reveals that CD4 triage remains a critical element of maternal HIV care and PMTCT.
We would like to thank the Zambian Ministry of Health and Lusaka District Health Management Team for their continued support of operational research. K. C. Liu wrote the first draft, supported interpretation of results, and was responsible for editing the final article. B. H. Chi, J. S. A. Stringer, and E. M. Stringer conceived the study, guided the analysis, and interpreted the data. M. J. Giganti performed the statistical analysis and heavily assisted with writing the Methods section. All authors contributed to subsequent drafts and approval of the final version.
1. 2009 AIDS Epidemic Update: Geneva. Joint United Nations Programme on HIV/AIDS
. Vol 1. Geneva, Switzerland; 2009:1-100.
2. Sturt AS, Dokubo EK, Sint TT. Antiretroviral therapy (ART) for treating HIV infection in ART-eligible pregnant women. Cochrane Database Syst Rev
. (3):CD008440. 2010.
3. Cooper ER, Charurat M, Mofenson L, et al. Combination antiretroviral strategies for the treatment of pregnant HIV-1-infected women and prevention of perinatal HIV-1 transmission. J Acquir Immune Defic Syndr
4. Ekouevi DK, Coffie PA, Becquet R, et al. Antiretroviral therapy in pregnant women with advanced HIV disease and pregnancy outcomes in Abidjan, Cote d'Ivoire. AIDS
5. WHO MTCT Collaborative Study Team. Rapid Advice: Use of Antiretroviral Drugs for Treating Pregnant Women and Preventing HIV Infection in Infants
. Vol 1. Geneva, Switzerland: WHO; 2009:1-26.
6. Bicego G, Boerma JT, Ronsmans C. The effect of AIDS on maternal mortality in Malawi and Zimbabwe. AIDS
7. Sewankambo NK, Gray RH, Ahmad S, et al. Mortality associated with HIV infection in rural Rakai District, Uganda. AIDS
8. McIntyre J. Maternal health and HIV. Reprod Health Matters
9. Kuhn L, Aldrovandi GM, Sinkala M, et al. Potential impact of new WHO criteria for antiretroviral treatment for prevention of mother-to- child HIV transmission. AIDS
10. Kuhn L, Aldrovandi GM, Sinkala M, et al. Differential effects of early weaning for HIV-free survival of children born to HIV-infected mothers by severity of maternal disease. PLoS One
11. Ministry of Health. National Protocol Guidelines: Integrated Prevention of Mother to Child Transmission of HIV/AIDS
. Lusaka, Zambia: Ministry of Health; 2008.
12. World Health Organization. Towards Universal Access: Scaling Up Priority HIV/AIDS Interventions in the Health Sector
. Geneva: WHO; 2008.
13. Mandala J, Torpey K, Kasonde P, et al. Prevention of mother-to-child transmission of HIV in Zambia: implementing efficacious ARV regimens in primary health centers. BMC Public Health
14. Stringer EM, Sinkala M, Stringer JS, et al. Prevention of mother-to-child transmission of HIV in Africa: successes and challenges in scaling-up a nevirapine-based program in Lusaka, Zambia. AIDS
15. Kuhn L, Aldrovandi GM, Sinkala M, et al. Effects of early, abrupt weaning on HIV-free survival of children in Zambia. N Engl J Med
16. Stinson K, Boulle A, Coetzee D, et al. Initiation of highly active antiretroviral therapy among pregnant women in Cape Town, South Africa. Trop Med Int Health
17. Carter RJ, Dugan K, El-Sadr WM, et al. CD4+ cell count testing more effective than HIV disease clinical staging in identifying pregnant and postpartum women eligible for antiretroviral therapy in resource-limited settings. J Acquir Immune Defic Syndr
18. Brown ER, Otieno P, Mbori-Ngacha DA, et al. Comparison of CD4 cell count, viral load, and other markers for the prediction of mortality among HIV-1-infected Kenyan pregnant women. J Infect Dis
19. Nguyen RH, Gange SJ, Serwadda D, et al. Screening HIV-positive pregnant women for antiretroviral therapy: utility of self-reported symptoms. Int J STD AIDS
20. Kishk NA. Knowledge, attitudes and practices of women towards antenatal care: rural-urban comparison. J Egypt Public Health Assoc
21. Kigozi IM, Dobkin LM, Martin JN, et al. Late-disease stage at presentation to an HIV clinic in the era of free antiretroviral therapy in sub-Saharan Africa. J Acquir Immune Defic Syndr
22. O'Brien ME, Kupka R, Msamanga GI, et al. Anemia is an independent predictor of mortality and immunologic progression of disease among women with HIV in Tanzania. J Acquir Immune Defic Syndr
23. Karon JM, Fleming PL, Steketee RW, et al. HIV in the United States at the turn of the century: an epidemic in transition. Am J Public Health
24. Zolfo M, De Weggheleire A, Schouten E, et al. Time for “test and treat” in prevention of mother-to-child transmission programs in low- and middle-income countries. J Acquir Immune Defic Syndr
25. Harrington M, Carpenter CC. Hit HIV-1 hard, but only when necessary. Lancet
26. Peter T, Badrichani A, Wu E, et al. Challenges in implementing CD4 testing in resource-limited settings. Cytometry B Clin Cytom
. 2008;74(suppl 1):S123-S130.
Keywords:© 2011 Lippincott Williams & Wilkins, Inc.
HIV; pregnancy; CD4; predictor; PMTCT; antiretroviral therapy eligibility