Maternal Virus Load and Perinatal Human Immunodeficiency Virus Type 1 Subtype E Transmission, Thailand

590

Maternal Virus Load and Perinatal Human Immunodeficiency Virus Type 1 Subtype E Transmission, Thailand Nathan Shaffer, Anuvat Roongpisuthipong, Wimol Siriwasin, Tawee Chotpitayasunondh, Sanay Chearskul, Nancy L. Young, Bharat Parekh, Philip A. Mock, Chaiporn Bhadrakom, Pratharn Chinayon, Marcia L. Kalish, Susan K. Phillips, Timothy C. Granade, Shambavi Subbarao, Bruce G. Weniger, and Timothy D. Mastro, for the Bangkok Collaborative Perinatal HIV Transmission Study Group1

HIV/AIDS Collaboration, Nonthaburi, and Siriraj Hospital Faculty of Medicine, Mahidol University, and Rajavithi Hospital and Children’s Hospital, Department of Medical Services, Ministry of Public Health, Bangkok, Thailand; Centers for Disease Control and Prevention, Atlanta, Georgia

To determine the rate and risk factors for human immunodeficiency virus (HIV)-1 subtype E perinatal transmission, with focus on virus load, pregnant HIV-infected women and their formula-fed infants were followed prospectively in Bangkok. Of 281 infants with known outcome, 68 were infected (transmission rate, 24.2%; 95% confidence interval, 19.3%–29.6%). Transmitting mothers had a 4.3-fold higher median plasma HIV RNA level at delivery than did nontransmitters (P ! .001 ). No transmission occurred at !2000 copies/mL. On multivariate analysis, prematurity (adjusted odds ratio [AOR], 4.5), vaginal delivery (AOR, 2.9), low NK cell percentage (AOR, 2.4), and maternal virus load were associated with transmission. As RNA quintiles increased, the AOR for transmission increased linearly from 4.5 to 24.8. Twothirds of transmission was attributed to virus load 110,000 copies/mL. Although risk is multifactorial, high maternal virus load at delivery strongly predicts transmission. This may have important implications for interventions designed to reduce perinatal transmission.

Although subtype E human immunodeficiency virus (HIV)1 infection has spread rapidly in Thailand and surrounding countries since 1989 [1], the rate and risk factors for subtype E perinatal transmission have not been described. While HIV infection rates in young men have decreased dramatically from peak levels [2, 3], HIV infection in women and children in Thailand has become epidemic. The 1996 median provincial HIV prevalence among pregnant women was 2.3%; ∼23,000 HIV-infected women deliver annually. We hypothesized that in Received 24 July 1998; revised 16 October 1998. Presented in part: XI International Conference on AIDS, Vancouver, Canada, July 1996 (abstract 4948). Written, informed consent was obtained from all study women for themselves and their newborn infants. The consent and study protocol were approved by the Ethical Review Committee, Ministry of Public Health, Thailand, and the Institutional Review Board, CDC, Atlanta. Human experimentation guidelines of the US Department of Health and Human Services were followed in the conduct of this clinical research. The use of trade names is for identification only and does not imply endorsement by the Public Health Service of the US Department of Health and Human Services. Financial support: CDC, as part of a bilateral HIV/AIDS research collaboration with the Thai Ministry of Public Health. 1 Members are listed after text. Reprints or correspondence: Dr. Nathan Shaffer, Centers for Disease Control and Prevention (Mailstop E-50), 1600 Clifton Rd., Atlanta, GA 30333 ([email protected]). The Journal of Infectious Diseases 1999; 179:590–9 q 1999 by the Infectious Diseases Society of America. All rights reserved. 0022-1899/99/7903-0008$02.00

a new, rapidly spreading epidemic, perinatal transmission might be high. We also wanted to investigate whether HIV-1 subtype E might be more transmissible perinatally than is subtype B, as has been suggested for heterosexual transmission [4]. In other settings worldwide, reported perinatal HIV-1 transmission rates range widely, from 12% to 48% [5]. However, wellconducted studies in the United States and Europe suggest that without antiretroviral interventions and in the absence of breast-feeding, the perinatal transmission rate for HIV-1 subtype B clusters at 20%–25% [6–9]. A variety of risk factors for transmission have been identified, including advanced maternal illness, low maternal CD41 T cell counts, premature delivery, prolonged rupture of membranes, and breast-feeding [10]. Cesarean section may be protective, but this remains unclear [11]. Recently, maternal plasma virus load has been associated with transmission [12–19], but the reported strength of this relationship also has been quite variable. Moreover, few studies have simultaneously examined maternal, obstetric, and virus load factors in a unified analysis. Zidovudine can markedly reduce perinatal transmission [6], but the AIDS Clinical Trials Group (ACTG) 076 analysis suggested that reduction in maternal virus load explains only a small part of this effect [15]. As efforts continue to further decrease perinatal transmission in developed countries, and to introduce simplified regimens in developing countries, it is important to better understand the relation of virus load and

JID 1999;179 (March)

HIV Load and Perinatal Transmission, Thailand

transmission. We therefore investigated the rate of and risk factors for perinatal transmission of HIV-1 subtype E in Thailand, with specific emphasis on maternal virus load.

Methods Patients and protocol. The study was conducted in Bangkok at Siriraj Hospital (18,000 deliveries per year), Rajavithi Hospital (17,000 deliveries per year), and Children’s Hospital. The setting and enrollment procedures have been described [20]. HIV counseling and testing of pregnant women is a routine part of antenatal care. In this study, no subjects were receiving antiretroviral therapy during pregnancy (enrollment was before the release of the ACTG 076 results). Consenting pregnant women, confirmed to be HIV-1– positive, residing in the Bangkok area, and willing to bring their children for study visits were eligible. Specific counseling was given not to breast-feed, in accord with Thai national guidelines for HIVpositive women. Enrolled mothers and children were provided individualized care and support from the research team and the clinical investigators. Incentives for participation included transportation reimbursement for study visits, partial payment of medical fees, and provision of infant formula. From November 1992 through March 1994, HIV-1–seropositive pregnant women were offered enrollment [20]. A baseline sociodemographic and risk factor questionnaire was administered, and a physical examination was done. Enrolled women were scheduled for repeat evaluations at each remaining trimester, at delivery, and at 6 and 12 months postpartum. Venous blood was collected for immunologic and virus studies and storage. Delivery blood samples were obtained 1–3 days after delivery. Newborn infants were examined by study pediatricians, and the delivery records were reviewed. The children were scheduled for study visits at 1, 2, 4, 6, 9, 12, 15, and 18 months, coinciding with standard well-child visits. Venous blood (1–3 mL) was drawn at birth and at 2, 6, 12, and 15 months; all birth and 2- and 6-month samples were systematically tested for HIV-1 by DNA polymerase chain reaction (PCR). Laboratory sample collection and processing. Serum and EDTA-anticoagulated venous blood samples were collected from mothers for serologic assays, lymphocyte phenotyping, and hematologic measurements. For the first half of the study, venous blood was also collected in heparinized tubes and processed by ficoll-hypaque density gradient centrifugation for peripheral blood mononuclear cells and plasma. For the second half, because of a consensus that citrate or EDTA were the preferred anticoagulants for virus load assays, venous blood was collected in citrated cell preparation tubes (Becton Dickinson Vacutainer Systems, Franklin Lakes, NJ) and centrifuged within 30 min. Cells and DNA lysate were frozen in liquid nitrogen, and all plasma samples, regardless of anticoagulant, were frozen at 2707C within 8 h of collection. The EDTA tube drawn from infants and children was processed within 6 h for complete blood cell count, flow cytometry, and plasma storage and within 24 h for DNA lysates. Serologic assays. Maternal enrollment serum and sequential child plasma samples were tested with the Genetic Systems HIV1/HIV-2 EIA (Genetic Systems, Redmond, WA) and Novapath HIV-1 Immunoblot (BioRad Laboratories, Hercules, Calif). Qualitative HIV DNA PCR. An aliquot of the infant sample

591

was prepared as a leukocyte DNA lysate and frozen in liquid nitrogen for later DNA PCR testing. Lysate was prepared by washing a 500-mL sample of whole blood with lysis buffer (Amplicor Specimen Wash Solution; Roche Diagnostics, Branchburg, NJ) and then with proteinase K extraction buffer. Diagnostic PCR was done at the Centers for Disease Control and Prevention (CDC) on duplicate coded samples, with negative and positive controls, with two gag primer pairs (SK38/SK39 and SK145/SK150) [21]. This PCR assay is estimated to detect 10 provirus copies/25 mL of DNA lysate, representing 150,000 leukocytes. HIV-1 RNA load. HIV-1 RNA load for coded maternal delivery specimens (first freeze-thaw) was determined with the Amplicor HIV-1 MONITOR Test (Roche Diagnostics) according to instructions, with one modification. A new set of gag primers (SK145 and SK151) provided by the manufacturer was added to the master mix reagent before amplification to ensure efficient quantification of RNA from HIV-1 subtype E [22, 23]. Plasma samples collected in heparin were pretreated with heparinase before RNA extraction and PCR testing. With a starting volume of 200 mL, the assay has a threshold sensitivity to detect 200 HIV RNA copies/mL and a linear quantitation range of 400–750,000 copies/ mL. Lymphocyte phenotyping and HIV-1 subtyping. Lymphocyte phenotyping was done on fresh mother or infant venous samples (EDTA anticoagulant) with the FACScan flow cytometer (Becton Dickinson Immunocytometry Systems, San Jose, CA), by use of a standard six-tube, two-color monoclonal antibody panel (Becton Dickinson) [24]. NK cells were defined as cells that were CD32, CD161, and CD561. Peptide EIAs were done on enrollment maternal serum samples by use of V3-loop envelope peptides [25]. Serum monoreactive to subtype E or B0 (Thai B) was considered to indicate subtype E or B infection. For women with dually reactive or nonreactive serum, later samples were retested. For specimens that remained dually reactive or nonreactive, the C2-V3 region of gp120 was amplified either from DNA lysates of peripheral blood mononuclear cells or from reverse-transcribed RNA from serum samples to determine HIV-1 subtype. Definition of infant HIV infection and timing of transmission. Infants were considered HIV-infected if they had two positive DNA PCR tests or one positive PCR test and an AIDS-defining condition according to CDC classifications [26]. Infants were defined as uninfected if they tested PCR-negative on 2 samples, including 1 obtained at >6 months of age, or seroreverted to HIV-negative status on EIA testing. Infants not meeting these criteria were considered to have unknown infection status. Following the proposed definition of Bryson et al. [27], with minor modification [28], we considered infected infants who tested DNA PCR-positive on a venous blood sample collected within 72 h of birth to be infected in utero and infants who tested PCR-negative at birth but later tested PCR-positive to be infected intrapartum. Infected infants who did not have a birth PCR result were omitted from analyses of timing of transmission. Statistical analysis. Transmission analyses were limited to transmitting mothers (mothers whose infants were infected) and nontransmitting mothers (mothers whose infants were uninfected). The transmission rate was calculated as the number of known infected children divided by the number of known infected and uninfected children. Minimum and maximum transmission rate

592

Shaffer et al.

estimates were made by assuming all unknown children to be either uninfected or infected. An alternative estimate of the transmission rate was made by use of the Kaplan-Meier modified life table method, with infection defined as any positive PCR test. Potential risk factors for transmission, including obstetric characteristics, HIV risk behavior, HIV illness severity, immunologic characteristics (CD41 and CD81 cell count, percentage, and ratio at third trimester and delivery; NK cell count and percentage), virologic characteristics (virus load and subtype), and partner HIV status were evaluated. We have separately reported on maternal sexually transmitted diseases in the full cohort (syphilis) and in a substudy (gonorrhea and chlamydial infection) [29]. Frequency data were analyzed by x2 or Fisher’s exact test for dichotomous variables and x2 for trend for ordered variables; relative risks for transmission and 95% confidence intervals (CIs) were calculated. The Wilcoxon rank sum test was used to compare differences for continuous variables. Continuous variables were also dichotomized at the median or at levels reported as significant by other investigators. Because there were no significant differences in the virus load findings for specimens collected in heparin (after treatment with heparinase) and citrate, and because the epidemiologic findings were consistent, virus load data were combined. Maternal virus levels were grouped into quartiles and quintiles and a nonparametric lowess smoother function [30] was used to further evaluate the relationship of maternal RNA level and risk of transmission. These analyses were also done controlling for the effect of anticoagulant. Associations between continuous variables were assessed by Spearman’s rank correlation. Multiple unconditional logistic regression was used to analyze variables found to be significant at P ! .1 on univariate analysis, and adjusted odds ratios and 95% CIs were estimated from the final logistic model. Population-attributable fractions were estimated for virus load categories by use of logistic regression, adjusted for other significant

JID 1999;179 (March)

risk factors. For this cohort analysis, this entailed estimating from the logistic model the observed number of cases and then estimating from the same model the expected number of cases assuming a reduction of virus load to a target value. The population-attributable fraction is estimated as 1 2 (expected/observed) number of cases [31, 32]. Virus load was grouped into quintiles and a 4-group categorization was done on the basis of log10 cut points for attributable fraction estimation. Category-specific attributable fraction estimates were calculated, along with a summary estimate (the sum of the category-specific estimates).

Results Enrollment. A total of 342 HIV-1–infected pregnant women were enrolled (table 1). Most women were young, primigravid, enrolled in the study before third trimester, and asymptomatic with regard to their HIV disease. The median CD41 lymphocyte count was 430 cells/mm3. Ten percent of women reported a history of commercial sex work, and 7% reported injection drug use or a partner who injected drugs. In addition to the characteristics shown, 60% of women were married and 95% reported one steady partner (median time with partner, 1.5 years). Of the women enrolled, 47 (14%) were lost to follow-up before delivery (31 lost, 11 miscarriages, 4 abortions, 1 stillbirth). The remaining 295 women all had singleton live births. Of these, definitive infection status was determined for all but 14 (4.7%) children; 3 died in the first month and 11 were lost before 2 months. Follow-up of the children was high: 92% at 6 months, 87% at 12 months, and 82% at 18 months. The 281 women included in the transmission analysis and the 61 women excluded did not differ (table 1).

Table 1. Characteristics of HIV-1–infected pregnant women at enrollment. a

Transmission analysis Characteristic

All (n 5 342)

Age, years, median (range) Gravidity, no. (%) 1 2 12 Gestational age, weeks, first antenatal clinic visit, median (range) Enrolled before third trimester, no. (%) HIV-related symptoms (any), no. (%) WHO clinical stage 1, no. (%) CD41 T cells/mm3 at enrollment, no. (%) !200 200–499 >500 Median (range) HIV subtype, no. (%) E B HIV risk, no. (%) CSW IDU Partner IDU Education, years, median (range)

22 (14–41)

Included (n 5 281) 22 (14–40)

194 106 42 18 256 96 333

(56.7) (31.0) (12.3) (4–39) (74.9) (28.1) (97.4)

156 88 37 19 204 81 273

(55.5) (31.3) (13.2) (6–39) (72.6) (28.8) (97.2)

6 220 116 430

(1.8) (64.3) (33.9) (100–1290)

5 175 101 440

(1.8) (62.3) (35.9) (100–1290)

326 (95.6) 15 (4.4) 35 4 22 6

(10.2) (1.2) (6.4) (1–16)

268 (95.7) 12 (4.3) 28 4 21 6

(10.0) (1.4) (7.8) (1–16)

NOTE. WHO, World Health Organization; CSW, commercial sex work; IDU, injection drug use. a All comparisons between women included and not included, P 1 .5.

Not included (n 5 61) 22 (17–41) 38 18 5 17 52 15 60 1 45 15 390

(62.3) (29.5) (8.2) (4–34) (85.2) (24.6) (98.4) (1.6) (73.8) (24.6) (160–790)

58 (95.1) 3 (4.9) 7 0 1 6

(11.5) (0) (1.8) (2–13)

JID 1999;179 (March)

HIV Load and Perinatal Transmission, Thailand

Transmission rate. Of the 281 children with known infection status, 68 were infected and 213 were uninfected, yielding a transmission rate of 24.2% (95% CI, 19.3%–29.6%). Of the uninfected children (defined primarily by PCR-negative testing), 212 (99.5%) were confirmed as HIV-negative on EIA testing at 12 or 15 months. When it was assumed that all unknown children were uninfected or infected, the minimum and maximum transmission estimates were 23.1% and 27.8%. When all children with at least one PCR test were included (280/281

593

children), the Kaplan-Meier estimate of transmission was 23% (95% CI, 18%–27%). Repeated inquiries confirmed that all the infants were formula-fed without difficulty; only 1 infant received some breast milk from his mother and 2 others received limited breast milk from a relative. Univariate risk factors for transmission. There was no difference in maternal age, gravidity, parity, HIV symptoms, HIV risk behavior, partner HIV status, hepatitis B or syphilis serologic status, or duration of membrane rupture or labor be-

Table 2. Univariate analysis of risk factors for perinatal HIV-1 transmission. Risk factor Maternal age > median (22 y) ! median Primigravida No Yes HIV-related symptoms No Yes CD41 T cells/mm3 at delivery > median (450) ! median HIV risk Heterosexual IDU/partner IDU CSW Gestational age, weeks >37 !37 Birth weight (g) >2500 !2500 Mode of delivery Cesarean elective Vaginal NK cell % at delivery 111 (median) 12 Viral subtype B E

a

No. of subjects (n 5 281)

No. infected (%) (n 5 68)

Relative risk (95% confidence interval)

P

168 113

38 (22.6) 30 (26.6)

1.0 1.2 (0.8–1.8)

.45

125 156

29 (23.2) 39 (25.0)

1.0 1.1 (0.7–1.6)

.73

200 81

45 (22.5) 23 (28.4)

1.0 1.3 (0.8–1.9)

.30

142 133

31 (21.8) 36 (27.1)

1.0 1.2 (0.8–1.9)

.31

232 21 28

58 (25.0) 3 (14.3) 7 (25.0)

1.0 0.6 (0.2–1.7) 1.0 (0.5–2.0)

.70

270 11

62 (23.0) 6 (54.6)

1.0 2.4 (1.3–4.3)

.02

253 28

57 (22.5) 11 (39.3)

1.0 1.7 (1.0–2.9)

.05

34 11 247

4 (11.8) 0 (0) 64 (25.9)

1.0

.07

119 156

22 (18.5) 45 (28.9)

1.0 1.6 (1.0–2.5)

.047

3 8 15 18 24

1.0 2.7 5.1 5.9 8.0

.001

56 56 55 57 56

(5.4) (14.3) (27.3) (31.6) (42.9)

2.2 (0.9–5.7)

(0.8–9.5) (1.6–16.6) (1.8–18.9) (2.6–25.1)

193 64

49 (25.4) 13 (20.3)

1.0 0.8 (0.5–1.4)

.41

25 181 55 7

6 41 14 3

(24.0) (22.7) (25.5) (42.9)

1.0 0.9 (0.5–2.0) 1.1 (0.5–2.4) 1.8 (0.6–5.4)

.42

93 117 41

26 (28.0) 27 (23.1) 7 (17.1)

1.0 0.8 (0.4–1.5) 0.5 (0.2–1.5)

.38

12 268

2 (16.7) 66 (24.6)

1.0 1.5 (0.4–5.3)

.74

NOTE. Subject nos. not totaling 281 or 68 indicate missing or unknown values, which were excluded from analysis. CSW, commercial sex work; IDU, injection drug use. a Overall x2.

594

Shaffer et al.

Figure 1. Maternal virus load at delivery (RNA copies/mL) for transmitting and nontransmitting mothers. Each symbol represents RNA level for given subject. Solid rule indicates median (transmitting 5 61,500 copies/mL, nontransmitting 5 14,400 copies/mL; P ! .001). Lower detection cutoff for assay (dashed line) was 200 copies/ mL.

tween transmitting and nontransmitting women (table 2). There was also no significant difference in transmission according to hepatitis B surface antigen status (not shown) or syphilis (VDRL test result >1:4), or other maternal sexually transmitted diseases [29]. The transmission rate did not differ statistically between women with HIV-1 subtype B (16.7%; 95% CI, 2.3%– 51.8%) and subtype E infection (24.6%; 95% CI, 19.5%– 30.2%), although the number of women with subtype B infection was small. Higher transmission rates were associated with prematurity, low birth weight, vaginal delivery, and low NK cell percentage. There was no transmission among the small number of women (11) with elective cesarean sections (before the onset of labor). Median CD41 T cell counts and percentage were slightly lower at third trimester and delivery among transmitting mothers than among nontransmitting mothers (delivery CD41 T cell counts shown), but the differences were not significant when

JID 1999;179 (March)

analyzed either as discrete or continuous variables. There were also no significant differences in CD81 T cell counts. Virus load and risk of transmission. For the 280 mothers with known transmission outcome and a delivery measurement, the median maternal plasma RNA level at delivery was 20,700 copies/mL (range, !200 [below detection] to 1,648,000). RNA levels were comparable for samples collected in citrate and heparin (not shown). Women with values above the median were at increased risk of transmission (relative risk, 3.24; 95% CI, 1.95–5.41; P ! .001). Median virus load was 4.3 times (0.63 log10) higher for transmitting women (61,500 copies/mL; range, 2200–581,000) than for nontransmitting women (14,400 copies/mL; range, !200– 1,648,000; P ! .001) (figure 1). Only 2 women (0.7%) had RNA levels below detection; both were nontransmitters. None of 20 women with RNA levels !2200 copies/mL transmitted infection. However, the 5 women with the highest RNA values (1600,000 copies/mL) also did not transmit (4 of the 5 had vaginal deliveries). When RNA levels were grouped by quintiles (table 2), the transmission rate increased from 5.4% to 42.9% from the lowest to the highest quintiles, and the relative risk of transmission increased from 2.7 to 8.0 from the second to the fifth quintiles, compared with the first quintile. There was no interaction of anticoagulant with quintile virus load groups, with respect to the probability of transmission. A similar pattern was seen when the data were analyzed by quartiles (not shown). Risk of transmission was 8.5% for women with virus load !10,000 copies/mL (4 log10) and 30.8% for women with virus load >10,000 copies/mL. Figure 2 shows a lowess smoothed curve of the probability of transmission with increas-

Figure 2. Risk of transmission according to maternal RNA level at delivery: lowess smoothed curve showing probability of transmission across full range of RNA values, according to percentiles and logtransformed copy number.

JID 1999;179 (March)

HIV Load and Perinatal Transmission, Thailand

Table 3. Multiple logistic regression analysis of risk factors for HIV1 perinatal transmission. Risk factor Prematurity (!37 weeks) Vaginal delivery NK cell % !11% (median) HIV RNA (copies/mL) b Quintile 2 Quintile 3 Quintile 4 Quintile 5

Univariate OR (95% CI) AOR (95% CI) 4.0 (1.0–17.2) 2.6 (0.8–9.2) 1.8 (1.0–3.4) 2.9 6.6 8.2 13.3

(0.7–15.0) (1.6–31.1) (2.1–37.6) (3.4–60.4)

4.5 (1.1–19.5) 2.9 (1.0–10.7) 2.4 (1.3–4.6) 4.5 10.9 11.5 24.8

P .03 .047 .006 a !.001

(1.0–31.0) (2.8–72.6) (3.0–75.9) (6.5–163.5)

NOTE. OR 5 odds ratio; AOR 5 adjusted odds ratio (adjusted for other covariates in model); CI 5 confidence interval. a 2 x for linear trend. b Quintile 2 is second lowest quintile grouping, quintile 5 is highest quintile.

ing maternal plasma delivery RNA level. As seen, the relationship is nearly linear throughout, except at the extremes. Virus load and CD41 T cell count. Maternal plasma RNA level was weakly negatively correlated with CD41 T cell count (r 5 2.28; P ! .001) but not with CD81 T cell count at delivery (r 5 .02; P 5 .7). Correlations were similar for both transmitting and nontransmitting mothers. Virus load and timing of transmission. For 49 transmitting mothers (72%) with a birth DNA PCR result for the child, the median RNA level was 1.8 times higher for 14 women (29%) presumed to have transmitted in utero (60,111 copies/mL) than for 35 women presumed to have transmitted intrapartum (33,850 copies/mL), but this difference was not significant (P 5 .38). Multivariate analysis of transmission risk factors. By multiple logistic regression, prematurity, vaginal delivery, low percentage of NK cells, and elevated maternal delivery RNA level were independently associated with transmission (table 3). Compared with the odds of transmission in the lowest (first) RNA quintile, the odds of transmission in quintiles 2 through 5 increased from 4.5 to 24.8. There was no association between maternal virus load and gestational age (not shown). Population attributable fraction estimates. To estimate the potential reduction in transmission that might be obtained with a reduction in virus load, population-attributable fraction estimates were calculated (table 4). According to this model, reducing virus load below the second quintile (!6500 copies/mL) in our population would reduce transmission by 84%, while reducing virus load below 4 log10 would reduce transmission by 67%.

Discussion In this non–breast-feeding Asian population with predominantly HIV-1 subtype E infection, we found the perinatal HIV transmission rate to be ∼24% and maternal plasma virus load at delivery to be an extremely strong predictor of transmission. Transmission risk was low (8.5%) at virus loads below 10,000

595

copies/mL, and no transmission occurred below 2000 copies/ mL. Other factors independently associated with transmission were prematurity, low NK cell percentage, and mode of delivery. These findings have important implications for the design and evaluation of intervention strategies to reduce perinatal transmission, both in Thailand and in other settings. Estimated perinatal transmission rates in given populations can differ substantially, because of differences in study methods, host or viral characteristics, obstetric practices, therapy, and breast-feeding. It is not known whether transmission risk differs among HIV-1 subtypes, although HIV-2 is known to have a low risk of perinatal transmission [33]. In our study, we were able to make a preliminary comparison between HIV-1 subtype E (24.6%) and subtype B0 (16.7%) transmission rates, but the small number of women with subtype B0 infection limited our power to conclude that the rates might be different. Nevertheless, our finding of an overall 24% transmission rate in a formula-fed cohort with predominantly HIV-1 subtype E infection suggests that the risk of perinatal transmission is similar to that for subtype B reported in North America and Europe [6–9]. Several recent studies have identified maternal virus load as a determinant of perinatal HIV-1 transmission (table 5). However, the strength of the findings differ, and few studies have adjusted for other maternal immunologic and obstetric risk factors in a single analysis. In addition, these studies differed from each other and from ours in the range of RNA values and the proportion of women without detectable HIV RNA (!1% in our study, compared with 4%–30% in other studies). Although the differences might be explained by different assays or subtypes, these differences might affect the strength of association found. Dickover et al. [14] found a high virus load among transmitters and a strong association with transmission at 150,000 copies/mL. In the ACTG 076 trial, Sperling et al. Table 4. Perinatal transmission population-attributable fraction estimates for maternal virus load. Virus load stratum Quintile analysis Q1 Q2 Q3 Q4 Q5 Summary Log10 analysis !4.0 (!10,000) 4.0–4.5 (10,000–31,623) 4.5–5.0 (31,623–100,000) 15.0 (1100,000) Summary

AF (95% confidence interval)

8.8 19.1 23.2 33.0 84.1

— (3–22) (11–30) (14–35) (24–44) (53–96)

17.9 21.9 27.3 67.0

— (8–34) (12–35) (18–39) (41–85)

NOTE. AF 5 attributable fraction—estimated % of transmission in population associated with various virus load levels, or amount of transmission that would theoretically be eliminated if virus load were reduced below these levels. AF estimates are adjusted for mode of delivery, prematurity, and NK cell %. Summary population-attributable fraction is sum of estimates for individual virus load strata for each of the two analyses.

596

Shaffer et al.

JID 1999;179 (March)

Table 5. Summary of major perinatal HIV transmission studies reporting on maternal virus load (plasma RNA).

Study [reference]

N (no. infected)

Subtype

Median virus load

% below detection

Transmitting

Nontransmitting

Association

Comment

Dickover et al. [14]

97 (20)

B

4%

94,054

4596

P ! .001

Lower threshold, !20,000; upper threshold, 180,000; for RNA 150,000, OR 5 53

Sperling et al. [15] Placebo

168 (38)

B

13%

8320

5370

P 5 .003

RNA levels at study entry; quartile analysis significant but nonlinear

Zidovudine a Cao et al. [16]

159 (12) 209 (19)

B B

10% 8%

19,330 15,000

4650 6000

P 5 .03 P 5 .06

Thea et al. [17]

105 (51)

B

30%

16,000

6600

P ! .01

Mayaux et al. [18]

236 (45)

A, B, C

15%

10,567

3574

P ! .01

Burns et al. [19]

160 (36)

B

NA

NA

NA

P 5 .006

Shaffer et al. (this analysis)

280 (68)

E

!1%

61,500

14,400

P ! .0001

85% of women used some zidovudine; no upper or lower thresholds, no linear relationship Nested case-control study, AOR 2.7 for CD41 cell count 1500 No thresholds; RNA lower in African women (methodology?); RNA significant on multivariate analysis Increased OR (2.9) with RNA >10,000; no transmission for plasma RNA !1000 New Roche Amplicor primers; quintile multivariate analysis highly significant; linear increased RR from 1 to 25; lower threshold !2000

NOTE. OR 5 odds ratio; AOR 5 adjusted odds ratio; RR 5 relative risk; NA 5 not available (not reported). Maternal virus load (plasma RNA) is shown as copies/mL. HIV-1 subtypes are presumed based on geographic distribution for various studies, except for present study, in which subtyping was done. All studies used Roche Amplicor Monitor assay (v. 1.0) with original primers, except Thea et al. (nucleic acid sequence-based amplification; Organon-Teknika, Rockville, MD) and present study (modified primer set; Roche). a % below detection and median virus loads were estimated from published figure.

[15] showed increasing transmission risk within quartiles of virus load, but the risk was not convincingly linear. Furthermore, only 10,000 copies/mL but found less association with women with advanced HIV disease and frequent sexual activity during pregnancy. They also found little or no transmission at !1000 copies/mL. In our study, by use of additional primers that have been shown to be highly sensitive for non-B subtypes and to yield nearly equivalent quantification among divergent group M HIV subtypes [22, 23], nearly all women had detectable plasma RNA levels at delivery, RNA levels were relatively high, and RNA level was the strongest predictor of transmission. Risk increased across nearly the full range of RNA values, as shown by the quintile analysis and the lowess smoothed curve plot. In the multivariate analysis, the adjusted odds of transmission increased to 25 for women in the highest RNA quintile level. If reduction in virus load at delivery translates directly into re-

duced perinatal transmission risk, then lowering virus load in this population below 4.5 log10 would reduce transmission by half, and lowering virus load below 4 log10 would reduce transmission by two-thirds. Several smaller perinatal virus load studies have reported upper and lower thresholds for transmission [12–14], while larger studies have not [15–18]. In our study, women in the highest virus load quintile still had a !50% risk of transmission, and several nontransmitting mothers had the highest virus loads, suggesting that there is no upper threshold above which transmission always occurs. Further investigation is needed to determine why some women with very high virus loads do not transmit. On the other hand, our data do suggest a lower threshold (!2000 copies/mL in our population) below which transmission is unlikely. While specific thresholds and cut points will differ in different populations, because of differences in risk factors, virus load assays, and subtypes, our very clear findings on virus load may be due in part to the fact that we could quantify the RNA level in nearly all women and in part to our collection and processing procedures. Before virus load measurements were available, CD41 cell count was used as a proxy for viral activity as well as an immunologic index. Several studies have found a strong relationship between CD41 T cell count or percentage and risk of perinatal transmission, although cutoffs and patterns have been variable [7, 8, 34–36], but other studies have not [16, 19]. In our study, CD41 T cell count and percentage were only weakly correlated with virus load and were not related to transmission. These differences might be explained by differences in the sample size of different studies, in the distribution of CD41 T cells in the specific study populations, or in other local risk factors.

JID 1999;179 (March)

HIV Load and Perinatal Transmission, Thailand

New markers, such as CD38 and CD8/DR, may provide more specific information on lymphocyte subsets and transmission risk [37]. A new finding in this study is that a low percentage of maternal NK cells is an independent risk factor for transmission. NK cells are effector cells that lyse tumor cells or virus-infected cells by nonspecific mechanisms [38, 39]. Low NK cell number or activity is associated with acquired immunodeficiencies, including AIDS, as well as with a variety of other conditions [40, 41]. However, the relationship between NK cell number and activity is not well defined. Our finding that NK cell depletion might lead to a higher risk of transmission should be examined in other studies, and functional studies of these cells should be conducted to investigate the immunologic basis for this association. Prematurity, one of the first risk factors identified with perinatal HIV transmission, remains one of the most consistently reported [34, 42]. It is not known, however, whether the high infection rate among premature infants reflects maternal or fetal factors and cause or effect. In our study, we also found prematurity to be an independent risk factor. Although only 4% of women in our study gave birth at !37 weeks’ gestation, the transmission rate in this group was 55%. This likely represents a subgroup of women for whom interventions to reduce perinatal transmission might be most difficult, especially for antiretroviral interventions targeted late in gestation or at the time of delivery. Since the report that first-born twins may be at increased risk for infection [43], attention has focused on preventable exposures during delivery and the possible protective effect of cesarean section. Although several analyses have suggested that cesarean section is protective, other studies have not found clear evidence for protection [8, 11, 36, 44]. In our study, with a low cesarean section rate (12%) and few elective cesareans, multivariate analysis indicated that cesarean section was protective, independent of virus load. Very recent information from a large meta-analysis indicates that elective cesarean section is protective [45], while the French Perinatal Cohort has reported a strong protective interaction between elective cesarean section and zidovudine prophylaxis [46]. Although our data clearly indicate the importance of virus load as a determinant of transmission, several caveats should be mentioned. The relatively small sample size or infrequent occurrence of certain factors in our study population (e.g., lack of maternal sexually transmitted diseases, few elective cesareans, few women with very low CD41 cell counts) may have led us to underestimate or not identify other risk factors. In addition, while our analysis indicates the theoretical reduction in transmission that might be achieved by lowering virus load to various target levels, the lack of direct knowledge regarding cause and effect of virus load at delivery and transmission precludes knowing whether a specific intervention that reduces virus load to a given target level will have the hypothesized

597

effect. This analysis would need to be done in the context of clinical trials. Furthermore, observational data and clinical trial data may not yield the same results, as evidenced by the fact that the ACTG 076 trial did not find the same causal relationship between transmission and virus load. Newer clinical trials may shed additional light on this. Finally, although we think of the perinatal HIV epidemic in global terms, the degree to which findings in one population with a given viral subtype, in this case subtype E, and specific risk factor profile may be generalizable to other populations needs further analysis. In summary, our study provides specific information on HIV perinatal transmission with a distinct genetic subtype, which may be generalizable to perinatal transmission worldwide. Transmission with HIV-1 subtype E appears to be similar to rates reported for subtype B. While the risk of transmission is multifactorial, our study suggests that maternal virus load is the single most important determinant of transmission. These findings may be useful in designing simplified interventions to reduce perinatal transmission in developing countries and may help explain the 50% reduction in perinatal HIV transmission recently announced at these same study hospitals, at which a short, oral zidovudine regimen late in pregnancy was used [47]. Acknowledgments We thank Veronique Batter, Linda Carr, and Meade Morgan (CDC) for assistance with data management and analysis, Jennifer Rapier for DNA PCR testing, Richard Respess (CDC) for guidance on virus load testing, Chou-Pong Pau (CDC) for assistance with peptide subtyping, R. J. Simonds and Martha Rogers (CDC) for critical review of the manuscript, Marie Morgan (CDC) for editorial assistance, Suchitra Nimmannitya (former director, Children’s Hospital, Bangkok) for encouragement to begin this project, and Timothy Dondero, Phillip Nieburg, and Helene Gayle (CDC) for their overall support. We also gratefully acknowledge the dedicated field work of the study nurses and social workers: Kanchana Neeyapun, Paichit Tothong, and Suchinda Pinyovanichkul (team leaders); Nantiya Chookaew, Sununta Henchaichon, Sujira Jalanchavanapate, Bongkoch Jetsawang, Kunyarat Klumthanom, Sukanya Phurksakusamesuk, Chariya Prasert, Supanee Samsukkree, Sumaleelak Sorapipatana, Suratsavadee Suwanmaitre, and Chanidapa Yuvasevee. Collaborative Group Members The Bangkok Collaborative Perinatal HIV Transmission Study Group comprises the following institutions and people who participated in this study: HIV/AIDS Collaboration, Nonthaburi, and CDC, Atlanta (Nathan Shaffer, Nancy Young, Timothy Mastro, Bruce Weniger); HIV/AIDS Collaboration, Nonthaburi (Khanchit Limpakarnjanarat, Philip Mock, Waranee Pokapanichwong, Jirasak Laosakkitiboran, Prayoon Yuentrakul); Faculty of Medicine, Siriraj Hospital, Department of Obstetrics/Gynecology, Bangkok (Sommai Toongsuwan, Prapas Bhiraleus, Chaiporn Bhadrakom, Pongsakdi Chaisilwattana, Anuvat Roongpisuthipong, Pattrawan Chaiyakul); Faculty of Medicine, Siriraj Hospital, Department of Pediatrics, Bangkok (Montri Tuchinda, Sanay Chearskul, Nirun Wanprapa); Faculty of Medicine,

598

Shaffer et al.

Siriraj Hospital, Department of Microbiology, Bangkok (Chantapong Wasi); Rajavithi Hospital, Department of Obstetrics/Gynecology, Bangkok, (Pratharn Chinayon, Wimol Siriwasin, Yunyong Mangclaviraj); Children’s Hospital, Bangkok (Tawee Chotpitayasunondh, Sunthorn Horpaopan, Varaporn Sangtaweesin, Wanida Suteewan).

References 1. Weniger BG, Takebe Y, Ou CY, Yamazaki S. The molecular epidemiology of HIV in Asia. AIDS 1994; 8(suppl 2):S13–8. 2. Mason CJ, Markowitz LE, Kitsiripornchai S, et al. Declining prevalence of HIV-1 infection in young Thai men. AIDS 1995; 9:1061–5. 3. Nelson KE, Celentano DD, Eiumtrakol S, et al. Changes in sexual behavior and a decline in HIV infection among young men in Thailand. N Engl J Med 1996; 335:297–303. 4. Soto-Ramirez LE, Renjifo B, McLane MF, et al. HIV-1 Langerhans’ cell tropism associated with heterosexual transmission of HIV. Science 1996; 271:1291–3. 5. Working Group on Mother-to-Child Transmission of HIV. Rates of motherto-child transmission of HIV-1 in Africa, America, and Europe: results from 13 perinatal studies. J Acquir Immune Defic Syndr Hum Retrovirol 1995; 8:506–10. 6. Connor EM, Sperling RS, Gelber RD, et al. Reduction of maternal-infant transmission of human immunodeficiency virus type 1 with zidovudine treatment. N Engl J Med 1994; 331:1173–80. 7. Thomas PA, Weedon J, Krasinski K, et al. Maternal predictors of perinatal human immunodeficiency virus transmission. Pediatr Infect Dis J 1994; 13:489–95. 8. Landesman SH, Kalish LA, Burns DN, et al. Obstetrical factors and the transmission of human immunodeficiency virus type 1 from mother to child. N Engl J Med 1996; 334:1617–23. 9. Mayaux MJ, Blanche S, Rouzioux C, et al. Maternal factors associated with perinatal HIV-1 transmission: the French Cohort Study; 7 years of followup observation. J Acquir Immune Defic Syndr Hum Retrovirol 1995; 8: 188–94. 10. Bryson YJ. Perinatal HIV-1 transmission: recent advances and therapeutic interventions. AIDS 1996; 10(suppl 3):S33–42. 11. Dunn DT, Newell ML, Mayaux MJ, et al. Mode of delivery and vertical transmission of human immunodeficiency virus type 1: a review of prospective studies. J Infect Dis 1994; 7:1064–6. 12. Fang G, Burger H, Grimson R, et al. Maternal plasma human immunodeficiency virus type 1 RNA level: a determinant and projected threshold for mother-to-child transmission. Proc Natl Acad Sci USA 1995; 92: 12100–4. 13. Weiser B, Nachman S, Tropper P, et al. Quantitation of human immunodeficiency virus type 1 during pregnancy: relationship of viral titer to mother-to-child transmission and stability of viral load. Proc Natl Acad Sci USA 1994; 91:8037–41. 14. Dickover RE, Garatty EM, Herman SA, et al. Identification of levels of maternal HIV-1 RNA associated with risk of perinatal transmission. JAMA 1996; 275:599–605. 15. Sperling RS, Shapiro DE, Coombs RW, et al. Maternal viral load, zidovudine treatment, and the risk of transmission of human immunodeficiency virus type 1 from mother to infant. N Engl J Med 1996; 335:1621–9. 16. Cao Y, Krogstad P, Korber B, et al. Maternal HIV-1 viral load and vertical transmission of infection: the Ariel Project for the prevention of HIV transmission from mother to infant. Nat Med 1997; 3:549–52. 17. Thea DM, Steketee RW, Pliner V, et al. The effect of maternal viral load on the risk of perinatal transmission of HIV-1. AIDS 1997; 11:437–44. 18. Mayaux MJ, Dussaix E, Isopet J, et al. Maternal virus load during pregnancy and mother-to-child transmission of human immunodeficiency virus type 1: the French perinatal cohort studies. J Infect Dis 1997; 175:172–5. 19. Burns DN, Landesman S, Wright DJ, et al. Influence of other maternal

JID 1999;179 (March)

variables on the relationship between maternal virus load and mother-toinfant transmission of human immunodeficiency virus type 1. J Infect Dis 1997; 175:1206–10. 20. Siriwasin W, Shaffer N, Roongpisuthipong A, et al. HIV prevalence, risk factors, and partner serodiscordance among pregnant women, Bangkok, Thailand. JAMA 1998; 280:49–54. 21. Schochetman G, Subbarao S, Kalish ML. Methods for studying genetic variation of the human immunodeficiency virus (HIV). In: Adolph KW, ed. Viral genome methods. Boca Raton, FL: CRC Press, 1996:25–41. 22. Michael N, Robb M, Birx D, et al. Performance of the Amplicor HIV-1 Monitor test and a modified HIV-1 Monitor test on HIV-1 subtypes A to F (abstract 279). In: 4th Conference on Retroviruses and Opportunistic Infections: program and abstracts (Washington, DC). Alexandria, VA: Infectious Diseases Society of America, 1997. 23. Parekh B, Phillips S, Granade TC, Baggs J, Hu D, Respess R. Impact of HIV-1 subtype variation on viral RNA quantitation. AIDS Res Hum Retroviruses 1999; 15:133–42. 24. Centers for Disease Control and Prevention. 1997 Revised guidelines for performing CD41 T-cell determinations in persons infected with human immunodeficiency virus (HIV). MMWR Morb Mortal Wkly Rep 1997; 46(RR-2):1–29. 25. Pau CP, Lee-Thomas S, Auwanit W, et al. Highly specific V-3 peptide enzyme immunoassay for serotyping HIV-1 specimens from Thailand. AIDS 1993; 7:337–40. 26. Centers for Disease Control and Prevention. 1994 Revised classification system for human immunodeficiency virus infection in children less than 13 years of age. MMWR Morb Mortal Wkly Rep 1994; 43(RR-12):1–19. 27. Bryson YJ, Luzuriaga K, Sullivan JL, Wara DW. Proposed definitions for in utero versus intrapartum transmission of HIV-1. N Engl J Med 1992; 327:1246–7. 28. Kuhn L, Abrams EJ, Matheson PB, et al. Timing of maternal-infant transmission: associations between intrapartum factors and early polymerase chain reaction results. AIDS 1997; 11:429–35. 29. Chaisilwattana P, Chuachoowong R, Siriwasin W, et al. Chlamydial and gonococcal cervicitis in HIV-seropositive and HIV-seronegative pregnant women in Bangkok. Sex Transm Dis 1997; 24:495–502. 30. Cleveland WS. Robust locally weighted regression and smoothing scatterplots. J Am Stat Assoc 1979; 74:829–36. 31. Deubner DC, Wilkinson WE, Helms MJ, Tyroler HA, Hames CG. Logistic model estimation of death attributable to risk factors for cardiovascular disease in Evans County, Georgia. Am J Epidemiol 1980; 112:135–43. 32. Greenland S, Drescher K. Maximum likelihood estimation of the attributable fraction from logistic models. Biometrics 1993; 49:865–72. 33. Adjorlolo-Johnson G, De Cock KM, Ekpini E, et al. Prospective comparison of mother-to-child transmission of HIV-1 and HIV-2 in Abidjan, Ivory Coast. JAMA 1994; 272:462–6. 34. European Collaborative Study. Risk factors for mother-to-child transmission of HIV-1. Lancet 1992; 339:1007–12. 35. St. Louis ME, Kamenga M, Brown C, et al. Risk for perinatal HIV-1 transmission according to maternal immunologic, virologic, and placental factors. JAMA 1993; 269:2853–9. 36. Simonds RJ, Steketee R, Nesheim S, et al. Impact of zidovudine use on risk and risk factors for perinatal transmission of HIV. AIDS 1998; 12:301–8. 37. Pitt J, Brambilla D, Reichelderfer P, et al. Maternal immunologic and virologic risk factors for infant human immunodeficiency virus type 1 infection: findings from the Women and Infants Transmission Study. J Infect Dis 1997; 175:567–75. 38. Moretta L, Ciccone E, Mingari MC, et al. Human natural killer cells: origin, clonality, specificity, and receptors. Adv Immunol 1994; 55:341–80. 39. Gumperz J, Parham P. The enigma of the natural killer cell. Nature 1995; 378:245–8. 40. Whiteside TL, Herberman RB. Role of human natural killer cells in health and disease. Clin Diagn Lab Immunol 1994; 1:125–33. 41. Lucia B, Jennings C, Cauda R, et al. Evidence of a selective depletion of a

JID 1999;179 (March)

42.

43. 44. 45.

HIV Load and Perinatal Transmission, Thailand

CD161 CD561 CD81 natural killer cell subset during HIV infection. Cytometry 1995; 22:10–15. Goedert JJ, Mendez H, Drummond JE, et al. Mother-to-infant transmission of human immunodeficiency virus type 1: association with prematurity or low anti-gp120. Lancet 1989; 2:1351–4. Goedert JJ, Duliege AM, Amos CI, et al. High risk of HIV-1 infection for first-born twins. Lancet 1991; 338:1471–5. European Collaborative Study. Vertical transmission of HIV-1: maternal immune status and obstetric factors. AIDS 1996; 10:1675–81. Read J. Mode of delivery and vertical transmission of HIV-1: a meta-analysis

599

from fifteen prospective cohort studies (the International Perinatal HIV Group) [abstract 23275]. Presented at the 12th World AIDS Conference, Geneva, June 1998. 46. Mandelbrot L, Le Chenadec J, Berrebi A, et al. Perinatal HIV-1 transmission: interaction between zidovudine prophylaxis and mode of delivery in the French Perinatal Cohort. JAMA 1998; 280:55–60. 47. Centers for Disease Control and Prevention. Administration of zidovudine during late pregnancy and delivery to prevent perinatal HIV transmission—Thailand, 1996–1998. MMWR Morb Mortal Wkly Rep 1998; 47: 151–4.