Original article Antiretroviral therapy in pregnancy: balancing the risk of preterm delivery with prevention of mother-to-child HIV transmission

Transcription

1 Antiviral Therapy 2 5: (doi:.385/imp63 Original article Antiretroviral therapy in pregnancy: balancing the risk of preterm delivery with prevention of mother-to-child HIV transmission Claire L Townsend *, Pat A Tookey, Marie-Louise Newell,2, Mario Cortina-Borja Medical Research Council Centre of Epidemiology for Child Health, UCL Institute of Child Health, University College London, London, UK 2 Africa Centre for Health and Population Studies, University of KwaZulu-Natal, Mtubatuba, South Africa *Corresponding author c.townsend@ich.ucl.ac.uk Background: Highly active antiretroviral therapy (HAART for pregnant HIV-positive women reduces the risk of mother-to-child transmission, but is associated with an increased risk of preterm delivery (<37 weeks gestation. We aimed to quantify the incremental risk benefit ratio for HAART compared with zidovudine monotherapy with respect to these outcomes. Methods: Two-stage Monte Carlo simulation methods were used to estimate the risk benefit ratio for HAART in pregnancy. Estimates of mother-to-child transmission and preterm delivery rates were obtained from UK and Ireland surveillance data collected through the National Study of HIV in Pregnancy and Childhood. Results: At a population level, HAART was associated with a more than sevenfold reduction in mother-to-child transmission compared with zidovudine monotherapy (adjusted odds ratio [AOR].3, 95% confidence interval [CI].6.27, but with a.4-fold increased odds of preterm delivery (AOR.43, 95% CI..86 and twofold increased odds of severe preterm delivery (<32 weeks; AOR 2.6, 95% CI The incremental risk benefit ratio for HAART in pregnancy compared with monotherapy was.63 (95% simulation interval.6.96 additional preterm births and.23 (95% simulation interval severe preterm births for each infection prevented. Conclusions: It is estimated that for every HIV transmissions prevented through the use of HAART (rather than monotherapy, 63 additional preterm deliveries would occur, including 23 at <32 weeks gestation. Interpretation of these ratios is context-dependent and requires additional information about morbidity, mortality and costs associated with the outcomes. Introduction The use of antiretroviral therapy in pregnancy substantially reduces the risk of mother-to-child HIV transmission (MTCT. In 994, zidovudine was shown in the AIDS Clinical Trial Group 76 trial to reduce the risk of transmission by two-thirds in the intervention arm (8% compared with the untreated control arm (25%, when taken in pregnancy and at delivery, and administered to the infant for the first 6 weeks of life []. Transmission rates in observational studies of non-breastfeeding women receiving zidovudine in the mid-99s ranged from 4 to % [2 4]. The risk of MTCT can be further reduced by the use of highly active antiretroviral therapy (HAART, which consists of three or more antiretroviral drugs, usually from at least two different classes; although there have been no clinical trials of HAART in pregnancy, transmission rates of 2% have been reported in large observational studies [3,5 7]. Despite the benefits of HAART, evidence suggesting an association with preterm delivery has emerged, and concerns about other adverse effects in pregnancy and early life have been raised [8]. A number of European studies demonstrated an increased risk of preterm delivery associated with combination antiretroviral therapy in comparison with less complex regimens [9 3], and a putative mechanism involving pregnancy- and HIVrelated changes in the cytokine environment has been suggested [4]. Although some studies from the US have not observed this association [5,6], others have reported a link with protease inhibitor-based HAART [7,8]. More recently, a study in Côte d Ivoire demonstrated an increased prevalence of low birth weight in infants exposed to HAART, compared with those exposed to zidovudine in an earlier time period [9]. Although information on gestational age was not 2 International Medical Press (print (online 775

2 CL Townsend et al. available in the Côte d Ivoire study, this observation was probably driven by preterm delivery. Indeed, findings from the French Perinatal Cohort suggest that HAART is not associated with a low birth weight z-score (that is, birth weight adjusted for gestational age [2]. Concerns about adverse effects in pregnancy have provided the rationale for reducing HAART exposure where possible. British guidelines suggest that zidovudine monotherapy can be used for prevention of MTCT (combined with elective caesarean section delivery for women who do not need HAART for their own health and have a baseline HIV RNA plasma viral load of <, copies/ml [2]. This approach appears to be effective, with no transmissions occurring among the 468 women who received zidovudine monotherapy and delivered by elective caesarean section between 2 and 26 in the UK and Ireland [7]. Quantification of the risks and benefits of HAART in pregnancy can inform the debate about optimal management of HIV infection in pregnancy, and has been explored using decision analysis to investigate the risks of mitochondrial toxicity, another possible HAARTassociated side effect, in an African context [22]. Alternatively, probabilistic simulation modelling has been proposed as a method for simultaneously estimating the risks and benefits of treatment and uncertainty around them [23,24]. In this paper, we estimate the risk benefit ratio for HAART in pregnancy in terms of preterm delivery and MTCT at a population level, based on UK and Ireland surveillance data, and use Monte Carlo simulation methods to quantify the joint variability around the estimates. Methods We used a two-stage probabilistic simulation approach for risk benefit analysis. The first stage involved obtaining estimates of the risk and benefit probabilities by treatment. In the second stage, Monte Carlo methods were used to simulate from the probability distributions, thus propagating the uncertainties around the estimates. Model estimates Estimates of preterm delivery and MTCT were based on data from the National Study of HIV in Pregnancy and Childhood (NSHPC, an ongoing, comprehensive, active surveillance study of pregnancies in HIV-positive women in the UK and Ireland (London Multi- Centre Research Ethics Committee approval, 24, MREC/4/2/9 [7,]. Logistic regression analyses were carried out in Stata version. (StataCorp LP, College Station, TX, USA. Full study methodology and details of similar analyses are described elsewhere [7,]. HAART was defined as three or more antiretroviral drugs. Preterm delivery estimates were adapted from previous NSHPC analyses [25,26]. Deliveries at <37 weeks gestation were defined as preterm and those at <32 weeks gestation as severely preterm [27]. Logistic regression models for the association between antiretroviral therapy and preterm delivery were fitted, adjusting for year of delivery, maternal ethnicity (White, Black or other, region of birth (UK/Ireland or elsewhere, injecting drug use (as the source of HIV acquisition and clinical status (asymptomatic or HIV-related symptoms. As there was no change over time in preterm delivery rates among women on monotherapy (trend test, P=.47 or on HAART (P=.2, all pregnancies occurring between 99 and 26 were included. MTCT rates were derived specifically for these analyses; although transmission rates in this population have been previously published [7], the appropriate comparisons required for these analyses have not been presented. In order to estimate the effect of HAART relative to zidovudine monotherapy at a population level (that is, assuming that all infected pregnant women are treated with either HAART or monotherapy regardless of other factors, the association between treatment type and MTCT was explored by comparing births to women on HAART at any time between 996 and 26 with those to women on monotherapy before 998, when HAART was uncommon (only five cases between 996 and 997. Births to women on monotherapy since 998 were excluded because of the selective use of this treatment in recent years for women who did not require HAART for their own health and delivered by elective caesarean section. Logistic regression models were adjusted for mode of delivery (elective caesarean section, emergency caesarean section or vaginal delivery, gestational age (categorized as 37 weeks, weeks, weeks or <32 weeks and sex of the child. Because of the differences in time periods between the preterm delivery and MTCT analyses, it was not possible to use a common dataset for risk and benefit estimates. For the baseline scenario, observed preterm delivery and MTCT rates in women on monotherapy were used. To estimate rates in the HAART group, adjusting for confounders, the definition of an odds ratio (OR used was OR=(p /[-p ]/(p /[-p ], where p is the rate in the unexposed or baseline group, and p is the rate in the exposed group [28]. Adjusted preterm delivery and transmission rates for the HAART group were derived by solving the equation for p and substituting the adjusted odds ratio (AOR estimated in logistic regression models for the OR, as follows: AOR p p = ( -p +(AOR p International Medical Press

3 Preterm delivery risk and MTCT Incremental risk benefit ratio Incremental risks and benefits were calculated as p -p (rate in HAART-exposed women minus rate in monotherapy-exposed women; R=R -R denoted incremental risks, where R corresponded to the risk in the exposed group and R to the risk in the baseline group; B=B -B denoted incremental benefits, where B corresponded to the benefit in the baseline group and B to the benefit in the exposed group [23]. The incremental risk benefit ratio (IRBR is defined as: IRBR= = R -R B -B (2 Monte Carlo simulation Calculating confidence intervals (CIs for IRBRs is problematic because ratios can be negative and tend to be non-normally distributed [29]; therefore, in order to incorporate (simultaneously the uncertainty from the risk and benefit estimates, Monte Carlo methods similar to those suggested by Lynd and O Brien [23] were used to simulate the joint probability densities of incremental benefit (MTCT and incremental risk (preterm/severe preterm delivery. Values for proportions of both preterm delivery and MTCT were generated from appropriate beta probability density functions [3] in order to account for the fact that normality cannot always be assumed when calculating CIs around proportions [23]. To model the uncertainty relating to the AOR, the estimate of its natural logarithm, that is ln(aor, was used because it tends to be normally distributed. Analyses were carried out in R version 2.8. (R Foundation for Statistical Computing, Vienna, Austria [3]. The simulations consisted of repeatedly sampling from the assumed parameters distributions by randomly selecting values from the beta distributions for the baseline proportions and from the normal distributions for ln(aor, using the algorithm shown in Figure. A value for the baseline risk (, denoting sampled R was sampled from the distribution of preterm delivery for monotherapy-exposed women, and a value for the baseline benefit ( was sampled from the MTCT distribution for monotherapy-exposed women. This was replicated 5, times in order to simulate the joint uncertainty around the estimates. The distribution of ln(aor for the risks and benefits was also randomly sampled 5, times, and rates for the HAART group ( and were calculated from Equation using the point estimates for risk and benefit rates in the baseline group (R and B, and the ln(aor s. For each of the 5, realizations, the incremental risk ( - and benefit ( - were calculated, and the IRBR was computed. The distribution of the IRBR was thus approximated. The same random number seed was fixed at the start of each simulation to enable simulations to be replicated. Simulation intervals were obtained by approximating the 2.5% and 97.5% quantiles of the distributions generated in the simulation. Assumptions Adverse effects other than preterm delivery, such as maternal drug-related toxicity, the effect of mode of delivery or paediatric outcomes (such as mitochondrial toxicity or anaemia were not considered in these models. Benefits other than the effect on MTCT, for instance, slowing of maternal disease progression, were also omitted. Although AORs were used in the simulation to control for other potential risk factors, baseline rates related to the population of HIV-infected pregnant women in the UK and Ireland. Furthermore, these models assumed a homogeneous population, and differential effects within population subgroups were ignored. Sensitivity analyses Although some studies have shown an increased risk of transmission in preterm infants, prematurity was not an independent risk factor for MTCT among women on HAART in this population [7]; however, the effect of an association between preterm delivery and MTCT was investigated in a sensitivity analysis. Models were adapted by incorporating a relative risk, varying from to 4, for the increase in MTCT resulting from preterm delivery; details of this methodology are available elsewhere [25]. Results Risk and benefit estimates Estimates for preterm delivery rates were obtained from previous analyses [25,26] and are shown in Table. HAART was associated with a.43-fold increased odds of preterm delivery (95% CI..86; P=. compared with monotherapy among women exposed between 99 and 26, after adjusting for year of delivery, maternal ethnicity, region of birth, injecting drug use and clinical symptoms [26]. This effect did not vary according to inclusion of a protease inhibitor (AOR.42 versus.45 for HAART without protease inhibitor. Estimates for MTCT rates were obtained from logistic regression models. HAART was associated with an 87% reduction in MTCT compared with pre-998 monotherapy (AOR.3, 95% CI.6.27; P<; n=4,39 after adjusting for mode of delivery, gestational age and sex (Table. Among women on monotherapy, the rate of MTCT did not vary significantly by year or by maternal HIV symptoms. Although Antiviral Therapy

4 CL Townsend et al. Figure. Flowchart showing the simulation algorithm for the Monte Carlo risk benefit model of antiretroviral therapy in pregnancy HAART simulation Preterm delivery rate (R MTCT rate (B Sample simulation Sample In(AOR for preterm delivery Equation In(AOR for MTCT HAART HAART Preterm delivery rate (Rˆ MTCT rate (Bˆ Preterm delivery rate (Rˆ MTCT rate (Bˆ Incremental preterm Incremental MTCT delivery rate (Rˆ -Rˆ rate (Bˆ -Bˆ Point estimate Sampled Calculated Incremental risk benefit ratio Rˆ -Rˆ ( Bˆ -Bˆ From Equation 2 (substituting sampled rates B, benefit in baseline group; Bˆ, sampled benefit in baseline group; Bˆ, sampled benefit in exposed group; HAART, highly active antiretroviral therapy; ln(aor, natural logarithm of the adjusted odds ratio; MTCT, mother-to-child transmission; R, risk in baseline group; Rˆ, sampled risk in baseline group; Rˆ, sampled risk in exposed group. Table. Preterm delivery and mother-to-child transmission estimates Measurement n Estimate (95% CI a se Preterm delivery (<37 weeks gestation Baseline rate, monotherapy (R 98/957.2 ( Unadjusted rate, HAART 596/4, ( Adjusted rate, HAART (R 4. AOR (HAART versus monotherapy b.43 ( Severe preterm delivery (<32 weeks gestation Baseline rate, monotherapy (R 3/957.4 ( Unadjusted rate, HAART 2/4, ( Adjusted rate, HAART (R 2.8 AOR (HAART versus monotherapy b 2.6 ( Mother-to-child transmission c Baseline rate, monotherapy (B ; pre-998 2/72 7. ( Unadjusted rate, HAART 42/4,282. ( Adjusted rate, HAART (R. AOR (HAART versus monotherapy d.3 ( a Estimates presented as percentages unless otherwise indicated. b Adjusted odds ratio (AOR based on comparison of highly active antiretroviral therapy (HAART at any time ( with monotherapy at any time (99 26, adjusted for year of delivery, maternal ethnicity (White, Black or other, region of birth (UK/Ireland or elsewhere, injecting drug use (as source of HIV acquisition and clinical status (asymptomatic or HIV-related symptoms. c Estimates were obtained from the same dataset as that used in Townsend et al. [7], but with the exclusion of births to women on monotherapy since 2 (n=638, and the addition of births to women on monotherapy before 998 (n=72 or HAART before 2 (n=62. d AOR based on comparison of HAART at any time (99 26 with monotherapy in the pre-haart era (before 998. AOR adjusted for mode of delivery, gestational age and sex. B, benefit in baseline group; CI, confidence interval; R, risk in baseline group; R, risk in exposed group International Medical Press

5 Preterm delivery risk and MTCT Figure 2. Beta distributions describing the probability of preterm delivery and MTCT in women on monotherapy A Beta distribution for preterm delivery <37 weeks B Beta distribution for MTCT Probability density Mean =.2 SD= Probability density Mean =.7 SD= Probability of preterm delivery (<37 weeks Probability of MTCT MTCT, mother-to-child transmission. the transmission rate in this analysis was higher in severely preterm infants (<32 weeks gestation than in full-term infants (6.7 versus 2.7%, there was no statistically significant association in the multivariable analysis (AOR.4 for severe preterm versus full-term, 95% CI ; P=.86. The beta probability distributions used to model baseline preterm delivery and MTCT rates are shown in Figure 2. Incremental risk benefit ratio The incremental benefit associated with exclusive HAART (versus exclusive monotherapy was a reduction in MTCT of 6.% ( B=7.%-.%, and the incremental risk was an increase in preterm delivery of 3.8% ( R=4.%-.2% for preterm and.4% for severe preterm delivery ( R=2.8%-.4%. The estimated IRBR (from Equation 2 was.63 preterm infants and.23 very preterm infants per HIV infection averted. Simulation The joint probability densities of preterm delivery and MTCT are shown in Figure 3A. A clear difference between the two treatment groups is apparent. The simulation was repeated for severe preterm delivery and a similar pattern was observed (Figure 3B. The incremental risk benefit pairs were then computed using the simulated rates. The median risk benefit ratios were.65 preterm and.25 very preterm infants for each HIV infection averted. Although these values were very close to the point estimates (.63 and.23, respectively, the means tended to be higher (.74 and.29, respectively because of the right-skewed distribution of the ratios. Figure 4 shows the incremental risks and benefits for the simulations on a risk benefit plane, with the x-axis representing the difference in the probability of MTCT ( B occurring with HAART relative to monotherapy, and the y-axis representing the difference in the probability of preterm or severe preterm delivery ( R. The red and blue lines in Figure 4 correspond to acceptability thresholds (µ, or the number of adverse events we would be willing to accept for one additional beneficial event. At the line µ=, one additional premature birth would be acceptable in order to prevent one additional transmission. This condition is only satisfied for the points lying below this line; thus, the proportion of points below the line estimates the probability that the condition (each beneficial event corresponding to one adverse event is satisfied. These lines can also be used to indicate confidence limits; the upper and lower red lines indicate the limits between which 95% of the points fall, and correspond to µ=.6 and µ=.96 in the preterm delivery model, and µ=-.2 and µ=.88 in the severe preterm delivery model. In the preterm delivery model, 95% of the points fall below the line with slope µ=.6; the corresponding line for the severe preterm delivery model has a slope of µ=.74. There is, therefore, a 95% chance that for each transmission avoided, the maximum number of additional preterm infants would be.6 and the maximum number of severely preterm infants would be.74. Sensitivity analyses Modelling up to a fourfold increased risk of transmission associated with preterm delivery produced only a small shift in the joint densities. The IRBRs did not change and CIs varied only marginally compared with the unadjusted model [25]. Antiviral Therapy

6 CL Townsend et al. Discussion Although awareness of the risks associated with any treatment is important, these must be considered in relation to the corresponding benefits in order for conclusions about the value of that treatment to be made. A risk benefit model based on two-stage Monte Carlo methods was developed to enable the observed risks of preterm delivery to be quantified in relation to the main benefit of HAART in pregnancy, a reduced risk of MTCT. Comparison was made between HAART (any time and zidovudine monotherapy in the pre-haart era because use of monotherapy in recent years was restricted to women with a low baseline risk of transmission. This comparison enabled the full extent of the benefits associated with HAART to be quantified at a population level in the context of the reported risks. These results, which were based on national surveillance data from the UK and Ireland, suggest that avoiding transmissions by using HAART rather than monotherapy would result in an additional 63 preterm births (95% CI 6 96, 23 of which would occur before 32 weeks gestation (95% CI -2 88, assuming a.4- and 2-fold increase in the risk of preterm and severe preterm delivery, respectively, and a 7-fold decrease in the risk of MTCT associated with HAART. These findings were not affected by allowing for an increased risk of transmission among preterm infants, as demonstrated in a sensitivity analysis. Interpretation of these findings depends on the choice of acceptability threshold, that is, how many preterm deliveries, and particularly severe preterm deliveries, are we willing to accept in order to prevent one perinatal transmission? In this UK and Ireland context, if these outcomes were equal in severity, HAART would be considered superior to exclusive monotherapy because more infants avoid infection with this approach than are born prematurely. However, further information is required to determine an appropriate risk benefit acceptability threshold for these outcomes. Preterm delivery is associated with a significantly increased risk of perinatal morbidity and mortality, although this is generally confined to very premature infants [32]. HIV infection has serious lifelong consequences, however, and ultimately results in progression to AIDS and death. Although the advent of effective antiretroviral drugs has improved prognosis for HIV-infected children in resource-rich countries, complications relating to toxicity or resistance might arise [33,34]. The appropriate balance of risks and benefits will probably vary by population and healthcare setting. In resource-poor countries, both premature birth and HIV infection will be associated with higher rates of paediatric morbidity and mortality than in resource-rich settings, and the relative consequences and costs of these outcomes will differ according to the quality of neonatal Figure 3. Joint densities of preterm delivery and MTCT by type of antiretroviral therapy resulting from 5, simulations Preterm delivery rate (<37 weeks Severe preterm delivery rate (<32 weeks A B HAART.5..5 MTCT rate HAART.5..5 MTCT rate (A Joint density of preterm delivery (<37 weeks and mother-to-child transmission (MTCT, and (B joint density of severe preterm delivery (<32 weeks and MTCT. HAART, highly active antiretroviral therapy. Note that the y-axis scales differ. care facilities and availability of antiretroviral therapy. Population-specific information on outcomes in preterm and HIV-infected children would ideally be factored into decisions regarding the acceptable risk benefit threshold. Finally, attitudes and preferences of both clinicians and patients towards different perinatal outcomes, which might be affected by the continued stigma of HIV, could also influence decision making. These findings relate to a population level approach assuming that either monotherapy or HAART was used 78 2 International Medical Press

7 Preterm delivery risk and MTCT Figure 4. Risk benefit plane showing the incremental risk of preterm delivery and severe preterm delivery relative to the incremental MTCT benefit for HAART compared with monotherapy Incremental risk (preterm delivery <37 weeks A Incremental benefit (MTCT Incremental risk (severe preterm delivery <32 weeks B Incremental benefit (MTCT..5 (A Preterm and (B severe preterm delivery incremental risks are shown relative to the mother-to-child transmission (MTCT incremental benefit. Points correspond to results of 5, simulations and lines represent acceptability thresholds (µ, with slope values shown. Red dots indicate point estimates of incremental risks and benefits. HAART, highly active antiretroviral therapy. Note that the y-axis scales differ. for all pregnant women. In practice, the range of treatments currently available means that in resource-rich countries, decisions on the type and timing of treatment can be based on individual clinical and immunological factors, desired mode of delivery and additional factors, such as past treatment experience and patient preference. Consequently, zidovudine monotherapy is now usually reserved for women with favourable baseline characteristics who do not require HAART for their own health [2,35]. Among pregnant women in the UK and Ireland, the MTCT rate in women on zidovudine monotherapy who delivered by elective caesarean section (%, 95% CI.8 was not significantly different from those among women on HAART who had a planned vaginal delivery (.7% or elective caesarean section (.7%; P=.5 [7]. With no significant difference in benefits, the reduced risk of preterm delivery associated with zidovudine monotherapy would suggest that this is an appropriate approach for preventing MTCT in this particular group of women who do not require HAART for their own health and deliver by elective caesarean section. Other factors, such as the risk of obstetric complications associated with different modes of delivery, might also need to be considered [36]. Because there have been no clinical trials of HAART in pregnancy, we used observational data from a large prospective surveillance study in the UK and Ireland to assess the odds of preterm delivery and MTCT associated with type of antiretroviral therapy. We used AORs to derive risk and benefit rates associated with HAART, which were adjusted for potential confounding. In a previous analysis of antiretroviral therapy and preterm delivery, we demonstrated that adjusting for maternal HIV-related symptoms or CD4 + T-cell count, in order to account for possible bias in indication for treatment by type of therapy (monotherapy or dual therapy versus HAART, did not substantially alter the association between HAART and preterm delivery []. Furthermore, the association between therapy and preterm delivery was of similar magnitude in as in 2 25, despite potential changes over time in other (unmeasured risk factors for preterm delivery. The effect estimates used in this analysis were adjusted for year of delivery, as well as for maternal HIV symptoms and demographic factors (ethnicity, region of birth and injecting drug use. Adjusting for year of delivery did not substantially alter the association between HAART and preterm delivery, suggesting that the observed association is unlikely to be the result of bias in treatment groups produced by changes over time in other risk factors for preterm delivery. Although the use of zidovudine monotherapy prior to 998 might have been more common among women with more advanced HIV disease, particularly in earlier years, the MTCT rate Antiviral Therapy

8 CL Townsend et al. in this group did not vary by year or maternal clinical status and was consistent with other reported rates [2 4]. Another limitation of this analysis is that we were unable to consider duration of HAART because of limited statistical power to detect a difference in preterm delivery rates according to trimester of exposure []. Determining optimal timing of initiation of HAART depends both on the potential for adverse effects and on the risk of MTCT, which is strongly associated with duration of treatment in pregnancy [6,7]. Further limitations of these surveillance data are discussed in detail elsewhere [7,]. The 95% simulation intervals around the risk benefit ratios were relatively wide, partly because the estimate of MTCT among women on monotherapy was based on only 72 women. Although we built our models using data from one population, estimates based on a summary of aggregated published data could be used, potentially resulting in tighter simulation intervals around the risk benefit ratios. In conclusion, we have estimated that the use of HAART in pregnancy at a population level would result in 63 (95% CI 6 96 additional preterm deliveries, 23 (95% CI at <32 weeks gestation, for every infections prevented, compared with the use of zidovudine monotherapy. However, among women who do not require treatment for their own health, monotherapy can be highly effective in reducing MTCT when combined with elective caesarean section delivery [7]. Further research is needed to clarify the optimal approach for preventing MTCT in this group of healthier women. Acknowledgements National surveillance of obstetric and paediatric HIV is undertaken through the NSHPC in collaboration with the Health Protection Agency Centre for Infections and Health Protection Scotland. We gratefully acknowledge the contribution of the midwives, obstetricians, genito-urinary physicians, paediatricians, clinical nurse specialists and all other colleagues who report to the NSHPC through the British Paediatric Surveillance Unit of the Royal College of Paediatrics and Child Health, and the obstetric reporting scheme run under the auspices of the Royal College of Obstetricians and Gynaecologists. We thank Janet Masters who coordinates the study and manages the data, and Icina Shakes and Hiwot Haile-Selassie for administrative support. We also thank Catherine Peckham for comments on this paper. CLT and MC-B wrote the programmes and carried out the analyses. CLT drafted the paper. All authors contributed to the interpretation of the results, commented on drafts of the paper and approved the final version. This paper was presented previously at the 6th Conference on Retroviruses and Opportunistic Infections (8 February 29, Montreal, QC, Canada. Disclosure statement CLT was funded by the Medical Research Council (MRC; grant number G5895. The NSHPC is currently funded by the Health Protection Agency (grant number GHP/3/3/3. This work was undertaken at the MRC Centre for Paediatric Epidemiology and Biostatistics, which benefits from funding support from the MRC in its capacity as the MRC Centre of Epidemiology for Child Health. The UCL Institute of Child Health, University College London, receives a proportion of funding from the Department of Health s National Institute for Health Research Biomedical Research Centres funding scheme. Any views expressed in this paper are those of the authors, and not necessarily those of the funders. The authors declare no other competing interests. References. Connor EM, Sperling RS, Gelber R, et al. Reduction of maternal-infant transmission of human immunodeficiency virus type with zidovudine treatment. N Engl J Med 994; 33: European Collaborative Study. HIV-infected pregnant women and vertical transmission in Europe since 986. AIDS 2; 5: Cooper ER, Charurat M, Mofenson L, et al. Combination antiretroviral strategies for the treatment of pregnant HIV--infected women and prevention of perinatal HIV- transmission. J Acquir Immune Defic Syndr 22; 29: Kind C, Rudin C, Siegrist C-A, et al. Prevention of vertical HIV transmission: additive protective effect of elective cesarean section and zidovudine prophylaxis. AIDS 998; 2: European Collaborative Study. Mother-to-child transmission of HIV infection in the era of highly active antiretroviral therapy. Clin Infect Dis 25; 4: Warszawski J, Tubiana R, Le Chenadec J, et al. Mother-tochild HIV transmission despite antiretroviral therapy in the ANRS French Perinatal Cohort. AIDS 28; 22: Townsend CL, Cortina-Borja M, Peckham CS, et al. Low rates of mother-to-child transmission of HIV following effective pregnancy interventions in the United Kingdom and Ireland, AIDS 28; 22: Thorne C, Newell ML. The safety of antiretroviral drugs in pregnancy. Expert Opin Drug Saf 25; 4: Thorne C, Patel D, Newell ML. Increased risk of adverse pregnancy outcomes in HAART-treated HIV infected women in Europe. AIDS 24; 8: Grosch-Woerner I, Puch K, Maier RF, et al. Increased rate of prematurity associated with antenatal antiretroviral therapy in a German/Austrian cohort of HIV--infected women. HIV Med 28; 9:6 3.. Townsend CL, Cortina-Borja M, Peckham CS, Tookey PA. Antiretroviral therapy and premature delivery in diagnosed HIV-infected women in the United Kingdom and Ireland. AIDS 27; 2: Lorenzi P, Spicher VM, Laubereau B, et al. Antiretroviral therapies in pregnancy: maternal, fetal and neonatal effects. AIDS 998; 2:F24 F International Medical Press

9 Preterm delivery risk and MTCT 3. Boer K, Nellen JF, Patel D, et al. The AmRo study: pregnancy outcome in HIV--infected women under effective highly active antiretroviral therapy and a policy of vaginal delivery. BJOG 27; 4: Fiore S, Newell ML, Trabattoni D, et al. Antiretroviral therapy-associated modulation of Th and Th2 immune responses in HIV-infected pregnant women. J Reprod Immunol 26; 7: Tuomala RE, Shapiro DE, Mofenson LM, et al. Antiretroviral therapy during pregnancy and the risk of an adverse outcome. N Engl J Med 22; 346: Watts DH, Balasubramanian R, Maupin RT, Jr., et al. Maternal toxicity and pregnancy complications in human immunodeficiency virus-infected women receiving antiretroviral therapy: PACTG 36. Am J Obstet Gynecol 24; 9: Cotter AM, Garcia AG, Duthely ML, Luke B, O Sullivan MJ. Is antiretroviral therapy during pregnancy associated with an increased risk of preterm delivery, low birth weight, or stillbirth? J Infect Dis 26; 93: Schulte J, Dominguez K, Sukalac T, Bohannon B, Fowler MG. Declines in low birth weight and preterm birth among infants who were born to HIV-infected women during an era of increased use of maternal antiretroviral drugs: pediatric spectrum of HIV disease, Pediatrics 27; 9:e9 e 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 28; 22: Briand N, Mandelbrot L, Le CJ, et al. No relation between in-utero exposure to HAART and intrauterine growth retardation. AIDS 29; 23: de Ruiter A, Mercey D, Anderson J, et al. British HIV Association and Children s HIV Association guidelines for the management of HIV infection in pregnant women 28. HIV Med 28; 9: Ciaranello AL, Seage GR, III, Freedberg KA, et al. Antiretroviral drugs for preventing mother-to-child transmission of HIV in sub-saharan Africa: balancing efficacy and infant toxicity. AIDS 28; 22: Lynd LD, O Brien BJ. Advances in risk-benefit evaluation using probabilistic simulation methods: an application to the prophylaxis of deep vein thrombosis. J Clin Epidemiol 24; 57: Spiegelhalter DJ, Abrams KR, Myles JP. Cost-effectiveness, policy making and regulation. Bayesian approaches to clinical trials and health-care evaluation. Chichester: John Wiley & Sons, Ltd 24; pp Townsend CL. Antiretroviral therapy and pregnancy outcome in HIV-infected women in the United Kingdom and Ireland. PhD thesis, University College London 29. (Accessed 3 July 2. Available from ac.uk/66/ 26. Townsend CL, Schulte J, Thorne C, et al. Antiretroviral therapy and preterm delivery a pooled analysis of data from the United States and Europe. BJOG 2. doi./ j x 27. Goldenberg RL, Culhane JF, Iams JD, Romero R. Epidemiology and causes of preterm birth. Lancet 28; 37: Kirkwood BR, Sterne JAC. Essential medical statistics. Oxford: Blackwell Publishing Briggs AH, O Brien BJ, Blackhouse G. Thinking outside the box: recent advances in the analysis and presentation of uncertainty in cost-effectiveness studies. Annu Rev Public Health 22; 23: Gupta AK, Nadarajah S. Handbook of the beta distribution and its applications. New York: Marcel Dekker R Development Core Team. R: a language and environment for statistical computing. Vienna: R Foundation for Statistical Computing Saigal S, Doyle LW. An overview of mortality and sequelae of preterm birth from infancy to adulthood. Lancet 28; 37: Giaquinto C, Morelli E, Fregonese F, et al. Current and future antiretroviral treatment options in paediatric HIV infection. Clin Drug Investig 28; 28: Judd A, Doerholt K, Tookey PA, et al. Morbidity, mortality, and response to treatment by children in the United Kingdom and Ireland with perinatally acquired HIV infection during : planning for teenage and adult care. Clin Infect Dis 27; 45: Panel on Treatment of HIV-infected Pregnant Women and Prevention of Perinatal Transmission. Recommendations for use of antiretroviral drugs in pregnant HIV--infected women for maternal health and interventions to reduce perinatal HIV transmission in the United States, 24 May 2. (Accessed 3 July 2. Available from aidsinfo.nih.gov/contentfiles/perinatalgl.pdf 36. Read JS, Newell ML. Efficacy and safety of cesarean delivery for prevention of mother-to-child transmission of HIV-. Cochrane Database Syst Rev 25; 4:CD5479. Accepted for publication 4 February 2 Antiviral Therapy