Trinh Duong, A E Ades, Diana M Gibb, Pat A Tookey, Janet Masters

Trinh Duong
statistician

A E Ades
reader

Pat A Tookey
senior research fellow

Janet Masters
research assistant

Diana M Gibb
senior lecturer in epidemiology

Correspondence to: A.E Ades a.ades@ich.ucl.ac.uk

Abstract

Objective To estimate and interpret time trends in vertical transmission rates for HIV using data from national obstetric and paediatric surveillance registers.

Design Prospective study of HIV infected women reported through obstetric surveillance. HIV infection status of the child and onset of AIDS were reported through paediatric surveillance. Rates of vertical transmission and progression to AIDS rate were estimated by methods that take account of incomplete follow up of children with indeterminate infection status and delay in AIDS reporting.

Setting British Isles

Subjects Pregnant women infected with HIV whose infection was diagnosed before delivery and their babies.

Main outcome measures Mother to child transmission of infection and progression to AIDS in children.

Results By January 1999, 800 children born to diagnosed HIV infected women who had not breast fed had been reported. Vertical transmission rates rose to 19.6% (95% confidence interval 8.0% to 32.5%) in 1993 before falling to 2.2% (0.0% to 7.8%) in 1998. Between 1995 and 1998 use of antiretroviral treatment increased significantly each year, reaching 97% of live births in 1998. The rate of elective caesarean section remained constant, at around 40%, up to 1997 but increased to 62% in 1998. Caesarean section and antiretroviral treatment together were estimated to reduce risk of transmission from 31.6% (13.6% to 52.2%) to 4.2% (0.8% to 8.5%). The proportion of infected children developing AIDS in the first 6 months fell from 17.7% (6.8% to 30.8%) before 1994 to 7.2% (0.0% to 15.7%) after, coinciding with increased use of prophylaxis against Pneumocystis carinii pneumonia.

Conclusions In the British Isles both HIV related morbidity and vertical transmission are being reduced through increased use of interventions.

Introduction

Randomised controlled trials have established that administration of zidovudine during pregnancy, at labour, and to newborn infants can significantly reduce the risk of vertical transmission of HIV. ⁽¹⁾ The protective effect of elective caesarean section has been confirmed by a recent meta-analysis of cohort studies ⁽²⁾ and a randomised controlled trial. ⁽³⁾ The combined effect of these interventions is reported to reduce transmission to 2% or less in some cohort studies. ^{(2) (4)}

However, the general population of infected women may differ from those recruited into trials or cohort studies in terms of adherence to antiretroviral treatment, previous exposure to antiretroviral treatment, and uptake of elective caesarean section. It is therefore essential to monitor uptake of interventions and vertical transmission rates in the wider population. Here we present estimates of HIV vertical transmission rates based on the combined 1984-98 obstetric and paediatric surveillance data in the British Isles. ⁽⁵⁾

Methods

Population studied

Since 1989 pregnant women in the British Isles known to be infected with HIV have been notified to the Royal College of Obstetricians and Gynaecologists. After notification, obstetricians are asked to report on outcome of pregnancy, and, since 1995, on mode of delivery and uptake of antiretroviral treatment. The child’s HIV infection status is ascertained through a parallel scheme run though the British paediatric surveillance unit. After initial notification of children born to infected mothers, paediatricians are asked for diagnostic and clinical data. Subsequently, children of indeterminate infection status are followed until infection status is known, usually within 12 months, and paediatricians are asked to report when an infected child develops AIDS or dies. Paediatric and obstetric reports can be linked, and the combined data form the basis for national paediatric HIV surveillance. Details have been published previously. ⁽⁵⁾ All reporting is voluntary and confidential. This analysis is confined to prospectively ascertained children whose mothers’ HIV infection was reported to have been diagnosed before delivery.

Statistical methods

DNA polymerase chain reaction tests now allow HIV-1 infection to be determined in the first 2 months of life. However, outside specialist centres diagnostic tests may be delayed, and in surveillance data their reporting is incomplete, so that many children are recorded as having indeterminate infection status, especially those born recently. Although children with indeterminate HIV status are often excluded in analyses of vertical transmission of HIV, ^{(6) (7) (8)} this can produce bias because the probability that a child of indeterminate infection status is infected depends on age. ⁽⁹⁾ We used an estimation-maximisation algorithm ⁽¹⁰⁾ to estimate the probability of infection for each indeterminate child based on when they were last known to be seropositive, and the fact that they had not yet been reported to have AIDS. This requires that the AIDS incubation time in infected children, the distribution of delay in AIDS reporting, and the time to antibody loss in uninfected children is known or estimated simultaneously. This is a modification and extension of earlier work, ⁽⁹⁾ and additionally allows for covariate effects on vertical transmission rate. Further details and a stata program are available from us.

The median time to antibody loss in uninfected children was assumed to be 9.8 months (interquartile range 7.8-12.2 months). ⁽⁹⁾ We estimated the distribution of delay in reporting AIDS in vertically infected children by Poisson regression, ⁽¹¹⁾ assuming that all AIDS diagnoses are reported within three years.

The rate of progression to AIDS among infected children was described by three variables, which were all estimated by the estimation-maximisation algorithm: the rate during the first 6 months of life before 1994, the rate during the first 6 months from 1994 onwards, and the overall rate after the first 6 months. Prophylaxis against Pneumocystis carinii pneumonia became routine during 1994. Data on P carinii prophylaxis in the British Isles are incomplete, but it was known to have been given to 21% of prospectively ascertained children before the beginning of 1994 and to 64% after. We therefore assumed that it was given from 6 weeks of age to all children from 1994 onwards according to national (and later US) guidelines, ^{(12) (13)} and to no children before 1994.

The estimation-maximisation algorithm was used to estimate yearly vertical transmission rates since 1984, and the effect of mode of delivery and antiretroviral treatment during 1995 to 1998. Confidence intervals were estimated by bootstrap sampling from the data (n=999). ⁽¹⁴⁾ Assumptions on date when P carinii prophylaxis became routine and on distribution of delay in AIDS reporting were subjected to sensitivity analysis.

Results

By the end of January 1999, a total of 822 prospectively ascertained children born to HIV infected women between 1984 and 1998 had been reported. Twenty two infants known to have been breast fed were excluded.

Vertical transmission rate

Of the remaining 800 children, 73 (9.1%) were known to be infected, 23 of whom were reported to have developed AIDS. An estimated 69% of AIDS reports were received within 6 months of the date of AIDS diagnosis; 87% within 12 months, and 99% within 24 months. In all, 488 (61%) children were uninfected and 239 (29.9%) had indeterminate infection status (table 1). Among the 120 infants born in 1998, 77 (64%) had indeterminate infection status. In all, 138 (66%) of the 239 infants of indeterminate infection had no antibody results at all and were assumed last known seropositive at birth. Among the remainder, age at the most recent antibody result ranged from 1 to 15 months, but most were under 3 months old. Most children recorded as having indeterminate status had been reported by an obstetrician, but the paediatric report had either not yet been received or had been forwarded before diagnosis. Thirty seven indeterminate children had left the British Isles or were otherwise lost to follow up; parents of another 12 refused testing.

Table 1  Infection status of infants born to diagnosed, HIV infected women between 1984 and 1998 who were not reported to have been breast fed and vertical transmission rates estimated with estimation-maximisation algorithm
 

*1984-6 estimate pooled, due to small numbers.
 

The crude overall transmission rate based only on children with known infection status is 13.0% (95% confidence interval 10.2% to 15.8%). If all indeterminate children were uninfected the transmission rate would be 9.1%. The overall estimation-maximisation algorithm estimate, which removes the bias caused by indeterminate children, is 12.1% (9.7% to 14.7%).

The transmission rate rose to 19.6% (8.0% to 32.5%) in 1993 and from 1995 decreased steeply to 2.2% (0.0% to 7.8%) in 1998 (fig 1). A cubic model gave a significantly better fit than the constant model (c²₃=14.1, P=0.003) and the linear model (14.1, P=0.0009), suggesting that transmission rate has changed over time but not in a linear fashion. The decline in transmission rate from 1995 onwards was significant (odds ratio of 0.51 a year; 95% confidence interval 0.27 to 0.77, c²₂=8.8, P=0.003)

Figure 1

(F1) Year on year estimates of vertical transmission rates and 95% confidence intervals for HIV among children born to mothers known to be infected with HIV and who did not breast feed

AIDS hazard

Of the 23 children reported to have developed AIDS, 11 were diagnosed by the age of 6 months, of whom 7 developed P carinii pneumonia. Infected children progressed to AIDS within the first 6 months of life at an estimated baseline rate of 17.7% (6.8% to 30.8%) before 1994 and 7.2% (0.0% to15.7%) from 1994 onwards. The difference did not reach significance (c² =2.51, P=0.11). After 6 months of age, an estimated 5.0% (2.4% to 8.5%) of children who had not yet developed AIDS would do so each year.

Trends in mode of delivery and antiretroviral treatment

Among 357 livebirths reported during 1995 to 1998, information on antiretroviral treatment taken by the mother during pregnancy and delivery, and by the baby after birth, was available for 349 (98%). Uptake of antiretroviral treatment increased significantly over this period (c²₁ =34.5, P<0.001 for trend), as did the proportion taking more than one antiretroviral drug (table 2). The proportion of mother-baby pairs receiving all three components of treatment ⁽¹⁾(antenatal, during delivery, and post partum to the baby) also increased significantly (c²=10.6, P=0.001 for trend), reaching 89% in 1998. All but five of the 292 mother-child pairs who had received any antiretroviral treatment took zidovudine either alone or in combination. Of the 346 infants with information on mode of delivery during 1995-8, the rate of elective caesarean section remained constant at around 40% up to 1997 and increased significantly to 62% in 1998 (c²₁ =30.8, P<0.001).

Table 2  Trends in mode of delivery and use of antiretroviral treatment 1995-8
 

*Antenatal, during labour, and post partum to baby.

Effects of elective caesarean section and antiretroviral treatment

Table 3 shows that the risk of HIV transmission adjusted for caesarean section was significantly lower in infants with exposure to antiretroviral treatment than in those without. The effect of caesarean section was not significant after antiretroviral treatment was adjusted for. The baseline transmission rate among infants delivered vaginally or by emergency caesarean section in the absence of antiretroviral treatment was estimated to be 31.6%. Elective caesarean section or antiretroviral treatment alone reduced the transmission rate to 15.3% and 10.1%, respectively, and when combined reduced transmission to 4.2%. There was no interaction effect observed between elective caesarean section delivery and antiretroviral treatment (c²₁ =0.2, P=0.7).

Table 3  Effects of elective caesarean section and antiretroviral treatment on vertical transmission rates of HIV, 1995-8 from logistic regression
 

* ²₁=6.9, P=0.009.

† ²₁=3.3, P=0.07.

Sensitivity analyses

Model assumptions were varied so that P carinii prophylaxis was begun in 1993 or 1995 rather than 1994, and AIDS reporting was treated as immediate or retarded so that only 55% were reported within 12 months rather than the estimated 85.5%. These variations affected the estimate of vertical transmission rate in 1997 (6.8%) or 1998 (2.2%) by no more than 0.9% absolute, and they did not change the significance of the fall in vertical transmission since 1995.

Discussion

HIV vertical transmission rates rose gradually from 1984 to 1994 and then fell rapidly between 1995 and 1998. The earlier rise could be accounted for by changes in the epidemiology of HIV not detailed in this report: mean maternal age increased from 20 to 30, while the proportion of mothers with HIV symptoms increased; the proportion of women infected through injecting drugs fell, and the proportion exposed abroad increased. ⁽⁵⁾

In the four years ending 1998, during which uptake of elective caesarean section and antiretroviral treatment was increasing, the estimated vertical transmission rate fell steeply to 2.2% (0% to7.8%) in 1998. This is comparable with a 2.0% estimate based on a meta-analysis, ⁽²⁾ and 0.8% (1/133) in the French cohort ⁽⁴⁾ among women taking antiretroviral treatment (mainly zidovudine only) and having an elective caesarean section. Bearing in mind the relatively wide confidence intervals, the effects of interventions were similar to the adjusted odds ratios of 0.4 for zidovudine use and 0.29 for elective caesarean section observed in the Swiss cohort. ⁽⁸⁾ We found no evidence for an interaction between the effects of zidovudine and elective caesarean section, which was consistent with Read et al ⁽²⁾ but not the French cohort study. ⁽⁴⁾

We also estimated the efficacy of P carinii prophylaxis in limiting progression to AIDS in infancy. Progression to AIDS (70% of which was P carinii pneumonia) within the first 6 months of life was reduced by nearly 60% in infants born after the beginning of 1994 compared with those born before. The effect was not significant, but a similar reduction of 66% in AIDS in infancy since 1993 was recently reported in the United States. ⁽¹⁵⁾ Because antiretroviral treatment was rarely started in the first 6 months of life, this change is likely to be due to prophylactic co-trimoxazole given to most children born to infected women from 6 weeks of age. ^{(12) (13)} Because individual data on P carinii prophylaxis was incomplete, calendar time was used as a proxy variable. The resulting "misclassification" would underestimate the effect of prophylaxis.

Prospective cohort studies tend to be set in specialist centres, whereas randomised trials may use exclusion criteria. National surveillance data have the advantage of being population based but are more likely to suffer from missing data and less frequent follow up. However, as long as bias caused by children of indeterminate status is removed, surveillance data should accurately reflect vertical transmission rates in clinical practice. Trials may, in any case, become increasingly difficult to carry out. Antiretroviral treatment is becoming increasingly diverse, more women will have received antiretroviral treatment before pregnancy, and the transmission rates now being reported are very low. ^{(16) (17) (18) (19)} If the possibility of teratogenic effects becomes a major concern, ⁽²⁰⁾ compliance with treatment regimens may fluctuate. Surveillance data may eventually become the most reliable way to monitor uptake of interventions and HIV vertical transmission rates.

We thank obstetric and paediatric respondents to the surveillance schemes of the Royal College of Obstetricians and Gynaecologists and the British paediatric surveillance unit of the Royal College of Paediatrics and Child Health, as well as colleagues at the Communicable Disease Surveillance Centre, Colindale, and the Scottish Centre for Infection Environment and Health, Glasgow. We also thank David Dunn for help in developing the statistical methods.

Key messages

· Reliable estimates of HIV vertical transmission rates can be derived from surveillance data

· Infected pregnant women are increasingly taking up elective caesarean section and antiretroviral treatment to reduce the risk of transmitting HIV to their babies

· Vertical transmission rates have fallen greatly over the past four years and progression to AIDS among infected children may also have slowed

· These benefits can only occur if infected women are diagnosed before or during pregnancy

Contributors: TD elaborated the statistical methods, programmed them, carried out the analyses, and drafted the methods and results section. AEA obtained funding for the surveillance (with Professor C S Peckham), conceptualised the analysis, and supported TD. TD and AEA are the guarantors. Introduction and discussion sections were drafted by DMG. PAT and JM assembled the surveillance data and assisted in its interpretation. All authors commented on all drafts of the paper.

Funding: TD, PAT, and JM were supported by grants from AVERT (Aids Education and Research Trust) and the Department of Health.

Competing interests: None declared.

1. Connor EM, Sperling RS, Gelber R, Kiselev P, Scott G, O'Sullivan MJ, et al. Reduction of maternal-infant transmission of human immunodeficiency virus type 1 with zidovudine treatment. Pediatric AIDS Clinical Trials Group Protocol 076 Study Group. N Engl J Med 1994;331:1173-80.
2. International Perinatal HIV Group. Mode of delivery and the risk of vertical transmission of human immunodeficiency virus type1. A meta-analysis of fifteen prospective cohort studies. N Engl J Med 1999;340:977-87.
3. European Mode of Delivery Collaboration. Elective caesarean-section versus vaginal delivery in prevention of vertical HIV-1 transmission: a randomised clinical trial. Lancet 1999;353:1035-9.
4. Mandelbrot L, Le Chenadec J, Berrebi A, Bongain A, Benifla JL, Delfraissy JF, et al. Perinatal HIV-1 transmission. Interaction between zidovudine prophylaxis and mode of delivery in the French perinatal cohort. JAMA 1998;280:55-60.
5. Nicoll A, McGarrigle C, Brady T, Ades AE, Tookey PA, Duong T, et al. Epidemiology and detection of HIV-1 among pregnant women in the United Kingdom: results from national surveillance 1988-1996. BMJ 1998;316:253-8.
6. Maguire A, Sanchez E, Fortuny C, Casabona J. Potential risk factors for vertical HIV-1 transmission in Catalonia, Spain: the protective role of cesarean section. Working Group on HIV-1 Vertical Transmission in Catalonia. AIDS 1997;11:1851-7.
7. European Collaborative Study. Vertical transmission of HIV-1: maternal immune status and obstetric factors. AIDS 1996;10:1675-81.
8. Kind C, Rudin C, Siegrist CA, Wyler CA, Biedermann K, Lauper U, et al. Prevention of vertical HIV transmission: additive protective effect of elective cesarean section and zidovudine prophylaxis. Swiss Neonatal HIV Study Group. AIDS 1998;12:205-10.
9. Dunn DT, Ades AE. Estimating the HIV vertical transmission rate and the pediatric AIDS incubation period from prospective data. J Am Stat Assoc 1997;91:935-43.
10. Dempster AP, Laird NM, Rubin D. Maximum likelihood estimation from incomplete data via the EM algorithm. J R Stat Soc 1977;B:1-38.
11. Rosenberg PS. A simple correction of AIDS surveillance data for reporting delays. J Acquir Immune Defic Syndr 1990;3:49-54.
12. Gibb DM, Walters S. Guidelines for management of children with HIV infection. Horsham: AVERT, 1993:1-29.
13. US Public Health Service recommendations for human immunodeficiency virus counselling and voluntary testing for pregnant women. MMWR 1995;44:1-14.
14. Efron B, Tibshirani J. Bootstrap measures for standard errors, confidence intervals, and other measures of statistical accuracy. Stat Sci 1986;1:54-77.
15. Lindegren ML, Byers B, Roger M, Wortley P, Fleming P. On the verge of elimination of perinatally acquired HIV/AIDS? Updated trends in the United States. 6th conference on retroviruses and opportunistic infections Chicago 1999; Abstract 228.
16. Fiscus S, Adimora AA, Schoenbach VJ, Wilfert C, Johnson VA. Can zidovudine monotherapy continue to reduce perinatal HIV transmission? The North Carolina experience 1993-1997. XII World AIDS Conference Geneva June 1998; Abstract 33162. [Q to A Please give editor, publisher, and place of publication]
17. Ciria LM. Impact of zidovudine on perinatal HIV transmission in Mallorca Island, Spain. XII World AIDS Conference Geneva June 1998; Abstract 23314.
18. Galvaao NO, Silva CL, Naud PSV, Chaves EBM, Zachia SA, Larangeira M, et al. Reduction of maternal-infant transmission of human immunodeficiency virus type 1 with zidovudine (ZDV) treatment. XII World AIDS Conference Geneva June 1998; Abstract 12155.
19. Rakusan TA, Temple V, Hart L, Loechelt B, Rana S, Young M, et al. Use of zidovudine to reduce the risk of perinatal transmission of HIV infection in the Washington metropolitan area. XII World AIDS Conference Geneva June 1998; Abstract 12157.
20. Lorenzi P, Spicher VM, Laubereau B, Hirschel B, Kind C, Rudin C, et al. Antiretroviral therapies in pregnancy: maternal, fetal and neonatal effects. AIDS 1998;12:F241-7.

Vertical transmission of HIV is falling in Britain

Pregnant women infected with HIV in the British Isles are increasingly taking up interventions to reduce the risk of infecting their baby. An analysis of surveillance data by Duong et al (p000) has shown large increases in the proportion of elective caesarean section deliveries and use of antiretroviral treatment to reduce transmission of HIV infection to the infant over the past four years. As a result vertical transmission of HIV has fallen from 19% in 1995 to 2% in 1998 among mothers whose diagnosis had been known during pregnancy.