Rubella seroprevalence in pregnant women in North Thames: estimates based on newborn screening samples

Abstract

Objectives

Routine screening for rubella susceptibility is recommended in the UK so that women found to be susceptible can be offered immunization in the post partum period. We demonstrate the use of newborn dried blood spot samples linked to routine vital statistics datasets to monitor rubella susceptibility in pregnant women and to investigate maternal characteristics as determinants of rubella seronegativity.

Setting

North Thames region of England (including large parts of inner London).

Methods

Maternally acquired rubella IgG antibody levels were measured in 18882 newborn screening blood spot samples. Latent class regression finite mixture models were used to classify samples as seronegative to rubella. Data on maternal country of birth were available through linkage to birth registration data.

Results

An estimated 2.7% (95% CI 2.4%–3.0%) of newly delivered women in North Thames were found to be seronegative. Mothers born abroad, particularly in Sub-Saharan Africa and South Asia, were more likely to be seronegative than UK-born mothers, with adjusted odds ratios of 4.2 (95% CI 3.1–5.6) and 5.0 (3.8–6.5), respectively. Mothers under 20 years were more likely to be seronegative than those aged 30 to 34.

Conclusion

Our findings highlight the need for vaccination to be targeted specifically at migrant women and their families to ensure that they are protected from rubella in pregnancy and its serious consequences.

Introduction

Before rubella vaccination programmes were introduced, at least 200 children were born each non-epidemic year in the UK with defects attributed to congenital rubella infection, including profound hearing loss, cataracts and congenital heart defects.¹ Due to a successful national vaccination programme, introduced initially for schoolgirls in 1971 and for all children aged 12 to 15 months in 1988 as part of measles-mumps-rubella (MMR) vaccination, circulation of rubella virus in the community has virtually ceased² and the number of rubella cases and children born with congenital rubella has declined dramatically. Between 1990 and 2007, 55 children with congenital rubella defects were reported to the National Congenital Rubella Surveillance Programme, which covers England, Scotland and Wales. Two thirds of these children were born to women who themselves were born abroad. Overall, about one third of mothers of affected children acquired their infection abroad.³

During the last decade the UK has experienced large-scale migration from Asia, Africa and Europe. Many countries in these regions have either no national rubella vaccination programme, or sub-optimal vaccine uptake as in parts of Southern and Eastern Europe.^4,5 As a large proportion of international migrants to the UK reside in London,⁶ the proportion of women susceptible to rubella infection is likely to have increased. Studies of rubella immunity during pregnancy have shown that in the UK women from minority ethnic groups are more likely to have low or undetectable rubella antibody levels than those from other groups.^7–9

The uneven uptake of MMR vaccine across the UK and especially in London is cause for concern. Since 1999, MMR uptake among 24-month-old children in London has remained below 80%,¹⁰ the minimum uptake required in the long term to prevent rubella virus circulation.^11,12 In some localities within London, first-dose MMR uptake was less than 70% in 2006–2007.¹⁰ Low uptake of MMR vaccine has already been linked to several measles outbreaks in London (see for example Atkinson et al.¹³ and Ashmore et al.¹⁴), and one death in England in 2006.¹⁵ The European region of the World Health Organization has set a target for elimination of rubella and reduction of the number of congenital rubella cases to less than 1/100 000 by 2010.¹⁶ In order to achieve and maintain this low incidence, rubella control measures, particularly MMR uptake among children in the UK and in other European countries needs to be improved.

All pregnant women in the UK are offered antenatal screening for rubella susceptibility.^17,18 However, data routinely collected from the antenatal screening programme are not currently disaggregated by maternal characteristics,¹⁹ and reliable data on maternal country of origin are not available. Furthermore, as the cut-off values used in the antenatal screening programme to indicate rubella susceptibility in pregnancy are designed to identify all women requiring vaccination, using such data may overestimate prevalence of seronegativity as it includes false positives.

Data on rubella antibody levels for women of reproductive age, disaggregated by age group, country of birth and area of residence, are required to identify groups at increased risk of infection. As maternal IgG antibodies are transferred to offspring across the placenta,^21,22 newborn antibody levels can be used as a proxy for maternal antibody concentration. We carried out a seroprevalence study among newly delivered women using residual dried blood spot samples from routine newborn screening. As coverage of screening is very high, these samples provide a large and nearly unbiased sampling frame for serosurveys and eliminate the need for further sample collection, which may be costly and introduce bias. The samples used in this study were collected in the North Thames area, which includes a large part of north and central London, and is a region with low MMR uptake and a substantial non UK-born population.

Methods

All newborns in the UK are offered screening for rare disorders at five to eight days of age, using whole blood spot samples collected on filter paper cards.²³ Coverage of newborn screening is estimated to be close to 100%.²⁴ Newborn screening samples collected in the North Thames region of England were used in this study. After completion of routine newborn screening a further blood spot is punched from the screening cards into 96-well plates for the HIV unlinked anonymous surveillance programme.²⁵

For logistical reasons the sampling unit was the well plate, rather than individual samples. In order to achieve 95% confidence intervals for the prevalence of rubella susceptibility with a precision to within 2.5% of the value of the estimated prevalence among Sri Lankan pregnant women (the groups with the highest prevalence of rubella susceptibility in a previous study⁹), around 20 000 newborn dried blood spot samples were required. A random sample of 220 plates was taken, containing 19,781 of the approximately 55,600 samples tested for HIV in the first and third quarter of 2004. Samples which were repeats, HIV-positive or those for which parents had declined screening or HIV testing were excluded before analyses (n = 899).

As part of the HIV testing protocol, dried blood spots are eluted in 170 μl PBS-Tween 20 solution. The residual eluates from HIV testing were tested for rubella IgG antibody according to the manufacturers’ instructions using a commercial enzyme-linked immunosorbent assay (ELISA; Diesse, Siena, Italy), which has been validated for use on dried blood spot samples.²⁶ Antibody concentrations are given as optical density (OD) ratios (i.e. the optical density of the sample, divided by a reference sample with a known antibody level of 10 IU/ml), which are adjusted for inter-plate variation.

Statistical methods

The specificity of antibody assays may be compromised in serosurveys in which alternative specimens to serum are used, and cut- off values provided by the manufacturer to identify seronegative samples may therefore not be applicable.^27,28 Instead, we constructed latent class regression finite mixture models to the distribution of rubella IgG levels to estimate the proportion of seronegative samples. Similar models have been applied in analyses of data from other serosurveys to define a seronegative population.^27–29 Details of these models are described in the Appendix. The samples allocated to the component with the lowest mean of the optimal mixture model were assumed to be seronegative.

In order to take account of sampling design, the standard error of the estimate of the proportion of seronegative women was adjusted using the svy functions in STATA 9.0. Maternal age and PCT of maternal residence at delivery are available from screening card data. A robust algorithm, linking screening card data and birth registration records held by the Office for National Statistics, has been established for the HIV unlinked anonymous surveillance programme^25,30 to provide additional information on maternal country of birth. As maternal age was missing from 31% of samples after completion of data linkage, it was imputed using multiple imputation by chained equations with the Mice package³¹ in the R software, assuming that maternal age data were missing at random. Five imputations were undertaken, creating five multiply imputed datasets.

In order to evaluate the independent effect of maternal characteristics on the odds of rubella seronegative status, defined as allocation to the component of the mixture model with the smallest mean based on posterior probabilities, logistic regression analyses were carried out, also using STATA version 9.O.³² The logistic regression model was constructed as a mixed effects model, with plate as a random effect to take account of variation in assay conditions by plate. These analyses were undertaken separately for each of the five imputed datasets; coefficients were then combined using the mijoin and mini commands in STATA^33,34

For the purpose of logistic regression analyses, maternal age was coded into age groups (<20, 20–24, 25–29, 30–34 and 35 years and over) and maternal country of birth into seven world regions (UK, other Europe, Sub-Saharan Africa, South Asia, Americas, North Africa and West and Central Asia, and East and South East Asia and Oceania). Maternal residence at delivery within the North Thames area was coded into a five category variable; North West, North Central and North East London, Essex, and Bedfordshire and Hertfordshire. The former three are urban areas, whereas the latter two mainly cover suburban and rural localities.

This study was approved by the London Multi-Centre Research Ethics Committee (MREC).

Results

Rubella antibody concentrations from 18882 newborn blood spot samples received in the screening laboratory and anonymously tested for HIV between January and March, and July to September 2004 were included in the analysis, representing around 33% of the approximately 56600 screening samples received in the screening laboratory and tested for HIV in this period. Record linkage to birth registration data was achieved for 18410 (97.5%) samples. Maternal area of residence at delivery was available for 18750 samples (99.3%). After linkage, maternal country of birth was available for 18324 mothers (97.0%), 11195 (61.1%) of whom were born in the UK. The median maternal age (interquartile range) was 30 years (25 to 34 years), before imputation and in the five multiply imputed datasets.

From the mixture model it was estimated that 509 of 18882 samples were allocated to the component with the smallest mean; thus we estimated that 2.7% of specimens in our study sample (95% confidence interval [CI] 2.4%-3.0%) were seronegative to rubella IgG antibody. Table 1 shows the proportion of seronegative samples, as defined by the latent class mixture model, by maternal characteristics and adjusted odds ratios for the odds of rubella seronegative status.

Table 1

The number of seronegative samples by maternal characteristics and adjusted odds ratios by maternal country of birth, age and SHA where registered at delivery

Maternal characteristics	Number of samples	Number seronegative (%)	95% CI for% seronegative	Adjusted odds ratio (95% CI)^*
Age group †
<20	689	36 (5.2)	3.8,7.1	2.5 (1.7,3.7)
20–24	2124	74 (3.5)	2.7, 4.5	1.4(1.0, 2.1)
25–29	3308	82 (2.5)	1.9, 3.2	1.0 (0.7, 1.4)
30–34	3937	81 (2.1)	1.6,2.6	1
35 +	2755	62 (2.2)	1.7,2.9	1.0(0.6, 1.5)
Maternal region of birth
UK	11 195	128 (1.1)	0.9, 1.4	1
Rest of Europe	1340	34 (2.5)	1.9, 3.4	2.1 (1.4, 3.1)
North Africa, West and Central Asia	868	31 (3.6)	2.5, 5.2	2.7(1.8, 4–1)
Americas	536	22 (4.1)	2.6, 6.5	3.4(2.1,5.5)
East and South East Asia and Oceania	536	21 (3.9)	2.3, 6.5	3.6 (2.2, 5.8)
Sub-Saharan Africa	1732	88 (5.1)	4.1, 6.3	4.2 (3.1,5.6)
South Asia	2117	131 (6.2)	5.2, 7.4	5.0 (3.8, 6.5)
Maternal area of residence
North East London	4336	137(3.2)	2.7, 3.8	1
North West London	3834	172 (4.5)	3.7, 5.5	1.4(1.1, 1.8)c
North Central London	3164	81 (2.6)	2.0, 3.2	0.9(0.7, 1.3)
Bedfordshire and Hertfordshire	3696	65 (1.8)	1.4,2.3	0.8 (0.6, 1.1)
Essex	3380	41 (1.2)	0.9, 1.7	0.8 (0.5, 1.1)
Random effect on plate				σ_u = 0.19 (0.09, 0.36)
				ρ = 0.05 (0.02, 0.09)

Wald test P values

P < 0.001

p = 0.044

p = 0.009

†

Number & proportion seronegative by age group based on maternal age distribution before imputation (n = 12813)

Odds ratios have been combined from the five multiply imputed datasets are adjusted for each of the other covariates and random variation by plate (n = 17985)

Recently delivered mothers aged less than 20 years of age, were more likely to be seronegative to rubella than those aged 30–34. Mothers born abroad, particularly those born in Sub-Saharan Africa and South Asia, were more likely to be seronegative than those born in the UK. Women resident in North West London were also at increased risk of being seronegative, compared with the baseline area of North East London. The combined estimate of the intraclass correlation coefficient ρ was 0.05, indicating that around 95% of the variation in maternal seronegative status was due to variation in maternal characteristics, and 5% due to variation between plates.

Discussion

We used dried blood spot samples left over from newborn screening to estimate maternally-derived rubella IgG antibody level as a proxy for maternal seronegative status in the North Thames area of England, a setting of high international migration and suboptimal MMR uptake among children. An estimated 2.7% of newly delivered women in North Thames were categorized as seronegative to rubella. This definition was derived from a latent class regression mixture model. Women born abroad were more likely than UK born women to be seronegative, with those born in Sub-Saharan Africa and South Asia at highest risk. Women under 25 years, particularly those under 20 years, were significantly more likely to be seronegative than women giving birth in their early thirties. We also found that women resident in North West London were at significantly greater risk of being rubella seronegative; however there were no significant differences between the two suburban/rural regions (Essex and Bedfordshire and Hertfordshire) and the baseline area of North East London, which is highly urbanised and includes parts of central London. The increased odds of being seronegative to rubella among women born abroad can be explained by the absence of rubella vaccination programmes in their country of origin. The majority of countries in Sub-Saharan Africa and South Asia have not implemented selective or universal rubella immunization programmes,³⁵ and MMR vaccination has only recently been included in the national schedules in countries in Central and South America and East Asia. In some countries in Eastern and Southern Europe, although rubella vaccination has been in place, uptake has been suboptimal.^36–38

Giraudon et al.³⁹ reported that 4% of women screened negative to rubella antibody in London through the antenatal screening programme in 2004. Our estimate is slightly lower than this. However these two prevalence estimates may not be directly comparable, as the aim of the antenatal screening programme is to identify women in need of vaccination, and hence the cut-off used to define rubella seronegative status may be higher than that used in the present study.²⁰ It has also been reported in previous British studies that a high proportion of children with congenital rubella have mothers of Asian background⁸, and that women who are from a minority ethnic background are more likely than white British women to be seronegative to rubella.^7,9 The use of data linkage in this study allows us to estimate the prevalence of rubella susceptibility in pregnant women by country of birth thus reflecting the effect of different immunization policies.

Country of birth was used as an indicator of maternal migrant status, but the date of entry to the UK is not known for women who have migrated. It was therefore not possible to estimate the prevalence of rubella seronegative status separately for newly arrived migrants and long-term residents although it is likely that the prevalence in the former group is higher.

Residual newborn screening samples are a useful resource for monitoring maternal susceptibility to rubella in the general population. Due to the very high coverage of newborn screening, these samples provide a large and near-universal sampling frame with which to estimate maternally acquired rubella IgG antibody using relatively inexpensive methods. Although it was not possible in this study to make a direct comparison between rubella IgG levels using paired samples of maternal sera and newborn dried blood spots, a high level of agreement between maternal and newborn rubella IgG levels has been demonstrated elsewhere.^22,40,41

Latent class regression finite mixture models allow estimation of the proportion of women with low antibody levels who are at increased risk of infection, but avoid using cut-off values that may be less accurate when alternative dried blood spot samples rather than serum are used.²⁷ The proportion of seronegative samples by maternal region of birth could be estimated as the samples were linked to birth registration data. This is the first study of rubella seroprevalence in pregnant women in which such data linkage is used, providing a more specific indicator of maternal background than those used in previous studies. It is unlikely that a study of this size with minimal bias and relevant demographic information would have been feasible without data linkage between health service and vital statistics datasets.

The main limitation of this study is the lack of data on parity, as it is not on the public record of the birth registration form in England and Wales and is not recorded for unmarried women. Parity is associated with rubella immunity in pregnancy⁹ as post-partum vaccination is recommended for women who screen rubella susceptible in pregnancy. Thus women delivering their second or subsequent child would be expected to be immune if the screening programme is effective. The observed relationship between maternal age and rubella seronegative status may therefore be attributed to the antenatal screening programme, as older mothers are more likely to be giving birth to their second or subsequent child. However, this would only apply to women previously pregnant in the UK, and no reliable information is available regarding the proportion of women who take up post-partum rubella vaccination. The inclusion of parity on the public record of the birth registration form would allow future studies to explore further the influence of maternal age and parity on rubella seronegative status in pregnant women.

The aim of the current UK antenatal screening programme for rubella antibody is to identify susceptible women who require post-partum vaccination. However, post-partum vaccination only protects women who have delivered previous children in the UK, and susceptible migrant women arriving in the UK pregnant or prior to becoming pregnant will remain at increased risk of infection. As childhood vaccination uptake has been suboptimal in London for several years, outbreaks are now more likely, and spread to pregnant women is a risk, particularly if other family members are susceptible. Supplementary vaccination activities targeted particularly at newly arrived migrants and their families, preferably at their first point of contact with UK health-care services, are required to prevent a rubella outbreak spreading to susceptible pregnant women.

Contributions: All laboratory testing was carried out by DW. Data management and analysis was carried out by PH and MCB. All authors contributed to the preparation and checking of the manuscript.

Footnotes

Acknowledgements

This study was funded by the Medical Research Council, grant number G0300654. This work was undertaken at the UCL Institute of Child Health which receives a proportion of funding from the Department of Health's NIHR Biomedical Research Centres funding scheme. In addition, The Centre for Paediatric Epidemiology and Biostatistics benefits from funding support from the Medical Research Council in its capacity as the MRC Centre of Epidemiology for Child Health.

Latent class regression finite mixture models

Latent class regression finite mixture models⁴² were fitted to the full dataset (n = 18882) of rubella IgG antibody levels to characterise unidentified subpopulations according to their susceptibility. Before the models were fitted, natural logarithms were applied to the optical density ratio values to adjust for a right-skewed distribution.

The mixture models fitted were: (1)

f (y) = \sum_{k = 1}^{K} π_{k} f_{k} (y),

where 0 ≤ π_k ≤ 1, (i = 1,…,K), and (2)

\sum_{k = 1}^{K} π_{k} = 1

where y denotes log antibody levels and π _k . is the weight of the k-th component in the mixture. Normal distributions were assumed for the mixture model components. The models were fitted using the flexmix library⁴³ in the R software.⁴⁴ The Bayesian Information Criterion (BIC) was used as the basis of the model selection strategy to determine the number of components in the optimal mixture model.

Models were fitted assuming up to five components. The optimal mixture model, based on minimization of the BIC, had four components, as shown in Table 2.

Figure 1 shows the densities of the four components of the optimal model, the total density of the rubella IgG levels in the 18882 samples, and the means (μ) and component weights (π) for the four components. Figure 1

The density of the components of the optimal mixture model, with their means and weights, and the total density function of rubella IgG antibody levels (expressed as natural logarithms of the optical density ratios). Total n = 18882

By using the Flexmix library to construct the mixture models, it is possible to calculate the probability of belonging to each component for every observation, known as posterior probabilities. The posterior probabilities can be estimated as follows: the probability that observation j is allocated to component i is given by:

We used equation (3) to assign each observation to the mixture model component for which the posterior probability was maximized. The individuals with observations assigned to the component with the smallest mean were assumed to be seronegative.

References

Peckham

. Congenital rubella in the United Kingdom before 1970: the prevaccine era. Rev infect Dis 1985; 7 (Suppl 1): S11-6

Health Protection Agency. Rubella notifications (confirmed cases) England and Wales, 1995-2007 by quarter. Available from: http://www.hpa.org.uk/infections/topics_az/rubella/data_rub_not.htm. Last accessed 11 January 2008

Tookey

. Congenital rubella BPSU 20th Annual Report 2006-2007. Available from: http://bpsu.inopsu.com/publications/annual_reports/BPSU%20Annual%20report%202006-7.pdf. Last accessed 28 June 2008

Gabutti

, Rota

, Salmaso

, Bruzzone

, Bella

, Crovari

. Epidemiology of measles, mumps and rubella in Italy. Epidemiol Infect 2002; 129: 543-50

Panagiotopoulos

, Antoniadou

, Valassi-Adam

. Increase in congenital rubella occurrence after immunisation in Greece: retrospective survey and systematic review. BMJ 1999; 319: 1462-7

Office for National Statistics. International Migration: Migrants entering or leaving the United Kingdom and England and Wales 2006. Available from: http://www.statistics.gov.uk/downloads/theme_population/MN33.pdf. Last accessed 28 August 2006

Miller

, Waight

, Rousseau

. Congenital rubella in the Asian community in Britain. BMJ 1990; 301: 1391

Miller

, Nicoll

, Rousseau

. Congenital rubella in babies of south Asian women in England and Wales: an excess and its causes. Br Med J (Clin Res. Ed) 1987; 294: 737-9

Tookey

, Cortina-Borja

, Peckham

. Rubella susceptibility among pregnant women in North London, 1996-1999. J Public Health Med 2002; 24: 211-6

10.

NHS Information Centre. NHS Immunisation Statistics, England 2006-07. Available from: http://www.ic.nhs.uk/webfiles/publications/immstats2005to2006/Final%202006-07%20lmmunisation%20Bulletin%20%28with%20copyright%29.pdf. Last accessed 27 June 2007

11.

Anderson

, Grenfell

. Quantitative investigations of different vaccination policies for the control of congenital rubella syndrome (CRS) in the United Kingdom. J Hyg (Lond) 1986; 96: 305-33

12.

Nokes

, Anderson

. Measles, mumps, and rubella vaccine: what coverage to block transmission? Lancet 1988; 2: 1374

13.

Atkinson

, Cullinan

, Jones

, Fraser

, Maguire

. Large outbreak of measles in London: reversal of health inequalities. Arch Dis Child 2005; 90: 424-5

14.

Ashmore

, Addiman

, Cordery

, Maguire

. Measles in North East and North Central London, England: a situation report. Euro Surveill 2007; 12: E070920

15.

Health Protection Agency.

Increase in measles in 2006, in England and Wales.

CDR Weekly 2006; 16: 3

16.

World Health Organisation. Eliminating measles and rubella and preventing congenital rubella infection: WHO European Region strategic plan 2005-2010. Available from: http://www.euro.who.int/Document/E87772.pdf. Last accessed 27 June 2008

17.

Department of Health. Screening for infectious diseases in pregnancy Available from: http://www.dh.gov.uk/assetRoot/04/09/20/49/04092049.pdf. Last accessed 28 June 2008

18.

National Institute of Clinical Excellence. Antenatal care: routine care for the healthy pregnant woman. Available from: http://www.rcog.org.uk/resources/Public/pdf/Antenatal_Care.pdf. Last accessed 28 June 2008

19.

Anderson

, Righarts

, Maguire

. Surveillance of antenatal infections-HIV, hepatitis B, syphilis and rubella susceptibility in London. Commun Dis Public Health 2004; 7: 251-7

20.

Ades

. Evaluating screening tests and screening programmes. Arch Dis Child 1990; 65: 792-5

21.

Leineweber

, Grate

, Schaad

, Heininger

. Transplacentally acquired immunoglobulin G antibodies against measles, mumps, rubella and varicella-zoster virus in preterm and full term newborns. Pediatr Infect Dis J 2004; 23: 361-3

22.

Goncalves

, Nascimento

MSJ

, Reu

, Cutts

. Levels of rubella antibody among vaccinated and unvaccinated Portuguese mothers and their newborns. Vaccine 2006; 24: 7142-7

23.

UK Newborn Screening Programme Centre. Newborn blood spot screening in the UK: policies and standards. Available from: http://www.ich.ucl.ac.uk/newborn/download/policies_standards.pdf. Last accessed 12. March 2007

24.

Ades

, Walker

, Jones

, Smith

. Coverage of neonatal screening: failure of coverage or failure of information system. Arch Dis Child 2001; 84: 476-9

25.

Cortina-Borja

, Cliffe

, Tookey

, Williams

, Cubitt

, Peckham

. HIV prevalence in pregnant women in an ethnically diverse population in the UK: 1998-2002. Aids 2004; 18: 535-40

26.

Hardelid

, Williams

, Dezateux

. Agreement of rubella IgG antibody measured in serum and dried blood spots using two commercial enzyme-linked immunosorbent assays. J Med Virol 2008; 80: 360-4

27.

Gay

, Vyse

, Enquselassie

, Nigatu

, Nokes

. Improving sensitivity of oral fluid testing in IgG prevalence studies: application of mixture models to a rubella antibody survey. Epidemiol Infect 2003; 130: 285-91

28.

Vyse

, Gay

, Hesketh

, Pebody

, Morgan-Capner

, Miller

. Interpreting serological surveys using mixture models: the seroepidemiology of measles, mumps and rubella in England and Wales at the beginning of the 21st century. Epidemiol. Infect. 2006; 134: 1303-12

29.

Talukder

, Kafatos

, Pinot de

. The seroepidemiology of varicella zoster virus among pregnant Bangladeshi and white British women in the London Borough of Tower Hamlets, UK. Epidemiol Infect 2007; 135: 1344-53

30.

Ades

, Walker

, Botting

, Parker

, Cubitt

, Jones

. Effect of the worldwide epidemic on HIV prevalence in the United Kingdom: record linkage in anonymous neonatal seroprevalence surveys. Aids 1999; 13: 2437-43

31.

van Buuren

, Oudshoorn

. Flexible multivariate imputation by MICE. TNO report PG/VGZ/99.054 Available from: http://web.inter.nl.net/users/S.van.Buuren/mi/docs/rapport99054.pdf. Last accessed 28 June 2008

32.

StataCorp. Intercooled Stata 9.0

33.

Carlin

, Li

, Greenwood

, Coffey

Carolyn

. Tools for analyzing multiple imputed datasets. The Stata Journal 2003; 3: 226-44

34.

Carlin

, Galati

, Royston

. A new framework for managing and analysing multiply imputed data in STATA. The Stata Journal 2008; 8: 49-67

35.

World Health Organisation. WHO vaccine-preventable diseases monitoring system: 2006 global summary. Available from: http://www.who.int/vaccines-documents/GlobalSummary/GlobalSummarypdf. Last accessed 27 June 2008

36.

Rafila

, Marin

, Pistol

, Nicolaiciuc

, Lupulescu

, Uzicanin

. A large rubella outbreak, Romania-2003. Euro Surveill 2004; 9: 7-9

37.

Panagiotopoulos

, Georgakopoulou

. Epidemiology of rubella and congenital rubella syndrome in Greece, 1994-2003. Euro Surveill 2004; 9: 17-9

38.

Ciofi degli

, Filia

, Revello

, Buffolano

, Salmaso

. Rubella control in Italy. Euro Surveill 2004; 9: 19-21

39.

Giraudon

, Trail

, Arnold

, Maguire

. Antenatal infection screening surveillance in London 2000 to 2004. Health Protection Agency

40.

Linder

, Sirota

, Aboudy

. Placental transfer of maternal rubella antibodies to full-term and preterm infants. Infection 1999; 27: 203-7

41.

Griffiths

, Berney

, Argent

, Heath

. Antibody against viruses in maternal and cord sera: specific antibody is concentrated on the fetal side of the circulation. J Hyg (Lond) 1982; 89: 303-10

42.

McLachlan

, Peel

. Finite mixture models. New York: John Wiley and Sons, 2000

43.

Leisch

. FlexMix: A general framework for finite mixture models and latent class reduction in R. J Stat Soft 2004; 11: 1-18

44.

R Development Core Team. R: A language and environment for statistical computing. Available from: http://www.r-project.org. Last accessed 28 June 2008

45.

Kuha

. AIC and BIC – Comparisons of assumptions and performance. Sociological Methods & Research 2004; 33: 188-229