Dr R. Gilbert, Centre for Paediatric Epidemiology and Biostatistics, Institute of Child Health, 30, Guilford Street, London WC1N 1EH, UK.
Objective To determine the association between congenital toxoplasmosis and preterm birth, low birthweight and small for gestational age birth.
Design Multicentre prospective cohort study.
Setting Ten European centres offering prenatal screening for toxoplasmosis.
Population Deliveries after 23 weeks of gestation in 386 women with singleton pregnancies who seroconverted to toxoplasma infection before 20 weeks of gestation. Deliveries after 36 weeks in 234 women who seroconverted at 20 weeks or later, and tested positive before 37 weeks.
Methods Comparison of infected and uninfected births, adjusted for parity and country of birth.
Main outcome measures Differences in gestational age at birth, birthweight and birthweight centile.
Results Infected babies were born or delivered earlier than uninfected babies: the mean difference for seroconverters before 20 weeks was −5.4 days (95% CI: −1.4, −9.4), and at 20 weeks or more, −2.6 days (95% CI: −0.5, −4.7). Congenital infection was associated with an increased risk of preterm delivery when seroconversion occurred before 20 weeks (OR 4.71; 95% CI: 2.03, 10.9). No significant differences were detected for birthweight or birthweight centile.
Conclusion Babies with congenital toxoplasmosis were born earlier than uninfected babies but the mechanism leading to shorter length of gestation is unknown. Congenital infection could precipitate early delivery or prompt caesarean section or induction of delivery. We found no evidence for a significant association between congenital toxoplasmosis and reduced birthweight or small for gestational age birth.
Congenital toxoplasmosis is caused by transplacental transmission of the parasite Toxoplasma gondii in women that acquire the infection during pregnancy. As few women have symptoms, maternal infection is detected by testing for seroconversion of toxoplasma specific antibodies. The risk of transmission increases steeply with gestational age at maternal seroconversion.1 Congenital toxoplasmosis is associated with signs of intracranial calcification, hydrocephalus and/or retinochoroiditis in approximately one in six infants, and with perinatal death or severe disseminated infection in 1–2% of cases.1–3 Reported rates of congenital toxoplasmosis range from less than 1 per 10,000 live births in Massachusetts,4 Austria,5 Sweden6 and Norway7; 2–3 per 10,000 live births in Poland8 and Brazil9; and 1 per 1000 in France.10
Congenital toxoplasmosis has long been considered a cause of intrauterine growth retardation. Serological testing for congenital toxoplasmosis is still recommended as one component of the TORCH screen (acronym for toxoplasmosis, other infections, rubella, cytomegalovirus and herpes simplex virus) for unexplained intrauterine growth retardation.11 However, evidence for an association between growth retardation and congenital toxoplasmosis is limited to reports of case series referred for clinical problems.12,13 To date, no population-based studies have investigated the effect of congenital toxoplasmosis on size or weeks of gestation at birth.
In this report, we analysed a large prospective cohort of women who seroconverted to toxoplasma infection during pregnancy to determine whether congenital toxoplasmosis is associated with preterm delivery, low birthweight or being small for gestational age.
The study was based on women enrolled in the European Multicentre Study on Congenital Toxoplasmosis (EMSCOT) for which methods have been reported in detail elsewhere.14 Women were identified by serological testing which was repeated in susceptible women each month in France, and three monthly elsewhere. The analysis was restricted to women who seroconverted during pregnancy and had a singleton birth. Women who had a spontaneous miscarriage or termination of pregnancy before 24 weeks of gestation were excluded as they were not routinely investigated for maternal and congenital toxoplasmosis, and birthweight was not always recorded. We also excluded mother–child pairs identified by neonatal screening for congenital toxoplasmosis because perinatal deaths and preterm deliveries would be under-represented. Stillbirths were excluded from the analyses because the date of death was unknown and may have occurred several days before delivery. Inclusion of stillbirths could therefore introduce bias in favour of lower birthweight centile. Terminations of pregnancy were excluded from analyses of gestational age at delivery and birthweight as these outcomes were determined by the timing of termination. We also excluded terminations from analyses of birthweight centile standardised for gestational age and sex as information was lacking on the baby's sex (Fig. 1). Congenital infection status was measured as previously defined.14 The study complied with the research ethics requirements in each participating country.
We hypothesised that fetal toxoplasma infection could cause premature onset of labour due to inflammatory mediators, and it could inhibit fetal growth. The primary outcomes were therefore gestational age at birth, and the gestational age- and sex-standardised birthweight centile reflecting babies born small for gestational age. We used the clinician's best estimate of the expected date of birth based on an early prenatal dating ultrasound scan, if available, and date of last menstrual period. Standardised scores (z-scores) were calculated for birthweight using sex- and gestational-age-specific growth reference standards.15 Secondary outcomes were birthweight, and each of the outcomes dichotomised.
Prenatal treatment was analysed as the interval between seroconversion and the start of any type of treatment or, if untreated, delivery (treatment delay). We also analysed the type of initial treatment (spiramycin, pyrimethamine–sulphonamide or no treatment). Treatment regimens have been described in detail elsewhere.14 Subsequent treatment changes, for example, from spiramycin to pyrimethamine–sulphonamide, could not be considered, as they were conditional on the results of prenatal diagnosis for congenital infection status.
Maternal seroconversion was estimated to occur at the midpoint between the last negative IgM test and the first positive IgM test unless no IgG was found when IgM was first positive, in which case the date of seroconversion was considered to be 14 days before the date of the first positive IgM test.1,14 Women identified by tests for recent infection (for example, with specific IgG and IgM positive tests and low IgG avidity at the first prenatal test) who gave birth to an infected child were assumed to have their last negative IgM test at conception. Other covariates were country (France vs Italy or Austria), maternal age and parity. Other factors associated with low birthweight and preterm delivery, such as socio-economic status and maternal smoking, were not recorded for the study. These factors would confound the analyses only if they are associated with mother-to-child transmission. We know of no experimental or observational data to indicate whether this is the case or not. No information was collected on the mode of delivery.
Analyses were constructed to avoid the inherent correlation between gestational age at seroconversion and gestation at birth. For example, pregnant women that seroconvert at 38 weeks of gestation are not at risk of preterm delivery as seroconversion must necessarily precede birth. We made an a priori decision to separately analyse seroconverters from the first (less than 20 weeks) and second halves of pregnancy as the effect of congenital toxoplasmosis on outcome is likely to be greater the earlier fetal infection occurs. Multiple regression models were derived for the three continuous outcomes to determine the significance of differences between infected and uninfected births accounting for covariates. Logistic regression models were derived for these outcomes dichotomised as follows: preterm delivery (<37 weeks of gestation), low birthweight (<2500 g) and small for gestational age birth (<10th centile for age and sex standardised birthweight). Resulting odds ratios represent the risk of an unfavourable outcome given an infected versus uninfected fetus. A backward stepwise approach was used for all multivariate analyses. Country and parity were kept in all models because of established differences in population norms among countries and the association between parity and birthweight.11,16 Other potential confounders included in initial models were gestation at seroconversion and its interaction with congenital infection status, maternal age and its square, treatment delay and its interaction with congenital infection status, and type of first treatment. Variance inflation factors were computed for each variable, and potentially confounding variables and variance inflation factor values exceeding 10 were excluded from further model development.17 Variables were retained in the final model if significant at P < 0.05. Main effects were included if interactions were significant.
The second analysis involved women who seroconverted at or after 20 weeks of gestation. We limited the analysis to women who had their first positive IgM test before 37 weeks and delivered at 37 or more weeks of gestation in order to ensure that all women were at risk of the outcome. The statistical approach was the same as that used above except that for logistic regression, gestation at birth was dichotomised near the median (<40 weeks). Statistical analyses were performed using SAS version 8.2.
Figure 1 shows the number of women included in the analyses.Tables 1 and 2 show the prevalence of preterm delivery, low birthweight and small for gestational age birth according to congenital infection status, as well as results of univariate and multivariate analyses.
Table 1. Mean differences and odds ratios for gestation at birth, birthweight and birthweight centile according to congenital infection status in women who seroconverted before 20 weeks of gestation.
Table 2. Mean differences and odds ratios for gestation at birth after 37 weeks, birthweight and birthweight centile according to congenital infection status in women who seroconverted at 20 weeks or more of gestation.
Of the 386 women who seroconverted before 20 weeks, 42 (11%) gave birth to infected babies (Table 1). Five women had terminations at 24 weeks or more. All five were performed for toxoplasmosis and all the fetuses were infected (weeks of gestation at termination were 24, 27, 27, 28, 33). There were three stillbirths and all were uninfected (weeks of gestation at delivery were 31, 33 and 40). The mean maternal age was 27.6 years (SD 4.8) and almost half (n= 163, 45%) had no other children. The majority of women (n= 278, 72%) were enrolled from French centres. The median gestation at seroconversion was 13.8 weeks (range: 1.5, to 19.9) and almost all (n= 383, 99%) were treated, mostly with spiramycin as the first treatment (n= 315, 82%). The median interval between seroconversion and the start of treatment was 31days (range: 7, to 262 days).
Infants with congenital toxoplasmosis had a reduced length of gestation relative to uninfected infants (mean reduction 5.4 days; 95% CI: 1.4, 9.4), and were significantly more likely to be born before 37 weeks (Fig. 2A). Preterm birth occurred in 11/42 (25%) infected babies (the earliest at 34.6 weeks), compared with 31/342 (9.1%, earliest at 30.4 weeks; 10 were born before 34 weeks). Country of birth was significantly associated with a shorter length of gestation for babies born in France versus elsewhere (mean difference 4.6 days; P= 0.001). The length of gestation did not differ significantly in women with a positive versus negative prenatal diagnosis (mean difference −4.5 days; 95% CI: −10.1, 1.2), or in women with infected versus uninfected fetuses who had no prenatal diagnosis (mean difference −2.7days; 95% CI: −9.2, 3.7).
There was no significant association between congenital toxoplasmosis and birthweight or being small for gestational age. Parity was a significant confounding factor for birthweight (P= 0.01) and parity and country were significant confounders for smallness for gestational age (P= 0.04 and P= 0.01, respectively).
There was no significant interaction between gestational age at seroconversion and congenital infection status for any of the outcomes.
Seroconversion at 20 weeks of gestation or later
A total of 234 women seroconverted at 20 weeks or later, had their first positive IgM test before 37 weeks and gave birth after 36 weeks of gestation. Of these, 104 (44.4%) had infected babies, one had a stillbirth and none had a termination. The average maternal age was 29.2 years (SD 4.9), and the majority (n= 193, 82.5%) were enrolled from French centres. Almost all women were treated (n= 225, 96.2%), with spiramycin as the first treatment in 181 cases (77.4%). The median gestation at seroconversion was 27.2 weeks (range: 20–35 weeks) and the median treatment delay was 24.3 days (21, 143 days).
Congenital toxoplasmosis significantly reduced gestational age at birth (mean difference −2.6 days; 95% CI: −4.7, −0.5) (Fig. 2B). Babies in France were born significantly earlier (mean −6.4 days; P < 0.0001).
There was no significant association between congenital toxoplasmosis and mean birthweight or with low birthweight. However, confidence intervals for these analyses were wide due to the small number of infants with low birthweight. Parity was a significant confounder when birthweight was analysed as a continuous variable (P= 0.02). There was no significant association between congenital toxoplasmosis and standardised birthweight. Country of birth and parity were significant confounding factors (P < 0.01).
Congenital toxoplasmosis was associated with reduced length of gestation in women seroconverting before and after 20 weeks. We found no significant association between congenital toxoplasmosis and birthweight or small for gestational age birth.
Strengths of the study include the fact the women were identified prospectively by prenatal screening for maternal infection, and enrolled before any attempt was made to diagnose fetal infection. This reduced the potential for selection bias due to investigation of congenital infection in small for gestational age or preterm births. A limitation is that information was not available on whether labour was spontaneous or induced, or whether delivery was by caesarean section. In addition, we found no clear evidence that length of gestation was shorter in women with a positive versus negative prenatal diagnosis. Consequently, it is not possible to determine whether the association between congenital toxoplasmosis and length of gestation is due to induction of labour or caesarean section or due to spontaneous onset of labour. Obstetric intervention may partly explain the finding of earlier gestational age at delivery in the French centres. In addition, although we used the most accurate estimate available for the expected date of delivery, methods for computing this date may have varied among countries.
A further limitation was that we could not investigate the effects of treatment versus no treatment as there were few untreated women. However, we found no significant effect of the timing or type of treatment on any of the outcomes. Well-known risk factors for preterm delivery or small for gestational age birth, such as maternal smoking, socio-economic status and ethnicity, were not measured in this study. At the time the study was initiated, we knew of no association between these factors and mother-to-child transmission of toxoplasmosis; no evidence has emerged since.
As there was no comparison group of uninfected women, we were not able to determine whether there was any independent effect of maternal infection on gestation or small for gestational age birth. However, a recent large cohort study found no evidence of an association between fever in early pregnancy and fetal death.18
We did not analyse the effect of congenital toxoplasmosis on spontaneous miscarriage or stillbirth. This was because of the small number of fetal losses, the possibility of selection bias favouring infected fetuses and the fact that fetal losses were often not investigated for congenital toxoplasmosis. Instead, we compared total fetal losses using a subset of pregnant women (10/448) from the EMSCOT study with a first positive IgM test in the first trimester, standardising for maternal age using the age distribution reported in a study of unselected women with viable pregnancies at 10 weeks of gestation.19 The rate of spontaneous fetal loss before 28 weeks among toxoplasma-infected women in the EMSCOT cohort was 2.48% (95% CI: 0.04, 4.03),20 similar to the rate reported by Gilmore and McNay21 of 2.1% (1.50, 2.84). Out of 10 fetal losses in this subset of the EMSCOT cohort, two had congenital infection based on a positive prenatal diagnosis (PCR positive) and autopsy, two had a negative prenatal diagnosis (PCR negative) and six were not investigated.
The study is the first population-based study of the effect of congenital infection on gestation and smallness at birth. Sever et al.22 previously reported an association between low birthweight and high titres (>1024) for toxoplasma-specific IgG in women post-delivery (relative risk 1.4; P < 0.01). However, congenital infection status was not measured nor was there any exploration of confounding due to factors that affect maternal exposure to infection.
For babies with congenital infection, the clinical consequences of being born on average 5.4 days earlier are unlikely to be serious unless delivery takes place before 30 weeks of gestation. Although 25% (95% CI: 12.5, 39.9) of infected babies were born preterm, compared with 9% (95% CI: 6.1, 12.1) of uninfected babies, none of the infected babies were born before 34 weeks. However, further studies are required that include information on the onset of labour and mode of delivery to determine whether shorter length of gestation is related to obstetric intervention or a consequence of fetal infection.
Congenital toxoplasmosis was associated with shorter length of gestation at birth but the mechanisms underlying this association require further exploration. We found no evidence to support the view that congenital toxoplasmosis retards fetal growth. Nevertheless, given that there were only 146 infected infants in the study, we cannot exclude a possible effect of T. gondii on fetal growth in individual cases.
The project was funded by the European Commission (BIOMED II No. BMH4-CT98-3927 and QLG5-CT-2000-00846). The authors would like to thank Pat Doyle for advice on the analyses.
Members of the European Multicentre Study on Congenital Toxoplasmosis (EMSCOT)
M.-H. Bessieres, W. Buffolano, H. Dumon, R. Gilbert, E. Petersen (chairperson), A. Pollak, P. Thulliez, M. Wallon.
K. Freeman, L. Oakley, A. Pollak, W. Buffolano, E. Petersen, E. Semprini, A. Salt, R. Gilbert.
(number of patients contributed to this report, centre): P. Thulliez, S. Romand (155; Institut de Puericulture, Paris), M. Wallon, F. Peyron (140; Hôpital de la Croix Rousse, Lyon), A. Prusa, M. Hayde, A. Pollak (105; University Children's Hospital, Vienna), J. Franck, H. Dumon, P. Bastien, E. Issert (60; Hôpital de la Timone, Marseille), M.-H. Bessieres (55; Hôpital de Rangueil, Toulouse), W. Buffolano (37; Universita di Napoli, Naples), N. Ferret, P. Marty (27; Hôpital de l'Archet, Nice), C. Chemla, (23; Hôpital Maison Blanche, Reims), H. Pelloux, H. Fricker-Hidalgo, C. Bost-Bru (21, Centre Hospitalier Universitaire de Grenoble), E. Semprini, V. Savasi (7; Milan), B. Evengard, G. Malm (0; Huddinge Hospital, Stockholm), M. Paul (0; University Medical Sciences, Poznan), E. Petersen (0; Statenseruminstitut, Copenhagen).
Study design and co-ordination
R. Gilbert (Principal Investigator), L. Gras, J. Rickett, A. Salt, L. Valenti (Institute of Child Health, London).
K. Freeman, L. Oakley, L. Gras (Institute of Child Health, London).