Bulletin of the World Health Organization

Untreated maternal syphilis and adverse outcomes of pregnancy: a systematic review and meta-analysis

Gabriela B Gomez a, Mary L Kamb b, Lori M Newman c, Jennifer Mark b, Nathalie Broutet c & Sarah J Hawkes d

a. Amsterdam Institute for Global Health and Development, Trinity Buildings, Building C, Pietersbergweg 17, PO Box 22700, 1100 DE Amsterdam, Netherlands.
b. Division of STD Prevention, Centers for Disease Control and Prevention, Atlanta, United States of America.
c. Department of Reproductive Health and Research, World Health Organization, Geneva, Switzerland.
d. Institute for Global Health, University College London, London, England.

Correspondence to Gabriela B Gomez (e-mail: g.gomez@aighd.org).

(Submitted: 19 May 2012 – Revised version received: 06 November 2012 – Accepted: 16 November 2012 – Published online: 17 January 2013.)

Bulletin of the World Health Organization 2013;91:217-226. doi: 10.2471/BLT.12.107623

Introduction

In 2008, the World Health Organization (WHO) estimated that 1.86 million cases of syphilis occur globally among pregnant women each year and that a large proportion of them are untreated or inadequately treated. Up to one third of the women attending antenatal care (ANC) clinics are not tested for syphilis.1 If syphilis is left untreated during pregnancy, it can lead to fetal loss or stillbirth or, in a liveborn infant, neonatal death, prematurity, low birth weight or congenital syphilis. Programmes that include syphilis testing coupled with appropriate, prompt penicillin treatment for pregnant women who test positive for Treponema pallidum infection have been shown to be efficacious in reducing adverse pregnancy outcomes.24 In addition, these interventions have been estimated to be highly cost-effective, even in settings where the burden of syphilis among pregnant women is moderate or low.58

Existing barriers to scaling up these programmes can only be overcome through active involvement from policy-makers. Evidence-based estimates of the burden of congenital syphilis at the global, national and subnational levels help make the case for allocating resources to these effective programmes, increasing access to interventions and making progress towards elimination. Calculating the burden relies on precise estimates of the local prevalence of syphilis and adverse pregnancy outcomes among untreated women with syphilis. Previous estimates of adverse pregnancy outcomes in women with syphilis have been based on point estimates from single studies. To improve the quality of these estimates in the context of WHO’s global initiative for the elimination of congenital syphilis,9,10 we performed a systematic review and meta-analysis of reported estimates of adverse pregnancy outcomes among women with untreated syphilis and women without syphilis.

Methods

We performed a systematic review and meta-analysis that accorded with MOOSE guidelines11 and PRISMA requirements.12

Search strategy and inclusion criteria

We systematically searched the published literature without date or language restrictions to identify studies assessing pregnancy outcomes in the presence of maternal syphilis. We used combinations of the following terms to search PubMed, EMBASE and Cochrane Libraries: syphilis/congenital syphilis, pregnancy, antenatal/prenatal, neonate/newborn, infant, birth/pregnancy outcome, mortality, death, stillbirth/fetal death, neonatal death, infant death, preterm/low birth weight and perinatal death/mortality. The last search was performed in December 2011. We included literature published in any language and on any date. We reviewed references in seminal papers, review articles and medical textbooks. We canvassed experts in the field to identify additional studies, particularly older studies that may have been published before the availability of online databases. The grey literature and conference abstracts were not searched.

Because our goal was to estimate the range of possible birth outcomes associated with untreated syphilis during pregnancy, our population of interest was pregnant women who had untreated syphilis. We included studies that described pregnancy outcomes among women presumed to have syphilis (i.e. women who were seroreactive for T. pallidum infection, irrespective of the test used), as well as uninfected (i.e. control) women. We focused on fetal, neonatal and infant outcomes because maternal outcomes of syphilis would not be expected for several years after disease onset and were unlikely to have been reported in the same studies. We included studies of at least 30 patients that described the sampling strategy used to recruit patients from the community, hospitals or ANC clinics. We excluded studies describing partially treated populations, unless adverse pregnancy outcomes were reported specifically among women who lacked syphilis testing or treatment (i.e. at least 2.4 million U of intramuscular penicillin). We considered a broad range of study designs, including clinical trials, observational studies, programme reviews and case series. We assessed potential studies to ensure that there was no duplication of case series.

The data extracted included study characteristics such as study year, geographical location, diagnostic test used, sample size for cases and controls, design, length of follow-up, type of outcome and frequency estimates for T. pallidum–associated seroreactivity in the study population or a comparable population from the same setting and period. Length of follow-up varied among the studies. The tests used to define seroreactivity varied across settings and over time, from the Wasserman and Kahn tests (for which sensitivity and specificity data are not available) in early studies, to the Venereal Disease Research Laboratory or rapid plasma reagin tests (sensitivity, 71–100%; specificity, 98%), the fluorescent treponemal antibody absorption test (sensitivity, 84–100%; specificity, 97%) and the microhaemagglutination assay for T. pallidum (sensitivity, 76–100%; specificity, 99%) in more recent studies.13 The early literature did not always define birth outcomes (e.g. dates used to define stillbirth, fetal loss or prematurity) and in such cases we relied on the outcome terminology in the original papers.

Statistical analysis

We calculated crude proportion estimates and standard errors (SEs) for all adverse pregnancy outcomes in women with untreated syphilis and women without syphilis. We then calculated crude proportion estimates and SEs for the following select adverse pregnancy outcomes: fetal loss or stillbirth (combined), neonatal death (defined as a death occurring from birth up to the age of 28 days), prematurity or low birth weight (combined) and clinical evidence of syphilis. We separately calculated crude proportion estimates of infant death (defined as a death occurring between ages 29 and 365 days) to allow for the differentiation between neonates and infants.

Proportions were transformed into logits. The SEs of the logits were calculated for the meta-analysis procedure. Both the logits and the SEs of the logits resulting from the meta-analysis routines were transformed back to percentages for the forest plots.14,15 We used random-effects meta-analyses to pool estimates (with 95% confidence intervals [CIs]) and calculated measures of heterogeneity between studies (i.e. I2 values and a P-value of < 0.05 was defined as indicative of a statistically significant difference in results).

We performed group-specific analyses for all adverse outcomes and for specific adverse outcomes to explore the sources of heterogeneity. We defined three groups of studies. Group 1 included studies calculating the frequency of an adverse pregnancy outcome involving women whose reproductive history was recorded before the existence of penicillin treatment. Group 2 comprised studies in which the frequency of an adverse pregnancy outcome was calculated for women attending an ANC clinic that offered no treatment or testing for syphilis. Group 3 comprised studies that examined the frequency of an adverse pregnancy outcome involving women attending an ANC clinic in which recruitment in the study was done at the time of delivery (and syphilis treatment was only available at that time). All data were prepared and analysed using Stata/SE version 11.0 (StataCorp. LP, College Station, United States of America).

Results

Overall, 3258 citations were retrieved. Six articles6,1620 were considered eligible for the study (Fig. 1) and extracted data are shown in Table 1. All articles presented the findings of observational studies that included a “control” arm assessing adverse pregnancy outcomes among women without syphilis. Of these studies, Harman’s investigation of family histories of infants with blindness predated the availability of syphilis treatment and was classified as a group 1 study.16 Harman tabulated all pregnancy outcomes and child deaths from birth through the first year of life over the reproductive histories (as many as 18 pregnancies) of 150 mothers with syphilis and 150 “healthy” mothers living in the same impoverished London community. Wammock, Ingraham and McDermott et al. looked at pregnancy outcomes among women attending ANC clinics in which either screening or treatment was not available; all three investigations were classified as group 2 studies.1719 The studies by Ingraham and Wammock were conducted before the availability of penicillin treatment in the United States of America. The study by McDermott et al. had a primary goal of assessing malaria-associated adverse pregnancy outcomes in women attending a rural ANC clinic in Malawi. No antenatal syphilis testing programme existed at that time and seroreactivity was identified later on the basis of analysis of stored blood samples. The intervention study in Zambia, by Hira et al., and the retrospective study of women tested for syphilis at delivery in the United Republic of Tanzania, by Watson-Jones et al., were ANC evaluation studies conducted in the context of poorly implemented syphilis screening/treatment programmes; both investigations were considered to be group 3 studies.6,20 All studies reported on stillbirth or fetal loss. Only three studies reported on neonatal death.1719 Five studies reported on clinical evidence of congenital syphilis in children6,1618,20 and four studies reported on prematurity or low birth weight.6,17,18,20 Only McDermott et al. looked at infant death (i.e. deaths among subjects aged 29 days to one year).19

Fig. 1. Flow diagram of study selection for a systematic review and meta-analysis of studies of adverse pregnancy outcomes among women without syphilis and women with untreated syphilis
Fig. 1. Flow diagram of study selection for a systematic review and meta-analysis of studies of adverse pregnancy outcomes among women without syphilis and women with untreated syphilis

All adverse pregnancy outcomes

All studies consistently reported a higher proportion of adverse pregnancy outcomes among untreated women with syphilis (range: 53.4–81.8%) than among women without syphilis (range: 10.2–20.8%). Fig. 2 shows the study-specific and summary estimates for all adverse pregnancy outcomes.

Fig. 2. Study-specific and summary estimates of the proportion (%) of all adverse pregnancy outcomes (APOs) among women with untreated syphilis and women without syphilis
Fig. 2. Study-specific and summary estimates of the proportion (%) of all adverse pregnancy outcomes (APOs) among women with untreated syphilis and women without syphilis
CI, confidence interval.

Selected adverse pregnancy outcomes

Fig. 3 and Fig. 4 show study-specific estimates for select adverse pregnancy outcomes and their summary statistics in women with and without syphilis, respectively. The pooled estimates of neonatal death were 12.3% (95% CI: 9.3–16.2) among women with syphilis and 3.0% (95% CI: 2.1–4.3) among women without syphilis, for an absolute difference of 9.3%.

Fig. 3. Study-specific and summary estimates of the proportion (%) of selected adverse outcomes among women with untreated syphilis
Fig. 3. Study-specific and summary estimates of the proportion (%) of selected adverse outcomes among women with untreated syphilis
APO, adverse pregnancy outcome; CI, confidence interval; d, days.
Fig. 4. Study-specific and summary estimates of the proportion (%) of selected adverse outcomes among women without syphilis I2
Fig. 4. Study-specific and summary estimates of the proportion (%) of selected adverse outcomes among women without syphilis <em>I</em><sup>2</sup>
APO, adverse pregnancy outcome; CI, confidence interval.

Only the study by McDermott et al. reported data that allowed calculation of the proportion of infants who died.19 An absolute difference of 11.2% in the frequency of infant death was observed among infants of women with and without syphilis: 21.3% (16 of 75 infants; 95% CI: 13.6–31.9) versus 10.1% (256 of 2530 infants; 95% CI: 9.0–11.4), respectively. Harman reported an absolute difference of 11.5% in the frequency of infant death between infants born to women with and without syphilis: 22.9% (229 of 1001 births; 95% CI: 20.3–25.5]) versus 11.4% (94 of 826 births; 95% CI: 9.2–13.6), respectively. However, these data could not be subclassified into neonatal and infant deaths because Harman reported deaths during the period from birth to one year of age.16

Pooled estimates and heterogeneity

Meta-analysis of the estimates of all adverse outcomes across the six studies revealed substantial heterogeneity for both women with syphilis (I2 = 91.8%; P < 0.001) and women without syphilis (I2 = 95.9%; P < 0.001; Fig. 2). A group-specific analysis of all adverse outcomes revealed a marked difference between estimates for women with syphilis, with the highest summary statistic observed for group 2 and the lowest observed for group 3 (Appendix A, available at: www.who.int/reproductivehealth/publications/rtis/syphilis_in_pregnancy/en/m). Heterogeneity was acceptable only for group 3 (56.4% [95% CI: 50.8–61.9]; I2 = 0.0%; P = 0.547). Among women without syphilis, the summary statistic was highest for group 1 and lowest for group 3 (Appendix A). Again, heterogeneity was acceptable only for group 3 (10.3% [95% CI: 9.2–1.5]; I2 = 0.0%; P = 0.658).

We also analysed summary estimates of selected adverse pregnancy outcomes by study group in women with and without syphilis (Appendix B and Appendix C, both available at: www.who.int/reproductivehealth/publications/rtis/syphilis_in_pregnancy/en/m). In pregnant women with syphilis, the summary statistic for stillbirth and fetal loss was highest for group 2 and lowest for group 1. Heterogeneity was acceptable only for group 3 (22.5% [95% CI: 18.1–27.5]; I2 = 0.0%; P = 0.609). The summary statistic for clinical evidence of syphilis was highest for group 2 and lowest for group 3; heterogeneity was acceptable for group 2 (39.5% [95% CI: 34.0–45.4]; I2 = 0.0%; P = 0.343) and group 3 (2.8% [95% CI: 1.4–5.4]; 0.0%; P = 0.508). For prematurity or low birth weight, heterogeneity was acceptable for group 2 (30.2% [95% CI: 22.4–39.3]; I2 = 26.4%; P = 0.244) and group 3 (3.8% [95% CI: 1.3–10.4]; I2 = 55.3%; P = 0.135).

In women without syphilis, the summary statistic for stillbirth and fetal loss was highest for group 1 and comparably low for groups 2 and 3, but heterogeneity was unacceptable for the latter groups (Appendix C). The frequency of prematurity or low birth weight was comparable for groups B and C but heterogeneity was acceptable only for group 3 (7.6% [95% CI: 4.8–11.7]; I2 = 68.6%; P = 0.074).

Discussion

In this article, we quantified the proportion of all adverse pregnancy outcomes and specific adverse pregnancy outcomes among untreated women with syphilis and women without syphilis using data from six articles that met eligibility criteria for inclusion in our systematic review and meta-analysis. On average, pooled estimates of fetal loss or stillbirth, neonatal death and prematurity or low birth weight showed significantly higher rates among the offspring of women with syphilis than among the offspring of women without syphilis. The absolute differences were 21% for fetal loss or stillbirth, 9% for neonatal death and 6% for prematurity or low birth weight. Signs and symptoms of syphilis were found in 15% of the infants born to untreated women with syphilis. The frequency of any adverse pregnancy outcomes was 52% higher among women with syphilis than among women without syphilis.

Our estimates are consistent with previously published data. Holder23 quoted a 1949 review by Thomas et al. in which 70% of women with syphilis reported adverse pregnancy outcomes and Rutgers et al.24 reported adverse outcomes in 53% of women with syphilis. Both estimates are consistent with our summary estimate of 66.5%. Similar estimates have also been reported for specific pregnancy outcomes in women with syphilis. Thomas et al. estimated that 25–30% of pregnancies in women with syphilis resulted in neonatal death and Rabut et al.25 reported that 39% resulted in perinatal death (although the latter outcome was not defined, it probably included neonatal deaths, stillbirths and fetal losses). Rutgers et al.24 reported that 27% of infants were “admitted” because they were born before term (2.2%), had a low birth weight (13.5%) or presented with symptoms of congenital syphilis (11.3%).

Only one study evaluated in this report estimated death during the first year of life. It showed that infants delivered by untreated mothers with syphilis had 10% more deaths than those delivered by mothers without syphilis. This suggests that a considerable mortality burden is currently overlooked because of short follow-up periods in most cohort studies.

Early investigations evaluating syphilis testing and treatment preceded the use of experimental study designs; later studies have been limited by moral restrictions on the use of randomized controlled designs that withhold an effective intervention from a control group. Therefore, all studies included in our analysis had an observational design. Nevertheless, one strength of this review is that each study included a comparison group to assess adverse pregnancy outcomes among mothers without syphilis. This gave us the opportunity to estimate the excess adverse outcomes in the presence of maternal syphilis and give a broad idea of the risk of some of the adverse pregnancy outcomes in women with untreated syphilis.

Owing to the inherent differences between experimental and observational study designs and to biases commonly seen in observational data,11 appropriate caution should be taken in interpreting our results. We relied on the outcome definitions used in the original papers, many of which were published before a consensus definition was available. There were also unacceptable heterogeneity in estimates across studies, variations in diagnostic tests across settings and periods, and a lack of control for potential confounders. Several potential sources of bias are present as well. Reporting bias is possible because infants born to mothers known to have syphilis are more likely than infants born to mothers not known to have syphilis to be diagnosed with congenital syphilis. Group 2 studies included only asymptomatic women and therefore may have underestimated adverse pregnancy outcomes in women with untreated primary or secondary syphilis, both of which make transmission to offspring more likely and are associated with a fatality rate of nearly 100% (as noted among seven cases included in the Wammock cohort). Primary and secondary syphilis during pregnancy are not estimated to be common, so this potential bias is unlikely to change the magnitude of the findings. Another limitation associated with group 2 is that, although each study followed infants for at least a short period, the length of follow-up varied. Therefore, the burden of congenital syphilis might have been underestimated because congenital syphilis may have been missed in some infants. The lack of a search for grey literature sources could have introduced a reporting bias, but because of the nature of the research assessed we expect this to be minimal.

In the group 1 study, which assessed the lifetime reproductive histories of women infected with T. pallidum before the availability of anti-syphilis treatment, the timing of infection was not reported. This study probably underestimated the proportion of adverse pregnancy outcomes: some pregnancies included in the study may have occurred before women were infected and others may have occurred years after exposure, when syphilis was inactive and less likely to be transmitted to the fetus. Indeed, the risk of adverse pregnancy outcomes is recognized to be directly related to higher maternal antibody titres and such titres decrease over time in untreated women with syphilis.26

Finally, group 3 studies occurred in settings in which pregnant women might not have attended an ANC clinic, might have attended an ANC clinic but not been tested for syphilis, or were tested for syphilis but did not receive treatment before delivery. These studies probably underestimated the frequency of adverse pregnancy outcomes for several reasons. First, women with stillbirths and fetal losses who did not seek care would not have enrolled in a study. For example, the group 3 study by Watson-Jones et al. showed no fetal losses among 73 seroreactive women with rapid plasma reagin titres of ≥ 1:8. This is surprising given the large number of fetal losses in group 2 studies. Second, because penicillin is a highly effective intervention, treating neonates born to seroreactive mothers at birth probably prevented clinical infection or further progression of disease in the infants of infected mothers whose infections were not identified earlier.2 This bias was important since congenital syphilis in live born infants represented a small fraction of pregnancy outcomes in group 3 studies, compared with group 1 and 2 studies. Third, women in group 3 who partially or fully completed a regimen of syphilis treatment during pregnancy might not have had this information reflected on their maternity card. This would have resulted in the misclassification of some women with syphilis as untreated. Finally, because group 3 studies did not follow up study subjects after delivery, neonatal deaths, infant deaths and clinical manifestations that occurred after this period would have been missed.

Our study suggests that, unless testing and treatment of syphilis in pregnancy are universally available, over half of pregnancies in women with syphilis will result in an adverse outcome. This is a preventable burden on mothers, families and health systems that highlights the need to prioritize global efforts to eliminate the mother-to-child transmission of syphilis. This paper also reminds policy-makers charged with resource allocation that the elimination of congenital syphilis is a public health priority that will enable immediate progress towards Millennium Development Goal 4: reducing the mortality rate among children younger than 5 years by two thirds.


Competing interests:

None declared.

References

Share