Lives Saved Tool supplement detection and treatment of syphilis in pregnancy to reduce syphilis related stillbirths and neonatal mortality

In 2001, twelve million people globally were estimated to be infected with the causative agent of syphilis, the spirochete Treponema pallidum (T.pallidum) [7]. Syphilis classically presents in its primary stage with a painless, spontaneously resolving ulcer (chancre), lasting 3 – 6 weeks, that typically occurs around 21 days, but which may occur up to 3 months post infection. This is followed, 6 – 8 weeks later, by the secondary stage, which consists of a more diffuse inflammatory response characterised by an exanthema which manifests variably and for which healthcare is often not sought even if available. This stage also resolves spontaneously and is followed by an asymptomatic ‘latent period’. Untreated, 15 – 40% of infected individuals develop late ‘tertiary’ complications that may occur years after the initial infection: e.g. granulomas, neuropathic joint disease, cardiovascular and neurological complications.

Syphilis can be difficult to diagnose, since T pallidum cannot be cultured in vitro, and PCR-based tests are unavailable outside research settings. While direct diagnosis of characteristic treponemes with darkfield microscopy is highly sensitive and specific, this requires exudate obtained from a lesion during the brief primary stage. Thus most infections are identified using serologic tests that support, rather than establish, a syphilis diagnosis. Serologic tests fall into 2 broad categories: non-treponemal tests (e.g., RPR, VDRL) measure antibody to a component of mammalian cell membranes modified by T pallidum. Positive RPR tests are reported as the highest titre for which tests are still positive after dilution, with titre correlating with recentness of infection and stage of disease and falling over time or with treatment. In general, titres >1:4 have been considered to represent “active” infections, more likely to be more recently acquired and infectious. With treatment, titres typically return to negative. Treponemal tests (e.g., TPHA, MHA-TP) detect T pallidum antibody, and are reported as either positive or negative. Because these tests typically remain positive even after treatment, a positive treponemal test could indicate either recent active or old (or previously treated) infection. Currently available, rapid point-of-care tests (e.g., ICS strip) are treponemal tests.

The risk of syphilis to the foetus is dependent upon the stage of maternal infection and on the stage of pregnancy at which the foetus is exposed. Primary, secondary, early latent, and in some cases late latent, stage maternal syphilis infection can lead to haematogenous spread to the foetus, resulting in a systemic inflammatory response [8]. Although spirochetes can cross the placenta to the foetus from early pregnancy [9], foetal immune systems are not mature enough to mount a consistent immune response until about 18 to 22 weeks gestation. After this time the characteristic features of congenital syphilis may be seen. In addition, placental infiltration with reduced blood flow to the foetus can lead to growth restriction which, if severe, may result in foetal death [8]. Foetal involvement occurs most consistently in pregnant women with ‘active’ infections (i.e. RPR titre >1:4), that is inadequately or not treated infections acquired in the five years preceding the pregnancy. This review therefore focuses on this group of pregnant women with active infection. Based on the available data, an expert panel has estimated , that in the absence of effective treatment, 25% of pregnancies to such women will result in a 2^nd trimester miscarriage or stillbirth, 11% of pregnancies in a neonatal death at term, 13% in a preterm or low birth weight infant and an additional 20% with clinical signs of congenital syphilis that are attributable to syphilis. Hence, of the 75% of the pregnancies resulting in a liveborn infant, 15% would be expected to result in a neonatal death at term and 17% in a preterm or low birth weight infant, with a further 27% of babies who survive to 28 days developing clinical signs of congenital syphilis [3]. Congenital syphilis in the first 2 years of life may present with hepato-splenomegaly, anaemia, jaundice, rash, snuffles and pseudo-paralysis [10]. Later manifestations of congenital syphilis, after the first 2 years of life may be an important treatable cause of neurological, developmental, and musculoskeletal problems and death in infants and children in low-income settings [11]. In view of the ongoing morbidity and mortality associated with congenital syphilis, and because of the challenges of making a specific diagnosis, both CDC and WHO now recommend a 10 day course of intravenous or intramuscular penicillin treatment for all infants born to mothers with untreated or inadequately treated syphilis in pregnancy , regardless of symptoms or clinical findings [12, 13].

Parenteral penicillin was identified as effective in the treatment of syphilis shortly after its introduction in 1943, and is still the recommended drug for treatment of syphilis, whether acquired sexually or through transmission from mother to child during pregnancy [14]. Studies have established that penicillin is treponemocidal at fairly low serum concentrations (>0.1 mug/ml), and that longer duration regimens are more effective than shorter duration regimens, even those achieving higher concentrations [15]. These provide the therapeutic rationale for use of relatively low doses of longer acting penicillin G formulations (e.g. 2.4 million units benzathine penicillin). Current CDC and WHO guidelines recommend a single dose of long acting penicillin to treat early syphilis, and three injections of benzathine penicillin at weekly intervals for infections of > 1 year (or unknown) duration for infected adults, including pregnant women [14]. For women with infection of > 1 year duration (i.e. latent infection), clinical studies suggest that a single penicillin dose effectively treats the foetus in at least 95% of cases [16, 17]; while subsequent doses treat the mother. Other antibiotics, such as oral tetracyclines and macrolides (e.g., erythromycin, azithromycin) have been used to treat syphilis in adults, but are not recommended either due to their toxicity in the foetus (tetracyclines) or because they do not cross the placental barrier (macrolides) and thus do not treat foetal infection at all. Macrolide antibiotics have also been associated with development of resistance [15, 18]. Parenteral ceftriaxone has been used to treat syphilis in penicillin-allergic patients, but as no controlled studies among pregnant women have assessed its efficacy in treating the foetus, this is not recommended in pregnancy [13, 14].

Many factors limit the quality of evidence available concerning the magnitude of the effects of penicillin treatment on foetal and neonatal mortality. Syphilis in pregnancy is a dynamic disease; the timing of infection, re-infection and treatment can all vary, and there are a variety of possible non-specific outcomes (including stillbirth, preterm birth and neonatal and infant death) that are associated with many different, and potentially confounding factors (e.g. other infections, maternal well-being and adherence to recommendations). Given the longstanding recommendations regarding identification and treatment of maternal syphilis randomised controlled trials of the effect of penicillin compared to no treatment would not be ethical. Many observational studies control for potential confounders inadequately or not at all. In summary, despite evidence and widespread policy acceptance of the importance of identifying and treating syphilis during pregnancy, there are limited data regarding the magnitude of the effect such detection and treatment has on foetal and neonatal mortality.

We used the PICO approach (Population, Intervention, Comparison, and Outcome) to identify the studies to be included. The population of interest is pregnant women with active syphilis and the intervention being reviewed is serologic detection of syphilis in pregnant women and treatment of women with active syphilis (e.g RPR>1:4) with at least 2.4 million units penicillin given at least 28 days prior to delivery. The comparison group is pregnant women with active syphilis who do not receive at least 2.4 million units at least 28 days prior to delivery. The outcomes of interest are adverse pregnancy outcomes including stillbirth, preterm delivery, congenital syphilis and neonatal mortality associated with congenital syphilis (figure 1). Within this structure we considered both randomised trials and observational studies (figure 2). For duplicate reports of trials or studies we counted outcome data only once, although we have included data from secondary analyses if they added pertinent information. Possible adverse effects of penicillin treatment were not addressed as part of this review.

All studies meeting the criteria above were abstracted onto a standardised abstraction form for each outcome of interest [20]. Each study was assessed for limitations and graded according to the CHERG adaptation of the GRADE technique [21]. The evidence was summarised by outcome including a qualitative assessment of study quality and sources of bias adapted from the Cochrane review handbook. CHERG Rules for Evidence Review were applied to the collective evidence to provide an estimate for reduction in stillbirth and congenital syphilis related neonatal mortality.

Separate meta-analyses were planned for the effect of penicillin on incidence of stillbirth, preterm delivery, congenital syphilis, and neonatal mortality. Meta-analyses were conducted with STATA version 10.0 statistical software [22]. Heterogeneity was assessed using I² and the chi-squared test. When evidence of heterogeneity was present (p<0.10), a random effects model was used, otherwise a fixed effect was assumed. Summary risk ratios and corresponding 95% confidence intervals (CI) are reported.

Eight observational studies reported the incidence of stillbirths in treated, infected women compared with untreated infected women. None of the studies made any attempt to control this comparison for systematic differences between treated and untreated women (confounding). Women not attending antenatal clinics or not adhering to complex penicillin treatment regimes may differ in their risk profiles for stillbirth, preterm delivery and neonatal death from fully compliant infected women.

Two early, observational studies from the United States report very large reductions in stillbirths to pregnant women identified as ‘syphilitic’ by physical examination and serological testing and treated with at least 2.4 million units of penicillin G compared to untreated syphilis positive women diagnosed at delivery (RRs of 0.15 (95%c.i. 0.10 – 0.25) and 0.06 (95%c.i. 0.01 – 0.27)) [24, 25]. Four further studies reported RRs of 0.65 (95% c.i 0.29 – 1.45)[26], 0.17 (0.02 – 1.25) [27] , 0.19 (95% c.i. 0.10 – 0.38)[28] and 0.53 (95%c.i. 0.03 – 9.68)[29] for stillbirth. One study reported no stillbirths in a group of 111 adequately treated RPR-positive women compared to 5 stillbirths in 45 untreated pregnancies RR=0.04 (95%c.i. 0.00 – 0.66) [30]. Two Tanzanian cohorts paired in setting and time [17, 31] found three stillbirths amongst 133 women with high titre syphilis who received treatment [17], compared to 18 stillbirths amongst 73 women with high titre syphilis who were untreated[31] RR 0.09 (95% c.i. 0.03 – 0.30).

There is some evidence of heterogeneity across these 8 studies (P=0.02; I²=56.8%). A random-effects meta-analysis of all eight studies produced an estimated RR of 0.18 (95% c.i. 0.10 - 0.33) (figure 3). The heterogeneity observed appears to be largely attributable to a large South African study [26]. Excluding this study from the analysis there is no longer strong evidence of heterogeneity (P=0.65 I²=0%). The summary risk ratio is little changed (RR 0.15: 95% c.i. 0.10 – 0.21 ). The outlying study was relatively large, with a very low number of stillbirths in the untreated group. This may be partly due to some women being recruited in antenatal clinic in the 3^rd trimester, and a relatively high loss to follow-up (15%). Since none of these studies controlled for confounding, other factors may have contributed to the observed differences between treated and untreated women.

Five studies compared stillbirth rates in uninfected women compared to infected women who were treated during pregnancy. In three of the studies the observed stillbirth rate were slightly higher among infected women receiving treatment than among uninfected women (RR 1.3; 95% c.i. 0.81 – 2,13[25]: RR 1.4; 95% c.i. 0.34 – 5.75[24]: RR 1.4; 95% c.i. 1.05 – 1.86 [32]) while two studies found no elevated stillbirth risk (RR 0.94: 95% c.i. 0.90 – 0.98 [27]: RR 0.68; 95% c.i. 0.2 – 2.4 [17]).

Four studies investigated the effect of syphilis detection and penicillin treatment on perinatal mortality (i.e. stillbirths and live born infants who died in the first week of life) without distinguishing stillbirths from early neonatal deaths. Two of these studies attempted to control confounding. In one cohort of 158 women with syphilis diagnosed at first ANC visit in rural South Africa, 5/26 untreated pregnancies with known outcomes resulted in perinatal death, compared to 10/116 with at least 2.4 MU penicillin (adjusted RR=0.45: 95%c.i. 0.17 – 1.12) [33]. Another South African study reported a reduction in perinatal mortality of 35% with each dose of penicillin received (adjusted RR = 0.65: 95% c.i. 0.5 – 0.85) [26]. These results are broadly consistent with the two further observational studies that did not control for confounding. A study from South Africa reported a RR of 0.34 (0.11 – 1.03) [34] while a study from Swaziland observed 7/24 perinatal deaths in untreated pregnant women, 0/8 in treated women and 4/145 in those negative for syphilis [35].

Two early studies from the United States, report very large reductions in all-cause neonatal mortality in infants of pregnant women with syphilis (RR = 0.14, 95%c.i. 0.07 – 0.27 and RR = 0.18 95%c.i. 0.03 – 0.95) with the use of penicillin therapy [24, 25]. The dosing schedules of penicillin administered in these early studies were different to those in current use, using regular dosing with short acting penicillin. Neither study adjusted for potential confounding factors, but both had an additional control group of uninfected women and observed no excess risk of neonatal death in infected women receiving penicillin compared with uninfected women. Three more recent observational studies that also did not adjust for confounding were identified. One study in South Africa recruited women at their first ANC visit and compared those RPR positive women who were treated with penicillin with those who were not treated. To receive treatment women had to return to get their test result. For analysis purposes, only those completing treatment >28 days prior to delivery were considered to have been treated. An all-cause neonatal mortality risk ratio of 0.54 (95%c.i. 0.21 – 1.41) was reported [26]. The treatment effect observed in this study may have been attenuated as screening was based on RPR test positivity without confirmatory testing and will therefore have included some women without active syphilis infection. This is likely to result in an underestimate of the effectiveness of the invention for infected women. The second study, in The Russian Federation, recruited women at delivery, tested them for syphilis using a non-treponemal test for syphilis (Wassermann test) or a treponemal test (fluorescent treponemal antibody absorption) and, if positive, ascertained previous penicillin treatment by examining their treatment cards [36]. This study reported a risk ratio of 0.32 (95% c.i. 0.09 – 1.13) for neonatal death in those treated with penicillin in pregnancy. The third, cross-sectional, study from Zimbabwe used the RPR test to diagnose active syphilis and reported a RR of 0.14 (95% c.i. 0.02 – 1.32) for women treated with 1 dose of benzathine penicillin at least 28 days prior to delivery compared to untreated women [30]. A meta-analysis of these five studies results in an estimated RR of 0.20 (95%c.i. 0.13 – 0.32). (figure 4) There was not strong evidence of heterogeneity (I-squared=40.5%, p=0.15). Given the very large effect size and the biological plausibility of a treatment effect it is unlikely that the differences between treated and untreated women are entirely explained by uncontrolled confounding.

Seven observational studies which did not adjust for confounding were retrieved [24, 27, 29, 34, 37‐39] and two Tanzanian cohorts paired in time and place [17, 31]. One early cohort study from the United States reported a RR 0.18 (95% c.i. 0.03 – 0.95) for preterm birth associated with penicillin treatment but was excluded from further analysis as no definition of preterm birth or how it was assessed was given [24]. The remaining seven studies defined preterm birth as birth before 37 weeks completed gestation and used ultra-sound or last menstrual period and clinical examination at birth to estimate gestational age. A South African cohort reported a RR of 0.47 (0.24 – 0.94) for preterm birth with 1 – 3 doses of 2.4 million units penicillin G compared to no treatment [34]. A second, hospital-based cohort from Kenya which recruited mothers at delivery reported a RR of 0.50 (95% c.i.0.26 – 0.95) for preterm birth with penicillin treatment compared to no treatment [27]. A multi-centre retrospective cohort study in Russia reported a RR for preterm birth of 0.32 (95% c.i. 0.26 – 0.95) comparing adequately treated women to untreated women [38]. One small South African study observed a RR 0.08 (0.01 – 1.24) comparing treated to untreated women [37]. In China one study found a RR 0.23 (0.1 – 0.53) for preterm birth with treatment compared to untreated[39], whilst another small study reported a RR 0.41 (0.02 – 0.72).[29] Re-analysis of the paired Tanzanian cohorts give a RR 0.56 (95% c.i. 0.26 – 1.23). Combining these studies gives a RR of 0.36 (95% c.i. 0.27 – 0.47) for preterm birth with penicillin treatment (figure 5). There was no evidence of heterogeneity p=0.49, I² =0.0%.

Since none of these studies controlled for confounding in this analysis, other factors may have contributed to the association between penicillin treatment and reduced risk of preterm delivery. However the consistency and strength of association and further analyses of the Tanzania cohorts paired in setting and time, which adjusted for potential confounders such as reproductive tract infections, HIV, malaria and anaemia, provide some evidence that the effect seen is not entirely explained by uncontrolled confounding factors. The Tanzania study reported the risk of preterm birth to be 6.1 (95% c.i. 2.5 – 15.3) times higher in women with high titre RPR positivity (i.e., active syphilis) compared to uninfected women[31], while women with high titre RPR positivity who were treated with penicillin had no increased risk of preterm birth (RR =0.58; 95% c.i. 0.3 – 1.1) [17].

Three studies were identified addressing the effect of penicillin for pregnant women with active syphilis on the incidence of congenital syphilis. Two early cohort studies in the United States report risk ratios of 0.05 (95% c.i. 0.01 – 0.16) and 0.05 (95% c.i. 0.03 – 0.09) for congenital syphilis [24, 25].

A large, population-based study in 6 districts in China screened 477,656 pregnant women using Toluidine Red Unheated Serum Test, a non-treponemal serological test similar to RPR, for detecting current infection [40, 41]. Among 2208 (0.5%) test-positive women, the risk of congenital syphilis was 0.004 (7/1855) in those treated with at least 2.4MU IM benzathine benzylpenicillin, and 0.23 (81/353) in those not receiving treatment (RR = 0.02 (95% c.i. 0.01 – 0.04)) [4].

These three studies showed some evidence of heterogeneity (p=0.08 I² =61.3%) although all three studies indicate very large effects of treatment with penicillin. This heterogeneity may be partly explained by differences in diagnosis of congenital syphilis between the studies. A random effects meta-analysis produced an estimated risk ratio for congenital syphilis of 0.03 (95%c.i. 0.02 – 0.07) in those women with syphilis in pregnancy who were treated with penicillin, compared to those untreated. (Figure 6)

These results are consistent with a small prospective uncontrolled series examining 11 VDRL positive women identified between fourteen and nineteen weeks gestation. Amniocentesis confirmed foetal infection in four out of 11 cases. All four mothers were treated with 2 doses of 2.4 million units Penicillin G 1 week apart. No infant had serological evidence of congenital infection [9]. Another larger case series of penicillin-treated antenatal attendees with active syphilis reported that 98.2% (95% c.i 96.2 – 99.3%) of pregnancies resulted in an uninfected infant [42]. Timing of treatment is very important in subsequent risk of congenital syphilis. In a large prospective cohort study in China, women commencing treatment before 20 weeks gestation had a lower rate of congenital syphilis RR=0.50 (95%c.i. 0.38 – 0.64) [43] than women treated after 20 weeks.

The CHERG Rules for Evidence Review were applied [20] (table 1)

This work was supported in part by a grant to the US Fund for UNICEF from the Bill &Melinda Gates Foundation (grant 43386) to “Promote evidence-based decision making in designing maternal, neonatal and child health interventions in low- and middle-income countries”, and by a grant to Save The Children USA from the Bill & Melinda Gates Foundation (Grant 50124) for "Saving Newborn Lives".