Interpretation
Our finding that C. trachomatis and N. gonorrhoeaescreening and treatment reduced preterm birth or low birthweight among
nulliparous women is similar to the AuTop trial,(22) which found that
Bacterial vaginosis (BV) screening and treatment reduced preterm birth
among nulliparous, but not multiparous low-risk pregnant women in
France. Authors of the AuTop trial hypothesized that the differential
effect may be due to nulliparous women having an unknown risk for
preterm birth, whereas multiparous participants had a low risk as women
with a history preterm birth or late miscarriage were excluded from
enrollment. However, the Maduo study did not exclude participants based
on risk factors and 13% of multiparous women had a documented history
of preterm birth or low birth weight outcomes. AuTop authors also
speculated that the results could be due to higher treatment uptake
among nulliparous compared to multiparous women (85% vs 79%). In
Maduo, the prevalence of C. trachomatis and N. gonorrhoeaeinfections and treatment uptake were similar by parity. Another reason
for the differential impact in our study could be that nulliparous women
were less likely to be living with HIV compared to multiparous women
(5% vs. 24%) suggesting a lower risk of previous exposure to C.
trachomatis and N. gonorrhoeae infections. Some limited research
supports the idea that a degree of protective immunity could develop as
a result of prior chlamydial infection, which would make the screening
and treatment intervention less effective.(23)
Our finding that STI screening and treatment could reduce adverse birth
outcomes compared to syndromic management adds some support to previous
systematic reviews and meta-analyses. Adachi et al., found eight studies
that evaluated the effect of screening on adverse pregnancy outcomes.
(8) Five studies found that treatment with erythromycin reduced adverse
birth outcomes.(24-28) Two systematic reviews and meta-analyses assessed
the relationship between antenatal infection and adverse birth outcomes.
One review found 12 relevant observational studies, (20) and the
meta-analysis results concluded that C. trachomatis was
associated with preterm labor (unadjusted OR =1·29; 95% CI 1·11–1·50)
and low birth weight (unadjusted OR = 1·80; 95% CI 1·20–2·71).
However, that review found substantial heterogeneity across studies and
only two studies reported whether women had received treatment. Another
systematic review and meta-analysis assessing N. gonorrhoeaefound that women with N. gonorrhoeae were more likely to
experience preterm birth (OR 1·55, 95% CI 1·21 to 1·99, n=18 studies)
and low birth weight (OR 1·66, 95% CI 1·12 to 2·48, n=8). (7) Included
studies were all observational, most did not control for confounding,
and most women with N. gonorrhoeae had received treatment.
A recent study investigated the relationship between C.
trachomatis, N. gonorrhoeae , and Trichomonas (T.) vaginalisscreening and treatment among pregnant women living with HIV in South
Africa(29) and found no association between screening and preterm birth
or low birth weight. However, this study was limited to women living
with HIV, did not exclude symptomatic participants, found the testing
group was less likely to be on anti-retroviral therapy, and treated
participants for T. vaginalis . Further, participants in that
study had high levels of persistent infections as 27% were positive forC. trachomatis at repeat testing.(30) No participants in the
current study were positive at test-of-cure.(31)
Given the equipoise around the impact of STI screening on adverse
pregnancy outcomes, future research may consider individual
randomization or other methods for addressing confounding, measurement
bias, and lack of power. As discussed by Low in a recent commentary,(32)
observational studies face major threats to internal validity as many
factors are associated with both STIs and pregnancy outcomes.
Case-control and cohort studies are subject to measurement bias when
receipt of treatment is not recorded because treatment likely reduces
the risk of the outcome.(32) Cluster trials, such as the current study
and the upcoming WANTAIM Trial(33) seek to reduce confounding and
measurement bias; however, confounding can still occur through
unmeasured baseline imbalance and clustering has a negative impact on
the power to detect an effect and thus cluster designs typically require
larger sample sizes.(34)