Dear Editor,

I read with concern the review on the effects of periodontal treatment on pregnancy outcomes by George et al.1 recently published in the journal. The idea of infection been able to cause preterm delivery and low birth weight originates in the notion that bacterial vaginosis may be responsible for poor pregnancy outcomes. This hypothesis was popular during two decades and dismissed when the evidence originating from several clinical trials did not substantiate the existence of benefits from interventions aimed to treat vaginosis on the incidence of poor pregnancy outcomes.2

The results of many observational studies could be taken to indicate the existence of an association between periodontal infection and poor pregnancy outcomes, but many voices have been raised questioning the quality of that evidence with critical aspects like the lack of consideration for key confounding and the existence of selection bias been pervasive in these studies. Even though the quality of the epidemiological evidence today does not recommend conducting intervention studies on the topic, several clinical trials have been accomplished and enormous amounts of resources have been invested in research to evaluate whether periodontal treatment can improve pregnancy outcomes.

Two systematic reviews1,3 of intervention studies of the topic have been published during the last couple of years and their conclusions vary considerably. These reviews1,3 share nine original clinical trials and while George et al.1 additionally included a trial from an Hungarian population,4 Polyzos et al.3 included two additional trials conducted in American5 and Brazilian6 populations, respectively. One may wonder whether this can be the explanation for the differences reported. However, the answer can be found in the stark contrast in the strategies used to appraise the quality of the data and the attempts to disentangle its influence on the conclusions. In the review by George et al., key aspects of a trial like blinding of participants and allocation concealment were rarely confirmed in the trials included in the analyses and the studies in which the method of allocation was not reported received 'the benefit of the doubt'. No information on the quality of the trials was included in the analyses. On the other side Polyzos et al.3 found that only five trials could be considered of high methodological quality,5,7-10 with a low risk of bias. When meta-analysis was restricted to 'high quality trials' no effects of periodontal intervention on preterm birth, low birth weight or spontaneous abortions/stillbirths could be discerned; whereas significant beneficial effects of periodontal treatment on the rates of preterm birth and low birth weight were apparent when meta-analysis was conducted for 'low quality trials'.3 A closer look at the meta-analyses restricted to high quality studies reveals a tendency for periodontal interventions to result in higher rates of preterm deliveries (fig. 3) and overall poor pregnancy outcomes (fig. 6).3

The authors also documented the occurrence of publication bias with smaller studies reporting higher odds ratios than did larger studies and trials perceived to be missing in the review corresponding to the field of non-significant findings.3

George et al. conclude, 'The cumulative evidence suggests that periodontal treatment during pregnancy may reduce preterm birth and low birth weight incidence. However, these findings need to be further validated through larger more targeted randomised control trials'. One may wonder about how many additional expensive studies need to be conducted to decide whether the evidence is 'controversial' or not?

References