Volume 196, Issue 2 , Pages 93-94, February 2007
Periodontal disease and preterm birth: do the data have enough teeth to recommend screening and preventive treatment?
Article Outline
In recent years there has been a mounting body of evidence that indicates an association between periodontal disease and preterm birth. Drs Jean-Noel and Michel, in the current issue of the Journal, report the results of a comprehensive and well-executed meta-analysis that summarizes the available observational studies that explore this relationship.1 Corroborating prior studies, their meta-analysis indicates a “likely association” between periodontitis and preterm birth or low birthweight, with approximately a 2-fold increase risk for preterm birth in patients with periodontal disease. We applaud the authors for explicitly identifying the unavoidable limitations of the analysis and caution the reader to consider these limitations before using these results to alter clinical practice. Three issues limit our ability to clinically apply findings from this area of research. First, the validity of the meta-analytic results is limited by the quality of the studies that comprise the analysis. Second, as the authors mention, association does not always correspond to causation, and third, we can never leap from observation of an association to proof of efficacy of an intervention.
See related article, page 135
The first issue restricting the clinical application of these findings is the inherent methodologic limitations of meta-analysis of observational studies. In meta-analysis, results from independent studies that are considered “combinable” are integrated in a statistical procedure to arrive at conclusions about a body of research. However, if heterogeneity across studies exists, combining the results may underestimate the variance and the size of the confidence interval,2 thereby overestimating the precision of the pooled estimate (odds ratio or relative risk) or producing a biased pooled estimate. Jean-Noel and Michel carefully considered the role of heterogeneity in their meta-analysis and found significant heterogeneity across studies. Among the 17 studies included in this meta-analysis, there are great variations in the study designs (ie, case-control, cohort, and cross-sectional studies), assessment of exposure (ie, periodontitis) and adverse pregnancy outcome (ie, preterm birth, low birthweight, preterm low birthweight, and intrauterine growth restriction), and control of confounding. In an attempt to delineate factors that may account for the heterogeneity, the authors conducted subgroup analysis and meta-regression by study characteristics, including study design and measurement of exposure and outcomes. On the basis of statistically nonsignificant test results, the authors concluded that differences in study designs and assessment of exposures and outcomes could not have accounted for the heterogeneity. But, among the 17 observational studies included in the meta-analysis, a standardized outcome measure was clearly lacking. Hence, concluding that heterogeneity is not creating a biased pooled odds ratio may not be warranted until further exploration of variations is conducted.
A major limitation with statistical testing of heterogeneity is that the statistical tests have notoriously weak power and often fail to reject the null hypothesis of homogeneity even if substantial differences between studies exist.2, 3 Small P values suggest that the heterogeneity should not be ignored, but large P values should not be unequivocally interpreted as justifying the homogeneity assumption.4 Regarding subgroup analysis, if there is important within-group heterogeneity (heterogeneity not accounted for by the grouping variable), the standard error tends to be underestimated; consequently, statistical tests based on standard error will not be valid.4 Therefore, when heterogeneity is suspected, it should not simply be ignored after a statistical test is applied; rather, it should be scrutinized, with an attempt to explain it.3
In this meta-analysis, Jean-Noel and Michel suggested that 1 possible source of heterogeneity was the quality of studies based on statistical testing results. As a common practice, the authors weighted studies included in the meta-analysis based on a quality score. Quality scoring is subjective; it submerges important information by combining unequal study features into a single score.4 Item-specific analyses of the impact of each factor (eg, study design) would allow the data, rather than the investigators, to indicate the importance of each item in determining the estimated effect.4 Notwithstanding limitations, we commend Jean-Noel and Michel for their efforts in such a comprehensive and well considered meta-analysis and we strongly agree with their conclusion that the pooled estimate from their meta-analysis may be biased.
The later 2 issues, proving causation and demonstrating treatment efficacy, can be interrelated or they can function independently. The most definitive conclusion that we can draw from decades of preterm birth research is that the pathogenesis of premature birth is multifactorial, highly complex and variable. Even when study is restricted to spontaneous preterm birth, excluding indicated preterm birth, etiologic factors can include environment (eg, economic status), infection (eg, bacterial vaginosis), genetic variation (eg, cytokine polymorphisms), social habit (eg, smoking), and many others. As Jean-Noel and Michel point out, periodontal disease may not cause premature birth, but may share a common etiologic mechanism, such as a genetic predisposition for a hyperinflammatory response. Thus far, we simply cannot prove that periodontal disease causes the etiologic complex disease of preterm birth without further exploring the association. As a related example, others have shown that bacterial vaginosis (BV) and genetic variation at the gene locus for TNF-α each affect the risk of preterm birth independently and as interacting factors.5 We should heed lessons from past false assumptions that strong associations will translate to effective therapy. Recall that multiple studies revealed an association between BV and preterm birth, but antibiotic treatment of BV in controlled clinical trials did not reduce the risk of prematurity, and in one sub-population may have even increased risk.6, 7 Similarly, associations between various inherited thrombophilias and adverse pregnancy outcomes (eg, early pregnancy loss or fetal growth restriction) have been revealed in observational studies. The presence and strength of association vary significantly based on the specific thrombophilia and the outcome of interest.8 However, many clinicians and topic experts have adopted antithrombotic agents as standard therapy in an effort to prevent adverse pregnancy outcomes even though there are no reliable intervention trials to prove benefit or safety.9, 10, 11, 12
We identify only 3 published studies that applied a dental intervention for the prevention of preterm birth—2 randomized controlled trials and 1 nonrandomized cohort study.13, 14, 15 One group of investigators at a single center performed the 3 studies on a low socioeconomic status Chilean clinical population with periodontal disease. Initial results are promising, with an apparent reduction (66% or more) in preterm birth risk for patients treated with periodontal therapy. However, we view these studies with caution. Because the Chilean study population is fairly homogeneous, the results may not be generalizable to other populations. Further, each of the three studies had disparity between the treatment and control group for at least one of the following potential confounding variables: history of preterm birth, maternal age, marital status, and maternal weight gain. These factors may affect the accuracy of the estimate of the treatment effect even in a randomized trial.
In summary, dental interventions can not be recommended as preventive therapy for preterm birth based on the research currently available. Further observational research and clinical trials are needed to better characterize the association between periodontal disease and preterm birth and to explore the affects of dental therapy on pregnancy outcome. Where randomized clinical trials are impractical, meta-analysis allows us to draw clinically relevant conclusions if done correctly. In particular, the thorough consideration of heterogeneity between observational study results in a meta-analysis will generally provide more insights than the mechanistic calculation of a pooled estimate, which often is biased.16
References
- . Preterm low birth weight and maternal periodontal status: a meta-analysis. Am J Obstet Gynecol. 2007;In press
- . Metaanalyses are observational studies: how lack of randomization impacts analysis. Am J Gastroenterol. 2005;100;:1233–1236
- . Meta-analysis: principles and procedures. BMJ. 1997;315:1533–1537
- . Meta-Analysis. In: Rothman KJ, Greenland S editor. Modern epidemiology. Philadelphia, PA: Lippincott-Raven Publishers; 1998;p. 643–673
- Association of polymorphism within the promoter of the tumor necrosis factor alpha gene with increased risk of preterm premature rupture of the fetal membranes. Am J Obstet Gynecol. 1999;180:1297–1302
- . Antibiotics for treating bacterial vaginosis in pregnancy. Cochrane Database Syst Rev. 2005;CD000262
- Failure of Metronidazole to prevent preterm delivery among pregnant women with asymptomatic Trichamonas vaginalis infection. N Engl J Med. 2001;345:487–493
- . Thrombophilic disorders and fetal loss: a meta-analysis. Lancet. 2003;361:901–908
- The management of thrombophilia during pregnancy: a Canadian survey. J Obstet Gynaecol Can. 2002;24:946–952
- . Anticoagulants in pregnancy. Obstet Gynecol Surv. 2005;61:59–69
- . Anticoagulants for the treatment of recurrent pregnancy loss in women without antiphospholipid antibody syndrome. Cochrane Database Syst Rev. 2005;CD004734
- Thrombophilia in pregnancy: a systematic review. Br J Haematol. 2005;132:171–196
- . Periodontal therapy may reduce the risk of preterm low birth weight in women with periodontal disease: a randomized controlled trial. J Periodontol. 2002;73:911–924
- . Higher risk of preterm birth and low birth weight in women with periodontal disease. J Dent Res. 2002;81:58–63
- . Periodontal therapy reduces the rate of preterm low birth weight in women with pregnancy-associated gingivitis. J Periodontol. 2005;76(Suppl.):2144–2153
- . Meta-analysis spurious precision? (Meta-analysis of observational studies). BMJ. 1998;316:140–144
PII: S0002-9378(06)01227-0
doi:10.1016/j.ajog.2006.09.029
© 2007 Mosby, Inc. All rights reserved.
Refers to article:
- Preterm low birth weight and maternal periodontal status: a meta-analysis
Volume 196, Issue 2 , Pages 93-94, February 2007
