History of periodontal treatment and risk for intrauterine growth restriction (IUGR)

This retrospective cohort study examined insurance records of women concurrently enrolled in medical and dental insurance plans through Aetna Inc., a nationwide, private health insurer. Aetna’s data warehouse holds claim information on members for a four-year period. We therefore restricted the analysis to women that delivered a singleton live-birth between January 2010 and December 2011, and then included claims data for those birth events up to one year before gestation and up to one year after the birth event. We chose to restrict the study to singleton gestations, since the etiology and risk factors for IUGR vary between singleton and multiple births. The study cohort was restricted to women who had both medical and dental insurance, who were between 13 and 50 years of age at the time of delivery, and for whom zip code level data and other covariates were available. The analytic sample included 32,168 women.

Using data from 212,427 dental claim records for the period January 2009 through December 2012, procedures performed before, during and after pregnancy were identified and classified using the Code on Dental Procedures and Nomenclature Current Dental Terminology (CDT) 2009–10 edition for procedures occurring before 2011, and the CDT 2011/2012 edition for procedures from 2011 onwards. Maternal oral treatment type included periodontal treatment, prophylaxis, other dental treatment, or no oral treatment of any kind during the study period. Periodontal treatment was provided by general dentists and periodontists. Oral prophylaxis procedures can be provided by dentists or dental hygienists, however in the United States most oral prophylaxis procedures are provided by dental hygienists. Periodontal treatment included surgical and non-surgical codes (Additional file 1: Table S1). Because the insurance database did not include indicators of conception, the date of conception was estimated from the date of birth. For analytic purposes, each dental treatment category was considered in relation to the time period in which it occurred: pre-conception, during gestation, or after birth. The insurance dataset contained diagnosis codes and treatment codes for medical care, however it was limited to treatment codes for dental services. In the United States, dentists are only required to provide treatment codes for reimbursement via a dental claim. Given that periodontal disease is a chronic condition, periodontal treatment occurring after delivery was considered to indicate that periodontitis was present and untreated during pregnancy.

Using data extracted from 2,622,764 medical claims records, for the period January 2009 through December 2012, presence of IUGR was assumed if one or more International Classification of Diseases (ICD) 9th revision codes indicating slow fetal growth and fetal malnutrition (764.x), or poor fetal growth affecting management of mother (656.5x) had been used. Fetal growth restriction was defined as an IUGR diagnosis within 14 days of delivery; thus, in this study, the term IUGR refers to growth restriction defined accordingly. In each record, both primary and secondary claims were searched. Multiple births were identified and eliminated from the analytic file using supplementary delivery codes V27.2 to V27.9.

Maternal income, race and ethnicity were not available in the insurance database. We therefore imputed these characteristics based on the zip code of residence for each woman at the time of insurance enrollment, and the 2010 United States Census zip code level data for these variables. We identified complications of pregnancy using ICD-9 codes (Additional file 1: Table S2). Because not all women had dental coverage for the entire study period, months of enrollment for each woman was included as a covariate. The sample included women from all 50 states, the District of Columbia, and Puerto Rico.

The risk of IUGR was associated with types of dental and periodontal treatment, as previously defined. We fit logistic regression models from which we derived odds ratios (OR) and 95% confidence intervals (CI) to assess the magnitude of the effect. In these models, we adjusted for potential confounding factors including pregnancy complications, maternal age at delivery, primiparity, zip code level income, race, and ethnicity; the models were also adjusted for duration of continuous dental coverage during the study period. Annual income was analyzed in quintiles, and categorized in US dollars as ≤28,125; 28,126 to 33,500; 33,501 to 39,283, 39,284 to 48,831; and ≥ 48,832. To assess non-linear effects of maternal age on the odds of incident IUGR, we included a quadratic term for age in the regression models.

To determine whether risk of IUGR increases with severity of periodontitis, we examined the rate of IUGR in relation to the number and type of periodontal treatment procedures, and compared women who received surgical as opposed to non-surgical periodontal treatment.

Since the association between periodontal treatment and IUGR is likely affected by unmeasured confounding, we undertook a sensitivity analysis to evaluate the extent to which unmeasured confounding may have impacted the associations [11]. This sensitivity analysis was based on the following assumptions: (i) the prevalence rates of the unmeasured confounder among pregnancies with and without IUGR were 3% and 6%, respectively; and (ii) the odds ratio of IUGR comparing the presence versus absence of unmeasured confounding was allowed to vary between 0.1 (protective) to 1.0 in 0.1 increments, and 2.0 (increased risk) to 10 in 1.0 increments. The evaluation of the impact of unmeasured confounding was based on fairly conservative assumptions.

The odds ratios corrected for unmeasured confounder(s) are shown in Fig. 1. For instance, if the odds ratio of IUGR comparing the presence versus absence of an unmeasured confounder was 2.0, the bias-corrected odds ratio for each of the three scenarios were 0.7 (95% CI 0.5, 1.1) for pre-gestation periodontal treatment, 1.2 (95% CI 0.9, 1.8) for periodontal treatment during pregnancy, and 1.5 (95% CI 1.3, 1.8) for periodontal treatment post-gestation. For odds ratios of the unmeasured confounder over 5, the bias-corrected odds ratio were enhanced for both periodontal treatment during pregnancy and post-gestation. These findings confirm the confounder-adjusted odds ratios reported earlier (Table 3), and when unmeasured confounding is taken into account, the associations between periodontal treatment both during pregnancy and post-gestation are associated with increased odds of IUGR.

Diagnosis of IUGR was based on ICD coding, and this may have introduced some misclassification. Women carrying IUGR fetuses, particularly those that are diagnosed as not being severely growth restricted, are less likely to undergo clinician-initiated obstetrical intervention (labor induction or a prelabor cesarean) and less likely to have a diagnosis of IUGR recorded [12]. However, for severe IUGR (e.g., estimated fetal weight below the third or the first percentiles), misclassification of IUGR status is very unlikely since growth restriction serves as a sentinel cause for obstetrical intervention, and is therefore billed for insurance reimbursement. For the same reason, we believe the recording of the exposure is accurate in this data system because periodontal treatment is the basis for a reimbursable claim.

Second, despite adjustment for several confounders, we lack data on smoking and maternal pre-pregnancy body-mass index. However, the sensitivity analysis conducted to determine the potential effect of unmeasured confounding indicates that our models are robust. Finally, dental data is based on CDT codes that reflect treatment of periodontal disease rather than diagnosis. However, we believe that the treatment codes are sufficiently specific to infer presence of periodontal disease. While women included in this study are from virtually all states in the US, the insured populations are from middle to higher-income socioeconomic strata. This should be considered while generalizing the results from the study; however, it is unlikely that the association between IUGR and infections in general, and periodontal disease in particular, would be any lower among poorer women than among those we studied.

Finally, while we report that there is an increase in the rate of IUGR as a function of increasing number of treatments for periodontal disease, there are many factors that determine frequency of treatment; therefore, this finding is only suggestive of a relationship between the severity of periodontal disease and increased risk of IUGR.

In this study, a large sample of integrated medical and dental claims data provided a unique opportunity to explore the association between IUGR and periodontitis. In addition, the findings appear robust following adjustment for observed confounders, in fact correction for unmeasured confounding makes the associations stronger. Conducting secondary analyses using insurance data to shed light on the possible causes of negative birth outcomes is highly economical, and valuable in suggesting directions for future research.

The finding that periodontal treatment post-gestation was associated with an increased risk of IUGR in a large national sample add to the growing body of literature indicating a relationship between periodontal infection and related inflammation with adverse birth outcomes [4‐8, 13]. Periodontal treatment in the period immediately following gestation is interpreted as signifying that periodontal disease was present during gestation. The inflammatory process associated with periodontal disease and the presence of periodontal pathogens in the blood can affect the fetus and the placenta [4].

This study utilized periodontal treatment as a proxy for the presence of periodontal disease. Periodontal treatment during gestation was expectedly rare and was also low during the pre-gestational period; our sample size was therefore too small to evaluate effects of treatment during pre-gestation and gestation. We expect that treatment during the pre-gestational and gestational periods to have limited adverse impact on birth outcomes. Tonetti and colleagues observed a short term increase in the systemic inflammatory response immediately following periodontal treatment which was then followed by a decrease in inflammation [14]. The finding that deleterious effects associated with periodontal therapy are short-lived is consistent with our finding of no statistically significant effect of treatment in the period prior to gestation and during gestation. However, periodontal treatment provided immediately following birth, which we found to be significantly related to IUGR appears to be a marker of disease during gestation.

Boggess and colleagues also observed that the incidence of small for gestational age increased with periodontal disease severity [15]. These findings are consistent with observations by several other investigators. In a study of Brazilian women, Siqueria and colleagues found increased odds of IUGR (adjusted OR 2.06, 95% CI 1.07, 4.19) among women diagnosed with periodontitis [16]. Similarly, Kumar and colleagues reported an increased association between periodontitis and IUGR, which was attenuated after adjusting for confounders [17]. The associations that we report are very similar to those of the Brazilian study.

The insured and employed population in our analyses is in the upper quartile of income in the United States and would be expected to have better oral hygiene and prevention practices. In addition, it is expected that utilizing treatment as a proxy for periodontal disease to some extent underestimates the true incidence of periodontal disease. In our analytical sample, maternal periodontal disease, indirectly assessed through the delivery of periodontal treatment in the immediate post-partum period, affected 9% of the women. By comparison, earlier studies have reported prevalence rates of 56–61% for maternal periodontitis [18, 19]. It should be noted that while we and Siquiera and colleagues [16] observed an increased odds for growth restriction with periodontal disease, other studies did not find such a relationship [20].