Study design and setting
A retrospective unmatched case-control study was conducted to test the association link between periodontitis and premature birth. The study is considered retrospective because both the exposure and the outcome happened prior to subject enrolment. In this study the outcome was preterm delivery and the hypothesized exposure was presence of maternal periodontitis. Participants were selected through hospital registers whereby the research team would verify cases of preterm deliveries and the controls would be selected from the next 2 full-term births on the register. An a priori estimate of the required sample size was calculated using G*Power 3.1.9.6. by Faul [19], Kiel, Germany. The researcher used a small effect size, an alpha of 0.05 and a power of 0.95 with 10 predictors. The estimated sample of cases of 185 was combined with a 1:2 ratio of unmatched controls, totalling 370, thus the total required sample size was 555. The research team completed the exams and questionnaires with study subjects until the required sample was reached. The study was conducted at 12 health facilities in in 6 districts of the Southern Province of Rwanda from February to August 2018, including Kamonyi, Muhanga, Ruhango, Nyanza, Huye and Gisagara. Corresponding district hospitals and nearby health centers were selected for the study as well as a referral hospital in Huye.
Data sources and measurement
A structured clinical exam and standardized questionnaire were used to collect information regarding the presence of periodontal infection among pregnant women attending antenatal care clinics in the Southern Province of Rwanda. The study adapted the questionnaire from the WHO Oral Health Assessment Tool for Adults of 2013 [20] for use in the Rwandan context. The questionnaire was sent to experts for content validation.
Prior to use in the study in southern Rwnda, the questionnaire was piloted in Nyamata District Hospital in the Eastern Province to ensure that it captured all the required information and that the questions were clear to respondents. After the pilot study, the feedback from the participants was considered and any questions that were not clear were corrected accordingly. The questionnaire was translated in Kinyarwanda using forward and backward translation whereby this questionnaire was first translated in Kinyarwanda from English and then it was again translated back to English by another translator to determine whether the content remained consistent.
The questionnaire collected information on the following variables: age of the respondent, education level, health and lifestyle behaviors (e.g. tobacco use), socio-economic status, mother’s weight, number of previous pregnancies, previous preterm delivery, weight gain during pregnancy, illnesses during pregnancy and stress during pregnancy. The main outcome variable of interest was premature birth.
Participants and inclusion/exclusion criteria
The inclusion criteria were postpartum mothers aged 18–35 who delivered singleton infants in all selected health facilities within 1 to 5 days before recruitment. Those women who delivered premature were recuited as cases and those who delivered at term were recruited as controls. Maternal registers were checked in order to determine who could be recruited as case or controls based on the above criteria. Cases and controls were enrolled in a ratio of 1:2 and each enrolled case of premature birth was followed by 2 control subjects who delivered at term gestation that were next on the register. A case was defined as any delivery below 37 weeks and a control was defined as any delivery of 37 weeks and above. We used ratio of 1:2 in order to increase the study power and reduce data collection time as there were limited cases of premature birth. Mothers with twin infants, those with systemic conditions like uncontrolled diabetes, HIV infection and those without teeth in one or more sextants were excluded from the study. Women established risk factors for prematurity, like abnormal placentation, eclampsia, uterine abnormalities, and other pregnancy complications that may easily lead to prematurity were also excluded from the study so that they would not bias the results. Other variables that were likely to cause prematurity were documented via the study questionnaire and were controlled during the multivariate analysis.
Data collection procedures
A periodontal examination was performed on all women enrolled in the study. The study examiners used a Williams graduated periodontal probe to perform the periodontal clinical examination [21]. Six data collectors were calibrated by a qualified dentist on how to perform a periodontal examination to ensure that they all understood it in the same way, to help reduce the chance of reporting different findings for the same patients. The calibration process focused on probe design, gentle and constant force of probing and proper angulations. While no formal agreement analysis was conducted, the examiners were required to examine at least 2 patients and the oral examination training was conducted until all examiners were in agreement with each other and with the trainer. The calibration results was considered valid only when there was no descripancy in results and when the results between the two examiners were consistent. The examinations assessed bleeding on probing, probing depth and clinical attachment loss measured in mm at six different sites on each tooth (buccal-mesial, mid-buccal, buccal-distal, lingual-mesial, mid-lingual and lingual-distal). The study subjects were examined at their bedsides by the lead researcher and calibrated research assistants using a periodontal probe, intraoral mirror and headlights. A gentle probing force was applied to guide the tip into the periodontal pocket until the resistance was felt. The pocket depth was measured using the gingival margin as a reference point. The mothers in the study group were also asked about their dental care practices and their smoking habits. To help prevent bias, the research team members were blind to the study status of the cases and controls.
The current study defined periodontitis as presence of pocket depth greater than 3 mm on either maxilla or mandible or both and presence of interdental clinical attachment loss (CAL) on ether maxilla, mandible or both of 2 mm or above and buccal or oral CAL of 3 mm or above [22].
Based on Jati et al [23], gingival recession was defined as “apical migration of marginal gingiva and characterized by gradual displacement of gingiva away from the cemento-enamel junction that results in the root surface exposure to the oral environment”.
Clinical attachment loss was measured as follows: when the gingival margin was at the cemento-enamel junction and there was no recession, then the CAL was equal to the pocket depth; when the gingival margin was apical to the cemento-enamel junction, CAL was equal to pocket depth plus gingival recession; when the gingival margin was on the anatomical crown in case of gingival overgrowth, CAL was equal to pocket depth minus gingival recession. CAL was not considered in some of the specific cases that were of non-periodontal cause. For example, when the gingival recession was of traumatic origin, dental caries extending in the cervical area of the tooth and in the cases of recession by malposition of the tooth [22].
The current study defined some of the questionnaire variables as follows; i) physical trauma as a wound on the body that was caused by a sudden physical injury, for example, an accident; ii) violence as any behavior or action that intends to hurt someone, physical or verbal and iii) stress during pregnancy as anything that causes emotional strain or tension to the pregnant women.
Data analysis
IBM SPSS for windows version 21.0 (IBM corp., Armonk, New York, USA) was used for all data analyses [24]. Descriptive statistics, including chi-square analysis was conducted as part of the background to the main hypothesis testing analysis that used multiple logistic regression. The study regression model was built using a three-step modelling approach. Hierarchical multivariate logistic regression analysis was used with the study variables being entered into the analysis using three groups: the demographic variables that were significant in the univariate analysis (age and employment status) were entered first in the regression model as step 1, followed by the second group in step 2, which included other known potential risk factors and confounders in this study such as ever used tobacco, mother’s weight, inter-conception period, whether premature delivery was experienced before, whether stress was experienced during pregnancy, malaria during pregnancy, urinary tract infection, physical trauma and violence during pregnancy. The third and final step of the regression analysis, added periodontitis status to the model, as it was hypothesized as the study’s main exposure variable.
The odds ratio was obtained with 95% confidence intervals, thus statistical significance was defined as p < 0.05. The fit of the logistic regression models was assessed using regression diagnostic information, such as the − 2 log likelihood and Hosmer-Lemshow chi square values.