This is a retrospective cohort study that followed STROBE checklist. It was approved by the office of human research administration, Harvard Faculty of Medicine, [45 CFR 46.101(b) (4)], #IRB 16-1838. We collected records of all patients that reported receiving oral BIS from 2008–2015 (N=30), to identify suitable radiographs for analysis. To be included in the study, each patient should have at least two exposures of Complete Mouth Radiographic Series (CMRS) or repeated bitewing (BW) radiographs at least a one-year interval. Furthermore, each BW radiograph had to clearly show the alveolar bone crest and cement-enamel junction, as well as show at least two posterior approximating teeth to be included. Exclusion criteria used were: 1) patients that were not within the specified age range, 2) patients with no BW radiographs, 3) patients with radiographs in which the cement-enamel junction (CEJ) and alveolar bone crest were not visible, 4) patients who did not have at least 2 approximating teeth or where the interproximal space was too narrow to observe the bone crest. Teeth were excluded if dental restorations obliterated the CEJ, rendering the distance between CEJ and alveolar crest questionable. Additionally, cases in which a tooth was found adjacent to an edentulous site with alveolar bone levels greater than 2mm from the CEJ were not considered pathological due to possible surgical trauma. Any records indicating sites receiving osseous surgery or bone grafts were excluded. Patients were excluded as well due to closed electronic files or because their BW radiographs could not be calibrated with the measuring tool. Third molar teeth were not included due to their tendency of not being captured by BW radiographs. Non-functional teeth were excluded for the possibility of super eruption. For longitudinal data analysis, we required that eligible subjects for inclusion to have at least two exposures of complete mouth survey radiographs or repeated BW radiographs with at least one-year interval.
Primary predictor:
The main predictor was whether or not the subjects had reported taking oral BIS. Other variables included in the model were age, sex (although we did not expect any confounding by age or sex, since the two groups were matched on them, we included them to account for any residual confounding), smoking status, median house income, race, diabetes and hypertension. All data were collected from the electronic health record of the Harvard Dental Center using AxiUm® software. Due to the small number of this sample, categorization to different age categories resulted in groups with very few subjects (presented in the descriptive statistics section). Hence, age was used as a continuous predictor for the multivariable analysis. Furthermore, we categorized body mass index (BMI) into two groups of Underweight/Normal weight and Overweight/Obese, with the former group as the reference group for the same reason of scarce data.
In this sample, no one had reported as being current smoker so we created a binary smoking variable for analysis by coding everyone who had ever smoked (former smoker) as ever smoker (=1) and those who had never smoked as never smoker (=0).
Primary outcome:
The primary outcome is the mean of the alveolar bone level on mesial and distal sites of posterior teeth in millimeters between the group that was taking oral BIS and the group that was not. The bone levels at the follow-up visits were compared to the baseline mean of both groups. Interproximal bone loss occurs when the distance between the cement-enamel junction (CEJ) and the alveolar bone crest is greater than or equal to 2 mm, as determined on a bitewing radiograph (26-31). We also classified the amount of bone loss based on the American Academy of Periodontology (AAP) case definition into mild, moderate and severe periodontitis to estimate the prevalence (26). One trained examiner (MH) carried out the measurement of the outcome using the calibrated measuring tool of Emago® after conducting the inter-examiner reliability test. An intra class correlation coefficient (ICC) test was performed with an excellent average score of 0.96 (0.93-0.97).
Sample size:
This is a subpopulation of a large sample size (N=1131) collected to estimate prevalence of periodontal diseases in HSDM. After identifying patients who reported using bisphosphonates, we implemented very strict exclusion criteria (described before). We further selected patients on the availability of repeated radiographs that affected the total number of eligible patients. We collected records of all patients that reported receiving oral BIS from 2008–2015 (N=30), to identify suitable radiographs for analysis. We identified 26 patients out of the 30 identified earlier that satisfied inclusion criteria described before. The 26 patients who were taking BIS were then matched on age and sex to another 26 patients who did not report receiving BIS at any point of their life. Radiographs of a total of 52 patients (26 patients of each group) were analyzed over a two-year period.
It is true sample size is quite small due to the low number of patients reported taking BIS. To address this problem we calculated the power of detecting at least 0.5 mm difference between the two groups. The main sample of this subpopulation had an average mean alveolar bone level of 1.38 mm (±0.7). Given these parameters, with a set to 0.05, we have more than 80% power to detect a real difference. This could also mean that the two groups might differ in less than 0.5 mm of mean alveolar bone level. However such a difference might be considered clinically insignificant.
Statistical Analyses:
Descriptive statistics of categorical data as well as the prevalence of each periodontitis case definition were calculated. A mixed-effect linear regression model with a multi-level design was performed to estimate the difference of change in mean bone level in millimeters (mm). We included the time term to adjust for the amount of change across the years of follow up for both groups.