Association between coronal caries and malocclusion in an adult population

The aim of the population-based SHIP was to estimate the prevalence of a broad range of diseases, risk factors, and health-related factors for the Northeast German population. The baseline examination SHIP‑0, whose sampling method was adopted from the World Health Organization MONICA (Monitoring Trends and Determinants in Cardiovascular Disease) Project in Augsburg, Germany, was approved by the local ethics committee and performed between 1997 and 2001 [28]. The net sample (without migrated or deceased subjects) comprised 6265 subjects with an age range from 20 to 79 years. Finally, 4308 subjects—all were Caucasian—gave written, informed consent and participated in SHIP‑0, which corresponded to a response rate of 68.8%. SHIP‑0 comprised a medical examination, a clinical dental examination (including periodontal, orthodontic, functional, and cariologic components), an interview, and a questionnaire completed by each participant [26, 28].

The occlusal status was assessed according to selected occlusal parameters including the sagittal intermaxillary relationship in the canine region. This relationship was registered separately for the right and left canine regions and determined as neutral, distal by the width of ½ premolar and 1 premolar, and mesial by at least a ½ premolar width [25]. The following signs were recorded as being either present or absent: frontal and lateral crowding, ectopic position of canines, widely spaced teeth without approximal tooth contact, frontal and lateral crossbite, buccal nonocclusion, excessive overjet and overbite, edge-to-edge bite, open bite, negative overjet and retruded position of maxillary incisors. Orthodontic status was not recordable when in 2 or more sextants of the dentition (2 anterior and 4 posterior tooth regions), 3 or more teeth per sextant were missing, regardless of whether the gaps were restored or not. Third molars were not included in the evaluation.

According to the WHO recommendations [42], coronal caries findings (cavitated carious defects into the enamel and dentine), fillings, secondary caries on the surface level, and missing teeth, were registered by surface with the exception of third molars according to the half-mouth method (quadrants 1 and 4, or quadrants 2 and 3 in alternating sequence) using a periodontal probe (PCP 11, Hu Friedy, Frankfurt am Main, Germany) [26, 39]. Cavitated carious lesions (D component) were subdivided into lesions confined to enamel and those involving dentine. The number of cavitated lesion solely in enamel was absolutely minimal (n = 72). Initial caries lesions without cavitation were not recorded or counted for the caries scores. In detail, caries was defined in the manual of SHIP‑0 as follows:

This was the basis for the calculation of the DMFS index to characterize the SHIP sample in Tables 1 and 2, and to analyze the data using four ordered outcome levels on tooth level as described in more detail in the statistical analyses section.

Visual inspection and probing with the dental probe PCP11 determined the presence or absence of plaque and calculus on test teeth 1, 3, and 6 in the selected quadrants, and the proportion of sites with plaque was calculated per participant. If a test tooth was missing, the distal adjacent tooth was examined instead. Each of these teeth was scored at four sites: distobuccal, midbuccal, mesiobuccal, midlingual.

Eight experienced and calibrated dentists performed the dental examinations. Training of examiners and consensus discussions were carried out before the study started and training/calibration sessions were repeated twice yearly while the study was ongoing. Orthodontic calibration of the examiners was based on the examination of 30 pairs of casts showing complex symptoms of malocclusion, examination was repeated after several days. Intra- and interexaminer agreement were measured by Cohen’s kappa (κ) [25, 26]. Cohen’s κ values ranged from 0.66–0.81, meaning “good agreement” [41]. The calibration exercises for the caries scores consisted of each examiner performing two examinations on each of 10 and 5 test participants one to two weeks apart. Examiners applied the eight categories for caries as described in the manual for SHIP‑0. On surface level, which was the basis for calibration and certification, very good Cohen’s κ values were reached for intra- and interexaminer reliability (0.9–1.0 and 0.93–0.96, respectively [26, 39]). On the tooth level as used herein, good κ values were reached for intra- and interexaminer reliability (0.69–1.0 and 0.70–1.0, respectively).

To avoid selection bias, subject’s age range was restricted to 20–39 years; older subjects have a higher proportion of missing orthodontic variables due to missing teeth. As shown for the relationship between malocclusion and periodontal disease [4], confounding by tooth type across jaws required modelling on subject, jaw, and tooth levels. As is common in multilevel analyses [16], the outcome (caries) is measured on the tooth level, whereas some covariates are at the subject level, for example gender, and other covariates are at the tooth level, including all malocclusion variables except distal and mesial occlusion [4]. Thus, the 33 malocclusion variables on the subject level were transformed into 18 corresponding variables on the tooth level [4]. Thus, ectopic canines on the tooth level could occur only at 13, 23, 33, or 43 [4]. For crowding (and spacing as well), a single variable instead of two variables for anterior and posterior regions may be desirable. We addressed this coding scheme only in sensitivity analyses because the six joint tests for the global malocclusion conditions, including space conditions in the anterior region and lateral malocclusions, were clearly of clinical and statistical interest. Moreover, crowding was assessed differently in the anterior and posterior regions. The malocclusion variables were simultaneously fitted in ordinal logistic multilevel models using the “meologit” procedure (Stata software, release 14.2; Stata Corporation, College Station, TX, USA). The four ordered outcome levels were (1) sound, (2) carious defects into the enamel, (3) caries (dentine caries ≤3 mm, dentine caries >3 mm, filling, or secondary caries), and (4) tooth loss. Because pitfalls of ignoring the hierarchy in dental research (subject, tooth, surface; subject, jaw, tooth) have been well-known for 20 years [17], multilevel models have been widely used for answering complex research questions, especially when the tooth type is a confounder on a level different from the subject level [4, 18]. Herein, the three hierarchical levels subject, jaw, and tooth were included as random effects [36]; age, gender, school education (3 levels in accordance with the former east German school system), marital status (5 categories), jaw, tooth type (7 levels), the interaction between jaw and tooth type [21], and monthly household equivalence income (1 € = 1.956 German marks) were included as fixed effects [30]. Restricted cubic splines with three knots were used to allow for departures from linearity for age and income. Income was considered only in additional analyses because, unlike school education, it was linked with adulthood rather than childhood and, therefore, not assumed to be a confounder. As orthodontic treatment is part of the effect to be studied, it was not included into the model because “a confounder must not be an effect of the exposure” [37]. Odds ratios (OR) with 95% confidence intervals (CI) and p-values are provided. For any cut point of the outcome on four levels, ORs in ordinal logistic regression models can be interpreted as those in binary logistic regression models; note that the ordinal logistic regression model has fewer assumptions than the ordinary least squares regression model [22].

On the tooth level, the following malocclusions were associated with an increased odds ratio for caries, or more exactly, for tooth loss versus no tooth loos; or tooth loss or caries versus no caries; or tooth loss, caries, or enamel caries versus sound (Table 4): anterior open bite ≤3 mm (OR = 2.08, CI: 1.19–3.61, frequency among all incisors 2.9%) and increased sagittal overjet of 4–6 mm (OR = 1.31, CI: 1.05–1.64, frequency among all incisors 25.0%). Increased sagittal overjet of >6 mm (OR = 1.45, CI: 1.00–2.11, frequency among all incisors 8%) displayed a p-value of <0.1. Distal occlusion according to the sagittal intermaxillary relation in the canine region also displayed higher odds for caries with distal ½ premolar width (OR = 1.27, CI: 1.05–1.53, frequency among all teeth 28.9%) and distal 1 premolar width (OR = 1.31, CI: 1.06–1.63, frequency among all teeth 19.4%). For negative overjet, the data are consistent with a true OR between 0.84 and 5.62 (frequency among all incisors 1.1%). Some malocclusions were associated with a significantly reduced odds for caries: anterior spacing (OR = 0.24 CI: 0.17–0.33, frequency among all incisors: 10.4%), posterior spacing, (OR = 0.69 CI: 0.50–0.95, frequency among all posterior teeth 4.7%), posterior crowding (OR = 0.57 CI: 0.49–0.66 frequency among all posterior teeth 28.0%) and buccal nonocclusion (OR = 0.54 CI: 0.33–0.87, frequency among all posterior teeth: 1.7%), (Table 4).

Joint effects occurred for space conditions in the anterior region (p < 0.0001 for the global test with 5 degrees of freedom; Table 4), space conditions in the posterior region (p < 0.0001), vertical overbite (p = 0.0412), sagittal overjet (p = 0.0325), lateral malocclusions (p = 0.0051), and sagittal intermaxillary relationship in the canine region (p = 0.0200). The joint effect for increased sagittal overjet and distal occlusion, which were correlated, was statistically significant (p = 0.0011 for the global test with 4 degrees of freedom).

Whereas anterior and posterior spacing can be combined into a single spacing variable in a natural way, posterior crowding can be combined with different levels of anterior crowding. Counting posterior crowding as the lowest level of the presence of anterior crowding, the ORs were 0.65 (95% CI: 0.58–0.74; p < 0.0001), 0.64 (95% CI: 0.43–0.95; p < 0.0255), and 0.60 (95% CI: 0.17–2.14; p = 0.4348) from the lowest to the highest crowding level, respectively. The OR of spacing was 0.38 (95% CI: 0.30–0.48; p < 0.0001). Counting posterior crowding as the middle level of anterior crowding, the OR of the middle level was 0.56 (95% CI: 0.49–0.65; p < 0.0001). Of note, the 95% CIs for anterior and posterior spacing did not overlap in the main analysis (Table 4).

Including household income did not lead to a change >10% in the ORs of malocclusion variables in the reduced sample of 1171 subjects.