Abstract
Objectives:
Clinical measures of periodontal disease such as attachment loss (CAL) and probing depth (PD) vary considerably between and within individuals with periodontitis and are known to be influenced by person-level factors (e.g., age and race/ethnicity) as well as intraoral characteristics (e.g., tooth type and location). This study sought to characterize site-level disease patterns and correlations using both person-level and intraoral factors through a model-based approach.
Methods:
This study used full-mouth, 6 sites per tooth, periodontal examination data collected from 2,301 Hispanic/Latino adults aged 60–74 years in the Hispanic Community Health Study/Study of Latinos (HCHS/SOL). The presence of site-level CAL ≥3 mm and PD ≥4 mm was estimated using generalized estimating equations (GEE), explicitly modeling pairwise periodontal site correlations, while adjusting for number of teeth, sex, and Hispanic/Latino background. Subsequently tooth- and tooth-site patterns of intraoral CAL ≥3 mm and PD ≥4 mm were estimated and visualized in the HCHS/SOL population.
Results:
The findings showed that posterior sites had the highest odds of CAL ≥3 mm and PD ≥4 mm. Sites located in the interproximal space had higher odds of PD ≥4 mm but lower odds of CAL ≥3 mm than non-interproximal sites. Mexicans had the lowest odds of CAL ≥3 mm among all Hispanic/Latino backgrounds. While Mexicans had lower odds of PD ≥4 mm than Central Americans and Cubans, they had higher odds than Dominicans and Puerto Ricans. Site-level proportions and pairwise correlations of PD ≥4 mm were generally smaller than those of CAL ≥3 mm.
Conclusions:
Patterns of site-level probabilities of clinical measures of periodontal disease can be defined based on tooth, site, and individual-level characteristics. Intraoral correlation patterns, while complex, are quantifiable. Risk factors for site-level CAL ≥3 mm may differ from those of PD ≥4 mm. Likewise, participant risk factors for site-level clinical measures of periodontal disease are distinct from those that affect individual-level periodontitis prevalence. Future epidemiological investigations should consider model-based approaches when examining site-level disease probabilities to identify intra-oral patterns of periodontal disease and make inferences about the larger population.
Keywords: Periodontal disease, intraoral correlation, generalized estimating equations, statistical methodology
Introduction
Descriptive evaluations of clinical measures of periodontal disease, such as attachment loss (CAL), have clearly established that intraoral patterns of occurrence, severity, and progression exist. In a study of adults with chronic periodontitis, the highest incidence of CAL ≥2.5 mm occurred on mandibular and maxillary molars and at interproximal sites.1 Novel overall and by age group visualizations of a nationally representative sample of U.S. adults corroborated these findings by demonstrating that the highest mean CAL occurred on maxillary and mandibular molars with some indication of higher mean CAL among sites in the interproximal space.2 While clinically recognizable, these patterns and intraoral correlations are typically not accounted for or used appropriately in epidemiologic and clinical investigations of periodontitis.
Much periodontal research is often based upon data obtained in a full mouth oral examination that consists of an evaluation of six sites [distobuccal (DB), buccal (B), mesiobuccal (MB), mesiolingual (ML), lingual (L), and distolingual (DL)] on 28 fully erupted permanent teeth (excluding third molars), resulting in a maximum of 168 periodontal sites. Typically, measurements of CAL, probing depth (PD), and bleeding on probing (BOP) are recorded at each site. Since these sites are naturally clustered within the mouth, standard statistical assumptions of independence are violated. Instead of analyzing site-level data, many oral epidemiological investigations focus on the prevalence of periodontitis, where individuals are identified as having periodontitis based on case definitions that aggregate information across available sites in the mouth, as well as the extent of periodontitis, which is often defined as the percent of measured sites meeting a particular CAL or PD criterion.
Two such investigations were performed by Jiménez et al3 and Sanders et al4 in the first large-scale study of U.S. Hispanics and Latinos since 1982. Both established that the burden of periodontitis varies by age, sex, and Hispanic/Latino background using the Centers for Disease Control and Prevention and the American Academy of Periodontology (CDC-AAP) periodontitis surveillance definition.5,6 Jiménez et al further examined the prevalence of CAL and PD severity using case definitions with thresholds such as having one or more sites with CAL ≥3 mm or having one or more sites with PD ≥4 mm.3 However, like most epidemiological studies, knowledge of site-level probabilities of disease and within-mouth correlations was lacking.
Thus, this study sought to expand the previous studies by characterizing site-level patterns of disease within the mouth and identifying effects of tooth-level and site-level characteristics on periodontal disease measurements, CAL ≥3 mm and PD ≥4 mm. The paired estimating equations method of Prentice,7 an extension to the usual GEE method that allows for modeling of both means (i.e., probabilities for binary site-level indicators of disease) and correlations, was applied to full mouth periodontal examination data collected in the Hispanic Community Health Study/Study of Latinos (HCHS/SOL). Specifically, this study aimed to determine whether there are differences across Hispanic/Latino backgrounds for site-level CAL and PD, adjusting for sex and accounting for variations in the pattern of existing teeth. Furthermore, this study compared model results for CAL and PD to provide further insight into the patterns of site-level disease for these fundamental clinical measures of periodontal health and disease.
Methods
Study design, setting, and selection of participants
The HCHS/SOL is a multicenter longitudinal population-based cohort study of Hispanic and Latino adults in the United States which aims to describe the prevalence of risk and protective factors for chronic conditions and disease incidence over time.8 The study used a stratified two-stage area probability design to recruit participants. Stratified simple random samples of census block groups, the primary sampling units (PSUs), were selected from each of four HCHS/SOL field centers located in the Bronx, NY; Chicago, IL; Miami, FL; and San Diego, CA. Stratified samples of household addresses were selected within each PSU. Details of the complex survey design have been previously published.9
This study examined older adults who self-identify as Hispanic/Latino, a population in which periodontitis is prevalent. Specifically, baseline clinic examinations of 2,792 Hispanic/Latino (Central American, Cuban, Dominican, Mexican, Puerto Rican, South American, or Mixed/Other) participants, ages 60–74, enrolled in the HCHS/SOL were conducted between 2008 and 2011. As a part of the baseline examination, eligible participants underwent a full mouth periodontal examination. At each study center, trained dental examiners measured PD, CAL, and BOP using a periodontal probe (UNC-12, Hu-Friedy, Rotterdam, Netherlands) with graduated 1 mm increments. Primary teeth and dental implants were excluded from assessment. Participants were excluded from this analysis if no teeth were eligible for periodontal measurements, i.e., no fully erupted permanent teeth were present or if clinical measurements could not be assessed for any tooth, resulting in an analytic sample of 2,301 (82%) participants. Participants provided written consent and the study was approved by the Institutional Review Boards of all participating institutions.
Site-level measures of periodontal disease
In this study, CAL ≥3 mm (CAL3+) and PD ≥4 mm (PD4+) were examined as clinical indicators of sites with evidence of periodontal disease experience. These site-level dichotomous outcome variables were defined because they correspond to minimum thresholds of CAL and PD used to indicate disease in various periodontitis surveillance case definitions, including the CDC/AAP5,6 and European Federation of Periodontology and American Academy of Periodontology (EFP/AAP) 2018 case definitions of periodontitis.10
Statistical methods
Descriptive statistics for the number of eligible teeth, number of CAL and PD sites measured, age, and sex were computed in the baseline HCHS/SOL population aged 60–74 using weighted analyses to account for complex sampling. Additionally, overall proportions of CAL3+ and PD4+ at each site and Pearson correlations for all site pairs were computed and plotted using heat map visualizations. Overall normalized HCHS/SOL participant sampling weights, adjusted for non-response and calibrated to the 2010 Decennial Census9 were used for all analyses to reflect unequal participant probabilities of selection and to obtain estimates applicable to the target population of Hispanic and Latino adults in the four communities.
Logistic regression models were fitted separately for both CAL3+ and PD4+ including effects for tooth location (anterior versus posterior), site location [sites located in the interproximal space versus sites not located in the interproximal space], and participant-level characteristics for the log of the total number of teeth, sex, and Hispanic/Latino background. The correlation structure for both CAL3+ and PD4+ was modeled considering correlations between pairs of sites on the same tooth that are next to each other (adjacent sites) and share the same interproximal space, pairs of adjacent sites on the same tooth that do not share the same interproximal space, non-adjacent pairs of sites on the same tooth, pairs of adjacent sites on adjacent teeth on the same jaw that share the same interproximal space, non-adjacent pairs of sites on adjacent teeth on the same jaw, pairs of sites on teeth directly above and below each other (vertically adjacent) that have the same site location, pairs of sites on vertically adjacent teeth that have different site locations, and finally sites on non-adjacent teeth.
Models were fitted using the Prentice GEE method,7 consisting of a set of paired estimating equations and implemented using the SAS macro, GEECORR.11 The macro provides parameter estimates and empirical (sandwich) standard errors for both marginal mean and correlation covariates. Sampling weights were included as “cluster weights” for the joint estimation of mean and correlation models where a cluster consists of all site-specific observations in a mouth. Mean models were specified using the logit link (i.e., logistic regression) and correlation models were specified using the identity link. Odds ratios (ORs) and 95% confidence intervals (CIs) were computed from mean model estimates and standard errors. The inverse logit function was applied to mean model estimates to obtain adjusted site-level probabilities of disease. Finally, estimated site-level probabilities of disease and site-pair correlations were plotted using heat map visualizations. Analyses were performed using SAS/STAT and SAS/IML software Version 9.4. Copyright ©2002–2012 SAS Institute.
Results
Population estimates of baseline demographic and clinical characteristics are shown overall and by Hispanic/Latino background in Table 1. The average age was 66 years and there were slightly more females than males in the study population. Age and sex were consistent between Hispanic/Latino backgrounds, with the exception the overrepresentation of females among South Americans. Overall, the average number of fully erupted permanent teeth was 18.2 (95% CI: 17.6, 18.8; range 1–28). Clinical measurements of CAL and PD were taken at approximately 104 (95% CI: 100, 107; range 2–168) sites per participant. A greater proportion of posterior sites had CAL3+ and PD4+ than anterior sites. Non-interproximal sites had more CAL3+ but less PD4+ than interproximal sites. Crude overall site-level proportions of CAL3+ and PD4+ are also shown in Figure 1 and suggest that the highest proportion of CAL3+ occurred on the buccal site of the maxillary first molar. Other areas with high proportions of CAL3+ were lingual sites of mandibular anterior teeth and buccal and lingual sites of maxillary posterior teeth. In contrast, the highest proportions of PD4+ occurred on distolingual and mesiolingual sites of posterior teeth.
Table 1:
Characteristics of the HCHS/SOL Population Aged 60–74, 2008–2011
| Overall (N=2031) |
Central American (N=164) |
Cuban (N=342) |
Dominican (N=168) |
Mexican (N=779) |
Puerto Rican (N=406) |
South American (N=137) |
Other or More Than One (N=35) |
|
|---|---|---|---|---|---|---|---|---|
| Number of Eligible Teetha | 18.2 (17.6, 18.8) |
18.1 (16.5, 19.7) |
16.0 (15.0, 16.9) |
15.6 (14.2, 17.1) |
22.3 (21.7, 22.9) |
17.2 (16.0, 18.3) |
16.4 (14.7, 18.2) |
18.2 (14.8, 21.7) |
| Number of CAL Sites Measureda | 103.8 (100.5, 107.1) |
102.4 (93.3, 111.5) |
92.0 (86.5, 97.5) |
89.9 (81.2, 98.6) |
126.2 (122.7, 129.7) |
99.4 (92.7, 106.1) |
88.7 (78.0, 99.3) |
101.0 (82.1, 119.9) |
| Agea | 65.9 (65.6, 66.1) |
66.1 (65.2, 67.0) |
66.7 (66.2, 67.1) |
65.3 (64.6, 66.1) |
65.1 (64.7, 65.5) |
65.9 (65.3, 66.6) |
65.9 (65.0, 66.8) |
65.7 (63.9, 67.6) |
| Sexb | ||||||||
| Male | 43.7 (40.8, 46.6) |
45.6 (35.0, 56.2) |
47.9 (41.9, 53.8) |
43.4 (33.4, 53.3) |
42.4 (37.3, 47.6) |
41.4 (33.6, 49.2) |
30.2 (21.1, 39.3) |
48.9 (27.8, 70.0) |
| Female | 56.3 (53.4, 59.2) |
54.4 (43.8, 65.0) |
52.1 (46.2, 58.1) |
56.6 (46.7, 66.6) |
57.6 (52.4, 62.7) |
58.6 (50.8, 66.4) |
69.8 (60.7, 78.9) |
51.1 (30.0, 72.2) |
| Percent of Site-level CAL3+c | ||||||||
| Overall | 23.7 (22.1, 25.3) |
26.4 (22.1, 30.8) |
29.9 (26.4, 33.3) |
20.5 (15.5, 25.4) |
21.0 (18.2, 23.8) |
20.3 (17.1, 23.4) |
23.9 (18.2, 29.6) |
20.8 (13.6, 28.0) |
| Tooth Location | ||||||||
| Anterior | 20.4 (18.7, 22.1) |
20.8 (16.9, 24.7) |
27.1 (23.6, 30.6) |
16.0 (11.8, 20.2) |
18.0 (15.2, 20.8) |
15.3 (12.1, 18.5) |
20.9 (15.3, 26.5) |
19.2 (11.8, 26.6) |
| Posterior | 27.4 (25.6, 29.3) |
32.3 (26.4, 38.3) |
33.5 (29.4, 37.7) |
25.8 (18.7, 32.9) |
23.7 (20.7, 26.7) |
26.3 (22.7, 29.9) |
27.7 (20.8, 34.5) |
22.7 (14.2, 31.2) |
| Site Location | ||||||||
| Interproximal | 21.3 (19.7, 22.9) |
22.3 (18.1, 26.6) |
26.2 (22.8, 29.5) |
16.8 (12.3, 21.3) |
20.3 (17.5, 23.1) |
17.5 (14.4, 20.6) |
21.1 (15.7, 26.6) |
17.1 (9.6, 24.7) |
| Non-interproximal | 28.5 (26.6, 30.4) |
34.5 (29.3, 39.8) |
37.2 (32.9, 41.5) |
27.7 (21.2, 34.2) |
22.3 (19.3, 25.3) |
25.7 (22.1, 29.3) |
29.3 (22.7, 35.9) |
27.9 (19.4, 36.4) |
| Percent of Site-level PD4+c | ||||||||
| Overall | 7.3 (6.6, 8.1) |
9.8 (7.4, 12.2) |
7.2 (5.8, 8.7) |
6.8 (3.9, 9.8) |
8.0 (6.7, 9.3) |
5.7 (4.3, 7.0) |
7.7 (5.6, 9.8) |
6.6 (3.6, 9.6) |
| Tooth Location | ||||||||
| Anterior | 3.9 (3.4, 4.5) |
4.7 (2.9, 6.5) |
3.1 (2.2, 4.1) |
5.1 (2.0, 8.2) |
4.9 (3.9, 5.9) |
3.1 (2.0, 4.1) |
4.2 (2.7, 5.8) |
1.7 (0.3, 3.2) |
| Posterior | 11.1 (10.1, 12.2) |
15.2 (11.6, 18.7) |
12.6 (10.2, 15.0) |
8.9 (5.4, 12.5) |
10.8 (9.1, 12.5) |
8.8 (6.8, 10.8) |
12.0 (8.7, 15.3) |
12.4 (7.0, 17.8) |
| Site Location | ||||||||
| Interproximal | 9.9 (9.0, 10.9) |
13.3 (10.1, 16.6) |
10.3 (8.2, 12.4) |
8.2 (4.9, 11.4) |
10.8 (9.1, 12.5) |
7.2 (5.5, 8.9) |
10.6 (7.7, 13.5) |
9.9 (5.2, 14.6) |
| Non-interproximal | 2.3 (1.9, 2.6) |
2.8 (1.6, 4.0) |
1.3 (0.8, 1.7) |
4.3 (1.7, 6.8) |
2.6 (2.0, 3.2) |
2.6 (1.8, 3.3) |
1.9 (1.0, 2.8) |
0.3 (0.0, 0.8) |
Note: All values are weighted for study design and nonresponse. The mean number of PD sites measured was the same as the mean number of AL sites measured for all background groups, except for Mexicans which differed by 0.1.
Mean, (95% confidence interval).
Column percent, (95% confidence interval).
Percent out of all sites, (95% confidence interval).
Figure 1:
Site-Level Probabilities of CAL3+ and PD4+ in the HCHS/SOL Population Aged 60–74. Odd rows are maxillary teeth; even rows are mandibular teeth. Denoted by FDI upper tooth number/lower tooth number.
With respect to clinical characteristics by Hispanic/Latino background, Mexicans had the highest number of eligible teeth and sites measured for periodontal disease, while Cubans, Dominicans, and South Americans had lower numbers of eligible teeth and sites measured for periodontal disease. Cubans had the highest overall proportion of site-level CAL3+, while Central Americans had the highest overall proportion of site-level PD4+. For all Hispanic/Latino backgrounds, patterns in site-level proportions by tooth and site location were the same as in the overall population.
Model estimated ORs and 95% CIs for tooth, site location, and patient characteristics of CAL3+ and PD4+ are shown in Table 2. Mexicans had the lowest odds of CAL3+. Meanwhile, Central Americans, Cubans, South Americans, and participants with other or more than one background had higher odds of PD4+ than Mexicans, whereas Dominicans and Puerto Ricans had lower odds. Females had lower odds of CAL3+ and PD4+ than males. A larger number of teeth was associated with lower odds of site-level CAL3+, but higher odds of PD4+. For both CAL3+ and PD4+, sites located on anterior teeth had lower odds of disease than sites on posterior teeth [CAL3+: OR=0.59 (95% CI: 0.57, 0.62); PD4+: OR=0.39 (95% CI: 0.35, 0.42)].
Table 2:
Odds Ratios (95% Confidence Intervals) and Correlation Estimates (95% Confidence Intervals) for CAL3+ and PD4+ in the HCHS/SOL Population Aged 60–74
| AL3+ | PD4+ | |
|---|---|---|
| Mean Model | OR (95% CI) | OR (95% CI) |
| Tooth Location | ||
| Anterior | 0.59 (0.57, 0.62) | 0.39 (0.35, 0.42) |
| Posterior | reference | reference |
| Site Location | ||
| Interproximal | 0.67 (0.64, 0.70) | 4.70 (4.22, 5.24) |
| Non-interproximal | reference | reference |
| Log number of teeth | 0.44 (0.37, 0.51) | 1.76 (1.35, 2.29) |
| Sex | ||
| Female | 0.57 (0.49, 0.66) | 0.78 (0.64, 0.94) |
| Male | reference | reference |
| Hispanic/Latino Background | ||
| Central American | 2.29 (1.59, 3.30) | 1.66 (1.12, 2.46) |
| Cuban | 2.40 (1.93, 2.97) | 1.61 (1.28, 2.03) |
| Dominican | 1.64 (1.21, 2.22) | 0.19 (0.06, 0.60) |
| Mexican | reference | reference |
| Puerto Rican | 1.58 (1.28, 1.95) | 0.56 (0.39, 0.80) |
| South American | 2.23 (1.59, 3.14) | 1.50 (0.98, 2.30) |
| Other or More than One | 1.67 (1.13, 2.47) | 1.64 (0.93, 2.89) |
| Correlation Model | Estimate (95% CI) | Estimate (95% CI) |
| Adjacent, same tooth, and share interproximal space | 0.521 (0.493, 0.549) | 0.376 (0.338, 0.414) |
| Adjacent, same tooth, and do not share interproximal space | 0.562 (0.537, 0.588) | 0.306 (0.257, 0.354) |
| Non-adjacent, same tooth | 0.442 (0.418, 0.466) | 0.290 (0.253, 0.326) |
| Adjacent, adjacent teeth, same jaw, and share interproximal space | 0.450 (0.421, 0.478) | 0.345 (0.306, 0.385) |
| Non-adjacent, adjacent teeth on the same jaw | 0.396 (0.372, 0.421) | 0.250 (0.213, 0.287) |
| Vertically adjacent teeth and same site location | 0.322 (0.295, 0.350) | 0.295 (0.257, 0.332) |
| Vertically adjacent teeth and different site locations | 0.264 (0.237, 0.291) | 0.212 (0.178, 0.245) |
| Non-adjacent teeth | 0.281 (0.256, 0.306) | 0.166 (0.134, 0.198) |
Meanwhile, interproximal sites had lower odds of CAL3+ [OR=0.67 (95% CI: 0.64, 0.70)], but higher odds of PD4+ [OR=4.70 (95% CI: 4.22, 5.24)] than non-interproximal sites. These tooth location and site comparisons for males of Mexican background with 18 teeth are visualized in Figure 2, where maxillary teeth are depicted on the odd rows and mandibular teeth are depicted on the even rows and numbered according to the Federation Dentaire Internationale (FDI) numbering system on the x-axis. Rows are grouped by tooth site (DB, B, MB, ML, L, and DL). Figure 2 is colored by the magnitude of model estimated site-level probabilities of CAL3+ and PD4+, reflecting assumptions of symmetry built into the model. Plots for other sex-Hispanic/Latino background groups are shown in the supplementary material and have similar patterns with varying intensity due to the common main effects model structure.
Figure 2:
Model Estimated Site-Level Probabilities of CAL3+ and PD4+ in the HCHS/SOL Population Aged 60–74 for Males of Mexican Background with 18 Teeth. Odd rows are maxillary teeth; even rows are mandibular teeth. Denoted by FDI upper tooth number/lower tooth number.
Patterns in pairwise correlations estimated by the model (Table 2) were similar for CAL3+ and PD4+, but comparatively higher for CAL3+ than PD4+. Adjacent sites on the same tooth that do not share the same interproximal space were the most correlated for CAL3+. In contrast, adjacent sites on the same tooth that share the same interproximal space were the most correlated for PD4+. Pairs of sites that were in close proximity to each other, such as pairs of non-adjacent sites on the same tooth and pairs of adjacent sites on adjacent teeth on the same jaw, had the second highest correlations. Site pairs on non-adjacent teeth and site pairs on vertically adjacent teeth with different site locations had the lowest correlation. Heat maps of the estimated pairwise correlations for CAL3+ and PD4+ are shown in Figure 3.
Figure 3:
Model Estimated Pairwise Site Correlations of CAL3+ and PD4+ in the HCHS/SOL Population Aged 60–74. Sites are numbered 1–168 where for FDI tooth #17: 1=DB, 2=B, 3=MB, 4=ML, 5=L, 6=DL; for FDI tooth #16, 7=DB, 8=B, 9=MB, 10=ML, 11=L, 12=DL; etc.
Discussion
This study has detailed intraoral patterns and intensity of clinical measures of periodontal disease measured by CAL3+ and PD4+ within a large population-based sample of U.S. Hispanic and Latino older adults using statistical modeling for clustered data. The findings determined that (1) the probability of periodontal disease clinical indicators within the mouth presents with a quantifiable pattern based on site location (interproximal versus non-interproximal) and tooth location (anterior versus posterior), and (2) the spatial intraoral correlation pattern of disease across sites while complex, is well-described by an eight-parameter model based on different combinations of site-pairs. Novel heat maps from this study offer population-averaged interpretations of site-level probabilities of both CAL3+ and PD4+. Finally, there appears to be site-level heterogeneity of periodontal disease between sexes and Hispanic/Latino backgrounds.
The person-level findings of this study are consistent with previous studies of periodontitis which found that men had higher prevalences of CAL3+ and PD4+ than women.12–14 However, the site-level results are comparatively different than those for individual-level periodontitis prevalence patterns among Hispanic and Latino backgrounds. Eke et al14 reported Mexican Americans in the 2009–2014 National Health and Nutrition Examination Survey (NHANES) as having the highest prevalence of periodontitis, defined using the CDC/AAP case definition,5,6 among adults over age 65. Meanwhile, the estimates in this study indicate a lower risk of site-level CAL3+ among Mexicans than all other Hispanic/Latino backgrounds. Likewise, Central Americans and Cubans had a higher risk of site-level PD4+ than Mexicans. Granted this study examined an older population than NHANES, and HCHS/SOL is generalizable to the four HCHS/SOL communities, not the U.S. population like NHANES, site-level patterns by Hispanic/Latino background still differ from the participant-level patterns shown by Jiménez et al3 for similar age groups when prevalence is defined as one or more sites with CAL3+ or one or more sites with PD4+. Moreover, differences between Hispanic/Latino backgrounds identified in the present study may be influenced to some degree by tooth loss, which differed significantly among these groups. Tooth loss is likely to influence periodontitis measurement, as teeth with high CAL and PD measurements may end up being lost, and its relative importance tends to increase with age.15 Accordingly, this suggests that participant risk factors for site-level clinical measures of periodontal disease are distinct from those that affect periodontitis prevalence.
An argument can also be made that determinants of CAL3+ may differ from those of PD4+. CAL offers evidence of past periodontal disease experience, whereas PD may provide a better indication of current disease status,16 and both measures are subject to errors intrinsic to periodontal probing.17 Moreover, toothbrush abrasion may contribute to gingival recession, and therefore CAL.18 This study examined CAL3+ and PD4+ separately as outcomes of interest, and for comparison purposes the same mean and within-mouth correlation models were fitted to both. However, Figure 1 suggests that patterns of the observed effects of tooth-level and site-level characteristics may differ for CAL3+ and PD4+ and that site-level proportions of PD4+ are small in comparison to CAL3+. The latter point is often the case because CAL is the sum of PD and gingival recession. Furthermore, these findings for CAL3+ differ from Grbic and Lamster1 and Billings et al.2 Thus, different models could be applied for each clinical measure of periodontal disease and such models may be different for various race and ethnicity groups. A notable limitation is that, in consideration of the primary goal to compare Hispanic/Latino subgroups following Jiménez et al,3 this analysis did not account for lifestyle, behavioral, demographic and social factors of periodontal disease, nor did it consider non-Hispanic populations, which would need to be addressed in future research.
Computational issues in GEE estimation may preclude obtaining valid results in complex models. For example, PSUs from the HCHS/SOL study design were ignored during the model fitting for computational reasons related to a scenario where PSUs were defined as the clusters, although intraclass correlations could be computed for individual-level measures of periodontal disease. Additionally, when site-level proportions are small, such as those for PD4+, models with many covariates may yield convergence issues and violations of bounds for site-pair correlations that are unique to correlated binary data.7,19,20 Future research could consider the application of alternating logistic regressions21,22 for the analysis of multi-level periodontal data that model within-mouth association via pairwise odds ratios, which are not subject to pairwise bounds.
A number of studies have suggested the use of paired estimating equations approaches when there is scientific interest in the correlation structure of tooth- or site-level disease in the mouth,21–24 but this may be the first study to apply the Prentice estimation method7 for population-averaged marginal mean and pairwise correlation models in the context of periodontal disease. This study illustrates a marginal regression-based approach for the analysis of multi-level data that adjusts for confounding due to sex and the number of teeth, and identifies tooth-, tooth site-, and individual-level risk factors for cardinal periodontal disease measurements.
Supplementary Material
Acknowledgements
The authors thank the staff and participants of HCHS/SOL for their important contributions. A complete list of staff and investigators has been provided by Sorlie et al8 and is also available on the study website http://www.cscc.unc.edu/hchs/.
Funding
The Hispanic Community Health Study/Study of Latinos was carried out as a collaborative study supported by contracts from the National Heart, Lung, and Blood Institute (NHLBI) to the University of North Carolina (N01-HC65233), University of Miami (N01-HC65234), Albert Einstein College of Medicine (N01-HC65235), Northwestern University (N01-HC65236), and San Diego State University (N01-HC65237). The following Institutes/Centers/Offices contribute to the HCHS/SOL through a transfer of funds to the NHLBI: National Center on Minority Health and Health Disparities, the National Institute of Deafness and Other Communications Disorders, the National Institute of Dental and Craniofacial Research, the National Institute of Diabetes and Digestive and Kidney Diseases, the National Institute of Neurological Disorders and Stroke, and the Office of Dietary Supplements.
Footnotes
Disclosure
None of the authors had any financial or other conflicts of interest.
Contributor Information
Tracie L. Shing, Department of Biostatistics, Gillings School of Global Public Health University of North Carolina at Chapel Hill, Chapel Hill, NC 27599-7420
John S. Preisser, Department of Biostatistics, Gillings School of Global Public Health University of North Carolina at Chapel Hill, Chapel Hill, NC 27599-7420
Daniela Sotres-Alvarez, Department of Biostatistics, Gillings School of Global Public Health University of North Carolina at Chapel Hill, Chapel Hill, NC 27599-7420.
Kimon Divaris, Division of Pediatric and Public Health, Adams School of Dentistry University of North Carolina at Chapel Hill, Chapel Hill, NC 27599-7420.
James D. Beck, Division of Comprehensive Oral Health/Periodontology, Adams School of Dentistry University of North Carolina at Chapel Hill, Chapel Hill, NC 27599-7420
Data Availability Statement
The data that support the findings of this study are available on request from the corresponding author or through the study website http://www.cscc.unc.edu/hchs/. The data are not publicly available due to privacy or ethical restrictions.
References
- 1.Grbic JT, Lamster IB. Risk Indicators for Future Clinical Attachment Loss in Adult Periodontitis. Tooth and Site Variables. Journal of Periodontology. 1992;63(4):262–269. doi: 10.1902/jop.1992.63.4.262 [DOI] [PubMed] [Google Scholar]
- 2.Billings M, Parascandola M, Iafolla T, Dye BA. Data visualization of the relationship between smoking and periodontal site-specific effects across the lifespan in the U.S. adult population. J Periodontol. 2020;92(8):1126–1135. doi: 10.1002/JPER.20-0564 [DOI] [PubMed] [Google Scholar]
- 3.Jiménez MC, Sanders AE, Mauriello SM, Kaste LM, Beck JD. Prevalence of periodontitis according to Hispanic or Latino background among study participants of the Hispanic Community Health Study/Study of Latinos. The Journal of the American Dental Association. 2014;145(8):805–816. doi: 10.14219/jada.2014.31 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Sanders AE, Campbell SM, Mauriello SM, et al. Heterogeneity in periodontitis prevalence in the Hispanic Community Health Study/Study of Latinos. Annals of Epidemiology. 2014;24(6):455–462. doi: 10.1016/j.annepidem.2014.02.018 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Eke PI, Page RC, Wei L, Thornton-Evans G, Genco RJ. Update of the Case Definitions for Population-Based Surveillance of Periodontitis. Journal of Periodontology. 2012;83(12):1449–1454. doi: 10.1902/jop.2012.110664 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Page RC, Eke PI. Case Definitions for Use in Population-Based Surveillance of Periodontitis. Journal of Periodontology. 2007;78(7s):1387–1399. doi: 10.1902/jop.2007.060264 [DOI] [PubMed] [Google Scholar]
- 7.Prentice RL. Correlated Binary Regression with Covariates Specific to Each Binary Observation. Biometrics. 1988;44(4):1033–1048. doi: 10.2307/2531733 [DOI] [PubMed] [Google Scholar]
- 8.Sorlie PD, Avilés-Santa LM, Wassertheil-Smoller S, et al. Design and Implementation of the Hispanic Community Health Study/Study of Latinos. Annals of Epidemiology. 2010;20(8):629–641. doi: 10.1016/j.annepidem.2010.03.015 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.LaVange LM, Kalsbeek WD, Sorlie PD, et al. Sample Design and Cohort Selection in the Hispanic Community Health Study/Study of Latinos. Annals of Epidemiology. 2010;20(8):642–649. doi: 10.1016/j.annepidem.2010.05.006 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Papapanou PN, Sanz M, Buduneli N, et al. Periodontitis: Consensus report of workgroup 2 of the 2017 World Workshop on the Classification of Periodontal and Peri-Implant Diseases and Conditions: Classification and case definitions for periodontitis. J Clin Periodontol. 2018;45:S162–S170. doi: 10.1111/jcpe.12946 [DOI] [PubMed] [Google Scholar]
- 11.Shing TL, Preisser JS, Zink RC. GEECORR: A SAS macro for regression models of correlated binary responses and within-cluster correlation using generalized estimating equations. Computer Methods and Programs in Biomedicine. 2021;208:1–8. doi: 10.1016/j.cmpb.2021.106276 [DOI] [PubMed] [Google Scholar]
- 12.Dye B, Tan S, Smith V, Louis B, Barker L, Thorton-Evans G. Trends in Oral Health Status: United States, 1988–1994 and 1999–2004. Vol 248. U.S. Department of Health and Human Services, Centers for Disease Control and Prevention, National Center for Health Statistics; 2007. [Google Scholar]
- 13.Eke PI, Dye BA, Wei L, Thornton-Evans GO, Genco RJ. Prevalence of Periodontitis in Adults in the United States: 2009 and 2010. J Dent Res. 2012;91(10):914–920. doi: 10.1177/0022034512457373 [DOI] [PubMed] [Google Scholar]
- 14.Eke PI, Thornton-Evans GO, Wei L, Borgnakke WS, Dye BA, Genco RJ. Periodontitis in US Adults National Health and Nutrition Examination Survey 2009–2014. The Journal of the American Dental Association. 2018;149(7):576–588.e6. doi: 10.1016/j.adaj.2018.04.023 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Haworth S, Shungin D, Kwak SY, et al. Tooth loss is a complex measure of oral disease: Determinants and methodological considerations. Community Dent Oral Epidemiol. 2018;46(6):555–562. doi: 10.1111/cdoe.12391 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Kingman A, Albandar JM. Methodological aspects of epidemiological studies of periodontal diseases: Methodological aspects of epidemiological studies of periodontal diseases. Periodontology 2000. 2002;29(1):11–30. doi: 10.1034/j.1600-0757.2002.290102.x [DOI] [PubMed] [Google Scholar]
- 17.Leroy R, Eaton KA, Savage A. Methological issues in epidemiological studies of periodontitis - how can it be improved? BMC Oral Health. 2010;10(8):7. doi: 10.1186/1472-6831-10-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Litonjua LA, Andreana S, Bush PJ, Cook RE. Toothbrushing and gingival recession. Int Dent J. 2003;53(2):67–72. doi: 10.1111/j.1875-595x.2003.tb00661.x [DOI] [PubMed] [Google Scholar]
- 19.Qaqish BF. A family of multivariate binary distributions for simulating correlated binary variables with specified marginal means and correlations. Biometrika. 2003;90(2):455–463. doi: 10.1093/biomet/90.2.455 [DOI] [Google Scholar]
- 20.Preisser JS, Qaqish BF. A comparison of methods for simulating correlated binary variables with specified marginal means and correlations. Journal of Statistical Computation and Simulation. 2014;84(11):2441–2452. doi: 10.1080/00949655.2013.818148 [DOI] [Google Scholar]
- 21.Ananth CV, Kantor ML. Modeling Multivariate Binary Responses with Multiple Levels of Nesting Based on Alternating Logistic Regressions: an Application to Caries Aggregation. J Dent Res. 2004;83(10):776–781. doi: 10.1177/154405910408301008 [DOI] [PubMed] [Google Scholar]
- 22.Vanobbergen J, Lesaffre E, Garcia-Zattera M, Jara A, Martens L, Declerck D. Caries Patterns in Primary Dentition in 3-, 5- and 7-Year-Old Children: Spatial Correlation and Preventive Consequences. Caries Research. 2007;41:16–25. doi: 10.1159/000096101 [DOI] [PubMed] [Google Scholar]
- 23.DeRouen TA, Hujoel PP, Mancl LA. Statistical Issues in Periodontal Research. J Dent Res. 1995;74(11):1731–1737. doi: 10.1177/00220345950740110301 [DOI] [PubMed] [Google Scholar]
- 24.Gunsolley JC, Getchell C, Chinchilli VM. Small sample characteristics of generalized estimating equations. Communications in Statistics - Simulation and Computation. 1995;24(4):869–878. doi: 10.1080/03610919508813280 [DOI] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The data that support the findings of this study are available on request from the corresponding author or through the study website http://www.cscc.unc.edu/hchs/. The data are not publicly available due to privacy or ethical restrictions.