Suitability of machine learning models for prediction of clinically defined Stage III/IV periodontitis from questionnaires and demographic data in Danish cohorts

AIM

To evaluate if, and to what extent, machine learning models can capture clinically defined Stage III/IV periodontitis from self-report questionnaires and demographic data.

MATERIALS AND METHODS

Self-reported measures of periodontitis, demographic data and clinically established Stage III/IV periodontitis status were extracted from two Danish population-based cohorts (The Copenhagen Aging and Midlife Biobank [CAMB] and The Danish Health Examination Survey [DANHES]) and used to develop cross-validated machine learning models for the prediction of clinically established Stage III/IV periodontitis. Models were trained using 10-fold cross-validations repeated three times on the CAMB dataset (n = 1476), and the resulting models were validated in the DANHES dataset (n = 3585).

RESULTS

The prevalence of Stage III/IV periodontitis was 23.2% (n = 342) in the CAMB dataset and 9.3% (n = 335) in the DANHES dataset. For the prediction of clinically established Stage III/IV periodontitis in the CAMB cohort, models reached area under the receiver operating characteristics (AUROCs) of 0.67-0.69, sensitivities of 0.58-0.64 and specificities of 0.71-0.80. In the DANHES cohort, models derived from the CAMB cohort achieved AUROCs of 0.64-0.70, sensitivities of 0.44-0.63 and specificities of 0.75-0.84.

CONCLUSIONS

Applying cross-validated machine learning algorithms to demographic data and self-reported measures of periodontitis resulted in models with modest capabilities for the prediction of Stage III/IV periodontitis in two Danish cohorts.

Effectiveness of a Comprehensive School-Based Oral Health Intervention Program on Periodontal Disease Outcomes: A Randomized Controlled Trial

Background

Periodontitis is a prevalent oral health issue in rural India, contributing to tooth loss and systemic health complications. School-based oral health education programs offer a promising prevention and early intervention strategy.

Objective

To assess the impact of a comprehensive school-based oral health program on periodontal disease prevalence and severity among school-aged children in rural Bhopal.

Materials and Methods

A randomized controlled trial was conducted among 5000 children (10-15 years) in rural Bhopal. The intervention group received a 6-month program comprising oral hygiene instruction, dietary counseling, information on periodontitis, and facilitated dental checkups. The control group received no additional oral health education. Pre- and post-test assessments included clinical examinations (Community Periodontal Index, Gingival Index, Plaque Index) and oral health knowledge and practices questionnaire.

Results

After six months, the intervention group showed significant reductions in periodontal indices (CPI: -0.45, P < 0.001; GI: -0.30, P < 0.001) and plaque levels (PI: -0.65, P < 0.001) compared to the control group. They also demonstrated significant improvements in oral health knowledge (P < 0.001) and reported oral hygiene practices (P < 0.001).

Conclusion

A comprehensive school-based oral health education program is efficacious in improving periodontal health, knowledge, and practices among rural Indian children. The program's success supports its potential for broader implementation as a sustainable and cost-effective public health intervention in similar settings.

Utilisation of a think-aloud protocol to validate a self-reported periodontitis questionnaire

OBJECTIVES

The gold standard approach to diagnose periodontitis is based on clinical examination and radiographic investigations. This, however, is expensive, tedious, and not feasible in population-level assessments. The self-reported periodontitis questionnaire offers great benefit to facilitate larger epidemiological surveys. There is limited evidence on cognitive validation of self-reported periodontitis questionnaire. This study employed a think aloud approach to investigate if participants interpreted, comprehended, and understood the items in a self-reported periodontitis questionnaire, in the same way as researchers' intended.

METHODS

20 adults, resident of the UK and fluent in English participated in online recorded think aloud interviews. The self-reported periodontitis questionnaire consists of 15 questions and 2 open ended questions. The interviews were then transcribed and coded by two independent researchers according to predefined categories representing comprehension and conventional content analysis was used to analyse open-ended data.

RESULTS

The think aloud approach revealed that most of the questions in the self-reported periodontitis questionnaire were well understood by the participants. Two items, however, were identified as problematic: one was misinterpreted, and another question was not understood by most of the participants. Qualitative conventional content analysis of open-ended questions reiterated and complemented the findings of the think aloud study. Further questions coded as problematic for 3 or more participants were then considered for rephrasing.

CONCLUSIONS

This study revealed the interpretation and understanding of self-reported periodontitis questions by English speaking UK residents and highlighted the probable reason for lower sensitivity values of the self-reported periodontitis questionnaire.

CLINICAL SIGNIFICANCE

This study employed think-aloud approach to capture the thought process of the participants as they answered questions on self-reported periodontitis questionnaire. Overall, the questionnaire was well received by the participants, however, some questions were misunderstood/misinterpreted. This study highlights the potential information bias if participants do not understand the questions in epidemiological surveys.

Validity of individual self-report oral health measures in assessing periodontitis for causal research applications

BACKGROUND

Self-reported measures of periodontitis developed for use in population surveillance are increasingly used in causal research. Numerous studies evaluate the validity of these measures against clinical parameters of periodontitis, yet few include validation parameters outside of multivariable models. Individual item validity is necessary to adequately inform use of these measures in causal research.

METHODS

We used data from the National Health and Nutrition Examination Survey 2011-2014 in which dentate participants completed full-mouth periodontal examinations (N = 6966). We evaluated six self-report questionnaire items related to periodontal disease status against periodontitis case definitions developed by the Centers for Disease Control and Prevention and American Academy of Periodontology (CDC-AAP). We estimated the sensitivity and specificity of individual items using severe and moderate-to-severe periodontitis classifications. We additionally combined items to evaluate the validity of joint measures.

RESULTS

Sensitivity was highest when measures were evaluated against severe periodontitis. Self-rated oral health of fair/poor demonstrated the highest sensitivity for severe (0.60) and moderate-to-severe periodontitis (0.48). Specificity was highest when evaluated against moderate-to-severe periodontitis, with self-reported history of tooth mobility as the most specific measure (0.87 for severe disease; 0.92 for moderate-to-severe) followed by a history of bone loss (0.88 for severe; 0.91 for moderate-to-severe). Combining questions generally improved specificity at the expense of sensitivity.

CONCLUSIONS

Our findings related to item-specific validity and the associated clinical profiles facilitate needed considerations for the use of self-reported measures of periodontitis in causal research. Additionally, item-specific validity can be used to inform assessments of misclassification bias within such investigations.

Association of Gum Treatment with Cognitive Decline and Dementia Risk among Older Adults with Periodontal Symptoms: A 12-Year Prospective Cohort Study

INTRODUCTION

This study examines the associations of gum treatment with cognitive decline and dementia risk among older adults with periodontal symptoms in the USA.

METHODS

A cohort of 866 adults aged ≥50 with periodontal symptoms was recruited for the 2008 Health and Retirement Study "Dental Health Experimental Module" and followed until 2020. Cognitive function was assessed with the Telephone Interview for Cognitive Status (TICS). Dementia status was ascertained with the Langa-Weir algorithm based on TICS scores and proxy assessments. Linear mixed-effects model and multivariable Cox regression models were utilized to analyze the associations of gum treatment with cognitive decline and the risk of dementia, respectively.

RESULTS

Of 866 participants (mean age 67.7, 61.4% women), 105 (12.1%) developed dementia with a median follow-up of 9 (IQR, 6-10) years. The dementia incidence rates were lower in the group with gum treatment (7.4 vs. 12.9 per 1,000 person-years). Compared with participants who did not have gum treatment, those with gum treatment experienced a decline in TICS score that was on average 0.025 (95% CI, 0.005-0.044) points less per year and a 38% lower incidence of dementia (hazard ratio, 0.62; 95% CI, 0.41-0.93). These associations were consistent across participants with a different severity of periodontal symptoms and sociodemographic characteristics (age, sex, race, ethnicity, and education) except for income levels.

CONCLUSION

Prompt gum treatment for older adults with periodontal symptoms may be beneficial for their cognitive health.

Prevalence of self-reported severe periodontitis: Data from the population-based CONSTANCES cohort

AIM

To assess the prevalence of severe periodontitis based on the population-based CONSTANCES cohort using a validated self-reported questionnaire.

MATERIALS AND METHODS

Individuals were selected from the adult population in France using a random sampling scheme. Analyses were restricted to those invited in 2013-2014 who completed the periodontal health questionnaire at the 2017 follow-up. The risk of severe periodontitis was assessed using the periodontal screening score (PESS) and weighting coefficients were applied to provide representative results in the general French population.

RESULTS

The study included 19,859 participants (9204 men, mean age: 52.8 ± 12.6 years). Based on a PESS ≥ 5, 7106 participants were at risk of severe periodontitis, corresponding to a weighted prevalence of 31.6% (95% confidence interval: 30.6%-32.7%). This prevalence was higher among participants aged 55 and over, those with lower socio-economic status as well as current smokers, e-cigarette users and heavy drinkers. Among individuals at risk of severe periodontitis, only 18.8% (17.3%-20.4%) thought they had gum disease, although 50.5% (48.6%-52.5%) reported that their last dental visit was less than 6 months.

CONCLUSIONS

The present survey indicates that (1) self-reported severe periodontitis is highly prevalent with marked disparities between groups in the general French adult population, and (2) periodontitis could frequently be under-diagnosed given the low awareness.

Validity of self-reported male sexual function scales in a young Chinese population: a comparative study with clinician-assisted evaluation

Psychometric scales, commonly used to gauge sexual function, can sometimes be influenced by response biases. In our research from June 2020 to April 2021, we examined the accuracy of self-reported sexual function scales. We invited patients from the Department of Infertility and Sexual Medicine at the Third Affiliated Hospital of Sun Yat-sen University (Guangzhou, China), who have male sexual dysfunction, to participate by filling out a self-reported version of a specific questionnaire. In addition, they went through a clinician-assisted version of this questionnaire, encompassing tools such as the Premature Ejaculation Diagnostic Tool (PEDT), the 6-item International Index of Erectile Function (IIEF-6), the Erection Hardness Scale (EHS), and the Masturbation Erection Index (MEI). Using the clinician-assisted version as a reference, we categorized patients and applied various statistical methods, such as the Chi-square test, intraclass correlation coefficient (ICC), logistic regression, and the Bland-Altman plot, to gauge reliability. In our study with 322 participants, we found that while there were no notable discrepancies in error rates based on our categorization, certain scales showed significant differences in terms of overestimation and underestimation, with the exception of the PEDT. The positive diagnosis rate consistency between the self-reported and clinician-assisted versions was observed. High ICC values between the two versions across the scales were indicative of remarkable reliability. Our findings show that the self-reported versions of tools such as EHS, IIEF-6, MEI, and PEDT are credible and hold clinical reliability. However, employing a dual-diagnosis approach might be more prudent to circumvent potential misdiagnoses.

Type 1 Diabetes Mellitus Patients' Self-perception of Periodontal Diseases

OBJECTIVES

 The study aimed to evaluate type 1 diabetes mellitus (T1DM) patients' self-perceived periodontal health status and to identify the association between periodontal disease (PD) and DM.

MATERIALS AND METHODS

 This cross-sectional study included 113 T1DM children between 3 and 18 years old from the Universiti Teknologi MARA and the University of Malaya. Periodontal health parameters, including plaque index, gingival index, probing pocket depth, simplified basic periodontal examination, and clinical attachment loss, were recorded. Self-perceived periodontal health status was assessed with questionnaires.

STATISTICAL ANALYSIS

 Statistical analysis was performed to evaluate the sensitivity of the questionnaire and the relationship between T1DM and periodontal parameters.

RESULTS

 The median age was 11.4 years. Half of them (50.4%) were females. A total of 83.5% rated their oral condition as good, whereas 27.5% reported a history of gingival bleeding. Clinical examination revealed that 48.7% had healthy gingiva, whereas 47.8% had gingivitis. The question "Do you have bleeding when brushing, flossing, or eating food?" showed good accuracy in the evaluation of PD (p < 0.001).

CONCLUSION

 The questionnaire has a high potential to be used by medical professionals in identifying T1DM patients at risk of PD to guide nondental health care providers in making appropriate referrals to dental services.

Apical periodontitis and associated factors in a rural population of southern Brazil: a multilevel analysis

OBJECTIVES

This study aims to evaluate the association between apical periodontitis (AP) and sociodemographic/clinical factors in a probability sample of individuals living in a rural area of southern Brazilian.

MATERIALS AND METHODS

A cross-sectional study was conducted involving 584 non-edentulous adult individuals who had undergone a full-mouth radiographic survey. Periapical status was analysed using the periapical index (PAI). Endodontic status was evaluated considering the occurrence of voids in the filling material and the root filling length. Crown status was classified based on the presence of caries, restorations, and prosthetic crowns. Sociodemographic variables, frequency of dental care, and periodontal disease were also registered. The data were analysed using hierarchical multilevel Poisson regression analysis. The multilevel structure was composed of three models: sociodemographic variables, clinical variables, and clinical variables adjusted by sociodemographic variables (α = 5%).

RESULTS

The prevalence of AP in the sample was 60.45%. AP was significantly associated with age, skin colour, schooling, periodontal disease, and frequency of dental care (P < 0.005). Among the 10,396 teeth evaluated, 868 (8.35%) had AP, which was significantly associated with tooth group, dental arch, crown status, and endodontic treatment (P < 0.005).

CONCLUSIONS

The prevalence of AP was high in the population studied. An older age, black/brown skin colour, low level of schooling, infrequent dental care, severe periodontal disease, mandibular teeth, posterior teeth, inadequate crown status, and having undergone endodontic treatment were significantly associated with the outcome. Clinical relevance This study about a rural probability sample reinforces that AP is still a recurrent oral health problem.

Mendelian randomization highlights the causal association of obesity with periodontal diseases

AIM

The underlying mechanisms connecting obesity and periodontal diseases remain unclear. This study investigates the potential causal association of obesity with periodontal diseases using Mendelian randomization (MR).

MATERIALS AND METHODS

Single-nucleotide polymorphisms of obesity traits including body mass index (BMI), waist circumference (WC), and WC adjusted for BMI (WCadjBMI) from large-scale genome-wide association studies were screened for instrumental variables. The single trait periodontitis and the combined trait comprising periodontitis and loose teeth were adopted as surrogates for periodontal diseases. Inverse-variance weighted (IVW), series of sensitivity analyses and multivariable MR were employed to determine the association of obesity with periodontal diseases.

RESULTS

IVW results showed that per 1-SD increment in BMI (odds ratio, OR = 1.115; 95% confidence interval [CI] = 1.064-1.169; p < .001) and WC (OR = 1.117; 95% CI = 1.052-1.185; p < .001), but not WCadjBMI, were significantly associated with an increased risk of periodontitis/loose teeth. Moreover, the MR estimates were consistent across other MR sensitivity analyses and multivariable MR. However, a causal association of obesity with the single trait periodontitis was not identified.

CONCLUSIONS

The presented evidence supports previous epidemiological findings by showing a potential causal association of genetic liability to obesity with periodontal diseases. The biological mechanisms underlying this association warrant further investigation.

Who to Be Treated: Nomogram Using Self-Reported Periodontal Screening Instrument among English-Speaking Adults in Multi-Ethnic Singapore

Periodontal disease is a major public health problem. This study aimed to develop a nomogram using a self-reported periodontitis screening instrument in predicting severe periodontitis (SP), defined by the World Workshop on Classification of Periodontal and Peri-Implant Diseases and Conditions, and evaluate its utility in clinical setting. An Akaike information criterion selected multivariable model was developed to predict SP using a self-reported questionnaire, with a nomogram developed based on its regression coefficients. Discriminatory capability was evaluated by Receiver-operating characteristic curve. Ability to predict SP of individual patients was evaluated with bootstrapping. Decision curve analysis (DCA) was performed to evaluate its potential clinical utility by evaluating clinical net benefit at different thresholds. 58.1% of 155 participants were classified with SP. Older males without tertiary education, with ‘loose teeth’, ‘bone loss’ and ‘mouth rinse use’ had higher SP risk. The nomogram showed excellent discriminatory capability with Area under Curve of 0.83 (95% CI = (0.76, 0.89)), good calibration (intercept = 0.026) and slight overestimation of high risk and underestimation of low risk (slope = 0.834). DCA showed consistent clinical net benefit across the range of thresholds relative to assumption of ‘no patient’ or ‘all patient’ with SP. Our nomogram using a self-reported periodontitis instrument is useful in SP screening in English-speaking Singaporean adults.

Self-reported questionnaire on periodontal condition validated for use in Brazil

There is a current expectation of instruments for periodontal condition surveillance worldwide. The present study aimed to validate the Oral Health Questions Set B (OHQB) for the Brazilian Portuguese and evaluate its temporal stability. This is a sequential mixed-method investigation. After the forward-backward translation process to the Brazilian Portuguese, the OHQB Brazil (OHQB-Br) was applied to 156 participants (39.5 ± 14.14 years; 51.9% males). In sequence, through a full-mouth six-sites/teeth examination and in accordance with the original instrument, the periodontal diagnosis was obtained (March 2020). In January 2021, the OHB-BR was reapplied (n = 71). Ordinal alpha and McDonald's omega tested the internal consistency of the OHQB-Br. Temporal stability was investigated [Spearman correlation, intraclass correlation coefficient (ICC), and the Bland-Altman]. The concurrent validity was also verified, considering the periodontal clinical diagnosis (Kruskal Wallis). The ordinal alpha (0.69) and McDonald's omega (0.73) coefficients showed an adequate internal consistency of the OHQB-Br. The OHQB-Br temporal stability was high, as demonstrated by the Spearman coefficient (0.80) and ICC (0.79) and by the Bland-Altman plot. A concurrent validity showed a direct relationship between the OHQB-Br and the clinical condition of no periodontitis, mild, moderate, and severe periodontitis (p < 0.05). Because the OHQB-Br shows internal validity, temporal stability, and adequately identifies periodontal health and moderate/severe periodontitis, the instrument might represent an important tool, at the public level or other settings, for periodontal surveillance in Brazil.

Diagnostic accuracy of severe periodontitis case definitions: Comparison of the CDC/AAP, EFP/AAP, and CPI criteria

BACKGROUND

This cross-sectional study aimed to evaluate the diagnostic accuracy of severe periodontitis by three different case definitions.

METHODS

Patients aged >35 years (n = 243, mean age = 47.8 ± 9.3 years) were evaluated by periodontal examinations and classified in patients with severe periodontitis according to the case definition proposed by the Centers for Disease Control and Prevention (CDC) in collaboration with the American Academy of Periodontology (AAP)-CDC/AAP, the new periodontal disease classification proposed jointly by the AAP and the European Federation of Periodontology (EFP/AAP), and the Community Periodontal Index (CPI). The primary outcome was to compare the diagnostic accuracy of severe periodontitis between the three case definitions. Sensitivity (SS), specificity (SP), and area under the receiver operating characteristic (ROC) curve were used to analyze the accuracy.

RESULTS

The proportion of subjects presenting the most severe definitions of each classification was: severe periodontitis (CDC/AAP) = 47.3%, type III and IV periodontitis (EFP/AAP) = 43.2% and CPI code 4 = 44.4%. When comparing the CDC/AAP case definitions with the EFP/AAP classification, SS, SP, and ROC of 81.7%, 91.4%, and 0.866 (CI, 95%: 0.816 to 0.916), respectively, were obtained. Similarly, when comparing CDC/AAP to CPI code 4, SS, SP, and ROC of 81.7%, 89.1%, and 0.854 (CI, 95%: 0.802 to 0.906), respectively, were obtained. When type III and IV periodontitis (EFP/AAP) was compared with CPI code 4, the values were 92.4%, 92%, and 0.922 (CI, 95%: 0.883 to 0.961), respectively.

CONCLUSIONS

The EFP/AAP classification and the CPI demonstrated high diagnostic accuracy with the CDC/AAP case definition to diagnose severe periodontitis.

Diagnostic accuracy of self-reported measures of periodontal disease: A clinical validation study using the 2017 case definitions

AIM

To clinically validate a self-reported questionnaire for periodontal disease and assess its accuracy for differentiating periodontal health and different stages of periodontitis.

METHODS

A Chinese (Cantonese) version of a self-reported questionnaire was prepared by translating and validating the original English questions proposed by the Centers for Disease Control and Prevention/American Academy of Periodontology (CDC/AAP). The utility of the CDC/AAP questionnaire and its individual questions was assessed against a full-mouth periodontal examination. Periodontal case definition was based on the 2017 World Workshop classification of periodontal diseases. Multivariable logistic regression and the area under the receiver operating characteristic curve (AUROC) analysis were performed to assess the accuracy of the questionnaire.

RESULTS

408 subjects enrolled in this study, including those with periodontal health (16.2%), gingivitis (15.2%), Stages I/II periodontitis (31.8%) and Stages III/IV periodontitis (36.8%). Overall, the questionnaire had poor accuracy in detecting the presence of Stages I/II periodontitis with an AUROC 0.608. While it showed moderate to high accuracy for identifying periodontal disease (gingivitis and periodontitis), periodontitis and Stages III/IV periodontitis with an AUROC of 0.837, 0.803 and 0.870, respectively. Self-reported measures in combination with age and tobacco smoking showed excellent performance for identifying Stages III/IV periodontitis with a high AUROC of 0.953, a sensitivity of 95.7%, and a specificity of 89.0%. Specific questions and combinations provided greater utility to discriminate the various periodontal case definitions.

CONCLUSIONS

The self-reported CDC/AAP questionnaire may be a feasible tool for detecting periodontitis, and its combination with demographic and lifestyle factors is useful for the identification of individuals with Stages III/IV periodontitis. This questionnaire seems less helpful in screening of Stages I/II periodontitis. Further studies are needed to test the validity in larger community-based populations.

Alcohol use disorders are associated with higher prevalence of periodontitis in a rural area of Brazil

OBJECTIVE

The aim of the present cross-sectional study was to determine whether alcohol use disorders (AUD) are independently associated with severe and extent cases of periodontitis in individuals living in a rural area of Brazil.

METHODS

A representative probability sample (N = 585) was evaluated using six-site full-mouth periodontal examination. AUD was assessed using the alcohol use disorders identification test (AUDIT) and considered in its unidimensional and bidimensional factor structures. The first address hazardous alcohol consumption [HAC (overall scores ≥8)], and the second deals with two distinct constructs, harmful use [HU (items 1-3 scores ≥4 for men or ≥3 for women)], and alcohol-related consequences [ARC (items 4-10 scores ≥1)]. Periodontal outcomes were defined based on the criteria proposed by the CDC-AAP (2012) and an adaptation of the EFP-AAP (2018) definition of generalized stages III or IV periodontitis (GP). Adjusted multilevel multinomial and binary logistic regression analysis were used with a conceptual hierarchical approach to calculate the odds ratios (OR) of periodontal outcomes.

RESULTS

In the adjusted model, HU (OR = 2.69; 95% CI: 1.29-5.58) and ARC (OR = 3.79; 95% CI: 1.51-9.51) were significantly associated with higher prevalence of severe periodontitis (SP). Higher occurrence of GP was detected in individuals presenting HAC (OR = 1.88; 95% CI: 1.05-3.37) and ARC (OR = 2.90; 95% CI: 1.61-5.24).

CONCLUSION

AUD was independently associated with higher prevalence of SP and GP in individuals living in a rural area.

Development and validation of a latent, multidimensional, self-report periodontal disease measure

BACKGROUND

Self-report measures of periodontal disease have utility for screening, but have not capitalized on a latent variable approach based on psychometric theory to validate such measures. This study aimed to develop a psychometrically valid self-report measure of periodontal disease using latent variable factor analysis and other evidence-based psychometric analyses.

METHODS

Likert-type items reflecting periodontal disease were administered to a sample of adults (n = 828) in the United States via an online survey. Items were adapted from prior self-report measures or were newly developed based on psychometric item development theory and theoretical knowledge of periodontal disease. Psychometric analyses included exploratory and confirmatory factor analysis, parallel analysis, and a calculation of internal consistency.

RESULTS

Exploratory factor analysis (EFA) was indicative of the goodness-of-fit with two factors (root mean square error of approximation (RMSEA) = 0.08; comparative fit index (CFI) = 0.97; Tucker Lewis index (TLI) = 0.96; standardized root mean squared residual = 0.06); five of the 22 original survey questions were eliminated based on the results of this EFA. Parallel analysis supported a two-factor model to represent the similarities across items-one factor reflecting physiologic components and another reflecting functional components of periodontal disease. Confirmatory Factor Analysis also indicated adequate model fit (RMSEA = 0.07; CFI = 0.98; TLI = 0.98; and weighted root mean square residual = 1.20).

CONCLUSIONS

Psychometric analyses of a new 17-item periodontal disease self-report measure provided initial evidence of construct/factor validity. This approach to developing self-report periodontal disease measures may facilitate useful and cost-effective estimates of periodontal disease and provide a testable scale. Future work should include clinical validation.