Muscle weakness, greater fat mass, lower hematocrit levels, and advanced age in a diagnostic prediction model of periodontitis in adults with obesity: a cross-sectional study Authors: Karina Sarno Paes Alves Dias, Virgílio Santana-Júnior, Luciana Mara Barbosa Pereira, Felipe Oliveira Bittencourt, Hérika Maria Silveira Ruas, Gefter Thiago Batista Correa, Stênio Fernando Pimentel Duarte, Juciane Fagundes Durães Benitez, Renato Sobral Monteiro Junior, Sérgio Henrique Sousa Santos, Desirée Sant’Ana Haikal, Alfredo Maurício Batista de Paula Published: 2026 ABSTRACT BackgroundPeriodontitis is a chronic inflammatory disease that shares key biological pathways with obesity, particularly low-grade systemic inflammation and immune dysregulation. This study aimed to develop and internally validate a diagnostic prediction model for periodontitis in adults with obesity using routinely available clinical and laboratory parameters.MethodsThis cross-sectional study included 115 Brazilian adults with obesity (body mass index ≥ 30 kg/m2) receiving care in public ambulatory health services. Data collection included demographic characteristics, anthropometric and body-composition measures, handgrip strength (HGS), hematological and biochemical parameters, oral-health behaviors and status, and a comprehensive periodontal examination. Periodontitis was diagnosed using clinical and radiographic criteria consistent with the 2017 World Workshop framework. Candidate independent predictors were pre-specified based on biological plausibility and feasibility in public primary health care settings. A pre-specified forced-entry multivariable logistic regression model was developed with a maximum of four predictors and internally validated using 1,000 bootstrap resamples.ResultsPeriodontitis was diagnosed in 71.3% of participants (82/115). The final model retained age (adjusted OR = 1.13 per year; 95%CI = 1.08–1.20), fat mass (adjusted OR = 1.04 per kg; 95%CI = 0.99–1.09), hematocrit (adjusted OR = 0.89 per percentage point; 95%CI = 0.76–1.02), and reduced HGS (adjusted OR = 5.22; 95%CI = 1.80–17.30) as independent predictors. Apparent discrimination was excellent (AUC = 0.873; 95%CI = 0.807–0.938), with an optimism-corrected AUC of 0.865. Calibration was acceptable (Hosmer–Lemeshow p = 0.749), and overall accuracy was 80.0%, with sensitivity of 76.8% and specificity of 87.9%. A LASSO sensitivity analysis confirmed predictor robustness. A simplified exploratory screening score and a provisional nomogram were derived to improve interpretability.ConclusionsA parsimonious diagnostic prediction model identified age, fat mass, hematocrit, and HGS as independent predictors of prevalent periodontitis in adults with obesity. The model should be regarded as exploratory and hypothesis-generating. It may inform future risk-stratification strategies in public primary health care settings, but external validation and recalibration are required before clinical implementation.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12903-026-08601-2. Introduction Periodontitis is a multifactorial, chronic, biofilm-associated inflammatory disease characterized by dysbiosis of the subgingival microbiota in susceptible individuals, progressive destruction of the tooth-supporting tissues, and, in advanced stages, tooth loss [1, 2]. Its pathogenesis involves interactions between periodontal pathogens and the host response, modulated by genetic and behavioral factors such as smoking and inadequate oral hygiene [3, 4]. Clinically, periodontitis is characterized by increased probing depth, bleeding on probing, and clinical attachment loss [2]. Beyond its local effects, periodontitis has systemic relevance because microbial products and inflammatory mediators may disseminate into the circulation and contribute to low-grade inflammation [5]. Observational studies and systematic reviews have reported associations between periodontitis and multiple comorbidities, particularly cardiometabolic disorders [6–8]. Obesity is a highly prevalent chronic disease associated with important metabolic, immunological, and inflammatory disturbances [9]. In addition to increasing the risk of multiple chronic diseases and being associated with higher all-cause mortality, obesity imposes a substantial economic burden on healthcare systems [10, 11]. Excess adiposity promotes chronic low-grade systemic inflammation through adipokines, cytokines, and immune-cell activation, thereby shaping susceptibility to several chronic non-communicable diseases [12]. A growing body of evidence suggests that obesity and periodontitis are biologically linked [13, 14]. Despite this background, clinically pragmatic diagnostic prediction tools capable of identifying adults with obesity who are more likely to have periodontitis remain scarce. Prediction models based on routinely obtainable variables may be useful to prioritize periodontal assessment in resource-constrained health care settings. The present study therefore aimed to develop and internally validate a parsimonious diagnostic prediction model for prevalent periodontitis among adults with obesity using demographic, body-composition, oral health status and health behaviors, functional, and laboratory measures that are potentially available in public primary health care settings. Materials and methods Study design and setting This single-center observational cross-sectional study was conducted between April 2022 and December 2024 in public ambulatory health care services with oral-health teams in Vitória da Conquista, Bahia, Brazil. The study was designed and reported in accordance with the TRIPOD statement for the development of a diagnostic prediction model [15] (Table 1S. Material Supplementary). All procedures were approved by the relevant ethics committees under CAAE No. 50251721.0.0000.5578 (Process No. 4.931.111), and all participants provided written informed consent. Participants Eligible participants were adults aged 18 years or older, of either sex, with obesity defined as BMI ≥ 30 kg/m2, who were physically independent and had preserved cognitive and functional abilities at enrolment. Recruitment was prospective and consecutive, using a pragmatic non-probability strategy that included all eligible adults who attended the participating services during the study period and agreed to participate. Exclusion criteria were acute local or systemic inflammatory conditions at assessment, untreated comorbidities likely to impair physical performance, medications with major effects on body composition, relevant orthopedic or locomotor limitations, recent oncological treatment, or diagnosed neurological disease. Outcome definition The primary outcome was prevalent periodontitis, diagnosed using clinical and radiographic criteria aligned with the 2017 World Workshop case definition [16]. A case was defined as interdental clinical attachment loss at two or more non-adjacent teeth, or buccal/oral clinical attachment loss of at least 3 mm with probing depth greater than 3 mm at two or more teeth. Radiographic evidence of alveolar bone loss was used to support classification when appropriate. Examiners were trained and calibrated before fieldwork; inter-examiner agreement for key periodontal measures was high (kappa = 0.87). Variables used to define the outcome were not considered candidate predictors. Predictors and data collection Distinct trained teams collected demographic, anthropometric/body-composition, functional, periodontal diagnosis, oral health status and behaviors, and laboratory data to reduce measurement bias. Demographic and behavioral variables included age and sex, and tobacco and alcohol consumption. Anthropometric measurements were obtained using standardized and calibrated instruments [17]. Body composition was assessed using multi-frequency bioelectrical impedance analysis (BIA) (InBody120, ROK), estimating absolute fat mass—FM (kg), body fat percentage (%), and skeletal muscle mass—SMM (kg) [18]. Height was measured with a fixed stadiometer (Cescorf, BRA), and body weight with a calibrated electronic scale (HBF222‑T, Omron®, JPN). Obesity was defined as BMI ≥ 30 kg/m2 according to World Health Organization (WHO) criteria. Handgrip strength (HGS) was measured with a hydraulic dynamometer, and reduced HGS was operationalized using a sample-based threshold (values below the 50th percentile), reflecting relative functional weakness within the adult study population rather than a clinical diagnosis of sarcopenia [19]. Hematological and biochemical variables included red blood cell count, hemoglobin, hematocrit, total leukocytes, fasting glucose, total cholesterol, triglycerides, albumin, and high-sensitivity C-reactive protein (CRP). hs‑CRP values ≥ 10 mg/L were interpreted as suggestive of acute inflammation and analyzed separately [20–22]. Oral-health behaviors included toothbrushing frequency and dental-floss use. The oral examination was performed by previously trained and calibrated dentists, following standard clinical procedures [23]. The number of remaining, decayed, and missing natural teeth was recorded to characterize participants’ dental status. Periodontal status was evaluated using comprehensive clinical and radiographic assessments. All remaining natural teeth were examined at six sites per tooth, with standardized recording of key periodontal parameters, including probing depth, clinical attachment level, bleeding on probing, visible biofilm, gingival recession, and tooth mobility. Periapical radiographs were obtained and independently interpreted by calibrated examiners blinded to predictor data, to support diagnostic classification and the evaluation of alveolar bone loss. Statistical analysis and model development Continuous variables were summarized as mean ± standard deviation or median and interquartile range, as appropriate, and categorical variables as absolute and relative frequencies. Candidate predictors were pre-specified according to biological plausibility, prior evidence, and routine feasibility in primary care rather than selected solely based on univariable statistical significance. Missingness was restricted exclusively to biological variables; anthropometric, body-composition, and demographic variables had no missing data (0%). Among laboratory parameters, most hematological and biochemical variables had one missing observation each (0.9%; n = 1/115), attributable to hemolyzed or insufficient samples. High-sensitivity C-reactive protein had two missing values (1.7%; n = 2/115) and fasting glucose had four missing values (3.5%; n = 4/115); the latter was not included in the final model. Little's MCAR test applied to biological variables indicated data were not missing completely at random (χ2 = 100.78, df = 73, p = 0.017); accordingly, a missing-at-random assumption was adopted, and expectation–maximization imputation was applied to affected laboratory variables prior to modelling. A multivariable logistic regression model was developed using a pre-specified forced-entry approach, with all four candidate predictors entered simultaneously into the model regardless of individual statistical significance. This strategy was adopted in alignment with current recommendations for diagnostic prediction modelling, which prioritize clinical and biological pre-specification over data-driven variable selection. Model complexity was constrained a priori to a maximum of four predictors based on events-per-variable considerations. Multicollinearity was assessed using tolerance and variance inflation factors (all VIF < 1.4). Model performance was evaluated by discrimination (AUC-ROC with 95% CI via DeLong method), calibration (Hosmer–Lemeshow goodness-of-fit test and calibration plot), Nagelkerke R2, and classification metrics at the optimal probability threshold identified by the Youden Index. Internal validation used 1,000 bootstrap resamples to estimate optimism and assess coefficient stability. As a pre-planned sensitivity analysis to assess the robustness of the pre-specified predictor set, a LASSO-penalized logistic regression model was fitted using tenfold cross-validation optimizing discriminative performance, with the regularization parameter selected at the minimum cross-validated loss (λ = 0.0049). All statistical analyses were performed in R version 4.5.1 [24–27]. Results The participant flowchart, including eligibility assessment, refusals, exclusions, and the final analytical sample, are presented in Fig. 1.Fig. 1Flowchart of the selection process for participants Flowchart of the selection process for participants A total of 115 adults with obesity were included. Mean age was 41.1 ± 12.4 years, 63.5% were women, and periodontitis was diagnosed in 82 participants (71.3%). The sample had a mean BMI of 35.7 ± 5.5 kg/m2, mean fat mass of 42.0 ± 10.6 kg, mean handgrip strength of 31.8 ± 10.5 kgf, and mean hematocrit of 43.1 ± 3.6%. Reduced handgrip strength was present in 48.7% of participants. Table 1 presents in detail the anthropometric, laboratory, and oral health characteristics of the sample.Table 1Baseline characteristics of adults with obesity included in the study (n = 115)Variablesn (%)Mean ± SDP25P50P75Age (years)-41.1 ± 12.431.841.650.0Women73 (63.5%)----Weight (kg)-98.0 ± 17.885.092.4109.5Height (m)-1.6 ± 0.91.61.61.7BMI (kg/m2)-35.7 ± 5.531.834.138.0FM (kg)-42.0 ± 10.632.743.049.2SMM (kg)-32.7 ± 13.022.928.636.2HGS (kgf)-31.8 ± 10.525.630.436.2Reduced HGS56 (48.7)----Hematocrit (%)-43.1 ± 3.640.043.046.0Triglycerides (mg/dL)-225.1 ± 129.0119.0192.0297.0Albumin (g/dL)-4.4 ± 0.34.14.44.7Inflammatory biomarker hs-CRP (mg/dL)-0.8 ± 1.40.20.30.7 Acute inflammation (hs-CRP ≥ 10 mg/L)26 (22.6)---- Regular tooth brushing frequenc107 (93.0)---- Regular dental floss use82 (71.3)---- Remaining natural teeth-24.8 ± 5.222.026.028.0 Periodontitis82 (71.3)----Data are presented as n (%) for categorical variables and mean ± standard deviation (SD) for continuous variables. P25, P50, P75: 25th, 50th, and 75th percentilesBMI body mass index, BIA bioelectrical impedance analysis, FM fat mass, SMM skeletal muscle mass, HGS handgrip strength, hs-CRP high-sensitivity C-reactive protein Table 1Baseline characteristics of adults with obesity included in the study (n = 115)Variablesn (%)Mean ± SDP25P50P75Age (years)-41.1 ± 12.431.841.650.0Women73 (63.5%)----Weight (kg)-98.0 ± 17.885.092.4109.5Height (m)-1.6 ± 0.91.61.61.7BMI (kg/m2)-35.7 ± 5.531.834.138.0FM (kg)-42.0 ± 10.632.743.049.2SMM (kg)-32.7 ± 13.022.928.636.2HGS (kgf)-31.8 ± 10.525.630.436.2Reduced HGS56 (48.7)----Hematocrit (%)-43.1 ± 3.640.043.046.0Triglycerides (mg/dL)-225.1 ± 129.0119.0192.0297.0Albumin (g/dL)-4.4 ± 0.34.14.44.7Inflammatory biomarker hs-CRP (mg/dL)-0.8 ± 1.40.20.30.7 Acute inflammation (hs-CRP ≥ 10 mg/L)26 (22.6)---- Regular tooth brushing frequenc107 (93.0)---- Regular dental floss use82 (71.3)---- Remaining natural teeth-24.8 ± 5.222.026.028.0 Periodontitis82 (71.3)----Data are presented as n (%) for categorical variables and mean ± standard deviation (SD) for continuous variables. P25, P50, P75: 25th, 50th, and 75th percentilesBMI body mass index, BIA bioelectrical impedance analysis, FM fat mass, SMM skeletal muscle mass, HGS handgrip strength, hs-CRP high-sensitivity C-reactive protein Baseline characteristics of adults with obesity included in the study (n = 115) Data are presented as n (%) for categorical variables and mean ± standard deviation (SD) for continuous variables. P25, P50, P75: 25th, 50th, and 75th percentiles BMI body mass index, BIA bioelectrical impedance analysis, FM fat mass, SMM skeletal muscle mass, HGS handgrip strength, hs-CRP high-sensitivity C-reactive protein In bivariate analyses restricted to candidate predictors, prevalent periodontitis was more frequent among participants who consumed alcohol, had reduced handgrip strength, and reported irregular flossing. Participants with periodontitis were older and showed lower mean hematocrit, higher fat mass, higher triglycerides, and lower albumin levels, with small to moderate effect sizes (all p > 0.05). Mean HGS was lower in the periodontitis group with a moderate effect size. BMI and total body weight did not differ significantly between groups (p > 0.05) (Table 2).Table 2Comparison of categorical variables between diagnosis of periodontitis and obesityVariables (n / %)Periodontitis diagnosisp valueEffect size (φ/V)No (n = 33)Yes (n = 82)Alcohol consumption11 (50.0)11 (50.0)0.0140.229Reduced HGS10 (17.9)46 (82.1)0.0120.233Irregular dental-floss use17 (20.7)17 (51.5)0.0030.278Variables (mean +/- SD)Periodontitis diagnosisp valueEffect size (φ/V)No (n = 33)Yes (n = 82)Age (years)31.5 +/- 9.844.9 +/- 11.2<0.001-1.234FM (kg)39.0 +/- 9.843.2 +/- 10.70.058-0.392Hematocrit (%)44.2 +/- 4.242.6 +/- 3.20.0320.445Triglycerides (mg/dL)202.1 +/- 150.8234.3 +/- 118.90.035-0.252Albumin (g/dL)4.6 +/- 0.34.3 +/- 0.30.0110.306Only candidate predictors relevant to model development are shown. Variables constituting the clinical definition of periodontitis were intentionally excluded from comparative modelling tables. Data are presented as n (%) or mean +/- standard deviation (SD)FM fat mass, HGS handgrip strength Table 2Comparison of categorical variables between diagnosis of periodontitis and obesityVariables (n / %)Periodontitis diagnosisp valueEffect size (φ/V)No (n = 33)Yes (n = 82)Alcohol consumption11 (50.0)11 (50.0)0.0140.229Reduced HGS10 (17.9)46 (82.1)0.0120.233Irregular dental-floss use17 (20.7)17 (51.5)0.0030.278Variables (mean +/- SD)Periodontitis diagnosisp valueEffect size (φ/V)No (n = 33)Yes (n = 82)Age (years)31.5 +/- 9.844.9 +/- 11.2<0.001-1.234FM (kg)39.0 +/- 9.843.2 +/- 10.70.058-0.392Hematocrit (%)44.2 +/- 4.242.6 +/- 3.20.0320.445Triglycerides (mg/dL)202.1 +/- 150.8234.3 +/- 118.90.035-0.252Albumin (g/dL)4.6 +/- 0.34.3 +/- 0.30.0110.306Only candidate predictors relevant to model development are shown. Variables constituting the clinical definition of periodontitis were intentionally excluded from comparative modelling tables. Data are presented as n (%) or mean +/- standard deviation (SD)FM fat mass, HGS handgrip strength Comparison of categorical variables between diagnosis of periodontitis and obesity Only candidate predictors relevant to model development are shown. Variables constituting the clinical definition of periodontitis were intentionally excluded from comparative modelling tables. Data are presented as n (%) or mean +/- standard deviation (SD) FM fat mass, HGS handgrip strength Table 3 shows the final multivariable model retained age, fat mass, hematocrit, and reduced handgrip strength as pre-specified predictors. Age (adjusted OR 1.13 per year; 95% CI 1.08–1.20; p > 0.001) and reduced handgrip strength (adjusted OR 5.22; 95% CI 1.80–17.30; p = 0.004) reached conventional statistical significance. Fat mass (adjusted OR 1.04; 95% CI 0.99–1.09; p = 0.157) and hematocrit (adjusted OR 0.89; 95% CI 0.76–1.02; p = 0.110) did not reach significance at the 0.05 threshold when modelled simultaneously with the dominant predictors, but were retained based on pre-specified clinical rationale consistent with TRIPOD recommendations. The model yielded a Nagelkerke R2 of 0.474, Hosmer–Lemeshow p = 0.749, and an AUC of 0.873 (95% CI 0.807–0.938), indicating excellent apparent discrimination. At the optimal probability threshold of 0.78 (Youden Index), overall accuracy was 80.0% (95% CI 71.8–86.3%), sensitivity 76.8% (95% CI 66.6–84.6%), specificity 87.9% (95% CI 72.7–95.2%), positive predictive value 94.0% (95% CI 85.6–97.7%), and negative predictive value 60.4% (95% CI 46.3–73.0%).Table 3Final multivariable diagnostic prediction model for prevalent periodontitis in adults with obesityPredictorAdjusted OR95% CIp valueAge (per year)1.131.08–1.20 < 0.001Fat Mass (per kg)1.040.99–1.090.157Hematocrit (per %)0.890.76–1.020.110Reduced handgrip strength5.221.80–17.300.004Model performance: AUC 0.873 (95% CI 0.807–0.938); optimism-corrected AUC 0.865 (95% CI 0.815–0.942); Hosmer–Lemeshow p = 0.749; Nagelkerke R2 = 0.474 Table 3Final multivariable diagnostic prediction model for prevalent periodontitis in adults with obesityPredictorAdjusted OR95% CIp valueAge (per year)1.131.08–1.20 < 0.001Fat Mass (per kg)1.040.99–1.090.157Hematocrit (per %)0.890.76–1.020.110Reduced handgrip strength5.221.80–17.300.004Model performance: AUC 0.873 (95% CI 0.807–0.938); optimism-corrected AUC 0.865 (95% CI 0.815–0.942); Hosmer–Lemeshow p = 0.749; Nagelkerke R2 = 0.474 Final multivariable diagnostic prediction model for prevalent periodontitis in adults with obesity Model performance: AUC 0.873 (95% CI 0.807–0.938); optimism-corrected AUC 0.865 (95% CI 0.815–0.942); Hosmer–Lemeshow p = 0.749; Nagelkerke R2 = 0.474 Internal validation using 1,000 bootstrap resamples demonstrated high model stability and minimal overfitting, revealing an optimism of only 0.008. The optimism-corrected AUC remained excellent at 0.865 (95% CI: 0.815–0.942). As a pre-planned sensitivity analysis, a LASSO-penalized logistic regression model was fitted with tenfold cross-validation (λ = 0.0049). All four pre-specified predictors were retained without zero-shrinkage, with coefficients directionally consistent with the forced-entry model: age (β = 0.112), fat mass (β = 0.030), hematocrit (β = − 0.103), and reduced HGS (β = 1.439). The LASSO model achieved an AUC of 0.875, essentially identical to the forced-entry model (AUC = 0.873), confirming that the pre-specified predictor set is robust and not an artefact of the modelling procedure. The final prediction equation is: logit(P) = − 0.614 + 0.121 × age (years) + 0.035 × fat mass (kg) − 0.121 × hematocrit (%) + 1.652 × reduced handgrip strength, where reduced handgrip strength is coded 1 (reduced) or 0 (preserved). The optimal cutoff point (red dot) was identified using the Youden Index (0.78), corresponding to the best balance between sensitivity and specificity. The dashed diagonal line represents the line of no discrimination (Fig. 2).Fig. 2Discriminative performance and cutoff point determination of the predictive model for periodontitis in adults with obesity. a Plots illustrate the trade-off between sensitivity (blue line) and specificity (purple line) across probability thresholds. For clinical application prioritizing early detection, a final cut-off of Youden's Index = 0.780 was selected, balancing sensitivity (76.8%) with specificity (87.9%) at the optimal threshold. b The model achieved an AUC of 0.873 (95% CI 0.807–0.938), indicating good‑to‑excellent discrimination capacity for the diagnosis of periodontitis in adults with obesity. The diagonal line represents no‑discrimination. Abbreviations: AUC: area under the curve, ROC: Receiver operating characteristic curve Discriminative performance and cutoff point determination of the predictive model for periodontitis in adults with obesity. a Plots illustrate the trade-off between sensitivity (blue line) and specificity (purple line) across probability thresholds. For clinical application prioritizing early detection, a final cut-off of Youden's Index = 0.780 was selected, balancing sensitivity (76.8%) with specificity (87.9%) at the optimal threshold. b The model achieved an AUC of 0.873 (95% CI 0.807–0.938), indicating good‑to‑excellent discrimination capacity for the diagnosis of periodontitis in adults with obesity. The diagonal line represents no‑discrimination. Abbreviations: AUC: area under the curve, ROC: Receiver operating characteristic curve The calibration plot depicts the agreement between predicted probabilities and observed event rates across deciles of predicted risk. The dashed diagonal line represents perfect calibration, while the solid line reflects the model's observed performance. Overall calibration was acceptable across most of the probability range; however, some deviation from the ideal calibration line was observed at intermediate predicted probabilities (approximately 0.30–0.55), which may reflect limited sample size in that decile range and reinforces the need for recalibration in external datasets. (Fig. 3).Fig. 3Calibration curve of the diagnostic prediction model for periodontitis in adults with obesity.The blue points and solid line represent the observed proportions of periodontitis across deciles of predicted risk, and the red dashed line represents perfect calibration (ideal agreement between predicted and observed probabilities) Calibration curve of the diagnostic prediction model for periodontitis in adults with obesity.The blue points and solid line represent the observed proportions of periodontitis across deciles of predicted risk, and the red dashed line represents perfect calibration (ideal agreement between predicted and observed probabilities) To support clinical interpretability, a simplified exploratory screening score was derived from the regression coefficients of the final pre-specified forced-entry model and is presented in Table 4. Point values for each predictor category were calculated proportionally from the incremental logit contribution of each category relative to the lowest-risk reference combination, using a common scaling factor of 0.35 logit units, and rounded to the nearest integer.Table 4Exploratory simplified screening scores derived from the final predictive diagnostic model of periodontitis in adults with obesityPredictorCategoryPointsAge (years) < 35035–44445–548 ≥ 5512Fat Mass (kg) < 35035–44.9145–54.92 ≥ 553Hematocrit (%) ≥ 45042–44.9139–41.92 < 393Reduced handgrip strengthNo0Yes5Point values were derived proportionally from the regression coefficients of the final model estimated by pre-specified forced entry in R, using the midpoint of each category relative to the lowest-risk reference. The threshold of ≥ 10 points corresponds to an approximate predicted probability of ≥ 0.80, consistent with the optimal diagnostic cutoff identified by the Youden Index. Suggested exploratory interpretation: 0–5 points = lower screening priority; 6–9 points = intermediate priority; ≥ 10 points = higher priority. These categories are exploratory and do not constitute validated clinical thresholds. Maximum possible score: 23 points Table 4Exploratory simplified screening scores derived from the final predictive diagnostic model of periodontitis in adults with obesityPredictorCategoryPointsAge (years) < 35035–44445–548 ≥ 5512Fat Mass (kg) < 35035–44.9145–54.92 ≥ 553Hematocrit (%) ≥ 45042–44.9139–41.92 < 393Reduced handgrip strengthNo0Yes5Point values were derived proportionally from the regression coefficients of the final model estimated by pre-specified forced entry in R, using the midpoint of each category relative to the lowest-risk reference. The threshold of ≥ 10 points corresponds to an approximate predicted probability of ≥ 0.80, consistent with the optimal diagnostic cutoff identified by the Youden Index. Suggested exploratory interpretation: 0–5 points = lower screening priority; 6–9 points = intermediate priority; ≥ 10 points = higher priority. These categories are exploratory and do not constitute validated clinical thresholds. Maximum possible score: 23 points Exploratory simplified screening scores derived from the final predictive diagnostic model of periodontitis in adults with obesity Point values were derived proportionally from the regression coefficients of the final model estimated by pre-specified forced entry in R, using the midpoint of each category relative to the lowest-risk reference. The threshold of ≥ 10 points corresponds to an approximate predicted probability of ≥ 0.80, consistent with the optimal diagnostic cutoff identified by the Youden Index. Suggested exploratory interpretation: 0–5 points = lower screening priority; 6–9 points = intermediate priority; ≥ 10 points = higher priority. These categories are exploratory and do not constitute validated clinical thresholds. Maximum possible score: 23 points To facilitate individualized risk estimation, a nomogram was constructed based on the final multivariable logistic regression model (Fig. 4). The nomogram incorporates age, FM, hematocrit, and reduced HGS, assigning a weighted point value to each predictor according to its relative contribution to the model. The sum of these values yields a total score, which corresponds to the predicted probability of prevalent periodontitis. The graphical representation demonstrates coherent scaling across predictors, consistent with their relative regression coefficients: age and reduced handgrip strength dominate the nomogram score, reflecting their higher absolute contribution to the predicted log-odds, followed by hematocrit and fat mass. The nomogram uses a continuous internal point scale derived directly from the regression model coefficients, which differs from the simplified categorical scoring in Table 4; both tools are intended as complementary interpretability aids and should not be used interchangeably.Fig. 4Nomogram derived from the final diagnostic prediction model of periodontitis in adults with obesity Nomogram derived from the final diagnostic prediction model of periodontitis in adults with obesity How to use this nomogram: (1) For each predictor, locate the patient's value on its corresponding axis. (2) Draw a perpendicular vertical line upward from each predictor value to the "Points" axis and record the individual point value. (3) Sum the point values obtained for all four predictors (Age, Fat mass, Hematocrit, and Reduced handgrip strength) to obtain the "Total Points." Note that the Hematocrit axis is displayed in descending order (left to right), reflecting its inverse association with periodontitis probability. (4) Locate the Total Points value on the "Total Points" axis and draw a perpendicular vertical line downward to the "Probability of Periodontitis" axis to read the estimated individual probability of prevalent periodontitis. Predicted probabilities ≥ 0.78, corresponding to the optimal threshold identified by the Youden Index, indicate higher screening priority for periodontal evaluation. This nomogram is an exploratory implementation aid derived from the development ample and has not been externally validated; it should not be used as a standalone clinical decision tool. Discussion In this cross-sectional study of adults with obesity receiving care in public ambulatory services, prevalent periodontitis was common and a parsimonious diagnostic prediction model identified age, FM, hematocrit, and reduced HGS as independent predictors of disease probability. Taken together, the findings reinforce the view that periodontitis in obesity may reflect multisystem dysregulation rather than an isolated oral condition. Collectively, these results support the concept that periodontal health is closely intertwined with systemic health and reinforce the need for integrated approaches to the management of obesity and related systemic chronic diseases [28–30]. The association between older age and periodontitis is consistent with epidemiological evidence showing that cumulative exposure to biofilm, chronic inflammation, and systemic or behavioral risk factors contributes to greater periodontal burden across adulthood [31–33]. Within a population composed entirely of adults with obesity, age likely acts as a proxy for cumulative biological and clinical susceptibility. Regarding periodontitis, this pattern has been attributed to cumulative periodontal damage over time, resulting from prolonged exposure to dental biofilm, chronic inflammation, and multiple systemic and behavioral risk factors [28, 34]. Greater FM remained independently associated with periodontitis after adjustment. Excess adiposity is widely recognized as a state of chronic low-grade systemic inflammation characterized by sustained secretion of pro-inflammatory cytokines and adipokines, all of which may intensify periodontal vulnerability [9, 12, 35, 36]. Adiposity-related inflammation and metabolic dysfunction have been implicated in increased susceptibility to periodontitis and worse periodontal outcomes [13, 14, 37–40]. Our findings suggest that the degree of adiposity may confer an incremental inflammatory burden that exacerbates periodontal vulnerability. Nevertheless, the cross-sectional design precludes causal inference. Importantly, the use of FM rather than BMI alone provides a more specific representation of adiposity burden in this population. Chronic low-grade systemic inflammation—commonly observed in both obesity and periodontitis—has been shown to disrupt erythropoiesis, impair iron homeostasis, and alter micronutrient metabolism, thus contributing to subclinical hematological changes consistent with stages preceding the clinical manifestation of anemia [41, 42]. In the present study, lower hematocrit was independently associated with prevalent periodontitis, but the present data do not support a diagnosis of anemia. Instead, the result should be interpreted as a subtle hematological difference that may reflect inflammation-related alterations in erythropoiesis or iron handling. Because average hematocrit values remained within broadly normal ranges, overinterpretation should be avoided. Obesity has been consistently linked to impaired muscle quality, reduced relative strength, and the development of sarcopenia or sarcopenic obesity, particularly in older individuals [43, 44]. Reduced HGS was the strongest functional status predictor retained in the model. This finding should be interpreted cautiously. HGS is best viewed here as a marker of overall functional reserve and systemic health rather than as a direct causal determinant of periodontal disease. Both reduced SMS and periodontitis may arise from shared pathways involving chronic inflammation and metabolic dysregulation. The relatively small number of participants with reduced HGS in the non-periodontitis group resulted in a wide confidence interval, underscoring limited statistical precision. Therefore, our findings should be interpreted with caution and require confirmation in larger, adequately powered prospective studies. From a clinical perspective, the association between muscle weakness and periodontitis in adults with obesity suggests that functional indicators such as HGS may provide complementary information for risk stratification in this population [19, 44–46]. Several methodological considerations warrant emphasis. First, the cross-sectional design precludes causal or temporal inference; the model estimates the probability of existing disease, not future incidence. Second, the modest sample size, despite the pre-specified inclusion of only four clinically grounded predictors, may have limited the precision of coefficient estimates [24, 25]. A formal sample size calculation was not performed prior to data collection. Post-hoc adequacy assessment was performed using the sample size framework [25] and, based on the observed C-statistic (0.873), four predictors, and outcome prevalence of 71.3%, indicated that criteria for overfitting control (minimum N = 84) and intercept precision (minimum N = 92) were both met by the analytical sample of 115 participants. However, the criterion for C-statistic precision estimation would require N = 315, reinforcing the need for external validation. Third, reduced HGS was defined with a sample-based threshold rather than consensus sarcopenia criteria, which limits comparability with other studies and may influence effect estimates. Fourth, external validation was not performed. Accordingly, model performance may be optimistic and generalizability to independent populations remains unknown. Fifth, participants were recruited from public primary health care settings and may differ from the broader population of adults with obesity in terms of socioeconomic background, comorbidity burden, and health-care utilization. These features may introduce selection bias and limit extrapolation to other settings. Finally, the simplified score and the provisional nomogram are exploratory implementation tools and should not be interpreted as ready-for-use clinical decision aids. Even with these limitations, the model has practical value as an initial framework for integrated oral-systemic screening research. It relies on variables that are conceptually accessible in multidisciplinary obesity care and suggests that future implementation studies may be able to combine anthropometric, functional, and hematological information to prioritize referral for periodontal evaluation. Larger external datasets, graphical calibration assessment, and recalibration will be essential next steps. Conclusion This study developed a parsimonious diagnostic prediction model for prevalent periodontitis in adults with obesity using age, FM, hematocrit, and HGS parameters. The model showed encouraging apparent performance after internal bootstrap validation, but it should be regarded as exploratory. External validation, calibration refinement, and implementation studies are required before the model, its simplified score, or its nomogram can be recommended for routine clinical use. Supplementary Information Supplementary Material 1. Supplementary Material 1.