Research Article
Using NHANES 1999–2006 data, we evaluate how a composite inflammation–metabolism index (CTI) relates to endometriosis, characterize a risk threshold around CTI 8.53, compare CTI with other inflammatory indices, and provide a reproducible, survey-weighted protocol with internal validation for population-level risk stratification.
Endometriosis is a prevalent gynecological disorder associated with chronic low-grade inflammation, yet diagnosis is often delayed because symptoms are non-specific and definitive confirmation typically relies on invasive procedures. Composite indices that jointly capture inflammatory and metabolic disturbance may improve population-level risk stratification and inform earlier clinical evaluation. Using cross-sectional data from 1,805 women aged 20–54 years in NHANES 1999–2006 (146 self-reported endometriosis; 1,659 controls), we examined survey-weighted associations between a C-reactive protein–triglyceride–glucose composite index and endometriosis risk. Survey-weighted multivariable logistic regression was used to estimate odds ratios and 95% confidence intervals for the index modeled as a continuous exposure and by quartiles. Restricted cubic splines and piecewise regression were applied to evaluate non-linear patterns and to identify potential threshold effects. To contextualize the incremental value of the composite index, we compared its performance with established inflammatory indices and with component markers, including C-reactive protein and the triglyceride–glucose index, using receiver operating characteristic analyses. This workflow provides a reproducible, survey-weighted analytic framework for evaluating composite inflammation–metabolism indicators and their threshold behavior in relation to endometriosis risk.
Endometriosis is a chronic, inflammation-related gynecological condition that affects an estimated 5%–10% of women of reproductive age1. Diagnosis traditionally depends on laparoscopic and histologic confirmation, which, although definitive, is invasive and costly and is often preceded by many years of non-specific symptoms and repeated consultations2. Epidemiologic evidence suggests that the interval from symptom onset to diagnosis commonly spans several years and can extend to nearly a decade in many settings3. These diagnostic delays can exacerbate pain, reduce quality of life, and increase the risk of infertility, underscoring the need for less invasive approaches that can support earlier risk stratification and trigger timely clinical evaluation4.
Inflammation and metabolic dysregulation are both implicated in the development and progression of endometriosis. C-reactive protein (CRP) reflects systemic low-grade inflammation, whereas triglyceride–glucose (TyG) indices capture insulin resistance and metabolic stress. The C-reactive protein–triglyceride–glucose composite index (CTI) integrates these inflammatory and metabolic axes into a single metric and may therefore represent the systemic inflammatory–metabolic milieu more comprehensively than any single parameter alone5. Prior work has linked CTI to cardiometabolic disorders, cancer outcomes, and psychosocial health, highlighting its potential as a multi-system risk indicator6. A recent NHANES-based study has also reported an association between CTI and endometriosis, indicating that CTI may have relevance in gynecologic contexts7.
Despite this emerging evidence, several gaps limit interpretation and reproducibility. Existing work has not provided a transparent, stepwise workflow that starts from raw NHANES variables and explicitly documents how CTI is computed, how covariates are coded, and how survey design features are incorporated into weighted regression. Moreover, prior analyses have not consistently clarified whether CTI offers incremental information beyond its component markers (CRP and TyG), nor have they systematically characterized non-linear dose–response patterns and threshold behavior using reproducible modeling choices. Finally, the stability of CTI-related associations across independent NHANES cycles has not been explicitly evaluated, leaving uncertainty about robustness across non-overlapping samples.
Accordingly, the present study leverages a large, nationally representative dataset to evaluate whether CTI is independently associated with endometriosis risk, to characterize dose–response and potential threshold effects, and to compare discriminatory performance across CTI, conventional inflammatory indices, and CTI component markers. By explicitly documenting an end-to-end analytic workflow and conducting internal validation using non-overlapping NHANES cycles, this work aims to provide a reproducible protocol and to clarify the most defensible role for CTI as a risk-stratification feature within epidemiologic and multi-biomarker assessment strategies rather than as a standalone diagnostic test.
Access restricted. Please log in or start a trial to view this content.
The NHANES protocol was approved by the National Center for Health Statistics (NCHS) Research Ethics Review Board, and written informed consent was obtained from all participants. This work was a secondary analysis of de-identified public-use data; therefore, no additional institutional ethical approval was required. All authors read and approved the final manuscript.
1. Study design and data source
This study was conducted as a secondary analysis of NHANES, a series of cross-sectional, nationally representative surveys administered by the U.S. Centers for Disease Control and Prevention and overseen by the NCHS Research Ethics Review Board. Public-use NHANES datasets were fully de-identified and were accessed for secondary analysis. Data from the 1999–2000, 2001–2002, 2003–2004, and 2005–2006 cycles were used.
Public NHANES component files were downloaded for each cycle, including (i) demographic files containing the participant identifier (SEQN) and survey design variables, (ii) reproductive health questionnaire files containing endometriosis self-report, and (iii) laboratory files required for the composite index computation (C-reactive protein, triglycerides, and fasting plasma glucose). Examination/anthropometry files (e.g., body mass index) and laboratory measures required for comparator indices (e.g., neutrophils, lymphocytes, platelets) were additionally obtained when those indices were analyzed. Within each 2-year cycle, component files were merged using SEQN, and the merged dataset was checked to ensure one record per SEQN. Cycle-level datasets were then appended to construct the combined 1999–2006 analytic file.
The analytic sample was restricted to women aged 20–54 years. Participants were excluded if endometriosis status was missing, if any composite-index components (C-reactive protein, triglycerides, or fasting plasma glucose) were missing, or if essential covariates required for the fully adjusted model were missing under a complete-case strategy. Complex survey design variables (strata and primary sampling units) were retained, along with the fasting laboratory subsample weights required for analyses incorporating fasting measures. When multiple NHANES cycles were combined, multi-cycle weights were created according to NHANES analytic guidance by dividing the 2-year subsample weight by the number of combined cycles, and the resulting weight, strata, and PSU variables were applied in all analyses. The steps for participant inclusion and exclusion were documented in a flow diagram (Figure 1).
2. Definition of endometriosis
Endometriosis status was defined using the reproductive health questionnaire item: “Have you ever been told by a doctor or other health professional that you have endometriosis?” Participants who responded “Yes” were classified as endometriosis cases, and those who responded “No” were classified as controls. Because this definition was based on self-report rather than laparoscopic or histologic confirmation, potential misclassification was addressed as a study limitation.
3. Definition of the Composite Index (CTI)
The C-reactive protein–triglyceride–glucose composite index was operationalized to jointly reflect systemic inflammation and metabolic disturbance. Laboratory measurements of C-reactive protein (mg/L), triglycerides (mg/dL), and fasting plasma glucose (mg/dL) were extracted from NHANES laboratory files. The triglyceride–glucose index was calculated as the natural logarithm of [triglycerides × fasting plasma glucose/2]. CTI was calculated using the following formula: CTI = 0.412 × ln(CRP) + TyG. Higher CTI values indicate a higher combined burden of low-grade inflammation and insulin resistance8.
If any C-reactive protein values required handling prior to log transformation (e.g., values at or below the detection limit), a single prespecified rule was applied consistently across all cycles and was documented to support replicability (for example, replacing non-positive values with the smallest positive measurable value observed prior to log transformation). Quartile cut points were determined from the weighted distribution in the full analytic sample and were applied consistently across categorical analyses, with Quartile 1 used as the reference category.
4. Covariates
Covariates were prespecified to mitigate confounding based on epidemiologic reasoning and prior literature. Demographic variables included age, race/ethnicity, education level, and marital status. Lifestyle variables included smoking history (≥100 cigarettes in lifetime vs. <100) and alcohol consumption (≥12 drinks/year vs. <12). Comorbidity history included self-reported hypertension, diabetes, stroke, coronary heart disease, and cancer. Anthropometric and laboratory variables included body mass index, hemoglobin, neutrophil count, lymphocyte count, and platelet count; these measures also supported computation of comparator inflammatory indices where applicable (e.g., neutrophil-to-lymphocyte ratio, platelet-to-lymphocyte ratio, systemic immune-inflammation index, systemic inflammation response index). Reproductive variables (e.g., gravidity and parity) were included when available in the selected cycles and were coded according to NHANES documentation. Categorical covariates were converted to indicator variables prior to model entry.
Because C-reactive protein was a component of the composite index, it was not entered as an independent covariate in multivariable regression models to avoid overadjustment and collinearity. Instead, C-reactive protein and the triglyceride–glucose index were evaluated as comparator markers in the discrimination analyses.
5. Statistical analysis
All analyses accounted for the NHANES complex survey design to generate nationally representative estimates. The survey design was specified by linking the multi-cycle subsample weight, strata, and PSU variables to the analytic dataset. Continuous variables were summarized as weighted means with standard deviations, and categorical variables were summarized as weighted counts and percentages. Baseline characteristics were compared between cases and controls using survey-weighted procedures appropriate for NHANES, and baseline characteristics were summarized in Table 1.
Associations between the composite index and endometriosis were evaluated using survey-weighted logistic regression. Three sequential models were fitted to demonstrate adjustment: an unadjusted model, a model adjusted for age and race/ethnicity, and a fully adjusted model including demographic factors, lifestyle variables, comorbidity history, anthropometric/laboratory covariates, and reproductive history variables. The composite index was analyzed both continuously (per 1-unit increase) and categorically (quartiles, with Quartile 1 as the reference), and regression estimates were summarized in Table 2. Linear trend across quartiles was tested by assigning each quartile its weighted median value and modeling that term continuously.
Non-linear dose–response relationships were assessed using survey-weighted restricted cubic splines with prespecified knot placement, and spline curves were plotted in Figure 2. Threshold effects were evaluated using survey-weighted segmented (piecewise) logistic regression by comparing model fit between segmented and single-slope specifications, and the estimated inflection point and slope parameters on each side of the inflection point were reported in Table 3.
Subgroup analyses were conducted to explore effect modification by prespecified factors (e.g., age group, race/ethnicity, education level, marital status, and selected lifestyle factors). Interaction was tested by including cross-product terms between the continuous composite index and subgroup indicators within the survey-weighted framework, and subgroup associations were summarized in Figure 3.
Discriminatory performance was evaluated using receiver operating characteristic analyses based on model-predicted probabilities derived from survey-weighted logistic models. Area under the curve estimates were obtained for the composite index, commonly used inflammatory indices, and component markers, and AUC summaries were provided in Supplementary Table 1; an expanded ROC comparison was provided in Supplementary Figure 1.
Missing data were handled using complete-case analysis after excluding participants with missing endometriosis status, missing composite-index components, or missing essential covariates required for the fully adjusted model. When a robustness assessment was performed, multiple imputation was applied for covariates with missingness under a prespecified imputation model, and imputed estimates were compared with complete-case estimates.
Analyses were performed using R (version 4.4.1) and additional statistical software as listed in the Table of Materials. Key packages used for survey inference, spline modeling, segmented regression, and ROC estimation were recorded, and session information (operating system and R session details) was retained to support replication.
6. Procedure end point and outputs
The analytic workflow was considered complete once the harmonized multi-cycle dataset was constructed with the prespecified inclusion/exclusion criteria (Figure 1), the composite index and covariates were generated according to documented rules, and the prespecified survey-weighted regression, non-linearity/threshold assessment, subgroup, and discrimination analyses were executed under the same survey design specification. The primary outputs of this workflow were organized as a baseline summary (Table 1), regression estimates across sequential adjustment models (Table 2), threshold model parameters (Table 3), spline visualization (Figure 2), subgroup summary visualization (Figure 3), and discrimination summaries (Supplementary Table 1 and Supplementary Figure 1).
7. Internal independent validation
Internal independent validation was performed by splitting the combined dataset into a derivation cohort and a non-overlapping validation cohort based on NHANES cycles. Participants from the 1999–2000 and 2001–2002 cycles were assigned to the derivation cohort, and participants from the 2003–2004 and 2005–2006 cycles were assigned to the validation cohort. The same inclusion/exclusion criteria, composite index computation, covariate coding rules, and survey-weighting strategy were applied independently within each cohort.
Within the derivation cohort, survey-weighted logistic regression models were fitted using the fully adjusted specification. The non-linearity and threshold assessment procedures used in the main analysis were applied in the derivation cohort, and discrimination was evaluated using ROC/AUC methods based on model-predicted probabilities. The same modeling strategy was then repeated in the validation cohort without modifying variable definitions, coding rules, or weighting specifications. Derivation-versus-validation estimates were summarized as a cohort-to-cohort comparison of association estimates (Figure 4) and as derivation-versus-validation ROC curves (Figure 5), with corresponding numerical summaries provided in Table 4.
Access restricted. Please log in or start a trial to view this content.
Baseline characteristics
A total of 1,805 women were included in the analysis, of whom 146 (8.1%) reported a diagnosis of endometriosis and 1,659 (91.9%) did not. Women without endometriosis served as the control group. Baseline characteristics are summarized in Table 1. Women with endometriosis were older than controls (40.7 ± 8.3 vs. 36.8 ± 9.6 years, P < 0.001). The distribution of race/ethnicity differed between groups: 70.5% of women with endometriosis...
Access restricted. Please log in or start a trial to view this content.
This study demonstrates that the C-reactive protein–triglyceride–glucose index (CTI), a composite measure of inflammatory and metabolic status, is independently and positively associated with the odds of endometriosis in a nationally representative sample of women (Table 2). This interpretation is consistent with previous evidence implicating oxidative stress, immune activation, and metabolic dysregulation in the development and progression of endometriosis9. By integr...
Access restricted. Please log in or start a trial to view this content.
The authors declare that they have no competing financial interests. Artificial intelligence tools (including large language models) were used for language polishing and clarity improvement only. All analyses, study design decisions, interpretation, and final wording were produced and verified by the authors, who take full responsibility for the accuracy and integrity of the manuscript.
This study was supported by the Jiangsu Vocational College of Medicine School–Local Collaborative Innovation Project (Grant No. 202491016).
Access restricted. Please log in or start a trial to view this content.
Access restricted. Please log in or start a trial to view this content.
Request permission to reuse the text or figures of this JoVE article
Request Permission