Data for this current study was from a prospective multi-centre study across eight tertiary medical centres in Korea. Institutional review boards across study sites including, Chonnam National University Hospital (CNUH-2021-250), Chungnam National University Sejong Hospital (CNUSH 2021-08-002), Dongguk University Hospital (DUIH 2021-03-030-005), Kyungpook National University Hospital (KNUH 2021-05-011), Chungbuk National University Hospital (CBNUH 2021-07-027-001), Kangwon National University Hospital (KNUH-A-2021-05-011-012), Eulji University Hospital (EMCS 2022-12-015), and Jeju National University Hospital (2021-06-005). This study followed the ethical principles of medical research involving human participants as outlined in the Declaration of Helsinki. All participants were prospectively enrolled after IRB approval at each study site and provided informed consent before participation in the study.

Trained personnel recruited participants from individuals undergoing gastrointestinal endoscopy (esophagogastroduodenoscopy and/or colonoscopy) at the above-mentioned medical centres between July 2021 and September 2023. Comprehensive data encompassing demographic and lifestyle factors, medical history, and metabolic parameters were obtained from medical records. Information on dietary intake was collected using a standardized food frequency questionnaire. From the initial pool of 912 participants, 271 were excluded based on the following criteria: absence of complete colonoscopy information (n = 228), presence of gastrointestinal dysfunction (n = 43), incomplete dietary data (n = 11), and implausible energy intake (defined as an energy intake of more or less than three standard deviations from the mean log-transformed energy intake; n = 14). The remaining IBD patients were matched with IBD-free non-cases on a 1:1 basis according to age and sex using propensity score matching, resulting in a final cohort of 127 case-matched pairs (Fig. 1).

IBD, which includes CD, UC, and unclassified IBD, was diagnosed by trained specialists using a comprehensive diagnostic imaging assessment (including endoscopy, computed tomography, and magnetic resonance imaging), histological examinations, clinical symptom evaluation, blood and faecal test results, medication history and treatment response at each hospital site following standard guidelines [23]. Furthermore, CD and UC were classified according to the Montreal classification for IBD [24].

Participants completed questionnaires on their sociodemographics, lifestyle information, and medical histories. Participants reported age (in years) and sex as either ‘male’ or ‘female’. Women’s menopausal status was presented as ‘premenopausal’ or ‘post-menopausal,’ and the highest education completed was reported and classified as ‘elementary to high school’ or ‘college graduate and above’. In addition, current smoking was defined as having smoked a cigarette at least once in the last 12 months. Physical activity in metabolic equivalent tasks (MET-hours/week) was estimated based on the average hours and the number of days spent on moderate to intense levels of physical activity or walking [25] and classified as ‘none’, ‘< 14 MET-hours/week,’ and ‘≥ 14 MET-hours/week’. Body mass index (BMI) was estimated as a function of weight (kg) divided by the square of the height (m), and obesity status was defined as BMI ≥ 25 kg/m² [26]. Participants provided their dietary intake information using validated food frequency questionnaires. Information on the typical food consumption and portion sizes in the last 12 months preceding the study was reported and documented [27]. There were nine options (ranging from ‘never’ or ‘less than once/month’ to ‘three times/ day’ and three portion size options; ‘one-half of a standard serving’, ‘one standard serving,’ and ‘one and half of standard serving’) for each food and drink item, and average daily consumption of food (in grams/day) and nutrients was determined by multiplying the frequency of consumption with the reported amount. Details of the food frequency questionnaire assessment and food intake estimation have been reported previously [21,28]. Alcohol intake (in grams/day) was estimated as the total ethanol weight, including the multiplication of quantities and frequencies of the types of liquors/alcohol. Also, participants were asked if they had been previously diagnosed with hypertension [29] and diabetes mellitus [30]. Serum levels of hs-CRP were measured in IBD patients using clinical chemical analyzers, specifically the cobas 8000 (Roche Diagnostics System, Basel, Switzerland) and the Beckman Au5821 automatic biochemical analyzer (Beckman Coulter Inc., Brea, CA, USA). The coefficients of variation of these biomarkers were < 2%.

The predicted pro-inflammatory hs-CRP score model used in this study was developed [31] and validated [21] in the Korean population. Briefly, pro-inflammatory factors previously reported to be associated with the blood profiles of hs-CRP in the literature were applied to model the predicted pro-inflammatory hs-CRP score through an empirical constellation of these factors in a linear regression model. These factors include food groups, nutrients, alcohol intake, BMI, smoking status, physical activity, educational level, and menopausal status. The predicted pro-inflammatory hs-CRP score was computed as the total score of the multiplication of the beta coefficients (from sex-combined and sex-specific predicted pro-inflammatory hs-CRP score derived from empirical models) with subject-level pro-inflammatory demographic and lifestyle factors and food intake (grams/day) (Supplementary Table 1). The predicted pro-inflammatory hs-CRP score was categorized into tertiles to include a reasonable number of participants in each category for statistical comparisons.

Patients with IBD were matched with IBD-free non-cases by propensity scores. Propensity scores were estimated using a logistic regression model and inverse probability treatment weighting, with age (continuous) and sex (male vs. female) as covariates. IBD cases were matched with IBD-free non-cases in a 1:1 ratio, with the nearest neighbours having propensity scores within the exact boundaries, and unmatched participants were excluded. After matching, the propensity scores and covariates distributions were reviewed to match quality and consistency.

Characteristics of participants by IBD status and across tertile distribution of the predicted pro-inflammatory hs-CRP score were presented using mean ± standard deviation (SD) or median and (range) as applicable for continuous data and frequencies (percentages) for categorical data. Conditional logistic regression models were used to estimate the odds ratios (ORs) and 95% confidence intervals (CIs) for the prevalence of IBD according to tertiles of sex-combined and sex-specific predicted pro-inflammatory hs-CRP score. Differences in ORs were assessed when determining the variables to be included in the final model. The regression model in the sex-combined predicted pro-inflammatory hs-CRP scores was adjusted for age (continuous, years), sex (male or female), alcohol intake (continuous, grams/day), smoking (no, past, or current), highest education completed (high school or less, or college education and above), physical activity (continuous, MET-minutes/week), dietary energy intake (continuous, kcal/day), diabetes mellitus status (no or yes), and hypertension status (no or yes). The final regression model of the male-specific predicted pro-inflammatory hs-CRP score was adjusted for the same covariates, excluding sex. Also, the final regression model for the female-specific predicted pro-inflammatory hs-CRP score was adjusted for the same covariates in the sex-combined model and post-menopausal status (no or yes), excluding sex. The median value of the tertile distribution of the predicted pro-inflammatory hs-CRP score was assigned in a continuous model to assess linear trends in the relationship between the predicted pro-inflammatory hs-CRP score and IBD.

In the subgroup analyses by IBD subtype (CD or UC), obesity status (< 25 or ≥ 25 kg/m²), hypertension status (no or yes), and post-menopausal status (no or yes for females), the predicted pro-inflammatory hs-CRP score was dichotomized by median values due to the small sample size. Effect modification was evaluated by including interaction term(s) and comparing the nested models with or without the cross-product using the likelihood ratio test. In addition, the receiver operating characteristic (ROC) curve was plotted in sex-combined and sex-specific regression models by treating the predicted pro-inflammatory hs-CRP score in a continuous model. The area under the curve (AUC) and 95% CI were estimated to determine the predictive ability of the predicted pro-inflammatory hs-CRP score in discriminating IBD. All statistical analyses were conducted using SAS (version 9.4; SAS Institute Inc., Cary, NC, USA), with a two-sided p < 0.05.

Of the 127 case-non-case pairs of patients with IBD included in this study (Fig. 1), 59 (46.5%) had CD, and 68 (53.7%) had UC. The characteristics, medication history, and results of blood and faecal tests for IBD cases only (including UC and CD) are presented in Supplementary Table 2. Overall, 51 (40.2%) were females, and the median (range) for CRP was 0.1 (0–5.0) mg/dL. Also, the characteristics of the 127 case-non-case pairs are in Supplementary Table 3. Overall, the mean age was 44.6 ± 14.1 years, 102 (40.2%) were females, 56 (22.1%) were current smokers, and 150 (59.1%) reported alcohol use (Table 1). Also, the median (interquartile range) for physical activity was 14.0 (5.3, 28.8) MET-hour/week, the mean BMI was 25.3 ± 4.6 kg/m², and 119 (46.9%) had a BMI of at least 25 kg/m². Mean energy intake was 2,108.6 ± 972.9 kcal/day, mean glycated haemoglobin was 13.6 ± 1.7%, 21 (8.3%) had diabetes mellitus, and 48 (18.9%) were hypertensives.

When the characteristics of participants were stratified by IBD status (Supplementary Table 3), age and sex did not differ, but IBD patients vs. non-IBD cases presented a lower prevalence of current smoking (19.7% vs. 24.4%, respectively), alcohol use (48.0% vs. 70.1%, respectively), and hypertension (17.3% vs. 20.5%, respectively). In contrast, the prevalence of obesity (58.3% vs. 35.4%) was higher among IBD patients than non-IBD cases. In Table 1, participants in the third tertile of the predicted pro-inflammatory hs-CRP score presented with a higher prevalence of IBD, current smoking, alcohol use, obesity, diabetes mellitus, and hypertension, but a lower proportion of females compared to those in the first tertile of the predicted pro-inflammatory hs-CRP score. This trend was similar to the sex-specific models of the pro-inflammatory predicted hs-CRP score (Supplementary Table 4).

The ORs and 95% CIs of IBD prevalence by tertile distribution of the predicted pro-inflammatory hs-CRP score are presented in Table 2. In the age-adjusted conditional logistic regression model, a higher predicted pro-inflammatory hs-CRP score was associated with a higher prevalence of IBD: 1.00, 0.81 (0.38, 1.72), and 2.66 (1.08, 6.54; p for trend = 0.02). The ORs were robust when the model was additionally adjusted for sex, alcohol intake, current smoking status, education, physical activity, dietary energy intakes, diabetes mellitus status and hypertension status; 1.00, 0.88 (0.38, 2.07) and 8.11 (2.07, 31.81; p for trend = 0.006). A similar but statistically insignificant trend was observed for the male-specific predicted pro-inflammatory hs-CRP score: 1.00, 2.09 (0.61, 7.16), and 3.65 (0.85, 15.67; p for trend = 0.08) and female-specific predicted pro-inflammatory hs-CRP score; 1.00, 4.07 (0.82, 20.13), and 2.88 (0.48, 17.42; p for trend = 0.25).

We dichotomized the sex-combined predicted pro-inflammatory hs-CRP scores for the subgroup analysis by IBD subtype, obesity, hypertension, and menopausal status (for the female-specific model only) – Table 3. A higher predicted pro-inflammatory hs-CRP score was associated with a higher prevalence of UC [< median – 1.00 and ≥ median – 5.78 (1.29, 25.89)], but not for CD [< median – 1.00 and ≥ median – 1.44 (0.31, 6.69)]. A similar trend was observed in the overall sample; however, a higher predicted pro-inflammatory hs-CRP score was not significantly associated with a higher prevalence of IBD, independent of obesity, hypertension, or menopausal status, with no evidence of interaction (Table 3).

Furthermore, the predicted pro-inflammatory hs-CRP score was treated as a continuous variable to evaluate its ability to classify IBD (Fig. 2), and the AUC (95% CI) was 0.67 (0.61, 0.74) for the sex-combined model, 0.71 (0.63, 0.80) for the male-specific model, and 0.65 (0.54, 0.75) for the female-specific model. Similarly, the AUC (95% CI) results for the IBD subtype were 0.68 (0.58, 0.77) for CD and 0.72 (0.64, 0.81) for UC, thereby suggesting that the predicted pro-inflammatory hs-CRP score possibly demonstrated a modest predictive ability in discriminating IBD with a portentous likelihood of viability for population-level risk screening to identify at-risk groups for early timely intervention.