Development and validation of a risk prediction model for small intestinal mucosal injury in nonsteroidal anti-inflammatory drug users

Abstract

BACKGROUND

Despite the enhanced detection capability of capsule endoscopy for small intestinal lesions, the independent risk factors and distinct clinical-endoscopic characteristics in high-risk populations remain incompletely elucidated.

AIM

To investigate the risk factors for small intestinal mucosal injury (SIMI) in nonsteroidal anti-inflammatory drug (NSAID) users and to establish a corresponding predictive model.

METHODS

We conducted a retrospective analysis of clinical data from patients undergoing capsule endoscopy at our institution and Shanxi Bethune Hospital from August 2012 to January 2025. Multivariate analysis was performed to identify independent risk factors for SIMI. A nomogram was subsequently developed to predict the risk of SIMI in NSAID users. Receiver operating characteristic curve analysis, calibration curve and decision curve analysis were performed to evaluate the predictive accuracy of the nomogram.

RESULTS

In the primary cohort, SIMI was identified in 114 out of 181 NSAID users. Multivariate analysis revealed that advanced age, smoking, proton pump inhibitor use, elevated body mass index, high triglyceride levels, and increased low-density lipoprotein were independent risk factors for SIMI. The research team developed a risk prediction model for estimating the risk of mucosal injury, which achieved an area under the curve of 0.775 (95% confidence interval: 0.700-0.849) in the derivation set. The model exhibited an area under the curve of 0.797 (95% confidence interval: 0.643-0.952) in the validation cohort.

CONCLUSION

We identified risk factors for SIMI in NSAID users and established a predictive model, which may facilitate early identification of high-risk populations and guide clinical interventions.

Key Words: Small intestinal mucosal injury; Nonsteroidal anti-inflammatory drug; Capsule endoscopy; Risk factors; Prediction model

Core Tip: This study identified advanced age, smoking, proton pump inhibitor use, elevated body mass index, high triglyceride, and low-density lipoprotein levels as independent risk factors for small intestinal mucosal injury in nonsteroidal anti-inflammatory drug users. A predictive nomogram was developed based on these factors. The model demonstrated good predictive accuracy, with an area under the curve of 0.775 in the derivation cohort and 0.797 in the validation cohort. This tool can facilitate early identification of high-risk individuals and guide clinical interventions to prevent small intestinal mucosal injury.

INTRODUCTION

With the aging population and increasing incidence of cardiovascular and cerebrovascular diseases, nonsteroidal anti-inflammatory drugs (NSAIDs), including aspirin, have become one of the most widely used medications worldwide. NSAIDs are extensively employed for analgesia and anti-inflammatory purposes, while low-dose aspirin is commonly used for the prevention and treatment of cardiovascular events and stroke[1]. It is well-established that NSAID use is associated with an elevated risk of upper gastrointestinal mucosal injury, manifesting as peptic ulcers and/or bleeding[2,3]. Studies indicate that such NSAID-related gastrointestinal complications are not confined to the upper tract - NSAIDs are also recognized as a leading cause of obscure gastrointestinal bleeding, with lesions frequently localized in the small intestine[4].

Unlike upper gastrointestinal injury, which can be promptly assessed via conventional endoscopy, small intestinal mucosal damage remains diagnostically challenging due to its anatomical inaccessibility. Capsule endoscopy is a highly useful and noninvasive method for evaluating small bowel diseases and is considered one of the best modalities for assessing small intestinal mucosal lesions[5-8]. With advancements in capsule endoscopy, small bowel pathology has garnered increasing attention among clinicians.

NSAID-induced small intestinal injury is a clinically significant yet underrecognized condition[9]. The risk factors for NSAID-associated small intestinal mucosal injury (SIMI) remain unclear, and its endoscopic and clinical features are often difficult to delineate accurately due to overlapping pathologies such as inflammatory bowel disease or intestinal tuberculosis. Of particular concern is that NSAID-related small intestinal injury frequently progresses asymptomatically until severe complications arise, including occult bleeding, protein-losing enteropathy, or iron-deficiency anemia[10]. Therefore, early identification of high-risk individuals and precise characterization of injury patterns are crucial for reducing morbidity. Accordingly, this study aimed to develop and validate a risk prediction model for SIMI in NSAID users by identifying key risk factors and characterizing endoscopic features based on capsule endoscopy findings.

MATERIALS AND METHODS

Study population

This retrospective study reviewed medical records of patients who underwent capsule endoscopy at Tongji Hospital and Shanxi Bethune Hospital between August 2012 and January 2025. Exclusion criteria included: (1) Age < 18 years; (2) Diagnosed inflammatory bowel disease; (3) Intestinal tuberculosis; (4) Henoch-Schönlein purpura; and (5) Incomplete clinical data. We collected comprehensive clinical data (including demographic characteristics, medical history, medication use, and laboratory tests) and capsule endoscopy findings. NSAID users were defined as patients who had received continuous oral NSAID treatment for at least 2 weeks prior to the capsule endoscopy procedure. The Human Investigation Committee of Tongji Medical College, Huazhong University of Science and Technology approved this study, approval No. 2024-S079. Details that might disclose the identity of the patients had been omitted.

Capsule endoscopy procedure

The examination was performed using OMOM, PillCam SB 3, or PillCam SB 2 capsule endoscopy systems. Although these systems differ in technical specifications such as image resolution and field of view, they are all clinically validated for the detection of significant mucosal injury. Patients fasted overnight prior to the procedure and underwent bowel preparation with oral polyethylene glycol electrolyte powder. After capsule ingestion, clear liquids were permitted at 2 hours and light meals at 4 hours.

Capsule endoscopy image review

All previous capsule endoscopy records were retrospectively evaluated, with reinterpretation of the original images. Two expert endoscopists, each with experience in over 300 capsule endoscopy interpretations, independently reviewed all recordings. A third senior endoscopist with over 20 years of experience and more than 500 interpretations served as the adjudicator for any discrepancies. All identified lesions were systematically documented, and corresponding images were archived for further analysis.

Small intestinal lesions were categorized into seven distinct phenotypes: (1) Erythema: Defined as focal red spots or crimson mucosal areas with preserved villous architecture[11]; (2) Erosions: Discrete lesions featuring central pallor surrounded by congested mucosa with villous loss[12]; (3) Ulcers: Circumscribed mucosal breaks with exudate-covered bases and typically elevated margins[13]; (4) Protrusions: Including inflammatory hyperplasia, lymphoid follicles, stromal tumors, polyps, extrinsic compressions, or indeterminate masses; (5) Vascular malformations; (6) Diverticula: Outpouchings from the small bowel wall; and (7) Lymphangiectasia[7]. Representative images are presented in Figure 1.

Figure 1 Representative image of small intestine lesions. A: Erythema; B: Erosion; C: Ulcer; D: Vascular malformation; E: Diverticulum; F: Lymphangiectasia; G-I: Protrusion. White arrow: The location of the lesions in each panel.

Following image review, subjects presenting with erythema, erosions, or ulcers in the small intestine were classified into the SIMI group, while all remaining subjects were assigned to the non mucosal injury group[7].

Statistical analysis

Categorical variables were expressed as n (%) and compared using the χ² test or Fisher’s exact test, as appropriate. Continuous variables were assessed for normality. Normally distributed data are presented as mean ± SD and compared using the independent samples t test, while non-normally distributed data are expressed as median (interquartile range) and compared using the Mann-Whitney U test. Variables with a statistically significant association in the univariate analysis (P < 0.05) were identified as candidates and were all simultaneously entered into the multivariate logistic regression model. Independent influencing factors were identified through multivariate regression, and a nomogram drawn was constructed primarily based on these results. The predictive performance of the nomogram model was evaluated by the area under the curve (AUC), calibration curve, and decision curve analysis. Additionally, internal validation was carried out. The nomogram drawn was generated using R software (version 4.3.3), while all other statistical analyses were conducted with SPSS version 26.0 (IBM Corp.). A two-tailed P value < 0.05 was considered statistically significant.

RESULTS

Clinical characteristics and endoscopic findings of enrolled patients

Between August 2012 and January 2025, 924 patients underwent capsule endoscopy at two participating hospitals. After excluding 59 patients aged < 18 years, 27 with incomplete clinical data, 38 with confirmed inflammatory bowel disease, 2 with intestinal tuberculosis, and 3 with anaphylactoid purpura, 795 participants were ultimately included in this study (Figure 2).

Figure 2 Flowchart of patient enrollment and exclusion criteria in the current study. NSAIDs: Nonsteroidal anti-inflammatory drugs; SIMI: Small intestinal mucosal injury.

The cohort had a median age of 55 years, with 459 males (57.74%) and a mean body mass index (BMI) of 22.87 ± 3.39 kg/m². Current smokers accounted for 43.65% of participants, while 28.05% reported alcohol consumption. Among the 795 participants, 747 (93.96%) presented with gastrointestinal symptoms - predominantly gastrointestinal bleeding (the most common manifestation), followed by abdominal pain, diarrhea, and bloating. The remaining 48 participants were asymptomatic individuals undergoing routine screening (Table 1). Table 1 also summarizes the endoscopic characteristics of small intestinal lesions. Significant small bowel pathology was detected in 546 patients (68.68%), with erosions and protrusions being the most prevalent findings. The majority of lesions involved both jejunal and ileal segments.

Table 1 Baseline characteristics of enrolled patients and comparison between nonsteroidal anti-inflammatory drug users and nonusers, n (%).

Variables	n = 795	NSAID group (n = 181)	Non-NSAID group (n = 614)	P value
Sex, male	459 (57.74)	108 (59.67)	351 (57.17)	0.549
Age (years)	55.00 (43.00, 64.50)	60.00 (52.00, 66.00)	52.50 (39.00, 62.25)	< 0.001
BMI (kg/m2)	22.87 ± 3.39	23.45±2.73	22.39 ± 3.80	0.003
Smoking habit	347 (43.65)	83 (45.86)	264 (43.00)	0.495
Alcohol consumption	223 (28.05)	55 (30.39)	168 (27.36)	0.426
Comorbidity
Hypertension	305 (38.36)	92 (50.83)	213 (34.69)	< 0.001
Diabetes mellitus	111 (13.96)	29 (16.02)	82 (13.36)	0.363
Cardiovascular disease	153 (19.25)	106 (58.56)	47 (7.65)	< 0.001
Stroke	59 (7.42)	46 (25.41)	13 (2.12)	< 0.001
Liver cirrhosis	15 (1.89)	3 (1.66)	12 (1.95)	0.796
Renal disease	14 (1.76)	5 (2.76)	9 (1.47)	0.224
Rheumatic/autoimmune disorders	23 (2.89)	19 (10.50)	4 (1.09)	< 0.001
Presenting symptoms				< 0.001
Abdominal pain	191 (24.02)	28 (15.47)	163 (26.55)
Diarrhea	53 (6.67)	4 (2.21)	49 (7.98)
Gastrointestinal bleeding	483 (60.75)	140 (77.35)	343 (55.86)
Abdominal distension	20 (2.52)	0 (0.00)	20 (3.26)
Asymptomatic (health check-up)	48 (6.04)	9 (4.97)	39 (6.35)
Hb (g/L)	97.92 ± 31.31	91.05 ± 26.53	99.20 ± 32.00	0.058
WBC (× 109/L)	6.04 ± 4.70	5.85 ± 1.97	6.07 ± 4.98	0.788
Alb (g/L)	39.26 ± 5.46	38.70 ± 5.54	39.37 ± 5.45	0.440
TG (mmol/L)	1.46 ± 0.52	1.50 ± 0.59	1.41 ± 0.43	0.103
LDL (mmol/L)	2.74 (1.69, 3.29)	2.86 (1.88, 3.42)	2.49 (1.59, 3.22)	0.190
HDL (mmol/L)	1.27 (1.07, 1.43)	1.25 (1.07, 1.42)	1.29 (1.16, 1.45)	0.335
Presence of significant small intestinal lesions	546 (68.68)	153 (84.53)	393 (64.01)	< 0.001
Location of lesions				< 0.001
Predominantly jejunal	177 (22.26)	85 (46.96)	92 (14.98)
Predominantly ileal	173 (21.76)	31 (17.13)	142 (23.13)
Both jejunal and ileal	196 (24.65)	37 (20.44)	159 (25.90)
Endoscopic manifestations
Erythema	59 (7.42)	34 (18.78)	25 (4.07)	< 0.001
Erosions	209 (26.29)	87 (48.07)	122 (19.87)	< 0.001
Ulcers	90(11.32)	32 (17.68)	58 (9.45)	0.002
Protrusions	131 (16.48)	17 (9.39)	114 (18.57)	0.003
Vascular malformations	115 (14.47)	17 (9.39)	98 (15.96)	0.027
Diverticula	42 (5.28)	9 (4.97)	33 (5.37)	0.832
Lymphangiectasia	21(2.64)	7 (3.87)	14 (2.28)	0.242

NSAID: Nonsteroidal anti-inflammatory drug; BMI: Body mass index; Hb: Hemoglobin; WBC: White blood cell; Alb: Albumin; TG: Triglyceride LDL: Low density lipoprotein; HDL: High-density lipoprotein.

Clinical and capsule endoscopic characteristics of NSAID users

For further characterization of NSAID-related effects, patients were stratified by NSAID use into the NSAID group (n = 181) and non-NSAID group (n = 614). Compared with non-NSAID users, NSAID users were significantly older (P < 0.001) and had higher BMI values (P = 0.003). Gastrointestinal bleeding was more prevalent in the NSAID group, whereas abdominal pain, diarrhea, bloating, and asymptomatic cases were more common in the non-NSAID group (P < 0.001). The NSAID group also showed higher rates of hypertension, cardiovascular disease, cerebral infarction, and rheumatic/autoimmune disorders (all P < 0.001). No significant intergroup differences were observed in gender distribution, smoking/alcohol habits, diabetes, liver cirrhosis, renal diseases, white blood cell count, hemoglobin or lipid levels.

The NSAID group demonstrated a significantly higher incidence of small intestinal lesions compared to the non-NSAID group (P < 0.001), with predominant involvement of the jejunum. Erythema, erosions, and ulcers are the primary manifestations of NSAID-associated small intestinal injury. In contrast, the non-NSAID group showed significantly higher rates of protrusions and vascular malformations (Table 1).

Analysis of risk factors for SIMI in NSAID users

The 181 NSAID-exposed patients were stratified into two subgroups based on the presence of SIMI: The NSAID SIMI group (n = 114) and NSAID non mucosal injury group (n = 67). Compared to the non mucosal injury subgroup, patients with SIMI were significantly older (P = 0.009), had higher smoking rates (P = 0.017), elevated BMI (P = 0.011), higher triglyceride (TG) levels (P = 0.007), higher low-density lipoprotein (LDL) levels (P = 0.004), and a greater rate of proton pump inhibitor (PPI) use (P = 0.025). No significant differences were observed between groups regarding gender, alcohol consumption, clinical symptoms, indications for NSAID use, NSAID types or duration of use, leukocyte count, hemoglobin or albumin levels (Table 2). Multivariate analysis identified age, smoking, PPI use, BMI, TG, and LDL as independent risk factors for mucosal injury in NSAID users (Table 3).

Table 2 Comparison of clinical characteristics between nonsteroidal anti-inflammatory drug exposed patients with and without small intestinal mucosal injury, n (%).

Variables	SIMI group (n = 114)	Non-SIMI group (n = 67)	P value
Sex, male	70 (61.40)	38 (56.72)	0.535
Age (years)	60.64 ± 10.22	56.29 ± 9.48	0.009
Smoking habit	60 (52.63)	23 (34.33)	0.017
Alcohol consumption	36 (31.58)	19 (28.36)	0.649
BMI (kg/m2)	23.87 ± 2.67	22.73 ± 2.72	0.011
Presenting symptoms			0.627
Abdominal pain	15 (13.16)	13 (19.40)
Diarrhea	2 (1.75)	2 (2.99)
Gastrointestinal bleeding	91 (79.82)	49 (73.13)
Asymptomatic (health check-up)	6 (5.26)	3 (4.48)
Indications for NSAID use			0.644
Cardiovascular disease	77 (67.54)	43 (64.18)
Other	37 (32.46)	24 (35.82)
NSAID types			0.404
Aspirin	88 (77.19)	48 (71.64)
Other NSAIDs	26 (22.81)	19 (28.36)
Prolonged use (≥ 3 months)	96 (84.21)	54 (80.60)	0.533
PPI	69 (60.52)	29 (43.28)	0.025
WBC (× 109/L)	5.83 ± 2.07	5.95 ± 1.55	0.887
Hb (g/L)	89.69 ± 24.28	98.19 ± 37.49	0.412
Alb (g/L)	38.36 ± 5.50	39.73 ± 5.77	0.461
TG (mmol/L)	1.60 ± 0.62	1.34 ± 0.49	0.007
LDL (mmol/L)	3.03 (1.95, 3.51)	2.36 (1.58, 2.97)	0.004
HDL (mmol/L)	1.25 (1.07, 1.42)	1.25 (1.05, 1.44)	0.802

SIMI: Small intestine mucosal injury; BMI: Body mass index; NSAID: Nonsteroidal anti-inflammatory drug; PPI: Proton pump inhibitor; WBC: White blood cell; Hb: Hemoglobin; Alb: Albumin; TG: Triglyceride LDL: Low density lipoprotein; HDL: High-density lipoprotein.

Table 3 Multivariate analysis identified the risk factors for small intestinal mucosal injury in patients using nonsteroidal anti-inflammatory drugs, n (%).

Characteristic	Univariate analysis		Multivariate analysis
	OR (95%CI)	P value	OR (95%CI)	P value
Age	1.045 (1.010-1.082)	0.011	1.043 (1.003-1.084)	0.033
Smoking	2.126 (1.139-3.968)	0.018	2.352 (1.088-5.086)	0.030
BMI	1.174 (1.035-1.331)	0.013	1.176 (1.019-1.358)	0.026
PPI	2.009 (1.089-3.705)	0.025	2.099 (1.003-4.390)	0.049
TG	2.279 (1.232-4.215)	0.009	2.427 (1.196-4.924)	0.014
LDL	1.557 (1.107-2.188)	0.011	1.606 (1.089-2.367)	0.017

CI: Confidence interval; OR: Odds ratio; BMI: Body mass index; PPI: Proton pump inhibitor; TG: Triglyceride LDL: Low density lipoprotein; NSAID: Nonsteroidal anti-inflammatory drug.

The prediction model grounded on independent risk factors

The logistic regression model demonstrated good fit using the Hosmer and Lemeshow test (χ² = 2.759, P = 0.949). Based on the results of the multivariate logistic regression analysis, a risk prediction nomogram model for SIMI in NSAID users was constructed (Figure 3). A receiver operating characteristic curve was drawn to estimate the predictive accuracy of the nomogram, and the AUC was 0.775 (Figure 4A). The calibration curve demonstrated a strong agreement between the observed and predicted probabilities (Figure 4B). Decision curve analysis showed that the predictive model provided substantial net clinical benefit across a wide range of threshold probabilities (approximately 0.2 to 0.88) (Figure 4C). For validation, we utilized a cohort of 37 patients who underwent capsule endoscopy at two hospitals between February and July 2025. Among them, 23 patients were diagnosed with SIMI, while 14 showed no evidence of injury. The results demonstrated that the risk scoring system exhibited good predictive accuracy, achieving an AUC of 0.797 (Figure 4D).

Figure 3 A nomogram for assessing the risk of small intestinal mucosal injury in nonsteroidal anti-inflammatory drug users. BMI: Body mass index; PPI: Proton pump inhibitor; TG: Triglyceride LDL: Low density lipoprotein.

Figure 4 Evaluation of the risk prediction model for small intestinal mucosal injury in both the development and validation sets. A: Receiver operating characteristic curve of the development set, with an area under the curve value of 0.775; B: Calibration curve of this nomogram model; C: Decision curve analysis of this nomogram model; D: Receiver operating characteristic curve of the validation set, with an area under the curve value of 0.797. AUC: Area under the curve; CI: Confidence interval; DCA: Decision curve analysis.

DISCUSSION

SIMI represents a clinically significant pathological condition with multifactorial etiology and diverse manifestations, particularly with the growing recognition of NSAID-induced enteropathy. Extensive prior research has established the association between NSAID use and small intestinal mucosal damage[7,14,15]. This study systematically analyzed the risk factors and clinical endoscopic characteristics of mucosal injury in NSAID users. By rigorously excluding patients with confounding factors - such as inflammatory bowel disease and tuberculosis - we identified key risk factors associated with SIMI and developed a predictive model.

Consistent with previous clinical trials indicating that 55% to 75% of chronic NSAID users exhibit mucosal injury[16,17], our study also found a higher incidence of small intestinal lesions among NSAID-exposed patients. Notably, in agreement with the findings reported by Ehrhard et al[18], small intestinal lesions in NSAID users predominantly presented as erythema, erosions, and ulcers, and occurred more frequently in the jejunum. This pattern may be attributed to the richer villous architecture, larger absorptive surface area, and higher local concentrations of drugs and bile acids in the jejunal mucosa.

Multivariate analysis identified six independent risk factors: Advanced age, smoking, PPI use, elevated BMI, higher TG levels, and increased LDL. Previous studies have indicated that advanced age is associated with diminished mucosal repair capacity and increased susceptibility to NSAIDs[19,20].

Cigarette-derived reactive oxygen species may directly damage gastrointestinal mucosa, compromising small intestinal barrier integrity, while smoking-induced gut microbiota alterations could indirectly impair small intestinal epithelium[21,22]. The association between PPI use and SIMI is particularly noteworthy. Although PPIs effectively reduce upper gastrointestinal damage, they may alter gut microbiota and increase intestinal permeability, thereby potentially exacerbating NSAID-related lower gastrointestinal injury[23]. However, the potential for confounding by indication must be considered. Specifically, patients with a higher baseline risk for gastrointestinal complications (e.g., those with a history of ulcers) are more likely to be co-prescribed NSAIDs and PPIs. While our multivariate model adjusted for PPI use as a covariate, the inherent channeling of PPI to higher-risk patients may introduce residual confounding. The relationship between PPI co-therapy and SIMI risk warrants further investigation in prospective studies designed to better control for this confounding.

Elevated BMI demonstrated significant association with SIMI in our study, aligning with Sun et al’s findings[21] identifying obesity as a SIMI risk factor. Pathophysiologically, obesity may indirectly contribute to SIMI through gut microbiota alterations, characterized by increased Firmicutes/Prevotella ratios and decreased beneficial bacteria (Bacteroides, Bifidobacterium, Lactobacillus), which reduce short-chain fatty acid production critical for epithelial barrier integrity[24]. Additionally, adipose-derived proinflammatory mediators (leptin, tumor necrosis factor-α) may exacerbate NSAID-induced mucosal damage[25,26]. Obese patients may also develop intestinal barrier dysfunction through downregulation of tight junction proteins[27], which may exacerbate NSAID-induced epithelial apoptosis and bacterial translocation. Hyperlipidemia is associated with intestinal barrier dysfunction and gut microbiota dysbiosis[28]. In rat models, hyperlipidemia has been shown to induce intestinal barrier injury, accompanied by downregulated expression of tight junction proteins ZO-1 and occludin, as well as reduced villus height and thinning of the muscular layer in the small intestine[29]. Furthermore, studies have demonstrated significant alterations in the gut microbiota of hyperlipidemic mice, which subsequently compromise the intestinal mucosal barrier[30]. Additionally, elevated levels of deoxycholic acid resulting from a high-fat diet can impair intestinal stem cells by reducing interleukin-22 expression in the ileum[31]. Moreover, studies have indicated that the lipid-lowering agent fenofibrate can ameliorate intestinal injury and promote mucosal recovery in rat models[32]. Similarly, atorvastatin has been shown to alleviate radiation-induced intestinal damage through activation of autophagy and exertion of antioxidant effects[33]. These findings underscore the necessity of weight and lipid management in high-risk populations.

The nomogram developed in this study demonstrated good discriminatory ability, with an AUC of 0.775 in the derivation cohort and 0.797 in the validation cohort. These results indicate that the model exhibits satisfactory predictive accuracy and generalizability. Furthermore, bootstrap validation revealed excellent calibration performance, further supporting its potential for clinical application.

A key strength of this model lies in its reliance on readily available clinical parameters - including demographic characteristics, lifestyle habits, routine laboratory tests, and medication history. This facilitates the early identification of high-risk patients prior to initiating long-term NSAID therapy. For such individuals, clinicians may consider implementing preventive strategies.

In recent years, machine learning-based artificial intelligence technologies for medical image recognition have advanced significantly[34]. Building on the clinical risk factors and practical model identified in our study, artificial intelligence presents a transformative avenue for future research. For instance, deep learning algorithms could be applied to automatically analyze entire capsule endoscopy videos, capturing subtle, sub-visual mucosal features imperceptible to the human eye. Furthermore, multi-modal data fusion can integrate these imaging phenotypes with clinical and microbiome data to construct next-generation, personalized prediction tools, promising a deeper understanding of NSAID-induced enteropathy and enabling precise risk stratification.

This study has inherent limitations due to its retrospective design, including potential selection bias and unmeasured confounders such as dietary habits. Furthermore, the relatively small sample size of NSAID users, particularly in the validation cohort, may affect the model's stability and the reliability of the validation results, which also precluded detailed comparisons of different NSAID subtypes’ enterotoxic effects and limited statistical power to detect rare outcomes (e.g., strictures or perforations). Moreover, the model’s performance lacks external validation, and its generalizability to other populations requires further confirmation in future independent studies. The absence of biomarker data (e.g., fecal calprotectin) also restricted our ability to correlate endoscopic findings with subclinical inflammation or barrier dysfunction. Future prospective studies should incorporate multi-omics approaches (metagenomics, metabolomics) to investigate the mechanistic interplay between gut microbiota and NSAIDs, thereby identifying novel therapeutic targets.

CONCLUSION

In summary, we identified key risk factors and constructed a reliable predictive model for SIMI in NSAID users. This tool could assist clinicians in identifying high-risk individuals who may benefit from closer monitoring, alternative therapies, or prophylactic measures.