Anti-TNF therapy attenuates the diagnostic utility of leucine-rich α2-Glycoprotein 1 for endoscopic activity in ulcerative colitis Authors: Akihiro Dejima, Hajime Takatori, Masaki Miyazawa, Hirofumi Okafuji, Tomoyuki Hayashi, Akihiro Seki, Kouki Nio, Noriho Iida, Shinya Yamada, Tetsuro Shimakami, Tadashi Toyama, Taro Yamashita Published: 2026 ABSTRACT BackgroundSerum leucine-rich α2-glycoprotein1 (LRG1) is a promising biomarker for mucosal inflammation in ulcerative colitis (UC). However, because systemic LRG1 expression is partly tumor necrosis factor-α (TNF-α)-dependent, anti-TNF therapy may pharmacologically suppress its levels independently of mucosal status.ObjectivesTo evaluate whether anti-TNF therapy is associated with false-negative serum LRG1 results, attenuates diagnostic accuracy for endoscopic activity (MES ≧ 2), and impairs diagnostic granularity to differentiate strict healing targets (MES 0 vs. 1) compared to other therapies.DesignA retrospective, single-center diagnostic accuracy study conducted in accordance with the STARD 2015 guidelines.MethodsWe analyzed 127 paired colonoscopy–LRG1 evaluations (49 under anti-TNF, 78 without) from 50 patients receiving advanced therapies (April 2021–March 2025). Endoscopic activity was defined as Mayo Endoscopic Subscore (MES) ≥2. Diagnostic performance was evaluated using ROC analysis; AUC differences were compared using a cluster bootstrap approach. Factors associated with false-negative LRG1 (<16 μg/mL despite MES ≥2) were identified using generalized estimating equations.ResultsOverall, LRG1 demonstrated satisfactory discrimination (AUC 0.81), but performed significantly lower under anti-TNF than non-anti-TNF therapy (AUC 0.44 vs. 0.91; ΔAUC=0.46). Among active evaluations, 45% (24/53) showed false-negatives. Anti-TNF was a strong independent predictor of false-negatives (OR 9.3, 95% CI 1.68–51.9), while JAK inhibitor use was protective (OR 0.09, 95% CI 0.01–0.80, P<0.03) with high accuracy (AUC 0.95 vs. 0.75, P<0.01). LRG1 failed to differentiate MES 0 and MES 1 in the anti-TNF cohort (p=0.63), whereas a marginal trend was observed in the non-anti-TNF cohort (p=0.06).ConclusionAnti-TNF therapy may selectively attenuate serum LRG1 performance, potentially causing frequent false-negatives and obscuring the distinction between complete healing (MES 0) and mild residual inflammation (MES 1). Clinicians must consider the therapeutic context when interpreting LRG1 for disease monitoring. Introduction Ulcerative colitis (UC) is a chronic inflammatory bowel disease characterized by continuous mucosal inflammation of the colon. Over the past two decades, treatment strategies for UC have evolved substantially with the introduction of biologic agents and small-molecule therapies. Current therapeutic paradigms emphasize treat-to-target approaches aimed at achieving objective endpoints such as endoscopic remission, which is associated with improved long-term outcomes including reduced hospitalization, colectomy, and colorectal cancer risk.1–3 Although endoscopy remains the gold standard for assessing mucosal inflammation, repeated endoscopic examinations are invasive, costly, and burdensome for patients. Consequently, reliable non-invasive biomarkers that accurately reflect mucosal inflammation are increasingly important for disease monitoring in clinical practice. C-reactive protein (CRP) is the most widely used blood biomarker in inflammatory bowel disease. However, CRP levels frequently fail to correlate with endoscopic activity in UC, and many patients with active mucosal inflammation exhibit normal CRP levels. 4 Therefore, alternative biomarkers capable of reflecting intestinal inflammation more accurately are needed. Leucine-rich α2-glycoprotein 1 (LRG1) has recently emerged as a promising biomarker for inflammatory diseases including inflammatory bowel disease. LRG1 is a 50-kDa glycoprotein produced by hepatocytes, neutrophils, macrophages, and intestinal epithelial cells. 5 Unlike CRP, whose production is largely dependent on interleukin-6 signaling, LRG1 expression is induced by multiple inflammatory cytokines including interleukin-1β, interleukin-22, and tumor necrosis factor-α (TNF-α). Previous studies have demonstrated that LRG1 correlates with endoscopic disease activity in UC and may reflect mucosal inflammation more accurately than conventional biomarkers.6,7 However, an important mechanistic question remains unresolved. Because TNF signaling contributes to the induction of LRG1 expression, pharmacological blockade of TNF may alter circulating LRG1 levels independently of mucosal inflammation. In patients receiving anti-TNF therapy, suppression of TNF-mediated inflammatory pathways could potentially reduce systemic LRG1 production despite persistent intestinal inflammation. Given the widespread use of anti-TNF agents for the treatment of moderate-to-severe UC, understanding the impact of anti-TNF therapy on LRG1 performance is clinically important. If anti-TNF therapy attenuates the relationship between LRG1 levels and mucosal inflammation, LRG1 may underestimate disease activity in this treatment context. Therefore, the aim of the present study was to investigate the frequency and determinants of false-negative LRG1 results in UC and to evaluate whether anti-TNF therapy potentially modifies the diagnostic performance of LRG1 for detecting endoscopic disease activity. Methods This retrospective single-center study was conducted at Kanazawa University Hospital over a 4-year duration between April 2021 and March 2025, in accordance with the Declaration of Helsinki and the ethical guidelines for medical research involving human subjects. The study protocol was approved by the institutional review board (No.1935-11). The reporting of this diagnostic accuracy study was prepared, structured, and conforms to the Standards for Reporting Diagnostic Accuracy (STARD) 2015 guidelines. 8 The completed STARD 2015 checklist has been provided as Supplementary Material. Detailed statements regarding the institutional ethical approval and the patient opt-out process are provided in the ‘Declarations’ section at the end of the manuscript. Consecutive patients with steroid-resistant or steroid-dependent UC who initiated advanced therapies at our institution during this designated study period were retrospectively identified. Eligible therapies included anti-TNF agents, integrin inhibitors, interleukin-12/23 inhibitors, anti-IL-23 p19-specific inhibitors (mirikizumab, risankizumab), Janus kinase inhibitors, calcineurin inhibitors, and tacrolimus. Sphingosine-1-phosphate receptor modulators and other unapproved agents were not included due to non-availability during the enrollment period. Cases with concomitant use of advanced therapies or those in the transitional phase of switching targeted agents were permitted, provided that the treatment status at the exact time of colonoscopy and LRG1 measurement was clearly documented. To be included in the final analysis, patients were required to have paired evaluations consisting of both a colonoscopic examination and a serum LRG1 measurement performed during their post-treatment follow-up period. Specific exclusion criteria were strictly applied to ensure data integrity. Evaluations were excluded if the interval between colonoscopy and LRG1 measurement exceeded one month, if any treatment modifications or dosage adjustments occurred between these two assessments, or if the patient was under 18 years of age. Furthermore, patients with incomplete clinical or laboratory records at the time of evaluation were excluded from the analysis. Endoscopic disease activity was assessed using the Mayo Endoscopic Subscore (MES) 9 and the Ulcerative Colitis Endoscopic Index of Severity (UCEIS) 10 (Supplement). For the MES, endoscopic remission was defined as a score of 0 or 1, and active disease as a score of ≧2. Similarly, for the UCEIS, remission was defined as a score of 0 or 1, and active disease as a score of ≧2. Both endoscopic scores were independently evaluated by two expert endoscopists who were blinded to the LRG1 levels and other biomarker results. Any disagreements between the two evaluators were resolved by consensus. Serum LRG1 levels were measured using a latex agglutination immunoassay (Nanopia LRG; Sekisui Medical Co., Ltd., Tokyo, Japan). CRP levels were measured simultaneously as conventional laboratory biomarkers. For diagnostic performance evaluations, CRP was analyzed both as a continuous variable and as a categorical variable dichotomized at the institution’s clinical reference value of 0.14 mg/dL (with values <0.14 mg/dL considered normal and ≥0.14 mg/dL considered elevated). Continuous variables are presented as medians with ranges, and categorical variables are expressed as percentages. Differences between groups were assessed using the Mann–Whitney U test or Student’s t-test as appropriate. The diagnostic performance of LRG1, CRP for detecting endoscopic activity was evaluated using receiver operating characteristic (ROC) curve analysis. Area under the curve (AUC) values were calculated for the overall cohort and separately for evaluations performed with and without anti-TNF therapy. To account for repeated measurements within patients, the difference in AUC between anti-TNF and non-anti-TNF evaluations was estimated using a cluster bootstrap approach with resampling at the patient level (2000 iterations). To evaluate whether other molecular targeted therapies also impact LRG1 performance, subgroup ROC analyses were performed for JAK inhibitors, IL-23 inhibitors, and integrin inhibitors. The difference in AUC between users and non-users for each drug class was compared using a cluster bootstrap approach (2000 iterations). False-negative LRG1 results were defined as LRG1 levels <16 μg/mL despite active endoscopic inflammation (MES ≥2). Factors associated with false-negative LRG1 results were evaluated using generalized estimating equations (GEE) with an exchangeable correlation structure to account for repeated observations within patients. All statistical analyses were performed using EZR software (version 1.54), 11 which is a graphical user interface for R (The R Foundation for Statistical Computing). Specifically, the ‘geepack’ package was utilized for generalized estimating equations (GEE) modeling, and customized R scripts were applied to perform the cluster bootstrap resampling (2000 iterations) at the patient level. A two-sided P value <0.05 was considered statistically significant. To ensure the scientific rigor and validity of our findings, several strategies were implemented to minimize potential biases inherent in this retrospective study. First, to mitigate selection bias, we identified all consecutive patients who met the inclusion criteria at Kanazawa University Hospital throughout the designated study period, ensuring a representative sample within a setting of standardized clinical protocols. To address information bias and the risk of misclassification due to fluctuations in disease activity, we strictly limited the temporal gap between clinical assessments as described above. This ensured that the biomarker levels directly reflected the mucosal status captured during the endoscopic evaluation. Furthermore, to reduce assessment bias, endoscopic activity was graded using the MES, a widely validated and standardized scoring system, while serum LRG1 levels were measured using a consistent latex agglutination immunoassay to maintain objectivity and reproducibility across the cohort. Finally, to correct for potential statistical bias arising from the clustering effect of repeated measurements within the same individuals, we employed advanced analytical techniques. We utilized GEE for regression modeling and a cluster bootstrap approach with 2000 iterations for the comparison of AUC values. These methods accounted for the non-independence of intra-patient observations, thereby providing less biased and more appropriate statistical estimates than conventional analysis. Methodologically, our sample size of 50 patient clusters with 127 clustered evaluations is fully adequate to avoid biased standard errors and ensure mathematical stability in both the GEE regression models and cluster bootstrap distributions. Because this study utilized a fixed historical cohort, no formal a priori power calculation was performed. However, the final sample size of 50 patient clusters (encompassing 127 total evaluations) was methodologically justified and sufficient to allow for advanced statistical modeling. In longitudinal data analysis, a sample size of 30 to 50 independent clusters is widely recognized as the standard threshold required to ensure the asymptotic stability of generalized estimating equations (GEE) and cluster bootstrap distributions while preventing biased standard errors. This sample size successfully enabled exploratory subgroup analyses, particularly for the comparison between anti-TNF and non-anti-TNF groups, and provided acceptable statistical estimates while accounting for intra-patient correlations. To facilitate clinical interpretation and define diagnostic categories, certain quantitative variables were categorized into binary groups. Most notably, serum LRG1 levels were dichotomized using a cutoff value of 16.0 μg/mL. This specific threshold was chosen based on previously established clinical guidelines and large-scale validation studies which identified 16.0 μg/mL as the optimal cutoff for detecting endoscopic activity in patients with UC6,7. Furthermore, endoscopic findings were grouped into “remission” (MES 0 and 1) and “active disease” (MES ≥2) to evaluate the discriminatory power of the biomarkers in a clinically relevant manner. There were no missing data for the primary variables of interest, including endoscopic scores, serum LRG1 levels, and types of molecular targeted therapies, as the study population was restricted to patients with complete clinical and laboratory records. Consequently, no statistical imputation methods were required, and all identified eligible cases were included in the final analysis. Patients were only included if they had complete data at the time of evaluation; therefore, cases that might have been lost to follow-up in a long-term longitudinal sense did not affect the integrity of the cross-sectional diagnostic performance analysis. All patients identified as meeting the inclusion criteria had available electronic medical records covering the required variables, ensuring that no cases were excluded due to attrition after the defined clinical assessment point. Sensitivity analyses were not performed because the inclusion criteria were strictly defined to ensure data consistency. Specifically, the exclusion of cases with treatment modifications between colonoscopy and LRG1 measurement ensured that the biomarker values directly reflected the mucosal status, reducing the potential confounding that would typically warrant additional sensitivity testing. Results A total of 128 patients who received advanced therapies during the study period were screened. Among 214 colonoscopy–LRG1 paired evaluations, 64 evaluations were excluded because serum LRG measurements were not obtained within one month of colonoscopy and 23 evaluations were excluded because age was under 18. Ultimately, 50 patients with 127 clinical evaluations were included in the final analysis (Figure 1). There were 49 evaluations performed under anti-TNF therapy and 78 evaluations without anti-TNF therapy.Figure 1.Flow diagram of participant selection. A total of 128 patients with 214 paired colonoscopy–LRG1 evaluations were initially screened. After excluding 64 evaluations due to an interval of >1 month between LRG1 measurement and colonoscopy, and 23 evaluations due to age under 18, a final cohort of 50 patients with 127 evaluations was included. Flow diagram of participant selection. A total of 128 patients with 214 paired colonoscopy–LRG1 evaluations were initially screened. After excluding 64 evaluations due to an interval of >1 month between LRG1 measurement and colonoscopy, and 23 evaluations due to age under 18, a final cohort of 50 patients with 127 evaluations was included. The baseline characteristics of the evaluations are shown in Table 1. In overall groups, the median age was 44 years (range 18–74), and 73% of evaluations were performed in male patients. The median disease duration was eight years. Pancolitis was present in 60% of evaluations. Multiple advanced therapeutic agents were used, reflecting the refractory nature of the study population. Notably, three evaluations in the anti-TNF cohort involved the concomitant use of tacrolimus, which was prescribed either as a short-term combination strategy for refractory status or during the transitional phase of switching between these advanced therapies.Table 1.Patient background.​Anti-TNF user n=49Non-user n=78Overall n=127Age, median (range)49(19-74)42(18-74)44(18-74)Sex, male: female27:2266:1293:34Type: left-sided colitis: total colitis19:3032:4651:76Steroid resistance/Steroid dependency24/2542/3666/61Duration of disease: years, median (range)11.0(0-25)5.5(0-29)8.0 (0-29)Drugs: IFX:ADA: GLM22:21:60:0:022:21:6UST: MRKZ:RZB0:022:7:322:7:3VDZ: CGM0:019:019:0Tofa: UPD: FIL0:0:013:7:113:7:1Tac31215MES score, median (range)0(0-3)1(0-3)1 (0-3)MES score = 0 or 1 (%)33(67)41(53)74 (58)LRG1, median (range)12.9(8.5-17.3)10.4(6.8-38.6)12.3 (6.8-38.6)CRP, median (range)0.06(0.02-0.35)0.06(0.02-4.57)0.06 (0.02-4.57)IFX=infliximab, ADA=adalimumab, GLM=golimumab, UST=Ustekinumab, MRKZ=Mirikizumab, RZB=Risankizumab, VDZ=vedolizumab, Tofa=tofacitinib,CGM=carotegrast methyl, UPD=Upadacitinib, FIL=Filgotinib, Tac=tacrolimus. MES=Mayo endoscopic subscore, LRG1=Leucine-rich α2 glycoprotein1, CRP=C-reactive protein. The three cases categorized under ‘Tac’ within the Anti-TNF user column represent evaluations performed during concomitant therapy or the treatment-switching phase between anti-TNF agents and tacrolimus. Table 1.Patient background.​Anti-TNF user n=49Non-user n=78Overall n=127Age, median (range)49(19-74)42(18-74)44(18-74)Sex, male: female27:2266:1293:34Type: left-sided colitis: total colitis19:3032:4651:76Steroid resistance/Steroid dependency24/2542/3666/61Duration of disease: years, median (range)11.0(0-25)5.5(0-29)8.0 (0-29)Drugs: IFX:ADA: GLM22:21:60:0:022:21:6UST: MRKZ:RZB0:022:7:322:7:3VDZ: CGM0:019:019:0Tofa: UPD: FIL0:0:013:7:113:7:1Tac31215MES score, median (range)0(0-3)1(0-3)1 (0-3)MES score = 0 or 1 (%)33(67)41(53)74 (58)LRG1, median (range)12.9(8.5-17.3)10.4(6.8-38.6)12.3 (6.8-38.6)CRP, median (range)0.06(0.02-0.35)0.06(0.02-4.57)0.06 (0.02-4.57)IFX=infliximab, ADA=adalimumab, GLM=golimumab, UST=Ustekinumab, MRKZ=Mirikizumab, RZB=Risankizumab, VDZ=vedolizumab, Tofa=tofacitinib,CGM=carotegrast methyl, UPD=Upadacitinib, FIL=Filgotinib, Tac=tacrolimus. MES=Mayo endoscopic subscore, LRG1=Leucine-rich α2 glycoprotein1, CRP=C-reactive protein. The three cases categorized under ‘Tac’ within the Anti-TNF user column represent evaluations performed during concomitant therapy or the treatment-switching phase between anti-TNF agents and tacrolimus. Patient background. IFX=infliximab, ADA=adalimumab, GLM=golimumab, UST=Ustekinumab, MRKZ=Mirikizumab, RZB=Risankizumab, VDZ=vedolizumab, Tofa=tofacitinib,CGM=carotegrast methyl, UPD=Upadacitinib, FIL=Filgotinib, Tac=tacrolimus. MES=Mayo endoscopic subscore, LRG1=Leucine-rich α2 glycoprotein1, CRP=C-reactive protein. The three cases categorized under ‘Tac’ within the Anti-TNF user column represent evaluations performed during concomitant therapy or the treatment-switching phase between anti-TNF agents and tacrolimus. Serum LRG1 levels were significantly higher in patients with active endoscopic inflammation than in those in remission. Median LRG1 levels were 16.2 μg/mL in the active disease group compared with 10.6 μg/mL in the remission group (p < 0.01). CRP levels were also significantly higher in the non-remission group (0.12 vs. 0.06 mg/dL, p<0.01). Furthermore, to address the strict clinical targets of the STRIDE-2 consensus, we evaluated whether serum LRG1 levels could differentiate between complete mucosal healing (MES 0) and near-healing (MES 1). In the overall cohort, no significant difference was observed between the MES 0 and MES 1 subgroups (median 10.6 vs. 10.8 μg/mL, p = 0.26). However, when stratified by treatment type, patients in the non-anti-TNF cohort demonstrated a notable marginal trend toward higher LRG1 levels in the MES 1 subgroup compared to the MES 0 subgroup (median 9.10 vs. 9.25 μg/mL, p = 0.06). Conversely, this trend was entirely abolished in the anti-TNF cohort, showing no significant difference between the two subgroups (median 12.90 vs. 13.75 μg/mL, p = 0.63). Despite this association, a substantial proportion of patients with active endoscopic inflammation exhibited low LRG1 levels. Among the 53 evaluations with MES ≥2, 24 evaluations (45%) had LRG1 values below the clinical cutoff of 16 μg/mL and were classified as false-negative. Clinical factors associated with false-negative LRG1 results are shown in Table 2. Anti-TNF therapy was strongly associated with the occurrence of false-negative results. In the multivariate analysis using generalized estimating equations, anti-TNF therapy remained independently associated with false-negative LRG1 (OR 9.3, 95% CI 1.68–51.9), whereas JAK inhibitor use was significantly associated with a lower risk of false-negative results (OR 0.09, 95% CI 0.01–0.80, P < 0.03).Table 2.Clinical factors associated with false-negative LRG1 results in evaluations with active endoscopic disease (MES ≥2).VariableFalse-negative LRG1 (n=24)True-positive LRG1 (n=29)P valueMultivariate OR (95% CI)P valueAge, years (median, range)42 (18–63)42 (21–73)0.82——Male sex, n (%)15 (63%)22 (76%)0.37——Pancolitis, n (%)12 (50%)19 (66%)0.28——Disease duration, years (median,range)7.5(0-29)4(0-13)0.26——Prior biologic exposure, n (%)14 (58%)23 (79%)0.14——Anti-TNF therapy, n (%)14 (58%)2 (7%)<0.019.3 (1.68–51.9)0.01Integrin inhibitor,n (%)4(17%)2(7%)0.39​​IL-23 inhibitor,n (%)4(17%)6(21%)1.00​​JAK inhibitor,n (%)1(4%)15(52%)<0.010.09(0.01-0.80)0.03Other advanced therapies, n (%)10 (42%)27 (93%)<0.01——LRG1, Leucine-rich α2-glycoprotein1; MES, Mayo Endoscopic Subscore; OR, odds ratio; CI, confidence interval. False-negative LRG was defined as serum LRG1 <16 μg/mL despite endoscopic activity (MES ≥2). Multivariate analysis was performed using a generalized estimating equation model to account for repeated measurements within patients. Table 2.Clinical factors associated with false-negative LRG1 results in evaluations with active endoscopic disease (MES ≥2).VariableFalse-negative LRG1 (n=24)True-positive LRG1 (n=29)P valueMultivariate OR (95% CI)P valueAge, years (median, range)42 (18–63)42 (21–73)0.82——Male sex, n (%)15 (63%)22 (76%)0.37——Pancolitis, n (%)12 (50%)19 (66%)0.28——Disease duration, years (median,range)7.5(0-29)4(0-13)0.26——Prior biologic exposure, n (%)14 (58%)23 (79%)0.14——Anti-TNF therapy, n (%)14 (58%)2 (7%)<0.019.3 (1.68–51.9)0.01Integrin inhibitor,n (%)4(17%)2(7%)0.39​​IL-23 inhibitor,n (%)4(17%)6(21%)1.00​​JAK inhibitor,n (%)1(4%)15(52%)<0.010.09(0.01-0.80)0.03Other advanced therapies, n (%)10 (42%)27 (93%)<0.01——LRG1, Leucine-rich α2-glycoprotein1; MES, Mayo Endoscopic Subscore; OR, odds ratio; CI, confidence interval. False-negative LRG was defined as serum LRG1 <16 μg/mL despite endoscopic activity (MES ≥2). Multivariate analysis was performed using a generalized estimating equation model to account for repeated measurements within patients. Clinical factors associated with false-negative LRG1 results in evaluations with active endoscopic disease (MES ≥2). LRG1, Leucine-rich α2-glycoprotein1; MES, Mayo Endoscopic Subscore; OR, odds ratio; CI, confidence interval. False-negative LRG was defined as serum LRG1 <16 μg/mL despite endoscopic activity (MES ≥2). Multivariate analysis was performed using a generalized estimating equation model to account for repeated measurements within patients. Receiver operating characteristic analysis demonstrated moderate overall discrimination of LRG1 for detecting endoscopic activity (AUC 0.81), CRP for detecting endoscopic activity (AUC 0.70). (Figure 2). However, diagnostic performance differed markedly according to treatment exposure. Similar findings were observed in the analysis using the UCEIS (Supplementary Figure S1).Figure 2.Diagnostic performance of serum biomarkers for endoscopic activity. Receiver operating characteristic (ROC) curves were used to evaluate the ability of serum biomarkers to detect endoscopic activity defined by Mayo Endoscopic Subscore ≥2. Diagnostic performance of serum biomarkers for endoscopic activity. Receiver operating characteristic (ROC) curves were used to evaluate the ability of serum biomarkers to detect endoscopic activity defined by Mayo Endoscopic Subscore ≥2. The distribution of serum LRG1 levels among the four groups is shown in Figure 3. In the Non-anti-TNF user groups, the median LRG1 level was higher in the active group than in the remission group. However, in the anti-TNF user groups, there was no marked difference in median LRG1 levels between the remission and active groups. When applying the cut-off value of 16 μg/mL (indicated by the red dashed line), a larger proportion of patients in the Non-anti-TNF user active group exceeded this threshold compared to the other groups. Notably, while the median LRG1 level in the Non-anti-TNF active group was above the cut-off, a substantial number of patients in the anti-TNF user active group remained below 16 μg/mL. This observation suggests that serum LRG1 levels may be suppressed in patients receiving TNF inhibitors, even during active disease states. Similar findings were observed in the analysis using the UCEIS (Supplementary Figure S2).Figure 3.Serum LRG 1 levels in patients with and without TNF inhibitors. Box-and-whisker plots show serum leucine-rich α2-glycoprotein1 (LRG1) concentrations (μg/mL) in four groups: non-anti-TNF remission, non-anti-TNF active, anti-TNF remission, and anti-TNF active. The horizontal red dashed line indicates the clinical cut-off value of 16 μg/mL. Boxes represent the interquartile range, horizontal lines indicate the median, and whiskers extend to 1.5 times the interquartile range. In the non-anti-TNF user group, the median LRG1 level was higher in the active disease group than in the remission group. However, in the anti-TNF user group, no marked difference in median LRG1 levels was observed between the two groups, indicating frequent false-negative LRG1 results under anti-TNF therapy. Serum LRG 1 levels in patients with and without TNF inhibitors. Box-and-whisker plots show serum leucine-rich α2-glycoprotein1 (LRG1) concentrations (μg/mL) in four groups: non-anti-TNF remission, non-anti-TNF active, anti-TNF remission, and anti-TNF active. The horizontal red dashed line indicates the clinical cut-off value of 16 μg/mL. Boxes represent the interquartile range, horizontal lines indicate the median, and whiskers extend to 1.5 times the interquartile range. In the non-anti-TNF user group, the median LRG1 level was higher in the active disease group than in the remission group. However, in the anti-TNF user group, no marked difference in median LRG1 levels was observed between the two groups, indicating frequent false-negative LRG1 results under anti-TNF therapy. In evaluations performed without anti-TNF therapy, LRG1 demonstrated good discrimination for endoscopic activity (AUC 0.91). In contrast, in evaluations performed under anti-TNF therapy, discrimination was markedly attenuated (AUC 0.44) (Figure 4). Similar findings were observed in the analysis using the UCEIS (Supplementary Figure S3).Figure 4.Impact of anti-TNF therapy on the diagnostic performance of serum biomarkers. ROC curves were stratified according to anti-TNF therapy. A cluster bootstrap method was used to estimate differences in AUC. Anti-TNF therapy significantly reduced the diagnostic accuracy of LRG1 for detecting endoscopic activity (ΔAUC = 0.46, 95% CI 0.13–0.54). Impact of anti-TNF therapy on the diagnostic performance of serum biomarkers. ROC curves were stratified according to anti-TNF therapy. A cluster bootstrap method was used to estimate differences in AUC. Anti-TNF therapy significantly reduced the diagnostic accuracy of LRG1 for detecting endoscopic activity (ΔAUC = 0.46, 95% CI 0.13–0.54). The diagnostic performance of LRG 1 was significantly lower in the anti-TNF group than in the non-anti-TNF group (ΔAUC = 0.46, 95%CI 0.13–0.54). To further investigate whether the attenuation of LRG1 performance was specific to anti-TNF therapy, we analyzed its diagnostic accuracy across other drug classes (Supplementary Figure S4). In contrast to anti-TNF therapy, JAK inhibitor use was associated with significantly higher LRG1 performance compared to non-users (AUC 0.95 vs. 0.75; ΔAUC = 0.20, P < 0.01). No significant differences in AUC were observed for IL-23 inhibitors (AUC 0.88 vs. 0.77, P = 0.35) or integrin inhibitors (AUC 0.67 vs. 0.82, P = 0.30). Discussion To our knowledge, this is the first study to specifically quantify the frequency of false-negative LRG1 results under anti-TNF therapy in patients with UC. Although LRG1 demonstrated good overall performance for detecting endoscopic activity, the vast majority (88%) of active evaluations performed under anti-TNF therapy showed LRG1 levels below the established clinical cutoff. Furthermore, anti-TNF therapy emerged as the strongest independent predictor of false-negative LRG1 results and may markedly attenuate the diagnostic performance of LRG1 for detecting active mucosal inflammation. These findings fill a critical gap in the literature by suggesting how systemic cytokine-targeted therapies can decouple a blood-based biomarker from local intestinal tissue inflammation, holding major implications for non-invasive treatment monitoring. The biological mechanism underlying this observation is likely related to the regulation of LRG1 expression. LRG1 production is induced by several inflammatory cytokines, including TNF-α, IL-1β, and IL-22. Pharmacological blockade of TNF-α may therefore suppress circulating LRG1 levels despite the persistence of mucosal inflammation. 12 Recently, Hayasaka et al. reported that several therapeutic agents, particularly anti-TNF agents, were associated with significantly lower serum LRG1 levels even among patients with endoscopically remitted UC. 13 Their findings suggested that LRG1 values are influenced not only by inflammatory activity but also by the mechanism of action of concomitant therapies. However, because their analysis was restricted to patients in endoscopic remission, the clinical implications of these treatment-associated changes in LRG1 remained uncertain. Our study extends these observations and adds several crucial insights to the current literature. First, while Hayasaka et al. 13 hypothesized a general treatment-associated effect based solely on a remission cohort, our data suggest that anti-TNF therapy not only lowers serum LRG1 levels but also substantially impairs the diagnostic performance of LRG1 for detecting active mucosal inflammation. Indeed, 14 of 16 (88%) of active evaluations under anti-TNF therapy showed LRG1 values below the established clinical cutoff, resulting in frequent false-negative findings. Furthermore, anti-TNF therapy was identified as a strong independent predictor of false-negative LRG1 results. Importantly, our multivariate analysis using GEE (Table 2) highlighted that this confounding impact on LRG1 kinetics is not a uniform feature shared by all advanced therapies. While specific pharmacological blockade of TNF-α strongly induced false-negative results, the use of JAK inhibitors was inversely and significantly associated with false-negative risk (OR 0.09), indicating that broader immune cascade modifications preserve the diagnostic utility of LRG1. Our additional class-specific ROC analysis (Supplementary Figure S4) further supported this pharmacological divergence. Specifically, the diagnostic performance of LRG1 remained favorable, or was even enhanced, under JAK inhibitor therapy (AUC 0.95), while no detrimental impact was observed under IL-23 or integrin inhibitors. These findings suggest that the suppression of circulating LRG1 and the resulting loss of its diagnostic granularity for mucosal inflammation may be specifically associated with the pharmacological blockade of TNF-α. This might be explained by the fact that LRG1 expression is multi-cytokine dependent; while JAK inhibitors and IL-23 inhibitors target specific signaling nodes, they may leave alternative LRG1-inducing pathways intact enough to reflect mucosal activity, whereas TNF blockade appears to blind the biomarker more comprehensively in the context of UC monitoring. Together, these findings suggest that suppression of TNF-dependent LRG1 production may lead to clinically meaningful underestimation of disease activity in patients receiving anti-TNF agents. These findings have important implications for treat-to-target strategies in UC. According to the STRIDE-2 consensus, the ultimate endoscopic target is endoscopic healing; however, recent evidence underscores a critical prognostic distinction between complete endoscopic healing (MES 0) and near-healing (MES 1), with the latter carrying a significantly higher risk of subsequent clinical relapse. Our further sub-analysis regarding these strict targets yielded notable insights. In patients not exposed to anti-TNF therapy, serum LRG1 levels demonstrated a marginal trend toward identifying subtle residual inflammation, with higher values in MES 1 than in MES 0 (p = 0.06). Remarkably, this granular differentiation was not observed in patients undergoing anti-TNF therapy (p = 0.63). This biological dampening suggests that systemic pharmacological blockade of TNF-α may suppress the subtle systemic shedding of LRG1 caused by minor local tissue inflammation. Therefore, LRG1 is increasingly used as a noninvasive biomarker to guide disease monitoring; however, our data suggest that low LRG1 values should be interpreted cautiously, especially because anti-TNF therapy may confound the biomarker regarding the critical clinical distinction between MES 0 and MES 1. In this setting, reliance on LRG1 alone may fail to confirm true deep mucosal healing, potentially masking ongoing minor endoscopic activity, and additional evaluation using endoscopy or complementary, gut-specific biomarkers such as fecal calprotectin (FC) may be required. Several limitations should be acknowledged. First, this was a retrospective single-center study with a relatively small sample size. Consequently, the exact effect sizes, including the reported odds ratios and diagnostic accuracy differences, should be interpreted with caution and require validation in larger cohorts. Second, the study population consisted primarily of patients receiving advanced therapies at a tertiary referral center, which may limit generalizability. Third, we could not evaluate or compare the performance of LRG1 with FC. As a stool-based, gut-specific biomarker, FC directly reflects intestinal neutrophilic inflammation and is less influenced by systemic cytokine pathways or specific systemic molecular targeted therapies than blood-based biomarkers. However, because FC was not routinely or simultaneously measured for the majority of our cohort during the study period, a meaningful comparison could not be performed. Given that CRP is inherently less sensitive in UC than in Crohn’s disease—as also observed in our study where the median CRP in the active group was only 0.12 mg/dL—future prospective studies comparing the diagnostic accuracy of LRG1 and FC under anti-TNF therapy are warranted. Fourth, this study did not evaluate intestinal histological activity. While serum LRG1 closely reflects neutrophilic infiltration, its precise resolution in detecting microscopic, low-grade histological activity—especially in patients who achieved endoscopic remission (MES 0 or 1)—remains undetermined. Fifth, we lacked data on intestinal ultrasound (IUS) activity. IUS has recently gained substantial recognition as a non-invasive modality to assess transmural inflammation. Given that anti-TNF therapy blunts serum LRG1 performance, evaluating whether a multi-modal non-invasive approach combining LRG1 and IUS can optimize disease monitoring under specific targeted therapies warrants further investigation in future prospective validation. Sixth, residual confounding cannot be completely excluded because treatment allocation was not randomized. Nevertheless, the consistency of the findings across both MES- and UCEIS-based analyses supports the reliability of the observed association. Conclusion Our findings suggest that anti-TNF therapy is potentially associated with a high frequency of false-negative LRG1 results and could substantially attenuate the ability of LRG1 to detect endoscopic activity in this cohort of patients with UC. Clinicians should therefore consider the specific treatment context when interpreting LRG1 values during disease monitoring, and further prospective, large-scale studies are warranted to confirm these observations. Supplemental material Supplemental material - Anti-TNF therapy attenuates the diagnostic utility of leucine-rich α2-Glycoprotein 1 for endoscopic activity in ulcerative colitis Supplemental material for Anti-TNF therapy attenuates the diagnostic utility of leucine-rich α2-Glycoprotein 1 for endoscopic activity in ulcerative colitis by Akihiro Dejima, Hajime Takatori, Masaki Miyazawa, Hirofumi Okafuji, Tomoyuki Hayashi, Akihiro Seki, Kouki Nio, Noriho Iida, Shinya Yamada, Tetsuro Shimakami, Tadashi Toyama and Taro Yamashita in Therapeutic Advances in Gastroenterology.