ABSTRACT
Background
Endoscopic and histologic evaluation is part of routine clinical practice to confirm the presence of active disease in ulcerative colitis. Faecal calprotectin (FCAL) levels show a positive correlation with endoscopic indices, relapse, and response to treatment. Existing studies have not clearly determined cut‐off values for endoscopic and histologic remissions.
Aims
Our study aims to determine an optimal FCAL threshold for discriminating between endoscopic and histologic active disease and remission in UC, based on the Mayo Endoscopic Score (MES) and the Geboes Histologic Score.
Methods
We performed a pooled analysis of retrospective and prospective studies of patients with UC from four academic centres. Key inclusion criteria were adult UC patients undergoing colonoscopy, with FCAL measurements. Receiver operating characteristic curves were computed on 80% of the data using 5‐fold cross‐validation to determine the optimal FCAL threshold for predicting active UC based on specificity (spec) and sensitivity. Optimal thresholds were then tested on the remaining 20% of the data. Active endoscopic disease was assessed using MES 0–1 versus 2–3 and MES 0 versus 1–3, and active histologic disease or remission was evaluated using GHS < 3.1 versus GHS ≥ 3.1 and GHS ≤ 2.0 versus GHS > 2.0.
Results
A total of 741 patients with UC were included in our analysis. For MES 0 versus 1‐2‐3, the AUC was 0.716 (95% CI: 0.701–0.731, p < 0.0001) with an optimal threshold of 154.3 μg/g (specificity: 0.69 [95% CI: 0.65–0.72], sensitivity: 0.64 [95% CI: 0.60–0.69]). For MES 0–1 versus 2–3, the AUC increased to 0.802 (95% CI: 0.797–0.807, p < 0.0001), with an optimal FCAL threshold of 234.6 μg/g (specificity: 0.74 [95% CI: 0.74–0.75], sensitivity: 0.69 [95% CI: 0.62–0.75]). For GHS ≤ 2.0 versus GHS > 2.0, the AUC was 0.656 (95% CI: 0.647–0.664, p < 0.0001), with an optimal threshold of 117.6 μg/g (specificity: 0.61 [95% CI: 0.56–0.66], sensitivity: 0.59 [95% CI: 0.56–0.63]). For GHS < 3.1 vs. GHS ≥ 3.1, the AUC was 0.752 (95% CI: 0.743–0.762, p < 0.0001), with an optimal threshold of 166.2 μg/g (specificity: 0.69 [95% CI: 0.66–0.73], sensitivity: 0.69 [95% CI: 0.63–0.76]). Given the high AUC for MES 0–1 vs. MES 2–3, we tested different FCAL thresholds from the literature and propose a threshold of 170 μg/g to maximize sensitivity (sensitivity: 0.801 [95% CI: 0.729–0.873], specificity: 0.648 [95% CI: 0.641–0.656], LR+ 2.311 [95% CI: 2.001–2.670], LR‐ 0.289 [95% CI: 0.196–0.427], accuracy 67.62% [95% CI: 65.95–69.30]) and limit the proportion of false negatives.
Conclusions
Our study demonstrates that FCAL can predict both active endoscopic and histological disease with acceptable sensitivities and specificities. The proposed cut‐off values will help guide clinical practice to achieve the recommended treatment outcomes. Further studies are warranted to validate our results.
In this pooled analysis of 741 ulcerative colitis patients across four academic centres, a faecal calprotectin threshold of 170 μg/g (AUC 0.802, sensitivity 80%) identifies endoscopically active disease, providing clinicians a validated, non‐invasive tool to determine when colonoscopy can be safely avoided (Some graphics are AI‐generated).
1. Introduction
Ulcerative colitis [1] is a chronic inflammatory disease characterized by relapsing and remitting phases. Periods of relapse may lead to uncontrolled colonic inflammation, leading to complications such as strictures, the need for surgical intervention or colectomy, as well as an increased risk of colorectal cancer [2, 3]. Treatment of UC revolves around attaining and maintaining remission. Numerous studies have demonstrated that achieving clinical remission alone is insufficient; attaining endoscopic remission or healing results in improved outcomes, reduced hospitalizations, and fewer surgical interventions [2, 4]. These findings prompted a committee of inflammatory bowel disease (IBD) specialists to create the Selecting Therapeutic Targets in Inflammatory Bowel Disease (STRIDE) program in 2015 [5]. Twelve expert recommendations were made to help establish treatment goals in a treat‐to‐target strategy in IBD. In STRIDE‐I, the primary targets for both ulcerative colitis [1] and Crohn's disease [6] were symptomatic and endoscopic remission, and the emerging role of histologic remission was highlighted in UC, but was not described as a target [5, 7]. The updated STRIDE‐II guidelines, which were published in 2021, introduced time‐dependent and treatment‐dependent targets. STRIDE‐II defined symptomatic (clinical) response as a short‐term target, symptomatic remission and biochemical (CRP and FCAL) normalization as intermediate targets, and endoscopic healing, normalization of QoL, and absence of disability as long‐term treatment targets [7]. Adjunctive goals included histological remission in UC [7]. The concept of histologic response has been increasingly recognized as a crucial therapeutic goal in UC clinical trials, often included as secondary or exploratory endpoints [8, 9]. Meta‐analyses from Gupta et al. and Yoon et al. have shown that patients in endoscopic remission (i.e., Mayo Endoscopic Score (MES) of 0) but with persistent histologic activity were two to three times more likely to relapse, respectively, compared to their counterparts who had achieved histological remission [10, 11].
Colonoscopy and biopsy remain the gold standard for monitoring endoscopic and histological disease activity in UC. However, there is interest in non‐invasive biomarkers that could predict disease activity without carrying the risk of recurrent colonoscopies [4]. Faecal calprotectin is a calcium‐bound protein released by neutrophils during inflammation, and its concentration in stool reflects intestinal mucosal inflammation [12]. High levels of faecal calprotectin have been associated with active inflammation in UC, making it a useful non‐invasive marker for assessing disease activity [13, 14]. Studies have shown that faecal calprotectin correlates well with endoscopic findings, with higher levels indicating more severe inflammation [15]. Existing single‐centre studies of UC patients are limited by the fact that they have not tested optimal FCAL thresholds in independent datasets leading to poor generalizability of the model to other data and overfitting causing a possible optimistic inflation of reported metrics [16]. In addition, these studies have proposed a wide range of FCAL cut‐off values for the assessment of active disease on endoscopy or histology [17, 18, 19, 20, 21]. For example, some studies have suggested thresholds of 100 μg/g [20], 150 μg/g [22], 170 μg/g [18], 192 μg/g [23] to discriminate between active and inactive endoscopic disease in patients with clinically‐remitted UC. The current American Association of Gastroenterology [24] guidelines on faecal calprotectin, published in 2023, suggest that for patients in clinical remission, a faecal calprotectin threshold of < 150 μg/g may be used to rule out active inflammation [22].
Indeed, no large‐scale, multi‐center studies have been conducted to determine FCAL thresholds that most accurately reflect endoscopic and histological evidence of disease. In addition, to the best of our knowledge, no studies have validated optimal FCAL thresholds in separate patient cohorts. In this study, we aimed to determine the optimal FCAL thresholds for predicting disease remission or activity in patients with UC, based on the Mayo score (endoscopy) and the Geboes score (histology).
2. Materials and Methods
2.1. Study Design
We conducted a systematic literature search of PubMed (between 1980 and March 31, 2021) for all relevant articles to evaluate the association and correlation of faecal calprotectin measurements with histological and endoscopic remission in ulcerative colitis. Key words in the search included a combination of “faecal calprotectin”, “ulcerative colitis”, and “remission”.
Studies were restricted to full‐text articles in English‐language verified journals that featured adult human participants. Two authors independently reviewed the titles and abstracts of studies identified in the primary search to exclude those that did not address the research question of interest. Full text review yielded nine studies, with data originating from eight different institutions. Case series, studies with insufficient data, studies lacking endoscopic or histological scores, and studies of UC patients who had undergone total colectomy were excluded. Corresponding authors of all eligible studies were contacted, and data were obtained from four of those institutions. We then conducted a pooled analysis of individual patient data from the databases of four centers: McGill University Health Center [25], Institute of Pharmacology and Therapeutics [26], Brooke Army Medical Center [25, 26], and Hospital Universitari de Bellvitge‐Institut d'Investigació Biomédica de Bellvitge [27]. Research ethics board (REB) approval was obtained from the McGill University Health Center REB committee 2022‐8271.
2.2. Patient Population
Included patients were adults with an official diagnosis of ulcerative colitis, currently in clinical remission. They had to have undergone a colonoscopy with an established Mayo score, a concurrent or recent FCAL measurement (within 1 week before or after the colonoscopy, excluding while on bowel preparation), and a histology assessment with a Geboes score. Patients who were scored exclusively with other scores, such as the Nancy or the Baron scores, were excluded from the specific analyses (either histologic or endoscopic) for homogeneity. Detailed inclusion and exclusion criteria for all included studies are provided in Table S1. NSAID exclusion was not systematically applied in the remaining three contributing datasets, which is acknowledged as a limitation (see Section 4).
2.3. Study Outcomes
The primary outcome was to determine the diagnostic performance of faecal calprotectin in identifying endoscopic remission and histological remission in patients with UC who are currently in clinical remission, and to validate these findings in held‐out test folds within cross‐validation. For the Mayo Endoscopic Score (MES), a score of 0 is considered endoscopic remission, and for the Geboes Histologic Score [28], a score of < 2 is considered histologic remission.
Specifically, we aimed to establish optimal cut‐off points for faecal calprotectin levels, calculate the sensitivity and specificity of FC for predicting both endoscopic and histological remission, and determine areas under the ROC curve (AUC) for FC as a predictor.
The secondary outcome was to determine FC performance in identifying endoscopic improvement (MES 1) and histological activity (GHS > 3.1). The MES was assigned based on the most severely affected segment identified during endoscopic evaluation; when multiple segments were assessed, the highest (worst) score was retained. Endoscopic assessments were performed by experienced IBD‐specialist endoscopists at each centre without central reading, and FCAL results were concealed from endoscopists where reported (Hart et al. and Guardiola et al. datasets). Procedures included full colonoscopy (Hart et al., Guardiola et al., Patel et al.) and flexible sigmoidoscopy (Magro et al.). Rectal biopsies were obtained in all patients. Additional biopsies were taken from other colonic segments at the endoscopist's discretion. The GHS was calculated for each biopsy specimen, and the highest (worst) score was retained for analysis. Histologic evaluation was performed locally by IBD‐expert pathologists at each contributing centre. FCAL results and endoscopic findings were concealed from pathologists in the Hart et al. and Guardiola et al. datasets; concealment was not formally reported for Patel et al. Interobserver reproducibility was not assessed and no formal calibration was performed across centres.
The following FCAL assay kits were used at each contributing centre: Hart et al. (McGill University, n = 185): Bühlmann ELISA; Guardiola et al. (Bellvitge Hospital, n = 59): Calprotectin Bühlmann ELISA; Patel et al. (Mt. Sinai Hospital, n = 68): ELISA‐based assay (manufacturer not specified in original publication); Magro et al. (Portuguese multicenter, n = 371): Quantum Blue (Bühlmann) and EliA (Phadia, ThermoFisher) used in parallel. Three of four contributing centers used Bühlmann‐based platforms. No cross‐assay harmonization or calibration was performed.
2.4. Statistical Analysis
FCAL levels were compared across diagnostic categories (MES 0 vs. MES 1–3, MES 0–1 vs. MES 2–3; GHS ≤ 2.0 vs. GHS > 2.0, GHS < 3.1 vs. GHS ≥ 3.1) using a two‐tailed unpaired Student's t‐test. Given the known skewed distribution of FCAL values, results are reported as both mean (±SD) and median (interquartile range [IQR]). Univariate regressions with FCAL as a continuous variable and the various diagnostic categories outlined above were also performed.
For threshold analyses, receiver operating characteristic [ROC] curve analyses were performed for FCAL levels using the MES and GHS diagnostic categories (MES 0 vs. MES 1–3, MES 0–1 vs. MES 2–3; GHS ≤ 2.0 vs. GHS > 2.0, GHS < 3.1 vs. GHS ≥ 3.1). Using a stratified k‐fold cross‐validation method, we split the data into five folds, with 20% of the data contained in each fold. Each fold has approximately an equal representation of both binary classes. We then considered four of the folds (80%) as the training set and the remaining fold (20%) for testing to assess prediction accuracy based on the selected threshold. This process was iterated 5 times (Tables S3–S5). The selection of thresholds was performed by optimizing sensitivity and specificity across each diagnostic category using the “closest to (0,1)” method, which selects the point on the ROC curve that is geographically nearest to the upper left corner of the graph. The Youden threshold selection method was also used. The mean threshold across the 5 folds was used to predict accuracy in the testing fold. Areas under the curve [AUC] were calculated for each ROC plot across the 5 folds, with the DeLong test used to compute p‐values. Positive and negative predictive values [PPV; NPV] and positive and negative likelihood ratios [LR+; LR‐] were also reported for each fold and for the overall assessment, as well as accuracy.
Data are visualized as mean ± s.e.m. and p‐values < 0.05 were considered statistically significant unless otherwise indicated. All statistical analyses were performed using RStudio version 4.3.1 and GraphPad Prism 10.4.1.
3. Results
3.1. Patient Characteristics
A total of 741 patients with UC were included, of which 50.6% were male [n = 375], with a mean age of 35.7 years (standard deviation [SD] ± 14.1 years). Pancolitis was noted on 39% of colonoscopies [n = 289]. Mean faecal calprotectin measurement was 407.4 μg/g (SD ±973.9) and median measurement was 126 (IQR 41.0–351.9). Based on Mayo Endoscopic Scores [MES], 52.0% [n = 385], 22.1% [n = 164], 10.5% [n = 78], and 4.6% [n = 34] patients had an MES of 0, 1, 2, and 3, respectively. Eighty patients did not have an available MES and were excluded from Mayo‐specific analyses.
For further analysis, we separated patients based on endoscopic remission (MES 0 vs. MES 1–3) and endoscopic improvement (MES 0–1 vs. MES 2–3). For these analyses, 74.1% [n = 549] and 15.1% [n = 112] patients had a MES 0–1 vs. MES 2–3, respectively, and 52.0% [n = 385] and 37.2% [n = 276] patients had a MES of 0 versus MES 1–3, respectively. As for Geboes Histological Score [GHS], patients were separated into two different possible analyses: either histological remission (Geboes ≤ 2.0 and > 2.0) or histologically active disease (Geboes < 3.1 and ≥ 3.1). For these analyses, 104 patients and 63 patients had to be excluded, respectively, due to a lack of available GHS. Based on GHS with a cut‐off of 2.0, 35.6% [n = 264], and 50.3% [n = 373] patients had a GHS of ≤ 2.0 and > 2.0, respectively. Based on GHS with a cut‐off of 3.1, 63.7% [n = 472], and 27.8% [n = 206] patients had a GHS of < 3.1 and ≥ 3.1, respectively. Demographic data is presented in Table 1.
TABLE 1.
Demographic data.
| Patients with ulcerative colitis (n, %) | |
|---|---|
| Patients | 741 (100%) |
| Sex | |
| Female | 366 (49.4%) |
| Male | 375 (50.6%) |
| Age (years) | |
| Mean (±standard deviation) | 35.6 ± 14.1 |
| Smoking status | |
| Smoker | 176 (23.8%) |
| Non‐smoker | 450 (60.7%) |
| NA | 115 (15.5%) |
| Disease extent | |
| Proctitis | 261 (35.2%) |
| Left‐sided | 99 (13.4%) |
| Pancolitis | 289 (39.0%) |
| NA | 92 (12.4%) |
| Mayo score | |
| 0 | 385 (52.0%) |
| 1 | 164 (22.1%) |
| 2 | 78 (10.5%) |
| 3 | 34 (4.6%) |
| NA | 80 (10.8%) |
| Geboes score (< 2.0 vs. ≥ 2.0) | |
| ≤ 2.0 | 264 (35.6%) |
| ≥ 2.0 | 373 (50.3%) |
| NA | 104 (14.0%) |
| Geboes score (< 3.1 vs. ≥ 3.1) | |
| < 3.1 | 472 (63.7%) |
| ≥ 3.1 | 206 (27.8%) |
| NA | 63 (8.5%) |
| Faecal calprotectin | |
| Mean (±standard deviation) | 407.3 ± 973.9 |
| Median (IQR Q1‐Q3) | 126 (41.0–351.9) |
3.2. Correlation Between FCAL, Endoscopy, and Histology
We plotted the distribution of FCAL levels based on our MES and GHS diagnostic categories (Figure 1). We found significantly greater levels of FCAL in the MES 1–3 (mean: 676.3; median 275.5 [IQR 81.5–773.3]) versus MES 0 (mean: 180.3; median 74 [IQR 32.85–199.5]) (p‐value < 0.0001, Figure 1A) cohort as well as in the MES 2–3 (mean: 1027; median 539 [IQR 201.3–1021.0]) versus MES 0–1 group (mean: 256.8; median 92.2 [IQR 35.0–245.5]) (p‐value < 0.0001, Figure 1B). Similarly, for histological remission assessment with Geboes, we find significantly greater levels of FCAL in GHS > 2.0 (mean: 596.4; median 180.0 [IQR 54.5–547.0]) versus GHS ≤ 2.0 (mean: 171.0; median 75 [IQR 31.25–209.0]) (p‐value < 0.0001, Figure 1C) and GHS ≥ 3.1 (mean: 788.6; median 300.0 [IQR 127.5–814.3]) versus GHS < 3.1 (mean: 236.6; median 74.0 [IQR 30.0–206.8]) (p‐value < 0.0001, Figure 1D).
FIGURE 1.
FCAL levels (μg/g) across (A) MES 0 versus MES 1–3, (B) MES 0–1 versus MES 2–3, (C) GHS ≤ 2.0 versus GHS > 2.0 and (D) GHS < 3.1 versus GHS ≥ 3.1.
Additionally, we performed a univariate regression analysis to examine the association between FCAL levels and our defined diagnostic categories. The comparison between MES 0–1 versus MES 2–3 yielded the highest estimate (770.38, 95% CI: 591.26–949.49, p < 2e‐16), suggesting a strong association. Similarly, the comparison of MES 0 versus MES 1–3 also showed a significant effect (496.01, 95% CI: 357.68–634.34, p = 4.82e‐12). For the Geboes comparisons, both < 3.1 versus ≥ 3.1 and ≤ 2.0 versus > 2.0, demonstrated significant estimates of 552.02 (95% CI: 392.23–711.81, p = 2.57e‐11) and 425.41 (95% CI: 267.69–587.40, p = 2.41e‐07), respectively. These findings indicate a significant association between increased FCAL levels and increased disease severity, as assessed by the Mayo and Geboes endoscopic and histological scales (Table 2).
TABLE 2.
Univariate regression analysis of the association between faecal calprotectin and the covariates below.
| Estimate | 95% CI | Significance (p) | ||
|---|---|---|---|---|
| Lower | Upper | |||
| Mayo 0_1 vs. 2_3 | 770.38 | 591.26 | 949.49 | < 2e‐16 |
| Mayo 0_vs._1_2_3 | 496.01 | 357.68 | 634.34 | 4.82e‐12 |
| Geboes < 3.1 vs. ≥ 3.1 | 552.02 | 392.23 | 711.81 | 2.57e‐11 |
| Geboes ≤ 2.0 vs. ≥ 2.0 | 425.41 | 265.43 | 585.39 | 2.41e‐07 |
3.3. Predicting Endoscopic Remission and Improvement
For Mayo Endoscopic Score (MES) 0 versus 1–3, that is, remission (Figure 2A; Table 3), the overall AUC was 0.716 (95% CI: 0.701–0.731, p < 0.0001), with an optimal FCAL threshold of 154.3 μg/g (specificity: 0.69 [95% CI: 0.65–0.72], sensitivity: 0.64 [95% CI: 0.60–0.69]). The positive predictive value (PPV) was 0.60 (95% CI: 0.54–0.65) and the negative predictive value (NPV) was 0.73 (95% CI: 0.68–0.77). The positive likelihood ratio (LR+) for MES 0 versus 1–3 was 2.07 (95% CI: 1.74–2.5) and the negative likelihood ratio (LR‐) was 0.52 (95% CI: 0.43–0.61). The accuracy to detect (MES) 0 versus 1–3 with this threshold was 67.03% (95% CI: 63.44–70.61).
FIGURE 2.
Receiver operating characteristic curves (ROC) on the training data (80%) using 5‐fold cross validation for (A) prediction of Mayo score (0 vs. 1–3) and (B) prediction of Mayo score (0–1 vs. 2–3) based on FCAL.
TABLE 3.
AUC and test sensitivity, specificity, PPV, NPV, LR+ and LR‐ for MES 0–1 versus MES 2–3, MES 0 versus 1–3, Geboes ≤ 2.0 versus ≥ 2.0, Geboes < 3.1 versus ≥ 3.1 using the closest (0,1) method.
| Threshold | AUC | p | Sens | Spec | PPV | NPV | LR+ | LR‐ | |
|---|---|---|---|---|---|---|---|---|---|
| Mayo 0_vs._1_2_3 | 154.3 | 0.716 (0.701–0.731) | < 0.0001 | 0.643 (0.597–0.689) | 0.687 (0.653–0.722) | 0.597 (0.541–0.651) | 0.73 (0.682–0.773) | 2.069 (1.742–2.458) | 0.516 (0.434–0.613) |
| Mayo 0_1 vs. 2_3 | 234.57 | 0.802 (0.797–0.807) | < 0.0001 | 0.686 (0.62–0.752) | 0.743 (0.736–0.751) | 0.353 (0.293–0.419) | 0.921 (0.892–0.943) | 2.677 (2.215–3.235) | 0.42 (0.318–0.556) |
| Geboes ≤ 2.0 vs. ≥ 2.0 | 117.6 | 0.653 (0.644–0.662) | < 0.0001 | 0.593 (0.554–0.632) | 0.608 (0.562–0.654) | 0.683 (0.631–0.731) | 0.516 (0.461–0.571) | 1.525 (1.284–1.812) | 0.664 (0.568–0.776) |
| Geboes < 3.1 vs. ≥ 3.1 | 166.2 | 0.752 (0.743–0.762) | < 0.0001 | 0.694 (0.628–0.759) | 0.693 (0.657–0.73) | 0.497 (0.439–0.554) | 0.838 (0.799–0.872) | 2.26 (1.92–2.66) | 0.441 (0.356–0.547) |
When comparing MES 0–1 versus 2–3, that is, improvement (Figure 2B; Table 3), the AUC increased to 0.802 (95% CI: 0.797–0.807, p < 0.0001), with an optimal FCAL threshold of 234.6 μg/g (specificity: 0.74 [95% CI: 0.74–0.75], sensitivity: 0.69 [95% CI: 0.62–0.75]). The PPV was 0.35 (95% CI: 0.29–0.42) and the NPV was 0.92 (95% CI: 0.89–0.94). The LR+ for MES 0 versus 1–3 was 2.68 (95% CI: 2.22–3.24) and the LR‐ was 0.42 (95% CI: 0.32–0.56). The accuracy to detect MES 0–1 versus 2–3 was 73.38% (95% CI: 72.17–74.59) with this threshold.
3.4. Predicting Histological Remission and Disease Activity
For Geboes Histopathology Score [28] ≤ 2.0 versus > 2.0, that is, remission (Figure 3A; Table 3), the AUC was 0.653 (95% CI: 0.644–0.662), with an optimal threshold of 117.6 μg/g (specificity: 0.61 [95% CI: 0.56–0.65], sensitivity: 0.59 [95% CI: 0.55–0.63]). The PPV was 0.68 (95% CI: 0.63–0.73) and the NPV was 0.52 (95% CI: 0.46–0.57). The LR+ for GHS ≤ 2.0 versus > 2.0 was 1.53 (95% CI: 1.28–1.81) and the LR‐ was 0.66 (95% CI: 0.57–0.78). The accuracy in detecting GHS ≤ 2.0 versus > 2.0 was 60.13% (95% CI, 58.46–61.80) with this threshold.
FIGURE 3.
Receiver operating characteristic curves (ROC) on the training data (80%) using 5‐fold cross validation for (A) prediction of Geboes score (GHS ≤ 2.0 vs. GHS > 2.0) and (B) prediction of Geboes score (GHS < 3.1 vs. GHS ≥ 3.1) based on FCAL.
For GHS < 3.1 versus ≥ 3.1, that is, disease activity (Figure 3B, Table 3), the AUC was 0.752 (95% CI: 0.743–0.762), with an optimal threshold of 166.2 μg/g (specificity: 0.69 [95% CI: 0.66–0.73], sensitivity: 0.69 [95% CI: 0.63–0.76]). The PPV was 0.50 (95% CI: 0.44–0.55) and the NPV was 0.84 (95% CI: 0.80–0.87). The LR+ was 2.26 (95% CI: 1.92–2.66) and the LR‐ was 0.44 (95% CI: 0.36–0.55). The accuracy in detecting GHS < 3.1 versus ≥ 3.1 was 69.33% (95% CI: 66.25–72.40) with this threshold.
3.5. Specificity and Sensitivity Analysis
Given that the comparison of MES 0–1 versus 2–3 had the highest AUC (0.802, 95% CI: 0.797–0.807, p‐value < 0.0001), we performed a 5‐fold cross‐validation analysis with different threshold selection methods and report the consequent sensitivities and specificities (Table 4). As reported above, we initially employed a selection method that maximizes both sensitivity and specificity equally by selecting the threshold nearest to the ideal point (0, 1) on the ROC curve. This method yields a threshold of 234.6 μg/g (specificity: 0.74 [95% CI: 0.74–0.75], sensitivity: 0.69 [95% CI: 0.62–0.75]). We then used the Youden index method, resulting in a threshold of 273.5 μg/g (specificity: 0.767 [95% CI: 0.752–0.782], sensitivity: 0.66 [95% CI: 0.62–0.705]). Finally, we selected FCAL thresholds reported in the literature for mucosal healing, defined as MES 0–1. Theede et al. [23] report a FCAL threshold of 192 μg/g, which in our data gives a specificity of 0.687 [95% CI: 0.674–0.699] and sensitivity of 0.750 [95% CI: 0.674–0.827]. Using Hart and colleagues' [29] reported FCAL threshold of 170 μg/g on our data to discriminate MES 0–1 versus MES 2–3 gave a specificity of 0.648 [95% CI: 0.641–0.656] and sensitivity of 0.801 [95% CI: 0.729–0.873]. Singh et al. [17] report a FCAL threshold of 150 μg/g which yields a specificity of 0.620 [95% CI: 0.607–0.635] and sensitivity of 0.809 [95% CI: 0.732–0.886] in our data. Finally, Dulai et al. [20] use 100 μg/g resulting in a specificity of 0.511 [95% CI: 0.491–0.532] and sensitivity of 0.889 [95% CI: 0.846–0.933].
TABLE 4.
Testing thresholds from literature, closest 0.1 and Youden for test sensitivity, specificity, PPV, NPV, LR+ LR‐, and accuracy for MES 0–1 versus MES 2–3.
| Threshold | Sens | Spec | PPV | NPV | LR+ | LR‐ | Accuracy | |
|---|---|---|---|---|---|---|---|---|
| Mayo 0_1 vs. 2_3 | 100 | 0.889 (0.846–0.933) | 0.511 (0.491–0.532) | 0.272 (0.229–0.319) | 0.959 (0.930–0.976) | 1.829 (1.643–2.036) | 0.209 (0.122–0.360) | 57.64 (56.71–58.56) |
| Mayo 0_1 vs. 2_3 | 150 | 0.809 (0.732–0.886) | 0.620 (0.607–0.635) | 0.307 (0.257–0.361) | 0.945 (0.916–0.964) | 2.168 (1.889–2.488) | 0.287 (0.192–0.430) | 65.51 (64.47–66.54) |
| Mayo 0_1 vs. 2_3 | 170 | 0.801 (0.729–0.873) | 0.648 (0.641–0.656) | 0.320 (0.269–0.377) | 0.949 (0.877–0.963) | 2.311 (2.001–2.670) | 0.289 (0.196–0.427) | 67.62 (65.95–69.30) |
| Mayo 0_1 vs. 2_3 | 192 | 0.750 (0.674–0.827) | 0.687 (0.674–0.699) | 0.331 (0.276–0.39) | 0.933 (0.905–0.954) | 2.422 (2.06–2.848) | 0.351 (0.252–0.490) | 69.90 (67.74–72.05) |
| Mayo 0_1 vs. 2_3 | 234.6 | 0.643 (0.597–0.689) | 0.687 (0.653–0.722) | 0.597 (0.541–0.651) | 0.73 (0.682–0.773) | 2.069 (1.742–2.458) | 0.516 (0.434–0.613) | 73.38 (72.17–74.59) |
| Mayo 0_1 vs. 2_3 | 273.5 | 0.66 (0.62–0.705) | 0.767 (0.752–0.782) | 0.366 (0.303–0.435) | 0.917 (0.888–0.939) | 2.834 (2.317–3.467) | 0.442 (0.340–0.575) | 74.89 (73.01–76.77) |
4. Discussion
To the best of our knowledge, this is the first and largest study to assess the role of FCAL thresholds in predicting endoscopic and histological remission by testing optimized thresholds in a separate cohort of patients not previously used to determine the thresholds. Previous studies evaluating the same question would assess the performance of their thresholds on the same data used for ROC analysis and threshold generation, leading to poor generalizability of the model to other data and overfitting, which causes an optimistic inflation of reported metrics [16]. Here, we describe the use of FCAL for the prediction of endoscopic remission, defined as MES 0 versus MES 1–3 or MES 0–1 versus MES 2–3, and histological remission, described as GHS ≤ 2.0 versus GHS > 2.0, or histological improvement, defined as GHS < 3.1 versus GHS ≥ 3.1. Our pooled data from four different academic centres comprises 741 patients (885 patients before the application of inclusion criteria), the most extensive study of this kind to date, enabling us to separate our data into training and testing sets.
Overall, we observe higher AUC values for the endoscopic classification of MES 0–1 versus MES 2–3 compared to MES 0 versus MES 1–3, as well as the Geboes histopathological scores. Indeed, we demonstrate that FCAL is most discriminatory to separate MES 0–1 vs. MES 2–3 in the prediction of endoscopic remission, given our overall AUC of 0.802 (95% CI: 0.797–0.807) across our 5‐fold cross‐validation analysis.
In terms of threshold selection, many studies [17, 20, 23, 29] have proposed different FCAL cut‐offs but without independent validation of said cut‐offs. Here, we have tested optimal thresholds selected by optimizing specificity and sensitivity and the Youden method yielding thresholds of 234.6 μg/g (specificity: 0.74 [95% CI: 0.74–0.75], sensitivity: 0.69 [95% CI: 0.62–0.75]) and 273.5 μg/g (specificity: 0.767 [95% CI: 0.752–0.782], sensitivity: 0.66 [95% CI: 0.62–0.705]) respectively. Given the risk that patients with active disease may not receive endoscopic evaluation based on an FCAL threshold, the goal should be to limit false negatives and thereby increase the sensitivity of the threshold at the expense of specificity. Based on our analysis of different thresholds, we propose a cut‐off of 170 μg/g to maximize sensitivity (0.801 [95% CI: 0.729–0.873]) without significantly compromising specificity (0.648 [95% CI: 0.641–0.656]) for the discrimination of endoscopic remission based on MES 0–1 versus MES 2–3. These findings indicate that FCAL thresholds can effectively distinguish endoscopic remission in UC, with an accuracy greater than 70% (73.38% [95% CI: 72.17–74.59]). In practical clinical terms, and consistent with the STRIDE‐II treat‐to‐target framework, we propose the following algorithmic use of this threshold: in a patient with clinically quiescent UC (partial Mayo score ≤ 2), an FCAL value < 170 μg/g can be used to safely defer routine surveillance colonoscopy, given the associated high negative predictive value for endoscopically active disease (MES 2–3). Conversely, an FCAL ≥ 170 μg/g should prompt clinical reassessment, which may include colonoscopy, to confirm endoscopic activity before consideration of treatment escalation. This threshold is therefore most applicable in the monitoring phase of disease management, supporting decisions about the timing of endoscopic re‐evaluation rather than guiding induction therapy.
Importantly, in our data, FCAL was less valuable to discriminate histological remission (GHS ≤ 2.0 vs. GHS > 2.0) as evidenced by the poor AUC (0.653 [95% CI: 0.644–0.662]) and accuracy (60.13% [95% CI: 58.46–61.80]). For histological activity (GHS < 3.1 vs. GHS ≥ 3.1), the AUC was 0.752 (95% CI: 0.743–0.762). Still, this value was not sufficiently high to discriminate between the two groups in a clinical context based on our data. Similarly, for the assessment of MES 0 versus MES 1–3, the AUC of 0.716 (95% CI: 0.701–0.731) is not sufficiently high. It is well known that asymptomatic UC patients can still present with endoscopic findings of inflammation [30, 31]. Thus, highlighting the importance of having a non‐invasive biomarker, such as FCAL, to track endoscopic remission. Based on our findings, FCAL is a better discriminatory test for endoscopic as compared to histological remission and could limit the need for invasive colonoscopy procedures. Importantly, the relatively high negative predictive value of faecal calprotectin in our cohort suggests that low FCAL levels can reliably exclude significant endoscopic or histologic activity. This finding supports the use of FCAL as a non‐invasive tool to safely reduce the need for additional endoscopic assessments in patients in clinical remission.
To the best of our knowledge, our study represents the first and largest assessment of FCAL thresholds in held‐out test folds within cross‐validation; however, it has several limitations that warrant discussion. First, the validation performed was internal, using stratified 5‐fold cross‐validation within the pooled dataset rather than an external independent cohort. A leave‐one‐centre‐out approach would provide stronger evidence of cross‐site transportability; however, the unequal sample sizes across contributing centres precluded this design. Second, although we exclusively pooled UC patients across the four studies, only three out of the four included UC patients with clinical remission, which potentially led to a heterogeneous patient population. Third, alternative causes of FCAL elevation such as use of non‐steroidal anti‐inflammatory drug (NSAID) use were not systematically excluded across all four contributing datasets. While NSAID exclusion was a formal criterion in the Guardiola et al. dataset, it was not uniformly applied in the remaining studies. Although patients under regular specialist follow‐up for UC are generally counselled to avoid NSAIDs, this remains a potential source of misclassification that future prospective studies should address by incorporating NSAID use as a formal exclusion or stratification variable. Fourth, we were unable to control for the sampling of stool FCAL across studies due to variations in the time of collection and duration of storage before analysis. Fifth, different FCAL assay kits were used across contributing centres (see Section 2.3). As noted by international consensus, FCAL measurement tests are not interchangeable across platforms, and this variability limits the direct transferability of absolute threshold values to laboratories using different assay systems. The consistency of our proposed 170 μg/g threshold with values reported independently in the contributing studies (Hart et al.: 170 μg/g; Magro et al.: 150–250 μg/g) provides some reassurance regarding robustness across platforms. Sixth, endoscopic and histologic assessments were performed locally at each centre without central reading, and interobserver agreement was not formally assessed. Seventh, the Mayo Endoscopic Score was used as the endoscopic reference standard, as it was the predominant instrument in clinical use across contributing centres at the time of data collection. The Ulcerative Colitis Endoscopic Index of Severity (UCEIS), a validated alternative that may offer greater discriminatory precision, particularly at the MES 0 vs. MES 1 boundary, was not uniformly available and could not be applied. Future prospective studies should incorporate both the MES and UCEIS to better characterize the relationship between FCAL and endoscopic activity across the full severity spectrum. Finally, the cross‐sectional design of this pooled analysis does not allow for longitudinal assessment of FCAL as a prognostic marker for disease relapse.
In summary, we highlight that FCAL is a vital biomarker for discriminating endoscopic healing (MES 0–1 vs. MES 2–3), yielding a sensitivity of 0.801 based on a threshold of 170 μg/g in a pooled cohort of 741 UC patients from four academic centres. Our pooled analysis highlights the generalizability of FCAL in assessing endoscopic remission, and our assessment of thresholds in held‐out test folds within cross‐validation limits the risks of overfitting, as seen in previous studies.
Author Contributions
Fernando Magro: investigation, writing – review and editing, resources. Anish Patel: investigation, writing – review and editing, resources. Matthieu Allez: investigation, writing – review and editing. Alain Bitton: investigation, writing – review and editing. Mark Sorin: formal analysis, writing – review and editing, visualization, investigation, methodology. Gary Wild: investigation, writing – review and editing. Laurie‐Rose Dubé: conceptualization, methodology, validation, formal analysis, investigation, data curation, writing – original draft, writing – review and editing, visualization. Peter L. Lakatos: investigation, writing – review and editing. Talat Bessissow: conceptualization, methodology, validation, formal analysis, investigation, resources, data curation, writing – original draft, writing – review and editing, funding acquisition, project administration. Waqqas Afif: investigation, writing – review and editing. Jordi Guardiola: investigation, writing – review and editing, resources.
Funding
The authors have nothing to report.
Conflicts of Interest
Talat Bessissow has received honorarium for Speaker, consultant and research support from Abbvie, Alimentiv, Bristol‐Myers‐Squibb, CSF Vifor, Celltrion, Eli Lilly, Ferring, Fresenius Kabi, Gilead, Iterative scope, Johnson and Johnson, Merck, Mirium, Pendopharm, Pentax, Pfizer, Roche, Sandoz, Sanofi, Takeda, Vial Pharma. Peter L. Lakatos has been a speaker and/or advisory board member: AbbVie, Amgen, Celltrion, Ferring, Fresenius Kabi, Gilead, Johnson Johnson, Eli Lilly, Organon, Pharmacosmos, Pendopharm, Pfizer, Roche, Sandoz, Sanofi and Takeda and has received unrestricted research grant: Gilead, Pfizer and Takeda. Jordi Guardiola has served as a speaker, consultant, or advisory board member, or has received research or educational funding from Lilly, Roche, MSD, AbbVie, Celltrion, Kern Pharma, Takeda, Johnson & Johnson, Pfizer, Galapagos, Sandoz, Bühlmann, GoodGut, and GE Healthcare. Anish Patel has served as a speaker or consultant for Abbvie, Eli Lilly, BMS, Pfizer, Takeda, and Johnson and Johnson.
Supporting information
Table S1: Inclusion and exclusion criteria of pooled studies.
Table S2: Threshold selection, training sensitivity, specificity and AUC and test sensitivity, specificity, PPV, NPV, LR+ and LR‐ for MES 0 versus MES 1–3 prediction.
Table S3: Threshold selection, training sensitivity, specificity and AUC and test sensitivity, specificity, PPV, NPV, LR+ and LR‐ for MES 0–1 vs. MES 2–3 prediction.
Table S4: Threshold selection, training sensitivity, specificity and AUC and test sensitivity, specificity, PPV, NPV, LR+ and LR‐ for Geboes score (GHS < 2.0 vs. GHS ≥ 2.0).
Table S5: Threshold selection, training sensitivity, specificity and AUC and test sensitivity, specificity, PPV, NPV, LR+ and LR‐ for Geboes score (GHS < 3.1 vs. GHS ≥ 3.1).
Figure S1: Receiver operating characteristic curves (ROC) on the training data (80%) using 5‐fold cross validation for (A) prediction of Mayo score (0 vs. 1–3) and (B) prediction of Mayo score (0–1 vs. 2–3) based on FCAL.
Figure S2: Receiver operating characteristic curves (ROC) on the training data (80%) using 5‐fold cross validation for (A) prediction of Geboes score (GHS < 2.0 vs. GHS ≥ 2.0) and (B) prediction of Geboes score (GHS < 3.1 vs. GHS ≥ 3.1) based on FCAL.
Dubé L.‐R., Sorin M., Allez M., et al., “The Association of Faecal Calprotectin Measurements With Endo‐Histological Remission in Ulcerative Colitis: A Pooled Analysis,” Alimentary Pharmacology & Therapeutics 64, no. 5 (2026): 615–624, 10.1111/apt.70735.
Handling Editor: Sree Subramanian
Data Availability Statement
The data that support the findings of this study are available from the corresponding author upon reasonable request.
References
- 1. Shen B., Kochhar G., Navaneethan U., et al., “Practical Guidelines on Endoscopic Treatment for Crohn's Disease Strictures: A Consensus Statement From the Global Interventional Inflammatory Bowel Disease Group,” Lancet Gastroenterology & Hepatology 5, no. 4 (2020): 393–405, 10.1016/s2468-1253(19)30366-8. [DOI] [PubMed] [Google Scholar]
- 2. Neurath M. F. and Travis S. P., “Mucosal Healing in Inflammatory Bowel Diseases: A Systematic Review,” Gut 61, no. 11 (2012): 1619–1635, 10.1136/gutjnl-2012-302830. [DOI] [PubMed] [Google Scholar]
- 3. Christensen B. and Rubin D. T., “Understanding Endoscopic Disease Activity in IBD: How to Incorporate It Into Practice,” Current Gastroenterology Reports 18, no. 1 (2016): 5, 10.1007/s11894-015-0477-6. [DOI] [PubMed] [Google Scholar]
- 4. Colombel J. F., Rutgeerts P., Reinisch W., et al., “Early Mucosal Healing With Infliximab Is Associated With Improved Long‐Term Clinical Outcomes in Ulcerative Colitis,” Gastroenterology 141, no. 4 (2011): 1194–1201, 10.1053/j.gastro.2011.06.054. [DOI] [PubMed] [Google Scholar]
- 5. Peyrin‐Biroulet L., Sandborn W., Sands B. E., et al., “Selecting Therapeutic Targets in Inflammatory Bowel Disease (STRIDE): Determining Therapeutic Goals for Treat‐To‐Target,” American Journal of Gastroenterology 110, no. 9 (2015): 1324–1338, 10.1038/ajg.2015.233. [DOI] [PubMed] [Google Scholar]
- 6. Roberts K. J., Cubitt M. F., Carlton T. M., et al., “Preclinical Development of a Bispecific TNFalpha/IL‐23 Neutralising Domain Antibody as a Novel Oral Treatment for Inflammatory Bowel Disease,” Scientific Reports 11, no. 1 (2021): 19422, 10.1038/s41598-021-97236-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Turner D., Ricciuto A., Lewis A., et al., “STRIDE‐II: An Update on the Selecting Therapeutic Targets in Inflammatory Bowel Disease (STRIDE) Initiative of the International Organization for the Study of IBD (IOIBD): Determining Therapeutic Goals for Treat‐To‐Target Strategies in IBD,” Gastroenterology 160, no. 5 (2021): 1570–1583, 10.1053/j.gastro.2020.12.031. [DOI] [PubMed] [Google Scholar]
- 8. Colombel J. F., D'Haens G., Lee W. J., Petersson J., and Panaccione R., “Outcomes and Strategies to Support a Treat‐To‐Target Approach in Inflammatory Bowel Disease: A Systematic Review,” Journal of Crohn's and Colitis 14, no. 2 (2020): 254–266, 10.1093/ecco-jcc/jjz131. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Dal Buono A., Roda G., Argollo M., Zacharopoulou E., Peyrin‐Biroulet L., and Danese S., “Treat to Target or ‘Treat to Clear’ in Inflammatory Bowel Diseases: One Step Further?,” Expert Review of Gastroenterology & Hepatology 14, no. 9 (2020): 807–817, 10.1080/17474124.2020.1804361. [DOI] [PubMed] [Google Scholar]
- 10. Yoon H., Jangi S., Dulai P. S., et al., “Incremental Benefit of Achieving Endoscopic and Histologic Remission in Patients With Ulcerative Colitis: A Systematic Review and Meta‐Analysis,” Gastroenterology 159, no. 4 (2020): 1262–1275.e7, 10.1053/j.gastro.2020.06.043. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Gupta A., Yu A., Peyrin‐Biroulet L., and Ananthakrishnan A. N., “Treat to Target: The Role of Histologic Healing in Inflammatory Bowel Diseases: A Systematic Review and Meta‐Analysis,” Clinical Gastroenterology and Hepatology 19, no. 9 (2021): 1800–1813.e4, 10.1016/j.cgh.2020.09.046. [DOI] [PubMed] [Google Scholar]
- 12. Røseth A. G., Schmidt P. N., and Fagerhol M. K., “Correlation Between Faecal Excretion of Indium‐111‐Labelled Granulocytes and Calprotectin, a Granulocyte Marker Protein, in Patients With Inflammatory Bowel Disease,” Scandinavian Journal of Gastroenterology 34, no. 1 (1999): 50–54, 10.1080/00365529950172835. [DOI] [PubMed] [Google Scholar]
- 13. Costa F., Mumolo M. G., Ceccarelli L., et al., “Calprotectin Is a Stronger Predictive Marker of Relapse in Ulcerative Colitis Than in Crohn's Disease,” Gut 54, no. 3 (2005): 364–368, 10.1136/gut.2004.043406. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Gisbert J. P., Bermejo F., Pérez‐Calle J. L., et al., “Fecal Calprotectin and Lactoferrin for the Prediction of Inflammatory Bowel Disease Relapse,” Inflammatory Bowel Diseases 15, no. 8 (2009): 1190–1198, 10.1002/ibd.20933. [DOI] [PubMed] [Google Scholar]
- 15. Walsham N. E. and Sherwood R. A., “Fecal Calprotectin in Inflammatory Bowel Disease,” Clinical and Experimental Gastroenterology 9 (2016): 21–29, 10.2147/ceg.S51902. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Gholamy A., Kreinovich V., and Kosheleva O., “Why 70/30 or 80/20 Relation Between Training and Testing Sets: A Pedagogical Explanation,” International Journal of Intelligent Technologies and Applied Statistics 11, no. 2 (2018): 105–111. [Google Scholar]
- 17. Singh A., Bhardwaj A., Sharma R., et al., “Predictive Accuracy of Fecal Calprotectin for Histologic Remission in Ulcerative Colitis,” Intestinal Research 23, no. 2 (2024): 144–156, 10.5217/ir.2024.00068. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Hart L., Chavannes M., Kherad O., et al., “Faecal Calprotectin Predicts Endoscopic and Histological Activity in Clinically Quiescent Ulcerative Colitis,” Journal of Crohn's and Colitis 14, no. 1 (2019): 46–52, 10.1093/ecco-jcc/jjz107. [DOI] [PubMed] [Google Scholar]
- 19. Cannatelli R., Bazarova A., Zardo D., et al., “Fecal Calprotectin Thresholds to Predict Endoscopic Remission Using Advanced Optical Enhancement Techniques and Histological Remission in IBD Patients,” Inflammatory Bowel Diseases 27, no. 5 (2020): 647–654, 10.1093/ibd/izaa163. [DOI] [PubMed] [Google Scholar]
- 20. Dulai P. S., Feagan B. G., Sands B. E., Chen J., Lasch K., and Lirio R. A., “Prognostic Value of Fecal Calprotectin to Inform Treat‐To‐Target Monitoring in Ulcerative Colitis,” Clinical Gastroenterology and Hepatology 21, no. 2 (2023): 456–466.e7, 10.1016/j.cgh.2022.07.027. [DOI] [PubMed] [Google Scholar]
- 21. Kawashima K., Oshima N., Kishimoto K., et al., “Low Fecal Calprotectin Predicts Histological Healing in Patients With Ulcerative Colitis With Endoscopic Remission and Leads to Prolonged Clinical Remission,” Inflammatory Bowel Diseases 29, no. 3 (2022): 359–366, 10.1093/ibd/izac095. [DOI] [PubMed] [Google Scholar]
- 22. Singh S., Ananthakrishnan A. N., Nguyen N. H., et al., “AGA Clinical Practice Guideline on the Role of Biomarkers for the Management of Ulcerative Colitis,” Gastroenterology 164, no. 3 (2023): 344–372, 10.1053/j.gastro.2022.12.007. [DOI] [PubMed] [Google Scholar]
- 23. Theede K., Holck S., Ibsen P., Ladelund S., Nordgaard‐Lassen I., and Nielsen A. M., “Level of Fecal Calprotectin Correlates With Endoscopic and Histologic Inflammation and Identifies Patients With Mucosal Healing in Ulcerative Colitis,” Clinical Gastroenterology and Hepatology 13, no. 11 (2015): 1929–1936.e1, 10.1016/j.cgh.2015.05.038. [DOI] [PubMed] [Google Scholar]
- 24. Rieder F., Bettenworth D., Ma C., et al., “An Expert Consensus to Standardise Definitions, Diagnosis and Treatment Targets for Anti‐Fibrotic Stricture Therapies in Crohn's Disease,” Alimentary Pharmacology & Therapeutics 48, no. 3 (2018): 347–357, 10.1111/apt.14853. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Magro F., Lopes S., Coelho R., et al., “Accuracy of Faecal Calprotectin and Neutrophil Gelatinase B‐Associated Lipocalin in Evaluating Subclinical Inflammation in UlceRaTIVE Colitis‐The ACERTIVE Study,” Journal of Crohn's and Colitis 11, no. 4 (2017): 435–444, 10.1093/ecco-jcc/jjw170. [DOI] [PubMed] [Google Scholar]
- 26. Patel A., Panchal H., and Dubinsky M. C., “Fecal Calprotectin Levels Predict Histological Healing in Ulcerative Colitis,” Inflammatory Bowel Diseases 23, no. 9 (2017): 1600–1604, 10.1097/mib.0000000000001157. [DOI] [PubMed] [Google Scholar]
- 27. Guardiola J., Lobatón T., Rodríguez‐Alonso L., et al., “Fecal Level of Calprotectin Identifies Histologic Inflammation in Patients With Ulcerative Colitis in Clinical and Endoscopic Remission,” Clinical Gastroenterology and Hepatology 12, no. 11 (2014): 1865–1870, 10.1016/j.cgh.2014.06.020. [DOI] [PubMed] [Google Scholar]
- 28. Kerut C. K., Wagner M. J., Daniel C. P., et al., “Guselkumab, a Novel Monoclonal Antibody Inhibitor of the p19 Subunit of IL‐23, for Psoriatic Arthritis and Plaque Psoriasis: A Review of Its Mechanism, Use, and Clinical Effectiveness,” Cureus 15, no. 12 (2023): e51405, 10.7759/cureus.51405. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Hart L., Chavannes M., Kherad O., et al., “Faecal Calprotectin Predicts Endoscopic and Histological Activity in Clinically Quiescent Ulcerative Colitis,” Journal of Crohn's and Colitis 14, no. 1 (2020): 46–52. [DOI] [PubMed] [Google Scholar]
- 30. Rosenberg L., Nanda K. S., Zenlea T., et al., “Histologic Markers of Inflammation in Patients With Ulcerative Colitis in Clinical Remission,” Clinical Gastroenterology and Hepatology 11, no. 8 (2013): 991–996, 10.1016/j.cgh.2013.02.030. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Bessissow T., Lemmens B., Ferrante M., et al., “Prognostic Value of Serologic and Histologic Markers on Clinical Relapse in Ulcerative Colitis Patients With Mucosal Healing,” American Journal of Gastroenterology 107, no. 11 (2012): 1684–1692, 10.1038/ajg.2012.301. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Table S1: Inclusion and exclusion criteria of pooled studies.
Table S2: Threshold selection, training sensitivity, specificity and AUC and test sensitivity, specificity, PPV, NPV, LR+ and LR‐ for MES 0 versus MES 1–3 prediction.
Table S3: Threshold selection, training sensitivity, specificity and AUC and test sensitivity, specificity, PPV, NPV, LR+ and LR‐ for MES 0–1 vs. MES 2–3 prediction.
Table S4: Threshold selection, training sensitivity, specificity and AUC and test sensitivity, specificity, PPV, NPV, LR+ and LR‐ for Geboes score (GHS < 2.0 vs. GHS ≥ 2.0).
Table S5: Threshold selection, training sensitivity, specificity and AUC and test sensitivity, specificity, PPV, NPV, LR+ and LR‐ for Geboes score (GHS < 3.1 vs. GHS ≥ 3.1).
Figure S1: Receiver operating characteristic curves (ROC) on the training data (80%) using 5‐fold cross validation for (A) prediction of Mayo score (0 vs. 1–3) and (B) prediction of Mayo score (0–1 vs. 2–3) based on FCAL.
Figure S2: Receiver operating characteristic curves (ROC) on the training data (80%) using 5‐fold cross validation for (A) prediction of Geboes score (GHS < 2.0 vs. GHS ≥ 2.0) and (B) prediction of Geboes score (GHS < 3.1 vs. GHS ≥ 3.1) based on FCAL.
Data Availability Statement
The data that support the findings of this study are available from the corresponding author upon reasonable request.