Factors Associated With Loss of Fecal Biomarker Accuracy During Routine Assessment of Inflammatory Bowel Disease

ABSTRACT

Background

Fecal calprotectin (fCal) and fecal myeloperoxidase (fMPO) have been proposed as biomarkers for monitoring inflammatory bowel disease (IBD); however, their reliability is not fully established. This study investigated clinical variables associated with the loss of accuracy of fecal biomarkers.

Methods

Post hoc analyses were performed on previously reported data (New indicators of disease activity in IBD (NIDA‐IBD) cohort), which examined fecal biomarkers in individuals with IBD undergoing colonoscopy. Positive and negative results were recorded using previously identified thresholds (fCal, 150 μg/g; fMPO, 8 μg/g). Univariable logistic regression analyses were used to identify clinical variables (patient demographics, disease location and duration, symptoms and medication use) associated with false‐positive and false‐negative results. Subgroup analyses (Mann–Whitney U test) were performed to establish if the presence of colonic pseudopolyps was associated with fCal and fMPO concentrations in those with endoscopically inactive disease.

Results

One hundred and seventy individuals from the NIDA‐IBD cohort were included for analyses (CD, n = 98; female, n = 90; median disease duration 13 years). No clinical variables investigated were associated with a false positive or false negative result. Pseudopolyp presence was not associated with fCal or fMPO concentrations in those with endoscopically inactive disease. Positive and negative predictive values were higher using fecal biomarkers in combination with symptom scores than using symptom scores alone in predicting/excluding endoscopic activity.

Conclusions

No clinical variables were found to be associated with loss of biomarker accuracy, including fecal myeloperoxidase at widely used thresholds. Fecal biomarkers improved prediction of disease activity.

1. Introduction

Inflammatory bowel disease (IBD), comprising ulcerative colitis (UC) and Crohn's disease (CD), is characterized by chronic and fluctuating inflammation of the gastrointestinal (GI) tract [1]. Personalized management of IBD requires periodic assessment of disease activity and commonly involves the use of non‐invasive biomarkers [2, 3]. These measures of inflammation are more readily accessible and affordable when compared with ileocolonoscopy (the gold standard measure for GI inflammation), especially in resource‐constrained settings [4]. Neutrophil‐derived fecal biomarkers such as fecal calprotectin (fCal) and fecal myeloperoxidase (fMPO) are associated with IBD activity and longitudinal IBD outcomes [5, 6, 7], and their use (fCal) is supported in international guidelines [2]. The accuracy of fCal can be affected by several factors including the use of non‐steroidal anti‐inflammatory drugs (NSAIDs), proton‐pump inhibitors (PPIs), non‐IBD inflammation and in cases of predominantly proximal small bowel inflammation [8, 9, 10, 11]. Inaccuracies of fecal biomarkers may lead to over or under‐treatment, exposing patients to unnecessary escalation of medical therapy or risks related to uncontrolled inflammation [12].

This study aimed to identify clinical variables associated with the loss of accuracy of fCal and fMPO in IBD when compared to colonoscopy [7]. Furthermore, this study aimed to investigate the additive value of fecal biomarkers to symptom indices in identifying gut inflammation in routine IBD assessment [2].

2. Methods

This study analyzed data from a previously recruited cohort (New indicators of disease activity in IBD (NIDA‐IBD) cohort) [7] which examined fecal biomarkers in individuals with established IBD who were undergoing colonoscopy for disease assessment. Fecal samples were collected prospectively prior to endoscopic assessment to measure fCal and fMPO using previously described methods [7, 13]. Endoscopic assessment was performed using validated indices (simple endoscopic score for CD, SES‐CD; UC endoscopic index of severity, UCEIS) by the gastroenterologist performing the colonoscopy who was trained in using these scores but was blinded to biomarker assessment. Endoscopic activity was scored as SES‐CD > 2 and UCEIS > 1, and moderate‐to‐severe activity was scored as SES‐CD > 6 and UCEIS > 4 [6]. Patient symptoms were assessed at the time of biomarker collection and active symptoms were defined as a Harvey‐Bradshaw Index (HBI) for CD > 4 or simple clinical colitis activity index (SCCAI) for UC > 5 [6]. Biomarker limits previously associated with endoscopic disease activity in this cohort (fCal, 150 μg/g; fMPO, 8 μg/g) were used to classify test results [7].

Univariable binary logistic regression analyses were used to identify baseline clinical variables associated with false positive and false negative results. Additional analyses were performed using biomarker thresholds associated with moderate‐to‐severe disease activity (fCal, 250 μg/g; fMPO, 50 μg/g) [7].

Multivariable binary logistic regression analyses were subsequently conducted to assess associations between baseline variables and false positive and false negative results, whilst adjusting for factors previously associated with loss of fecal biomarker accuracy: non‐steroidal anti‐inflammatory (NSAID) use, proton pump inhibitor use (PPI), and additionally isolated ileal disease (among individuals with Crohn's disease) [8, 9, 10, 11]. These multivariable analyses were exploratory in nature and the number of variables was limited given the small sample size of this study and the risk of overfitting of statistical models.

Subgroup analyses (Mann–Whitney U test) were performed to establish if the presence of colonic pseudopolyps was associated with fCal and fMPO concentrations in those with endoscopically inactive disease. Multivariable linear regression analyses were conducted to assess the effect of pseudopolyp presence on fecal biomarker concentration in those with endoscopically inactive disease when controlled for potential confounders such as age and disease duration.

Positive predictive values (PPV) for detecting endoscopically active disease were calculated for three test thresholds: active symptoms alone, active symptoms and fCal > 150 μg/g, and active symptoms and fMPO > 8 μg/g. Negative predictive values (NPV) for excluding the presence of moderate‐to‐severely active disease on endoscopy were similarly calculated using inactive symptoms alone, inactive symptoms and fCal < 150 μg/g, and inactive symptoms and fMPO < 8 μg/g.

Subgroup analyses based on IBD phenotype were unable to be performed in this study given the small sample size of the cohort.

Informed consent was gained from all participants and study procedures were conducted in accordance with the Declaration of Helsinki and this study received approval from the New Zealand Health and Disability Ethics Committee (18/NTA/197).

3. Results

3.1. Baseline Participant Characteristics

One hundred and seventy individuals from the NIDA‐IBD cohort [6] were included for these analyses, 98 with CD and 72 with UC (Table 1).

TABLE 1.

Baseline characteristics for participants included from the NIDA‐IBD cohort.

	CD (n = 98)	UC (n = 72)	All IBD (n = 170)
Median age (IQR, years)	45 (34–56.8)	48 (40–62.3)	47 (36–58.8)
Female (%)	52 (53.1)	38 (52.8)	90 (52.9)
Median time since IBD diagnosis (IQR, years)	13 (6.3–21)	12 (4–22.5)	13 (5–22)
New Zealand European ethnicity (%)	98 (100)	69 (95.8)	167 (98.2)
Montreal classification of CD (%)
A1	7 (7.1)
A2	73 (74.5)
A3	18 (18.4)
L1	18 (18.4)
L2	21 (21.4)
L3	59 (60.2)
B1	58 (59.2)
B2	20 (20.4)
B3	20 (20.4)
Perianal disease	16 (16.3)
Montreal classification of UC (%)
E1		6 (8.3)
E2		33 (45.8)
E3		33 (45.8)
Endoscopic remission (%) (SES‐CD ≤ 2, UCEIS < 2)	35 (35.7)	37 (51.4)	72 (42.4)
Endoscopically active IBD (%) (SES‐CD > 2, UCEIS > 1)	63 (64.3)	35 (48.6)	98 (57.7)
Pseudopolyps present (%)	13 (13.3)	12 (16.7)	25 (14.9)
Immunomodulator/biologic use (%)	50 (51)	31 (43.1)	81 (47.6)
PPI use (%)	16 (16.3)	15 (20.8)	31 (18.2)
NSAID use (%)	1 (1)	2 (2.8)	3 (1.8)
Median fCal ug/g (IQR)	115.6 (46.3–350.3)	146.1 (50.3–781.1)	132.2 (46.3–519.1)
Median fMPO ug/g (IQR)	10.7 (3–44.4)	15.6 (1.9–86.4)	11.2 (2.4–61)

Abbreviations: CD, Crohn's disease; fCal, fecal calprotectin; fMPO, fecal myeloperoxidase; IBD, inflammatory bowel disease; IQR, inter‐quartile range; NIDA‐IBD, new indicators of disease activity in inflammatory bowel disease; NSAID, non‐steroidal anti‐inflammatory drug; PPI, proton pump inhibitor; SES‐CD, simple endoscopic score for Crohn's disease; UC, ulcerative colitis; UCEIS, ulcerative colitis endoscopic index of severity.

3.2. Variables Associated With Misclassification of Disease State Based on Biomarkers

Misclassification of disease state was assessed using previously reported biomarker thresholds associated with endoscopic activity (fCal, 150 μg/g; fMPO, 8 μg/g) [7]. This threshold for fCal resulted in 64 (37.6%) true positive and 13 (7.6%) false positive cases; and 59 (34.7%) true negative and 34 (20%) false negative cases. For fMPO, there were 71 (41.8%) true positive and 23 (13.5%) false positive cases; and 49 (28.8%) true negative and 27 (15.9%) false negative results.

Univariable binary logistic regression analyses demonstrated no significant association between the presence of the tested baseline clinical variables and biomarker inaccuracies at thresholds of fCal 150 μg/g and fMPO 8 μg/g (Table 2). Identical analyses conducted using alternative biomarker thresholds of 250 μg/g fCal and 50 μg/g fMPO demonstrated increased odds of a negative test result being false among individuals with ileal disease as compared to those with colonic disease (OR = 6.76, 95% CI 1.47–31.1), and with steroid use in the last year (OR = 2.48, 95% CI 1.1–5.59) for an fCal threshold of 250 μg/g (Table S1). Odds of a positive result being false were increased among individuals with extensive UC as compared to limited disease (OR = 2.00, 95% CI 1.17–23.25) at an fCal threshold of 250 μg/g (Table S1).

TABLE 2.

Logistic regression analyses showing odds ratios to identify clinical variables associated with loss of biomarker accuracy in inflammatory bowel disease (IBD).

Baseline clinical variable (fecal biomarker threshold)	Odds ratio (95% CI)
	False positive result, n = 13 (fCal 150 ug/g)	False negative result, n = 34 (fCal 150 ug/g)	False positive result, n = 23 (fMPO 8 ug/g)	False negative result, n = 27 (fMPO 8 ug/g)
Age	1.01 (0.97–1.05)	1.00 (0.97–1.03)	1.01 (0.98–1.04)	0.99 (0.96–1.03)
Female gender	0.37 (0.11–1.24)	1.19 (0.50–2.82)	0.52 (0.21–1.29)	0.74 (0.29–1.91)
Colonic versus Ileal (CD) (reference colonic)	0.00	2.08 (0.54–8.11)	0.78 (0.16–3.89)	1.67 (0.25–11.13)
Disease extent, Limited vs Extensive (UC) (reference limited)	3.30 (0.60–18.30)	1.00 (0.15–6.53)	1.94 (0.50–7.58)	1.00 (0.16–6.42)
Current smoker	0.90 (0.11–7.61)	4.91 (0.90–26.89)	1.78 (0.22–14.47)	2.94 (0.46–18.79)
Steroids at baseline	0.40 (0.05–3.22)	1.44 (0.36–5.77)	0.44 (0.10–2.01)	1.96 (0.45–8.55)
Previous steroid use for IBD	0.56 (0.15–2.11)	0.70 (0.25–1.91)	0.82 (0.32–2.13)	1.08 (0.37–3.18)
Biologic or immunomodulator use at baseline	0.94 (0.30–2.91)	0.73 (0.31–1.74)	1.01 (0.42–2.43)	0.98 (0.38–2.53)
BMI	0.99 (0.89–1.10)	1.06 (0.98–1.42)	0.95 (0.86–1.04)	1.05 (0.96, 1.14)
Disease duration	1.01 (0.95–1.06)	1.01 (0.97–1.06)	0.95 (0.90–1.00)	1.03 (0.98–1.07)
Other systemic inflammatory condition	1.37 (0.16–11.74)	2.49 (0.52–11.86)	1.66 (0.33–8.33)	0.90 (0.08–10.45)
Pseudopolyps	1.81 (0.46–7.11)	1.57 (0.39–6.37)	2.34 (0.82–6.68)	0.86 (0.19–3.97)
Subjectively active (HBI > 4, SCCAI > 5)	0.45 (0.14–1.45)	1.81 (0.77–4.26)	0.82 (0.34–1.97)	1.97 (0.76–5.11)

Note: A false positive result indicates fecal biomarker results above the selected threshold with an absence of endoscopically active disease. A false negative result indicates fecal biomarker results below the selected threshold with the presence of endoscopically active disease. Endoscopically active disease = SES‐CD > 2; UCEIS > 1.

Abbreviations: BMI, body mass index; CD, Crohn's disease; CI, confidence interval; fCal, fecal calprotectin; fMPO, fecal myeloperoxidase; HBI, Harvey‐Bradshaw Index for CD; SCCAI, simple clinical colitis activity index for UC; UC, ulcerative colitis.

Multivariable binary logistic regression analyses controlling for NSAID and PPI use among all individuals and NSAID use, PPI use and ileal isolated Crohn's disease among individuals with Crohn's disease, demonstrated no significant association between the tested variables and receiving an inaccurate result at thresholds of fCal 150 μg/g and fMPO 8 μg/g (Table S2). Using higher thresholds of fCal 250 μg/g and fMPO 50 μg/g, smoking was significantly associated with receiving a false negative result among all patients with IBD, but did not reach significance in the subgroup analysis of individuals with Crohn's disease (Table S3).

3.3. Pseudopolyps and Fecal Biomarker Concentrations

In total, there were 72 individuals with endoscopically inactive disease in this cohort. There were no significant differences in biomarker concentrations between those with and without pseudopolyps who had endoscopically inactive disease for either fCal (43.9 μg/g vs. 46.7 μg/g, p = 0.86) or fMPO (9.91 μg/g vs. 3.51 μg/g, p = 0.33). Multivariable linear regression did not provide evidence of an association between pseudopolyp presence and biomarker concentration among individuals with inactive disease when controlled for potential confounders age and disease duration (Table S4).

3.4. Test Performance: Positive and Negative Predictive Values

The PPV in predicting endoscopically active disease was higher when using a combination of symptom indices and fecal biomarkers than when using symptoms alone (Table 3, Figure 1). Similarly, the NPV for excluding the presence of moderate‐to‐severely active IBD was higher when using fecal biomarkers in conjunction with symptoms than when using symptoms alone (Table 3, Figure 1).

TABLE 3.

Predictive values for endoscopic disease activity when using symptom indices alone and in combination with fecal biomarkers.

	CD (n = 98)	UC (n = 72)	All IBD (n = 170)
Active symptoms (%)	60 (61.2)	36 (50.0)	96 (56.5)
PPV a : Active symptoms	62	78	68
NPV b : Inactive symptoms	68	100	84
fCal > 150 ug/g (%)	41 (41.8)	36 (50.0)	77 (45.3)
PPV a : fCal > 150 ug/g	85	81	83
NPV b : fCal < 150 ug/g	88	100	92
PPV a : fCal > 150 ug/g + Active symptoms	88	92	90
NPV b : fCal < 150 ug/g + Inactive symptoms	90	100	96
fMPO > 8 ug/g (%)	56 (57.1)	39 (54.2)	95 (55.9)
PPV a : fMPO > 8 ug/g	74	76	75
NPV b : fMPO < 8 ug/g	91	100	95
PPV a : fMPO > 8 ug/g + Active symptoms	71	92	81
NPV b : fMPO < 8 ug/g + Inactive symptoms	89	100	95

Note: Active symptoms = Harvey‐Bradshaw index for CD > 4 or simple clinical colitis activity index for UC > 5.

Abbreviations: fCal, fecal calprotectin; fMPO, fecal myeloperoxidase; IBD, inflammatory bowel disease.

^a Positive predictive value (PPV) of predicting endoscopically active inflammatory bowel disease (simple endoscopic score for Crohn's disease (SES‐CD) > 2 or ulcerative colitis endoscopic index of severity (UCEIS) > 1).

^b Negative predictive value (NPV) for excluding moderate to severely active disease on endoscopy (SES‐CD ≤ 6 or UCEIS ≤ 4).

FIGURE 1.

Improved confidence in assessing for the presence or absence of inflammatory bowel disease activity when using fecal biomarkers in addition to patient‐reported symptoms. Active symptoms = Harvey‐Bradshaw Index for Crohn's disease > 4 and simple clinical colitis activity index for ulcerative colitis > 5.

4. Discussion

This study reiterates that the fecal biomarkers, fCal and fMPO, are powerful adjuncts to symptoms when considering real‐world implementation of treat‐to‐target strategies for the management of IBD. These non‐invasive biomarkers, when used in conjunction with patient‐reported symptoms, improve the detection of endoscopically active disease and also the exclusion of significant missed diagnoses such as moderate‐to‐severely active disease. However, as demonstrated in this and other studies [14], an optimal biomarker “threshold” is unclear [3] and a proportion of patients remain misclassified when using fecal biomarkers alone compared to the gold standard of colonoscopy. Thus, ongoing investigation is required in large global cohorts to assess for potential contributions to this misclassification bias, and whether this can be reduced through the introduction of a third‐line diagnostic non‐invasive modality such as intestinal ultrasound [15]. This may help to strengthen the validity of future non‐invasive treat‐to‐target strategies for IBD.

4.1. Clinical and Biological Factors Contributing to Biomarker Inaccuracy

Elevated fecal biomarkers in those with endoscopically inactive disease can reflect non‐IBD sources of intestinal inflammation and clinical variables such as NSAID use, PPI use, presence of adenomas, perianal fistulas and intestinal infection [8, 9, 10, 11, 16, 17]. On the other hand, little is known about patients with low or undetectable neutrophil‐derived fecal biomarkers, but active disease on colonoscopy. This group of patients is at risk of undertreatment and may be exposed to disease progression. Some of these individuals may have fibrotic post‐inflammatory CD, although this doesn't fully explain the false negatives seen in UC, or in those who have recently been diagnosed [18]. A proportion of these patients may have ileal predominant CD, which has been previously implicated in loss of biomarker accuracy [11]. Ongoing investigations such as those carried out in this study may help to develop individualized fecal biomarker thresholds for specific patient groups.

IBD is a heterogeneous group of inflammatory disorders, whose exact cellular mechanisms leading to inflammation are incompletely understood [1]. The observation that some individuals have low or undetectable fecal biomarkers despite endoscopically active disease may reflect variability in the inflammatory pathway that causes alterations in neutrophil signaling, and leads to lower fCal and fMPO despite mucosal inflammation. Further investigation of this group is required to determine the mechanisms behind this phenomenon, as well as the potential consequences related to untreated inflammation.

Although this study did not identify any clinical variables associated with loss of biomarker accuracy at thresholds of fCal 150 μg/g and fMPO 8 μg/g, at a threshold of 250 μg/g, two variables were associated with an increased odds of receiving a false negative result, and one variable with an increased odds of receiving a false positive result. After controlling for PPI and NSAID use in multivariable regression analyses however, none of the three variables remained significant, indicating that some of the observed associations may be attributable to these factors. The first of these findings relates to an increased odds of receiving a false negative result for a fecal calprotectin of 250 μg/g in those with ileal versus colonic disease. This is in keeping with previous literature and our own analyses that have demonstrated that fecal calprotectin is less elevated in those with small bowel disease [6, 11]. The second of these findings is the increased odds of receiving a false positive result in individuals with extensive versus limited ulcerative colitis. This observation may be accounted for by the influence of gastrointestinal bleeding without significant luminal inflammation influencing the fecal calprotectin [19, 20]. It should be noted however that the confidence interval is large, reflecting the small sample size, and thus this result requires further consideration in larger studies. The third significant result is an increased odds of a false negative result in those with steroid use in the last year, potentially reflecting suppression of neutrophil‐derived biomarkers related to the ongoing anti‐inflammatory properties of recent cortico‐steroids. Following multivariable regression analysis, smoking was associated with receiving a false negative result at thresholds of fCal 250 μg/g and fMPO 50 μg/g. This was a surprising finding given that previous studies have demonstrated an association between smoking and higher fCal levels among healthy subjects [21, 22], however the wide confidence interval reflects the limited sample size, and this finding should therefore be interpreted with caution and validated in larger cohorts.

4.2. Difficulties in Establishing Biomarker Thresholds

The current study highlights the difficulties in establishing universally accepted fecal biomarker thresholds for determining IBD activity. Various authors have investigated cut‐off values of fCal ranging from 50 to 315 μg/g to determine active disease [13, 18]. Lower thresholds are more likely to overestimate disease activity and result in potentially unnecessary treatment, whereas higher thresholds are more likely to misclassify individuals as inactive, who may have benefited from escalation of medical therapy. Additionally, fCal measurements exhibit significant variability depending on collection and analysis techniques [23]. On initial assessment, fMPO appears to be a promising biomarker to be added to the growing array of tools for the non‐invasive assessment of IBD activity, particularly given that it can be measured without the need for ELISA methodology [24]. However, optimal thresholds of fMPO require validation in a large cohort of patients with ethnic diversity. In this study, previously suggested targets of fCal 150 μg/g and fMPO 8 μg/g⁷ were selected and exhibited superior performance characteristics to symptom scores alone, suggesting that their use has real world application despite optimal thresholds being an ongoing research question.

Whilst no variables were significantly associated with inaccuracies at thresholds of fCal 150 μg/g and fMPO 8 μg/g, three variables became significant at 250 μg/g fCal, although the association did not persist when PPI and NSAID use were adjusted for during multivariate regression. These findings suggest that the relationship between clinical factors and biomarker test accuracy may be threshold‐dependent.

4.3. Enhanced Diagnostic Accuracy Through Combined Assessment

The present study demonstrates that there is added value in using a combination of symptoms and fecal biomarkers in assessing for the presence and absence of endoscopically active disease. The value of combining these indices when assessing longitudinal outcomes in IBD has been previously reported [7]. We believe that using a combination of symptoms and biomarkers most closely reflects real‐world practice and what is recommended in current guidelines such as STRIDE II².

4.4. Study Limitations

The NIDA‐IBD study limitations have been previously described [7] and these analyses include a small sample size with a heterogeneous patient cohort, and a lack of small bowel imaging to identify proximal small bowel disease in CD which may have contributed to a portion of misclassified cases. Given the exploratory nature of this study, large multivariable analyses were not performed. Including many predictors with the available sample size could lead to overfitting and reduce the reliability of any identified associations.

5. Conclusion

This study adds to the body of literature investigating clinical factors associated with false positive and false negative fecal biomarker results in a population with known IBD undergoing routine colonoscopic assessment. No clinical variables tested were found to be associated with false positive or false negative results at thresholds of fCal 150 μg/g and fMPO 8 μg/g. Fecal biomarkers improved positive and negative predictive values for identifying and excluding the presence of GI inflammation over symptom scores alone. Further study is required to investigate the underlying causes related to loss of accuracy in fecal biomarkers, in particular in individuals with false negative results.

Ethics Statement

Informed consent was gained from all participants and study procedures were conducted in accordance with the Declaration of Helsinki and this study received ethics approval from the New Zealand Health and Disability Ethics Committee (18/NTA/197).

Conflicts of Interest

A.S. has received honoraria for educational activities from Janssen and Celltrion, and travel assistance from Dr. Falk Pharma (unrelated to this manuscript). A.S.D. has served on advisory boards for Janssen, AbbVie, and Nestlé (all unrelated to this manuscript). A.S.D.'s research is supported by Cure Kids. R.B.G. has received research grants, served on advisory boards, and received honoraria for educational activities for Janssen, AbbVie and Zespri (unrelated to this manuscript).

Supporting information

Data S1: Supplementary Information.

Kendall C., Swaminathan A., Borichevsky G. M., Day A. S., Kettle A. J., and Gearry R. B., “Factors Associated With Loss of Fecal Biomarker Accuracy During Routine Assessment of Inflammatory Bowel Disease,” JGH Open 10, no. 3 (2026): e70301, 10.1002/jgh3.70301.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.