Edinburgh Research Explorer Association Between Level of Fecal Calprotectin and Progression of Crohn's Disease

Background Aims 10 Mucosal healing is associated with improved outcomes in patients with Crohn’s disease (CD), but 11 assessment typically requires ileocolonoscopy. Calprotectin can be measured in fecal samples to 12 determine luminal disease activity in place of endoscopy—this measurement is an important 13 component of the treat to target strategy. We investigated whether levels of fecal calprotectin 14 associate with subsequent CD progression. performed a study of 918 patients with (4218 patient-years of follow-up; median, 50.6 interquartile 32.8–76.0 if had 1 or more fecal calprotectin measurement made 3 or more following We collected clinical data and fecal calprotectin measurements and analyzed to identify 21 factors associated with a composite outcome of progression in Montreal behavior, hospitalization, and resection. 432 (IQR, 1365–998 in patients who reached the composite endpoint vs 180 4 (IQR, 50–665 in patients not. cutoff of µg/g the for complete resolution There is a strong correlation between FC, endoscopic disease activity and ulcer depth. Our data show more directly that elevated FC can be used as a marker of increased risk of progression.

Crohn's disease (CD), a form of inflammatory bowel disease (IBD), is characterized by relapsing 2 episodes of intestinal inflammation and the accumulation of irreversible digestive damage. Prognosis 3 is highly variable between individuals, 1 such that the identification of patients at greatest risk of poor 4 outcomes is an urgent research priority. Some clinical phenotypes, such as disease location and 5 environmental factors such as smoking, have been clearly associated with poorer outcomes. 2,3 6 However, accurate prediction remains difficult. Over the past decade, there has been a paradigm 7 shift away from treating until symptom resolution and towards mucosal healing as persistent 8 subclinical bowel inflammation leads to poorer outcomes. 4-8 However this has typically required 9 ileocolonoscopy, which is invasive, expensive and carries risk for patients. 9 10 Fecal calprotectin (FC) has become well-established as a biomarker of intestinal inflammation. 11 Calprotectin is a 36.5 kDa protein that constitutes 60% of the contents of granules in neutrophils. 10 12 Its use as a screening test to distinguish IBD from irritable bowel syndrome is well-supported by 13 multiple studies, with an AUROC of 0.95 in meta-analysis. 11 Several groups have demonstrated that 14 FC correlates well with endoscopic measures of disease activity. [12][13][14][15][16] There has been greater 15 uncertainty of its role in small bowel CD, but more recently FC has been shown to correlate well with 16 both MRI 17 and capsule endoscopy findings. 18, 19 17 The use of FC as a prognostic marker has been demonstrated in the context of medically-and 18 surgically-induced remission. [20][21][22] In both contexts, baseline FC predicts disease flare over a follow-up 19 period of two years, though there is also a rise notable in FC 3-4 months prior to clinical disease 20 flare. The recent CALM study has demonstrated the effectiveness of a treat to target strategy 21 incorporating FC in Crohn's disease. 23 However, it has still not yet been demonstrated whether 22 elevations in FC, irrespective of clinical symptoms, are associated with disease progression. This 23 information would provide further support to the principle of treating beyond symptoms.

M A N U S C R I P T A C C E P T E D ACCEPTED MANUSCRIPT
We aimed to use a large, extensively-phenotyped cohort of CD patients followed over time to 1 determine the value of FC to predict progression of disease. We focused on endpoints associated 2 with digestive damage: progression of Montreal behaviour 24 , surgical resection or hospitalization for 3 severe flare. 4

5
This was a retrospective cohort study of CD patients managed at the Western General Hospital, 6 Edinburgh, UK, a teaching hospital that cares for secondary-and tertiary-referred patients with IBD. 7 The primary inclusion criteria were a diagnosis of CD and at least one FC more than three months 8 post-diagnosis. The a priori primary endpoint was a composite of progression in Montreal luminal 9 disease behavior (B1 to B2/B3 or B2 to B3), hospitalization for flare and resectional surgery. These 10 individual components were also defined as separate secondary endpoints. In order to reduce the 11 possibility of merely measuring the FC at the time of the disease flare that caused the endpoint, any 12 events that happened within 90 days after the index FC were regarded as having already happened 13 and were not included in the endpoint analysis. 14 We obtained FC data from the Edinburgh FC Registry (EFCR), a record of every FC done in Edinburgh 15 since its introduction in 2003. Patients in this initial cohort had their first FC between 2003 and 2014 16 and were followed up until 2015. Fecal calprotectins were requested as part of routine monitoring 17 and also directed by patients' symptoms. These data represent a convenience sample, and include 18 all patients tested during that period who met our inclusion criteria. 19 We matched these data to existing research and clinical databases to identify patients with a known 20 diagnosis of CD. We then interrogated the electronic and paper medical records to obtain 21 information on demographics, symptoms, disease location and behavior over time Edinburgh. The manufacturer's reference range for distinguishing inflammatory bowel disease from 13 functional gut disorders is >50 µg/g. 14 Statistical analysis was done using R 3.5.1 (R Foundation for Statistical Computing, Vienna, Austria). 15 The Mann Whitney U test was performed for continuous non-parametric data, while Fisher's exact 16 tests were done for categorical data. Survival analysis was performed using Kaplan Meier and Cox 17 proportional hazards models. 26 For the survival models, we have reported the outcome as the 18 proportion with maintained digestive health, i.e. the inverse of our primary endpoint. Patients were 19 excluded from the specific analysis of progression in Montreal behavior if they were already B3 at 20 baseline. 21 FC was analyzed using log-transformed data and using a predefined threshold of 250 µg/g. The 22 optimum threshold for FC on survival analysis was then explored by examining the p values of the 23 likelihood ratio test and the Akaike Information Criteria for Cox proportional hazard models. Variable 24 M A N U S C R I P T A C C E P T E D ACCEPTED MANUSCRIPT 8 selection for multivariable models was done using a stepwise backwards method based on Akaike 1 Information Criterion. We performed Cox proportional hazards analyses of the effect of drug therapy 2 up to 3 months pre or 6 months post fecal calprotectin on the primary outcome; for this analysis, 3 patients who had disease progression within the first six months or who were censored in that 4 period were excluded from analysis. The multistate transition data for disease progression in the 5 overall cohort was done using the empirical transition matrix method. 27 6 The principal analysis was done using the first FC for each patient where there was more than one. 7 Owing to the retrospective nature of this dataset, these were not taken at uniform intervals. 8 Exploratory analysis of multiple FCs was performed using the median for each rolling six-month 9 period centered on each month following diagnosis and stratified by progression in Montreal 10 behavior. FCs were excluded from this analysis where the patient was symptomatic at the time of 11 sampling. 12 This study was conducted as a service evaluation using data collected routinely as part of clinical 13 care, and therefore following guidance from the UK Health Research Authority did not require 14 specific ethical approval or consent. 15

16
We identified 918 CD patients meeting our inclusion criteria ( Figure 1). 61.1% were female, and 17 median age at the index FC measurement was 40.7 years (interquartile range [IQR] 28.5-54.8) (Table  18 1). Median follow-up time was 50.6 months (IQR 32.8-76.0), with a total of 4218 patient-years of 19 follow-up across the cohort. At diagnosis, 81% had an inflammatory (B1) phenotype, 12% stricturing 20 (B2) and 8% penetrating (B3). By 30 years post-diagnosis, the proportions of B1, B2 and B3 were 21 estimated as 29%, 36% and 36% respectively ( Figure 2 Demographic and biomarker data on the cohort stratified by whether the patients reached the 1 composite endpoint or not are shown in table 2. On univariable cox proportional hazards analysis, FC 2 was strongly associated with an elevated risk of reaching the primary endpoint (Table 3), with a 3 hazard ratio (HR) of 1.79 (95% CI 1.50 -2.14, p = 1.9×10 -10 ) for log 10 (FC). The only other blood tests 4 nominally associated with FC on univariable analysis were CRP (p=0.016), hemoglobin (p=0.011) and 5 platelets (p=0.003). There were also associations with younger age at diagnosis (p=0.010), female 6 sex (p=0.021), prior immunomodulator use (p=0.012), symptoms at index visit (p=1.2×10 -7 ). Smoking 7 status, previous intestinal resection, previous anti-TNF and time period of FC measurement 8 (pre/post 2008) use were not associated with the primary endpoint, nor was there a significant 9 difference in the time since diagnosis at the index FC. 10 On multivariable Cox proportional hazards analysis, disease progression was independently 11 associated with elevated FC, female sex, younger age, ileal/ileocolonic disease, previous 12 immunomodulator use and symptoms (Table 3). 13 A further analysis was performed to explore the effect of changes in treatment before and after 14 measurement of calprotectin (Supplementary Table 1). This was restricted to patients who did not 15 have disease progression and were not censored within the first six months. There were no 16 significant associations with changes in medication in the three months leading up to the 17 measurement of fecal calprotectin. Use of steroids in the six months following calprotectin was 18 significantly associated with disease progression (HR 1.5 [95% CI 1.16 -2.03], p=0.003). However, 19 this was no longer significant in a multivariable analysis that also included the FC result 20 (Supplementary Table 2). 21 Above a threshold FC of 250 µg/g, the hazard ratio for reaching the primary endpoint was 1.9 (95% 22  Figure 4). 4 Using the Kaplan-Meier estimates, the positive predictive value of an index FC >115 µg/g was 28%, 5 43%, 52% and 59% at 2, 4, 6 and 8 years respectively. The negative predictive value of an index FC 6 ≤115 µg/g was 88%, 80%, 74% and 65% at 2, 4, 6, and 8 years respectively. 7 In a sensitivity analysis by quartiles of time from diagnosis to first fecal calprotectin, the association 8 between calprotectin and disease progression was seen for quartiles 2 to 4, but not for the patients 9 in the first quartile; these patients had 3 to 15.5 months between diagnosis and first fecal 10 calprotectin (Supplementary Figure 5). 11 We performed an exploratory analysis using all of the available CD FC data and excluding FC taken 12 when patients had symptoms. This analysis included 1456 FCs from 396 patients. The rolling median 13 FC can clearly be seen to differ between those 35/396 patients with a subsequent progression in 14 Montreal behavior and those that did not (Supplementary Figure 6). 15

16
This study demonstrates that elevated FC is associated with increased disease progression, both as 17 defined by a composite primary endpoint of advance in Montreal luminal behavior, surgical 18 resection and hospitalization and by each of these endpoints when considered individually. 19 Mucosal healing is recognized as a target for therapy in Crohn's disease, with poorer prognosis and a 20 higher risk of surgery associated with increased endoscopic disease activity 4-8 There is a strong 21 correlation between FC, endoscopic disease activity and ulcer depth. 12,28 Our data show more 22 directly that elevated FC can be used as a marker of increased risk of progression. 23

M A N U S C R I P T A C C E P T E D ACCEPTED MANUSCRIPT
Although absolute index FC levels were lower in L1 patients, FC better predicted poorer outcomes in 1 patients with L1/L3 rather than L2 disease distribution. Patients with active colonic disease may be 2 more likely to exhibit symptoms, and thus have earlier intervention. In contrast, patients with active 3 ileal disease may tolerate a higher level of subclinical inflammation, resulting in delay of treatment 4 with a greater risk of progression and complications. 5 Other variables associated with an adverse outcome in our analysis included younger age, which has 6 previously been identified as an adverse prognostic factor 1 , and previous immunomodulator use 7 which is likely to be a marker for a more aggressive prior disease course. Symptomatically active 8 disease was associated with an increased rate of disease progression, independently of elevated FC. 9 This validates a treat-to-target approach aiming for a combination of resolution of symptoms as well 10 as mucosal healing, with FC a marker of the latter. 11 Thresholds for prediction of disease relapse have varied across the literature, influenced by the 12 disease cohort being studied and the assay used. Several studies have identified a cut-off of 250 µg/g 13 as being useful to distinguish active from inactive disease. 20,22,29 In the present study, the optimal 14 separation between survival curves for progression of disease was seen using a lower threshold of 15 115 µg/g, suggesting that lower levels of inflammatory activity may still be associated with an 16 adverse outcome. However, any such threshold needs to be interpreted in the context of the Together, these data now clearly support a treat-to-target strategy combining a patient-reported 2 symptom score with FC as a marker of mucosal inflammation. 3 Strengths of the present study include the large number of patients and duration of follow-up, with 4 a median follow-up time following index FC of greater than four years. A clinically relevant definition 5 of disease progression was selected a priori, and rich phenotype information was available. 6 Restricting measurement of endpoints to at least 90 days after the index FC should reduce bias from 7 measuring disease activity associated with an exacerbation that went on to cause hospital admission 8 or surgical resection. It can also be observed that the survival curves in figures 3-6 continue to 9 separate for many months after the index FC. This suggests that identification of mucosal 10 inflammation at any point in patient follow up, even at relatively modest levels previously 11 considered acceptable (i.e. FC 115-250ug/g), should warrant careful monitoring and low threshold 12 for treatment escalation decisions. 13 Limitations of this study relate to its retrospective nature. FCs were not collected at fixed intervals, 14 but as determined by the treating clinician. However, routine monitoring of FC including in 15 asymptomatic patients was established quite early on in Edinburgh after the full roll-out of the test 16 in 2005. The study was also performed at a single centre, which may reduce heterogeneity but at the 17 expense of generalizability. Nonetheless, although the Western General Hospital is a referral centre, 18 it also has a large secondary care population from the local catchment. Finally, medication data were 19 completed as accurately as was possible, but it is possible some courses of steroids, particularly 20 those in primary care, may have been missed. This is unlikely to have introduced any systematic bias. 21 In conclusion, we have shown in this study that elevated fecal calprotectin is associated with an

Background and Context
Fecal calprotectin is a marker of luminal Crohn's disease (CD) activity. We investigated whether fecal calprotectin associates with subsequent CD progression.

Findings
We have now shown that an increased fecal calprotectin is associated with a long-term increase in disease progression, including hospitalisation, surgery and advance in Montreal behaviour.

Implications for patient care
It is important to screen asymptomatic patients for mucosal inflammation and pursue complete resolution of inflammation.