Objective To determine the diagnostic accuracy of calprotectin to diagnose inflammatory bowel disease (IBD) in children in whom general practitioners (GPs) suspected IBD.
Design Prospective observational cohort study of a new calprotectin-based primary care referral pathway.
Setting 48 GP practices and gastroenterology secondary care services at the Royal Devon and Exeter NHS Foundation Trust in the South-West of England, UK.
Patients 195 children aged between 4 and 18 years referred on the pathway between January 2014 and August 2017 for investigation of gastrointestinal symptoms were included.
Interventions Primary-care-driven faecal calprotectin testing. Primary and secondary care records over 12 months from the point of calprotectin testing were used as the reference standard.
Main outcome measures Diagnostic accuracy of calprotectin testing to detect IBD.
Results 7% (13/195) tested patients were diagnosed with IBD. Using our prespecified cut-off of 100 µg/g, calprotectin had a diagnostic accuracy of 91% (95% CI 86% to 95%) with a sensitivity for distinguishing IBD from non-IBD of 100% (95% CI 75% to 100%), a specificity of 91% (95% CI 85% to 94%), a positive predictive value of 43% (95% CI 25% to 63%) and a negative predictive value of 100% (95% CI 98% to 100%). Calprotectin testing had no effect on the time to diagnosis, but a negative test contributed to saved referrals and was associated with fewer diagnostic tests in secondary care.
Conclusions Calprotectin testing of children with suspected IBD in primary care accurately distinguishes IBD from a functional gut disorder, reduces secondary care referrals and associated diagnostic healthcare utilisation.
- primary care
- diagnostic tests
Statistics from Altmetric.com
What is already known on this topic?
Distinguishing paediatric inflammatory bowel disease (IBD) from non-IBD in patients presenting to their general practitioner (GP) for the first time with lower gastrointestinal symptoms can be difficult.
Currently, GPs use a combination of ‘red-flag’ symptoms such as rectal bleeding, as well as blood biomarkers, including C reactive protein (CRP), to guide onward specialist referral.
Faecal calprotectin is routinely used in the secondary care setting to rule out IBD, but there is a paucity of UK data exploring its real-world use in paediatric primary care referral pathways.
What this study adds?
In this real-world application in children in primary care, we have shown that a calprotectin ≥100 µg/g discriminates accurately IBD from non-IBD and outperforms ‘red-flag’ symptoms and/or CRP.
Calprotectin testing did not influence the time to diagnosis of IBD, but a negative test may have contributed to a reduction in outpatient referrals and secondary care investigations.
Most children who present to their general practitioner (GP) with gastrointestinal symptoms are diagnosed with a self-limiting infection or functional gut disorder that can be managed in primary care;1–3 a minority, however, will have inflammatory bowel disease (IBD).
Because of the practical constraints of venepuncture, in particular in young children, primary care physicians often rely on presenting symptoms alone to formulate a differential diagnosis. The presence of ‘red-flag’ symptoms including weight loss, rectal bleeding and a family history of IBD should prompt secondary care referral; however, these symptoms are common and have poor discriminative power.4–6 Where C reactive protein (CRP) is available, its diagnostic accuracy for IBD is limited by moderate sensitivity, particularly for diagnosing terminal ileal Crohn’s disease and left-sided ulcerative colitis.7 8
Faecal calprotectin is a stool biomarker that distinguishes paediatric IBD and other organic intestinal disease from functional gut disorders.9 10 We, like others, have shown that in adults with suspected IBD, calprotectin can be used by GPs to risk stratify patient referrals and permit timely diagnoses.11–14 There is a paucity of data exploring the real-world use of calprotectin testing in paediatric primary care referral pathways.5
We hypothesised that calprotectin testing would distinguish IBD from non-IBD with clinically useful positive predictive value (PPV) and negative predictive value (NPV), save secondary care referrals and reduce the time to diagnosis from GP presentation.
We aimed to:
Compare the diagnostic accuracy of primary care calprotectin testing in children to distinguish IBD from non-IBD diagnoses, stratified by ‘red-flag’ symptoms.
Define whether calprotectin testing alters primary care referral behaviour, reduces the time to diagnosis from presentation to GP and influences the use of secondary care investigations.
Compare the phenotype and time to diagnosis of patients referred on and off our primary care calprotectin pathway.
Design and clinical setting
We designed a prospective observational cohort study to describe the diagnostic accuracy of calprotectin in children with suspected IBD in primary care.
The Royal Devon & Exeter (RD&E) NHS Trust provides paediatric secondary care services to the Eastern locality of the Northern, Eastern and Western Devon Clinical Commissioning group and serves a local population of 378 000 people, of whom 75 000 are under 18 years old. A working group comprising representation from primary to secondary care devised a new calprotectin referral pathway in January 2014 based on the 2013 NICE guidance.11 Of 49 primary care practices in the catchment area for the RD&E, 48 agreed to take part in this study (online supplementary methods).
Children aged between 4 and 18 years referred on the calprotectin pathway between January 2014 and August 2017 for investigation of gastrointestinal symptoms were included and followed up for at least 12 months. Exclusion criteria were a previous diagnosis of IBD, suspicion of cancer and use of non-steroidal anti-inflammatory drugs within the previous 6 weeks. This was a convenience sampling series. Use of calprotectin was not mandated in patients meeting our eligibility criteria; the test was used at the discretion of the treating GP.
Variables and data acquisition
We used a purpose-designed request form that captured patient demographic data, presenting symptoms and relevant family history at the point of calprotectin testing (online supplementary appendix 1). Patients with one or more of rectal bleeding, unintentional weight loss and a family history of IBD were deemed to have ‘red-flag’ symptoms. GPs were also asked the following hypothetical referral question to assess expected referral behaviour: ‘Would you have referred this patient if calprotectin had been unavailable?’
Using secondary care databases, we recorded healthcare utilisation data in the year after calprotectin testing including: outpatient clinic referrals (paediatrics, gastroenterology, colorectal surgeons, upper gastrointestinal surgeons, dieticians and private clinics); diagnostic imaging (ultrasound, CT and MRI) and endoscopy (colonoscopy, flexible sigmoidoscopy and gastroscopy).
The diagnosis of IBD was made on clinical, radiological and histopathological findings in line with the Paris classification and Porto criteria.15 16 Absence of IBD was defined as no radiological or histopathological evidence of IBD or no IBD diagnosis after a 12-month period of follow-up.2 We mandated this period of follow-up to allow sufficient time for organic pathology to evolve and any missed cases of IBD to be correctly diagnosed. The primary care notes of children not referred to secondary care were reviewed to see whether there had been any further contact with a healthcare practitioner.
Faecal calprotectin (index test)
Samples were analysed in accordance with our previously described methods.11 In brief, our quantitative ELISA (Immundiagnostik, Bensheim, Germany) reported levels in the range 6–2100 µg/g. In accordance with previous research, results ≥100 µg/g were deemed positive. Calprotectin results and recommended actions were returned to GPs within 10 days.
Because this study was designed as a service evaluation, a priori power calculations were not undertaken. Rather, we decided to allow our new pathway to become established and then assessed its usefulness over 3 years. All analyses were two tailed and p<0.05 were considered significant and were conducted in R 3.5.1 (R Foundation for Statistical Computing, Vienna, Austria). We included patients with missing clinical data in analyses for which they had data and specified the denominator for each variable.
We used χ² analyses to compare expected and observed GP referrals to determine whether calprotectin influenced GP referral behaviour. We used Fisher’s exact tests for categorical data and Mann-Whitney U tests for continuous data to identify baseline clinical variables and biomarkers associated with a diagnosis of IBD and to identify phenotypic differences, including the time to diagnosis between children diagnosed with IBD referred on and off our new pathway.
Receiver operator characteristic curves and area under the curve (AUC) analyses were undertaken to determine clinical validity of calprotectin as a continuous variable to diagnose IBD and organic intestinal disease: Youden’s formula was used to determine the optimal cut-off.17 We calculated sensitivity, specificity, PPV and NPV at our prespecified cut-off and then at various calprotectin cut-off thresholds to optimise its use in diagnosing IBD. We used stepwise forward multivariable logistic regression models to compare diagnostic strategies using calprotectin and CRP with ‘red-flag’ symptoms.
Ethical consideration and patient involvement
In accordance with UK Health Research Authority guidelines, we did not require formal ethical approval.18 Patients were not involved in the conception or design of this study.
Patient disposition through the study is outlined in figure 1. The laboratory received 260 calprotectin samples from 237 children between January 2014 and August 2017. We included 195 patients in the final analysis: 46% (90/195) were female with a median (IQR) age of 15.1 (12.0–17.1) years. At least one ‘red-flag’ criterion was present in 54% (104/192) of all patients.
Primary care referrals
Over half (53%, 104/195) of GPs indicated that they intended to refer their patient had calprotectin testing not been available; 33% (64/195) were unsure if they would have referred and 6% (11/195) stated that they would not have referred (table 1). On multivariable logistic regression, the final decision to refer was independently associated with pre-calprotectin intention to refer (OR=2.4 (95% CI 1.3 to 4.6) for ‘yes’ vs ‘unsure’ or ‘no’, p=0.006) and calprotectin ≥100 µg/g (OR=4.2 (95% CI 1.6 to 13, p=0.007).
Seven per cent (13/195) patients were diagnosed with IBD (8 Crohn’s disease, 2 ulcerative colitis, 3 IBD-unclassified). The most frequent non-IBD diagnoses was functional gut disorder/IBS (87%, 158/182) (online supplementary table 1).
Distinguishing IBD and non-IBD
Patients diagnosed with IBD more commonly reported diarrhoea (Bristol stool ≥type 6) (p=0.05), a change in stool appearance (p=0.039) and change in stool frequency (p=0.04) than patients with non-IBD (table 2). At the time of GP referral, 62% (8/13) of patients diagnosed with IBD and 54% (96/179) of patients subsequently diagnosed with non-IBD (p=0.78) had at least one ‘red-flag’ criterion.
Calprotectin levels were significantly greater in the IBD than the non-IBD group (online supplementary figure 1). Using a cut-off of 100 µg/g, calprotectin had a diagnostic accuracy of 91% (95% CI 86% to 95%) with a sensitivity for distinguishing IBD from non-IBD of 100% (95% CI 75% to 100%), a specificity of 91% (95% CI 85% to 94%), a PPV of 43% (95% CI 25% to 63%) and a NPV of 100% (95% CI 98% to 100%) (figure 1).
Increasing the threshold separating positive and negative calprotectin results from 50 µg/g (manufacturer’s cut-off) to 100 µg/g doubled the PPV of the test with no reduction in NPV (online supplementary tables 2 and 4). Further increases likewise increased the PPV but at the expense of a reduction in the NPV and a reduction in sensitivity due to missed IBD cases.
Thirteen patients had a false positive calprotectin; 11 were referred by their GP to secondary care. Of two patients not referred to secondary care, one (calprotectin=125 µg/g) was given an inferred diagnosis of non-IBD after 1.9 years’ follow-up and the other (calprotectin=102 µg/g) a diagnosis of gastroenteritis. Of the 11 referred patients, 9/11 underwent imaging and all 11 underwent upper and lower gastrointestinal endoscopy, with no other organic treatable diagnoses made. Of patients who did not have IBD, calprotectin was more likely to be positive in patients with rectal bleeding compared with those without (18% (7/39) vs 7% (10/139) p=0.06).
Calprotectin had a significantly better AUC when compared with albumin (p=0.010), haemoglobin (p=0.004), white blood cell count (p<0.001) and platelets (p<0.001) but not CRP (p=0.16) (figure 2). Using Youden’s method to maximise the difference between the true positive and false positive rate over all possible cut-off values, the optimal cut-off threshold for distinguishing IBD from non-IBD was 110 µg/g.
Secondary care investigations
Patients with a negative calprotectin were referred to secondary care less frequently and were less likely to present emergently than the positive calprotectin cohort (table 3). They also underwent fewer endoscopic investigations and less imaging.
IBD phenotype diagnosed on and off the calprotectin pathway
Over the study period, 42 children were diagnosed with IBD. Sixty-nine per cent (29/42) were diagnosed without primary care calprotectin testing, and of these, 73% (19/26) had a secondary care calprotectin prior to diagnosis. All had a calprotectin ≥100 µg/g.
Patients diagnosed without use of a primary care calprotectin test reported more rectal bleeding (75% (21/28) vs 31% (4/13), p=0.014) and unintentional weight loss (52% (13/25) vs 15% (2/13), p=0.04), but family history of IBD, disease phenotype, emergency hospital attendances and serological biomarkers, including CRP and calprotectin, were not significantly different between the two cohorts (online supplementary table 3 and online supplementary figure 2). There was no difference in time to diagnosis between patients who underwent primary care calprotectin testing and those who went straight to secondary care.
Key results and implementation
Based on the results of our predictive modelling, we would not recommend stratifying paediatric referrals using red-flag symptoms alone. Calprotectin, however, is a useful biomarker: the high NPV allows exclusion of IBD and over half of patients with a raised calprotectin were subsequently diagnosed with ulcerative colitis or Crohn’s disease despite the lower pretest probability in primary care6 19 than in referred cohorts.10 20 Our data suggest that GPs may be using calprotectin in primary care to determine whether secondary care referral is warranted. The optimal calprotectin cut-off threshold for distinguishing IBD from non-IBD was 110 µg/g. Although CRP and calprotectin both discriminated IBD from non-IBD, calprotectin was more sensitive, detecting IBD in three cases where CRP was within the normal range (≤5 mg/L). Calprotectin has the added benefit of being non-invasive and therefore widely applicable even in young children in primary care, particularly as other blood biomarkers, such as full blood count and white cell count, as demonstrated in our study, are often normal at diagnosis. In referred patients with a negative calprotectin, 10% underwent radiological and 13% endoscopic tests, which were normal, suggesting that secondary care clinicians can use this biomarker to reduce unnecessary investigations. Implementation of a primary calprotectin based-pathway is likely to reduce the risk incurred by unnecessary endoscopic procedures.21–23
Strengths and limitations
The strengths of our study lay in the prospective design and meticulous follow-up of all primary care records of non-referred patients. We acknowledge, however, some important limitations. The uptake of calprotectin testing in primary care was relatively poor with only approximately one-third children diagnosed with IBD during the time-period of the study referred by participating practices via the pathway. This small sample size limits the power of our analyses. With respect to our definition of ‘red-flag’ symptoms, only family history of IBD, rectal bleeding and unintentional weight loss were included, omitting other important factors, such as perianal lesions and linear growth failure.
Like previous studies, we did not mandate endoscopy in all patients, and partial verification bias may have overestimated the accuracy of calprotectin.11–13 We took various measures to limit the effect of this bias, including using a 12-month follow-up period to allow for IBD to evolve and capturing data across primary and secondary care. For non-referred patients, this should have avoided an erroneous diagnosis of functional gut disorder being assigned to patients who may have been diagnosed with IBD in other hospitals. Conversely, about one-fifth of patients with a raised calprotectin received invasive investigation without a subsequent IBD diagnosis. However, all patients were reviewed by a specialist prior to further investigation, emphasising the role of calprotectin as a tool which supports further investigation of referred children with suspected IBD.
Despite a relatively small sample size, we did detect IBD and report no false negatives unlike previous paediatric studies.5 6
It is unlikely that our proposed calprotectin cut-off is applicable in children age 0–4 years, as normal values of calprotectin are higher than children aged >4 years.24–26
Our conclusions are limited to a cohort of patients who received calprotectin at the discretion of their GP. Although we compare the IBD phenotype in patients diagnosed on and off the calprotectin pathway, we lack similar data for patients with non-IBD diagnoses made outside the pathway. Rectal bleeding remains an important symptom which prompts GP referral. However, we were unable to demonstrate a disease spectrum bias in the non-tested compared with calprotectin-tested patients with IBD, supporting our recommendations for calprotectin in all primary care patients with possible IBD not meeting criteria for emergent referral.
Contemporary UK data are lacking, but our time to diagnosis is shorter than previously reported in one UK cohort study of new paediatric IBD diagnoses where median time to diagnosis was 5 months.27 Conversely, in this study, median time to diagnosis was 63 days.
We acknowledge that our data come from a single secondary care centre in the UK. Whether our findings are applicable to other centres with less capacity is unknown, and our findings should be replicated in a multisite setting to allow for further differences in gender, ethnicity and age to be elucidated. It is likely, however, that the phenotype of paediatric IBD is similar to that presenting elsewhere in the UK27 and that calprotectin is likely to be useful when stratifying referrals.
Contributors GW, JH, RKR, TA, JRG and NAK participated in the conception and design of the work. CB was the project manager and coordinated recruitment. MHP and TJM coordinated all biochemical analysis and central laboratory aspects of the project. GJW, NC, AT, SL, LM, NMH, PH, MA, SM, JWH, RKR, TA, JRG and NAK were involved in the acquisition, analysis or interpretation of data. The data analysis was performed by GW and NAK. Drafting of the manuscript was conducted by GW, NC, AT, SL, JWH, RKR, TA, JRG and NAK. All the authors contributed to the critical review and final approval of the manuscript. TA obtained the funding for the study.
Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
Competing interests GJW has consulted for AbbVie and received honoraria from Falk and AbbVie for unrelated topics and a fellowship from NIHR. NC is funded by a PhD studentship by Crohn’s and Colitis UK. RKR reports honoraria from Abbvie, Ferring, Therakos and Celltrion, grants from Nestec, outside the submitted work. TA has received unrestricted research grants, advisory board fees, speaker honorariums and support to attend international meetings from AbbVie, Merck, Janssen, Takeda, Ferring, Tillotts, Ferring, Pfizer, NAPP, Celltrion, Hospira for unrelated topics. JRG received honoraria from Falk, Abbvie and Shield therapeutics for unrelated topics. NAK has consulted for Falk and received honoraria from Falk, Allergan, Pharmacosmos and Takeda for unrelated topics and is a deputy editor of Alimentary Pharmacology & Therapeutics Journal; no financial relationships with any organizations that might have an interest in the submitted work in the previous three years.
Patient consent for publication Not required.
Provenance and peer review Not commissioned; externally peer reviewed.
Data availability statement Data are available on reasonable request.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.