People with Crohn’s disease are likely to be very familiar with faecal calprotectin kits; these tests use a stool sample to determine the amount of gastrointestinal inflammation a person has. Gastroenterologists often use the results from these kits to assist with determining how active a person’s Crohn’s disease is without needing that person needing to undergo a colonoscopy.

We know it is crucial faecal calprotectin is kept low early after diagnosis to improve the likelihood a Crohn’s disease patient is going to have a good outcome. My colleagues at the University of Edinburgh and NHS Lothian have previously shown having a calprotectin below \(250 \; (\mu g/g)\) a year after diagnosis reduces the likelihood of the disease worsening and patients being hospitalised or requiring surgery (Plevris et al., 2021). However, this previous work did not consider calprotectin as a measurement which changes continuously over time and instead ony considered two measurements for each person measured a year apart.

We believed that across people with Crohn’s disease, there were likely common patterns of faecal calprotectin over time. To the best of our knowledge, no previous research had attempted to determine if we can group people with Crohn’s disease with other Crohn’s disease patients with similar calprotectin over time. Recently, colleagues and I have attempted to determine how many groups with their own shared calprotectin patterns can be reliably found, what these patterns look like, and if being in one of these groups are potentially affected by the data known at diagnosis (Constantine-Cooke et al., 2023).

For our study, we considered the calprotectin of 356 people with Crohn’s disease and followed them from diagnosis for five years. We have used highly flexible statistical models, known as latent class mixed models with natural cubic splines, to determine the number of groups with distinct calprotectin patterns and their appearance.

We found four distinct groups which are presented below and largely reflect what is seen by gastroenterologists in clinics. Three of the groups can be broadly described as having overall decreasing calprotectin over time with the fourth group (cluster 2) having consistently high calprotectin. It is highly likely there are patients in this latter group who would have benefitted from receiving more advanced therapies sooner and also patients who would have always had aggressive disease regardless of the treatment they received. We are interested in determining which patients this is the case for in the future.


The four distinct calprotectin profiles we found. The gray lines are the calprotectin measurements for each subject deemed to have a similar profile to the red curve.

Of the data which are typically available at diagnosis, only smoking status and the presence of upper gastrointestinal inflammation were found to be significantly associated with which group a patient was assigned to. Smokers were more likely to be assigned to class 1 or 2 whilst people with upper gastrointestinal disease were less likely to be assigned to class 1. Whilst smokers being assigned to class 2 makes sense as smoking is highly associated with worse disease outcomes in Crohn’s disease, we do not fully understand the association we found between smoking and class 1. However, this may be due to smokers in class 1 giving up smoking after being diagnosed. Unfortunately, we do not have data for whether patients gave up smoking, and this is simply a theory.

We found early biologic treatment being very highly associated with class membership with those in class 2 being less likely to receive early biologic treatment than the other groups. However, We were unable to accurately predict class membership using all variables available at diagnosis and data science techniques (random forest and multinomial logistic regression).

Overall, we have shown latent class mixed models are an excellent tool for finding subgroups of people with Crohn’s disease with shared calprotectin profiles and determining these profiles in a data-driven manner (rather than asking patients which diagram shown to them best described their disease).

We are currently working on expanding this work to include all ulcerative colitis and Crohn’s disease patients treated by the Edinburgh IBD Unit and model both CRP and calprotectin.

Our long-term goal is to develop a tool which will predict a patient’s risk of surgery, hospitalisation, or worsening disease based on symptoms, calprotectin, and other measures which will then update as additional measurements are taken. This could then help clinicians when making decisions such as which treatment plan to place a patient on and how closely a patient should be monitored.


References


  1. Normalization of fecal calprotectin within 12 months of diagnosis is associated with reduced risk of disease progression in patients with Crohn’s disease
    Nikolas Plevris, James Fulforth, Mathew Lyons, Spyros I. Siakavellas, Philip W. Jenkinson, Cher S. Chuah, Laura Lucaciu, Rebecca J. Pattenden, Ian D. Arnott, Gareth-Rhys Jones, and Charlie W. Lees
    Clinical Gastroenterology and Hepatology Sep 2021
  2. Longitudinal Fecal Calprotectin Profiles Characterize Disease Course Heterogeneity in Crohn’s Disease
    Nathan Constantine-Cooke, Karla Monterrubio-Gómez, Nikolas Plevris, Lauranne A.A.P. Derikx, Beatriz Gros, Gareth-Rhys Jones, Riccardo E. Marioni, Charlie W. Lees, and Catalina A. Vallejos
    Clinical Gastroenterology and Hepatology Mar 2023