Latent Class Analysis Offers Efficient Reference Standard for Active Crohn’s Disease

A comparative diagnostic evaluation has demonstrated that latent class analysis can serve as a viable and efficient reference standard for identifying active Crohn’s disease, offering an alternative to traditional panel consensus methods. In a retrospective study published in BMC Medical Research Methodology, researchers conducted the first direct comparative assessment evaluating whether a statistical modeling approach could replace or improve upon expert consensus panels in diagnostic test accuracy trials. Crohn’s disease activity typically manifests as persistent bowel inflammation that can culminate in chronic tissue damage, requiring diagnostic frameworks to accurately confirm both the presence of the disease and its inflammatory activity. Because individual diagnostic procedures exhibit suboptimal standalone accuracy, researchers historically depended on consensus panels to synthesize multi-test data. The newly evaluated statistical framework demonstrates that latent class modeling can reliably synthesize diagnostic information while significantly reducing the operational burdens associated with expert consensus.

Establishing an accurate reference standard remains a foundational hurdle in Crohn’s disease diagnostics because single modalities cannot reliably capture disease activity on their own. Evaluating terminal ileal Crohn’s disease, where evidence of inflammatory activity is most abundant, necessitates combining clinical, imaging, biochemical, endoscopic, and histopathological evaluations. Traditionally, diagnostic trials establish reference standards by convening panels of clinical experts who review all accessible patient test data to arrive at a joint classification. While panel consensus systematically pools disparate evidence, the process remains inherently time-consuming, resource-intensive, and susceptible to cognitive and group biases. Researchers recognized that latent class analysis offered a promising statistical mechanism to overcome these procedural vulnerabilities, but a lack of formal comparative evaluations against conventional panel consensus has historically hindered wider adoption in diagnostic accuracy research.

To directly compare these two reference standard methodologies, the research team conducted a retrospective analysis involving a cohort of 284 patients presenting either with a newly established diagnosis of Crohn’s disease or with a suspected clinical relapse. The investigators focused specifically on terminal ileal Crohn’s disease because inflammatory activity is most prevalent and thoroughly documented within this anatomical region. For the panel consensus arm of the study, expert panels reviewed and classified disease activity using all documented evidence gathered across a six-month window following patient recruitment. This extensive multi-test dossier incorporated clinical assessments, cross-sectional imaging, endoscopic observations, histopathological tissue findings, and relevant biochemical markers. Panels were tasked with synthesizing this diverse collection of longitudinal evidence into definitive determinations regarding the presence and activity of terminal ileal disease.

In parallel, the investigators constructed and estimated a Bayesian random effects latent class model designed to objectively integrate diagnostic classifications across multiple testing modalities. The statistical model systematically incorporated classification results generated from seven established diagnostic tests: magnetic resonance imaging, ultrasound, endoscopy, histopathological assessment, C-reactive protein, faecal calprotectin, and the Harvey-Bradshaw index. By combining these seven diagnostic components into a unified mathematical structure, the Bayesian random effects formulation allowed the investigators to model complex relationships and variability among disparate tests without relying on subjective panel deliberations. The primary objective was to evaluate the degree of diagnostic agreement between the latent class model classifications and the consensus panel classifications for active terminal ileal Crohn’s disease across the patient cohort.

When evaluating the comparative performance between the statistical model and panel consensus classifications, the authors reported a positive agreement of 87 percent, with a 95 percent confidence interval spanning 83 to 91 percent. Negative agreement between the two classification standards was measured at 70 percent, with a 95 percent confidence interval ranging between 65 and 75 percent. These statistical findings demonstrated substantial overall congruence in identifying active terminal ileal Crohn’s disease, while also highlighting notable areas of classification divergence. The researchers explored the potential underlying drivers of diagnostic disagreement between the model and the panel determinations, seeking to identify methodological factors that produced differing classifications across the evaluated patient cohort.

The authors determined that diagnostic disagreement between the reference standards primarily arose from structural differences in how cross-sectional imaging data were incorporated into the classifications. The Bayesian latent class model successfully combined and heavily weighted evidence obtained from magnetic resonance imaging and ultrasound. In contrast, the expert consensus panels were unable to utilize evidence from these specific imaging modalities during their classification workflow.

Beyond classification concordance, the comparative investigation revealed substantial disparities in labor, time, and logistical demands between the two reference standard strategies. Generating the latent class analysis reference standard required approximately one statistician working for 14 working days. Conversely, establishing the panel consensus reference standard required the operational equivalent of one panelist working for 25 working days, culminating in a considerably higher financial cost. Based on these findings, the study authors concluded that latent class analysis represents an efficient, viable reference standard for active terminal ileal Crohn’s disease compared to panel consensus, while emphasizing that further comparative evaluations are needed to foster broader adoption across diagnostic trials.

Subject of Research: Comparative evaluation of latent class analysis and panel consensus in active Crohn's disease

Article Title: Improving the active Crohn’s disease reference standard: comparative evaluation of latent class analysis and panel consensus

Article References: Parry, T., Dendukuri, N., Halligan, S., Taylor, S., & Mallett, S. (2026). Improving the active Crohn’s disease reference standard: comparative evaluation of latent class analysis and panel consensus. BMC Medical Research Methodology. https://doi.org/10.1186/s12874-026-03069-8

Image Credits: AI Generated

DOI: 10.1186/s12874-026-03069-8

Keywords: Crohn's disease, latent class analysis, panel consensus, reference standard, terminal ileal Crohn's disease, Bayesian random effects model, diagnostic test accuracy, magnetic resonance imaging, ultrasound, endoscopy, histopathology, C-reactive protein, faecal calprotectin, Harvey-Bradshaw index, bowel inflammation, positive agreement, negative agreement