Editorial: Right-Sizing Colonoscopy Referrals for Patients With Possible Signs and Symptoms of Colorectal Cancer

Abstract

Colorectal cancer (CRC) mortality reduction strategies may include primary prevention through risk reduction, early detection and prevention through screening, and timely diagnosis among individuals with potential signs and symptoms of CRC [1]. Although lower gastrointestinal symptoms are common among patients with CRC, they have poor positive predictive values of 0.2%–0.6% [2, 3]. Thus, identification of individuals at high risk for CRC based on signs/symptoms alone is a challenge.

As an adjunct to symptoms, the faecal immunochemical test (FIT) can correctly rule out CRC in 75%–80% of symptomatic patients at a faecal haemoglobin (f-Hb) cut-off of 10 μg Hb/g faeces, although specificity may be as low as 48.7% [4]. In settings with constrained colonoscopy resources, and among populations with low CRC prevalence (such as younger age adults), suboptimal specificity has significant implications for health system and patient burdens.

Prediction models combining FIT, patient characteristics and other tests may better optimise CRC risk stratification. However, a prior review has suggested that previous studies require extension with better methodologic rigour, including use of model derivation and validation datasets [5].

To address this evidence gap, Crooks et al. developed and validated a prediction model to estimate the 1-year risk of CRC for symptomatic patients by combining f-Hb, age, sex and blood cell indices—specifically platelet count and mean corpuscular volume (the COLOFIT model) [6]. A cohort of 34,435 adults age ≥ 18 referred between November 2017 and November 2021 with potential CRC symptoms and exposure to FIT was used to derive a prediction model for CRC diagnosis (the primary outcome), with CRC ascertained based on cancer registry data within 1 year of follow-up. Validation was performed with a December 2021–November 2022 cohort of 21,012. A net-benefit analysis compared trade-offs between CRC detection and colonoscopies generated based on COLOFIT versus a referral threshold of f-Hb ≥ 10 μg Hb/g faeces in isolation.

At a similar 0.6% CRC risk referral threshold as f-Hb ≥ 10 μg alone, COLOFIT had CRC sensitivities of 91.6% and 92.3%, and specificities of 79.7% and 82.2% for the derivation and validation datasets, respectively. At these referral thresholds, net-benefit analysis suggested a minimal 9 per 100,000 reduction in CRC case detection with a substantial 18% reduction in colonoscopies required for COLOFIT versus only FIT, confirming the COLOFIT model could achieve similar population benefits but with lower draw on colonoscopy resources.

Limitations include use of only registry—rather than colonoscopy—follow-up for CRC diagnosis, and potential for patients with measured blood count results to have had more severe symptoms and higher CRC risk than those without, potentially resulting in respective risks for spectrum bias and incorporation bias towards observing better model performance.

Nonetheless, this study demonstrates that a model incorporating clinical factors beyond only f-Hb has potential to achieve similar CRC detection with lower colonoscopy use than a FIT-based triage system alone. Future research should evaluate implementation of the COLOFIT model as a strategy for right-sizing colonoscopy referrals for patients with possible signs and symptoms of CRC.

Nicole P. Mirabadi: conceptualization, writing – review and editing. Samir Gupta: conceptualization, writing – review and editing, supervision.

The authors declare no conflicts of interest.

This article is linked to Crooks et al paper. To view this article, visit https://doi.org/10.1111/apt.18459.