Cost-effectiveness of FIT and a FIT-based model to optimise symptomatic diagnosis of colorectal cancer: health economic modelling for the COLOFIT project

Abstract

Introduction: Fecal immunochemical testing (FIT) at a threshold of 10 mg haemaglobin (Hb)/g is used in English primary care to prioritise urgent referral for colorectal cancer (CRC) investigation in symptomatic patients. The COLOFIT algorithm, based on FIT score, age, sex and blood results, performs better than FIT alone for identifying CRC. We assessed the cost-effectiveness of COLOFIT compared with FIT and investigated optimal risk thresholds.

Methods: An individual patient-level simulation model was developed, with synthetic populations constructed from data used to validate COLOFIT. Referral criteria based on different FIT scores and COLOFIT-assessed risk thresholds were modelled using probabilistic and scenario analyses. Outcomes included costs, quality-adjusted life years (QALYs) and cost-effectiveness measured using incremental net monetary benefit (INMB) based on a willingness to pay threshold of £20 000/QALY.

Results: COLOFIT at a CRC risk threshold of 0.64% has a 98% probability of being more cost-effective than FIT 10 mg Hb/g (INMB is £5.67 per person), while detecting similar numbers of cancers. Cost-effectiveness is achieved by cost savings from reducing referrals outweighing QALYs lost through reorienting expedited CRC diagnoses from younger (<50) to older (≥70) patients. Cost-effectiveness improves as risk thresholds rise. High structural uncertainty around cancer progression during diagnostic delay and diagnosis of other serious bowel diseases considerably affects cost-effectiveness.

Conclusions: COLOFIT is likely to be more cost-effective than FIT alone and could help alleviate pressure on diagnostic services. However, strategies to improve diagnosis in the under 50s would be necessary to mitigate potential harm. Further research should assess how COLOFIT impacts cancer survival and diagnosis of other serious bowel diseases.

Keywords: Risk Assessment; economics; statistics and numerical data.

Figure 1. Model diagram showing the structures of the short-term diagnostic model and the long-term Markov style model. Boxes in the short-term model represent events, and boxes in the long-term model represent health states. CRC, colorectal cancer; CTC, CT colonography; FIT, fecal immunochemical testing; GP, general practitioner; IBD, inflammatory bowel disease; TNM, tumour, node, metastases; USC, urgent suspected cancer.

Figure 2. Base case results comparing FIT and COLOFIT at different thresholds. (A) Distribution of PSA results on the cost-effectiveness plane for the FIT10:COLOFIT 0.64% comparison. (B) Cost-effectiveness acceptability curve indicating probability cost-effective at different WTP thresholds for the FIT10:COLOFIT 0.64% comparison. Plots of (C) USC referrals; (D) incremental net monetary benefit; (E) QALYs and (F) costs against USC CRC diagnoses for different thresholds of FIT and COLOFIT compared with sending all urgently. Graphs indicating net monetary benefit at different thresholds of (G) COLOFIT and (H) FIT compared with sending all urgently. CRC, colorectal cancer; FIT, fecal immunochemical testing; PSA, probabilistic sensitivity analysis; QALYs, quality-adjusted life years; USC, urgent suspected cancer; WTP, willingness to pay.

Figure 3. Plots showing FIT and COLOFIT at different thresholds compared against sending all urgently in subgroups defined by age and sex for (A, B) number of USC referrals plotted against all referrals; (C, D) number of USC CRC diagnoses; (E, F) incremental net monetary benefit; (G) QALYs and (H) costs. (C–H) All plotted against all USC CRC diagnoses. CRC, colorectal cancer; FIT, fecal immunochemical testing; QALYs, quality-adjusted life years; USC, urgent suspected cancer.