Clinical assessment to determine the risk of bowel cancer using Symptoms, Age, Mass and Iron deficiency anaemia (SAMI)

Abstract

Background: The aim of this study was to identify characteristics with independent predictive value for bowel cancer for use in the clinical assessment of patients attending colorectal outpatient clinics. Methods: This was a 22-year (1986–2007) retrospective cohort analysis of data collected prospectively from patients who attended colorectal surgical outpatient clinics in Portsmouth. The data set was split randomly into two groups of patients to generate and validate a predictive model. Multivariable logistic regression was used to create and validate a system to predict outcome. Receiver operating characteristic (ROC) curves and Hosmer–Lemeshow test were used to evaluate the model's predictive capability. The likelihood of bowel cancer was expressed as the odds ratio (OR). Results: Data from 29 005 patients were analysed. Discrimination of the model for bowel cancer was high in the development (C-statistic 0·87, 95 per cent c.i. 0·85 to 0·88) and validation (C-statistic 0·86, 0·84 to 0·87) groups. The most important co-variables in the final model were: age (OR 3·17–27·10), rectal (OR 31·48) or abdominal (OR 1·83–8·45) mass, iron deficiency anaemia (IDA) (OR 4·42–8·38), rectal bleeding and change in bowel habit in combination (OR 5·37), change in bowel habit without rectal bleeding, with or without abdominal pain (OR 2·12–2·52), and rectal bleeding with no perianal symptoms and without change in bowel habit (OR 2·91). Some 91·5 per cent of bowel cancers presented with these characteristics, 40·4 per cent with a mass and/or IDA. In patients with at least one of these characteristics the overall risk of having cancer was 10·0 (range 6·5–50·4) per cent, compared with 1·1 (0·3–2·3) per cent in patients without them. Conclusion: A clinical assessment that systematically identifies or excludes four symptom–age combinations, a mass and IDA (SAMI) stratifies patients as having a low and higher risk of having bowel cancer. This could improve patient selection for referral and investigation.