Skip to main content

Table 3 Test performance as measured by positive and negative predictive value (PPV, NPV), sensitivity, specificity, positive FITs per cancer detected, and cancer miss rate per 10,000 negative tests. FIT alone and model-based approach

From: Combining faecal immunochemical testing with blood test results for colorectal cancer risk stratification: a consecutive cohort of 16,604 patients presenting to primary care

Test criteria PPV (95% CI) NPV (95% CI) Sensitivity (95% CI) Specificity (95% CI) Positive FITs to detect one cancer  Cancer miss rate per 10,000 negative tests
FIT alone
FIT≥10 µg Hb/g 8.4% (7.1, 9.9) 99.9% (99.9, 100) 92.1% (86.4, 95.5) 91.5% (91.1, 91.9) 12 7
FIT≥10 µg Hb/g In subset with serum ferritin & C-reactive proteina 8.8% (6.9, 11.1) 99.9% (99.8, 100) 93.8% (85, 97.5) 90.0% (89.2, 90.7) 11 7
Multivariable model including FIT
Model Aa: continuous FIT PLUS continuous variables selected by stepwise procedure: age, sex, serum ferritin, platelets, c-reactive protein 1.7% (1.4, 2.2) 99.9% (99.6, 99.9) 93.8% (85, 97.5) 45.9% (44.7, 47.1) 57 14
Model B: FIT≥10µg Hb/g PLUS categorical variables selected by stepwise procedure: sex, low mean cell volume 7.4% (6.2, 8.7) 99.9% (99.9, 100) 93.5% (88.2, 96.6) 90.1% (89.6, 90.5) 14 6
Model C: FIT Spline PLUS categorical variables selected by stepwise procedure: sex, low mean cell volume 8.4% (7.1, 9.9) 99.9% (99.9, 100) 92.1% (86.4, 95.5) 91.5% (91.1, 91.9) 12 7
  1. In the model-based approach, a fixed threshold for a positive test was set to achieve the sensitivity of a FIT threshold of 10 in the FIT alone approach.
  2. CI Confidence interval
  3. a Serum ferritin and c-reactive protein tests were only conducted for a subset of patients. The model was applied to patients with non-missing values for both tests