Fecal immunochemical test in colorectal cancer screening : impact of screening strategy and gender on colonoscopy findings, missed lesions and costs

Abstract

In Sweden, colorectal cancer (CRC) is the second and third most common type of cancer in men and women respectively. The relative five-year survival is approximately 65%, but prognosis is better if diagnosed at an early stage of disease. Fecal Immunochemical Test (FIT) detects blood in the stool and is used in screening, and individuals with a positive test are referred for colonoscopy. Several studies have indicated a lower sensitivity for advanced neoplasia (AN; CRC and advanced adenomas) in women as compared to men. In the Stockholm-Gotland region, population-based screening was initiated in 2008, and from 2015 FIT screening with lower cut-off levels for a positive test in women (40μg/g) than in men (80μg/g) was applied. The aim of this thesis was to increase the knowledge of the performance of FIT in an average-risk Swedish screening population and to explore a gender-specific screening strategy regarding colonoscopy findings, screening costs and interval CRC (IC; CRCs detected between two screening rounds after a negative screening episode).

In Paper I the performance of two FIT samples at different cut-off levels was evaluated in a FIT-positive cohort from the randomized controlled study Screening of Swedish Colons (SCREESCO). The FIT level was significantly higher in individuals with CRC and AA as compared to other participants and correlated to adenoma size. CRC detection increased with lower cut-off level and multiple samples and was significantly higher with one sample at a low cut-off level than two samples at a higher cut-off level. The positive predictive value (PPV) for AA was significantly higher in men than in women for one and two samples at cut-off levels <40μg/g but PPV for CRC was equal between genders at all cut-offs and number of samples.

In paper II the accuracy of two FIT samples at different cut-off levels were evaluated in a colonoscopy cohort from the SCREESCO study. Sensitivity and specificity for AN ranged from 7-26% and 89-99% respectively depending on the number of samples and the cut-off level. There was no gain in sensitivity using two samples instead of one, for any of the cut-off levels. Specificity was significantly higher with one sample as compared to any of the two samples, at the lowest cut-off levels. In the 225 participants with adenomas, pedunculated shape and high-risk dysplasia was independently associated with FIT positivity at cut-off ≥10μg/g for any of the two samples. Sensitivity for AA was significantly higher in men vs women, but specificity was similar between genders.

In paper III the Stockholm-Gotland population-based screening program was evaluated regarding colonoscopy findings and costs in a screening cohort from 2015-2017. CRC was found in significantly more men than women, 138 (8.3%) vs 120 (5.8%). A normal colonoscopy was more common in women than in men (24% vs 17%, p-value <0.05). Had the cut-off level been 80μg/g in both genders, the PPV for CRC was estimated to be equalized between genders. However, in women with CRC, 28 (23%) had FIT level of 40-79μg/g and would thus have remained undetected at cut-off level 80μg/g in both genders. The gender-specific screening strategy was estimated to be 16% more expensive than the gender-equal strategy, corresponding to a 3% increment in costs per detected CRC.

In paper IV the ICs were evaluated in the first round of the Stockholm-Gotland population-based screening program and compared to the experienced incidence rate (EIR) prior to screening implementation. In the cohort 124 FIT ICs, 7 colonoscopy ICs, 3 ICs in individuals non-compliant to colonoscopy and 177 CRCs in non-participants were detected within 2 years. Test sensitivity was 0.75 in women and 0.62 in men (p-value 0.011), but would have been equal, had cut-off level been 80μg/g in both genders. The IC rate was significantly higher in men than in women, 12.6 vs 6.0 per 10,000 negatives. The rate ratio of the IC incidence/EIR was 0.30-0.44 and non-significantly lower in the women as compared to the men in each age group. In all the 568 CRCs including those in non-participants, proximal localization was significantly more common in women (42%) than in men (29%).