Faecal immunochemical test accuracy in patients referred for surveillance colonoscopy: a multi-centre cohort study

The use of colonoscopy to evaluate patients for the presence of colorectal cancer (CRC) has increased substantially in the past decade [1‐3]. This concerns both high-risk asymptomatic individuals and patients with signs or symptoms suggestive of CRC. At present, 22% of colonoscopies in the United States are performed for a family history of CRC and another 22% for surveillance after polypectomy or resection of CRC [4]. In The Netherlands, 33% of colonoscopies are performed for a personal history of adenomas or CRC or a family history of CRC [5]. With the planned introduction of a CRC screening program in 2013 in The Netherlands, not only the number of screening colonoscopies, but also the number of surveillance colonoscopies because of a personal history of colorectal neoplasia will increase [6]. Since the burden on endoscopic capacity is already high in some countries, it has been proposed to use faecal immunochemical tests (FITs) to postpone elective colonoscopy in asymptomatic high-risk subjects [7, 8]. Previous studies on guaiac-based faecal occult blood tests (g-FOBTs) have shown that the clinical use of this strategy was hampered by its relatively low sensitivity for advanced adenomas [9]. For that reason, high-risk subjects are recommended not to rely on g-FOBT screening, but to undergo colonoscopic evaluation at predetermined intervals [10‐12]. FIT, however, has been proven to be superior to g-FOBTs in detecting advanced adenomas and CRC [13‐20]. The higher sensitivity of FIT for advanced adenomas is particularly relevant since the goal of a surveillance program is to detect advanced adenomas before malignant transformation and in this way prevent CRC related death. A recent study in asymptomatic high-risk patients, showed sensitivities of 100% and 65% for the detection of CRC and all advanced neoplasia, respectively [7]. These results need confirmation in a different set of patients, preferably with a larger sample size and higher yield of advanced neoplasia, before contemplating on its use in surveillance programs. The present study evaluates the test accuracy of single and double FIT sampling to detect CRC and/or advanced adenomas in a colonoscopy-controlled high-risk population.

In the present study, test accuracy of one of the most commonly used FITs was evaluated in a large cohort of asymptomatic, high-risk individuals undergoing colonoscopy surveillance. It was found that 20% of CRCs and 72% of advanced adenomas were missed when using a single FIT sampling strategy at the lowest cut-off level. Moreover, providing a second FIT preceding colonoscopy did not result in a significant increase in sensitivity since still only 33% of all advanced neoplasia was detected. When the indication groups were separated into patients with a personal history of CRC, a personal history of adenomas and patients with a family history of CRC, a higher FIT positivity rate and a higher yield of advanced neoplasia was observed in the first two groups. Sensitivity, however, did not significantly differ between these three indication groups, showing that a once-only FIT sampling strategy is inadequate in detecting advanced neoplasia in either group of high-risk, asymptomatic individuals.

Limited data are available on the performance of FITs in asymptomatic high-risk patients. Three studies have assessed the efficacy of a first generation, qualitative FIT (OC light and Hemeselect) in volunteer first-degree relatives of patients with CRC, where sensitivities ranged from 50–83% for detection of advanced neoplasia [8, 30, 31]. However, small sample sizes and small numbers of target lesions hamper the generalization of these results. In patients with a personal history of CRC, one study showed a 100% sensitivity for detecting 9 recurrent CRCs [32]. These results were obtained from stools collected with digital examination and therefore not representative for at home FIT sampling. Moreover, test characteristics for advanced adenomas were not available. Another study, using a qualitative FIT in a larger population, showed sensitivities of 70% and 44% for detecting CRC and advanced adenoma, respectively [33]. Yet, incorrect criteria for defining advanced adenoma were used. In general, test characteristics from qualitative FITs should be interpreted with caution since the test performance can vary greatly between the different qualitative FITs, indicating that quality assurance is an issue [34].

The most promising results were shown in a recent study, providing three FITs preceding colonoscopy in asymptomatic high-risk patients [7]. Sensitivities of 100% and 65% for the detection of CRC and all advanced neoplasia, respectively, were found using the same FIT as was used in the present study at the same cut-off level [7]. Unfortunately, we were not able to confirm these promising findings in our study with a comparable sample size and a larger number of target lesions. Multiplicity of testing (three versus two FIT samples) and different prevalences of advanced adenomas (higher in the present study) may partly explain the differences between the two studies in FIT performance. Other potential explanations might be that the study by Hazazi et al. had a better defined high-risk population with more CRCs located in the distal colon or larger pedunculated advanced adenomas with a higher tendency to bleed or a better quality of bowel preparation since 23% of colonoscopies was rated “fair” bowel preparation in the present study. Moreover, in the present study a high percentage of patients have received their colonoscopy at an earlier time interval than recommended by the guidelines which might have resulted in smaller sized advanced adenomas.

Periodically performing non-invasive FIT sampling in the surveillance of high-risk individuals in order to triage individuals for invasive colonoscopy sounds appealing, as this could lead to a better utilization of colonoscopy as both a diagnostic and therapeutic procedure. However, when considering such an alternative surveillance scheme, a high sensitivity of FIT is important. Specificity is less of a concern, because at present colonoscopy is performed in all these patients. Even at the most sensitive cut-off level of 50 ng/ml in single and double FIT sampling, the majority of advanced neoplasia was missed in the present study. On the one hand it is debatable whether missing advanced adenomas is clinically relevant, since they could be detected in subsequent surveillance rounds (either with FIT or colonoscopy) while still being in a curable stage. On the other hand, a delay in detecting CRC can result in progression into advanced stage disease. Yet, the risk of progression from adenoma to carcinoma in the majority of patients with a family history of CRC and in the majority of patients with a personal history of adenomas may be limited and comparable to the risk in the general population [10, 35, 36]. Particularly in patients with only a few small tubular adenomas in the past or a non-significant family history of CRC, multiple rounds of FIT sampling may be a good alternative to colonoscopic surveillance. The present study was not designed to provide information on interval FIT testing in a surveillance program. Recent data from Australia showed that multiple rounds of FIT sampling within an existing surveillance program aided the detection of advanced neoplasia [29]. In that particular study, it was clearly demonstrated that repeated testing results in a compounding of sensitivity. It was also shown that in those patients who returned a negative FIT in multiple rounds of testing, the chance of finding advanced neoplasia was significantly reduced [29]. When we model cumulative detection in the same manner it can be calculated that a once-only FIT sensitivity of 80% for CRC compounds to 99.2% with 3 rounds of testing. Similarly for advanced adenoma detection, modelling a 28% once-only sensitivity compounds over 5 rounds of testing to 81%, which may be acceptable for the detection of an advanced adenoma. Although these computed sensitivities should be interpreted with caution, it allows us to explore the effect of multiple rounds of testing. Such a strategy of repeated FIT sampling in selected individuals may hold potential to postpone invasive colonoscopy and create more colonoscopic capacity. The effects of earlier diagnosis, particularly of advanced adenomas, on survival remain elusive and might only be clarified in a long-term prospective randomized trial. Since the yield of CRC is generally low in surveillance programs, it is questionable whether such a study is feasible.

There are some concerns that need to be addressed for proper interpretation of our results. First, the number of target lesions is relatively low. Although this study has a large sample size and the percentage of target lesions is higher than in most other studies, the absolute number of cancers is low [7, 8, 30‐33]. Potential explanations for the low number of CRCs may be over-usage of colonoscopy surveillance and inappropriate colonoscopy referral for patients with a non-significant family history of CRC [27, 28]. Second, there was substantial heterogeneity in the group of patients with a personal history of adenomas and a family history of CRC, making it difficult to draw firm conclusions for each subgroup.

The present study of high-risk individuals for colorectal neoplasia, shows that 67–70% of all advanced neoplasia are not detected by single or double FIT sampling before elective colonoscopy. Future studies should focus on multiple rounds of FIT sampling in different, high-risk patient groups in order to create a sensitive strategy to detect advanced neoplasia and potentially alleviate the burden on endoscopic capacity.