This study seems to be the first aimed at determining whether f-Hb found by a quantitative FIT can assist in deciding which patients with lower abdominal symptoms would benefit from endoscopy [51]. Among 739 invited patients, FIT and endoscopy were completed by 280 (median age 63 years, 59.6% women). Six (2.1%) participants had CRC, 23 (8.2%) had AA, 31 (11.1%), non-advanced adenoma (NAA) and 26 (9.3%) IBD as the most serious diagnosis. As with findings in screening, f-Hb was clearly related to severity of colorectal disease since those with CRC had a median f-Hb of >200 μg Hb/g faeces and those with CRC + AA + IBD having a median f-Hb of 15 μg Hb/g faeces, which were both significantly higher than that of all remaining participants without SCD. The NPV using a cut-off f-Hb of 10 μg Hb/g faeces were 100%, 94.4% and 93.9% for CRC, AA and IBD respectively. The AUC for all four clinical groups examined, with the more important colorectal diseases classified as positive, ranged from 0.734 to 0.671; these AUC suggest only a “fair” to “poor” rule-in diagnostic test. However, the high NPV, particularly for the group that is considered of most clinical interest, that is SCD (CRC + AA + IBD), was 88.1%. In consequence, a negative FIT result with f-Hb <10 μg Hb/g faeces means that SCD is unlikely to be present and, when used as a rule-out test, could facilitate patient reassurance of patients and saving of colonoscopy resources.

This study followed up that described above in a different part of Scotland and with a different FIT system [52]. Again designed to determine whether patients with lower abdominal symptoms can be investigated quickly using f-Hb, 909 consecutive patients referred from primary care for colonoscopy were invited: 507 submitted samples for f-Hb and a colonoscopy was completed in 484 patients. As with the study of McDonald et al. [51], those with CRC, AA or IBD had higher f-Hb than the group of 243 with normal colonoscopy plus the 196 patients with less significant clinical findings. Again, using a f-Hb cut-off of 10 μg Hb/g faeces, for the group with SCD, the test was a poor rule-in test with low sensitivity (68.9%) and PPV (26.3%,), but the NPV of 96.2% confirmed that FIT was a good rule-out test for SCD. The clinical characteristics at different f-Hb cut-offs were examined and, as expected, as the cut-off f-Hb was increased, sensitivity fell and specificity rose; while the PPV rose, the NPV remained very high at all f-Hb cut-offs. Interestingly, the 11 (2.2%) patients with CRC all had f-Hb >190 μg Hb/g faeces and sensitivity was therefore 100% for CRC, as found by McDonald et al. [51]. It was suggested that it might be that FIT could be of considerable value as either a rule-in for CRC on its own or a rule-out investigation for SCD.

This investigation studied the diagnostic accuracies of f-Hb and f-C in a group of symptomatic patients [53]; only the f-Hb component of this work will be addressed in this section of the Chapter. The PPV of referral symptoms for a diagnosis of CRC were not high and the values for specific symptoms were: palpable mass: 50.0%, weight loss: 14.3%, anaemia: 9.0%, rectal bleeding: 4.3%, abdominal pain: 3.6%, diarrhoea: 2.4%, and altered bowel habit: 2.2%. Rectal bleeding had a PPV of 21.0% for any SCD, but only one-third of these had undetectable f-Hb. As expected from the known overlap of symptoms in those with and without SCD, the most common findings at colonoscopy were: normal in 241 (33.2%), diverticular disease in 190 (25.2%), haemorrhoids in 98 (13.0%), NAA in 65 (8.6%), AA in 41 (5.4%), IBD in 34 (4.5%) and CRC in 28 patients (3.7%). In total, 1043 patients returned samples and f-Hb was detectable, that is greater than 0 μg Hb/g faeces, in 57.6%; 755 patients (median age 64 years, 54.6% women) returned faecal samples and completed colonic investigations and 103 patients were found to have SCD. Using a cut-off of “detectable” f-Hb, NPV were 100%, 97.8% and 98.4% for CRC, AA and IBD, respectively. It was concluded that undetectable f-Hb is a good rule-out test for SCD and could guide who requires further investigation. However, in laboratory medicine, the requirement is to fully conform to ISO 15189—“medical laboratories—particular requirements for quality and competence”—means that only results greater than the lower limit of the analytical working range laid down by the manufacturer, best described as the “limit of quantitation”, can be reported as concentrations [58]. Thus, while interesting for research studies, a cut-off of detectable f-Hb cannot be used in clinical practice, and only results for the OC-Sensor of <10 μg Hb/g faeces can be considered undetectable and only those ≥10 μg Hb/g faeces can be reported as numerical data. However, Mowat et al. also give comprehensive data using 10 μg Hb/g faeces as the cut-off. NPV for CRC, AA, IBD and SCD were 99.5%, 96.5%, 98.4% and 94.4%.: in contrast to previous studies in Scotland [51, 52], cases of CRC (3 of 28) would be missed using this cut-of as would 20 of 40 of AA and 9 of 34 of IBD, that is, 32 of 102 of those with SCD. Most interestingly, the three missed cases of CRC were all women (personal communication): this may reflect that fact that f-Hb is lower in women than in men [44, 45], possibly due to slower transit time of faeces passing through the colon [59].

In this study [54], a f-Hb cut-off of 20 μg Hb/g faeces was used, rather higher than that in the three studies done in Scotland [51–53]; the rationale probably being that this is the most common f-Hb cut-off used in asymptomatic screening programmes. The novel aspect of this study is that FIT to detect CRC was compared with the then current NICE [14] and SIGN [15] referral guidelines. In 787 symptomatic patients referred for colonoscopy, patients were assessed to see if they met NICE and SIGN referral criteria and all patients collected one sample for an estimate of f-Hb. CRC was detected in 97 (12.3%) patients; 241 (30.6%) had f-Hb ≥20 μg Hb/g faeces and 300 (38.1%) and 473 (60.1%) met NICE and SIGN referral criteria. As shown in other studies discussed earlier, the median f-Hb was found to be statistically significantly higher in individuals with a CRC than the others. The important finding here is that f-Hb had a higher sensitivity for CRC (87.6%) than NICE (61.9%) and SIGN criteria (82.5%). The specificity of f-Hb was also higher than NICE and SIGN criteria (77.4%, 65.2% and 42.7%). It was concluded that f-Hb is more accurate for the detection of CRC than the then current NICE and SIGN referral criteria in symptomatic patients referred for colonoscopy. In view of the overlap of symptoms in patients with and without SCD, this is probably hardly surprising, since NICE and SIGN referral criteria are based on age and symptoms and other clinical information rather than f-Hb. However, combinations of symptoms and f-Hb might have advantages and this will be discussed later in this Chapter. Finally, the authors of this study concentrated on the use of f-Hb as a rule-in test for CRC, but there are data on the NPV of f-Hb and NICE and SIGN referral criteria for both CRC and advanced neoplasia (AN), defined as CRC plus AA: these were 97.8%, 92.4% and 94.5% respectively for CRC and 90.8%, 82.1% and 83.6% for AN respectively. Thus, it was actually demonstrated that f-Hb is a good rule-out test for AN.

Auge et al. [55] provided a unique perspective on use of FIT in assessment of the symptomatic since two samples were obtained on each patient for FIT. The aims of this study were to evaluate the diagnostic yield for what was termed advanced colorectal neoplasia (ACRN), equivalent to CRC plus AA, in symptomatic patients, assessing the f-Hb of the first of two samples (FIT/1) and the higher concentration of two FIT samples (FIT/max). Samples were collected from two consecutive bowel motions from 208 symptomatic patients who required colonoscopy. Patients were categorised into two groups: patients with any ACRN and those with other diagnoses or normal colonoscopy. Colonoscopy detected ACRN in 29 patients. FIT/1 and FIT/max were significantly higher than in patients with NAA, other findings and normal colonoscopy, confirming the now well-established relationship between f-Hb and colorectal disease severity, This study also showed that higher f-Hb are found in men with lower abdominal symptoms compared with women, consistent with other studies [44–47], Similarly, a higher clinical sensitivity and PPV for ACRN was observed in men than in women and the NPV was lower in men, which perhaps implies that different f-Hb cut-offs should be used for the different sexes. The AUC of FIT/1 and FIT/max were 0.71 and 0.69, respectively, again showing that FIT are less than good tests for detection of disease. However, undetectable FIT/1 gave NPV of 95.0%, but increasing the FIT/1 cut-off to 10 μg Hb/g faeces decreased the NPV to 89.2%. Similar results were obtained using FIT/max with a cut-off of 20 μg Hb/g faeces, providing NPV of 89.3% and 89.0%. The authors concluded that undetectable FIT is a good strategy to rule out ACRN in symptomatic patients, but the problems of using this and complying with ISO 15189 standards in routine practice have already been discussed. Considering which of one or two samples is better, it was shown that the diagnostic yield for a rule-in test of collecting two samples for FIT can be achieved with one sample, but a lower faecal f-Hb cut-off is required. Interestingly, the FIT/max result was not concordant with the FIT/1 result in 39.2% of patients: this might indicate that f-Hb has considerable within-subject variation and this would be very worthy of investigation as has been done recently for f-C [60].

This was the second study done comparing quantitative FIT with NICE [14] and SIGN [15] guidelines for referral [56]. The authors prospectively studied 1054 symptomatic patients referred for colonoscopy from both primary and secondary healthcare. NICE and SIGN guidelines detected 46.7% and 43.3% of cases of CRC while f-Hb ≥15 μg Hb/g detected 96.7% of cases. Male sex, age and f-Hb ≥10 μg Hb/g were independent predictive factors of advanced neoplasia. As found in all studies, symptoms often used as referral criteria were poor predictors of both CRC and AN. The performance characteristics of NICE, SIGN and f-Hb concentration ≥ 15 μg Hb/g for the detection of AN were: sensitivity: 38.3%, 36.1% and 57.1% and specificity: 71.8%, 69.5 and 86.6%, respectively. This study explored the known relationship between f-Hb and severity of disease in considerable detail. AUC for f-Hb in the detection of CRC and AN were 0.94 and 0.76 and the optimal f-Hb cut-off for CRC and AN detection were ≥15 μg Hb/g and ≥10 μg Hb/g, respectively. If f-Hb ≥ 15 μg Hb/g was used as the criterion for referral, only 19.2% of the patients would need urgent assessment for detection of 96.6% of CRC, one of 30 patients being missed. At a cut-off f-Hb of ≥10 μg Hb/g, 22.5% of the patients would need urgent assessment for detection of 61.6% of the AN. It was concluded that a FIT-based strategy performs better than the then current high-risk symptoms-based approaches for fast-tracking referrals for investigation of suspected cancer referrals. The authors also considered that, with NPV of 99.9% for CRC and 93.0% for AN at a f-Hb cut-off of 10 μg Hb/g faeces, f-Hb could also provide a single and excellent rule-out test, thus sparing those with lower abdominal symptoms who do not have SCD from undergoing a colonoscopy. This would save many referrals to colonoscopy and would allow fast-track referrals to be seen more quickly. For the first time, risk-scoring involving f-Hb in assessment of the symptomatic, along with sex and age, was investigated and it was shown that this could accurately estimate the risk of AN. Risk-scoring will be discussed later in this Chapter.

Similar to Mowat et al. [52], the aim was to investigate f-Hb and f-C to identify patients for referral for colonic investigation who are most likely to have CRC; here, only the data on f-Hb will be discussed. This study involved patients recruited via the “two-week wait” clinics rather than directly in primary healthcare. Of 430 study participants, median age 67 years, 176 (64.1%), had change in bowel habit, 185 (43.0%) rectal bleeding, 129 (30.0%) abdominal pain, 74 (17.2%) anaemia, 68 (15.8%) weight loss and 95 (22.1%) family history of CRC. Twenty-four patients with CRC and one patient with high grade dysplastic adenoma were grouped together as cancer. At a cut-off of 7 μg Hb/g faeces, f-Hb returned a sensitivity of 84% and specificity of 93% for CRC detection with an AUC of 0.94 and NPV of 99%. For adenoma detection, with no other pathology on colonoscopy or CT, f-Hb had a sensitivity of 69% and specificity of 56% with AUC of 0.70 and NPV of 94%. As found in all other studies described here, the evidence was that f-Hb alone has considerable potential for use in risk stratifying symptomatic patients.

It would be of significant value for a meta-analysis to be performed on quantitative FIT in assessment of the symptomatic, but that is beyond the scope of this Chapter. However, the results of the seven published studies known to us to date provide evidence that unequivocally demonstrates the following:

It might appear that use of central laboratory-based quantitative FIT would cause some, albeit small, delay in reporting of f-Hb results as compared to POCT qualitative FIT. This might delay diagnosis and, more importantly, might cause some drop out and loss of patients from pathways after presentation since they might not wish to undertake faecal collection, in spite that quantitative FIT have easy to use, hygienic collection devices and only one sample is needed. It would therefore be of interest to undertake further research comparing qualitative and quantitative FIT. Further, in view of the contentious issue of application of tests for occult blood in faeces in patients presenting in primary care, NICE established a Diagnostics Assessment Committee on quantitative faecal immunochemical tests to guide referral for colorectal cancer in primary care; the final guidance (DG30) has been published very recently [61]: The recommendations include the following:

It is pleasing to see that, in general, the guidelines are in accord with the material in this Chapter.

As well described in a review by Vega et al. [69] on “CRC diagnosis: pitfalls and opportunities”, many factors affect incidence and mortality of CRC, including sex, age, socio-economic status, diet, alcohol consumption and tobacco use. Since the same spectrum of symptoms occur in those with and without SCD, it is attractive to think that risk-scoring systems which include both factors affecting incidence and mortality and/or symptoms might be effective. Williams et al. recently reviewed risk prediction models which combine multiple risk factors and symptoms that have the potential to improve timely diagnosis [70]. The aim was to systematically identify and compare the performance of models that predict the risk of primary CRC among symptomatic individuals; 18 papers describing 15 risk models were included. Models with good discrimination had been developed in both primary and secondary care populations and most contained variables which are easily obtainable in a single consultation. It was suggested that further research was needed to assess clinical utility before incorporation into practice. Some include results of gFOBT but, in this Chapter, only the few models to date which include f-Hb are discussed.

The study from The Netherlands using POCT FIT that evaluated the diagnostic value of history taking, physical examination, and simple blood testing and, subsequently, the added value of the POCT FIT, in that order, has been discussed previously [37], but models using quantitative estimates of f-Hb are of more current interest.

A very simple model was developed by Rodríguez-Alonso et al. [56] to provide a risk score for AN. The model incorporates age in five groups, sex and f-Hb as <10 μg Hb/g faeces or ≥10 μg Hb/g faeces. The risk score has a range of 0–11 points based on the sum of the scores according to the presence or absence of risk factors (see Table 3.2). The AUC for the risk score was 0.79. The internal validity of the model was assessed by a split sample procedure with 680 individuals (67.8%) in the training set, of whom 91 had AN, and 323 individuals (32.2%) in the validation set, of whom 42 had AN. Good agreement was found between the risk of AN predicted by the model and the observed prevalence of AN. The optimal cut-point for the score derived from the whole study population was ≥5, with 75.9% sensitivity and 72.0% specificity. Identical values were found for the score derived from the training set. The corresponding values in the validation set were 88.1% sensitivity and 63.3% specificity.