Finding the needle in the haystack: the diagnostic accuracy of the faecal immunochemical test for colorectal cancer in younger symptomatic patients

Detection of early onset colorectal cancer is challenging, and remains a rare diagnosis amongst younger people with gastrointestinal symptoms. We investigated whether faecal immunochemical testing (FIT) could identify younger patients at higher risk of colorectal cancer or serious bowel disease including colorectal cancer, inflammatory bowel disease or advanced adenomas.


INTRODUC TI ON
The identification of younger people at risk of colorectal cancer (CRC) represents a significant challenge in primary and secondary care.Whilst the absolute prevalence of CRC in younger people remains low, the disease incidence in this population is unfortunately increasing.Accurate risk assessment remains elusive given the poor predictive value of symptoms in younger people and as such novel approaches to risk stratification may be of significant clinical utility.
The incidence of early onset colorectal cancer (EOCRC), defined as a diagnosis of CRC under age 50 years, has significantly increased in recent years [1][2][3][4][5][6].A contemporary global analysis [1] has reported an increase in the incidence of EOCRC in 19 out of 36 countries confirming similar data previously reported from countries including the UK [7], USA [3], Norway [4], Australia [5] and Canada [6].Although the incidence of EOCRC is increasing, the absolute risk of CRC in this population remains low, with an incidence rate of 7.6 per 100 000 in patients in England aged between 30 and 39, or 2.8 per 100 000 in patients aged between 20 and 29 compared to 165.9 per 100 000 in patients aged between 60 and 69 [8].
The poor predictive value of gastrointestinal symptoms provides an additional challenge in this population.Bowel symptoms in younger patients may be attributed by clinicians to common benign colorectal disease, such as rectal bleeding from haemorrhoids or abdominal pain from irritable bowel syndrome [9][10][11].Unfortunately, younger patients who report non-specific symptoms but who have an underlying CRC may require multiple visits before referral for investigation, resulting in longer time to diagnosis [12][13][14][15].A survey of EOCRC patients in the UK reported that 43% of patients required three or more visits to their primary care physician (PCP) before referral to secondary care for further investigation, and that 65% were initially given an incorrect diagnosis [11].Patients themselves may not suspect their bowel symptoms may be due to cancer: in the same survey, 75% of patients did not suspect their symptoms were due to CRC, and 30% delayed seeing their PCP for up to 3 months [11].The interval from referral to diagnosis in younger patients is significantly longer in younger patients in national UK data, probably due to an absence of criteria indicating urgent referral for suspected cancer [12].National cohort data also report that EOCRC more commonly presents as an emergency, at a more advanced stage of diagnosis [12,14,16,17], although younger patients tend to have better survival than older patients [18,19].It has been suggested that advanced-stage disease at presentation in younger patients may be due to delay in diagnosis, but there is also evidence that this may be due to a higher prevalence of biologically 'aggressive' tumours [16].
Thus the diagnosis of CRC in younger patients is difficult, and improvements in early diagnosis in this population may improve outcomes.We postulate that faecal immunochemical testing (FIT) may assist PCPs to identify which younger patients with higher risk symptoms may require colonoscopy or other diagnostics.FIT is a noninvasive, quantitative immunoassay which detects the globin moiety of haemoglobin in faeces and can reliably measure the faecal haemoglobin concentration (f-Hb) to the nearest microgram of haemoglobin per gram of faeces (µg/g).To our knowledge, there are no published data specifically evaluating the diagnostic accuracy of FIT in younger patients referred urgently with suspected CRC symptoms.

ME THODS
The primary end-point of this analysis was to establish the sensitivity of FIT for CRC in younger patients with symptoms of suspected CRC.The secondary end-points were to establish the diagnostic accuracy of FIT for CRC, advanced neoplasia (AN) (i.e., CRC and advanced adenoma [AA]) and serious bowel disease (SBD) (CRC, AA and inflammatory bowel disease [IBD]) in younger patients.
The NICE FIT study was a multicentre, ethically approved, diagnostic accuracy study of FIT in English patients referred urgently from primary care with symptoms of suspected CRC that were investigated with colonoscopy [20].The study was designed with patient and public involvement, according to the standards of reporting of diagnostic studies (STARD) guidelines [21] and was registered prospectively The faecal immunochemical test may indicate urgent investigation of younger patients with bowel symptoms; higher faecal haemoglobin concentrations (above 150 µg/g) confer a 12% risk of colorectal cancer but also a 66% risk of serious bowel disease, including colorectal cancer, advanced adenomas and inflammatory bowel disease.
with multiple symptoms, a previously reported hierarchy was used to match each patient to one symptom [20].We defined younger patients as those referred urgently with symptoms under the age of 50 years ('<50'), with a comparator population ('≥50') of patients referred aged 50 years or older in line with published epidemiological data [2,3,5].
The f-Hb results on FIT measured on an HM-Jack analyser (Hitachi Chemical Diagnostics Systems Co.Ltd, Tokyo, Japan, supplied by Alpha Labs Ltd, Eastleigh, Hants, UK) were compared to colonoscopy findings.Data on patient demographics (age, sex, ethnicity, deprivation), serum haemoglobin and ferritin measurements and pathology results including tumour stage were also captured.
AAs were defined in the study protocol as adenomas greater than 10 mm, adenomas of any size with high grade dysplasia, and serrated polyps in the right colon.Patients with multiple findings at colonoscopy were re-categorized with one diagnosis in a hierarchy: CRC ranked highest followed by AAs and then IBD, followed by non-malignant findings.The diagnostic accuracy of FIT was calculated for CRC, AN and SBD.A full description of the methodology for the index test, reference standard, hierarchy of findings at colonoscopy and sample size calculation has been published previously [20].
Categorical data were compared with χ 2 tests; the Kruskal-Wallis test was used for ordinal or quantitative non-normal variables.
Sensitivity, specificity, positive predictive value (PPV) and negative predictive value were reported for each f-Hb cut-off, with 95% confidence intervals.Receiver operating characteristic (ROC) curves were plotted for f-Hb.These were done using an initial threshold of 0.1 to calculate sensitivity and specificity, and then recalculated with increments of 0.1 to plot the ROC curve.In every statistical analysis, P < 0.05 was considered significant.All analyses were performed

RE SULTS
In total, 9822 patients who underwent FIT for gastrointestinal symptoms were included in the final analysis, of whom 1103 were <50 years.The study flow is reported in Figure 1 and has been reported previously [20].

Population description
Patient demographics are reported in Table 1.The median age of patients in the <50 group was 44 years, and the median age in the ≥50 group was 67 years.Women constituted 59.8% of patients <50 and 54.3% of patients ≥50.Patients <50 were more ethnically diverse.Although the median index of deprivation was 6 in both groups, younger patients had a lower mean index of deprivation (5.7 vs. 6.3,P < 0.001) and were significantly more likely to have a lower index of deprivation.
The number needed to scope (NNS) to detect one CRC ranged from 23.8 in patients with detectable f-Hb (above the 2 µg/g cut-off) to one CRC per 8.8 colonoscopies with f-Hb measured above the 150 µg/g cut-off in patients <50.At the same FIT thresholds, the NNS was 10.9 and 2.9 in patients ≥50.
When f-Hb was undetectable (<2 µg/g), CRC prevalence (PPV) was less than 0.4% in all age groups, and the NNS ranged from 250 to 1000.
The NNS to detect one AN ranged from 9.6 in patients with detectable f-Hb (above the 2 µg/g cut-off) to one AN per 4.4 colonoscopies with f-Hb measured above the 150 µg/g cut-off in patients <50.
At the same FIT thresholds, the NNS was 5.9 and 2.1 in patients ≥50.

Diagnostic accuracy of FIT for SBD
The diagnostic accuracy of FIT for SBD is reported in Table 5.The PPV for SBD in patients <50 increased from 31.3% (26.3%-36.5%) to 65.6% (55.2%-75.0%)at cut-offs of 2 and 150 µg/g.The PPV for SBD in patients ≥50 increased from 24.2% (22.7%-25.7%) to 64.3% (60.5%-68.0%)at the same cut-offs.The NNS to detect one case of SBD ranged from 3.2 in patients with detectable f-Hb (above the 2 µg/g cut-off) to 1.5 with f-Hb measured above the 150 µg/g cutoff in patients <50.At the same cut-offs the NNS was 4.1 and 1.6 in patients ≥50.

DISCUSS ION
This is the first study to specifically report on the diagnostic accu- Colonoscopy remains the gold standard investigation of gastrointestinal symptoms and is increasingly employed in younger patients, predominantly to investigate symptoms [27].Through identification and excision of adenomas and other precursors it is also an effective CRC preventative intervention [28].In contrast, FIT is associated with a lower adenoma detection rate than colonoscopy, and thus is better designed to identify CRC rather than AA, as confirmed by our data where the false negative rate for AA was 54.6% at 10 µg/g [29].The sensitivity of FIT for AN, with both CRC and AA, remained low in both younger and older patients in our data as a result of its low sensitivity to AA.As such there is only a moderate estimated benefit of colonoscopy as a  patients with the highest risk symptoms have been captured in this study, and that younger patients at even lower risk may have quite reasonably been managed in primary care and not referred for investigations.The primary end-point of the original NICE FIT study was powered to establish the sensitivity of FIT for CRC in patients referred urgently with suspected CRC symptoms and was not age stratified [20].Consequently, this study is underpowered to detect the diagnostic accuracy of FIT in younger symptomatic patients.The prevalence of CRC may be lower in this group, and it is possible that triage for investigation with FIT may generate more false positive results.However, in the previous analysis of the NICE FIT study, triage of a group with lower risk (DG30) symptoms led to more true negative rather than false positive results [30].
There is no biological reason to suggest that the sensitivity of FIT will diminish in this population and fail to detect cancer.Further adequately powered research in a lower risk population may be required to determine the diagnostic accuracy of FIT for CRC or SBD, and also importantly the specificity and false positive rate, to avoid unnecessary anxiety to patients and unnecessary urgent referrals for investigation.
To establish accurate risk scores for younger symptomatic patients, prospective studies would need to be conducted in primary care to establish an accurate denominator, as only the highest risk patients will be referred on to secondary care.Due to the low prevalence of CRC in an unselected symptomatic younger population in primary care, this would require a very large study size.More objective measures alongside FIT, such as anaemia or thrombocytosis, may help to determine the risk of EOCRC [31,32].Further research into further non-invasive biomarkers for CRC such as breath [33] or urinary volatile organic compounds [34], or faecal/serum DNA samples, may help to identify patients at higher risk of CRC.Adding in faecal calprotectin into pathways for diagnosis of SBD may reduce false negative or false positive tests, although the detection of CRC and IBD was optimized by FIT without faecal calprotectin in a previous prospective study in primary care [35].
CRC diagnosis in younger symptomatic patients is difficult and FIT alone is unlikely to solve this problem.However, FIT could provide a diagnostic adjunct for PCPs to risk stratify referred patients in a similar way to how faecal calprotectin is currently being used.
The higher prevalence of SBD in patients with elevated f-Hb concentrations of 150 µg/g on FIT should indicate referral in younger patients and suggests that an urgent pathway should be set up for these patients, as this may avoid any delay to diagnosis in primary or secondary care [36].

CON CLUS ION
FIT holds promise in the investigation of younger patients with bowel symptoms.Detectable f-Hb should indicate referral for investigation of SBD, which will be detected in one in three patients.Higher f-Hb concentrations above 150 µg/g confer a 12% risk of CRC and a 66% risk of SBD indicating urgent referral for investigation.Further work is required to assess the utility of FIT to risk stratify younger symptomatic patients who are likely to benefit from invasive investigations such as colonoscopy.

(
ISRCTN 49676259) after ethical approval by the National Research Ethics Committee on 23 January 2017 (reference 16/LO/2174).Patients were recruited at 50 National Health Service hospitals across England and took part in the study with informed consent.Symptom criteria for an urgent referral described by the National Institute for Health and Care Excellence (NICE) NG12 and DG30 criteria include change in bowel habit, iron deficiency anaemia, abdominal or rectal mass, and rectal bleeding [22, 23].When patients were referred Conclusion: Detectable f-Hb on FIT in symptomatic younger patients may indicate referral for investigation of colorectal cancer and serious bowel disease.K E Y W O R D S colonoscopy, colorectal neoplasms, diagnostic accuracy, faecal immunochemical test, FIT What does this paper add to the literature?
This study was funded by NHS England awarded to RM Partners, the West London Cancer Alliance hosted by the Royal Marsden NHS Foundation Trust.The study is supported by the National Institute for Health Research Clinical Research Network Portfolio.Croydon University Hospital acted as study sponsor.The study sponsors and funders had no role in study design; in the collection, analysis and interpretation of data; in the writing of the report; and in the decision to submit the article for publication.

F I G U R E 1
Study flow diagram.Ineligible index test: FIT performed after colonoscopy, FIT sample inadequate, FIT sample >14 days old.Ineligible reference standard: incomplete colonoscopy, underwent other investigations, cancelled colonoscopy, withdrew from study or missing data curve analysis of FIT for CRC.Left: Patients aged <50 years.Right: Patients aged ≥50 years of FIT for CRC in younger patients with suspected bowel cancer symptoms.The negative predictive value of FIT for CRC in our data is very high due to the low CRC prevalence in younger patients of 1.5%.The sensitivity of FIT for CRC was moderate at lower cut-offs Diagnostic accuracy of FIT for CRC by age group [25]µg/g.This suggests that the presence of blood in faeces on FIT may indicate urgent investigation of symptomatic younger patients, particularly when f-Hb measures above 150 µg/g.Although our data may not support the use of FIT as a test to rule out CRC in younger patients, it can help clinicians in primary care to identify patients who require urgent referral and investigation for SBD, potentially an important application given the poor predictive value of symptoms alone for the identification of high risk patients.FIT at the time of CRC screening programme which incorporates all younger patients may not be feasible given the rarity of the disease in this population.While the American Cancer Society guidelines have advocated lowering the national screening age for CRC to 45[25], this is based on an assumed increased CRC risk due to birth cohort effects that is not yet seen in real-world data.Furthermore, according to modelling TA B L E 1 Patient characteristics a Other ethnicity: any other ethnic group, not specified.estimates, as the population screening age is lowered the relative risk of complications from colonoscopy increases alongside diminishing efficacy [26].Pragmatically, many countries and populations do not have the endoscopy or financial resources to adopt such a screening strategy.The vast majority of EOCRC may most likely be diagnosed after presentation with symptoms.a Perianal disease: anal fissure, anal fistula, haemorrhoids, solitary rectal ulcer.Miscellaneous: lipoma, melanosis coli, parasites.TA B L E 2 Findings at colonoscopy by age group TA B L E 3 a None is no FIT, i.e., all patients < or ≥ 50 years.b Number of CRC detected at cut-off/number of CRC diagnosed in age group.
Diagnostic accuracy of FIT for AN by age group None is no FIT, i.e., all patients < or ≥ 50 years.Diagnostic accuracy of FIT for SBD by age group Number of SBD detected at cut-off/number of SBD diagnosed in age group.
a a None is no FIT, i.e., all patients < or ≥ 50 years.b