Executive summary of recommendations

Randomised controlled trials

There have been no RCTs comparing time to diagnosis, stage at diagnosis or longer-term CRC outcomes between patients with and patients without FIT as part of their diagnostic pathway, nor any trials comparing FIT based pathways using different FIT thresholds.

Systematic review of cohort studies of primary care patients

A systematic review summarised the diagnostic performance of FITs for CRC across a range of thresholds, including 69 536 symptomatic adults from primary care from 23 cohort studies published between May 2018 and November 2020.26 Using the reported limit of detection (LoD), which ranged from ≥2 µg Hb/g faeces to ≥7 µg Hb/g faeces, meta-analysis of 11 studies (n=41 388 patients) resulted in a pooled sensitivity of 93.4% (95% CI 88.0% to 96.4%) and specificity of 76.9% (95% CI 67.7% to 84.0%). At a threshold of ≥10 µg Hb/g faeces (15 studies; n=48 872), pooled sensitivity was lower at 87.0% (95% CI 81.0% to 91.6%) with higher specificity 84.4% (95% CI 79.4% to 88.3%). Meta-analysis of five studies (n=24 187) reporting at ≥20 µg Hb/g faeces resulted in a reduced sensitivity of 84.1% (95% CI 78.6% to 88.4%) and an increased specificity of 86.6% (95% CI 75.6% to 93.1%). At a threshold of ≥150 µg Hb/g faeces meta-analysis of six studies (n=34 691) resulted in a sensitivity of 64.1% (95% CI 57.8% to 69.9%) and a specificity of 95.0% (95% CI91.2% to 97.2%).

Primary care cohorts with low prevalence of CRC

The underlying prevalence of CRC directly influences a test’s performance at a chosen threshold. Individual prospective and retrospective cohort studies reporting FIT performance in populations of patients tested in primary care report CRC prevalence ranging from 0.8% to 1.8%, highlighting the variation in symptomatic patient groups eligible for FIT across primary care settings.8 11 13 18 20 21 In a subgroup analysis of the review, at a threshold of ≥10 µg Hb/g faeces, sensitivity was lower at 86% (95% CI 78% to 93%) versus 89% (95% CI 82% to 96%) and specificity significantly higher of 87% (95% CI 82% to 92%) versus 81% (95% CI 74% to 88%) when eight studies with a combined prevalence <3% were compared with seven studies with a prevalence of ≥3%.26

Trade-offs between single FIT thresholds in primary care cohorts

Most cohort studies have reported the use of a single FIT threshold with some including statistical modelling to demonstrate the trade-offs at different FIT thresholds in terms of the numbers needed to scope (NNS) to detect one cancer and the number of missed cancers (NMC) per 1000 patients tested.

The review reported that for a prevalence of 1% and 2%, the NNS was 20 and 10 using a threshold of ≥10 µg Hb/g faeces, and the NMC was 1 and 3 per 1000 patients, respectively.26 Increasing the threshold to ≥20 µg Hb/g faeces reduced the NNS to 12 and 6 and the NMC increased to 2 and 4 per 1000 patients tested. At 150 µg Hb/g faeces the NNS is reduced further to 7 and 4, and the NMC increased to 4 and 8 per 1000 patients tested.

Based on a large retrospective cohort of 9896 patients tested in English primary care in the context of the DG30 NICE guidelines (CRC prevalence 1.1%), the authors illustrated the NNS and NMC for the thresholds of ≥7, 10, 20, 50, 100, 120 and 150 µg Hb/g faeces.27 The corresponding proportion of positive tests were 11, 10, 7, 4, 3, 3 and 2%, the proportion of cancers detected 91, 91, 85, 74, 61, 57 and 54%, the NNS to detect one cancer was 11, 10, 8, 6, 5, 5 and 4, and the NMC per 1000 FITs was 1, 1, 2, 3, 4, 5 and 5. Reducing the threshold of ≥10 to ≥2 µg Hb/g faeces resulted in increase in the NNS to 21 and a reduction in the NMC to 4 per 100 000 patients tested.23

At a higher prevalence of 1.6%, a smaller Spanish retrospective cohort of 4543 symptomatic patients reported the NMC per 1000 patients tested to be 3.7 (2.2–6.3) at a threshold of ≥10 µg Hb/g faeces compared with 4.1 (2.5–6.6) using ≥20 µg Hb/g faeces, and the NNS was 13.8 (10.8–17.7) compared with 10.9 (8.5–14.0).27 The authors concluded that the use of ≥20 µg Hb/g in preference to ≥10 µg Hb/g could reduce referrals for colonoscopy without missing more than one CRC per 1000 patients tested.

Increasing the threshold favours specificity reducing the NNS to detect one CRC but increases the NMC per 1000 patients tested. The opposite occurs when the threshold is reduced. The FIT threshold used in clinical practice is likely to be chosen based on a balance of tolerance of missed cancers and the diagnostic resources available to urgently investigate CRC.

Multiple thresholds for low-risk, intermediate-risk and high-risk populations

Multiple FIT thresholds have been introduced in some clinical settings to provide a rule-out threshold, a rule-in threshold and an intermediate range where the investigation of population subgroups and/or active safety netting is advised.

The Nottingham rapid CRC diagnosis (RCCD) service triages adult patients of any age, except those with rectal bleeding and rectal mass, combining low, intermediate and high thresholds (prevalence 1.6% (227/13 361)).10 The RCCD service considers ≥100 µg Hb/g faeces the ‘high risk’ positive contacting these patients directly for rapid investigation. Patients with a FIT result <4 µg Hb/g faeces, and with a FIT result of 4–10 µg Hb/g faeces but normal blood tests are considered ‘negative’. Patients with a FIT of 4–10 µg Hb/g faeces with anaemia, low ferritin or thrombocytosis, or with FIT ≥10 µg Hb/g faeces are considered ‘positive’ and investigated urgently via the 2WW. The cancer detection rate was 0.1% for <4 µg Hb/g faeces, 0.6% between 4 and 9.9 µg Hb/g faeces, 3.3% for 10–99.9 µg Hb/g faeces and 20.7% for ≥100 Hb/g faeces.

Cohort data from Tayside on FIT use in primary care patients with unselected GI symptoms was modelled to show that ‘reassurance thresholds’ of <2, 7, 10 and 20 µg Hb/g faeces were associated with CRC risks of 0.1, 0.3, 0.3 and 0.4%.18 Intermediate-risk populations were created to highlight those with a cancer risk below the 3% risk used by NICE to trigger urgent colorectal investigation. For example, an intermediate population defined by 10–99 µg Hb/g faeces had a risk of 2.7%, leaving a higher risk population ≥100 µg Hb/g faeces with a risk of 14.5%. An intermediate range of 10–149 µg Hb/g faeces resulted in an intermediate population risk of 3.2%, meaning all patients ≥10 µg Hb/g faeces would qualify for urgent investigation. The cancer risk for the intermediate 7–199 µg Hb/g faeces population was 2.8% with a risk of 17.2% in the ≥200 µg Hb/g faeces group. However, there was no intermediate population ≥20 µg Hb/g faeces with a risk below 3%. The patients with FIT ≥20 µg Hb/g faeces comprised 16.8% of the population tested compared with 21.9% for ≥10 µg Hb/g faeces and 25.4% for ≥7 µg Hb/g faeces.

Individualised FIT thresholds

All dichotomous FIT thresholds, from the LoD upwards, identify a population with CRC risk ≥3% as recommended by NICE for urgent referral. For example, the PPV for ≥2 µg Hb/g faeces was 4.7% (4.0% to 5.5%) in the Oxfordshire Primary Care cohort, rising to 8.4% (7.1% to 9.9%) using ≥10 µg Hb/g faeces.23 An analysis from the Southwest of England discussed moving from population risk to individual risk.11 The cancer risk in the ≥10 µg Hb/g faeces population was 7%, in line with larger data sets from low prevalence primary care populations.18 23 26 27 Although cautious about the uncertainty in their estimate, they calculated that the individualised risk was 3% at the threshold ≥37 Hb/g faeces (95% CI 26 to 50) suggesting safety netting may be warranted between 10 and 36 Hb/g faeces.

FIT in primary care: pre-referral

Four large retrospective cohort studies, in which not all individuals are investigated but have been followed-up, have described FIT usage for symptomatic patients in primary care: Northern Spain (n=38 675),21 Oxford (n=16 604),20 Tayside (n=5372)16 and Nottingham (n=24 855)10 where palpable rectal mass and bleeding were excluded.

In these large low prevalence cohorts, the cancer diagnosis rate at follow-up after reassurance without investigation based on very low fHb levels in primary care was 0.3% or less, regardless of variations in the platforms and cut-offs used. A further Danish study of 3462 patients evaluated FIT in patients without ‘alarm symptoms’ similarly found that the risk of CRC was <0.1% in those with fHb <10 µg Hb/g faeces.13

Evidence from these studies also demonstrates that primary care clinicians will still refer patients to secondary care where clinical concern persists via appropriate or alternate pathways. Unpublished data from Nottingham suggests one in three patients are seen in an alternate pathway after ‘negative’ FIT. One in seven FIT below the threshold patients were investigated in Denmark. A study from Southwest England,11 reported that GPs made a referral within 3 months for 1 in 10 negative FITs detecting more than half of the FIT below the threshold CRCs (5 of 8).

Some, not all, of these populations were included in a pooled analysis of 15 studies including 48 872 patients,26 yielding a sensitivity for CRC of 87.2% (95% CI 81.0% to 91.6%) when using a threshold of ≥10 µg Hb/g faeces. A threshold of ≥20 µg Hb/g faeces missed less than one additional CRC per 1000 patients (from a population of five studies; n=24 187, with CRC prevalence 2%).

FIT in secondary care: post-referral

Large UK cohort studies of patients preselected for referral by GPs describe the performance of FIT in populations with the majority fulfilling ‘high-risk’ NG12 criteria all receiving colonic investigation.

The NICE FIT study,28 a multicentre double-blinded study of 9822 patients undergoing colonoscopy, demonstrated that the risk of bowel cancer was around 0.2% in those with undetectable levels of fHb and 0.4% in those with fHb <10 µg Hb/g faeces. The Fast Track FIT study29 evaluated 5040 patients undergoing colonoscopy, CTC or colorectal telephone assessment pathway showed the risk of bowel cancer was 0.4% in undetectable fHb and 0.5% in those with fHb <10 µg Hb/g faeces. The quantitative FIT (qFIT) study30 reported on 3596 patients who underwent colonoscopy or CTC and the risk of CRC was 0.4% in those with undetectable fHb and 0.5% in those with fHb <10 µg Hb/g faeces. FIT kits were provided to patients in both primary care and hospital settings. A Scottish study of 4841 referred patients reported the risk of CRC when fHb <10 µg Hb/g faeces was 0.6%.31

Overall, this evidence suggests that FIT has an acceptable miss rate whether used in primary care prior to referral or in secondary care following referral.

FIT in primary care vs no FIT

A service evaluation from Nottingham including 1668 patients provides a small real-life uncontrolled comparison of an NHS Trust that adopted FIT for symptoms (excluding rectal bleeding and palpable rectal mass) covering half of the regions urgent 2WW CRC referrals and a private provider for the remainder of the referred population.9 FIT rollout increased 2WW referrals and the proportion of new CRC diagnoses made on 2WW pathways. Emerging differences were noted in the cost of investigations required to detect each CRC and the time to diagnosis favouring the pathway using FIT.32 Other regions of the East Midlands have restricted FIT usage to those over 60 years and demonstrated demand reduction14 similar to the experience in Scotland where the introduction of FIT achieved a 15% reduction in urgent referrals from primary care.16

In Northern Spain33 the clinical outcomes of 279 patients with symptomatic CRC diagnosed after a ‘positive’ FIT in primary care were compared with 1210 patients with symptomatic CRC without a primary care FIT. A higher proportion of Stage I and II cancers (51.3% vs 45.5%) and improved 3-year survival were found in the FIT group. The Nottingham group9 reported a pre-pandemic shift towards diagnosis at earlier stage. Juul et al 13 report 66.7% of cancers diagnosed at Stages I and II when evaluating FIT in symptomatic patients without ‘alarm symptoms’. Bailey et al 10 found 33% were detected at early stage when using FIT in those that satisfied DG30 criteria specifically. Turvill et al 29 described a higher proportion of Stage I and II CRC (52.7%) in their referred Fast Track study population with low fHb <18 µg Hb/g faeces although numbers are very small. Other studies have shown obstructing tumours and higher stage CRC are also common in FIT ‘negative’ CRC.

It is not possible to conclude that introducing FIT in primary care improves longer-term outcomes, but evidence is emerging that FIT testing in primary care could have this impact.

FIT only in secondary care

During the pandemic FIT was adopted widely across the UK given fears that endoscopy and CTC were aerosol generating procedures that increased risk of viral transmission.18 High fHb thresholds (100 µg Hb/g faeces in England and Wales, 400 µg Hb/g faeces in Scotland) were recommended with a pragmatic acceptance that some diagnoses would be missed, and a number of reports have described this.

High fHb could be used to identify referred patients for urgent/2WW/direct to colonoscopy pathways, with lower fHb directed to routine pathways. FIT could be used to ‘upgrade’ patients referred routinely and fHb might be used to determine the choice between colonoscopy and CTC, or colonoscopy and flexible sigmoidoscopy. An fHb could also become a component of all informed consent conversations for invasive colonic investigation. The current statute law underlying the 2WW system and the timed nature of pathways inhibits this individualised approach. ‘Local agreement’ is key to best practice in implementation.

It has been suggested that access to FIT should be restricted for use in secondary care but studies describing a favourable stage shift associated with FIT use suggest that a secondary care only approach may miss the opportunity to increase the proportion of cancers detected by identifying higher risk patients before referral.

Importance of FIT testing

Numerous studies and reviews have found strong associations between a FIT positive test and risk of CRC. Studies that compared the strength of this association with other risk factors have shown that a positive FIT test is usually more predictive of risk than demographic, clinical and laboratory criteria for referral to investigate possible CRC.28 34–36

Improving return of tests

A small number of articles were identified that provide insights into strategies for maximising return of FIT kits in patients who have been asked to have the test.

Coronado et al undertook surveys on worded versus non-worded instructions for performing FIT testing in an area with a population having high rates of non-English speakers. They reported preferences for non-worded instructions from both patients and professionals.37

A US study of FIT used for screening found that patients asked to complete a two-sample test were less likely to return than those completing a one-sample test; this difference was statistically significant though not numerically large (39.6% vs 43.3%).38

Haghighat et al described an initiative to encourage patients to submit their FIT kit as part of colorectal screening as soon as possible after this was offered, ideally before leaving the clinic. They reported significant improvements over the 6-month period of study (27.6% v. 20.6%, p<0.001).39

A survey of patients in the NICE FIT study examined 1151 questionnaires representing 30.6% of those mailed out (1151/3760), with lower percentage returns from London than outside London patients (17% and 43%, respectively). Most patients found FIT collection straightforward (90.2%), not unhygienic (76.3%) and preferable to colonoscopy (78.1%). People aged 40–64 years were less likely to prefer FIT to colonoscopy than older age groups. Patients from ethnic minority backgrounds were less likely to have found the test hygienic and to return the kit in a future test.40

Safety netting

Safety netting has come to be regarded as ‘best practice’ in relation to cancer diagnosis, especially in non-specialist settings.41 42 Its aim is to ensure patients do not drop through the healthcare net but are monitored until symptoms are explained, defined as a consultation technique to communicate uncertainty, provide patient information on red-flag symptoms and plan for future appointments to ensure timely re-assessment of a patient’s condition.43 NICE refer to safety netting as ‘the provision of support for patients in whom the clinician has some uncertainty as to whether the patient has a self-limiting illness and is concerned that their condition may deteriorate’.6 However, safety netting may also comprise administrative activities such as test result reconciliation and the follow-up of referrals.44 45 A key role for safety netting in FIT based pathways is the monitoring of FIT negative patients to ensure timely referral or investigation of those referred.

Absolute risk of missed cancers

In a recent meta-analysis pooling data from 35 925 patients from nine UK studies from primary and secondary care including a range of FIT thresholds from ≥2 to ≥19 µg Hb/g faeces and within the NICE NG12 context, the pooled percentage of missed CRCs due to a FIT below the selected threshold was 8.7% (95% CI 5.1% to 12.2%) equating to an NMC of 2.1 per 1000 patients tested.46 Pin-Vieito et al reported the NMC as 1 per 1000 patients tested at a CRC prevalence of 1% and 3 in 1000 at a prevalence of 2%, using a threshold of ≥10 µg Hb/g faeces, noting that the prevalence of CRC in primary care based studies ranged from 0.8% to 1.8%.26 Risks of 1 and 3 per 1000 patients equate to absolute cancer risks of 0.1% and 0.3%, respectively, both significantly below the current NICE threshold of ≥3% cancer risk used to warrant urgent cancer investigation.47 An effective safety netting strategy could identify FIT negative patients with an increased risk of cancer who may warrant further investigation.

Meta-analyses and randomised controlled trials

We found no meta-analyses or RCT of safety netting strategies to ensure CRCs are diagnosed in patients with a negative FIT, in primary or secondary care settings. A step-wedged cluster RCT is currently underway to test the effectiveness of an electronic safety netting toolkit embedded into major primary care clinical systems to facilitate patient follow-up in terms of the time and route to diagnosis.48

Observational studies

There were no observational studies evaluating safety netting strategies to promote re-consultation and onward referral among people with ongoing symptoms despite a negative FIT result.

Recent research has emphasised the importance of clinician ‘gut feeling’ in the diagnosis of cancer, conceptualised as the rapid summing up of multiple verbal and non-verbal patient cues.49 FIT pathways should allow clinicians to refer or investigate FIT negative patients if there are ongoing clinical concerns. In a cohort study from the Southwest of England, GPs still requested urgent investigation for five of the eight FIT negative cancers ‘probably because continuing symptoms allowed the GP to ‘overrule’ the negative test’.11

A common suggestion in the literature and guidelines was re-consultation within 4–6 weeks for patients with ongoing symptoms and a FIT below threshold. G27 guidance (2005), later replaced by NG12, recommended urgent referral for patients with symptoms persistent for 6 weeks. Underpinning evidence to inform the timing of any safety netting action is lacking including: what might be considered the ‘normal’ duration of a benign symptom, the time taken for progression of high-risk adenomas to cancer, or the interval of stage progression.

Many cohort studies have documented the clinical presentation of patients later diagnosed with FIT negative CRC, suggesting that these characteristics could be prioritised for referral or included in the communication of safety netting advice to patients.16 27 50 51 However, there is marked variation in the characteristics of FIT negative cancers between these studies and so relying on these characteristics to inform a safety netting strategy could be falsely reassuring.

Modelling studies

Modelling studies to date have not demonstrated the benefit of combining FIT with other clinical features and blood test results to enhance sensitivity by reducing false negative FITs. A comparison of FIT at ≥10 µg Hb/g faeces alone, with the FAST score (combining FIT age and sex), and ColonFlag (a machine learning algorithm using age, sex and FBC indices to derive a risk score), showed that FIT and ColonFlag missed a different 18% of CRCs, respectively, and FAST score missed 27.3%.52 Combining simple blood tests with FIT at best matches the sensitivity of FIT alone in patients tested in primary care, whether as pairs of results or within multivariable model.20

Introduction

Considerable emerging evidence has recently been published on the diagnostic accuracy of FIT in symptomatic patients using a range of thresholds.26 63–66 The the most common reported fHb triage threshold is 10 ug/g. Many of these reports assessed all patients presenting to their primary care practitioners with any bowel symptoms suspicious of CRC, but others categorised symptoms into low and high risk symptoms as defined by NICE.67 Only a few studies examined the diagnostic accuracy of FIT for individual symptoms.

In this section we a focus on those studies where all symptomatic patients that received FIT were linked with a reference standard investigation with full evaluation of the colon and rectum to exclude CRC with either colonoscopy or CTC. Where these evaluations were not performed, studies with other reference standards such as flexible sigmoidoscopy or CT scan and those with clinical follow-up/record linkage of a minimum of 3 months were examined. Where this was not available then studies with lesser follow-up and other linked investigations such flexible sigmoidoscopy or CT scan were included. In assessing the diagnostic accuracy of FIT we considered its accuracy in symptomatic patients in general, and also in ‘high-risk’ and ‘low-risk’ symptoms.

Early studies (table 2)

One of the first studies to examine the role of FIT in symptomatic patients was published in 2011 by a Dutch group who conducted a clinical study involving five centres testing FIT in a mixed cohort of symptomatic and asymptomatic patients who were scheduled for colonoscopy. A total of 2145 patients undifferentiated by colonoscopy indication were included which were then divided irrespective of their symptoms or lack of, into high and low risk groups for CRC. The overall sensitivity and specificity of FIT for CRC was 92.4% (95% CI 84.2% to 97.2%) and 86.4% (95% CI 84.8% to 87.9%).68 In 2013, the Tayside group in 2013 examined 280 participants who had both FIT and colonoscopy and found that patients with cancers had a median fHb of >1000 ng Hb/mL buffer (equivalent to >200 ug/g). Using a cut-off fHb concentration of 50 ng Hb/mL buffer (equivalent to 10 ug/g), the NPV for CRC was 100%. A year later the same group reported similar results but on a larger number of patients (n=569) and concluded that using FIT in primary care may help target colonoscopy more appropriately when patients present with colorectal symptoms.69 The same year, a group from Spain compared FIT with NICE 2005 and the Scottish Intercollegiate Guidelines Network (SIGN) referral criteria in 787 symptomatic patients FIT at >100 ng/mL (equivalent to 20 ug/g) had a higher sensitivity for CRC detection (87.6%) than NICE criteria (61.9%; p<0.001) and SIGN criteria (82.5%; p=0.4).24 The specificity of FIT was also higher than NICE and SIGN criteria (77.4%, 65.2%, 42.7%; p<0.001). In 2016 the Tayside group reported on 755 patients with FIT and endoscopy results. The sensitivity and specificity at a cut-off of 10 ug/g was 89.3% (95% CI 71.8% to 97.7%) and 79.1% (95% CI 75.9% to 82.0%), respectively.64 The same year another Scottish group reported on 484 patients who had FIT and colonoscopy and reported that all 11 cancers were detected at a cut-off of 10 ug/g (100% sensitivity).70

Later studies (table 3)

In 2017, NICE produced its DG30 guidelines which recommended the use of FIT in low-risk symptoms defined as ‘patients without rectal bleeding who have unexplained symptoms but do not meet the criteria for a suspected cancer referral’.5 The latter refers to high risk symptoms as defined in the updated NICE NG12 guidance in 201767 where use of FIT was not recommended. The DG30 guidance was based on a health technology assessment, commissioned by The National Institute for Health Research to produce a diagnostic accuracy report on FIT to triage symptomatic patients at low risk of CRC presenting in primary care.63 The report looked at 10 studies but summarised evidence from 5 studies only that reported on FIT as a rule-out test for CRC with a cut-off of 10 µg/g. Data were taken from one study (507 patients) for the HM-JACKarc analytical system,70 and four studies (4091 patients) for the OC-Sensor analyser.64 The summary estimate of sensitivity for the HM-JACKarc was 100% (95% CI 71.5% to 100%) and for the OC-Sensor was 92.1% (95% CI 86.9% to 95.3%).63 The corresponding specificity was 76.6% and 85.8%.

Two years later, a meta-analysis by the Warwick group included 17 studies of which 9 were on symptomatic cohorts (6755 patients).66 Five studies used the OC-sensor (4883 patients), three used HM-JACKarc (1499 patients) and one used the Actim Faecal Blood system. Five Studies (4603 patients; four OC-Sensor and one HM-JACKarc) examined FIT at a cut-off of 10 ug/g and the others looked at a range of cut-offs ranging from 7 to 50 ug/g. The overall pooled sensitivity and specificity for CRC were 0.90 (95% CI 0.87 to 0.92) and 0.87 (95% CI 0.83 to 0.90), respectively.

A subanalysis for studies that used OC sensor was performed. These studies examined multiple cut-off concentration values ranging from 10 to 40 µg/g. Analysis of the pooled sensitivity and specificity for an fHb cut-off range of 10–15 µg/g (4096 patients) showed sensitivity of 0.93 (95% CI 0.88 to 0.96) and specificity of 0.87 (95% CI 0.82 to 0.90). For the range between 20 and 40 µg/g, the pooled sensitivity and specificity of 0.87 (95% CI 0.84 to 0.90) and 0.89 (95% CI 0.84 to 0.92), respectively.

In the same year, a meta-analysis from Spain by Pin Vieito included 14 studies of which 7 were on symptomatic cohorts.65 Four studies (4035 patients) reported on FIT with a cut-off of 10 µg/g using the OC-Sensor; one of these studies was conducted in Scotland, and the others were conducted in Spain, or pooled data from studies from both countries. The pooled sensitivity for CRC was 94.1% (95% CI 90.0% to 96.6%) and specificity 66% (95% CI 47.1% to 80.1%).

A meta-analysis by the same group in 2021 included 23 studies (69 536 patients).65 The meta-analysis examined diagnostic accuracy of FIT at different thresholds. Fifteen studies (n=48 872) reported on a cut-off of 10 ug/g. The pooled sensitivity and specificity for CRC was 87.2% (95% CI 81.0% to 91.6%) and 84.4% (95% CI 79.4% to 88.3%), respectively. Five studies (n=24 187) reported on a cut-off of 20 ug/g and the pooled sensitivity and specificity were 84.1% (95% CI 78.6% to 88.4%) and 86.6% (95% CI 75.6% to 93.1%). Six studies (n=34 691) assessed FIT as rule in test cut-off of >150 ug/g showing a sensitivity of 64.1% (95% CI 57.8% to 69.9%) and a specificity of 95.0% (95% CI 91.2% to 97.2%). The group concluded that FIT is the test of choice to evaluate patients with new-onset lower GI symptoms in primary healthcare.

The most recent meta-analysis published in December 2021 from New Zealand included 15 studies with a cohort of 28 832 patients, all of whom were prospectively recruited.71 Thirteen studies (six HM-JACKarc, four OC-Sensor, one FOB Gold and two QuickRead Go analysers; n=25 500) reported on FIT at a cut-off 10 ug/g. The summary sensitivity and specificity were 88.7% (95% CI 85.2% to 91.4%) and 80.5% (95% CI 75.3% to 84.8%), respectively. At the lower cut-off of the LoD (three studies; 15 160 patients), the summary sensitivity increased to 96.8% (95% CI 91.0% to 98.9%) but specificity reduced to 65.6% (95% CI 59.0% to 71.6%).

A meta-analysis performed to inform these guidelines included 31 studies up to March 2022 with a cohort of 79 566 patients. For ‘all symptoms’, ‘all analyser’ analysis and a reference standard of >90% receiving either colonoscopy or CTC (16 studies, n=35 945), the summary sensitivity and specificity were 91.0% (95% CI 88.9% to 92.7%) and 75.2% (95% CI 69.6% to 80.1%), respectively, at a cut-off 10 µg/g.61

In the 2 years between the 2019 and 2021 meta-analyses, there has been an explosion in the number of studies reporting on FIT, with between 25 000 to 40 000 patients added to the latest meta-analyses.61 The Pin Vieito and Saw meta-analyses included two of the three largest diagnostic accuracy, multicentre studies that were conducted in England.28 62 The qFIT study (UCLH Cancer Collaborative) included 3596 patients with high-risk symptoms and reported a sensitivity of FIT for CRC at 83.3% (95% CI 75.6% to 91.0%) at cut-off of 10 µg/g using the OC-Sensor.62 The NICE FIT study included 9822 patients with high and low risk symptoms and at the same cut-off, reported a sensitivity for CRC of 90.9% (95% CI 87.2% to 93.8%).) using the HM-JACKarc analyser.28 The third largest research study published in 2021 by the York group included 5040 patients with high-risk symptoms and was included in a meta-analysis to inform these guidelines.29 61 The sensitivity and specificity of FIT for CRC at 10 ug/g using the HM-JACKarc analyser was 87.4% (95% CI 81.0% to 92.3%) and 80.9% (95% CI 79.7% to 81.9%), respectively. The group considered an optimal threshold between sensitivity and specificity and calculated this at 19 ug/g with a sensitivity of 85.4% (95% CI 78.8% to 90.6%) and specificity of 85.2% (95% CI 84.1% to 86.2%).

High and low risk symptoms

Specific symptoms

Although NICE has recommended in DG30 guidance use of FIT to triage low risk symptoms in primary care, it excluded patients with rectal bleeding from FIT testing.5 Moreover, early diagnostic accuracy and service development studies have recommended exclusion of certain symptoms from FIT testing including patients with rectal bleeding on the assumption of a high false positive rate. Patients with iron deficiency anaemia (IDA) or palpable abdominal mass were excluded after smaller studies reported false negative rates (undetected cancers) in patients meeting these symptoms. However, recent studies have confirmed that FIT can be used in specific high-risk symptoms, including IDA, rectal bleeding, and abdominal masses.

Rectal bleeding

There are several studies that reported on use of FIT in symptomatic patients with higher risk symptoms, including patient cohorts with rectal bleeding.16 27 31 78 However, three studies have focused solely on use of FIT in patients with rectal bleeding.17 77 79

The use of FIT in rectal bleeding was investigated recently in 462 patients in NHS Tayside, Scotland.17 The positivity rate was 63.3% at a cut-off of 10 µg/g. The prevalence of cancer was 8.5% (25/293) with an fHb >10 ug/g compared with 0.6% (1/168) when fHb <10 ug/g. The sensitivity and specificity for CRC is calculated as 96.2% (95% CI 80.4% to 99.9%) and 38.3% (95% CI 33.7% to 43.0%), respectively. The one CRC in the cohort with an fHb <10 ug/g was in the descending colon and would have been detected by flexible sigmoidoscopy. Indeed, flexible sigmoidoscopy detected the majority of serious bowel disease (SBD) pathology (CRC, inflammatory bowel disease and advanced adenomas) except for four advanced adenomas (10 out of 14 SBD out of 168 patients). The authors concluded that patients with an fHb <10 ug/g and persistent rectal bleeding, can be safely investigated with flexible sigmoidoscopy.

Högberg et al in 2020 reported on qualitative (using three FIT samples) rather than quantitative FIT in 606 patients with rectal bleeding.79 The positivity rate was 42% with a sensitivity of 96.2% and NPV of 99.7% for CRC.

The NICE FIT study investigated the diagnostic accuracy of FIT in 3143 patients with rectal bleeding either alone or in combination with other symptoms compared with 6679 patients with non-rectal bleeding symptoms.77 The positivity rate of 26.9% at 10 µg/g was lower than the other two studies above but higher than the non-rectal bleeding group at 15.2%. The sensitivity and specificity of FIT in the rectal bleeding group for CRC was 96.6% (95% CI 92.2% to 98.9) and 76.6% (95% CI 75.0% to 78.1%). The PPV was 16.8% (95% CI 15.9% to 17.9%) and NPV 99.8% (95% CI 99.5% to 99.9%). In the bleeding cohort, there were five CRCs with an fHb <10 ug/g, four of which would have been detected with flexible sigmoidoscopy. The group concluded that the use of flexible sigmoidoscopy in patients with rectal bleeding and an fHb <10 ug/g, would reduce the risk of CRC to 0.03%.

The Booth et al meta-analysis61 informing the guidelines identified three studies (n=3665) that reported specifically on rectal bleeding and used colonoscopy or CTC as a reference standard.17 77 80 The summary sensitivity at a cut-off of 10 µg/g was 96.6% (95% CI 92.8% to 98.8%) and specificity 71.7% (95% CI 70.2% to 73.2).

Iron deficiency anaemia

The Nottingham study, which used a postal FIT in both high and low risk groups described above as per NICE NG12 and DG30 (but excluding rectal bleeding), reported higher fHb levels in those with IDA at 4.8 (0.8–34.1) ug/g compared with 1.2 (0–6.4) ug/g in those without IDA. In this study 40 patients with CRC were identified and using a cut-off of 10 ug/g of fHb, CRC detection rate was 7.2 in those with IDA. In fact, the authors concluded that CRC detection was higher in those with IDA.74

Cunin et al reported in their cohort of patients where FIT was used in primary care applying the NICE NG12 criteria that 7/48 patients (14.6%) had CRC below the cut-off of 10 ug/g, that is, fHb below threshold cancers.81 Of the seven fHb below threshold cancers, five had anaemia as well as change in bowel habits and of these, four had true IDA. It was also observed that these six CRCs were right-sided (caecal). The FIT sensitivity for CRC was 80.0% (95% CI 55.7% to 93.3%) in patients with IDA compared with 89.0% (95% CI 70.0% to 97.1%) in those with a combination of other symptoms.

Earlier studies did not report influence of IDA on diagnostic accuracy of FIT.82 83 More recent data in abstract form applying a definition of IDA of ferritin under 15, reported sensitivity of 92.0% (95% CI 84.4% to 95.9%) and specificity of 63.2% (95% CI 59.1% to 67.4%)84 The prevalence of CRC in this cohort with IDA, as expected increased with bands of fHb levels; 1.2% in those with fHb under 9 ug/g, 13.5% in the band 10–200 ug/g and 38.9% in those with cut-offs >200 ug/g. Another study in abstract form, suggested that fHb offered similar discriminatory values to symptoms and even younger patients.85

Previous BSG guidelines do not recommend use of FIT in those with IDA.86 Reasons for their recommendation (evidence strength low; statement strength weak) was related to publication bias in that iron deficiency anaemia may be over-represented in those with fHb below a threshold of 10 ug/g CRCs. However, the larger and subsequently published NICE FIT and qFIT studies more recently have provided further consistent data which supports the use of FIT testing in people with IDA.28 62

The meta-analysis informing the guidelines identified two studies (n=724) that reported specifically on IDA and used colonoscopy or CTC as a reference standard.72 76 The summary sensitivity at a 10 µg/g cut-off was 96.7% (95% CI 88.7% to 99.6%) and specificity 73.6% (95% CI 70.1% to 76.9%).61

Change in bowel habit

Few studies have specifically reported on accuracy of FIT for CRC in patients with change in bowel habit (CIBH). Further, the definitions used make it more difficult to collate these together to form a unified consensus. In the Nottingham primary care study CIBH had lower CRC detection rates 4.5% versus 7.4% compared with those with IDA at a cut-off of 10 ug/g.74 The NICE FIT study suggested that the sensitivity of FIT for CRC at a cut-off of 10 ug/g in patients with CIBH, was higher in the older population (over 60 years of age) compared with those under 60 years of age (85.9% vs 60%), respectively.72

The large 38 765 participants in the Spanish primary care study,21 reported broadly similar findings even when applying cut-offs of 10 or 20 ug/g/faeces.78 In fact, the number needed to scope for those presenting with diarrhoea was 15 and constipation 16.2 at 10 ug/g compared with 12.8 with diarrhoea and 13 with constipation at 20 ug/g.

The meta-analysis informing the guidelines identified two studies (n=10 067) reporting specifically on CIBH symptom and used colonoscopy or CTC as a reference standard. The summary sensitivity and specificity at 10 µg/g were 85.6% (95% CI 79.0% to 90.8%) and 83.6% (95% CI 82.9% to 84.3%), respectively.61

Evidence summary

1. FIT is highly sensitive for CRC in symptomatic patients with most large studies reporting a sensitivity of >87% at the commonly used cut-off of 10 µg/g.

2. The diagnostic accuracy of FIT is similar in both high and low risk symptomatic patients, irrespective of the cut-off used.

3. FIT is not always detectable in patients with rectal bleeding and is a useful evaluation tool when CRC is suspected. FIT is highly sensitive for CRC in patients with rectal bleeding with a sensitivity of >90% at a cut-off of 10 ug/g. Patients with negative FIT and persistent rectal bleeding can be safely investigated with flexible sigmoidoscopy and appropriate safety measures in place.

4. The evidence for use of FIT for the detection of CRC in IDA supports its use at a cut-off of 10 µg/g. If a lower cut-off is applied, then the accuracy further improves.

5. The evidence for use of FIT for the detection of CRC in those with isolated change in bowel habit is less clear although UK data suggests greater benefit in those over 60 years (while the Spanish data are consistent with other symptoms).

FIT, demographics and colorectal cancer

Demographic variations in CRC incidence are well recognised87 as are variations in fHb from screening studies.88 89 CRC diagnoses rise with increasing age and it is known that fHb also rises with age, even in the absence of notable pathology. CRC incidence is higher in men overall but also in the male population referred by GPs for further investigation. Deprivation has been noted to be a risk factor for men rather than women.88 A number of studies on FIT in symptomatic patients report higher fHb in men than in women, although Bailey et al’s study11 in a restricted DG30 population notes high fHb levels in women aged 30–40 years specifically. Men with CRC show a predilection to the rectum and one group has suggested false negatives in palpable rectal mass where a bleeding tumour would present overtly; CRCs in women are associated with the right colon and most studies report lower fHb in this group of cancers, most likely due to distribution of blood throughout a formed stool reducing concentration and increasing risk of sampling error. Although many studies show some differences in fHb by age and gender, findings are inconsistent, and this is perhaps unsurprising given the complexity of the potential interactions described and the size of cohort needed to address all of these. Furthermore, FIT is a test for occult blood in stool which may be related to many other pathologies with different interactions with demographics, or no identifiable pathology at all, adding further complexity to this challenge.

Age

A number of papers have reported data relevant to diagnostic performance of FIT by age. Interpretation is complicated by the different age categorisations, FIT thresholds and outcomes considered, as well as variations in the study populations and endpoints. Moreover, not all of the studies reported on diagnostic performance in terms of sensitivity, specificity and/or area under the curve (AUC).

A subanalysis of the NICE FIT study28 assessed diagnostic accuracy in 1103 patients under 50 years of age preselected by GPs for urgent referral to secondary care.90 All patients were well enough to undergo colonoscopy, thus potentially excluding some older and frailer patients. At all FIT thresholds, sensitivity for the older group exceeded that for the younger age group. At a threshold of 10 µg Hb/g faeces sensitivity was 87.5% for those <50 years of age versus 97.4% in those ≥50 years of age. At the 2 µg Hb/g faeces threshold, specificity was higher for the younger group (70.4% vs 64.1%); at the 10 µg Hb/g faeces threshold, specificity was almost identical in the two groups (83.6% vs 83.5%); and at the 150 µg Hb/g faeces threshold, it was slightly higher in the older group (92.2% vs 94.9%). At all thresholds, the PPV for the older groups was more than double that for the younger group (PPV was 6.8% vs 17.1% in those over 50y at 10 µg Hb/g faeces). Despite these differences, further analysis of the 329 CRCs which were detected in this selected population found there was no association between age and FIT status (negative/positive) at a threshold of 2 or 10 µg Hb/g faeces. The CRC prevalence in the younger age group was 1.5% and the study was not powered to assess younger patients specifically. Furthermore, the authors point out the value of detecting inflammatory bowel disease (IBD) and other pathology in younger patients.

The Fast Track FIT study29 of 5040 patients preselected by GPs for urgent referral, all completing either colonoscopy, CT colonography or CT abdomen and pelvis, compared groups above and below 60 years of age. AUC was slightly lower for those aged 60+ years of age (0.88, 0.85–0.92) compared with 1217 younger patients (0.92, 0.88–0.96), but the 95% CIs for these estimates overlapped. In this cohort, sensitivity and specificity were slightly higher in the younger group (sensitivity: 90.0 vs 83.5; specificity: 87.4 vs 85.4); PPV and NPV did not differ. The authors describe an optimal cut-off threshold of 37 µg Hb/g faeces for those under 60 years of age, compared with a baseline of 20 µg Hb/g faeces for those over 60 years of age. In an earlier study53 this group also reported on 515 individuals similarly selected for investigation, in whom the FIT threshold to achieve optimal AUC for CRC detection again needed to be higher for younger patients (<65 years of age: FIT threshold ≥46 µg Hb/g faeces for AUC of 0.89 (0.722–1.000); 65+ years of age threshold ≥12 µg Hb/g faeces for AUC of 0.91 (0.842–0.981). However, these findings were based on only 7 cancers in the younger age group and 19 in the older group.

A small study of 404 patients,91 referred to secondary care for colonoscopy, in whom overall sensitivity and specificity of FIT at a threshold of 20 µg Hb/g faeces for CRC were 87.5% and 83.7%, respectively. FIT performance for relevant colonic pathology showed sensitivity of 50.6% and specificity of 69.6%; sensitivity was slightly lower and specificity slightly higher in those aged under 50, compared with those aged 50 and older. This study used a different FIT platform to most published studies (FOB Gold).

The largest study commenting on age reported on FIT in primary care use in Northern Spain and included 38 675 patients (CRC prevalence 1.7%) with a variety of symptoms.21 This study used 2 years registry-based follow-up as an endpoint and therefore the majority are not investigated in secondary care. However, the group reported diagnostic accuracy data based on follow-up and found no significant differences in sensitivity across three age strata:<50 years of age (93.1% at 10 µg Hb/g faeces; 91.8% at 20 µg Hb/g faeces), 50–69 years of age (91.5% at 10 µg Hb/g faeces; 88.7% at 20 µg Hb/g faeces) and >69 years of age (89.8% at 10 µg Hb/g faeces; 87.2% at 20 µg Hb/g faeces) although the values declined with increasing age at both cut-offs. They reported differences in specificity which fell significantly with rising age at a threshold of 10 µg Hb/g faeces (88.5% <50 years of age, 83.6% 50–69 years of age and 75% >69 years of age). In this study, the prevalence in the population under 50 years of age was only 0.3% and raising the threshold to 20 µg Hb/g faeces had the least detrimental effect on missed cancers in this age group; although the overall increase in missed CRC across the whole population at 20 µg Hb/g faeces was <1 in 1000.

Gender

Two large studies of FIT in high-risk cohorts selected for and completing secondary care investigation have reported diagnostic accuracy by gender. The NICE FIT study28 found no association between sex and FIT status using a cut-off of either ≥2 or ≥10 µg Hb/g faeces to define positivity. By contrast, the Fast Track FIT study29 found that, to achieve the optimum AUC for CRC, different FIT cut-offs would be needed for men and women; 21 µg Hb/g faeces for men and 16 µg Hb/g faeces for women gave AUC of 0.89 and 0.88, respectively. In their earlier, smaller, study of 515 patients, Turvill et al 53 reported that the optimal FIT cut-off for detecting CRC was ≥22 µg Hb/g faeces for men (AUC=0.909, 0.835–0.983) and ≥12 µg Hb/g faeces for women (AUC 0.891, 0.744–1.000); but these results were based on 18 cases of CRC in men and 8 cases in women. In a study of 928 patients78 (41% men), FIT at a threshold of 10 µg Hb/g faeces had lower sensitivity for the detection of bowel disease (CRC, high-risk polyps or colitis) in women than in men. Sensitivity was 95.4% (95% CI 75.1% to 99.7%) for men compared with 76.1% (95% CI 54.5% to 89.9%) for women. Specificity was slightly higher in women (men: 80.5% (95% CI 75.9% to 84.4%); women: 85.5% (95% CI 82.2% to 88.4%).

In their large primary care study (n=38 675) with follow-up, rather than full investigation, Pin Vieito et al 21 also reported higher sensitivity (91.6% vs 88.4%) but significantly lower specificity (79.9% vs 82.6%) in men at 10 µg Hb/g faeces. In a similar UK study of 9896 patients (41% men), where not all patients were investigated, Nicholson et al 27 reported that the AUC for both CRC was almost identical in men (0.933) and women (0.948).

Age and gender

Nicholson et al 27 also reported the AUC for FIT at a threshold of ≥10 µg Hb/g faeces for the detection of CRC if testing was restricted to different age groups. Restricting testing to people aged 80+ was the only instance where the AUC dropped below 0.90. If FIT testing, was to be restricted to those aged 40+, the AUC would be 0.944 (95% CI 0.899 to 0.988) for women and 0.934 (95% CI 0.897 to 0.972) for men; if limited to the 60+ age group, the AUC would be 0.919 (95% CI 0.856 to 0.982) for women and 0.921 (95% CI 0.870 to 0.971) for men; and if limited to those aged 70 and older, it would be 0.934 (95% CI 0.867 to 1.000) for women and 0.936 (95% CI 0.895 to 0.978) for men. In women, the AUC would be highest (0.708) when testing was to be restricted to those aged 50 and older; in men it would be highest if testing was restricted to those aged 70+.

Multivariate analyses including age and gender

A number of groups have reported multivariate analyses including demographics (with symptoms, blood results and other factors) as covariates and fHb measured by FIT is consistently the most predictive factor by some margin.14 25 92–95 No other factor reaches significance consistently, although increasing age and male gender appear most frequently.

The FAST score, combining age and gender with FIT, was developed in Northern Spain in patients undergoing colonoscopy,92 but has failed to show improved clinical effectiveness compared with FIT alone when trialled in a broader population in Tayside96 and when applied to the NICE FIT data set.28 It is not clear whether the use of different platforms affected these findings, as well as other intrinsic differences in the cohorts evaluated. The ColonFlag score52 takes this approach a step further, combining FBC results with fHb, age and sex, and showed improved specificity compared with FIT—taking an AND/OR approach sensitivity improved to 100%. However, this study is relatively small and needs further validation. A similar sized study (n=408) by Digby et al 34 found no value in other factors, including demographics, other than family history.

A much larger study by Withrow et al 93 reported on 16 604 patients tested in primary care (and includes the cohort from Nicholson et al)27 specifically focussing on the value of combining FIT with blood tests. However, the majority of patients were not investigated. In a variety of models, they found no value in including age. They also reject the value of models including gender, although they note that the threshold to reach a 3% PPV is slightly higher in men (25 µg Hb/g faeces) than in women (17 µg Hb/g faeces). Consistent with this, Rodriguez-Alonso et al 94 found that, after accounting for FIT, sex was statistically significantly associated with both CRC and advanced neoplasia and men had more than twofold increased risk of disease. They found age did not need to be included in a multivariable model for CRC, after FIT (and gender and IDA) had been included. However, when an outcome of advanced neoplasia was considered in a model including FIT result and gender, risk increased with increasing age.

There is insufficient evidence that diagnostic performance of FIT for the detection of colorectal neoplasia varies by ethnicity, deprivation and other factors.

Ethnicity and deprivation

Although some studies provide breakdown of populations by ethnicity and deprivation, none have looked at this as a primary outcome and there is no clear data on variations in diagnostic accuracy. Even in larger data sets the low event rate in non-white categories or specific deprivation groups appears too small and there are no pooled analyses that address this. The NICE FIT study28 compared the characteristics of 329 patients detected with CRC, all of who had undergone a FIT, were compared according to notional FIT status. At a positivity threshold of either ≥2 ug or ≥10 Hb/g of faeces, there was no statistically significant difference in the ethnic distribution of those who would have been classified FIT positive or FIT negative. The same observation was made for deprivation category of area of residence. A questionnaire based evaluation of the NICE FIT cohort suggests lower acceptability in non-white ethnic groups.40 The group in Nottingham have also reported some evidence that the non-return of FIT kits is associated with younger age, male sex, non-white ethnicity and higher deprivation populations (Abstract added to library). These findings mirror studies in screening and have implications for education, implementation and safety netting, rather than diagnostic accuracy per se.

Other factors

As noted earlier, Digby et al 34 reported on family history and after adjusting for FIT result (and rectal bleeding and folate level) a family history of polyps was associated with a more than eightfold increased risk of any significant bowel disease (CRC, advanced adenoma (AA) or IBD) (OR=8.21, 95% CI 1.74 to 38.78). Only nine people with a family history of polyps had significant bowel disease.

A single study60 has examined associations between smoking and body mass index and risk of advanced colorectal neoplasia, once FIT had been taken into account. There was a significant association between being a current or ex-smoker and risk of advanced colorectal neoplasia (ACN) (multivariate OR=1.51, 95% CI 1.02 to 2.29). Body mass index greater than 25 kg/m² was not significantly associated with ACN.

Evidence is too limited to conclude whether the diagnostic performance of FIT varies in people using specific medications.

Medications

In total five publications have reported on diagnostic accuracy of FIT in patients using particular medications. Two studies examined proton pump inhibitor (PPI) use. Rodriguez-Alonso et al 97 included 1002 patients referred for colonoscopy; 40% were PPI users. In total 133 patients had advanced neoplasia (AN) and 30 patients had CRC. There were no differences in sensitivity or specificity of FIT with a positivity threshold of ≥20 ug Hb/g of faeces for the detection of CRC among those who used PPIs and those who did not. When AN was considered, both sensitivity and specificity were significantly lower for PPI users (sensitivity: 43.0%; specificity: 86.9%) than non-users (sensitivity: 65.6%, p=0.009; specificity: 92.3%, p=0.010). The second, smaller, study of 612 patients98 published only as an abstract, reported that sensitivity of FIT at a threshold of 10 ug/g for advanced neoplasia (n=55) was lower in PPI users than non-users (54% vs 81%, p=0.05), while specificity did not differ. The area under the receiver operating characteristic curve was 74% (95% CI 0.58 to 0.91) for PPI users compared with 0.92 (95% CI 0.89 to 0.95) for non-users.

Three publications considered the possible influence of use of antiplatelet or anticoagulant medication on FIT diagnostic accuracy. Two reported on data from the COLONPREDICT study; the first, an abstract from 2014, included 1567 patients;92 the second, a full paper published in 2018, included 3052 patients.99 The smaller study reported that the diagnostic accuracy of FIT for CRC was significantly lower among those taking antiplatelet and/or anticoagulant medication (AUC: users 0.81; non-users, 0.88; p=0.04). The larger study focused on aspirin specifically in a study in which 16% of patients used aspirin. Continuous treatment with aspirin did not influence sensitivity, specificity of the AUC of FIT for either CRC detection or AN detection at a threshold for positivity of ≥20 ug of Hb/g of faeces. In a subgroup analysis of patients using ≥300 mg/day aspirin, sensitivity, specificity and AUC were lower, as was polyp prevalence, than among aspirin non-users, but this group included only 58 people and the differences were not statistically significant. The final study reported only multivariable modelling results99 and use of anti-coagulants, anti-platelets or non-steroidal anti-inflammatory drugs (NSAIDs) was not significantly associated with advanced colorectal neoplasia after adjusting for FIT results.

There is currently insufficient evidence to confirm whether diagnostic accuracy is impacted by the type of FIT analyser used.

  GRADE of evidence: low; Strength of recommendation: weak

There is no international standardisation of FIT methods meaning that different results could be obtained on different manufacturer systems.100 Despite this, in symptomatic testing, single thresholds for referral have been recommended.5

To date there are only two peer-reviewed publications that have directly compared results obtained on two different FIT analytical systems when patients have taken samples from the same bowel motion in a symptomatic pathway.75 101

In the first study75 732 patients returned both an OC-Sensor and an HM-JACKarc collection device. They had been instructed to collect samples into each device from the same bowel motion. Correlation of results was carried out and agreement at cut-offs of 4, 10 and 150 µg Hb/g faeces (µg/g) were assessed. To act as a control 114 patients collected two samples using two OC-Sensor devices. At thresholds of 4, 10 and 150 µg/g the Cohen’s kappa have concluded that there is not enough data to comment on the comparative performance is 0.74, 0.79 and 0.76, respectively, which is interpreted as substantial agreement. When two OC-Sensor devices are compared the Cohen’s kappa are 0.80 (substantial agreement), 0.91 (almost perfect agreement) and 1.00 (almost perfect agreement) respectively. This suggests that the referral rate will vary dependent on which of the two methods is used.

In terms of diagnostic accuracy, at thresholds of 4, 10 and 150 µg/g, OC-sensor had a higher sensitivity and lower specificity than HM-JACKarc for CRC. Thus, based on this study, employing the same thresholds for the two methods OC-Sensor will generate more referrals however it will also detect more CRCs than HM-JACKarc at the same thresholds.

In the second study101 the QuikRead Go (QRG), a quantitative point of care FIT test which had previously undergone independent analytical evaluation102 was compared with the FOB Gold wide method on the SENTiFIT laboratory analyser. Five hundred and fifty-three patients provided paired samples for both methods and underwent colonic investigations that were suitable to give definitive diagnostic outcomes. Fourteen patients were diagnosed with CRC. QRG reported one false negative. FOB Gold reported no false negatives. Thirty per cent of QRG results were >10 ug/g would have resulted in referral compared with 16.9% for FOB Gold wide.

In an unpublished study,103 233 patient returned FIT devices from four different FIT systems collected from a single bowel motion. To act as a control a further 189 patients returned two FIT devices from the same FIT system. Differences were observed in the referral rates for different methods and the categorisation according to Cohen’s kappa, specifically for one method more than the other three. There were only seven CRCs detected in the four FIT group so inadequate data to comment conclusively on the comparative diagnostic accuracy of the different FIT tests.

In addition to the three studies above, four systematic reviews have commented on the different FIT assays available.7 26 65 66 The conclusions were that there are a lack of studies directly comparing the performance of different FIT assays and that there are currently no data on the comparative performance of different FIT assays. In addition, it was reported that the limited number of studies, the majority of which were using OC-Sensor, along with high study heterogeneity, did not enable conclusions to be drawn from combining data from different studies.