Article Text

Download PDFPDF

Original article
Experience of adopting faecal immunochemical testing to meet the NICE colorectal cancer referral criteria for low-risk symptomatic primary care patients in Oxfordshire, UK
  1. Brian D Nicholson1,
  2. Tim James2,
  3. James E East3,
  4. David Grimshaw4,
  5. Maria Paddon2,
  6. Steve Justice2,
  7. Jason L Oke1,
  8. Brian Shine2
  1. 1 Nuffield Department of Primary Care Health Sciences, University of Oxford, Oxford, UK
  2. 2 Department of Clinical Biochemistry, John Radcliffe Hospital, Oxford University Hospitals Trust, Oxford, UK
  3. 3 Translational Gastroenterology Unit, Oxford NIHR Biomedical Research Centre, John Radcliffe Hospital, University of Oxford, Oxford, UK
  4. 4 Planned Care, Oxfordshire Clinical Commissioning Group, Oxford, UK
  1. Correspondence to Dr Brian D Nicholson, Nuffield Department of Primary Care Health Sciences, University of Oxford, Oxford OX26GG, UK; brian.nicholson{at}phc.ox.ac.uk

Abstract

Objective To compare the diagnostic performance of guaiac faecal occult blood (gFOB) testing with faecal immunochemical test (FIT) in a low-risk symptomatic primary care population to provide objective data on which to base local referral guidelines.

Design Stool samples from routine primary care practice sent for faecal occult blood testing were analysed by a standard gFOB method and the HM-JACKarc FIT between January and March 2016. Symptoms described on the test request were recorded. Patients were followed up over 21 months for evidence of serious gastrointestinal pathology including colorectal adenocarcinoma.

Results In 238 patients, the sensitivity and specificity for colorectal adenocarcinoma detection using gFOB were 85.7% and 65.8%, respectively, compared with 85.7% and 89.2% for FIT. The positive predictive value (PPV) for gFOB was 7.1% and negative predictive value (NPV) was 99.3%. Comparatively, the PPV for FIT was 19.4% and NPV 99.5%. The improved performance of FIT over gFOB was due to a lower false positive rate (10.8 vs 34.2, p≤0.01) with no increase in the false negatives rate. For any significant colorectal disease, the PPV for FIT increased to 35.5% with a reduction in NPV to 95.7%.

Conclusion In this low-risk symptomatic patient group, the proportion of samples considered positive by FIT was considerably lower than gFOB with the same rate of colorectal adenocarcinoma detection. One in three of those with positive FIT had a significant colorectal disease. This supports National Institute of Health and Care Excellence recommendation that FIT can be reliably used as a triage test in primary care without overburdening endoscopy resources.

  • colorectal cancer
  • primary care
  • stool markers
  • guaiac test

Statistics from Altmetric.com

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Introduction

The 2015 National Institute of Health and Care Excellence (NICE) recommended that UK general practitioners (GPs) use faecal occult blood (FOB) testing to investigate low-risk symptomatic patients in primary care with positive results triggering an urgent referral for investigation of colorectal cancer.1 At the time of the NICE guidance, many UK laboratories had ceased offering FOB to primary care services: only 54% of GPs had direct access to any FOB test.2 Test demand had steadily reduced prior to this point reflecting changes in practice that arose following the availability of the National Bowel Cancer Screening Programme, concerns over the diagnostic accuracy and well-documented analytical performance issues with the standard FOB method, which detects haemoglobin through its ability to oxidise the dye guaiac faecal occult blood (gFOB).3

The 2015 NICE NG12 guidance recommended GPs used FOB to decide whether the following patients should be referred for urgent investigation for colorectal cancer: aged ≥50 years with abdominal pain or weight loss, <60 years with a change in bowel habit or iron deficiency anaemia or aged ≥60 years with any type of anaemia. The guidance generated significant debate,4 5 which included concerns around FOB delaying cancer diagnosis due to false negatives and its potential to increase demand on already stretched endoscopy service due to false positives, a common problem associated with the gFOB method.3 Although the 2015 guidance recommended FOB, the method of FOB detection had not been defined. This was subsequently clarified in July 2017 NICE DG30 and the more specific immunoassay-based method for human haemoglobin, the faecal immunochemical test (FIT), is now recommended over gFOB ‘to guide referral for suspected colorectal cancer in people without rectal bleeding who have unexplained symptoms but do not meet the criteria for a suspected cancer pathway’.6

Increasing evidence demonstrates improved performance of FIT compared with gFOB in colorectal cancer screening programmes,7–9 but there is limited evidence to demonstrate the benefits of FIT in symptomatic primary care patients. NICE expressed concerns about the applicability of all 10 of the studies used to underpin this updated recommendation: none reported data on people with low risk symptoms of colorectal cancer6 and only one study was conducted in primary care.10 A threshold of 10 µg/g faeces was recommended to trigger referral but the literature reported a range of thresholds. A recent report combining three studies of FIT diagnostic accuracy supported its use in primary care but included patient populations referred for colonoscopy to assess referral guidance and not patients with low risk symptoms tested in the primary care setting in line with the NICE guidance.11

In early 2016, when considering how to react to the initial NICE guidance and to inform the redesign of the diagnostic pathways for colorectal cancer investigation in Oxfordshire, we analysed samples referred for FOB testing by both a conventional gFOB and by one of the new automated FIT methods and compared results obtained with subsequent diagnostic outcomes. We present data from this assessment together with the impact of changing FOB technique on demand test positive rates and outcomes.

Methods

The Oxford University Hospitals NHS Foundation Trust (OUHT) Clinical Biochemistry laboratory serves all primary care clinicians in the county of Oxfordshire with a population of approximately 660 000. The department is one of the largest in the UK, performing over 8 million tests a year, and has an ‘M’-based laboratory information management system (LIMS). All FOB testing is undertaken in a single laboratory at the John Radcliffe Hospital.

The patients included in the evaluation were consecutive samples sent to the laboratory from primary care in the period January to March 2016 for investigation of FOB. Leading up to this period, the change in NICE guidance and the indications for FOB testing were communicated to GPs in Oxfordshire by email and newsletter from the Oxfordshire Clinical Commissioning Group. To accurately reflect the clinical requesting patterns and standard laboratory practice, no samples were excluded. The assessment was registered as a service evaluation on our hospitals Datix register (CSS-BIO-3 4730).

Samples were collected into standard stool pots by patients in primary care and referred in for a standard FOB testing using a conventional in house method, the gFOB, using hydrogen peroxide and guaiac solution.12 Results were reported as either positive or negative dependent on blue/green colour development detected visually with the method quality monitored through participation in an external quality assessment programme. Only the gFOB result was reported to clinicians, consistent with historical practice, and this was use to guide subsequent patient management.

For the evaluation period, samples were additionally analysed for FIT using the HM-JACKarc analyser (Kyowa Medex, Tokyo, Japan), a method that has been independently evaluated with respect to analytical performance13 and is now recommended in the context of use for samples from primary care.6 The method had a calibration range of 7 to 450 µg Hb/g faeces and reproducibility across this range was determined to be between 2.2% and 7.3% when expressed as a percentage coefficient of variation. Sample preparation prior to analysis on the FIT instrument used the Extel Hemo-Auto MC device, a process that introduced additional analytical variation. Assessment of total variability, including the sampling stage, was determined to be 14.9% at 16 µg/g and 30.7% at 2 µg/g. External quality assurance samples were analysed to confirm test performance. The selection of the Hb concentration considered positive was made before the NICE recommendation to use 10 µg/g and was based on the methods lowest calibrant value of 7 µg/g.

In selecting the approach to faecal sample handling, we balanced the requirements to minimise sampling errors that may give rise to false positives if undertaken by patients with stability concerns,14 and we have highlighted this issue and balancing these risks in our contribution to the NICE FIT adoption resource.6 Where more than one sample result was available for any individual patient, any positive result within those samples tested was considered a positive outcome on the basis that a single positive would trigger referral. Where multiple samples on a single patient were collected, these were on sequential days, which precluded assessment of changes in FOB test results with disease progression.

To confirm the presence or absence of disease, OUHT clinical and diagnostic databases were searched for evidence of pathology for between 21 and 23 months following the FOB testing for all patients. Histology, endoscopy and CT colonoscopy reports were retrieved by searching by both hospital and NHS number. Patients were classified individually then by discussion between members of the research team (BS, BDN, TJ and JEE) as having adenocarcinoma; having significant colorectal disease (adenocarcinoma, high risk adenoma polyps larger than 10 mm and inflammatory bowel disease and one patient with neuroendocrine tumour); upper GI pathology; no significant pathology identified on endoscopy or CT colonoscopy; or no further follow-up investigation for 21–23 months. Patients who had no further investigation were categorised as negative for serious pathology as any serious pathology would be expected to have presented to secondary care within this time period.

Diagnostic accuracy of gFOB and FIT was summarised using sensitivity, specificity, negative and positive predictive values (PPVs) and exact 95% CI. Total test demand and numbers of positive results for FOB testing were assessed through reported results extracted from the LIMS results reporting system. Assessment of the change of positive rates was done using an interrupted time series analysis. We used a Poisson regression, corrected for overdispersion to model rate of positive tests in both the FOB era (1 January 2016 to 31 December 2016 and the FIT era (1 February 2017 to 31 October 2017). The model was specified so that we could test a step change in the number of positives as well as a longer term change in trend. A sensitivity analysis was preformed to remove patients with rectal bleeding as these patients have a potentially higher risk of colorectal adenocarcinoma and are excluded from the NICE guideline criteria.

Results

FOB testing by both FIT and gFOB was undertaken on 332 samples from 238 patients, median age 58 years, range 19–93 years, of whom 103 (43%) were male and 135 (57%) were female.

Clinical details

The majority of FOB requests had one or more clinical details (table 1) confirming lower abdominal symptoms (1,2): change in bowel habit (n=59), anaemia (62), abdominal pain/discomfort (45), blood in stools (23), rectal bleeding (9) and weight loss (4) symptoms. It was not possible to categorise the remaining requests (n=46) due to absent or uninterpretable clinical information. Significant colorectal disease was detected in 20 patients, 7 of which had adenocarcinoma (full details: online supplementary appendix 1). Subsequent investigations including gastroscopy only (n=17), colonoscopy only (30), gastroscopy and colonoscopy (13) and CT colonography (32).

Supplementary file 1

Table 1

Clinical details of patients tested

gFOB/FIT concordance

Figure 1 and online supplementary appendix 2 present results by diagnostic group by FIT concentration and by FOB positivity. The analytical agreement between gFOB and FIT was 77%, 202 tests being positive and 36 negative by both methods. In 65 samples, the gFOB was positive and the FIT was negative compared with 36 samples that were gFOB negative and FIT positive. The overall positive rate in this direct comparison was 30.4% (n=101) for gFOB compared with 14.2% for FIT (n=47). Seventy-four patients had more than one sample collected (from two up to five samples). For FOB, there was complete concordance in all samples in 73% of patients compared with 85% for FIT.

Figure 1

FIT and gFOB results of patients by outcome category. FIT, faecal immunochemical test; gFOB, guaiac faecal occult blood; GI, gastrointestinal.

Diagnostic accuracy

Table 2 presents the diagnostic accuracy of FIT and gFOB for adenocarcinoma and significant colorectal disease. Figure 2 presents receiver operator characteristics of the relationship between FIT threshold and the impact on diagnostic accuracy for both adenocarcinoma and significant colorectal disease.

Figure 2

Receiver operating characteristic (ROC) curves of FIT for the detection of adenocarcinoma (solid line) and significant colorectal disease (dashed line). Black dots represent a single estimated true positive/false positive rate of gFOB for adenocarcinoma (ACa) and significant colorectal disease (SCD). Cross-hatches correspond to the true positive/false positive rate estimates for thresholds of (7 µg/g, 10 µg/g, 20 µg/g and 50 µg/g). AUC, area under the curve; FIT, faecal immunochemical test; gFOB, guaiac faecal occult blood.

Table 2

Comparison of diagnostic accuracy of gFOB and FIT at different concentration thresholds for detection of adenocarcinoma and significant colorectal disease (adenocarcinoma, high-risk adenoma and inflammatory bowel disease)

At the haemoglobin concentration of 7 µg/g used for defining a FIT positive result, a sensitivity and specificity for detection of colorectal adenocarcinoma was 85.7% and 89.2% compared with 85.7% and 65.8% for gFOB. The PPV for FIT was 19.4% and negative predictive value (NPV) 99.5%. Comparatively, the PPV for gFOB was 7.1% and NPV was 99.3%. None of the patients with upper GI pathology (including one cancer) had a FIT above our threshold of 7 µg/g.

Changing the threshold for FIT to 10 µg/g had little effect on the diagnostic characteristics of FIT: sensitivity 85.7%; specificity 90.5%; PPV of 21.4% and NPV of 99.5%. Increasing the threshold further to a commonly used screening threshold (50 µg/g) increased specificity and PPV at the expense of sensitivity but with minimal impact on the NPV.

The diagnostic performance to detect any significant colorectal disease for gFOB were 65.0% sensitivity and 67.0% specificity compared with 55.0% and 90.8% for FIT. The PPV for gFOB was 15.3% and NPV 95.4% compared with a PPV of 35.5% and NPV of 95.7% for FIT. The area under the curve for FIT reduced from 89.5 (76.0–100.0) for colorectal adenocarcinoma to 79.8 (68.2–91.4) for any significant colorectal disease.

A sensitivity analysis removing the nine patients that had been referred with rectal bleeding resulted in no change in the sensitivity or specificity of gFOB or FIT for colorectal adenocarcinoma and a marginal decrease in the sensitivity and specificity of gFOB and FIT for the detection of significant colorectal disease (online supplementary appendix 3).

Changes in workload

Workload increased following the 2015 NICE guidance and the rate of positive results fell both in a slope and a stepwise manner when the FIT replaced gFOB (figure 3). Poisson regression modelling allowing for overdispersion indicated a significant step change in the number of positive tests after the introduction of FIT (rate ratio (RR)=0.53, p=0.004) and a borderline significant change in the long-term trend (RR=0.95, p=0.04992). Furthermore, the false positive rate of gFOB (34.2) was three times that of FIT (10.8, p≤0.01) with no increase in the false negative rate.

Figure 3

Changes in workload and positivity rates before and after the transition to FIT testing. The total number of tests (top panel) and the number of tests that were positive (bottom panel) for the period in which gFOB was used and the period in which FIT replaced gFOB. Lines in bottom panel are fitted trend lines estimated by Poisson regression. Shaded area represents the month in which the change from gFOB to FIT occurred. FIT, faecal immunochemical test; gFOB, guaiac faecal occult blood.

Discussion

The NICE guidance DG306 highlighted the need for further research into the use of FIT including auditing the use of the FIT in primary care. It has also been noted that the current evidence base for use of diagnostic tests in primary care for diagnosis of colorectal cancer is lacking.15 Our early experience with adopting FIT shows equivalent detection rates and high NPV for colorectal adenocarcinoma for gFOB and FIT in a population of low-risk symptomatic primary care patients. However, gFOB had many more false positive results: 1 in 5 patients with a positive FIT had cancer detected compared with 1 in 14 positive gFOBs, and 1 in 3 patients with positive FIT had serious colorectal disease detected compared with 1 in 6 positive gFOBs.

At the end of our initial evaluation, we selected a cut-off for 7 µg/g as this related to the lowest calibrant value for the FIT method used. Subsequently NICE has recommended a value of 10 µg/g.6 Our analysis suggests the difference in threshold had little impact on overall test performance, but this should be explored in the future as larger data sets become available as more laboratories start using this methodology.

The patient cohort included several patients with an upper GI pathology including one patient with cancer and none of these patients had FIT above our cut-off. This is consistent with a recent analysis from large Dutch FIT cohort that supports not performing upper GI investigations in patients with a positive FIT test16 as blood from an upper GI lesion does not remain intact through the GI tract and loses FIT assay immunoreactivity.

One patient in our group had a neuroendocrine tumour (NET), and this did not produce a positive FIT result. There is limited literature on tests of occult blood in the context of NET, but a recent report has suggested that their detection is possible through organised colorectal cancer screening17; however, we have limited evidence to confirm the characteristics of bleeding in these tumours.

Limitations

There are several potential limitations with the data presented. First, we have assumed that the samples referred from primary care accurately reflect the low-risk symptomatic population. The clinical details that GPs entered onto the FOB request suggest this is the case for most patients tested. As single symptoms, only rectal bleeding and anaemia would qualify for urgent investigation for colorectal cancer, but it can be assumed that GPs assessed the cases tested with FOB to be lower risk or they did not qualify for fast-track referral due to age. It is also likely that this is the population that the NICE guidance is aimed at as it accurately reflects the FOB requests received from routine primary care practice.

Second, the majority of patients did not have a gold standard investigation to rule in or rule out serious colorectal disease as this was not a controlled study but a retrospective assessment of diagnostic performance using routinely collected clinical data. However, all patients have been followed up for evidence of subsequent pathology in hospital clinical, laboratory, radiology, endoscopy and pathology databases for between 21 and 23 months after initial testing—a period that would reasonably allow significant pathology to be identified. The clearly defined geographical catchment area and the single central laboratory used in this study minimises the likelihood that serious disease was diagnosed during the study period and not captured.

Finally, there was a delay between faecal specimen collection and sampling into the devices provided by the supplier to preserve haemoglobin immunoreactivity. One report using exogenous haemoglobin added to faecal material showed that significant sample degradation can occur on storage18; however, endogenous haemoglobin is more heterogenous in this respect and most samples retain measurable quantities,14 above the 10 µg/g threshold, even 7 days after collection compared with using the suppliers sampling device. It has also been noted14 that when patients use the sampling device, it can lead to false positive results, presumably due to poor sampling technique. This is consistent with our observation that even when undertaken by experienced laboratory staff, use of the sampling device is the major factor contributing to preanalytical variability. We feel our current approach of advising laboratory delivery on the day of specimen collection mitigates degradation risk and should avoid the potential of false positives associated with patient collection. The current preliminary evaluation aimed to take a pragmatic approach in this respect but further work on sampling technique and identification of the optimum procedure using the collection device should be undertaken in future larger studies.

Comparison with existing literature

Several studies have reported the diagnostic accuracy of FIT in symptomatic primary care patients. Three studies have been reported by the same group from Dundee, Scotland, who combined these studies data resulting in a sensitivity for colorectal cancer at 93.3% and an NPV of 99.7%.10 A significant difference in our population was that it had not already been referred to secondary care for investigation, and all patients were tested using a single FIT technique. As we have observed similar specificity, 90.5%, and NPV, 99.5%, it would endorse the use of FIT in this context, although gFOB performed similarly. However, FIT clearly outperforms gFOB in sensitivity and PPV due to a reduced number of false positive results. For the detection of any significant colorectal disease, we observed very similar NPV at 95.7% compared with the Dundee group with an overall NPV of 96.0%. The sensitivity in our study was lower at 55.0% compared with 63.2%, possibly because the patients included in our study of routine practice may be a lower risk symptomatic group.

In a Dutch cross-sectional diagnostic accuracy study,19 810 patients referred for colonoscopy for suspicion of significant colorectal disease were investigated with point-of-care FIT using a haemoglobin threshold of >6 µg/g. The sensitivity for significant colorectal disease (colorectal cancer, inflammatory bowel disease, diverticulitis or advanced adenoma >1 cm) was 67%, specificity was 84%, PPV was 47% and NPV was 92%. In a further prospective English study of 430 non-consecutive patients20 referred for urgent lower gastrointestinal investigation, the sensitivity for colorectal cancer was 84% and specificity was 93%, similar to our data at 85.7% and 89.2%. Our data would suggest gFOB to be inferior in most measures of diagnostic performance to FIT. The difference in performance of FIT found in our study is most likely to be due to our study including a lower risk spectrum of unselected symptomatic primary care patients—the population that NICE intended GPs to use FOB testing.

Implications

Adoption of the NICE guidance NG121 for colorectal with respect to use of FOB in the patient pathway for colorectal cancer has been slow, probably due to the uncertainty of test methodology. The July 20176 guidance supporting the use of FIT methodologies to detect blood clarified this uncertainty, but there is limited data to show what impact the switch will have in a routine clinical setting.

It is possible that the initial reaction to NG124 5 may have influenced clinician behaviour. However, following the guidance, we have seen a rise in laboratory demand for FOB testing and the switch to FIT, with its lower rate of positive result, may attenuate the impact this workload growth may have on referral numbers for endoscopy. Within our service, the current laboratory cost of the FIT method is around £11 per sample, which is more expensive than gFOB (costing in the region of £7). This is low compared with the alternative options including colonoscopy, which costs £458 at 2017 NHS national tariffs, and enables fast tracking of those at greater risk of colorectal cancer. At a stable rate of 300 FOB samples per month, and assuming all positive FOB results progress to colonoscopy, compared with gFOB, a FIT strategy would save approximately £13 380 per month (£160 560 per year). A formal economic evaluation is underway at our institution.

Historically, more than one sample was recommended for FOB testing3 to overcome false positives associated with ingestion of a range of peroxidase containing dietary material. For FIT, this consideration is not relevant since it is a more specific analytical test based on immunoassay for human haemoglobin. Our observations that when more than one sample is taken the degree of concordance between samples is higher (73% for FOB and 85% for FIT) suggests greater test consistency, but clearly there is still some variability in the presence of blood in stools and the optimum number of samples for colorectal cancer detection also needs further consideration.

Areas of uncertainty where future research is warranted include: whether the utility of FIT can be improved in combination with symptoms associated with colorectal cancer21; how FIT can be used to influence the progression on to imaging techniques22; and how GP initiated FIT testing may complement bowel cancer screening programmes as there is evidence that longer intervals between symptomatic presentation to primary care testing can lead to more advanced stage cancer.23

Conclusion

To date, published studies have focused on those patients already referred for follow-up of lower gastrointestinal symptoms. Our data, although a small comparative study, provides confidence that the diagnostic accuracy of FIT in routine primary care practice translates successfully in the context of the lower risk population considered in NG12. We conclude that FIT offers a significant and safe methodological transition with improved diagnostic ability to reduce pressure on urgent referral pathways by identifying patients who do not require further investigation for colorectal adenocarcinoma, thereby controlling colonoscopy demand and reducing costs.

Significant of this study

What is already known on this topic

  • Faecal occult blood (FOB) testing is recommended by National Institute of Health and Care Excellence (NICE) as a triage test before further colorectal cancer investigation in patients with low-risk abdominal symptoms presenting to primary care. No study has assessed the comparative diagnostic accuracy of guaiac FOB (gFOB) and faecal immunochemical testing (FIT) in this setting.

What this study adds

  • In a consecutive sample of FOBs submitted by general practitioners to a large UK laboratory following local implementation of the NICE faecal occult blood testing (FOBT) criteria, gFOB and FIT had equivalent sensitivities and high negative predictive values for colorectal adenocarcinoma. However, the performance of FIT was far superior due to a significantly reduced false positive rate, resulting in higher positive predictive values and specificity.

How might it impact on clinical practice in the foreseeable future

  • Our results show that FIT can be reliably used as a triage test without overburdening endoscopy resources, supporting widespread implementation of the NICE recommendations for its use in low-risk patients in primary care.

Acknowledgments

We acknowledge Oxfordshire Clinical Commisioning Group (CCG).

References

Footnotes

  • Contributors BDN, TJ, DG, JEE and BS conceived the study. TJ, BS, MP and SJ collected the data. BDN, JLO, BS and TJ analysed the data. TJ and BDN prepared the first draft of the manuscript. All authors provided critical revisions to the manuscript.

  • Funding BDN receives funding from Macmillan Cancer Support and the NIHR. JEE was supported by the National Institute for Health Research (NIHR) Oxford Biomedical Research Centre (BRC).

  • Disclaimer The views expressed are those of the author(s) and not necessarily those of the NHS, the NIHR or the Department of Health.

  • Competing interests None declared.

  • Patient consent Not required.

  • Provenance and peer review Not commissioned; externally peer reviewed.