Throughout the developed world, population screening for colorectal cancer (CRC) is increasing at regional and national levels, with guaiac faecal occult blood tests (gFOBTs) being the most frequently used screening modality.1 Much of the evidence for the effectiveness of CRC screening involves the use of gFOBTs,2 but these have many disadvantages including poor sensitivity and specificity, false-positive and false-negative results from dietary constituents, and difficulty in reading the developed colour.3 For these reasons, there is considerable current interest in faecal immunochemical tests (FITs),4 and there is now much evidence to suggest that FITs should be used in population screening programmes as the first-line test.5 These tests are specific for human haemoglobin and so do not require any drug or dietary restriction, and they are easy to use: they are also more specific for lower gastrointestinal tract bleeding.3 Most FITs are qualitative, and the analytical detection limit, and thereby the clinical performance characteristics, is set by the manufacturers. However, a number of FITs are now available that permit quantitation of haemoglobin in faeces. This gives flexibility in allowing the haemoglobin concentration selected for use as the cut-off for classification as positive or negative to be set locally to give a positivity rate that is manageable in terms of the colonoscopy resource available.6

Most available qualitative and quantitative FITs available at present require the collection of faeces into tubes containing buffer solution; such tubes would probably be difficult and expensive to post to and from participants in a screening programme. However, a number of card collection systems are becoming available, and we have previously evaluated the characteristics of one of these in a reflex two-tier gFOBT/FIT approach to CRC screening.7 An automated FIT analytical system (analyser plus reagents, calibrators and controls) has recently become available and, since it does not require the use of specially designed tubes to present the samples for analysis, can be described as an “open” system. The three aims of the present study were to investigate the quantitation of haemoglobin in samples collected on cards with this new analytical system, the relationship between haemoglobin and pathology, and the optimum cut-off for this combination of collection device and analytical system.

MATERIALS AND METHODS

Population

The study group consisted of participants in the second 12 months of Phase 1 of the Scottish Bowel Screening Programme (SBoSP): the group here was not the same as in our previous studies.7 8 As in the Scottish arm of the UK pilot studies,9 the initial population comprised individuals aged 50–69 years living in NHS Grampian, NHS Tayside and NHS Fife invited to participate in the SBoSP. A conventional gFOBT kit (hema-screen, Immunostics, Ocean, New Jersey, USA, supplied by Alpha Laboratories, Eastleigh, Hampshire, UK) was used. Participants collected two samples of faeces without dietary restriction from each of three consecutive bowel movements directly on to filter paper containing guaiac gum through oval shapes in the gFOBT kit card. Participants who had five or six positive ovals on the gFOBT were classed as strong positive, and were offered colonoscopy without further testing. Participants with 1–4 positive ovals were classed as weak positive and asked to do a second gFOBT (again without dietary restriction) and, if any oval was positive, colonoscopy was offered.

Study design

The participants with a positive gFOBT were asked to participate in the study either at a face to face interview with a colorectal specialist nurse or by postal invitation, and were asked to provide two samples of faeces on a single card collection device (hema-screen DEVEL-A-TAB, Immunostics) and return this to the Scottish Bowel Screening Centre. The device was used to collect samples of faeces by stabbing an applicator at least four times into a bowel motion and completely filling one of the two rectangular windows on the card, and then repeating this process with the next bowel motion: drying of faeces on the card ensures stability for up to 30 days. Information on the purpose of the study was also provided. Invitees were told that the results of the FIT would not be used to influence the decision to proceed with colonoscopy and that the result would be unavailable to them so that their decision to proceed with further investigation would not be influenced. Our previous studies7 8 have shown that this approach has no sampling bias according to age, gender or degree of gFOBT positivity. The study was approved by the NHS Tayside Medical Ethics Committee and all participants gave written informed consent.

gFOBT and FIT analyses

gFOBTs were analysed in the Centre Laboratory which is accredited by Clinical Pathology Accreditation (UK) (CPA) to ISO 15189-based standards and only performs analyses of faecal samples. In general, four bowel screener staff performed the analyses, and they had all received detailed training as required by CPA10 and by the Bowel Screening Programme standards of NHS Quality Improvement Scotland.11 Qualitative FIT analyses were also performed in the Centre Laboratory exactly as described previously.7 The samples of faeces in buffer used for the qualitative FIT were then frozen. Prior to quantitative FIT analyses, the samples were allowed to thaw in a refrigerator, mixed and centrifuged to enhance optical clarity. The supernatants were then assayed, by immunoturbidimetry, on a SENTiFOB analyser using FOB Gold reagents, calibrators and controls (Sentinel Diagnostics, Milan, Italy). The analyser is a fully automated photometric analyser for haemoglobin measurement, designed to be used with FOB Gold reagents at around 100 analyses per hour, and is supplied with a computer with dedicated user-friendly software, monitor and printer. The FOB Gold reagents are based on an antigen–antibody agglutination reaction between the human haemoglobin contained in the sample of faeces in buffer and polyclonal antihuman haemoglobin antibodies absorbed on polystyrene particles. Any agglutination is measured as an increase in absorbance at 570 nm and is proportional to the concentration of human haemoglobin contained in the sample. The calibrator is a lyophilised material containing human haemoglobin, and this is used to generate a six-point calibration curve using serial dilutions of the reconstituted material. Quality control materials, at two haemoglobin concentrations, are also provided lyophilised.

Colonoscopy and pathology

Data on colonoscopy were recorded on a specific form with information on the quality of the investigation (quality of preparation, completeness of colonoscopy) and on the results including number, size and localisation of CRCs and adenomas, and whether biopsy was performed. Full pathological data were available on all excised/biopsy specimens including polyp type and presence or absence of malignancy.

Analysis of results

The participants were divided into groups according to colonoscopy findings. Since the distributions in all groups had significant kurtosis and skewness and were not normally distributed, the data were statistically assessed non-parametrically. Comparison of the medians of the groups was done with the Mann–Whitney U test. Statistical analyses were carried out and graphs of the distributions were prepared using GraphPad Prism, version 3 (GraphPad Software, San Diego, California, USA). Sensitivity and specificity (with 95% CIs) were calculated for identification of high risk adenoma and cancer, primarily to facilitate comparison of the results with data on the previously studied FIT.7 8 Receiver operating characteristic (ROC) analyses and curve generation were done with StatsDirect Statistical Software, version 2.6.6 (StatsDirect, Altrincham, Cheshire, UK).

RESULTS

The 376 participants were classified in the following groups: normal (167, 44.4%), diverticular disease (DD; 43, 11.4%), hyperplastic polyps (HP; 41, 10.9%), low risk adenoma (LRA) according to published criteria12—that is, 1 or 2 polyps <10 mm in diameter (63, 37.7%), higher risk adenoma (HRA)—that is, ⩾3 or any >10 mm diameter (35, 9.3%), and cancer (27, 7.4%). Also shown are the statistical non-parametric characteristics of the distributions of haemoglobin concentrations.

Figure 1 demonstrates the distribution of faecal haemoglobin concentrations in each of these groups.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1 Distribution of faecal haemoglobin concentrations by colonoscopy group. DD, diverticular disease; HP, hyperplastic polyps; LRA, low risk adenoma; HRA, higher risk adenoma. Horizontal bars show median concentrations.

Table 1 shows data on the number and percentages of participants in each group that had faecal haemoglobin concentrations that were l<10 ng/ml, and >25, 50, 75, 100 and 800 ng/ml.

To compare the results in this study with previous data,7 8 Table 2 shows the sensitivity and specificity for HRA plus cancer at four different haemoglobin concentrations.

Figure 2 shows the ROC curve generated using normal + DD + HP + LRA as the group classed as without disease, and HRA + cancer as the group classed as with disease: the area under the ROC curve was 0.7591 (95% CI 0.6910 to 0.8274) with a haemoglobin concentration of 26.7 ng/ml at optimum sensitivity and specificity, giving a sensitivity of 74.1% (95% CI 61.5 to 84.5) and specificity of 70.0% (95% CI 64.7 to 75.1). At the same cut-off concentration, with cancer as the group classed as with disease and all others classed as the group without disease, the area under the ROC curve was 0.8386 (95% CI 0.7830 to 0.8942), giving a sensitivity of 92.6% (95% CI 75.7 to 99.1) and specificity of 66.8% (95% CI 61.6 to 71.7).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2 Receiver operating characteristic curve with normal + DD + HP + LRA as the group classed as without disease, and HRA + cancer as the group classed as with disease. The circle denotes a haemoglobin concentration of 26.7 ng/ml, the optimum concentration for best combined sensitivity and specifity. DD, diverticular disease; HP, hyperplastic polyps; HRA, higher risk adenoma; LRA, low risk adenoma.

DISCUSSION

This study was performed to assess the feasibility of performing quantitative haemoglobin analyses on samples of faeces collected on simple card devices. Overall, 376 individuals with positive gFOBT results on whom useful complete colonoscopy data were available provided samples for this study, and the distribution of colonoscopy findings was typical for a population of gFOBT-positive individuals in the incident round of a screening programme, the positive predictive value being 7.2% for cancer and 26.1% for all adenoma. Our results again highlight the major disadvantage of gFOBTs—that is, that many individuals found to be positive do not have significant neoplasia on colonoscopy.

We have shown that haemoglobin concentration can be quantitated using faecal samples collected on simple cards (hema-screen DEVEL-A-TAB) using the SENTiFOB open automated photometric analytical system with FOB Gold immunoturbimetric reagents, calibrators and controls. As shown in fig 1, the haemoglobin concentrations found in all of the groups had large dispersions. There were no statistical differences in haemoglobin concentrations between the group with normal colonoscopy and the groups with DD, HP or LRA. However, those with HRA had significantly higher haemoglobin concentrations than those with normal colonoscopy, as did those with cancer. Those with cancer had significantly higher haemoglobin concentrations than those with HRA. Thus, we have confirmed the very important finding that haemoglobin concentrations in faeces do rise from normal through HRA through cancer.13^–17

One approach to simpler decision making is to divide the participants into those who warrant colonoscopy and those who do not. The haemoglobin concentrations in the former (those with normal colonoscopy + DD + HP + LRA) compared with the distribution in the latter (those with HRA + cancer), although there is overlap, are statistically highly significantly different. Interestingly, none of the participants with cancer had haemoglobin <10 ng/ml, but 6 (17.1%) of those with HRA and 106 (33.9%) in the group who do not warrant colonoscopy had this finding: in contrast, only 18 (5.8%) of those in the group that would not warrant colonoscopy had haemoglobin concentrations >800 ng/ml whereas 11 (17.7%) of the HRA + cancer group and 6 (22.2%) of those with cancer did, as shown in table 1. This is similar to the findings in a study on 1000 consecutive ambulatory patients, some of whom were at increased risk for colorectal neoplasia and others who were symptomatic.17 The cancers in our study comprised 7 Dukes’ A, 10 Dukes’ B, 8 Dukes’ C1 and 1 Dukes C2, and there was no relationship between stage and haemoglobin concentration.

The question then arises as to the haemoglobin concentration that should be used in practice to guide colonoscopy. A concentration of 50 ng/ml is the usual analytical detection limit of qualitative FITs such as the hema-screen SPECIFIC studied by us previously,7 75 ng/ml was the optimum cut-off concentration in a study of symptomatic individuals17 and 100 ng/ml is a widely used cut-off concentration in CRC screening.18 As shown in table 2, the lower the haemoglobin concentration used as the decision making cut-off, the greater the yield of true-positive results but, of course, also the greater the number of false-positive results, and vice versa. This is clearly shown in fig 2, a formal ROC curve analysis, and statistical analysis of the data shows that use of a haemoglobin concentration of 26.7 ng/ml would give simultaneous optimum sensitivity and specificity.

Unfortunately, no single cut-off concentration gives the ideal clinical characteristics for a screening test but, rather than use the optimum concentration as determined by ROC analysis, the cut-off concentration used in practice could be selected to give the population positivity rate that could be handled by the available colonoscopy resource. In the population studied here, a representative group of gFOBT-positive individuals, at 25, 50, 75 and 100 ng/ml haemoglobin, the positivity rates would be 39.6, 26.6, 22.1 and 18.4%, respectively. It should be noted that these results are not directly comparable with other published data on quantitative FITs since our study has examined quantitative FIT on gFOBT-positive individuals. Unlike an unselected population, gFOBT-positive individuals would be thought very likely to have haemoglobin in their faeces, but we have shown here that gFOBT is in fact a poor reflection of faecal haemoglobin, as is known from the low positive predictive values found for this particular test.

In our previous studies on qualitative FITs on gFOBT-positive individuals,7 8 for cancer, sensitivity was 95.9% and 95.0% and specificity 59.2% and 39.5%; for cancer plus HRA, sensitivity was 87.8% and 90.1% and specificity 65.3% and 47.8%. Similar characteristics would be gained using the 26.7 ng/ml haemoglobin as a cut-off concentration. This is lower than that used in other studies and programmes, and is undoubtedly related to the smaller amount of faeces obtained with the card collection device. It is vital to recognise that there is no universally applicable cut-off concentration for qualitative or quantitative FITs since the amount of faeces varies with collection device, and the amount of buffer in the sample preparation tubes also varies with different FITs.

In conclusion, we have shown for the first time that the haemoglobin in faeces sampled onto simple card FIT collection devices can be analysed quantitatively. Haemoglobin concentration rises in a statistically significant way from those who do not require colonoscopy, through those with HRA, to those with cancer. As with other quantitative FITs, the cut-off concentration can be set to give appropriate clinical characteristics.19 The advantages of the card collection device used here include simplicity and familiarity for participants, ease of posting to and from participants, a requirement for only two samples of faeces on a single card and the ability to use one specimen preparation tube containing buffer for either qualitative or quantitative analysis. The advantages of the automated FIT used here include the ability to use sample cups or tubes, the speed of analysis, the analytical reproducibility and, most importantly, the ability to vary the exact haemoglobin concentration used to trigger an invitation for colonoscopy on the basis of the resources available. Here we have examined the role of an automated FIT in a two-tier reflex gFOBT/FIT approach. Undoubtedly the automated immunochemical quantitation of haemoglobin in faeces collected on cards could be used as a first-line test in screening for CRC, but further studies are required to establish the exact relationships between faecal haemoglobin concentrations and health and disease in the population invited to participate so that such a programme could be designed to be efficient and effective.