Population screening for colorectal cancer using guaiac-based faecal occult blood testing (gFOBT) has consistently been shown to reduce disease-specific mortality in large-scale, high quality trials.1^–5 As a result, the UK Departments of Health commissioned a demonstration pilot with independent evaluation to assess the feasibility of introducing such a programme into the UK National Health Services.6 This pilot was carried out in two geographical regions, one in Scotland and one in England,7 and the successful outcomes have led to the introduction of screening programmes throughout the UK.8^–10

The combined English and Scottish pilot was carried out over 2 years (2000–2002) in order to simulate the first round of a biennial screening programme and, in Scotland, the original pilot site comprising three Scottish NHS Boards (Grampian, Tayside and Fife) went on to carry out second and third rounds over the subsequent 4 years. The data obtained from the first three rounds of the Scottish pilot provide a unique insight into the effect of a biennial gFOBT screening programme offered to an unselected population.

METHODS

The methodology employed to set up and run the pilot was mandated by the UK National Screening Committee and was based on the Nottingham randomised trial, as this was considered to be most relevant for a UK population. This has been described in detail elsewhere6 and, for the purposes of this paper, only a brief description is given. The Scottish pilot was carried out in Grampian, Tayside and Fife (total population around 1.3 million). Biennial gFOBT was used, and the test kit (hema-screen, Immunostics, Ocean, New Jersey, USA) had, as shown in an independent evaluation,11 identical biochemical performance characteristics to that used in the trials carried out in Nottingham, England,2 Funen, Denmark,3 and Göteborg, Sweden.5 The gFOBT was of the traditional style—that is, designed to collect two samples from each of three separate faeces directly on to filter paper impregnated with guaiac gum through oval windows in the test kit card. The test card was developed and read within 14 days, completing the first sample in an accredited laboratory specially set up for the pilots. Those participants who had 5/6 ovals positive on the initial non-dietary-restricted gFOBT (strong positive) were offered colonoscopy without any further testing. If 1–4 ovals were positive (weak positive), participants were asked to complete another gFOBT and, if any of the subsequent six ovals were positive, colonoscopy was offered. However, if all ovals were negative on this repeat gFOBT, a further repeat gFOBT was requested. Then, if any of the six ovals was positive, colonoscopy was offered. The repeat tests were done with dietary restriction in the first round and without restriction in the second and third rounds. In the third round, repeat testing of those who had a negative test following a weak positive initial test was not undertaken, as the yield of significant neoplastic pathology in this group was negligible.

All men and women aged 50–69 who were residents in the three pilot NHS Boards were identified by the Community Health Index (CHI) and invited to participate by having a test kit sent via the postal service from the Screening Centre in Dundee. The CHI is a unique identifier for each individual registered with a general practitioner in Scotland, and consists of date of birth followed by a four-digit number from which gender can be identified. The package sent in the post included information regarding the nature and purpose of the screening process and instructions as to how to complete the test and send it back to the Centre laboratory for processing. This information included details of both the possible benefits and adverse effects of screening. The laboratory was accredited by CPA (UK) to ISO15189-based standards. In the first round, if the first kit was not returned within 6 weeks, a second kit with a reminder letter was posted. In the second and third rounds, because of a poor response to the reminder in the first round, only a letter with an offer to supply a replacement kit was sent. At no time during the first three rounds was there a major public health campaign aimed at increasing uptake, although some local publicity took place in each of the NHS Board areas.

Colonoscopy was carried out by endoscopists who had attended evaluation sessions at the Endoscopy Unit at St Mark’s Hospital, Northwick Park, and who agreed to submit the results to a central database and quality assurance programme. Every patient in whom colonoscopy was incomplete was offered double-contrast barium enema (DCBE). Histopathological examination of biopsies, polypectomy specimens and resection specimens was carried out by selected specialist consultant gastrointestinal pathologists using an agreed pro forma for reporting biopsies and polypectomy specimens. Slide circulation among the screening pathologists was employed to assess the quality of histological interpretation.

In order to monitor the efficacy of the pilot, a series of key performance indicators (KPIs) were developed,and these are shown in table 1. The data for the KPIs were collected by screening data retrieval staff and analysed by Information Services Scotland, a division of NHS National Services Scotland. For the purposes of analysis, the term cancer was used to indicate any tumour with unequivocal histological evidence of invasive malignancy, and the term polyp cancer was used to denote a cancer that was removed by endoscopic snaring at the time of colonoscopy. If a patient had a cancer diagnosed, he or she was not included in the adenoma group whether or not synchronous adenomas were diagnosed. The term adenoma was used to describe the situation where any adenoma but no cancer was diagnosed, and the term high-risk adenoma followed the British Society of Gastroenterology guidelines12 in denoting any adenoma >1 cm in diameter or any situation where three or more adenomas of any size were found; these guidelines do not take the histology of the adenomas into account. Interval cancers were defined as cancers diagnosed within 2 years of a negative screening test and were identified by a record linkage combining information relating to the screening invitees with the Scottish cancer registry data, hospital admission records and death records.

RESULTS

Screening started on 29 March 2000 with the first dispatch of test kits, and the first round was completed on 31 December 2002. The start and finish dates for invitations for the second and third rounds were 2 December 2002 to 30 April 2005 and 6 May 2005 to 20 July 2007, respectively. The results from these screening rounds are presented in the form of the KPIs as described in table 1 along with the interval cancer rates.

Uptake and positivity

In the first round, 304 245 invitations were sent out, and an evaluable result (all six ovals readable) was obtained in 167 415 (uptake of 55.0%); these figures were 309 803 and 164 077, respectively (uptake of 53.0%), in the second round and 317 864 and 175 853 (uptake of 55.3%) in the third. The uptake was greater in females than in males (table 2). It is also important to note that the response rate to the first invitation sent out increased with each round, and that 13.8% of non-responders in the first round completed testing in the second round, but only 85.4% of first round responders did so; in the third round, 13.1% of non-responders in either of the previous rounds completed testing compared with 83.0% of those who had responded at least once before.

The proportion of those completing the test with a positive result requiring further investigation was 2.07% in the first round, 1.90% in the second round and 1.16% in the third round, with the values being higher in males than in females in all screening rounds (table 2). In the second and third rounds, the positivity rate was higher in those who had not responded before when compared with previous responders.

Colonoscopy

The target set for colonoscopy waiting time was 4 weeks, but this was achieved in only 44.1% in the first round, although it improved to 61.7% in the second round and to 79.1% in the third round. Not all of those with a positive gFOBT result attended for colonoscopy; the uptake was 85.5% in the first round, going up to 89.5% in the second round but dropping again to 81.3% in the third round, with little difference between males and females. Colonoscopy completion (as judged by caecal intubation) was achieved in 88.7% in the first round, 91.6% in the second round and 94.7% in the third round. In all rounds, completion was higher in males than in females. Complications requiring admission occurred in 0.3%, 0.4% and 0.1%, respectively, in the three rounds, and there were no colonoscopy-related deaths (table 3).

Neoplasia detection rate

Of those with a positive gFOBT result, cancer was diagnosed in 0.21% in the first round, 0.12% in the second round and 0.07% in the third round. In all rounds, the cancer detection rates were higher in males than in females. The data for polyp cancers, adenomas and high-risk adenomas are given in table 4.

Stage at diagnosis

In the first round, 49.2% of the cancers were diagnosed at Dukes’ stage A (T1/2, N0), with polyp cancers (T1, Nx) included in the A category. This fell to 40.1% in the second round and to 36.3% in the third round. There was also a small but progressive drop seen in the proportion diagnosed at stage D (M1) (table 5).

Positive predictive value (PPV)

In the first round, 12.0% of those with a positive gFOBT had a cancer diagnosed; this fell to 7.0% in the second round, but was maintained at 7.5% in the third round. The PPVs for cancer, adenoma and total neoplasia are given in table 6. All PPVs were higher in males than in females. In the second and third rounds, the PPVs in all categories were higher in those who had not responded before and lower in those who had.

Interval cancers

When all cancers diagnosed in the population responding to the invitation to be screened within 2 years of the first gFOBT result in the first round were examined, 367 (61.4%) were found to be screen-detected. The remaining 41.9% were broken down into true interval cancers (ie, those who had a negative gFOBT result—30.1%), missed cancers (ie, those whose cancer was diagnosed after a positive gFOBT result and a reportedly normal screening colonoscopy or DCBE—1.2%) and a miscellaneous group made up of those who did not attend for colonoscopy and those who started the test process but did not complete it, to obtain a final evaluable result—7.4%. Similar findings were obtained for the second round, with a slightly higher interval cancer rate of 36.9% (table 7).

DISCUSSION

This report is the first account of the initial three rounds of an operational population-based colorectal cancer screening programme outside of a trial setting. As such, it contains data that have important implications for the rollout of screening throughout the UK and for other countries contemplating comprehensive population screening based on gFOBT.

Perhaps the most important criterion for successful population screening, given an appropriate disease and infrastructure for managing the disease in question, is uptake. In the randomised trial carried out in Nottingham,2 53.4% completed the first screening and, over five screening rounds, 59.6% completed at least one gFOBT. There are, however, no published data that allow analysis of uptake by screening round. In Funen,3 67.0% completed screening in the first round, but since only those who agreed to take part in the first screening round were invited for further screening, there is no information on uptake for subsequent screening rounds. In the Swedish study, 63% accepted screening in the first round and this fell to 60% in the second round13; in Burgundy, however, the first round acceptance rate was 52.8% and this rose to 54% in the second round.14 Data from the Minnesota randomised trial1 are not presented in such a way as to permit scrutiny of uptake by round. In all five major studies, from which the effectiveness of gFOBT screening has been derived, a single cohort of subjects was studied, defined by age at first invitation. Thus, there are no previous data from which to estimate the uptake by screening round in the situation which pertains in “real-life” population screening, where the cohort invited for screening is identified by age at each invitation and is therefore different in each screening round.

For these reasons, the data from the present study are the first to illustrate uptake in a true population screening context. It is interesting to note that the uptake in the first round (55.0%) is very much in keeping with previous findings and, although uptake fell in the second round, it recovered in the third round despite the absence of any major initiatives designed to increase uptake. As expected from previous work, uptake in males was lower than that in females in both rounds, and this difference varied little from round to round. It was gratifying to note that the response to the first invitation increased round on round, indicating an increasing level of initial acceptance, possibly related to familiarity with the programme. It is also important to note that, of the non-responders in the previous rounds, 13.8% completed testing in the second round and 13.1% in the third round, vindicating the approach of inviting previous non-responders. On the other hand, this boost to uptake from previous non-responders was offset almost exactly by significant default rates from previous responders. Optimising uptake must be a priority in any screening programme, and requires scrutiny of invitation techniques, potential involvement of primary care and the construction of the test kit itself.

Turning to positivity rates, the level of 2.07% in the first round in Scotland was very similar to the 2.1% seen in both Nottingham3 and Burgundy.13 This is surprising given the higher upper age limit (74 years) adopted in these studies, but probably reflects the known high colorectal cancer prevalence in North East Scotland. These data are very different, however, from the 1% positivity rate seen in the incidence round in Funen,3 which might be explained by the use of strict dietary restriction from the outset in this study. The positivity rate in the second round was more difficult to predict from previous studies for the reasons given above, but evidence from the Nottingham2 and Burgundy13 studies suggested that it should drop by about half. This was not the case, however, and the magnitude of the drop (2.07–1.90%) was more in keeping with the Funen findings where it dropped from 1.0% to 0.8%. In the third round, on the other hand, positivity did drop, although this may have been due, at least in part, to the change in the algorithm whereby those with weak positive test results followed by a negative result were not offered further re-testing.

Positivity is open to a number of influences. Clearly, the analytical detection limit of the gFOBT for haemoglobin is crucial, but this was identical for the gFOBT used in the present study and in the previous studies being used for comparison.10 However, populations differ in disease prevalence and incidence, and, as gFOBT detects the pseudo-peroxidase activity of haemoglobin rather than haemoglobin itself, dietary considerations may be important. However, recent work has suggested that the positivity rate for low analytical sensitivity gFOBT as used in this and the other studies is not influenced by dietary restriction.15 Moreover, if the faeces on the gFOBT kit is allowed to dry for at least 48 h before testing, as was done in all rounds, plant peroxidases do not cause false-positive results.16

Faecal immunochemical tests (FITs) can be used to overcome this problem16 since these are specific for the globin moiety of human haemoglobin. At present, however, FITs are more expensive than gFOBT, and, in most instances, have much lower analytical detection limits so that their use in population screening leads to much higher positivity rates.17 18 For these reasons, it is likely that gFOBT will continue to be used for some time at least as the initial screening test, although a combination of gFOBT and FIT has significant advantages. Performing an analytically sensitive FIT on a subsequent specimen obtained from those who are gFOBT positive directs colonoscopy to those more likely to benefit, as a negative FIT in this context indicates a very low risk of significant neoplasia.19 For this reason, a sensitive FIT test has been introduced as a second-line test for weak positive gFOBT results for rollout of the Bowel Screening Programme in Scotland.20

Prompt, high quality colonoscopy is clearly an essential component of any colorectal screening programme, and, in the Scottish pilot, the percentage of procedures that were performed within the 4 week target increased from the first to the second round, and the completion rate similarly improved. This illustrates the effect of careful audit and setting targets, and can be seen as an important effect associated with the introduction of a screening programme. Surprisingly, a fairly high proportion of gFOBT-positive individuals declined colonoscopy in all three rounds. Previous analysis of the UK pilot first round data indicated that about half of these individuals were medically unfit or had other good reasons for not undergoing the diagnostic examination,6 but this still leaves a substantial number who chose not to have a colonoscopy in the face of a positive gFOBT result. However, subsequent work has indicated that attendance for colonoscopy is related to attendance for initial consultation after a positive test result, and that telephone consultation results in significantly higher uptake.21

Tumour stage at diagnosis is also an important issue in a screening programme and, although there are some data on the stage of screen-detected cancers in previous studies, there is little information on how this varies from one screening round to another. In the Nottingham study,2 46% of screen-detected tumours diagnosed on the first screen were at stage A, and this dropped to 37% in those diagnosed in subsequent screening rounds. Similar overall results for screen-detected cancer were obtained in the Danish,3 Swedish13 and French14 studies, but in none of these reports is stage by screening round documented. In the present study, 49.2% of the cancers diagnosed in the first round were Dukes’ stage A, and this dropped to 40.1% in the second round and to 36.3% in the third round, perhaps indicating that a proportion of the screen-detected cancers diagnosed in the second and third rounds had been missed in the previous rounds. Given the observed interval cancer rate of about 30%, this seems a plausible explanation.

Equally important is the PPV, as this reflects the false-positive rate and thus the number of unnecessary colonoscopies. In the previous studies, this was around 10% for cancer and around 40% for total neoplasia including all adenomas, but, for the reasons noted above, it is difficult to estimate how this would vary by round in an ongoing population screening programme. Given that the first round will decrease the prevalence of neoplasia in the population offered screening, but have no effect on the number of false-positive tests, when roughly the same population is screened again, the relative false-positive rate will go up, and hence the PPV will go down. This is indeed what was observed in this study, with the PPVs for cancer and adenomas dropping between the first and second rounds. However, as the PPVs for cancer and adenomas were maintained at the same levels from the second to third rounds, it might be that a steady state is reached by the third round.

Finally, the interval cancer rate is an essential performance indicator in any screening programme as it allows estimation of the practical (if not actual) clinical sensitivity of the test. In the Nottingham study,2 approximately half of the cancers arising in responders were classified as interval cancer, but the term “interval” was not defined—that is, there was no stipulated maximum interval between the last negative FOBT result and the diagnosis of the cancer. A similar problem with definition was encountered in the reports of all the other previous studies. In this study, an interval cancer was clearly defined as one that was diagnosed within 2 years of a negative FOBT result, and this accounted for 30.1% of all cancers arising in the population that responded to the invitation to be screened in the first round and 36.9% for the same population in the second round. This rise might be related to an accumulation of interval cancers over two rounds, but it is difficult to predict how this will change with time. Interval cancers clearly represent an important problem, and future work must focus on increasing the sensitivity of the screening process by investigating different screening intervals and testing modalities.

From a UK perspective, it is important to compare this report with the recent account of the second round of the English demonstration pilot.22 The findings reported from England mirrored the Scottish first and second round data in many respects; uptake rates were similar, as were the associations between uptake rate and gender, deprivation and previous participation. As in Scotland, the English pilot observed a drop in uptake in the second round, but it is reassuring that, without any significant intervention, uptake recovered in the third round in Scotland.

Positivity, cancer detection rates and PPV for cancer were all higher in Scotland for both the first and second rounds, and this probably reflects the high disease incidence in North East Scotland.23 In Scotland, positivity fell in the second and third rounds, as would be expected in a population in whom the majority were undergoing incidence screening, and as predicted by previous studies. In England, however, the second round positivity rate was higher than in the first round. This unexpected finding cannot be attributed to differences in the screening algorithms between the two countries as at this time they were identical, but may be related to the withdrawal of South Warwickshire from the second round; this, of course, could only provide the explanation if this region had a lower than average positivity rate.

In summary, many of the data from the randomised trials of gFOBT screening are difficult to extrapolate to an ongoing “routine” population screening programme. Thus, while the results from the first three rounds of the Scottish colorectal cancer screening pilot are very much in keeping with what might be expected from previous studies performed in research settings, this report provides a unique practical insight into what to expect from the first three rounds of a biennial gFOBT-based screening programme. It is likely that each round will bring variations in the KPIs as defined by the pilot until a steady state is reached, but it is still not clear when this will occur. For this reason, it is important to monitor and report the findings of colorectal screening programmes in the UK as they roll out, not only for local planning, but for other countries that choose to embark on similar exercises.