Results and modelling

The change in distribution of CRC over the period 1976–1978 to 1990–1997 is shown in table 1.18 These data show a significant increase in the percentage of tumours diagnosed proximal to the splenic flexure (23.6% in 1976–1978 v36.7% in 1990–1997; p<0.001, χ²=76.33, df =1). As there are a similar number of unknown sites in each period, these will be excluded in future discussion.

The NICR contained data on 2767 CRCs for the years 1993–1995. This larger number of CRCs per year (922.3/year in the database of NICRv 644.1/year in the Northern Ireland CRC registry) reflects a difference in the methods of collecting data. The NICR data were based on histology in 82.5% of cases, imaging in 3.3% of cases, and others (for example, death certificate, clinical diagnosis) in 14.2% of cases. Of the 2767 CRCs in the NICR, 764 (27.6%) were proximal to the splenic flexure, 560 (20.2%) distal colon, 895 (32.4%) rectal, and 548 (19.8%) of unknown site. The percentage of CRCs proximal to the splenic flexure appears small due to the large number of unknown sites (where the diagnosis was histologically proved in 62.2% of cases). If the latter are excluded, these data are broadly comparable with the Northern Ireland CRC registry (the percentage of CRCs of known site proximal to the splenic flexure is 34.4% in the NICR v 38.3% in the Northern Ireland CRC registry). If the actual site distribution among the cases with unknown sites were comparable with known sites, our conclusions would not be materially different. In the future discussion the CRC distribution will be based on the Northern Ireland CRC register.

Indications for carrying out flexible sigmoidoscopies undertaken at the Royal Victoria Hospital are shown in table 2. Table 3 shows the level reached during FS, as determined by the endoscopist. The sigmoid colon was reached 98.8% of the time with the acute bend of the sigmoid/descending colon successfully manoeuvred approximately 74.4% of the time and the terminal ileum actually reached on 1.6% of occasions when the initial aim of the endoscopy was FS. It is probable that a colonoscope was used in such extensive FS examinations. As the aim of the study was to anticipate the likely benefit of FS as a screening tool, the completion rate has not been adjusted for failure due to poor bowel preparation. Also, as examinations for colitis activity were excluded and only nine strictures identified during the study period, adjustment for failure due to the presence of disease (presumably of lower incidence in the general population than in this selected subgroup) would have little effect on the overall completion rates.

There was no significant difference in crude completion rates between men and women (289/375 v 376/504, respectively; p=0.40, χ₂=0.58, df=1) (table4).

It is possible to estimate the number of CRCs which could be detected by FS in Northern Ireland by excluding those where the site is unknown and assuming: (1) the observed FS completion rate in this study would apply to a population screening programme, (2) everyone with CRC underwent FS, (3) that FS is 100% sensitive for CRC, and (4) no other diagnostic tests were applied. Under these assumptions, the observed left to right shift in CRC distribution means that whereas FScould have directly visualised 68.8% (812.9/1182) of cancers between 1976 and 1978, only 56.0% (2784.5/4970) could have been similarly visualised between 1990 and 1997 (p<0.001, χ²=63.38, df=1) (table 5).

When estimating the potential benefit of FS as a screening tool, the following limitations need to be taken into account: (1) FS has a 96.5% sensitivity rate in the rectosigmoid area for CRC (one study examining the use of both FS and double contrast barium enema following a positive faecal occult blood test (FOBT) showed that two of 57 CRC in the rectosigmoid area viewed by FS were missed),20 and (2) FS screening programmes are likely to have a 38% compliance rate in Ireland21 and 49% in the UK.13 Such low compliance rates may be improved by appropriate educational programmes aimed at increasing public awareness of CRC and the advantages of screening combined with more accessible FS, such as community clinics offering complete FS with no unnecessary delay.22

Factors which may increase the diagnostic yield of an FS screening programme include follow up of any abnormality found during FS with a complete colonoscopy. One study has shown that 22.6% (81/358) of proximal CRCs have a synchronous distal abnormality23whereas another has shown it may be as high as 34.5% (40/116).24 Another method of increasing the diagnostic yield of an FS screening programme would be the additional use of FOBT to detect proximal CRCs. However, FOBT is unlikely to provide additional benefit to a FS screening programme25 because: (1) FOBT may have a low sensitivity for asymptomatic proximal lesions,21 ,26 (2) it may be necessary to repeat FOBT every one to two years, requiring intensive patient follow up, and (3) compliance is likely to be low. However, a Japanese study of 11 333 colonoscopies carried out as part of a health checkup suggested that FS as a screening tool on its own would have missed 73.7% of proximal high risk tumours but would have missed only 62.0% when combined with immunological FOBT.27 The use of FOBT will therefore be included in the following discussion.

To estimate the likely sensitivity of a FS screening programme in Northern Ireland, one might assume a 96.5% sensitivity rate in the rectosigmoid area,20 an expected 49% compliance rate,13 indirect detection of 22.6% of proximal CRCs during follow up of any synchronous distal abnormalities,23 and the possibility of an additional 25% of CRCs being detected by concurrent use of non-rehydrated FOBT.28 ,29 In a combined FS and FOBT screening programme it is difficult to know whether patients non-compliant with FS would be compliant with FOBT. For the purposes of the following discussion, FOBT will be assumed to detect 25% of the CRCs not detected (either directly or indirectly) by FS. Such a population screening programme applied to Northern Ireland could have detected significantly more CRCs in 1976–1978 than in 1990–1997 (51.7% v48.2%, respectively; p=0.03, χ²=4.41, df=1) (table6).13 ,23 ,28 ,29 This decrease in sensitivity of 3.5% means that in 1997, when there were 669 surgically treated CRCs in Northern Ireland, 23.4 fewer cancers would have been detected by such a screening programme as a result of the left to right shift in CRC distribution (345.9 v 322.5; p=0.20, χ²=1.50, df=1).

Discussion

This study has examined the potential number of CRCs which could be diagnosed by a screening programme consisting of FS and FOBT. It has not attempted to assess the detection rate of polyps which would be of greater interest to a long term screening programme. However, the changes observed in the distribution of CRC can realistically be assumed to reflect earlier changes in the distribution of potentially neoplastic polyps, allowing conclusions to be made regarding the potential usefulness of FS as a screening tool.

This study has also not attempted to identify the cause of the proximal shift in CRC distribution in Northern Ireland. What is clear is that clinicians in Northern Ireland are currently confronted with a group of patients where an increasing percentage of tumours are proximal to the splenic flexure. Whatever the reason for this shift, it has important implications for screening of CRC.

A problem with any such study on sigmoidoscopy/colonoscopy completion rates is that there is no way of confirming the exact anatomical site reached. However, our results are in keeping with other recently reported studies, for example, 75.2% (8671/11531) of FS carried out for CRC screening in an American study achieved a depth of >40 cm and only 34.2% (3942/11 531) achieved a depth of 60 cm.30Another American study revealed that 80% (144/180) of FS achieved a depth of >40 cm and this was reduced to 70% when the aim of the FS was screening.31 In the latter case the reasons for incomplete FS were technical 51%, disease 15%, and poor preparation/stool 34%. One method for increasing FS completion rates may be to use a colonoscope to ensure adequate scope length to reach the splenic flexure.

In addition to any change in the sensitivity of FS for detecting CRC, any programme leading to positive screening tests for CRC would require follow up colonoscopy to fully visualise the bowel and treat premalignant lesions. On top of the burden of the screening programme itself, this could have considerable implications for limited financial and endoscopy resources. Any endoscopy based screening programme will require a substantial investment in endoscopy facilities and trained staff. It has been estimated14 that FS will have a 10–20% positivity rate if repeated five yearly from 50 to 65 years old, that annual rehydrated FOBT from 50 to 65 years old will have a 30% positivity rate (with 15 positive tests per life saved),32 and that their combined use will have a 50% positivity rate. If 50% of people may ultimately need a colonoscopy, it may be more practical to offer everyone a once off diagnostic colonoscopy at a selected age.

Although CRC screening is an attractive proposition with a 30% reduction in mortality9 saving 5700 lives/year in the UK, this study confirms the previously documented left to right shift in CRC distribution in Northern Ireland18 and demonstrates that if this shift continues, FS will become less successful as a screening tool than is currently predicted.