Introduction

Colonoscopy is the gold standard modality for the detection of colorectal cancer and, along with radiological imaging, determines lesion localisation leading to optimal pre-operative surgical planning [1, 2]. However, the accuracy of colonoscopy for segmentally localising tumours within the bowel is unclear with previous publications, with varying methodology, stating accuracies from as low as 59.7 to 98.3 % [1–15]. Colorectal surgical resection in the modern era is influenced by the reduced tactility of more frequently performed laparoscopic surgery and the earlier detection of smaller colorectal lesions through the NHS Bowel Cancer Screening Programme (NHSBCSP) [16, 17]. To determine accurate lesion localisation at colonoscopy in this modern era, we recently performed a prospective multi-centred observational study that found inaccurate lesion localisation in 19 % of patients that led to an on-table alteration in surgical management in 6 % [18]. Furthermore, pre-operative CT imaging could not detect these smaller lesions in almost 30 % of cases, confirming the reliance of modern day surgical planning on colonoscopy.

The aim of this paper was to build on our initial work by performing an analysis of potentially influencing factors on accurate lesion localisation at colonoscopy.

Patients and methods

Patients were recruited to this prospective, multi-centred, observational study from five West of Scotland colorectal centres over a six-month period from October 2011 to March 2012 and over a seven-month period from October 2012 to April 2013. This study was registered with the Clinical Effectiveness Departments of NHS Greater Glasgow and Clyde and NHS Ayrshire and Arran health boards. All patients were undergoing colonoscopy because of a positive faecal occult blood test through NHSBCSP (screening patients) [17] or because they presented to primary care with symptoms which warranted referral for colonoscopic examination (symptomatic patients). All colonoscopies were performed or supervised by consultants who routinely performed symptomatic and screening colonoscopies.

Eligible patients were identified either at colonoscopy or upon discussion of their management at multi-disciplinary team (MDT) meetings of colorectal teams at each recruiting centre. Patients were included if they met all of the following criteria: patient had a primary colorectal tumour, tumour was identified at colonoscopy and patient underwent elective, curative, surgical resection. Patients were excluded if treated with neo-adjuvant chemoradiotherapy (as they would undergo multiple staging investigations), or surgical resection was palliative.

True tumour location was defined as the intra-operative location, and this was compared to the locations at pre-operative colonoscopy and on radiological imaging (computed tomography (CT)). The bowel was divided into nine segments/locations to standardise reporting. Colonoscopic location was reported at the time of the colonoscopy with the following patient details recorded: age, sex, screening or symptomatic referral, previous abdominal surgery (including colorectal resections) and previous colorectal surgery only. Colonoscopic factors recorded were as follows: complete colonoscopy, difficulties encountered (e.g. diverticular disease, sigmoid looping) and use of a magnetic endoscopic imaging guide (MEI). Intra-operative tumour location was reported directly after resection by the operating surgeon by means of a standardised questionnaire. In addition, the surgeon also recorded the following: details of any intra-operative changes and accuracy of tattoo placement at colonoscopy.

All categorical variables were analysed with chi-squared or Fisher’s exact tests. All numerical variables were analysed with t tests, with appropriate generation of 95 % confidence intervals. All analyses were performed on SPSS® software version 18 (SPSS, Chicago, Illinois, USA), and p values of <0.05 were considered statistically significant.

Results

Of 133 patients identified, 22 were excluded, leaving a final sample size of 111 patients (Fig. 1). The mean age was 67.4 years (range 27–89) with male:female ratio of 1.3:1. The majority of patients were symptomatic referrals (n = 78, 70.3 %) with only 27 (24.3 %) having previously had any abdominal surgery. Complete colonoscopy was recorded in 78 patients (70.3 %) with tumour stenosis accounting for the majority of incomplete colonoscopies (n = 23).

Fig. 1
figure 1

Sample size of study and specific reasons for exclusion

Full size image

One tumour was identified at colonoscopy in all 111 patients. In 88 patients, the tumour location at colonoscopy matched the tumour location intra-operatively, giving a colonoscopic lesion localisation accuracy of 79.3 %. Inaccurate lesion localisations were spread throughout the nine pre-determined anatomical segments (Table 1). A subdivision of these anatomical segments into super-segments (Fig. 2) identified significantly reduced accuracy of lesion localisation from the hepatic flexure to the sigmoid colon when compared to the ascending colon/caecum and the rectum (70.7 versus 90.6 and 90.0 %, respectively; p = 0.036).

Table 1 The relationship between colonoscopic localisation and true (intra-operative) location and the colonoscopic localisation accuracies of each bowel segment
Full size table
Fig. 2
figure 2

Colonoscopic localisation accuracies of each super-segment (n = 110, this figure does not include the one case of colonoscopically-localised anastomotic recurrence)

Full size image

Of 111 patients, 110 had a pre-operative CT scan (one was not imaged pre-operatively and was excluded from this analysis). CT imaging was unable to identify the tumour in 24 cases (21.8 %). Of the 86 patients with identified lesions on CT, tumour location matched the true intra-operative location in 74 cases giving an accuracy of 86.0 %.

Of the 23 tumours inaccurately localised by pre-operative colonoscopy, surgical management was altered intra-operatively in seven cases (6.3 %) (Table 2). Of these seven cases, all were planned open procedures with alterations in management varying from extending the dissection distally (e.g. sigmoid colectomy to anterior resection) to more significant changes (e.g. potential recurrence at ileo-caecal anastomosis at colonoscopy found to be metachronous sigmoid cancer with planned local resection altered to subtotal colectomy). Surgical management was not altered in the remaining 16 cases either because pre-operative radiological imaging correctly localised the tumour and this was known when planning surgical management (8 of 16 cases) or because true lesion location was within the anatomical boundaries of the planned surgical resection (8 of 16 cases).

Table 2 Details of inaccurately localised tumours and changes in surgical management
Full size table

Potential influencing patient and colonoscopic factors on accurate lesion localisation at colonoscopy found complete colonoscopy to be statistically significant (p = 0.008) (Table 3). None of the other patient or colonoscopic factors analysed influenced accurate lesion localisation.

Table 3 Analysis of potential influencing patient and colonoscopic factors on accurate lesion localisation at colonoscopy
Full size table

Discussion

This is the first prospective, multi-centred study to analyse the factors that potentially influence accurate lesion localisation at colonoscopy. With the NHS Bowel Cancer Screening Programme detecting smaller and earlier cancers, this study builds on recent work by the same authors confirming colonoscopic lesion localisation to be only 79 % accurate with potentially significant on-table alterations to the planned surgical management [18]. In addition, it highlights that radiological imaging is limited with over a fifth of lesions not seen on pre-operative imaging, increasing the reliance on colonoscopy for accurate lesion localisation in current colorectal practice. Therefore, identification of patient or colonoscopic factors that influence lesion localisation accuracy at colonoscopy is extremely important.

The first finding from this study is that an incomplete colonoscopy significantly decreases the colonoscopist’s ability to accurately localise a colorectal lesion. This is likely to be a direct result of the colonoscopist having visual exposure to only some of the anatomical landmarks that aid localisation. A complete colonoscopy should allow the practitioner to orientate the scope using the characteristic triangular appearance of the transverse colon, the bluish hue of the liver at the hepatic flexure, the ileo-caecal valve and the appendix orifice signifying completion of the colonoscopy, and ileal intubation to confirm completion, to name some of the key features. In addition, the complete colonoscopy allows two views, one on insertion and one on withdrawal, which increases the probability of accurately detecting and localising a lesion.

The second finding from this study is that use of the magnetic endoscopic imaging (MEI) device to aid localisation did not significantly improve lesion localisation. It may be that the main use of the scope guide is in helping to acknowledge and remove colonic loops early, allowing easy and comfortable passage of the scope to its end point. However, this study is limited in that it is not powered for MEI and further work, perhaps from large observational studies, would determine the role of MEI, particularly in the area where most inaccuracies were found (hepatic flexure to sigmoid).

Few previous studies have assessed potential factors that influence lesion localisation [1, 5, 7]. Of these three studies, with varying methodology, different influencing factors were proposed: Vaziri et al. found that increasing age of the patient increased the likelihood that lesions would be inaccurately localised [1]. This contrasts with Piscatelli et al. and Borda et al. who did not find any significant relationship with age and inaccurate lesion localisation at colonoscopy [5, 7]. In addition, patient’s gender was not an influencing factor. Piscatelli et al. did find that previous colorectal surgery significantly increased lesion inaccuracy. Although results from our study did not support this, only a small percentage of our patients had undergone previous resection (3.6 %), but it is worth highlighting that within this small group, one patient had a significant change to their planned surgical management due to their inaccurately localised lesion. The finding of incomplete colonoscopy to be significant is in agreement with that of Borda et al., who reported decreased localisation accuracy in “cases of obstructive cancer” (i.e. those which prevent the completion of colonoscopy) (p < 0.05), but is in disagreement with that of Piscatelli et al., who concluded no influence of complete colonoscopy on localisation accuracy [5, 7].

The reason for these alternate and conflicting views can be seen on Table 4 that displays the result of a literature review of colonoscopic lesion studies [1–15, 18]. From this table, it can be seen that there is a significant variation in the methodology of these studies, with the majority being single centred, if not single colonoscopist, retrospective and several of low numbers. In addition, comparisons are difficult as the pre-defined bowel segments that document the anatomical location of the lesion vary between studies (from 3 to 13), making an interpretation of lesion localisation accuracy at colonoscopy difficult.

Table 4 Sixteen studies identified at literature review summarised by methodology and results and ordered by publication date
Full size table

One of the highest values for accuracy was reported by Vaziri et al., who also had one of the largest patient cohorts [1]. Three hundred and seventy four patients were analysed, and a colonoscopic localisation accuracy of 96.0 % was found. However, only one surgeon performed all procedures, and their patient cohort was identified retrospectively. This current study had a smaller patient cohort; however, its patients were recruited from five colorectal centres prospectively with over fifty-five trained endoscopists involved, and this is a pragmatic study that reflects outcomes from the current clinical practice. Louis et al. also benefited from one of the largest sample sizes: four hundred patients were retrospectively analysed, and a colonoscopic localisation accuracy of 88.0 % was concluded [2]. Again, with methodological advantage, Louis et al. standardised the bowel into four segments, potentially biassing results towards improved accuracy.

The findings from this prospective study confirm what many surgeons will have experienced in their clinical practice: pre-operative lesion localisation at colonoscopy carries a degree of inaccuracy that the surgeon must be aware of, especially if performing laparoscopic surgical resection for impalpable colonic tumours. Further work is underway in a larger multi-centred observational study to build on this study’s findings and to particularly assess the role of magnetic endoscopic imaging devices. However, it is possible that alternative approaches to lesion localisation may have to be considered, for example colonoscopic placement of a magnetic clip allowing intra-operative lesion detection at laparoscopy [19].

Conclusion

This is the first study to analyse potentially influencing factors to improve lesion localisation at colonoscopy, leading to optimal surgical planning and patient outcomes. Colonoscopists should be aware that inaccurate lesion localisation can occur in any area of the bowel and that particular attention should be paid to the area from the hepatic flexure to the sigmoid colon. Incomplete colonoscopy should warn the colonoscopist to be extremely careful with localising lesions, and perhaps requesting a CT pneumocolon to aid localisation is to be considered.