Safe investigation of isolated change in bowel habit with a flexible sigmoidoscopy? A systematic review and meta-analysis

Abstract

Introduction

Public awareness campaigns have led to increasing referrals of patients to colorectal surgery for possible cancer. Change in bowel habit, is traditionally described as a symptom of a left sided bowel cancer. If this is the case in practice, it raises the potentially attractive option of investigating such patients with flexible sigmoidoscopy only. This study sought to systematically review the literature describing tumour location of patients with bowel cancer presenting with left-sided symptoms to establish the safety of potential investigation of these patients with flexible sigmoidoscopy alone.

Methods

A systematic review of studies reporting both the presenting symptoms of patients with bowel cancer and the location of their cancer in the bowel was prospectively registered (CRD42017072492). MEDLINE, EMBASE and CENTRAL were searched with no date or language restriction.

Results

Seven studies were included. Isolated change in bowel habit (with or without rectal bleeding) was a presenting symptom of 73% (95% CI 41–96%, I2 = 99%) of left-sided cancers but also in 13% (95% CI 2–30%, I2 = 96%) of right-sided cancers. In all patients with cancer who presented with isolated change in bowel habit (with or without rectal bleeding), the cancer was right sided in 8% (95% CI 4–12%, I2 = 69%).

Conclusions

There is a higher than expected risk that if a cancer is diagnosed in a patient presenting with either an isolated change in bowel habit or a combination of change in bowel habit with rectal bleeding, the cancer may be right sided.

Introduction

In order to improve the timely identification and treatment of symptomatic colorectal cancer in the UK, public awareness campaigns such as the ‘Be clear on cancer campaign’¹ were launched, together with attempts to speed up referral times through National Institute for Health and Care Excellence guidelines, with the target that patients are seen in secondary care within two weeks.² However, the promotion of public awareness and the use of targets for urgent cancer referrals have led to an unsustainable increase in the numbers of patients referred to colorectal services needing urgent investigation.³ This in turn has increased healthcare costs, but has yet to increase the numbers of malignancies detected or reduce emergency admissions.⁴

Investigation for possible colorectal cancer is through either flexible sigmoidoscopy, colonoscopy or cross-sectional imaging (with or without pneumocolon), with each of these modalities having advantages and disadvantages. Flexible sigmoidoscopy can safely and quickly image the lumen of the left side of the colon without the need for oral bowel preparation products and with minimal morbidity, with randomised controlled trials suggesting that it may even have utility as a screening investigation,⁵ which is currently being rolled out in the UK. In contrast, colonoscopy is a more technically challenging and a longer procedure, which requires oral bowel preparation and is associated with documented morbidity and even mortality, with a perforation rate approximately double that of flexible sigmoidoscopy,^6,7 However, colonoscopy does allow visualisation of the lumen of the entire colon and may have a greater diagnostic yield.

Common reasons for urgent referral for investigation of possible colorectal cancer include symptoms of change in bowel habit or rectal bleeding, clinical examination findings such as an abdominal mass or haematological results such as iron deficiency anaemia. The pathophysiology of these symptoms has long been postulated and recorded in undergraduate medical textbooks and would suggest that change in bowel habit and fresh rectal bleeding should be regarded as symptoms from a tumour in the left side of the colon.⁸ This theory is supported by several published case series and, in theory, could lead to the use of only flexible sigmoidoscopy for investigation of patients presenting with only a change in bowel habit and/or rectal bleeding.^9,10 For such a policy to become acceptable, there must be sufficient safety evidence that the proportion of right sided cancers that would be present in patients presenting with symptoms of change in bowel habit or rectal bleeding must be equal to or less than the rate of asymptomatic cancers that would be expected from screening an asymptomatic population.

This paper aimed to systematically review the literature describing confirmed cases of colorectal cancer (defined by anatomical location) presenting with an isolated change in bowel habit with or without rectal bleeding. The ultimate aim of this study was to provide evidence as to whether patients presenting with ‘left-sided symptoms’ could be safely investigated with flexible sigmoidoscopy alone.

Materials and methods

Study design

This systematic review was registered prospectively with PROSPERO (registration number CRD42017072492) and was carried out in accordance with the PRISMA statement.¹¹ Any study reporting the presenting symptoms of patients with symptomatic colorectal cancer, including the presence or absence of a change in bowel habit, together with tumour location was included. Presence of the following symptoms: anaemia, weight loss and abdominal pain, were considered not to be ‘left sided’, so patients with any of these symptoms were excluded. Any study which did not report both presenting symptoms and the location of the tumour were excluded.

Literature search

Literature searches were carried out by a trained clinical research librarian using the following databases: MEDLINE, EMBASE and CENTRAL (all inception to 5 September 2017). No language or date restrictions were applied to the searches. The Cochrane Library of Systematic Reviews was searched for relevant reviews and the abstracts from the conference proceedings of both the Association of Surgeons of Great Britain and Ireland and the Association of Coloproctologists of Great Britain and Ireland for the past 10 years were searched for relevant unpublished studies. References of identified potentially relevant studies were hand searched for further studies. Finally, all studies citing the identified potentially relevant primary studies identified on Google Scholar were screened for inclusion.

Abstracts were screened independently by two authors (EH and FH) with the aid of Rayyan systematic review software (2016, Qatar Computing Research Institute).¹² Full-text versions of potentially relevant primary studies were then independently screened against the inclusion and exclusion criteria by two authors (PH and HBC) and agreement to inclusion reached by consensus.

Data extraction

Study characteristics were extracted by two authors (PH and HBC) with outcome data also independently extracted and verified by two authors (PH and JB). Risk of bias for included studies was assessed independently by two authors using the QUADAS-2 tool¹³ (PH and HBC) with any disagreement resolved by consensus. In order to use the QUADAS-2 tool in this review, the method of assessment of the patient’s symptoms was taken as the index test, while the reference standard was the study’s method of assessing for malignancy. When required, data were only reported in graphical form, they were extracted using the online tool WebPlotDigitizer Version 3.12.¹⁴

Statistical analysis

We aggregated proportional outcomes using the metaprop user written command for STATA.¹⁵ Data were pooled using a random effects model and 95% confidence intervals (CI) were calculated using the exact method. Statistical heterogeneity was assessed using I2 statistic with values greater than 50% regarded as evidence of statistical heterogeneity. Publication bias was assessed using Funnel plots constructed with study size on the Y axis, as suggested previously if more than 10 studies were included.¹⁶ All analyses were performed using Stata Version 15.

Results

Search results

Figure 1 shows the PRISMA flowchart. No relevant unpublished studies were identified from the hand search of conference proceedings. Of the 2400 abstracts screened, 2341 were excluded as not being relevant, leaving 59 studies for full-text review. Of these 59 studies, 40 were excluded, leaving 19 studies potentially for inclusion in the qualitative/quantitative analysis.^9,10,17–33 Two studies of the 19 remaining represented data drawn from the same patient data set of an earlier study and so the latest publication was used in this review.^9,10

Figure 1.

PRISMA flow diagram (from: Moher D, Liberati A, Tetzlaff J, Altman DG, The PRISMA Group preferred reporting items for systematicreviews and meta-analyses: the PRISMA statement. PLoS Med 2009; 6(7): e1000097).

Study characteristics

The characteristics of the included studies are shown in Table 1. The earliest study meeting inclusion criteria was published in 1955 and the latest in 2017. Eighteen studies were published journal articles and one was a research letter.²⁴ Full texts were obtained for 18 of the 19 studies. The full text of one study²¹ was not obtainable, despite a request to the British Library.

Table 1.

Characteristics of included studies.

Study	Country	Sample size (n)	Sample demographics	Diagnostic test	Individualsa	Definition of location	Cancers (n)
Alexiusdottir et al, 201217	Iceland	768	422 male, 346 female. Median age 73 years	Unclear	N	Right: proximal to splenic flexure. Left distal to splenic flexure	768
Badiani et al, 201218	UK	340	Limited information on demographics	Unclear	Y	Right sided: proximal to splenic flexure	251
Bedir et al, 201619	Turkey	132	74 male, 58 female. Mean age 63 years	Unclear	N	Not stated	132
Ben-ishay et al, 201320	Israel	236	124 male, 94 female. Mean age 71.5 years	Unclear	N	Not stated	236
Bhati et al, 200421	Pakistan	50	Unclear	Colonoscopy	N	Not stated	17
Bloem et al, 198822	Netherlands	624	Male : female 1 : 1.04. Mean age 66 years	Unclear	N	Right: caecum to hepatic flexure. Left: descending colon to rectosigmoid	563
Christodoulidis et al, 201023	Greece	453	256 male, 197 female. Mean age 66.3 years	Colonoscopy	N	Right: caecum-transverse colon. Left: Splenic flexure to sigmoid colon.	453
Ingham Clark et al, 201024	UK	835	Unclear	Colonoscopy	Y	Not stated	21
Couch et al, 201525	UK	968	Unclear	Colonoscopy	Y	Right: proximal to splenic flexure. Left: distal to splenic flexure	35
Curless et al, 199426	UK	273	Median age 68 years; > 50% male	Unclear	N	Splenic flexure	273
Kent et al, 201027	UK	2404	76 female, 83 male. Similar ages between groups	Colonoscopy and flexible sigmoidoscopy	Y	Right: proximal to splenic flexure. Left: level of or distal to splenic flexure	153
Majumadar et al, 199928	USA	194	Male : female 1 : 1. Mean age 66 years.	Unclear	N	Distal: rectum, sigmoid, descending. Proximal: transverse, ascending or caecum.	194
Masson et al, 200729	UK	143	85 male, 58 female. Mean age 67.4 years	Unclear	N	Not stated	143
Nawa et al, 200830	Japan	3565	2030 male 1522 female. Mean age 67.7 years	Unclear	N	Right: caecum, ascending and transverse colon. Left: distal to splenic flexure.	3510
Royle et al, 201431	UK	1606	729 male; median age 68 years	Flexible sigmoidoscopy	Y	Splenic flexure	117
Saidi et al, 200832	Kenya	250	Males 65%; mean age 49.3 years	Histology	N	Right: caecum, ascending colon, transverse colon. Left: splenic flexure, descending and sigmoid colon	253
Shallow et al, 195533	USA	750	Males 56%; age range 16–92 years	Sigmoidoscopy barium enema, surgery	N	Not stated	750
Thompson et al, 200810	UK	16433	Unclear	Flexible sigmoidoscopy	Y	Left: proximal to sigmoid colon	946
Thompson et al, 20179	UK	29005	Unclear	Flexible sigmoidoscopy	Y	Left: proximal to sigmoid colon	1585

^a Reports individual numbers of patients with different symptom clusters.

Risk of bias

Table 2 displays the risk of bias of included studies scored using the QUADAS-2 tool.¹³

Table 2.

Risk of bias of included studies

Study	Risk of bias				Applicability concerns
	Patient selection	Index test	Reference standard	Flow and timing	Patient selection	Index test	Reference standard
Alexiusdottir et al, 201217	Low	Unclear	Low	Unclear	Low	Low	Low
Badiani et al, 201218	High	Low	Low	Unclear	Low	Low	Low
Bedir et al, 201619	Low	Unclear	Low	Unclear	Low	Low	Low
Ben-ishay et al, 201320	Low	Unclear	Unclear	Unclear	Low	Low	Low
Bhati et al, 200421	Unclear	Unclear	Unclear	Unclear	Low	Low	Low
Bloem et al, 198822	Low	Unclear	Unclear	Unclear	Low	Low	Low
Christodoulidis et al, 201023	Low	Unclear	Low	Unclear	Low	Low	Low
Ingham Clark et al, 201024	Low	Unclear	Low	Unclear	Low	Low	Low
Couch et al, 201525	Low	Unclear	Low	Unclear	Low	Low	Low
Curless et al, 199426	Low	Low	Low	Low	Low	Low	Low
Kent et al, 201027	Low	Unclear	Low	Unclear	Low	Low	Low
Majumadar et al, 199928	Low	Unclear	Unclear	Unclear	Low	Low	Low
Masson et al, 200729	Low	Unclear	Unclear	Unclear	Low	Low	Low
Nawa et al, 200830	Low	Unclear	High	Unclear	Low	Low	Low
Royle et al, 201431	Unclear	Unclear	High	Unclear	Low	Low	Low
Saidi et al, 200832	Low	Unclear	Low	Unclear	Low	Low	Low
Shallow et al, 195533	Unclear	Unclear	High	Unclear	Low	Low	Low
Thompson et al, 200810	Low	Low	High	Low	Low	Low	Low
Thompson et al, 20179	Low	Low	Low	Low	Low	Low	Low

Data synthesis

Of the 18 included studies, only 5 reported presenting symptoms and combinations of presenting symptoms in sufficient detail to allow a meta-analysis of all patients presenting with an isolated change in bowel habit, with or without rectal bleeding.^{9,18,24,25,27} Requests for individual patient data were made to the corresponding authors of the remaining articles where possible, with two authors able to provide raw data for inclusion in the analysis.^22,31

Left-sided cancer

Five studies contributed data for left-sided cancers. Overall, the proportion of patients with change in bowel habit with or without rectal bleeding was 73% (95% CI 41–96%; Fig 2). There was evidence of considerable statistical heterogeneity (I2 = 99%). There were too few studies to assess publication bias.

Figure 2.

Proportion of patients with a left-sided cancer presenting with an isolated change in bowel habit (with or without rectal bleeding); ES = effect size.

Right-sided cancer

For right-sided cancers, six studies contributed data. Overall, the proportion of patients with change in bowel habit with or without rectal bleeding was 13% (95% CI 2–30%; Fig 3). There was evidence of considerable statistical heterogeneity (I2 = 96%). There were too few studies to assess publication bias.

Figure 3.

Proportion of patients with a right-sided cancer presenting with an isolated change in bowel habit (with or without rectal bleeding); ES = effect size.

All cancers presenting with a change in bowel habit

For cancer patients with a change in bowel habit (with or without rectal bleeding; five studies), the proportion of patients with a right-sided cancer was 8% (95% CI 4–12%; Fig 4). There was evidence of statistical heterogeneity (I2 = 69%). There were too few studies to assess publication bias.

Figure 4.

Proportion of all colorectal cancer patients presenting with an isolated change in bowel habit (with or without rectal bleeding) who have a right-sided cancer; ES = effect size.

Discussion

This systematic review has demonstrated that there is a likelihood that if a patient attending colorectal clinic with an isolated change in bowel habit (with or without rectal bleeding) that if they do have a colorectal carcinoma, it will be a left-sided tumour. This is in keeping with the basic understanding of pathophysiology of the method by which colorectal cancers present symptomatically, as is taught in both undergraduate and postgraduate textbooks.^8,34 However, this review has also demonstrated that around 13% of right-sided tumours may also present with this symptomology. Further, this review demonstrates that 8% of cancer patients referred with isolated ‘left-sided symptoms’ may in fact have a right-sided cancer. This finding means that caution must be exercised in using flexible sigmoidoscopy as the sole method of investigation for patients presenting with this symptomology, as the proportion of right sided lesions is higher than would be expected.

The findings of this review may, however, add weight to policy of some colorectal services of straight-to-test to flexible sigmoidoscopy in patients referred with isolated change in bowel habit (with or without rectal bleeding) or the provision of flexible sigmoidoscopy on the same day as clinic attendance, as described in some of the papers in this review.³¹ The decision on proceeding to imaging the rest of the colon must then come down to a shared decision between the endoscopist/colorectal surgeon and the patient.

There remains the potential for confounding in this patient population, as patients may present symptomatically with symptoms caused by a benign condition such as diverticular disease, but are found to also have cancer on investigation. Indeed, a recent systematic review of the diagnostic value of colorectal cancer symptoms presenting to primary care concluded a combination of both change in bowel habit and rectal bleeding to be only 60% specific for cancer. However, the rate of 8% of right-sided cancers far exceeds the background rate that would be expected from screening an asymptomatic population which ranges from 1% to 2.3%.^35–37

Limitations

Substantial amounts of statistical heterogeneity were identified between the studies contained in this review, which does limit the precision of the findings of the meta-analysis. However there were insufficient studies to permit further investigation of the reasons for this heterogeneity.

This review was mainly limited by a lack of access to individual patient data from a large proportion of the studies that it included, thus reducing its ability to include all patients in the final analysis. The majority of studies were judged to be an unclear risk of bias in their method of assessing a patient’s symptoms, with very few reporting the use of standardised questions or a proforma, which may have influenced the proportion of patients reporting or not reporting particular symptoms.

This review may also be limited due to the use of barium enema as a reference test by four of the included studies, which has lower sensitivity for right sided cancers than colonoscopy.³⁸ However, only one of these included studies was included in the final quantitative analysis owing to lack of response to individual patient data requests, which could reduce the impact of this limitation.

Conclusion

There is a higher than expected risk that if a cancer is diagnosed in a patient presenting with either an isolated change in bowel habit or a combination of change in bowel habit with rectal bleeding, the cancer may be right sided. The use of flexible sigmoidoscopy only to investigate patients presenting in this way should be carefully considered as it may miss a higher number of malignancies than expected.