Association Between Constipation and Colorectal Cancer: Systematic ...

894

CLINICAL AND SYSTEMATIC REVIEWS

nature publishing group

REVIEW

CME

Association Between Constipation and Colorectal Cancer: Systematic Review and Meta-Analysis of Observational Studies Andrew M. Power, BChD, MBChB, MFDS1, Nicholas J. Talley, MD, PhD2 and Alexander C. Ford, MBChB, MD, FRCP1,3 OBJECTIVES:

Constipation is common in the community, and may affect survival adversely. An association between constipation and development of colorectal cancer (CRC) could be one possible explanation. We performed a systematic review and meta-analysis examining this issue.

METHODS:

We searched MEDLINE, EMBASE, and EMBASE Classic (through July 2012). Eligible studies were cross-sectional surveys, cohort studies, or case–control studies reporting the association between constipation and CRC. For cross-sectional surveys and cohort studies, we recorded number of subjects with CRC according to the constipation status, and for case–control studies, number of subjects with constipation according to CRC status were recorded. Study quality was assessed according to published criteria. Data were pooled using a random effects model, and the association between CRC and constipation was summarized using an odds ratio (OR) with a 95% confidence interval (CI).

RESULTS:

The search strategy identified 2,282 citations, of which 28 were eligible. In eight cross-sectional surveys, presence of constipation as the primary indication for colonoscopy was associated with a lower prevalence of CRC (OR = 0.56; 95% CI 0.36–0.89). There was a trend toward a reduction in odds of CRC in constipation in three cohort studies (OR = 0.80; 95% CI 0.61–1.04). The prevalence of constipation in CRC was significantly higher than in controls without CRC in 17 case–control studies (OR = 1.68; 95% CI 1.29–2.18), but with significant heterogeneity, and possible publication bias.

CONCLUSIONS: Prospective cross-sectional surveys and cohort studies demonstrate no increase in prevalence of

CRC in patients or individuals with constipation. The significant association observed in case–control studies may relate to recall bias. Am J Gastroenterol 2013; 108:894–903; doi:10.1038/ajg.2013.52; published online 12 March 2013

INTRODUCTION Symptoms attributable to the lower gastrointestinal (GI) tract are common in the community (1–5). Most of these, such as chronic idiopathic constipation and irritable bowel syndrome, are functional in nature. It is often assumed, despite their relapsing and remitting natural history and adverse impact on healthrelated quality of life (6,7), that these functional disorders run a benign course, and do not affect mortality. Data from two large longitudinal studies, with a considerable length of follow-up, suggest that this is the case for most of these conditions (8,9), but in one of these studies the presence of chronic constipation

at baseline appeared to be associated with a reduced likelihood of survival (8). The reasons for any reduction in survival in chronic constipation are unclear. The condition may be associated with the development of some organic anorectal and colonic pathologies, including rectal prolapse, hemorrhoids, anal fissure, and diverticular disease, perhaps due to straining and an increase in colonic transit time (10). This delay in colonic transit has also been proposed, by some, as a potential etiological mechanism in the development of colorectal cancer (CRC) (11), due to prolongation of contact between carcinogens in the stool, such as bile acids (12), and the colonic mucosa.

1

Leeds Gastroenterology Institute, St. James’s University Hospital, Leeds, UK; 2Faculty of Health, University of Newcastle, New South Wales, Newcastle, Australia; Leeds Institute of Molecular Medicine, University of Leeds, Leeds, UK. Correspondence: Alexander C. Ford, MBchB, MD, FRCP, Leeds Gastroenterology Institute, St. James’s University Hospital, Room 125, 4th Floor, Bexley Wing, Beckett Street, Leeds LS9 7TF, UK. E-mail: [email protected] Received 18 September 2012; accepted 19 January 2013 3

The American Journal of GASTROENTEROLOGY

VOLUME 108 | JUNE 2013 www.amjgastro.com

Constipation and Colorectal Cancer

METHODS

Box 1. Eligibility Criteria Cross-sectional surveys, cohort studies, or case–control studies Adults (aged ≥16 years). Presence of constipation recorded at study inclusion (using a symptom questionnaire, Rome I, II, or III criteria, or self-reported). Diagnosis of CRC recorded (after lower GI investigation (using lower GI endoscopy, barium enema, or CT colonography) in cross-sectional surveys, or cancer registry data or medical records in cohort studies or case–control studies). Prevalence of CRC reported according to constipation symptom status in cross-sectional surveys and cohort studies. Prevalence of constipation reported according to CRC status in case–control studies. Sample size of ≥50 participants.

Search strategy and study selection

We performed a literature search using MEDLINE (1946 to July 2012), EMBASE, and EMBASE CLASSIC (1947 to July 2012) to identify cross-sectional surveys, cohort studies, or case–control studies that examined the association between constipation and CRC in adults (aged 16 years and over). In order to be eligible, studies had to recruit at least 50 participants, and define constipation using a symptom questionnaire, Rome I, II, or III criteria, or self-report at interview. The diagnosis of CRC could be made using lower GI endoscopy, barium enema, or computed tomography colonography in cross-sectional surveys, and cancer registry data or medical records in cohort and case–control studies. These eligibility criteria, which were defined prospectively, are provided in Box 1. The medical literature was searched using the following terms: colon cancer, rectal cancer, colorectal cancer, bowel cancer, colon adj5 cancer, rectal adj5 cancer, colorectal adj5 cancer, bowel adj5 cancer, colon adj5 carcinoma, rectal adj5 carcinoma, colorectal adj5 carcinoma, bowel adj5 carcinoma, colon adj5 adenocarcinoma, rectal adj5 adenocarcinoma, colorectal adj5 adenocarcinoma, or bowel adj5 adenocarcinoma. These were combined using the set operator AND with studies identified with the terms: constipation, functional constipation, idiopathic constipation, chronic constipation, or slow transit. There were no language restrictions. All abstracts yielded by the search were then screened for potential suitability, and those that appeared relevant were retrieved and examined in more detail. We performed a recursive search using the bibliographies of all obtained articles. We translated foreign language articles, where required. Eligibility assessment was performed independently by two investigators, using pre-designed eligibility forms, with all disagreements resolved by consensus. Data extraction

Data were extracted independently by two investigators on to a Microsoft Excel spreadsheet (XP professional edition; Microsoft, Redmond, WA), again with any discrepancies resolved by consensus. The following data were collected for all studies: year conducted, country, number of centers (where applicable), study setting, type of study design, method of symptom data collection, method used to confirm the presence or absence of CRC, criteria © 2013 by the American College of Gastroenterology

used to define constipation, total number of subjects providing complete data, the number of subjects with constipation, and the number of subjects with CRC. For cross-sectional surveys, where all participants were patients reporting lower GI symptoms, we recorded the number of subjects with CRC among those with constipation as the primary indication for colonoscopy, compared with the number of subjects with CRC among those whose primary indication was for other lower GI symptoms. Several of these studies included patients who required endoscopic visualization of abnormalities detected at barium enema, or who were asymptomatic but undergoing follow-up or surveillance for previous colorectal carcinoma, polyps, or inflammatory bowel disease. These groups of patients were always excluded from our analyses, as they were not relevant to the clinical question we were addressing. We assessed quality of the identified cross-sectional surveys according to the QUADAS-2 tool, as we were using them as diagnostic studies (17). This does not give an overall quality score, but rather assesses the risk of bias of individual studies according to various domains. We deemed high quality studies to be ones that were at low risk of bias across six or more of these seven domains. For cohort studies we recorded the number of subjects with CRC among those with or without constipation, and for case–control studies the number of subjects with constipation among cases with and controls without CRC. For cohort and case–control studies we assessed study quality using the Newcastle-Ottawa scale (18), with a total possible score of 9, higher scores indicating higher quality studies. Data synthesis and statistical analysis

The degree of agreement between the two investigators, in terms of judging study eligibility, was measured using the Kappa statistic. For cross-sectional surveys the proportion of individuals with and without constipation as the primary indication for colonoscopy who were found to have CRC in each study were compared using an OR, with a 95% confidence interval (CI). For cohort studies, the proportion of individuals with and without constipation who were found to have CRC were compared using an OR, with a 95% CI. For case–control studies the proportion of individuals with The American Journal of GASTROENTEROLOGY

REVIEW

If this theory were correct, it may contribute to an excess in mortality in individuals with constipation, compared with those without such symptoms. A meta-analysis of nine case–control studies (13), published in 1993, suggested that this was the case, with a pooled odds ratio (OR) for CRC in subjects with constipation of 1.48. However, subsequent studies have not all replicated such an association (14–16), and there have been much data published in the intervening 20 years. We have, therefore, conducted a systematic review and meta-analysis of all available observational studies examining this issue. We hypothesized that CRC is not causally linked to chronic constipation.

895

896

Power et al.

constipation status, and the odds of constipation according to the presence or absence of CRC. StatsDirect version 2.7.2 (StatsDirect, Sale, Cheshire, England) was used to generate Forest plots of pooled ORs with 95% CIs. Evidence of publication bias was assessed for, by applying Egger’s test to funnel plots (21), where a sufficient number of studies were available (22).

REVIEW

Studies identified in literature search (n = 2282)

Excluded (title and abstract revealed not appropriate) (n = 2195)

RESULTS

Studies retrieved for evaluation (n = 87) Excluded (n = 59) because: • Case series = 13 • Review article = 12 • Prevalence of CRC according to constipation status not reported = 8 • Retrospective surveys = 7 • Systematic review = 6 • Diagnosis of CRC not reported = 5 • Prevalence of constipation not reported = 3 • Prevalence of colonic adenoma only = 2 • Data not extractable = 2 • Case report = 1

Studies reporting association between constipation and colorectal carcinoma (n = 28) • 8 Cross-sectional surveys • 3 Cohort studies • 17 Case–control studies

Figure 1. Flow diagram of assessment of studies identified in the systematic review and meta-analysis.

and without CRC who reported constipation were compared, again using an OR and a 95% CI. If there were no individuals with or without constipation who were found to have CRC in a single study, 0.5 was added to all four cells for the purposes of the analysis, as ORs cannot be calculated from zero values. Heterogeneity between studies was assessed using the I2 statistic with a cutoff of 50% (19), and the χ2-test with a P value < 0.10, used to define a statistically significant degree of heterogeneity. We planned to conduct sensitivity analyses, where sufficient studies existed, according to study quality, geographical region, criteria used to define constipation, and whether case–control studies tried to minimize recall bias by asking participants to only report symptoms that had been present for several years before the diagnosis of CRC, and to ignore any recent change in bowel habit. We also performed a subgroup analysis of the association between constipation and CRC according to gender, where individual studies reported these data. We compared the individual ORs between these subgroups using the Cochran Q statistic. Data were pooled using a random effects model (20), to give a more conservative estimate of the odds of CRC according to The American Journal of GASTROENTEROLOGY

The search strategy yielded 2,282 citations (Figure 1), of which 87 potentially relevant articles were retrieved and assessed in more detail. Of these, 28 met our eligibility criteria and were included. Agreement between reviewers was excellent (Kappa statistic = 0.77). There were eight cross-sectional surveys (23–30), three cohort studies (31–33), and 17 case–control studies eligible for inclusion. (14,16,34–48) Detailed characteristics of cross-sectional surveys, cohort studies, and case–control studies, including study quality, are provided in Tables 1, 2, and 3 respectively. Gender data were reported by six of the case–control studies (16,34– 36,38,40), but only one of the cohort studies (32), and none of the cross-sectional surveys. Prevalence of CRC in patients presenting with constipation in cross-sectional surveys

The eight cross-sectional surveys contained a total of 8,866 patients undergoing colonic investigation for lower GI symptoms (23–30). Symptoms were recorded before investigation in all studies. Five studies used a questionnaire (25–29), with symptoms recorded by the patient in three studies (27–29), and by a physician in the remaining two studies (25,26). In the other three studies constipation status was defined according to self-report (23,24,30). Three of the studies were higher quality according to the QUADAS-2 tool (Table 1) (26,27,30). In total, 1,497 (16.9%) patients underwent colonoscopy for constipation as the primary indication, and 585 (6.6%) patients were found to have CRC after investigation. There were 78 (5.2%) of 1,497 patients with constipation as the primary indication for colonoscopy who were found to have CRC, compared with 507 (6.9%) of 7,369 without constipation as the primary indication. The pooled OR for CRC in participants with constipation as the primary indication, compared with those without, was 0.56 (95% CI 0.36 to 0.89), with no significant heterogeneity detected between studies (I2 = 34.1%, P = 0.16) (Figure 2). There were too few studies to assess for publication bias. When only the three higher quality studies, containing a total of 3,931 patients, were included in the analysis, the OR for CRC in patients with constipation as the primary indication for colonoscopy was 0.57 (95% CI 0.42 to 0.79), with no significant heterogeneity detected between studies (I2 = 0%, P = 0.56). When data from the five lower quality studies were pooled, the OR in patients with constipation as the primary indication for colonoscopy, vs. those without, was no longer statistically significant (0.57; 95% CI 0.25 to 1.31), but with significant heterogeneity between studies (I2 = 58.0%, P = 0.05). This difference in ORs was not statistically significant (Cochran Q = 0, P = 0.99). VOLUME 108 | JUNE 2013 www.amjgastro.com


897

Study

Country

Setting and number of centers

Method of collection of constipation symptom data

Consecutive patients

Blinding

Number with constipation (% with CRC)

Number without constipation (% with CRC)

Tate and Royle (30)

UK

Tertiary (1)

Self-report

Not reported

No

16 (0)

114 (12.3)

Low

de Bossett et al. (26)

Switzerland

Secondary (5)

Physician-completed questionnaire

Yes

No

73 (1.4)

734 (5.2)

Low

Selvachandran et al. (29)

UK

Secondary (1)

Patient-completed questionnaire

Not reported

No

290 (1.4)

1,978 (4.6)

Unclear

Panzuto et al. (28)

Italy

Primary (159)


Yes

No

134 (15.7)

146 (13.7)

Unclear

Bersani et al. (25)

Italy

Secondary (1)

Physician-completed questionnaire

Yes

Yes

201 (1.5)

1,101 (5.7)

Unclear

Adler et al. (23)

Germany

Secondary (39)

Self-report

Yes

No

55 (0)

550 (1.8)

Unclear

Bafandeh et al. (24)

Iran

Tertiary (1)

Self-report

Yes

No

48 (4.2)

432 (3.2)

Unclear

Huang et al. (27)

China

Tertiary (1)


Not reported

No

680 (6.9)

2,314 (11.1)

Duration of follow-up (years)

Number with constipation (% with CRC)

Number without constipation (% with CRC)

Quality score

Risk of bias

Low

CRC, colorectal cancer.

Table 2. Characteristics of included cohort studies

Method used to ascertain CRC status


Stool frequency used to define constipation

Study

Country

Sample used

Dukas et al. (31)

USA

Nurses

Review of medical records

Questionnaire

< 1 Bowel movement per day

12

22,684 (0.65)

61,755 (0.75)

6

Kojima et al. (32)

Japan

General population

Cancer registry data

Questionnaire


6–9

14,115 (0.7)

48,814 (1.1)

6

Watanabe et al. (33)

Japan

General population


Questionnaire


7

9,269 (0.6)

32,401 (0.6)

5

CRC, colorectal cancer.

Only two studies reported the duration of constipation (24,28), and in one this was > 3 months (24), and in the other within the last 3 months (28). Excluding the latter study from the meta-analysis in a subgroup analysis, due to the relatively short-time frame of constipation, altered the pooled OR slightly (0.53; 95% CI 0.40 to 0.71), but this difference in ORs was not statistically significant (Cochran Q = 0.04, P = 0.84). Prevalence of CRC in individuals with constipation at baseline in cohort studies

There were a total of 189,038 participants in the three cohort studies (31–33), and constipation status was recorded at entry into the cohort in all three studies. Presence of CRC was confirmed using cancer registry data in two studies, (32,33) and medical records of participants in the third (31). Constipation was defined as less than one stool per day in all three studies. Two of the studies © 2013 by the American College of Gastroenterology

scored 6 on the Newcastle-Ottawa scale (31,32), and the third study scored 5 (Table 2) (33). There were 46,068 (24.4%) individuals with constipation at study entry, and 1,511 (0.8%) subjects developed CRC during follow-up. Among those with constipation at baseline, 302 (0.7%) developed CRC, compared with 1,209 (0.85%) of 142,970 without constipation. The pooled OR for CRC in participants with constipation, compared with those without, was lower at 0.80 but this difference was not statistically significant (95% CI 0.61 to 1.04), and there was significant heterogeneity detected between studies (I2 = 74.8%, P = 0.02). There were too few studies to assess for evidence of publication bias, or to perform detailed sensitivity analyses according to the study characteristics. However, as all three studies recorded stool frequency in some detail, we were able to conduct a sensitivity analysis using a frequency of one stool every 3 days to define the presence of The American Journal of GASTROENTEROLOGY

REVIEW

Table 1. Characteristics of included cross-sectional surveys

898

Power et al.

REVIEW

Table 3. Characteristics of included case-control studies Setting and number of centers

Source of controls

Method used to ascertain CRC status


Threshold used to define constipation

Onset of constipation

Number with CRC (% with constipation)

Number without CRC (% with constipation)

Quality score

Study

Country

Higginson (34)

USA

Secondary care (7)

Hospital patients

Medical records

Interview

Occurring more than weekly

> 2 Years before diagnosis

340 (20.0)

1,020 (16.7)

3

Wynder and Shigematsu (35)

USA

Secondary care (4)

Hospital patients

Medical records

Interview

Mild or more

1–10 Years before diagnosis

793 (34.7)

407 (25.3)

2

Wynder et al. (36)

Japan

Tertiary care (2)

Hospital patients

Medical records

Interview

Occurring 2–5 Years sometimes or before more often diagnosis

157 (24.2)

307 (16.6)

2

Haenszel et al. (37)

Hawaii

Secondary care (3)

Hospital patients

Medical records

Interview

Not reported

Not reported

179 (6.1)

357 (2.5)

2

Jain et al. (38)

Canada

General population and secondary care (19)

General population

Cancer registry data and medical records

Interview

Not reported

Not reported

542 (54.8)

542 (45.8)

4

Nakamura et al. (39)

USA

Tertiary care (1)

Hospital patients and spouses


Interview

Mild or more


100 (30.0)

151 (31.8)

3

Kune et al. (40)

Australia General population (N/A)a

General population

Medical records

Interview

Not reported

Before diagnosis

685 (31.4)

723 (26.4)

3

Kotake et al. (41)

Japan

Tertiary care (10)

Hospital patients

Medical records

Questionnaire

No bowel movement for 3 days or more

Not reported

363 (29.5)

363 (19.6)

2

Yang et al. (42)

China

Secondary care (Multiple)

General population

Medical records

Medical records

Not reported


1,328 (14.9)

1,451 (9.1)

1

Jacobs and White (16)

USA

General population (N/A)a

General population


Interview

Occurring more than never

Over last 10 years

424 (44.1)

414 (33.8)

5

Nascimbeni et al. (43)

Italy

Tertiary care (1)

Hospital patients

Medical records

Questionnaire

Rome I criteria

For at least 3 years

55 (30.9)

96 (18.8)

3

Roberts et al. (44)

USA


General population


Interview

< 3 stools per week

Not reported

643 (7.3)

1,048 (2.9)

3

Hamilton et al. (45)

UK

General population (21)

General population


Medical records

Not reported

Not reported

349 (26.1)

1,744 (14.8)

6

Hamilton et al. (46)

UK

Primary care (317)

Primary care patients

Medical records

Medical records

Not reported

Not reported

5,477 (27.0)

38,314 (10.6)

3

Promthet et al. (47)

Thailand Secondary care (2)

Hospital patients

Medical records

Interview

Occurring occasionally or more

1 Year before diagnosis

130 (59.2)

130 (28.5)

3

Simons et al. (14)

Holland


General population


Questionnaire

Occurring Not reported sometimes or more often

1,207 (9.9)

1,753 (12.1)

3

Tashiro et al. (48)

Japan

Secondary care (8)

General population

Medical records

Interview

Not reported

212 (21.7)

790 (16.5)

5

1 Year before diagnosis

CRC, colorectal cancer. a Not applicable: population-based case–control study.




899

Tate and Royle, 1988

0.210 (0.000, 1.841)

de bossett et al., 2002

0.254 (0.006, 1.556)

Selvachandran et al., 2002

0.290 (0.077, 0.778)

Panzuto et al., 2003

1.171 (0.571, 2.406)

Bersani et al., 2005

0.250 (0.050, 0.777)

Adler et al., 2007

0.464 (0.000, 3.959)

Bafandeh et al., 2008

1.298 (0.139, 5.930)

Huang et al., 2010

0.594 (0.421, 0.825)

Combined (random)

0.563 (0.358, 0.885)

0.001

0.01

0.1

0.2

0.5

1

2

5

10

Odds ratio (95% Confidence interval)

Figure 2. Odds ratio for colorectal cancer in patients with constipation vs. patients without in cross-sectional surveys.

constipation. Using this threshold, 10,035 (5.3%) subjects had constipation at study entry, of whom 69 (0.7%) developed CRC, compared with 1,442 (0.8%) of 179,003 without constipation. The pooled OR increased slightly (0.90; 95% CI 0.70 to 1.15), but the heterogeneity previously observed disappeared altogether (I2 = 0%, P = 0.59). This difference in ORs was not statistically significant (Cochran Q = 0.40, P = 0.53). Prevalence of constipation in patients with CRC in case–control studies

The 17 case–control studies included a total of 62,594 subjects (14,16,34–48) 12,984 (20.7%) of whom were cases with CRC and 49,610 controls without. Presence of CRC was confirmed using the medical records of participants in 11 studies (34–37,40–43,46–48), cancer registry data in five studies (14,16,39,44,45) and a combination of both in the remaining study (38). Only four studies scored 4 or more on the Newcastle-Ottawa scale (Table 3) (16,38,45,48). There were a total of 9,199 (14.7%) participants who were classified as having constipation. In total, 3,300 (25.4%) of 12,984 cases with CRC reported constipation, compared with 5,899 (11.9%) of 49,610 controls without CRC. The pooled OR for constipation in cases with CRC was 1.68 (95% CI 1.29 to 2.18), but with significant heterogeneity between studies (I2 = 93.8%, P < 0.001) (Figure 3). There was evidence of funnel plot asymmetry (Egger test, P = 0.005), or other small study effects, with a lack of smaller studies showing no difference in odds of constipation in cases with CRC compared with controls without. When data from the six studies that reported the association between CRC and constipation according to gender were pooled, the OR for constipation in males with CRC was 1.58 (95% CI 1.34 to 1.87), compared with 1.25 (95% CI 1.04 to 1.52) in females with CRC. This difference in ORs approached statistical significance (Cochran Q = 3.31, P = 0.07). When only the four studies that scored four or more on the Newcastle-Ottawa scale were included in the analysis (16,38,45,48), © 2013 by the American College of Gastroenterology

the significant association between constipation and CRC persisted (OR 1.60; 95% CI 1.35 to 1.89), with no heterogeneity between studies (I2 = 29.1%, P = 0.24). In the 13 lower quality studies, the pooled OR was similar (1.71; 95% CI 1.23 to 2.39), but heterogeneity between studies remained (I2 = 94.7%, P < 0.001). This difference in ORs was not statistically significant (Cochran Q = 0.12, P = 0.73). There were 11 studies conducted in Western populations (14,16,34,35,38–40,43–46), and six studies that recruited cases and controls from Eastern populations (36,37,41,42,47,48). When only Western studies were included in the analysis the OR was 1.56 (95% CI 1.09 to 2.22), and the heterogeneity observed persisted (I2 = 95.8%, P < 0.001). When data from Eastern studies were pooled the OR was 1.80 (95% CI 1.55 to 2.10), but the observed heterogeneity was only borderline statistically significant (I2 = 48.6%, P = 0.08). Again, this difference in ORs was not statistically significant (Cochran Q = 0.53, P = 0.47). Nine studies stated specifically that they asked participants to only report symptoms that had been present for several years before the diagnosis of CRC, and to ignore any recent change in bowel habit (16,34–36,39,42,43,47,48). In these studies the pooled OR for constipation in CRC was 1.59 (95% CI 1.32 to 1.90), again with borderline significant heterogeneity between studies (I2 = 54.0%, P = 0.03). In the eight studies that did not report this information (14,37,38,40,41,44–46), the pooled OR was higher (1.76; 95% CI 1.14 to 2.72), with significant heterogeneity between studies (I2 = 96.4%, P < 0.001). Once again, this difference in ORs was not statistically significant (Cochran Q = 0.18, P = 0.67).

DISCUSSION This systematic review and meta-analysis has assembled data from all identified published observational studies examining the association between constipation and CRC in adults. In eight The American Journal of GASTROENTEROLOGY

REVIEW

Odds ratio meta-analysis plot (random effects)

900

Power et al.

REVIEW

Odds ratio meta-analysis plot (random effects) Higginson, 1966

1.25 (0.90, 1.72)

Wynder and Shigemastu, 1967a

1.57 (1.19, 2.07)

Wynder et al., 1967b

1.60 (0.97, 2.64)

Haenszel et al., 1973

2.53 (0.93, 7.04)

Jain et al., 1980

1.44 (1.12, 1.84)

Nakamura et al., 1984

0.92 (0.51, 1.64)

Kune et al., 1988

1.27 (1.00, 1.62)

Kotake et al., 1995

1.72 (1.20, 2.46)

Yang et al., 1995

1.75 (1.38, 2.23)

Jacobs and White, 1998

1.54 (1.16, 2.06)

Nascimbeni et al., 2002

1.94 (0.83, 4.48)

Roberts et al., 2003

2.68 (1.64, 4.43)

Hamilton et al., 2005

2.03 (1.53, 2.69)

Hamilton et al., 2009

3.12 (2.92, 3.34)

Promthet et al., 2010

3.65 (2.11, 6.34)

Simons et al., 2010

0.80 (0.62, 1.01)

Tashiro et al., 2011

1.41 (0.94, 2.08)

Combined (random)

1.68 (1.29, 2.18)

0.5

1

2

5

10

Odds ratio (95% Confidence interval)

Figure 3. Odds ratio for constipation in cases with colorectal cancer vs. controls without in case–control studies.

cross-sectional surveys, the presence of constipation as the primary indication for colonoscopy was associated with a significantly lower prevalence of CRC, with a 44% reduction in odds of CRC among patients with constipation according to either self-report or a questionnaire, and no significant heterogeneity detected between studies. The data from the three cohort studies demonstrated a non-significant trend towards a lower OR for development of CRC in individuals reporting constipation at baseline, defined as less than one stool per day, with significant heterogeneity detected between studies. This heterogeneity disappeared altogether when the definition of constipation was changed to a frequency of one stool every 3 days. In contrast, when data were pooled from the 17 case–control studies, the prevalence of constipation in patients with CRC was significantly higher than in controls without CRC, but again with significant heterogeneity detected between studies, and possible publication bias. When sensitivity analyses were conducted, the OR for constipation in CRC remained significantly higher in all analyses, and heterogeneity was reduced when only high quality studies or those conducted in the East were considered. There are several strengths of this study. We carried out a comprehensive and contemporaneous literature search that identified sufficient studies to allow the pooling of data from over 250,000 recruited subjects. This was augmented by hand-searching of the bibliographies of retrieved articles, and one foreign language paper was translated. The judging of study eligibility and data extraction were carried out independently by two investigators, with The American Journal of GASTROENTEROLOGY

discrepancies resolved by consensus. The quality of each study was assessed according to published criteria (17,18). Data were pooled using a random effects model, in order to give a more conservative estimate of the odds of CRC according to constipation status in cross-sectional surveys and cohort studies, and the odds of constipation according to the presence or absence of CRC in case– control studies. There was no significant heterogeneity observed when the results of cross-sectional surveys were combined, but heterogeneity was observed among cohort studies and case-control studies. However, the use of sensitivity analyses allowed us to explore potential reasons for the observed heterogeneity, and in some cases this disappeared. Where possible, we also assessed for evidence of publication bias, or other small study effects, by testing funnel plots for obvious asymmetry. As with any systematic review and meta-analysis, limitations include the quality of the eligible studies, as well as the reporting of data within them. The majority of the cross-sectional surveys identified, while large in several cases, were of lower quality when assessed using the QUADAS-2 tool. However, when only low-risk of bias studies were considered in the analysis the summary estimate was identical to that for unclear risk of bias studies. While the cohort studies were of higher quality, with all scoring 5 or more of a possible 9 on the Newcastle-Ottawa scale, the case–control studies we identified scored poorly, with only four of the seventeen case–control studies scoring over 4. It should be remembered that there is no recommended threshold in use to define higher quality studies, but the consistently low scores among case–control studies VOLUME 108 | JUNE 2013 www.amjgastro.com

is of concern. There was also considerable heterogeneity between case–control studies in many of the analyses conducted. The reasons for this remain speculative, but may relate to variations in the definition of constipation used, methods of acquiring data, demographic characteristics of recruited individuals, which were not reported in the majority of studies, or cultural differences. The distinction between chronic and acute/new-onset constipation is very important, as this could be the manifestation of mechanical bowel obstruction resulting from CRC. We sought out this information for each study and described it where reported. Nine of the case–control studies specifically stated that they asked participants to ignore any recent change in bowel habit (16,34–36,39,42,43,47,48), allowing us to pool data in a subgroup analysis according to this study characteristic, but none of the cohort studies reported this information. This is probably less relevant for the cohort studies, as they were prospective, and followed up individuals who reported constipation at baseline for a long duration (anywhere from 6 to > 10 years), in order to detect new cases of CRC. For the cross-sectional surveys only two studies reported the duration of constipation (24,28), and in one this was > 3 months (24), and in the other constipation had occurred within the last 3 months (28), precluding any meaningful pooling of data according to symptom duration. A final limitation is that we calculated ORs using raw data reported by the studies, rather than pooling adjusted ORs as reported by the individual studies, which would have been adjusted for some potential underlying differences between study participants. However, as the majority of studies were not conducted with the primary aim of studying the relationship between constipation and CRC, very few reported adjusted summary statistics, and those that did failed to use ORs in all cases, meaning that pooling of these data would not have been possible. As a result of this, there may be residual confounding issues that explain our findings. For instance it may be that patients with constipation are more likely to undergo colonoscopy, and this is the reason for the negative association between constipation and CRC seen in cross-sectional surveys. However, given that the comparator in these studies was patients with other lower GI symptoms, rather than asymptomatic healthy individuals we feel this explanation is unlikely. In fact, the largest of the cross-sectional surveys that showed a significant negative association between constipation and CRC, and which accounted for 60% of the weight of the meta-analysis, specifically stated that patients were undergoing colonoscopy for the first time (27). There have been several previous systematic reviews examining the association between lower GI symptoms and CRC (13,49–53), although none of these have reported exclusively on constipation alone as a possible risk factor for CRC, and one did not study this issue at all (49). The earliest of these, published by Sonnenberg and Muller (13) in 1993, identified nine case–control studies reporting on the prevalence of constipation in patients with CRC, and demonstrated a significant association between the two, with an OR of similar magnitude to the one we observed. In 2003, Chen et al. (53) also examined this issue , but restricted their analysis to four case–control studies conducted in China. The pooled OR was 2.23 in this study, higher than that observed in either the present study or that by Sonnenberg et al. (13). © 2013 by the American College of Gastroenterology

More recently, Jellema et al. (52) summarized data from 47 studies that reported the accuracy of lower GI symptoms, blood tests, and fecal occult blood testing in predicting CRC in primary or secondary care . There were only four studies identified that reported on the diagnostic accuracy of constipation, and sensitivity ranged from 0% to 51%, and specificity from 53% to 90%. The authors concluded that the diagnostic performance of constipation in predicting CRC was poor. In a meta-analysis of observational studies that examined the diagnostic accuracy of lower GI symptoms only (50), constipation was not associated with the presence of CRC, with a positive likelihood ratio of 1.1, although the number of studies providing data for this analysis was not reported. Finally, Astin et al. (51) performed a systematic review of studies conducted in primary care, identifying only three that reported on the association between constipation and CRC , with positive and negative likelihood ratios of 1.74 and 0.84 respectively. In our systematic review and meta-analysis we were deliberately inclusive, and analyzed data from cross-sectional surveys, cohort studies, or case–control studies conducted in the general population and primary, secondary, or tertiary care. When data from the cohort studies were pooled, the presence of constipation at baseline was not associated with the development of CRC after 6 to 12 years of follow-up. The finding that constipation was commoner in cases with CRC, compared with controls without, is in keeping with the two aforementioned meta-analyses of case–control studies (13,53). It has been postulated that this significant association is due to reverse causation, because CRC leads to the development of constipation, precipitating presentation to a physician (10). However, data from the cross-sectional surveys we identified indicate that this is not the case, with the presence of constipation as the primary indication for colonoscopy being negatively associated with an ultimate diagnosis of CRC following complete lower GI investigation. We would, therefore, propose that the significant association observed in case–control studies relates to a combination of poor study quality, publication bias, and recall bias among participants. Only 9 of the 17 provided information about the duration of symptoms relative to when the diagnosis of CRC was made, and the OR was slightly lower when data from these studies were pooled. Our results, therefore, suggest that constipation alone does not warrant lower GI investigation unless other alarm symptoms or signs are present, although these also perform poorly in predicting a diagnosis of CRC (49). Despite the fact that constipation is not recognized as an alarm feature, according to current guidelines for the detection of suspected cancer (54), physicians continue to refer constipated patients for lower GI examination to rule out CRC as an underlying cause. Hopefully the data from this systematic review and meta-analysis will lead to an alteration in such behavior. In conclusion, data from this systematic review and meta-analysis demonstrate that in cohort studies constipation is not associated with the development of CRC. In addition, when patients are colonoscoped with constipation as the primary indication, a diagnosis of CRC is less likely than in patients being colonoscoped for other lower GI symptoms as the primary indication. The association between constipation and CRC observed in The American Journal of GASTROENTEROLOGY

901

REVIEW


REVIEW

902

Power et al.

case–control studies is likely to be due to a combination of poor study quality and recall bias among participants. The use of invasive lower GI investigations to exclude CRC in patients presenting with constipation, in the absence of other alarm features, should be discouraged. CONFLICT OF INTEREST

Guarantor of the article: Alexander C. Ford, MBChB, MD, FRCP. Specific author contributions: N.J.T. and A.C.F.: conceived and drafted the study. A.M.P. and A.C.F.: collected all data. A.C.F.: analyzed and interpreted the data. A.C.F. and A.M.P.: drafted the manuscript. All authors commented on drafts of the paper. All authors have approved the final draft of the manuscript. Financial support: Dr Ford’s time was reimbursed by an investigator-initiated grant from Almirall. Almirall had no input into the study conduct or design. Potential competing interests: None. ACKNOWLEDGMENTS

We thank Dr Cathy Yuhong Yuan for assisting us with the translation of foreign language articles. REFERENCES 1. Drossman DA, Li Z, Andruzzi E et al. U.S. householder survey of functional gastrointestinal disorders. Prevalence, sociodemography, and health impact. Dig Dis Sci 1993;38:1569–80. 2. Ford AC, Forman D, Bailey AG et al. Irritable bowel syndrome: a 10-year natural history of symptoms, and factors that influence consultation behavior. Am J Gastroenterol 2008;103:1229–39. 3. Lovell RM, Ford AC. Global prevalence of, and risk factors for, irritable bowel syndrome: a meta-analysis. Clin Gastroenterol Hepatol 2012;10: 712–21. 4. Suares NC, Ford AC. Prevalence of, and risk factors for, chronic idiopathic constipation in the community: Systematic review and meta-analysis. Am J Gastroenterol 2011;106:1582–91. 5. Talley NJ, Zinsmeister AR, Van Dyke C et al. Epidemiology of colonic symptoms and the irritable bowel syndrome. Gastroenterology 1991;101: 927–34. 6. Drossman DA, Whitehead WE, Toner BB et al. What determines severity amongst patients with painful functional bowel disorders? Am J Gastroenterol 2000;95:974–80. 7. Wald A, Scarpignato C, Kamm MA et al. The burden of chronic constipation on quality of life: results of a multinational survey. Aliment Pharmacol Ther 2007;26:227–36. 8. Chang JY, Locke III GR, McNally MA et al. Impact of functional gastrointestinal disorders on survival in the community. Am J Gastroenterol 2010;105:822–32. 9. Ford AC, Forman D, Bailey AG et al. Effect of dyspepsia on survival: a longitudinal 10-year follow-up study. Am J Gastroenterol 2012;107:912–21. 10. Talley NJ, Lasch KL, Baum CL. A gap in our understanding: chronic constipation and its comorbid conditions. Clin Gastroenterol Hepatol 2009;7:9–19. 11. Burkitt DP. Epidemiology of cancer of the colon and rectum. Cancer 1971;28:3–13. 12. Narisawa T, Magadia NE, Weisburger JH et al. Promoting effect of bile acids on colon carcinogenesis after intrarectal instillation of N-methyl-N’nitro-N-nitrosoguanidine in rats. J Natl Cancer Inst 1974;53:1093–7. 13. Sonnenberg A, Muller AD. Constipation and cathartics as risk factors of colorectal cancer: a meta-analysis. Pharmacology 1993;47 (suppl 1): 224–33. 14. Simons CC, Schouten LJ, Weijenberg MP et al. Bowel movement and constipation frequencies and the risk of colorectal cancer among men in the Netherlands Cohort Study on diet and cancer. Am J Epidemiol 2010;172:1404–14. 15. Otani T, Iwasaki M, Inoue M, et al. For the Japan Public Health Center-based Prospective Study Group. Bowel movement, state of stool,


and subsequent risk for colorectal cancer: The Japan public health center-based prospective study. Ann Epidemiol 2006;16:888–94. 16. Jacobs EJ, White E. Constipation, laxative use, and colon cancer among middle-aged adults. Epidemiology 1998;9:385–91. 17. Whiting PF, Rutjes AW, Westwood ME et al. QUADAS-2: a revised tool for the quality assessment of diagnostic accuracy studies. Ann Intern Med 2011;155:529–36. 18. Wells GA, Shea B, O’Connell D et al. The Newcastle-Ottawa Scale (NOS) for assessing the quality of nonrandomised studies in meta-analyses. http:// www ohrica/programs/clinical_epidemiology/oxford htmAccessed 24th May 2012–2009. 19. Higgins JPT, Thompson SG, Deeks JJ et al. Measuring inconsistency in meta-analyses. Br Med J 2003;327:557–60. 20. DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials 1986;7:177–88. 21. Egger M, Davey-Smith G, Schneider M et al. Bias in meta-analysis detected by a simple, graphical test. Br Med J 1997;315:629–34. 22. Sterne JA, Sutton AJ, Ioannidis JP et al. Recommendations for examining and interpreting funnel plot asymmetry in meta-analyses of randomised controlled trials. Br Med J 2011;343:d4002. 23. Adler A, Roll S, Marowski B et al. Appropriateness of colonoscopy in the era of colorectal cancer screening: a prospective, multicenter study in a private-practice setting (Berlin Colonoscopy Project 1, BECOP 1). Dis Colon Rectum 2007;50:1628–38. 24. Bafandeh Y, Khoshbaten M, Eftekhar Sadat AT et al. Clinical predictors of colorectal polyps and carcinoma in a low prevalence region: results of a colonoscopy based study. World J Gastroenterol 2008;14:1534–8. 25. Bersani G, Rossi A, Ricci G et al. Do ASGE guidelines for the appropriate use of colonoscopy enhance the probability of finding relevant pathologies in an open access service? Dig Liver Dis 2005;37:609–14. 26. de Bosset V, Froehlich F, Rey J-P et al. Do explicit appropriateness criteria enhance the diagnostic yield of colonoscopy? Endoscopy 2002;34:360–8. 27. Huang L, Wang X, Gong W et al. The comparison of the clinical manifestations and risk factors of colorectal cancer and adenomas: results from a colonoscopy-based study in southern Chinese. Int J Colorectal Dis 2010;25:1343–51. 28. Panzuto F, Chiriatti A, Bevilacqua S et al. Symptom-based approach to colorectal cancer: Survey of primary care physicians in Italy. Dig Liver Dis 2003;35:869–75. 29. Selvachandran SN, Hodder RJ, Ballal MS et al. Prediction of colorectal cancer by a patient consultation questionnaire and scoring system: a prospective study. Lancet 2002;360:278–83. 30. Tate JJT, Royle GT. Open access colonoscopy for suspected colonic neoplasia. Gut 1988;29:1322–5. 31. Dukas L, Willet WC, Colditz GA et al. Prospective study of bowel movement, laxative use, and risk of colorectal cancer among women. Am J Epidemiol 2000;151:958–64. 32. Kojima M, Wakai K, Tokudome S et al. Bowel movement frequency and risk of colorectal cancer in a large cohort study of Japanese men and women. Br J Cancer 2004;90:1397–401. 33. Watanabe T, Nayaka N, Kurashima K et al. Constipation, laxative use and risk of colorectal cancer: The Miyagi Cohort Study. Eur J Cancer 2004;40:2109–15. 34. Higginson J. Etiological factors in gastrointestinal cancer in man. J Natl Cancer Inst 1966;37:527–45. 35. Wynder EL, Shigematsu T. Environmental factors of cancer of the colon and rectum. Cancer 1967;20:1520–61. 36. Wynder EL, Kajitani T, Ishikawa S et al. Environmental factors of cancer of the colon and rectum. II. Japanese epidemiological data. Cancer 1969;23:1210–20. 37. Haenszel W, Berg JW, Segi M et al. Large-bowel cancer in Hawaiian Japanese. J Natl Cancer Inst 1973;51:1765–79. 38. Jain M, Cook GM, Davis FG et al. A case-control study of diet and colo-rectal cancer. Int J Cancer 1980;26:757–68. 39. Nakamura GJ, Schneiderman LJ, Klauber MR. Colorectal cancer and bowel habits. Cancer 1984;54:1475–7. 40. Kune GA, Kune S, Field B et al. The role of chronic constipation, diarrhea, and laxative use in the etiology of large-bowel cancer. Data from the Melbourne Colorectal Cancer Study. Dis Colon Rectum 1988;31:507–12. 41. Kotake K, Koyama Y, Nasu J et al. Relation of family history of cancer and environmental factors to the risk of colorectal cancer: a case-control study. Jpn J Clin Oncol 1995;25:195–202.


42. Yang G, Zhang X-L, Zheng S. The role of common symptoms in screening of large bowel cancer. A population-based case-control study in Shanghai. Chin J Clin Oncol 1995;22:403–7. 43. Nascimbeni R, Donato F, Ghirardi M et al. Constipation, anthranoid laxatives, melanosis coli, and colon cancer: a risk assessment using aberrant crypt foci. Cancer Epidemiol Biomarkers Prev 2002;11:753–7. 44. Roberts MC, Millikan RC, Galanko JA et al. Constipation, laxative use, and colon cancer in a North Carolina population. Am J Gastroenterol 2003;98:857–64. 45. Hamilton W, Round A, Sharp D et al. Clinical features of colorectal cancer before diagnosis: a population-based case-control study. Br J Cancer 2005;93:399–405. 46. Hamilton W, Lancashire R, Sharp D et al. The risk of colorectal cancer with symptoms at different ages and between the sexes: a case-control study. BMC Med 2009;7:17. 47. Promthet SS, Pientong C, Ekalaksananan T et al. Risk factors for colon cancer in Northeastern Thailand: Interaction of MTHFR codon 677 and 1298 genotypes with environmental factors. J Epidemiol 2010;20:329–38.

© 2013 by the American College of Gastroenterology

48. Tashiro N, Budhathoki S, Ohnaka K et al. Constipation and colorectal cancer risk: the Fukuoka Colorectal Cancer Study. Asian Pac J Cancer Prev 2011;12:2025–30. 49. Ford AC, Veldhuyzen Van Zanten SJO, Rodgers CC et al. Diagnostic utility of alarm features for colorectal cancer: systematic review and meta-analysis. Gut 2008;57:1545–53. 50. Adelstein B-A, Macaskill P, Chan SF et al. Most bowel symptoms do not indicate colorectal cancer and polyps: a systematic review. BMC Gastroenterol 2011;11:65. 51. Astin M, Griffin T, Neal RD et al. The diagnostic value of symptoms for colorectal cancer in primary care: a systematic review. Br J Gen Pract 2011;61:e231–43. 52. Jellema P, van der Windt DAWM, Bruinvels DJ et al. Value of symptoms and additional diagnostic tests for colorectal cancer in primary care: systematic review and meta-analysis. Br Med J 2010;340:c1269. 53. Chen K, Qiu J-L, Zhang Y et al. Meta-analysis of risk factors for colorectal cancer. World J Gastroenterol 2003;9:1598–600. 54. Referral guidelines for suspected cancer in adults and children. www nice org uk/nicemedia/pdf/CG027fullguideline pdf 2007.


903

REVIEW
