Digestive Diseases and Sciences

, 53:3049

Screening Colonoscopy for Colorectal Cancer in Asymptomatic People: A Meta-Analysis

Authors

    • Department of GastroenterologyRabin Medical Center, Beilinson Hospital
    • Sackler Faculty of MedicineTel Aviv University
  • Rachel Hazazi
    • Department of GastroenterologyRabin Medical Center, Beilinson Hospital
    • Sackler Faculty of MedicineTel Aviv University
  • Zohar Levi
    • Department of GastroenterologyRabin Medical Center, Beilinson Hospital
    • Sackler Faculty of MedicineTel Aviv University
  • Gerald Fraser
    • Department of GastroenterologyRabin Medical Center, Beilinson Hospital
    • Sackler Faculty of MedicineTel Aviv University
Review

DOI: 10.1007/s10620-008-0286-y

Cite this article as:
Niv, Y., Hazazi, R., Levi, Z. et al. Dig Dis Sci (2008) 53: 3049. doi:10.1007/s10620-008-0286-y

Abstract

Objective The preferred method for screening asymptomatic people for colorectal cancer (CRC) is colonoscopy, according to the new American guidelines. The aim of our study was to perform a meta-analysis of the prospective cohorts using total colonoscopy for screening this population for CRC. We looked for the diagnostic yield of the procedure as well as for its safety in a screening setting. Methods We included papers with more than 500 participants and only those reporting diagnostic yield of adenoma (and/or advanced adenoma) and CRC. Nested analysis were performed for secondary endpoints of complications and CRC stages when this information was available. All analyses were performed with StatDirect Statistical software, version 2.6.1 (http://www.statsdirect.com). Results Our search yielded ten studies of screening colonoscopy conducted in asymptomatic people that met our inclusion criteria, with a total of 68,324 participants. Colonoscopy was complete and reached the cecum in 97% of the procedures. Colorectal cancer was found in 0.78% of the participants (95% confidence interval 0.13–2.97%). Stage I or II were found in 77% of the patients with CRC. Advanced adenoma was found in 5% of the cases (95% confidence interval 4–6%). Complications were rare and described in five cohorts. Perforation developed in 0.01% of the cases (95% confidence interval 0.006–0.02%) and bleeding in 0.05% (95% confidence interval 0.02–0.09%). Conclusions Our findings support the notion that colonoscopy is feasible and a suitable method for screening for CRC in asymptomatic people.

Keywords

ColonoscopyScreeningColorectal cancerAdenomaFecal occult blood test

Introduction

The preferred method of screening for colorectal cancer (CRC) in average and high-risk populations is colonoscopy, according to the recently published new guidelines, a joint effort of the American Cancer Society Colorectal Cancer Advisory Group, the US Multi-Society Task Force, and the American College of Radiology Colon Cancer Committee [1, 2]. These recommendations rely on evidence-based research demonstrating the efficacy, safety, and cost effectiveness of colonoscopy [313]. This is the first time colonoscopy is recommended for early detection and prevention of CRC and polyps, while fecal occult blood is recommended only for early detection of CRC (not for prevention and not for early detection of polyps) [1].

Research into screening colonoscopy is inferior to that into fecal occult blood testing because of lack of randomized, controlled studies, and longitudinal follow-up demonstrating effect on mortality. However, colonoscopy has all that is required for a screening project: ability to visualize the entire colon, biopsy lesions, and remove all polyps; being effective in a lower compliance rate in comparison with other modalities; having a clear advantage over sigmoidoscopy since proximal distribution of colonic neoplasm has recently been described in the latter, especially in women [9]; and the fact that the fecal occult blood test (FOBT) can prevent fewer cancer deaths than colonoscopy [14], and that a positive FOBT leads in any case to total colonoscopy. Colonoscopy, starting at age 50 years, every 10 years, was found to be as cost-effective as all the other possibilities suggested by the recent American guidelines for early detection and prevention of CRC [1, 13]. There is also a firm recommendation to shift colonoscopic potential from overused surveillance, after polypectomy or cancer surgery, to screening as part of the solution for future shortage of facilities.

Lack of prospective, randomized, controlled study of colonoscopy for prevention of CRC, demonstrating reduction in mortality, is a good argument for governments and health organizations that are reluctant to reimburse colonoscopy for this particular indication. This policy ignores the natural history of CRC, and the adenoma–carcinoma sequence. Still, in countries such as Austria, Germany, Poland, Italy, Luxemburg, and the USA, screening colonoscopy is advocated and partially reimbursed [15].

Winawer and colleagues demonstrated a decrease in incidence of 70–90% in a cohort of 1400 patients after polypectomy in comparison with three large historical controls [16]. Gupta and colleagues demonstrated a decrease of annual mortality from 25.2/100,000 to 21.4/100,000 in Olmsted County following increased polypectomy rate from 86/100,000 to 320/100,000 [17].

The objective of this study was to perform a meta-analysis of the prospective cohorts using total colonoscopy for screening CRC. We looked for the diagnostic yield of the procedure for adenoma, advanced adenoma, and colorectal cancer, and the complication rate of screening colonoscopy in this setting, particularly bleeding, perforation, and death.

Methods

Formulation of Questions

The primary question was: What is the diagnostic yield for adenoma, advanced adenoma, and colorectal cancer in screening colonoscopy cohorts, performed in asymptomatic people. The secondary question was: What is the complication rate of screening colonoscopy in this setting, particularly bleeding, perforation, and death. In addition, as stage shifting towards favorable stages is expected in a screening program, the stages of CRC diagnosed were computed when available.

Asymptomatic people were defined as men and women aged 40–75 years, without symptoms of abdominal pain, rectal bleeding or a change in bowel habits. Advanced adenoma was defined as adenoma of size 10 mm or greater, or containing villous histology or with high-grade dysplasia.

Data Source and Search Strategy

We conducted a bibliographic search of the PubMed (Medline), EMBASE, Cochrane Library (Cochrane Database of Systemic Reviews and the Cochrane Controlled Trial Register), CINAHL, and AMED databases up to October 31, 2007 using the following keywords: “screening colonoscopy” or “screening colorectal cancer”. The search was directed at papers appearing in English-language medical journals. All papers identified by the electronic database search were examined and additional references were identified from the references listed in each paper.

Study Selection

Our analysis was based on prospective cohort studies in adults using total colonoscopy for screening, looking for polyps and CRC. We included only papers with more than 500 participants, and only when diagnostic yield of adenoma (or advanced adenoma) and CRC were reported. Nested analyses were performed for the secondary endpoints of complications and CRC stages for papers with this information available. Where available the risk for CRC (average versus high) was computed (Table 1).
Table 1

List of ten cohort studies of screening colonoscopy for CRC in the average-risk population, fulfilling the inclusion/exclusion criteria (number of patients = 68,324)

Author

Country

Year

n/HR

Men

Mean age (y)

Complete colonoscopy

Adenoma

Adv. adenoma

CRC

Perforation

Bleeding

Rogge [3]

USA

1994

627/NM

NA

NA

NA

0.225

NA

0.0079

0

0.0015

Lieberman [4]

USA

2000

3196/444

0.968

62.9

0.977

0.357

0.095

0.01

0

0.0018

Imperiale [5]

USA

2000

1994/NM

0.589

59.8

0.97

NA

0.0496

0.006

0.0005

0.00015

Imperiale [6]

USA

2002

906/NM

0.61

44.8

0.99

0.087

0.035

0

NA

NA

Ibanez [7]

Spain

2004

2210/0

0.746

57.9

NA

0.149

0.053

0.0049

NA

NA

Soon [8]

Taiwan

2005

1456/NM

0.62

53.5

NA

0.147

0.04

0.003

0

0

Soon [8]

USA

2005

3403/230

0.489

58.7

NA

0.207

0.049

0.003

0.00029

0.00029

Schoenfeld [9]

USA

2005

1483/230

0

58.7

0.987

0.155

0.048

0.001

0

0

Chiu [10]

Taiwan

2005

1741/0

0.597

52.5

0.981

0.151

0.03

0.004

NA

NA

Regula [11]

Poland

2006

50148/10435

0.359

55.2

0.911

0.238

0.056

0.08

0.0000997

0.00025

Strul [12]

Israel

2006

1177/0

0.471

NA

NA

0.209

0.063

0.011

0

0

HR, number of high-risk screenees (positive family history); NM, not mentioned; NA, not available

Exclusion Criteria

We excluded studies in which virtual colonoscopy, sigmoidoscopy, or fecal occult blood tests were used for screening CRC, or when the study population was heterogeneous, using another method in addition to colonoscopy, or when colonoscopy was performed as a second-line examination for symptomatic patients as a surveillance procedure (after polypectomy or surgery for CRC or in patients with inflammatory bowel diseases) or important data, such as diagnostic yield, were missing.

Data Synthesis and Analysis

For the meta-analysis, the following data were extracted into an Excel file: author, journal, year of publication, country of study, number of participants, sex, mean age, completeness of colonoscopy, diagnostic yield of adenoma, advanced adenoma, CRC, CRC stage, and complications (perforation, bleeding, and death).

All analyses were performed with StatDirect Statistical software, version 2.6.1 (http://www.statsdirect.com).

Results

Our search yielded 10 studies (out of 31) of screening colonoscopy that were conducted in asymptomatic people and that met our inclusion criteria [312]. The study of Soon et al. [8] described two different populations (from Taiwan and the USA) and was dealt with as two different papers. Table 1 summarizes the studies, and the available data in each study. The number of screenees included ranged from 627 up to 50,148, with a total of 68,324 participants, in a 12-year period. Six cohorts were North American, two were from Tailand, one was from Israel, and the largest was from Poland. Sex distribution was in favor of men but was somewhat similar between men and women, except in the Veteran Administration study conducted by Lieberman [4] where 96.8% of the participants were men and the study by Schoenfeld [9] where 100% of the participants were women. These studies contributed important information about adenoma distribution along the colon, where advanced adenomas were more prevalent in right colon of women. Except for the second study by Imperiale [6] where screening started at the age of 40 years, all the studies examined populations of 50 years or older. Adenoma detection rates could not be compared between average and high-risk individual, since there was no stratification for this result between these groups.

In three studies [7, 10, 12] it was mentioned that potential screenees with a family history of CRC were excluded. In three studies [4, 9, 11] 6.7–20.8% of the screenees had such a family history. In four studies [3, 5, 6, 8] this factor was not mentioned at all.

The results of the meta-analysis of screening colonoscopy diagnostic yield is shown in Fig. 1, and complication rates in Fig. 2. Colonoscopy was complete and reached the cecum in 97% of the procedures (95% confidence interval 94–99%). Colorectal cancer was found in 0.78% of the participants (95% confidence interval 0.13–2.97%). A favorable stage shifting was demonstrated: stage I 41% (95% confidence interval 39–42%), stage II 36% (95% confidence interval 20–54%), stage III 18% (95% confidence interval 8–31%), and stage IV 8.6% (95% confidence interval 7.8–9.5%). Thus, stages I and II together were found in 77% of the cancers diagnosed. At least one adenoma was found in 19% of patients (95% confidence interval 15–23%) and advanced adenoma in 5% of cases (95% confidence interval 4–6%).
https://static-content.springer.com/image/art%3A10.1007%2Fs10620-008-0286-y/MediaObjects/10620_2008_286_Fig1_HTML.gif
Fig. 1

Screening colonoscopy for CRC in the asymptomatic population, diagnostic yields of adenoma, advanced adenoma, and CRC: a meta-analysis (random effect)

https://static-content.springer.com/image/art%3A10.1007%2Fs10620-008-0286-y/MediaObjects/10620_2008_286_Fig2_HTML.gif
Fig. 2

Screening colonoscopy for CRC in the asymptomatic population, complication rate of perforation and bleeding: a meta-analysis (random effect)

Complications were very rare and described in only five cohorts. Perforation developed in 0.01% of cases (95% confidence interval 0.006–0.02%), and bleeding in 0.05% of cases (95% confidence interval 0.02–0.09%). Three cases of unrelated death, out of 3,196 participants, within 30 days of the procedure, were described in only one cohort [4].

Discussion

More than 30 studies have been published in the medical literature to date describing a cohort of screening colonoscopy for asymptomatic people. This is the first meta-analysis published, establishing the diagnostic yield of colonoscopy, mortality, and morbidity in the screening setting. We chose to include only papers with more than 500 participants, since a smaller cohort cannot be a good representative of the population when diagnostic yield or complication rate are concerned. To the best of our knowledge this is the first meta-analysis assessing screening colonoscopy diagnostic yield. This could be particularly useful in defining quality parameters such as adenoma detection rate, and help universal adoption of a recommendation for colonoscopy as the method of choice for screening.

The findings of this meta-analysis are not surprising and support the worldwide acceptance of colonoscopy for screening for CRC. The diagnostic yield is almost 1% and a clear favorable stage shift is achieved. This was clearly described recently by Gupta et al. [17]. They found in Olmsted County, USA, that following screening efforts over 20 years the incidence and mortality from CRC decreased significantly, and there was a clear favorable stage shifting. They also demonstrated an increase in 5-year survival for patients diagnosed in a screening program (73%) in comparison with symptomatic patients (46%). This encouraging advance is related to increased rate of colonoscopy and adenoma resection from 86/100,000 to 320/100,000 over this time period. The complication rates are also acceptable and better than complication rates of diagnostic and therapeutic colonoscopies performed in symptomatic patients [18]. The perforation rate is 1:10,000 and the bleeding rate 1:2,000. Only three cases of deaths, unrelated directly to the colonoscopy, were described within 30 days of the procedure in one series [4]. In the paper of Regula et al. [11], colonoscopy performed in 50,148 patients resulted in no cases of mortality.

The weakness of our study is the heterogenous population of high- and average-risk screenees. Most of the asymptomatic participants belong to an average-risk population, but in some of the studies high-risk individuals (with family history of CRC) were included. A separate calculation for these two groups could not be performed, since the data presented uniformly for both. It may be true that the diagnostic yield may be higher for high-risk individuals, but the number of these in the cohorts was small and overwhelmed by the average-risk screenees. In addition there is one major publication that included 50,000 participants, while together the remaining nine papers add only 18,000 more participants to the study. We could not answer all our questions for all the papers. One paper reported only 627 screenees [3]; one paper included only women [9] and one mainly men [4]; four papers did not provide data on completeness of the procedure [3, 7, 8, 12]; three papers lacked information about complications [6, 7, 10]; four paper gave no information specifying the distribution of patients into average or high risk [3, 5, 6, 8]; one paper did not specify the detection of advanced adenoma [3]; and one paper included individuals form the age of 40 years [6].

Inspite of all the above, we believe that our findings support the notion that colonoscopy is feasible and a suitable method for screening for CRC in the setting of an asymptomatic population. We conclude that colonoscopy may be offered as a screening tool for high- as well as average-risk asymptomatic population for CRC since the diagnostic yield for polyps and cancer is high, and the complication rate is low.

Copyright information

© Springer Science+Business Media, LLC 2008