Interval Colorectal Cancers: Risk Factors and Outcomes
Interval Colorectal Cancers: Risk Factors and Outcomes
We conducted and reported this systematic review according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines, and followed a priori established protocol. The quality of evidence was rated using the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) approach.
In this systematic review, we included studies that met the following criteria: (i) population-based or multi-center studies; (ii) reported CRCs occurring within 3 years of a colonoscopy (for any indication); and (iii) reported the overall and/or site-specific prevalence of CRCs occurring within 3 years of a colonoscopy vis-à -vis the total number of CRC cases. Both retrospective and prospective studies were included, provided complete (or near-complete (>90%)), minimum 5-year follow-up was available, to allow accurate estimation of interval CRCs. We included case–control studies, in which the cases included all patients with CRC and a subset of patients with interval CRCs could be identified. Multicenter studies were only included if they were deemed to be representative of the majority of general population in a geographic area. We defined interval CRCs as those occurring after an index colonoscopy, regardless of whether an alternative screening strategy (flexible sigmoidoscopy, stool-based tests, etc.) had been used before or after index colonoscopy.
We excluded the following studies from the meta-analysis: (i) single-center studies, as these may not be able to accurately identify all interval CRCs, (ii) referral-center studies, because of risk of selection bias, (iii) studies with inadequate information on total number of CRCs identified during the time period, as prevalence of interval CRCs could not be determined, (iv) prospective studies with incomplete follow-up, and (v) randomized controlled trials given highly controlled environment in these with risk of Hawthorne effect. Although these studies were excluded from quantitative synthesis, key findings from these studies were included in the systematic review. In case of multiple studies from the same cohort, we included data from the most recent comprehensive report; if there was minimal overlap of time period, then both studies were included.
We conducted a comprehensive search of multiple electronic databases from each databases' inception to 8 October 2013, with no language restrictions. The databases included: Ovid MEDLINE, Ovid EMBASE, Ovid Cochrane Central Register of Controlled Trials, Ovid Cochrane Database of Systematic Reviews, Web of Science, and Scopus. The search strategy was designed and conducted by an experienced medical librarian with input from the study investigators, using controlled vocabulary supplemented with keywords to search for studies of CRCs diagnosed after colonoscopy. The details of the search strategy are included in the Supplementary Appendix online http://www.nature.com/ajg/journal/v109/n9/suppinfo/ajg2014171s1.html. Two authors, independently, reviewed the title and abstract of studies identified in the search to exclude studies that did not address the research question of interest, based on pre-specified inclusion and exclusion criteria (see above). The full text of the remaining articles was examined to determine whether it contained relevant information; discrepancy in article selection was resolved by consensus, in conjunction with the senior investigator (N.J.S.). Next, a recursive search of reference lists of all relevant articles was also conducted to search for additional studies. Third, a manual search of conference proceedings of major gastroenterology and oncology conferences (Digestive Diseases Week, American College of Gastroenterology annual meeting, American Society of Clinical Oncology annual meeting as well as the Gastrointestinal Research Forum; European Society of Medical Oncology annual meeting, and World Congress on GI Cancer) between 2010 and 2013 was conducted to identify additional studies published only in the abstract form.
Data on the following study- and patient-related characteristics were independently abstracted onto a standardized form by two investigators (and discrepancies were resolved by consensus): (i) study characteristics—primary author, time period of study/year of publication, geographic location of the population studied, study design; (ii) CRC characteristics—total number of patients with CRC (overall, proximal, and distal CRC), total number of patients with interval CRCs (overall, proximal, and distal CRC), definition and method of ascertainment of interval CRC (i.e., those patients with CRCs diagnosed within 6–36 months of an index colonoscopy) as well as detected CRCs (i.e., those patients with CRCs diagnosed during or within 6 months of index colonoscopy); (iii) risk factors (clinical and endoscopy-related) associated with interval CRCs, as compared with detected CRCs in individual studies; (iv) outcomes—stage (I/II categorized as 'early' stage, III/IV categorized as 'advanced' stage), grade (well- and moderately differentiated grouped into G1/G2, poorly differentiated and undifferentiated grouped into G3/G4), as well as differences in survival of interval and detected CRCs. In addition, where reported, the difference in the genetic profile of interval and detected CRCs was also abstracted. First or senior authors of studies were contacted to provide additional information where required.
The methodological quality of observational studies was assessed using published, non-validated, criteria for prevalence studies, by two study investigators independently. This scale grades studies according to eight methodological criteria, with a total possible score from 0 to 8.
Primary Outcome. The primary outcome of interest was the prevalence of interval CRCs, i.e., what proportion of CRCs are interval CRCs. These were separately assessed for any CRC site, proximal CRCs (cecum, ascending colon, transverse colon to the splenic flexure), and distal CRCs (descending colon, sigmoid colon, and rectum).
A priori hypotheses to explain potential heterogeneity in the direction and magnitude of effect among different observational studies included location of study (North America vs. Europe) and study setting (population-based vs. multicenter studies). Sensitivity analysis based on modifying the definition of interval CRCs (diagnosed within 60 months of previous colonoscopy, as opposed to within 36 months of colonoscopy) was also performed.
Secondary Outcomes. (i) Risk factors for interval CRCs: In order to identify risk factors associated with interval CRCs, we performed a meta-analysis of differences in the clinical, technical (or endoscopy-related), and tumor biology-related factors (genetic profile of tumors) by comparing patients who developed interval CRCs and those with detected CRCs; (ii) Outcomes of interval CRCs: We compared the stage, grade (at diagnosis), and differences in survival of patients with interval and detected CRCs.
We combined the proportions of patients with interval CRCs in individual studies to estimate a pooled prevalence of interval CRCs among all patients diagnosed with CRC, along with a 95% confidence interval (CI). We compared pooled prevalence of proximal interval CRCs, with the pooled prevalence of distal interval CRCs, using an odds ratio (OR) with 95% CI. We identified risk factors associated with development of interval CRCs relative to detected CRCs, by performing a meta-analysis of difference in distribution of clinical and endoscopy-related factors among the two groups by pooling maximally adjusted ORs reported in the respective studies as well as pooling unadjusted univariate analysis (using 2 × 2 tables, when the required data were available from the individual studies), and expressed the results as OR with 95% CI. We also performed a time-trend analysis by abstracting data on the proportion of patients identified as having interval CRCs based on time period of patient identification (1990s, 2000–2005, and beyond 2005); we did not perform an analysis based on the year of publication as these were not reflective of time period of patient recruitment. To give a more conservative estimate of the prevalence of interval CRCs, we pooled data using the random effects model described by DerSimonian and Laird.
We assessed heterogeneity between study-specific estimates using inconsistency index (I statistic). This estimates what proportion of total variances across studies was due to heterogeneity rather than chance; a value of >50% was considered suggestive of substantial heterogeneity. Once heterogeneity was noted, we investigated between-study sources of heterogeneity using subgroup analyses by stratifying original estimates according to study characteristics as described above. We planned to ascertain publication bias, qualitatively, by visual inspection of Funnel plot, and quantitatively, by Egger's regression test. For all tests (except for publication bias), a probability level <0.05 was considered statistically significant. All analyses were performed using Comprehensive Meta-Analysis version 2 (Biostat, Englewood, NJ).
Methods
We conducted and reported this systematic review according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines, and followed a priori established protocol. The quality of evidence was rated using the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) approach.
Selection Criteria
In this systematic review, we included studies that met the following criteria: (i) population-based or multi-center studies; (ii) reported CRCs occurring within 3 years of a colonoscopy (for any indication); and (iii) reported the overall and/or site-specific prevalence of CRCs occurring within 3 years of a colonoscopy vis-à -vis the total number of CRC cases. Both retrospective and prospective studies were included, provided complete (or near-complete (>90%)), minimum 5-year follow-up was available, to allow accurate estimation of interval CRCs. We included case–control studies, in which the cases included all patients with CRC and a subset of patients with interval CRCs could be identified. Multicenter studies were only included if they were deemed to be representative of the majority of general population in a geographic area. We defined interval CRCs as those occurring after an index colonoscopy, regardless of whether an alternative screening strategy (flexible sigmoidoscopy, stool-based tests, etc.) had been used before or after index colonoscopy.
We excluded the following studies from the meta-analysis: (i) single-center studies, as these may not be able to accurately identify all interval CRCs, (ii) referral-center studies, because of risk of selection bias, (iii) studies with inadequate information on total number of CRCs identified during the time period, as prevalence of interval CRCs could not be determined, (iv) prospective studies with incomplete follow-up, and (v) randomized controlled trials given highly controlled environment in these with risk of Hawthorne effect. Although these studies were excluded from quantitative synthesis, key findings from these studies were included in the systematic review. In case of multiple studies from the same cohort, we included data from the most recent comprehensive report; if there was minimal overlap of time period, then both studies were included.
Search Strategy
We conducted a comprehensive search of multiple electronic databases from each databases' inception to 8 October 2013, with no language restrictions. The databases included: Ovid MEDLINE, Ovid EMBASE, Ovid Cochrane Central Register of Controlled Trials, Ovid Cochrane Database of Systematic Reviews, Web of Science, and Scopus. The search strategy was designed and conducted by an experienced medical librarian with input from the study investigators, using controlled vocabulary supplemented with keywords to search for studies of CRCs diagnosed after colonoscopy. The details of the search strategy are included in the Supplementary Appendix online http://www.nature.com/ajg/journal/v109/n9/suppinfo/ajg2014171s1.html. Two authors, independently, reviewed the title and abstract of studies identified in the search to exclude studies that did not address the research question of interest, based on pre-specified inclusion and exclusion criteria (see above). The full text of the remaining articles was examined to determine whether it contained relevant information; discrepancy in article selection was resolved by consensus, in conjunction with the senior investigator (N.J.S.). Next, a recursive search of reference lists of all relevant articles was also conducted to search for additional studies. Third, a manual search of conference proceedings of major gastroenterology and oncology conferences (Digestive Diseases Week, American College of Gastroenterology annual meeting, American Society of Clinical Oncology annual meeting as well as the Gastrointestinal Research Forum; European Society of Medical Oncology annual meeting, and World Congress on GI Cancer) between 2010 and 2013 was conducted to identify additional studies published only in the abstract form.
Data Extraction and Quality Assessment
Data on the following study- and patient-related characteristics were independently abstracted onto a standardized form by two investigators (and discrepancies were resolved by consensus): (i) study characteristics—primary author, time period of study/year of publication, geographic location of the population studied, study design; (ii) CRC characteristics—total number of patients with CRC (overall, proximal, and distal CRC), total number of patients with interval CRCs (overall, proximal, and distal CRC), definition and method of ascertainment of interval CRC (i.e., those patients with CRCs diagnosed within 6–36 months of an index colonoscopy) as well as detected CRCs (i.e., those patients with CRCs diagnosed during or within 6 months of index colonoscopy); (iii) risk factors (clinical and endoscopy-related) associated with interval CRCs, as compared with detected CRCs in individual studies; (iv) outcomes—stage (I/II categorized as 'early' stage, III/IV categorized as 'advanced' stage), grade (well- and moderately differentiated grouped into G1/G2, poorly differentiated and undifferentiated grouped into G3/G4), as well as differences in survival of interval and detected CRCs. In addition, where reported, the difference in the genetic profile of interval and detected CRCs was also abstracted. First or senior authors of studies were contacted to provide additional information where required.
The methodological quality of observational studies was assessed using published, non-validated, criteria for prevalence studies, by two study investigators independently. This scale grades studies according to eight methodological criteria, with a total possible score from 0 to 8.
Outcomes Assessed
Primary Outcome. The primary outcome of interest was the prevalence of interval CRCs, i.e., what proportion of CRCs are interval CRCs. These were separately assessed for any CRC site, proximal CRCs (cecum, ascending colon, transverse colon to the splenic flexure), and distal CRCs (descending colon, sigmoid colon, and rectum).
A priori hypotheses to explain potential heterogeneity in the direction and magnitude of effect among different observational studies included location of study (North America vs. Europe) and study setting (population-based vs. multicenter studies). Sensitivity analysis based on modifying the definition of interval CRCs (diagnosed within 60 months of previous colonoscopy, as opposed to within 36 months of colonoscopy) was also performed.
Secondary Outcomes. (i) Risk factors for interval CRCs: In order to identify risk factors associated with interval CRCs, we performed a meta-analysis of differences in the clinical, technical (or endoscopy-related), and tumor biology-related factors (genetic profile of tumors) by comparing patients who developed interval CRCs and those with detected CRCs; (ii) Outcomes of interval CRCs: We compared the stage, grade (at diagnosis), and differences in survival of patients with interval and detected CRCs.
Statistical Analysis
We combined the proportions of patients with interval CRCs in individual studies to estimate a pooled prevalence of interval CRCs among all patients diagnosed with CRC, along with a 95% confidence interval (CI). We compared pooled prevalence of proximal interval CRCs, with the pooled prevalence of distal interval CRCs, using an odds ratio (OR) with 95% CI. We identified risk factors associated with development of interval CRCs relative to detected CRCs, by performing a meta-analysis of difference in distribution of clinical and endoscopy-related factors among the two groups by pooling maximally adjusted ORs reported in the respective studies as well as pooling unadjusted univariate analysis (using 2 × 2 tables, when the required data were available from the individual studies), and expressed the results as OR with 95% CI. We also performed a time-trend analysis by abstracting data on the proportion of patients identified as having interval CRCs based on time period of patient identification (1990s, 2000–2005, and beyond 2005); we did not perform an analysis based on the year of publication as these were not reflective of time period of patient recruitment. To give a more conservative estimate of the prevalence of interval CRCs, we pooled data using the random effects model described by DerSimonian and Laird.
We assessed heterogeneity between study-specific estimates using inconsistency index (I statistic). This estimates what proportion of total variances across studies was due to heterogeneity rather than chance; a value of >50% was considered suggestive of substantial heterogeneity. Once heterogeneity was noted, we investigated between-study sources of heterogeneity using subgroup analyses by stratifying original estimates according to study characteristics as described above. We planned to ascertain publication bias, qualitatively, by visual inspection of Funnel plot, and quantitatively, by Egger's regression test. For all tests (except for publication bias), a probability level <0.05 was considered statistically significant. All analyses were performed using Comprehensive Meta-Analysis version 2 (Biostat, Englewood, NJ).