Systematic Review With Meta-analysis

The Appropriateness of Colonoscopy Increases the Probability of Relevant Findings and Cancer While Reducing Unnecessary Exams

Leonardo Frazzoni; Marina La Marca; Franco Radaelli; Cristiano Spada; Liboria Laterza; Rocco Maurizio Zagari; Franco Bazzoli; Cesare Hassan; Marzio Frazzoni; Mario Dinis-Ribeiro; Lorenzo Fuccio


Aliment Pharmacol Ther. 2020;53(1):22-3. 

In This Article


We based our methodology on the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) recommendations (see Table S1)[13] and on the GRADE rating approach.[14] Data sources and search strategy, the selection process, data extraction and quality assessment are reported in Appendix S1.

Inclusion and Exclusion Criteria

To be included in the systematic review, we considered all studies comparing appropriate vs. inappropriate colonoscopies in terms of diagnostic yield. Appropriateness was defined according to criteria provided by ASGE in 1992[10] and 2000,[11] and by EPAGE-I[8] and EPAGE-II[9] (see Table 1 and Table S2). Prospective and retrospective studies, published as full text, including at least 20 patients were considered for inclusion, without language restriction. In order to be included, studies had to present results on a per-patient basis, while those presenting data on a per-diagnosis basis were excluded (eg a patient had to be considered only once even though he had multiple diagnoses at colonoscopy, thus avoiding duplicates).

Outcome Assessment

The primary outcome was the probability of relevant findings, CRC and IBD at colonoscopy on a per-patient basis, according to whether the indication to perform the examination was appropriate or not. The secondary outcome was the diagnostic accuracy of appropriateness reported by the included studies according to guidelines (ie ASGE and EPAGE), considered as a diagnostic test for relevant findings, CRC and IBD. Other secondary outcomes were the diagnostic accuracy of appropriateness for IBD, the rate of appropriate colonoscopies, the overall rate of relevant findings and CRC, the rate of the various appropriate and inappropriate indications and the rate of adverse events related to appropriate or inappropriate colonoscopies. Relevant findings were variably defined between studies as detailed in Table 2.

Statistical Analysis

We pooled proportions and risk ratios (RR) along with 95% confidence interval (CI) by random effects model. Statistical heterogeneity was computed according to the I2 statistic and was considered high when I 2 > 50%. Summary estimates of sensitivity, specificity, positive likelihood ratio (LR+) and negative likelihood ratio (LR-) of appropriateness as a diagnostic test, along with 95% CI were computed by a bivariate mixed-effects regression model. A summary receiving operating characteristic (SROC) curve was generated. Likelihood ratios for appropriateness were applied to the pooled prevalence of the outcome of interest (ie pre-test probability), in order to compute the post-test probability.

On the one hand, we calculated the number of 'more findings per 100 procedures' for appropriate vs. inappropriate colonoscopy by multiplying the pooled proportion of relevant findings and CRC in the 'inappropriate' group by the pooled RR with 95% CI. On the other hand, the number of saved 'unnecessary' procedures (ie inappropriate procedures with negative result, or true negatives) per 100 colonoscopies corresponded to the specificity of 'appropriateness' as a diagnostic test.

In order to explain heterogeneity, sensitivity analyses based on subgroup meta-analyses and metaregression were performed; for the analysis on diagnostic accuracy, we added one covariate at a time to the bivariate model. The following variables were selected a priori: (a) criteria applied to define appropriateness (eg ASGE-1992, ASGE-2000, EPAGE-I and EPAGE-II), (b) publication year, (c) period of enrolment, (d) monocentre vs. multicentre studies, (e) proportion of male patients, (f) mean patient age, (g) studies including only advanced adenomas or polyps ≥1 cm among relevant findings, (h) studies included in the previous meta-analysis by Hassan et al[12] vs. more recent studies and (i) the percentage of each reported indication (eg haematochezia, abdominal pain, etc) among appropriate and inappropriate exams respectively.

The publication bias was assessed by funnel plot and by the regression test by Harbord, Egger and Sterne.[15] All the analyses were performed with STATA version 16 (StataCorp).