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

Disclosures

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

In This Article

Results

Overall, 21 studies for a total of 24 comparison arms and 19,822 patients (mean age 44–64 years, male gender 39%-63%) were included[16–36] (see Figure S1). The criteria for appropriateness of colonoscopy according to ASGE and EPAGE are detailed in Table 1 and Table S2. Nineteen studies[16–29,31–33,35,36] were prospective, and 17[16–19,35] were conducted in Europe; eight studies[29–36] were published since 2007 and were not included in the previous meta-analysis by Hassan et al; mean patient age ranged from 52 to 64 years, whereas the proportion of male patients varied from 45% to 63% (see details in Table 2). The ASGE-1992, ASGE-2000, EPAGE-I and EPAGE-II criteria were applied in four,[16,17,26,29] six,[20,22,23,25,28,31] nine[18,19,21,24,26,27,32,35,36] and five[30,32–35] study arms respectively. Relevant findings were mostly consistent among the included studies and comprised CRC, adenomatous polyps, IBD and other colitis, stenosis and angiodysplasia (see Table 3). Appropriate colonoscopies were 15,162 (71%, CI 64%-78%). The most commonly reported appropriate indications were: haematochezia 17% (CI 13%-20%), CRC screening 11% (CI 7%-15%), chronic abdominal pain 8% (CI 5%-11%), iron deficiency anaemia 7% (CI 5%-8%), post-polypectomy surveillance 7% (CI 5%-9%), lower GI symptoms 7% (CI 3%-11%), positive FOBT 6% (CI 3%-9%), CRC surveillance 6% (CI 4%-7%), change in bowel habit 5% (CI 3%-8%), chronic diarrhoea 5% (CI, 4%-7%), IBD follow-up and evaluation of an imaging abnormality 2% (CI 2%-3%) each. The most common inappropriate indications were as follows: chronic constipation 8% (CI 5%-11%), abdominal pain 6% (CI 4%-9%), post-polypectomy or post-CRC follow-up at different interval 6% (CI 5%-7%), lower GI symptoms 5% (CI 3%-7%), haematochezia 4% (CI 2%-5%), acute or uncomplicated diarrhoea 1% (CI 1%-2%), transitory change in bowel habit 1% (CI 0%-2%) and undefined anaemia 1% (CI 0%-1%). Publication bias and quality of evidence assessment are reported in Appendix S2.

Relevant Findings

The pooled prevalence of relevant findings was 28% (CI 24%-32%) overall, 34% (CI 31%-38%) among appropriate colonoscopies and 18% (CI 15%-21%) among inappropriate examinations. Appropriateness significantly increased the probability of relevant findings at colonoscopy as compared to inappropriate indication, with high heterogeneity (4,755/15,162 vs. 904/4,852; RR 1.81, CI 1.53–2.14; I 2 81%; see Figure 1A). In absolute terms, performing colonoscopy with appropriate indication would lead to find on average 15 more relevant findings per 100 patients, while the true effect could lie between 10 and 21 more per 100 patients, when compared to inappropriate indication.

Figure 1.

Pooled risk ratio (RR) of appropriate vs. inappropriate colonoscopies for relevant findings (A) and cancer (B). ASGE, American Society for Gastrointestinal Endoscopy. EPAGE, European Panel for Appropriateness in Gastrointestinal Endoscopy

Colonoscopy appropriateness had sensitivity 88% (CI 85%-91%) and specificity 24% (CI 20%-29%) for relevant findings (see Figure 2A); LR+ was 1.16 (CI 1.12–1.21) and LR- was 0.49 (CI 0.40–0.60). At the pooled 34% prevalence of relevant findings (ie pre-test probability), performing a colonoscopy with appropriate indication would increase the post-test probability of relevant findings to 37%, whereas inappropriate colonoscopy would result in a 20% post-test probability. In absolute terms, performing colonoscopy with appropriate indication would save on average 24 examinations per 100 patients, while the true effect could lie between 20 and 29 per 100 patients, when compared to inappropriate indication.

Figure 2.

Pooled sensitivity and specificity of appropriate colonoscopy for relevant findings (A) and cancer (B) according to the summary receiver operating characteristic (SROC) curve. Sensitivity and specificity for inappropriate colonoscopies can be derived as complementary (eg for relevant findings: sensitivity 1-0.88 = 0.12 and specificity 1-0.24 = 0.76)

Colorectal Cancer

The pooled prevalence of CRC was 5% (CI 4%-6%) overall, 7% (CI 6%-8%) among appropriate colonoscopies and 2% (CI 1%-2%) among inappropriate examinations.

Appropriateness significantly increased the probability of discovering CRC at colonoscopy as compared to inappropriate indication, with low heterogeneity (998/15,162 vs. 86/4,852; RR 3.62, CI 2.44–5.37; I 2 48%; see Figure 1B). In absolute terms, performing colonoscopy with appropriate indication would lead to find on average five more CRCs per 100 patients, while the true effect could lie between three and nine more per 100 patients.

Appropriateness of colonoscopy had sensitivity 97% (CI 93%-98%) and specificity 22% (CI 18%-26%) for CRC (see Figure 2B); LR+ was 1.23 (CI 1.18–1.29) and LR- was 0.16 (CI 0.08–0.31). At the pooled 5% prevalence of CRC (ie pre-test probability), performing a colonoscopy with appropriate indication would increase the post-test probability of CRC to 6%, whereas inappropriate colonoscopy would result in a 1% post-test probability. In absolute terms, performing colonoscopy with appropriate indication would save on average 22 examinations per 100 patients, while the true effect could lie between 18 and 26 per 100 patients. The analyses in absolute terms are summarised in Figure 3.

Figure 3.

Consequences of performing colonoscopy with appropriate indication on efficiency, in terms of diagnostic gain and exams saving, as compared to inappropriate indication (pooled average effects per 100 patients are displayed)

Inflammatory Bowel Disease

The pooled prevalence of IBD was 5% (CI 4%-6%) overall, 6% (CI 5%-8%) among appropriate colonoscopies and 4% (CI 2%-6%) among inappropriate examinations. Appropriateness significantly increased the probability of IBD at colonoscopy as compared to inappropriate indication, with high heterogeneity (575/11,341 vs. 128/3,992; RR 1.86, CI 1.09–3.19; I2 81%; see Figure S6). In absolute terms, performing colonoscopy with appropriate indication would lead to find on average three more relevant findings per 100 patients, while the true effect could lie between one and nine more per 100 patients, when compared to inappropriate indication.

Colonoscopy appropriateness had sensitivity 89% (CI 80%-94%) and specificity 24% (CI 20%-28%) for IBD; LR+ was 1.2 (CI 1.1–1.3) and LR- was 0.45 (CI 0.24–0.84). At the pooled 5% prevalence of relevant findings (ie pre-test probability), performing a colonoscopy with appropriate indication would increase the post-test probability of relevant findings to 6%, whereas inappropriate colonoscopy would result in a 2% post-test probability. In absolute terms, performing colonoscopy with appropriate indication would save on average 24 examinations per 100 patients, while the true effect could lie between 20 and 28 per 100 patients, when compared to inappropriate indication.

Sensitivity Analysis

A sensitivity analysis was performed in order to assess whether the different appropriateness criteria, namely ASGE-1992, ASGE-2000, EPAGE-I and EPAGE-II had a different impact on the probability of relevant findings, CRC and IBD, or achieved a different diagnostic accuracy. We found a significant decreasing trend in the association with relevant findings from ASGE-2000 (RR 2.56, CI 1.36–4.85) to ASGE-1992 (RR 2.41, CI 1.93–3) to EPAGE-I (RR 1.55, CI 1.37–1.76) to EPAGE-II (RR 1.34, CI 0.99–1.81) (P < 0.01; see Figure 1A), whereas no significant impact was found on CRC probability (P = 0.63; see Figure 1B). Likewise, we detected a significant decreasing trend in the association with IBD from ASGE-2000 (RR 5.24, CI 2.53–10.82) to ASGE-1992 (RR 2.21, CI 1.24–3.62) to EPAGE-I (RR 1.01, CI 0.50–2.03) to EPAGE-II (RR 0.94, CI 0.51–1.74) (P < 0.01; see Figure S6).

A slight effect on diagnostic accuracy for relevant findings was found for the different appropriateness criteria: sensitivity and specificity were 90% (CI 85%-93%) and 27% (CI 20%-36%) for ASGE-1992, 92% (CI 83%-96%) and 24% (CI 18%-32%) for ASGE-2000, 83% (CI 75%-89%) and 28% (CI 20%-37%) for EPAGE-I, and 89% (CI 84%-92%) and 16% (CI 13%-20%) for EPAGE-II (see Figure S2).

No substantial impact on diagnostic accuracy for CRC was found for the various appropriateness criteria, even though EPAGE-II criteria had the lowest specificity (15%, CI 12%-19%) and the highest sensitivity (98%, CI 93%-99%) (see Figure S3).

A slight effect on diagnostic accuracy for IBD was found for the different appropriateness criteria: sensitivity and specificity were 94% (CI 89%-96%) and 51% (CI 50%-52%) for ASGE-2000, 94% (CI 64%-99%) and 26% (CI 22%-31%) for ASGE-1992, 71% (CI 60%-80%) and 28% (CI 22%-36%) for EPAGE-I, and 92% (CI 63%-98%) and 26% (CI 21%-32%) for EPAGE-II.

Other sensitivity analyses including subgroup meta-analyses and metaregression are reported in Appendix S2.

Adverse Events

Overall, seven adverse events were reported in four studies on 3,418 patients (0.2%, CI 0.05%-0.36%). Data were insufficient and not reported according to the appropriateness of indication, thus preventing us from performing a pooled analysis.

Comments

3090D553-9492-4563-8681-AD288FA52ACE

processing....