Table 1. Potential sources of heterogeneity of histologic en bloc (R0) resection rate among 60 studies of patients that underwent colorectal endoscopic submucosal dissection.
Variable | Studies, n | Tumors, n | R0 resection rate (95 % CI), % | P value1 |
Type of article | 0.23 | |||
Full-text journal | 41 | 6006 | 84 (80, 87) | |
Abstract | 19 | 2306 | 81 (72, 87) | |
Study design | 0.04 | |||
Retrospective | 36 | 6738 | 85 (81, 88) | |
Prospective | 7 | 531 | 75 (62, 85) | |
Setting | 0.11 | |||
Single center | 49 | 6876 | 84 (80, 87) | |
Multicenter | 4 | 1079 | 73 (58, 83) | |
Start year of data collection | 0.31 | |||
< 2005 | 14 | 1586 | 77 (70, 83) | |
2005 – 2009 | 30 | 4835 | 85 (81, 88) | |
≥ 2010 | 11 | 826 | 86 (71, 93) | |
Continent | 0.004 | |||
Asia | 40 | 7392 | 87 (84, 90) | |
Europe | 16 | 806 | 70 (62, 77) | |
South America (Brazil) | 2 | 18 | 83 (59, 95) | |
North America (USA) | 2 | 96 | 65 (55, 73) | |
Average age, years2 | 0.47 | |||
≤ 64 | 14 | 1798 | 84 (77, 88) | |
65 – 67 | 14 | 3563 | 82 (77, 87) | |
> 67 | 14 | 1444 | 87 (78, 93) | |
Female, %2 | 0.33 | |||
≤ 36 | 15 | 1613 | 84 (79, 88) | |
37 – 43 | 14 | 2172 | 88 (81, 93) | |
≥ 44 | 14 | 2066 | 80 (72, 86) | |
Number of tumors2 | 0.71 | |||
< 40 | 20 | 418 | 86 (78, 91) | |
40 – 90 | 20 | 1291 | 80 (73, 86) | |
> 90 | 20 | 6603 | 84 (79, 88) | |
Average tumor size, mm2 | 0.09 | |||
≤ 27 | 16 | 1844 | 85 (81, 89) | |
28 – 34 | 16 | 2409 | 85 (78, 90) | |
≥ 34 | 16 | 2061 | 80 (70, 88) | |
Histology | ||||
Carcinoid | 7 | 221 | 85 (79, 89) | 0.19 |
Non-carcinoid | 48 | 5051 | 82 (78, 86) | |
Length of the procedure, min§ | 0.009 | |||
≤ 61 | 15 | 2141 | 89 (84, 93) | |
62 – 101 | 15 | 2954 | 84 (79, 88) | |
> 101 | 15 | 1564 | 78 (68, 85) | |
N, number; R0, histologic en bloc resection rate
Potential sources of heterogeneity was assessed with metaregression. P < 0.05 indicates that the variable significantly explains part of the between study heterogeneity (i. e. an effect mofier). Differences in continent, lenth of the procedure, study design and average tumor size explains 18 %, 15 %, 8 %, and 4 % of the heterogeneity respectively.
Indicates variables that were cut at tertiles in order to ensure comparability of number of studies between groups.