A flow chart showing the study selection is presented in Figure 1. The study by Choi et al. performed open and laparoscopic cystectomy, published its result in median and range rather than mean and SDs, and was subsequently excluded as we were unable to obtain these data from the authors. The study by Matros et al. published median end points rather than the mean, but we can obtain appropriate data for time to first flatus and LOS from the published meta-analysis. We could not get the full text or abstract of one citation mentioned in the previous meta-analysis to extracted original data, so it was excluded. Finally, 17 studies[12–14,16–21,23,24,26–31] were included in our meta-analysis.
Quorum flow diagram depicting process of study selection. ENT, ear, nose, and throat; RCTs, randomized controlled trials.
The characteristics of the included 17 RCTs are presented in Table 1. These studies were published between 2002 and 2012. Five studies were conducted in the United States, three in Europe, seven in Asia, and two in Africa. A total of 686 patients were in the treatment group and 688 in the control. Operations were performed via open surgery,[13,14,17–20,23,24,27,28,30] laparoscopic techniques, or both,[26,31] not clearly stated.[16,29] Studies by McCormick et al. and Crainic et al. reported their open and laparoscopic surgery results separately, enabling calculation of their outcomes independently. All the included trials informed primary outcomes of time to first flatus; 14 trials documented time to first bowel movement and 13 trials documented LOS. The units of outcomes reported were variable and all were converted to days to allow uniform analysis. There was variation in the indications of surgery; seven studies were colectomy,[12–14,26–28,31] four studies were cesarean section,[17,18,20,23] and the remaining five studies (gastrectomy, gastrointestinal surgery, intestinal resection, appendectomy, ileostomy) were defined as others for small trial size. In two trials,[16,30] the gastrointestinal surgeries were done on children.
Risk of Bias
The results of the risk of bias assessments are reported in Figure 2. Overall, all studies had low risk of bias. Only one trial was at high risk of bias for sequence generation (not truly random), which was randomized based on hospital record number. But the method of randomization was unclear (not reported) in five trials.[12,13,16,26,30] Allocation concealment was stated clearly only in five trials.[17,20,23,27,28] Blinding of observers was part of the trial design in five studies,[20,23,27,28,30] but only one trial was double blinded with a placebo group. Only one trial reported incomplete outcome data.
Risk of bias summary: Review authors' judgments about each risk of bias item across all included studies. Low risk; High risk; Unclear risk.
Quantitative Synthesis of Data
Time to First Flatus A total of 686 patients were in the gum-chewing group and 688 in the control. The overall effect of the meta-analysis favored chewing gum, with a WMD of 0.31 days reduction (95% CI, ™0.42 to ™0.19; P = 0.000). However, it should be noted that nine trials have 95% CI crossing the zero level, and had a large heterogeneity for I2 = 73.5%; P = 0.000. (Fig. 3)
Overall and subgroup analysis of weighted mean difference (WMD) for time to first flatus (days) using random-effects model. CI, confidence interval.
Time to First Bowel Movement A total of 573 patients were in the gum-chewing group and 567 in the control. The overall effect of the meta-analysis favored chewing gum, with a WMD of 0.51 days reduction (95% CI, ™0.73 to ™0.29; P = 0.000). However, seven trials had 95% CI crossing the zero level, and also had a large heterogeneity for I2 = 86.4%; P = 0.000. (Fig. 4)
Overall and subgroup analysis of weighted mean difference (WMD) for time to first bowel movement (days) using random-effects model. CI, confidence interval.
LOS A total of 600 patients were in the gum-chewing group and 592 in the control. The overall effect of the meta-analysis favored gum chewing, with a WMD of 0.72 days reduction (95% CI, ™1.02 to ™0.43; P = 0.000). However, 10 trials had 95% CI crossing the zero level, and also had a large heterogeneity for I2 = 87.2%; P = 0.000. (Fig. 5)
Overall and subgroup analysis of weighted mean difference (WMD) for time to length of hospital stay (days) using random-effects model. CI, confidence interval.
Postoperative complications were reported in all but two studies.[16,29] A summary of complications rate is shown in Table 1. It shows that gum chewing was associated with a lower incidence of postoperative complications. But complications were varied in each trail and not related to individual patients experiencing them; hence, statistical analysis of significance between these groups was not valid.
Firstly, subgroup analysis was performed to explore heterogeneity between studies and assess the robustness of our findings (Table 2). The studies that underwent colectomy yielded a WMD of 0.3 days reduction in time to flatus but without significance (95% CI, ™0.68 to 0.08; P = 0.119); a significant reduction of 0.74 days in time to bowel movement (95% CI, ™1.37 to ™0.11; P = 0.038); a WMD of 1.10 days reduction in LOS but without significance (95% CI, ™2.37 to 0.177; P = 0.089). All of them had a large heterogeneity with I 2 = 73.5%, 86.4%, and 87.2%, respectively. Subgroup analysis for cesarean section demonstrated a significant decrease in all outcomes: time to flatus with WMD, ™0.26 (95% CI ™0.31 to ™0.22; P = 0.000); time to bowel movement with WMD, ™0.26 (95% CI ™0.43 to ™0.08; P = 0.004); LOS with WMD, ™0.21 (95% CI ™0.39 to ™0.03; P = 0.021). All had a large heterogeneity except for time to flatus analysis (I 2 = 1.0%, I 2 = 85.8%, and I 2 = 86.7%, respectively). Studies that underwent other abdominal surgery yielded a robust finding, associated with a significant reduction in all outcomes (P = 0.000) and demonstrated no heterogeneity (I 2 < 60%, P > 0.05). Studies evaluating open surgery demonstrated similar results with the overall effect. However for the laparoscopic, results were completely opposite, with all outcomes having no significance.
Even though a subgroup analysis has been used, the heterogeneity of colectomy studies remained large. Sensitivity analyses were performed to examine whether the effect estimate was robust by sequential omission of individual studies using random-effects estimates. The colectomy studies, excluding one study that reported 17 missing values and the duration time in treatment group longer than control, yielded opposite results in time to flatus (WMD, 0.49; 95% CI, ™0.88 to ™0.11; P = 0.012), but with substantial evidence of heterogeneity (I2 = 69.7%, P = 0.002). There was no change in time to bowel movement and LOS. In this base, excluding other two studies[12,13] in which patients were less than 30 yielded significant reduction of 0.24 days in time to flatus (P = 0.049), 0.69 days in time to bowel movement (P = 0.000), and demonstrated no heterogeneity (I2 = 0.0% and I2 = 32.5%, respectively). However, there was 1.11 days reduction in LOS without statistical significance (P = 0.106), which also demonstrated large heterogeneity (I2 = 81.3%, P = 0.000). Further exclusion of other single studies did not materially alter the overall combined WMD. Sensitivity analysis was not performed on laparoscopic surgery for the small number of studies.
Publication bias was assessed by Begg's funnel plots and Egger's tests. The shapes of the Begg's funnel plots revealed no obvious asymmetry (Fig. 6). The Egger's test was then used to statistically assess funnel plot symmetry. The funnel plot was relatively symmetrical, suggesting that publication bias was not present. (t = ™0.38, P = 0.580 for time to first flatus; t = ™1.05, P = 0.287 for time to bowel movement; t = ™1.56, P = 0.078 for LOS). These indicated that the results of these meta-analyses were relatively stable and that publication bias was unlikely to affect the results of the meta-analysis.
Funnel plots for publication bias in the studies of the meta-analysis of chewing gum following abdominal surgery. s.e., standard error; WMD, weighted mean difference.
A statistically significant effect of chewing gum on abdominal surgery was first observed after publication of the first trial in 2002 (1.1 days reduction in time to flatus, P < 0.01; 2.7 days reduction in time to bowel movement, P < 0.01). Although subsequent trials had increased the precision of the point estimate, no substantive change had occurred in the direction or magnitude of the treatment effect. A statistically significant effect on time to first flatus of chewing gum following abdominal surgery was consistently observed after publication of the seventh trial in 2006, third trial in 2006 on time to bowel movement, and seventh trial in 2006 on LOS. (Fig. 7)
J Gastroenterol Hepatol. 2013;28(7):1122-1132. © 2013 Blackwell Publishing