Meta-analysis comparing chewing gum versus standard postoperative care after colorectal resection

Background Previous incomplete studies investigating the potential of chewing gum (CG) in patients undergoing colorectal resection did not obtain definitive conclusions. This updated meta-analysis was therefore conducted to evaluate the effect and safety of CG versus standard postoperative care protocols (SPCPs) after colorectal surgery. Results Total 26 RCTs enrolling 2214 patients were included in this study. The CG can be well-tolerated by all patients. Compared with SPCPs, CG was associated with shorter time to first flatus (weighted mean difference (WMD) −12.14 (95 per cent c.i. −15.71 to −8.56) hours; P < 0.001), bowl movement (WMD −17.32 (−23.41 to −11.22) hours; P < 0.001), bowel sounds (WMD −6.02 (−7.42 to −4.63) hours; P < 0.001), and length of hospital stay (WMD −0.95 (−1.55 to −0.35) days; P < 0.001), a lower risk of postoperative ileus (risk ratio (RR) 0.61 (0.44 to 0.83); P = 0.002), net beneficial and quality of life. There were no significant differences between the two groups in overall complications, nausea, vomiting, bloating, wound infection, bleeding, dehiscence, readmission, reoperation, mortality. Materials and Methods The potentially eligible randomized controlled trials (RCTs) that compared CG with SPCPs for colorectal resection were searched in PubMed, Embase, Cochrane library, China National Knowledge Infrastructure (CNKI), and Chinese Wanfang databases through May 2016. The trial sequential analysis was adopted to examine whether a firm conclusion for specific outcome can be drawn. Conclusions CG is benefit for enhancing return of gastrointestinal function after colorectal resection, and may be associated with lower risk of postoperative ileus.


INTRODUCTION
Postoperative ileus is an important complication after colorectal surgery, which is characterized mainly by nausea, vomiting and abdominal distension [1]. It is associated with delayed postoperative recovery, prolonged length of hospital stay (LOS) and increased healthcare costs [2]. In the United States, the annual medical expenditures of managing postoperative ileus have been estimated to be 1 billion dollars [3]. Accordingly, it is quite important to prevent and reduce this given condition. To date, many methods (e.g. fluid restriction, early mobilization and nutrition) have been increasingly investigated in order to alleviate postoperative ileus [4][5][6], of which chewing gum (CG) has become a promising option. However, the efficacy of CG for patients after colorectal surgery is still debatable [1,7,8].
Considering these aspects, we therefore undertaken this updated meta-analysis to comprehensively evaluate the effect and safety of CG versus SPCPs for patients undergoing colorectal resection.

Literature search
The identification and selection of studies was graphically depicted in Figure 1. Electronic database searches captured 202 records, and 4 records were obtained from reference lists of relevant reviews. Sixty duplicate records were removed, and 113 records were eliminated by checking the titles and abstracts. The remaining 33 full-text articles were assessed for eligibility. After application of the inclusion criteria, a total of 26 studies [7, comprising 2214 participants were eligible for our inclusion criteria and included to perform meta-analysis.

Risk of bias
Details of risk of bias for individual trials were exhibited in Figure 2a, and the summary of risk of bias of included studies in Figure 2b. Most studies [7, 20-24, 27-29, 33, 36, 39, 40] generated appropriately the random sequence, and 6 studies [20,23,29,30,42,43] conducted adequately allocation concealment. Because it is extremely difficult to blind the participants and surgeons, almost all of these studies were valued as unclear risk of bias for this domain.

Level of evidence
We documented the GRADE evidence profile for all outcomes in Supplementary Table S1. In this systematic review, we assessed 20 outcomes, of which time to first flatus, bowel movement, bowel sounds, feeding, and postoperative ileus were listed as critical outcomes and remaining 15 outcomes such as LOS and overall complications were viewed as important outcomes. The level of evidence was moderate for time to first bowel sounds and complications related to CG, while the level of evidence for remaining outcomes was low or very low.  (Figure 3). We also examined the robust of pooled results through excluding conference abstract and studies with less than 20 patients per arm respectively, and shown that all results did not changed significantly (Supplementary Figure S1, Supporting Information).

Outcome
For time to first flatus, time to first bowel movement, time to first bowel sounds, and LOS, the accumulative Z-curve crossed the sequential monitory boundary, which suggested that the firm conclusions can be drawn based on the present accumulated information size and additional resources should not be wasted to plan further studies (Supplementary Figure S2, Supporting Information). For time to first feeding, trial sequential analysis was not performed due to the fact that finite studies (only 2 trials) were accrued, and thus further studies may be warranted to detect the difference between CG and SPCPs.

Secondary outcomes
CG significantly reduced the risk of postoperative ileus (RR 0.61, 95 per cent c.i. 0.44 to 0.83; P = 0.002) (I2 = 0 per cent), but did not decrease the risk of remaining secondary outcomes compared to SPCPs (Supplementary Figure S3, Supporting Information). The tolerability of CG was reported in 8 studies [19,22,27,29,30,36,41,43], and the qualitative findings suggested that CG was well-tolerated by all patients in active group. Atkinson et al.' study [20] reported economic effect and QoL, and the results suggested that patients who were informed to consume CG had a lower net benefit (MD -€173, 95 per cent c.i. -€1103 to 757) and worse QoL (no difference on day after operation 4 days, but worse at 6 and 12 weeks).
For postoperative ileus, the accumulative Z-curve crossed the conventional monitory boundary, but did not surpass the sequential monitory boundary, which suggested that a false positive conclusion was generated result from chance (Supplementary Figure  S4, Supporting Information). For remaining secondary outcomes, the accumulative Z-curve did not cross the conventional or sequential monitory boundaries (Supplementary Figure S4), which suggested that the accumulative information size too insufficient power to draw firm conclusions.

Subgroup analysis
We conducted subgroup analyses for outcomes with extremely heterogeneity according to the type of surgical approaches, and shown that CG decreased significantly time to first flatus in patients underwent laparoscopic and open colectomy, time to first bowel movement in all patients received various surgical approaches, and time to first bowel sounds, LOS and postoperative ileus in patients received open colectomy. The details of subgroup analyses were documented in Supplementary Table S2.
It is noted that, however, all significant heterogeneities were not decreased or omitted due to implementation of subgroup analyses. After carefully reviewed included trials, Total 60 of records excluded as duplicates  we detected a fact that all included trial generated different coverages due to incorporation of trials with large and small sample size in the present systematic review. And because of this, we confound these extremely heterogeneities. However, we can easily found that the effect estimates with consistent direction were generated by most included trials in each meta-analysis on outcome of interest, and thus we can reasonably establish that these extremely heterogeneities cannot impair the robust and validity of corresponding summary effect size.

Publication bias
The funnel plots, which often be drawn to inspect the existence of publication bias, were depicted in Supplementary Figure S5 (Supporting Information). For most primary (time to first flatus, time to first bowel movement and length of hospital stay) and secondary outcomes, the asymmetric funnel plots were constructed, which indicated the existence of publication bias.

Main findings
Our meta-analysis of 26 RCTs involving 2214 patients suggested that CG was well-tolerated by all patients in active group, and significantly decreased the time to first flatus, time to first bowel movement, time to first bowel sounds, and shortened LOS compared with SPCPs. Furthermore, the patients were informed to consume CG may experience the lower risk of postoperative ileus.

Comparison with other studies
In order to determine the effect and safety of CG in patients undergoing colorectal surgery, 11 meta-analyses [8][9][10][11][12][13][14][15][16][17][18] have been published previously. The characteristics and outcomes of these all meta-analyses have been summarized in Table 2.    RCTs involving 271 patients to conduct meta-analysis that compared the CG with SPCPs in colorectal surgery to carried out meta-analyses respectively. Of 5 metaanalyses, all [11-13, 15, 18] found that CG was associated with shorter time to first bowel movement, and 4 [11,13,15,18]  that patients who were informed to consume CG did not experience shorter LOS. It is noted that, however, a fatal limitation in these meta-analyses is that small sample size were accumulated. Small sample size is not enough power to draw a true inference due to chance (also termed as random error). Therefore, conclusions drawn from these meta-analyses may not be considered as definitive.
Three meta-analyses [14,16,17] with relatively large accumulated sample size (686, 993, and 1668 patients respectively) also evaluated the potential of CG on resumption of gastrointestinal function after colorectal surgery, and found that CG significantly decreased time to first flatus, time to first bowel movement, and LOS. However, these three meta-analyses [14,16,17] did not consider time to first feeding, economic effect and QoL as outcomes, which were useful for clinicians and policy makers. Furthermore, meta-analysis is an important technique to determine the magnitude and significance of an intervention, but these authors did not perform pooled quantitative analyses of most adverse outcomes   (e.g. postoperative ileus and mortality). As a result, the use of CG in patients undergoing colorectal resection still remains controversial. Compared with 11 previous meta-analyses [8][9][10][11][12][13][14][15][16][17][18], the present meta-analysis has several strengths. The period of our meta-analysis was until May 2016, which was longer than the other periods (the lasted study was until December 2014), and thus more studies and sample were accrued in our meta-analysis (26 RCTs involving 2214 patients). Language and publication status restrictions of the studies were not imposed in our study, and thus additional 2 Chinese studies (with 110 and 104 patients respectively) were captured. We quantitatively pooled data of those complications which sufficient information can be extracted, which may help clinicians to objectively assess the safety of CG. More outcomes were evaluated in our meta-analyses, and the effect and safety of CG can be more comprehensively assessed based on our pooled results. Furthermore, we adopted the trial sequential analysis method to determine whether further studies are warranted to detect differences between CG and SPCPs, and confirmed the evidence which CG significantly reduced the time to first flatus, time to first bowel movement, time to first bowel sounds, and LOS.

Limitations
We must not fail to acknowledge the several limitations in our meta-analysis. First, substantial heterogeneity across studies was detected, which can perhaps be explained by a variety of colorectal pathologies and SPCPs in primary studies. Second, most trials in our study were rated as to be high risks of bias, which may overestimated the benefits and harms of CG. Third, we did not excluded the Zaghiyan et al.'s trial, in which patients in active group were informed to chewed sugared gum, because this two types of gum can produce similar effect [3]. Forth, limited data were available on economic effect of CG and QoL; the conclusions may be changed if further studies were added. Fifth, although we performed subgroup analyses according to the different surgery approaches, the significant heterogeneity in each outcome which has been explained to have no impact on robust and validity of corresponding pooled result was not significantly decreased or omitted, and thus further investigation on this issue are warranted. Sixth, we did not carry out additional analyses according to different quantities, frequencies, and durations of consuming CG, which may contribute to substantial heterogeneity.

MATERIALS AND METHODS
We designed this meta-analysis according to the recommendations of Cochrane handbook for systematic reviews of interventions [44] and reported the pooled results in accordance with preferred reporting item for systematic review and meta-analysis statement (PRISMA) [45]. There was no formal protocol for this meta-analysis.

Search strategy
We assigned two investigators (Y.-H.D. and X.T.) searched independently PubMed, Cochrane Central Register of Controlled Trials (CENTRAL), Embase, China National Knowledge Infrastructure (CNKI) and WanFang databases from inception to January 31, 2016, and the last search was updated on May 31, 2016. All search algorithms were structured using Exploded Medial Subject Heading and appropriate keywords, including "chew*", "gum", "colorectal", "resection" and "random*". No language and publication status were imposed. We also checked manually the reference lists of relevant reviews and included studies to capture additional potentially eligible studies.

Study selection
Two independent investigators (Y.-H.D. and X.T.) removed duplicate records, checked the relevance based on titles and abstracts, and reviewed eventually full-text to determine which studies were eligible for our inclusion criteria after the electronic searches were completed. The following criteria were used to examine the eligibility of published RCTs: (i) patients: adult patients undergoing colorectal resection, regardless of surgical approach (open, laparoscopic, hybrid, hand assisted); (ii) intervention: use of CG, irrespective of category (sugarless or sugared) and method of usage (the quantity, frequency, and duration of CG); (iii) comparison: SPCPs; and (iv) reporting one or more of the outcomes described below. We excluded studies without outcomes of interest. Experimental trials and non-original articles including comment, editorial, and letter to the editor were also excluded from our study. Any divergences on eligibility of studies were resolved by consulting a third investigator (G.-M.S.).

Data extraction
Two investigators (Y.-H.D. and X.T.) extracted the following information from each study independently using a standardized Excel (Microsoft Corporation) file: first author, year of publication, number of patients, surgical approach, details of chewing gum and SPCPs, demographic characteristics, and outcomes. When we found duplicate reports of the same study in preliminary abstracts and articles, we analyzed data from the most complete dataset. Discrepancies were resolved by discussion between the two investigators.

Outcome variables and definitions
The primary outcomes were time to first flatus (defined by authors of individual trials), time to first bowel movement (defined by authors of individual trials), time to first bowel sounds (defined by authors of individual trials), time to first feeding (defined by authors of individual trials), and LOS (defined as the time from admission to surgical care unit to hospital discharge or death). Secondary outcomes included: overall complications, other complications including pulmonary infarction, cholecystitis and delirium, postoperative ileus (defined as lack of passage of flatus or stool and intolerance to oral intake for at least 24 h), nausea, vomiting, abdominal distension, overall infections, wound infection, other infections, bleeding, wound dehiscence, anastomotic leak, complications related to GC, readmission rate, reoperation rate, mortality, tolerability of chewing gum, economic effect, and QoL. All secondary outcomes were defined by authors of individual trials apart from postoperative ileus.

Assessment of risk of bias
Two investigators (J.-G.Z. and X.T.) adopted the Cochrane risk of bias tool to appraise independently risk of bias [46,47]. The following each item described in this tool was valued to be 'low', 'unclear', or 'high' based on the matching level between extracted information and assessment criteria: random sequence generation; allocation concealment; blinding of participants and personnel; blinding of outcome assessment; incomplete outcome data; selective reporting; and other bias. As dictated by the Cochrane method, trials were rated to be low risk of bias when all key domains are valued low, while trials were rated to be high risk of bias when any one or more key domains are valued high. Otherwise, trials were rated to be unclear risk of bias. The consensus principle was used to resolve any discrepancies.

Statistical analysis
We calculated risk ratio (RR) with 95 per cent c.i. and weighted mean difference (WMD) with 95 per cent c.i. to present the dichotomous and continuous data respectively. For continuous data, we also used the median values to perform meta-analysis when mean values were not available [48]. We calculated the Cochrane Q to describe qualitatively the heterogeneity, and I2 was used to quantitate it; a value of over 50 per cent indicated significant heterogeneity [49]. Random-effects model was used to perform all analyses regardless of heterogeneity in the present study. We also conducted sensitive analysis by excluding abstract and study with less than 20 participants per arm respectively for time to first flatus, time to first bowel movement and length of hospital stay. Subgroup analysis was also designed to investigate the effects for different surgical approaches including laparoscopic colectomy, open colectomy, and hybrid method. The funnel plot was drawn to examine the publication bias [50]. P < 0.05 indicated statistical significance. We completed all statistical analyses using RevMan 5.3 (The North Cochrane Centre, Copenhagen, Denmark).

Trial sequential analysis
The risk of yielding spurious statistical inferences in a cumulative meta-analysis was increasing result from repeated significance testing on sparse and accumulated data [51]. The trial sequential analysis, which is comparable to interim analysis in a single trial, was therefore used to examine whether the accumulative evidence was sufficient power to draw a firm conclusion that is an intervention yielded anticipated effect before required information size was accrued, and thus determine whether the trial should be terminated early [52][53][54][55]. Construction of sequential monitory boundary and calculation of required information size were at the core of performing trial sequential analysis [56]. We concluded that further studies were not needed if the trial sequential analysis boundary or the futility zone is crossed [52,55].
A false positive error of 0.05, a false negative error of 0.20 (corresponding to power of 80 per cent), and an anticipated risk ratio reduction of 20 per cent were used to conduct trial sequential analysis in the present study. For binary outcomes, a control event proportion was obtained from the result of the meta-analysis. For continuous outcomes, the mean difference, variance and diversity were estimated empirically based on all eligible trials entered into software. The analyses were done with trial sequential analysis version 0.9 beta (www.ctu.dk/tsa) [55].

Quality of evidence
Grading of recommendations assessment, development and evaluation (GRADE) method was used to rate the evidence in order to facilitate decision-making [57]. In this method, the evidence from RCTs was firstly established to has high quality and five down-grading factors including risk of bias, imprecision, indirectness, inconsistency, and publication bias can reduce the level to moderate, low and very low [57,58].

CONCLUSIONS
In patients undergoing colorectal surgery, implementation of CG have the potential to enhance resumption of gastrointestinal function through decreased the time to first flatus, bowel movement, and bowel sounds, and shorten LOS. However, whether CG may reduce risk of postoperative ileus would require to further study because insufficient evidences were accrued and objective endpoints such as gastrin should be applied. Because sufficient information sizes have confirmed effect of CG on time to first flatus, time to first bowel movement, time to first bowel sounds, and LOS, and thus more studies should be planned to investigate the safety www.impactjournals.com/oncotarget of CG. Furthermore, further studies focusing on optimal quantity, frequency, and duration of consuming CG should also be planned.