Meta-analysis of randomized clinical trials of colorectal surgery with or without mechanical bowel preparation

Transcription

1 Meta-analysis Meta-analysis of randomized clinical trials of colorectal surgery with or without mechanical bowel preparation K. Slim 1,E.Vicaut 2,Y.Panis 3 and J. Chipponi 1 1 Department of General and Digestive Surgery, Hôtel-Dieu, Clermont-Ferrand, 2 Department of Clinical Research, Hôpital F. Widal, Paris and 3 Department of General and Digestive Surgery, Hôpital Lariboisiere, Paris, France Correspondence to: Dr K. Slim, Department of General and Digestive Surgery, Hôtel-Dieu, Boulevard Leon Malfreyt, F Clermont-Ferrand, France ( kslim@chu-clermontferrand.fr) Background: Mechanical bowel preparation is used routinely before colorectal surgery, but some randomized clinical trials have suggested that it is of no benefit. This study assesses whether such bowel preparation may safely be omitted before elective colorectal surgery. Methods: A search of the literature was performed; the inclusion criteria were randomized clinical trials comparing bowel preparation with no preparation in colorectal surgery. The methodological quality of included trials was assessed. The primary outcome was anastomotic leakage; secondary outcomes were other septic complications. The meta-analysis was conducted using the Peto one-step method. Results: Eleven trials were retrieved, of which seven, containing 1454 patients, were included in the meta-analysis. There was no heterogeneity between the trials. Significantly more anastomotic leakage was found after mechanical bowel preparation (5 6 versus 3 2 per cent; odds ratio 1 75 (95 per cent confidence interval 1 05 to 2 90); P = 0 032). All other endpoints (wound infection, other septic complications and non-septic complications) also favoured the no-preparation regimen, but the differences were not statistically significant. Sensitivity analysis showed that these results were similar when trials of poor quality were excluded. Subgroup analysis showed that anastomotic leakage was significantly greater after bowel preparation with polyethylene glycol () compared with no preparation, but not after other types of preparation. Conclusion: There is good evidence to suggest that mechanical bowel preparation using should be omitted before elective colorectal surgery. bowel preparations should be evaluated by further large randomized trials. Presented to the 18th Congress of the French Society of Digestive Surgery, Lyons, France, December 2003 Paper accepted 8 April 2004 Published online in Wiley InterScience ( DOI: /bjs.4651 Introduction Until recently, preparation and cleansing of the bowel before colorectal surgery was a major surgical dogma. The most recent survey in the USA showed that more than 99 per cent of colorectal surgeons routinely employed mechanical bowel preparation 1, one-third of them using polyethylene glycol () exclusively. The main reason for this practice is a belief that postoperative morbidity is related to septic bowel content 2. Nevertheless a randomized trial over 30 years ago questioned this dogma, suggesting that vigorous mechanical bowel preparation was unnecessary 3. Several randomized clinical trials have been published since then to assess the omission of preoperative bowel preparation, but all have been underpowered 4. Two meta-analyses of randomized trials have also been published previously 5,6. The first, by Platell and Hall 5, included three trials with 497 patients, and the second, by Guenaga et al. 6, included six trials with 1159 patients. Since these publications, some abstracts have been published as full articles and further randomized trials have become available. The greater the number of patients in a metaanalysis, the greater is its power to detect a possible treatment effect, and so it is appropriate to perform a new analysis that includes all the information currently available on this subject. The present meta-analysis aims to assess whether mechanical bowel preparation should be Copyright 2004 British Journal of Surgery Society Ltd British Journal of Surgery 2004; 91:

2 1126 K. Slim, E. Vicaut, Y. Panis and J. Chipponi omitted before colorectal surgery by analysing published data in randomized clinical trials that compared bowel preparation before colorectal surgery with no preparation. Methods Search strategy and selection of trials This study is a systematic review of published data. The search was performed in March 2003 using the following electronic databases without years or language restrictions: Cochrane Library, CRD database, Embase, Lilacs, Medical Editors Trials Amnesty, Medline, Pascal Biomed and Scisearch. A manual search was made of the last six issues of major surgical journals (American Journal of Surgery, Annals of Surgery, Archives of Surgery, British Journal of Surgery, Diseases of the Colon and Rectum, Journal of the Americal College of Surgery and Surgery) and of the proceedings of major meetings from 1995 (Annual Meeting of the American Society of Colon and Rectal Surgeons, Annual Congress of the American College of Surgeons, Annual Congress of the American Gastroenterological Association, Annual Congress of the Association of Coloproctology of Great Britain and Ireland, and the European Gastroenterological Week). In addition a questionnaire was sent to French experts in colorectal surgery asking them whether they were aware of trials in this field. Finally, the references lists of published papers were checked. Two independent reviewers selected the trials to be included by reading the abstracts. Only studies designed as randomized trials comparing bowel preparation with no preparation before elective colorectal surgery were included. Critical assessment of trials and collection of data The methodological quality of the included trials was assessed by means of a validated scale 7 by two independent reviewers; when there was discordance the final score was established by consensus. Briefly, the quality scale involved three items: randomization, double-blinding, and withdrawals or dropout. The score ranges from 0 to 5 : 0 2 for randomization, 0 2 for blinding, and 0 1 for withdrawals or dropout. In this meta-analysis double-blinding was not feasible and single-blinding was considered appropriate. According to Moher et al. 7,the methodological quality of a trial is considered poor when the score is 2 or less. Data were collected independently by the two reviewers and cross-checked. When data were missing, particularly in abstracts of proceedings, the authors of trials published in abstract form were contacted to obtain any unpublished data. The following data were collected for each trial: type of randomization, sample size, blinding, withdrawals and dropout, type of bowel preparation (anterograde or oral versus retrograde or enema), the solution used (polyethylene glycol (), mannitol, sodium picophosphate), antibiotics given, type of colonic or rectal procedure, level of the anastomosis (peritoneal versus infraperitoneal), definition of anastomotic leakage (assessed objectively or not), other postoperative complications and postoperative mortality. Confounding factors, such as intraoperative peritoneal contamination (spillage of liquid bowel content), surgical technique (stapled or handsewn anastomosis) and abdominal drainage, were also recorded. Outcome measurements and sensitivity analysis Data were collected independently by the two reviewers and cross-checked. Outcomes were analysed on an intention-to-treat basis. The primary outcome was overall anastomotic leakage. Secondary outcomes were abdominal septic complications (peritonitis, pelvic abscess, reoperation, wound abscess, wound dehiscence, diarrhoea), extra-abdominal septic complications (bronchopulmonary complications, urinary infection) and other non-septic complications. Surgical site infections were considered as abdominal septic complications. Sensitivity analysis assessing the robustness of the results included heterogeneity analysis and subgroups analyses (poor-quality versus good-quality trials, abstract versus full article papers, versus othersolutions,peritonealversus infraperitoneal anastomosis). Statistical analysis Overall estimates of treatment effect and their 95 per cent confidence intervals (c.i.) were calculated using Peto onestep estimators for fixed effects (odds ratio; OR). The Q estimator was calculated to test heterogeneity between studies. Where possible the impact of a moderator variable was assessed by ANOVA. All calculations were made using comprehensive meta-analysis software from Biostat Inc. (Engelwood, New Jersey, USA). Results Characteristics of trials Eleven trials 3,8 17 were retrieved, of which three were excluded for the following reasons. In an early trial 3 including 97 patients, results were reported in insufficient

3 Meta-analysis of mechanical bowel preparation 1127 detail (no clear definition of surgical complications), making formal analysis impossible. In another trial 8 patients in the control group received total parenteral nutrition and were fasted for 7 days, and this cannot be considered as a no preparation regimen according to the commonly accepted definition. In the third trial 9 it was not possible to separate data relating to bowel preparation as two randomizations had been used: antibiotics versus placebo with low-residue preparation versus elemental diet. One trial was published twice 14,15 but has been considered only once, so that seven trials were eventually included, containing a total of 1454 patients: 720 who had mechanical bowel preparation and 734 who had no preparation. The characteristics of these trials are summarized in Table 1. Two trials 10,17 were published in abstract form, and so the authors were contacted. The authors of the earlier trial 10 responded that the data were no longer available, and the authors of the later trial 17 did not provide additional detailed data. None of the authors reported a calculation of sample size. In all included studies the patients received prophylactic antibiotics. All but two studies 10,17 (published in abstract form) reported that the groups were similar with regard to age, sex, diagnosis and type of procedure. Elective colorectal surgery was an inclusion criterion in all studies; one trial 11 included children and two 11,14,15 included patients without anastomosis. All trials except one stipulated no rectal cleansing in the no-preparation group; in one trial 16 patients undergoing rectal surgery were given a fleet enema on the day of operation. In all trials, confirmation of anastomotic leakage was obtained only when a leak was suspected clinically; no trial reported routine radiographic examination. The funnel plot (not shown) for the main endpoint of anastomotic leakage was symmetrical. The Q test did not show any heterogeneity: The Q value (Peto one-step method) for fixed values was 4 41 (P = 0 660). As can be seen below, for all outcomes considered there was no analysis that led to a conclusion in favour of bowel preparation. Meta-analysis of anastomotic leakage All trials reported the rate of anastomotic leakage. There was significantly more leakage in the bowel preparation group (5 6 versus 3 2 per cent; OR 1 74 (95 per cent c.i to 2 90); P = 0 032). The data from the seven trials and the meta-analysis are shown in Fig. 1. Table 1 Characteristics of randomized trials included in the meta-analysis of bowel preparation versus no preparation Reference Preparation (n = 720) No preparation (n = 734) Quality score* Exclusion criteria Malignancy (%) Solution tested Diet Infraperitoneal anastomosis Follow-up (days) No anastomosis 92 n.a. n.a. n.a Previous antibiotics 46 Mannitol Low residue 48 versus No anastomosis 79 Pico Normal 39 versus Colon not opened 46 Normal 54 versus , No exclusion 50 Mannitol Normal n.a No anastomosis or stoma 78 Normal 50 versus Normal *Quality score includes three items: randomization, blinding, and withdrawals or dropout. In groups having no preparation. Preparation versus no preparation., polyethylene glycol; pico, sodium picophosphate; n.a., not available Reference , Combined (7) Preparation 8 of 67 7 of 72 3 of 82 5 of of 30 7 of of of 701 Control 1 of 67 4 of 77 4 of 87 3 of of 30 4 of of of wt. OR (1 36, 20 14) 1 93 (0 57, 6 57) 0 79 (0 18, 3 58) 1 56 (0 38, 6 36) 1 99 (0 20, 19 94) 1 81 (0 55, 5 99) 1 18 (0 39, 3 58) 1 75 (1 05, 2 90) Favours preparation Favours controls Fig. 1 Meta-analysis of the effect of bowel preparation versus no preparation on anastomotic leakage. The diamond denotes the pooled odds ratio (OR) with the fixed-effects model. Values in parentheses are 95 per cent confidence intervals. The OR significantly favours no preparation (P = 0 032)

4 1128 K. Slim, E. Vicaut, Y. Panis and J. Chipponi Meta-analysis of other septic complications The rate of wound infection was reported in all trials and was higher after bowel preparation, although not significantly so (7 4 versus 5 7 per cent; OR 1 33 (95 per cent c.i to 2 03); P = 0 175). The Q test did not show any heterogeneity (Q = 3 31, P = 0 770). When other extra-abdominal septic complications were considered (three trials 12,13,15 ), the analysis still favoured the no-preparation regimen, although the difference was not statistically significant (10 7 versus 9 6 per cent; OR 1 12 (95 per cent c.i to 1 83); P = 0 631). Confounding factors (peritoneal septic contamination, technique of anastomosis and drainage) were not reported in three trials 10,14,15,17. These factors were similar in both groups in three trials The only trial 16 that reported more spillage of liquid bowel content after (16 6 versus 9 3 per cent; P = 0 040) showed a similar rate of septic complications in the preparation and no-preparation groups (10 2 versus 8 8 per cent respectively; P = 0 720). Meta-analysis of other endpoints Postoperative mortality was reported in three trials 11,12,16 ; the mortality rate was higher in the bowel preparation group, but the difference was not statistically significant (1 4 versus 0 8 per cent; OR 1 56 (95 per cent c.i to 5 45)). The reoperation rate (four trials ) was also higher, although not significantly so, in the preparation group (4 3 versus 2 7 per cent; OR 1 56 (0 68 to 3 59)). Finally, the rate of non-septic complications (four trials 12 14,16 ) was higher with the bowel preparation, but again the difference was not significant (11 9 versus 10 per cent; OR 1 23 (0 81 to 1 89); P = 0 334). Effect Reference Preparation Control OR (3) 12 14, of 82 1 of of of of 86 9 of of of 87 2 of 30 9 of of of 93 7 of of (0 32, 6 47) 0 50 (0 05, 5 02) 2 28 (0 98, 5 29) 1 79 (0 89, 3 61) 0 76 (0 24, 2 45) 1 31 (0 48, 3 59) 1 56 (0 38, 6 36) of of (0 49, 2 63) (4) 31 of of (0 67, 1 92) Combined (7) 53 of of 734 Favours preparation Favours controls 1 34 (0 88, 2 04) Fig. 2 Meta-analysis of the effect of bowel preparation versus no preparation on wound infection. The diamonds denote the pooled odds ratios (ORs) with the fixed-effects model. Values in parentheses are 95 per cent confidence intervals. The top three trials used preparation regimens other than polyethylene glycol (): sodium picophosphate 11 and mannitol 12,14,15. The OR favours no preparation, but not significantly so (P = 0 175) Effect (3) Reference 12 14, Preparation 3 of 82 2 of 30 7 of of of 67 7 of of 138 Control 4 of 87 1 of 30 4 of 77 9 of of 67 6 of of OR 0 79 (0 18, 3 58) 1 99 (0 20, 19 94) 1 93 (0 57, 6 57) 1 43 (0 59, 3 45) 5 23 (1 36, 20 14) 1 18 (0 39, 3 58) 1 56 (0 38, 6 36) 16 7 of of (0 55, 5 99) (4) 27 of of (1 03, 3 60) Combined (7) 39 of of 708 Favours preparation Favours controls 1 75 (1 05, 2 90) Fig. 3 Meta-analysis of the effect of bowel preparation versus no preparation on anastomotic leakage. The diamonds denote pooled odds ratios (ORs) with the fixed-effects model. The top three trials used preparation regimens other than polyethylene glycol (): sodium picophosphate 11 and mannitol 12,14,15. The OR significantly favours no preparation when compared with (P = 0 039), but not in comparison with the other regimens (P = 0 422)

5 Meta-analysis of mechanical bowel preparation 1129 Sensitivity analysis Two trials 12,13 stratified anastomotic leakage according to level of anastomosis, but the subgroup of patients with an infraperitoneal anastomosis was too small to allow a formal statistical analysis (48 patients in the bowel preparation group and 50 with no preparation). When the poor-quality trials 10,17 (quality score of 2 in Table 1) were excluded from the meta-analysis, the results did not change substantially and still favoured the no-preparation regimen. The rate of wound infection was higher in the bowel preparation group, but the difference was not statistically significant (OR 1 49 (95 per cent c.i to 2 47); P = 0 000). The rate of anastomotic leakage was also higher in the preparation group but, in contrast to the finding when all trials were analysed, the difference was no longer statistically significant (OR 1 55 (0 82 to 2 92); P = 0 000). When trials assessing preparation with were considered separately from the others, the rate of wound infection showed no substantial change (Fig. 2). However, the anastomotic leak rate was significantly higher after preparation than after no preparation (OR 1 92 (95 per cent c.i to 3 60); P = 0 039) (Fig. 3). Although the results still favoured no preparation over the other bowel preparation regimens, the difference was not statistically significant (OR 1 43 (0 59 to 3 45); P = 0 422). Discussion This meta-analysis clearly calls into question the value of preoperative bowel preparation in colorectal surgery; it has included the largest number of patients yet available and provides stronger evidence than previous studies 5,6.No result favoured the bowel preparation regimen; indeed, the main finding was that bowel preparation may be detrimental by increasing the risk of anastomotic leakage. It should also be appreciated that the data used in this meta-analysis appear to be comprehensive, even though three trials had to be excluded. Exclusion of these trials did not alter the results of the present meta-analysis, as the funnel plot applied to the seven studies showed that there was no publication bias. As the majority of the trials evaluated the use of as bowel preparation, it could be considered that the regimen is the factor actually being studied in this metaanalysis, rather than bowel preparation itself. Indeed, the only significant effect related to preparation. Still, the results of the other regimens in all cases favoured no preparation, and the lack of statistical significance may well be related to lack of power (378 patients in three trials). Having said this, it is probably unwise for the conclusions of the present meta-analysis to be extrapolated to all types of bowel preparation because only three regimens were actually evaluated in this study: in four trials, sodium picophosphate in one, and mannitol (now abandoned by most surgeons because of side-effects) in two. preparation regimens, such as senna (together with a povidone iodine-based enema), have been shown in randomized trials to be associated with a postoperative septic complication rate similar to that of. In one trial 18 the anastomotic leak rate was 5 3 per cent after senna and 5 7 per cent after, which is close to the rate of 5 6 per cent observed in the present meta-analysis. Similarly, two trials 19,20 compared sodium phosphate with and showed similar complication rates. Both of the above regimens are better tolerated by the patients than. Although it may be hypothesized that no preparation may also be superior to senna and sodium phosphate, such an assumption would need to be verified by further large randomized trials. For the present, the main conclusion of this meta-analysis is that no preparation appears to be superior to preparation, and that is probably detrimental (Fig. 3). Do the present findings apply to rectal surgery? This question cannot be answered because stratification according to the level of anastomosis was performed in only two trials 11,12, including few patients. With such insufficient data, the value of rectal preparation using, for example, an enema requires a new trial. Yet another point is that all the trials included here involved open colorectal procedures; the findings of this meta-analysis cannot properly be extrapolated to laparoscopic surgery, although in practice dilated bowel (as it usually is following bowel preparation) can hamper laparoscopic vision and make mobilization of the gut more difficult. Finally, in most of the included trials the no-preparation arm involved a normal diet. From a practical viewpoint, no preparation of the colon and allowing the patient to eat normally would comprise a more patient friendly approach. The authors of this meta-analysis agree with Hughes 3, who wrote 30 years ago that Omission of enemas and bowel washes from the preoperative procedures will be welcomed by both patients and nursing staff. References 1 ZmoraO,WexnerSD,HajjarL,ParkT,EfronJE, Nogueras JJ et al. Trends in preparation for colorectal surgery: survey of the members of the American Society of Colon and Rectal Surgeons. Am Surg 2003; 69: Nichols RL, Condon RE. Preoperative preparation of the colon. Surg Gynecol Obstet 1971; 132:

6 1130 K. Slim, E. Vicaut, Y. Panis and J. Chipponi 3 Hughes ES. Asepsis in large-bowel surgery. Ann R Coll Surg Engl 1972; 51: Slim K, Panis Y, Chipponi J. Mechanical colonic preparation for surgery or how surgeons fight the wrong battle. Gastroenterol Clin Biol 2002; 26: Platell C, Hall J. What is the role of mechanical bowel preparation in patients undergoing colorectal surgery? Dis Colon Rectum 1998; 41: Guenaga KF, Matos D, Castro AA, Atallah AN, Wille-Jorgensen P. Mechanical bowel preparation for elective colorectal surgery. Cochrane Database Syst Rev 2003; (2)CD Moher D, Pham B, Jones A, Cook DJ, Jadad AR, Moher M et al. Does quality of reports of randomised trials affect estimates of intervention efficacy reported in meta-analyses? Lancet 1998; 352: Makino M, Hisamitsu K, Sugamura K, Kimura O, Kaibara N. Randomized comparison of two preoperative methods for preparation of the colon: oral administration of a solution of polyethylene glycol plus electrolytes and total parenteral nutrition. Hepatogastroenterology 1998; 45: Matheson DM, Arabi Y, Baxter-Smith D, Alexander-Williams J, Keighley MR. Randomized multicentre trial of oral bowel preparation and antimicrobials for elective colorectal operations. Br J Surg 1978; 65: Brownson P, Jenkins SA, Nott D, Ellenbogen S. Mechanical bowel preparation before colorectal surgery: results of a prospective randomized trial. Br J Surg 1992; 79: (Abstract). 11 Santos JCM Jr, Batista J, Sirimarco MT, Guimaraes AS, Levy CE. Prospective randomized trial of mechanical bowel preparation in patients undergoing elective colorectal surgery. Br J Surg 1994; 81: Burke P, Mealy K, Gillen P, Joyce W, Traynor O, Hyland J. Requirement for bowel preparation in colorectal surgery. Br JSurg1994; 81: Miettinen RPJ, Laitinen ST, Mäkelä JT,Pääkkönen ME. Bowel preparation with oral polyethylene glycol electrolyte solution vs. no preparation in elective open colorectal surgery: prospective, randomized study. Dis Colon Rectum 2000; 43: Fillmann EEP, Fillmann HS, Fillmann LS. Elective colorectal surgery without preparation. Revista Brasilieira de Coloproctologia 1995; 15: Fillmann LS, Perondi F, Fillmann HS, Fillmann EEP. The elective resection for colo-rectal cancer without mechanical bowel preparation. Revista Brasilieira de Coloproctologia 2001; 21: Zmora O, Mahajna A, Bar-Zakai B, Rosin D, Hershiko D, Shabtai M et al. Colon and rectal surgery without mechanical bowel preparation: a randomized prospective trial. Ann Surg 2003; 237: Fa-Si-Oen PR, Buitenweg JA, van Geldere D, de Waard JW, Swank X, Putter H et al. The effect of preoperative bowel preparation with polyethylene glycol on surgical outcome in elective open colorectal surgery a randomized multicentre trial. Abstract presented to the Fourth Belgian Surgical Week, Ostende, Valverde A, Hay JM, Fingerhut A, Boudet MJ, Petroni R, Pouliquen X et al. Senna versus polyethylene glycol for mechanical preparation the evening before elective colonic or rectal resection: a multicenter controlled trial. French Association for Surgical Research. Arch Surg 1999; 134: Oliveira L, Wexner SD, Daniel N, DeMarta D, Weiss EG, Nogueras JJ et al. Mechanical bowel preparation for elective colorectal surgery. A prospective, randomized, surgeon-blinded trial comparing sodium phosphate and polyethylene glycol-based oral lavage. Dis Colon Rectum 1997; 40: Neijenhuis PA, den Hollander SMR. Mechanical bowel preparation with sodium phosphate lavage solution in colorectal surgery. A patient friendly method. Eur J Gastroenterol Hepatol 1998; 10: A51 (Abstract).