Meta-Analysis of the Placebo Rates of Clinical Relapse and Severe Endoscopic Recurrence in Postoperative Crohn s Disease

Transcription

1 GASTROENTEROLOGY 2008;135: Meta-Analysis of the Placebo Rates of Clinical Relapse and Severe Endoscopic Recurrence in Postoperative Crohn s Disease SARA RENNA,* CALOGERO CAMMÀ,, IRENE MODESTO,* GIUSEPPE CABIBBO, DANIELA SCIMECA,* GIUSEPPE CIVITAVECCHIA,* FILIPPO MOCCIARO,* AMBROGIO ORLANDO,* MARCO ENEA, and MARIO COTTONE* *Dipartimento di Medicina, Pneumologia e Fisiologia della Nutrizione Umana, Università di Palermo; Cattedra e Unità Operativa di Gastroenterologia, University of Palermo, Italy; Consiglio Nazionale delle Ricerche, IBIM, Palermo, Italy; Dipartimento di Scienze Statistiche e Matematiche S. Vianelli, University of Palermo, Italy See Jones J et al on page 1218 in CGH; see CME quiz on page 1761 Backgrounds & Aims: The benefit of therapy for prevention of postoperative recurrence of Crohn s disease (CD) is limited. Clinical relapse and severe endoscopic recurrence are the main outcomes in the evaluation of trials on prevention of recurrence. The aim of this metaanalysis was to focus on knowledge of the placebo rates of relapse and recurrence in postoperative CD and to identify factors influencing these rates. Methods: We performed a meta-analysis of placebo-controlled, randomized clinical trials, evaluating therapies for postoperative maintenance of CD identified on MEDLINE from 1990 to Primary outcomes were clinical relapse and severe endoscopic recurrence. Results: The pooled estimate of the placebo relapse rate was 23.7% (95% confidence interval [CI], 13 35; range 0 78). There was a statistically significant heterogeneity among studies (P <.0001). Heterogeneity in clinical relapse was present even if the trials were stratified according to the time of outcome. The pooled estimate of the severe endoscopic recurrence rate was 50.2% (95% CI, 28 73; range, 30 79). There was significant heterogeneity among the studies (P.00038). This heterogeneity was less apparent in studies carried out within 12 months. The logistic analysis identified only duration of follow-up as a variable associated with different placebo relapse rates. No variable was identified as a predictor of a placebo endoscopic recurrence rate. Conclusions: There is significant heterogeneity among placebo rates in postoperative CD. No single design variable was identified that explained the heterogeneity in placebo outcomes for clinical or endoscopic recurrence. In patients with Crohn s disease (CD) operated on for resection, an accurate estimation of rate of relapse and recurrence among patients treated with placebo is essential for evaluating natural history, for assessing treatment effect size, and for both calculating sample size and interpreting results of clinical trials. 1,2 Knowledge of factors that influence the outcome among patients receiving placebo may be also important for designing clinical trials. In CD, recurrence after surgery may be defined as clinical, endoscopic, and surgical outcomes. 3 Recurrence rates vary depending on the outcome that has been chosen. Among studies evaluating drugs for the prevention of recurrence rates, clinical relapse, and (in the last few years) endoscopic recurrence have both been evaluated as main outcomes. Surgical recurrence has not been an end point in trials evaluating drugs for prevention of recurrence because of the relative rarity of this event, and because of the long follow-up needed to obtain reliable data. The rate of clinical relapse, defined as either a Crohn s Disease Activity Index (CDAI) of 150 or an increase from 60 to 100 points, and the proportion of severe endoscopic recurrence defined according to the Rutgeerts (score 2), 4 vary among the studies Different drugs have been evaluated for prevention of recurrence, having as outcome clinical relapse and severe endoscopic recurrence The rate of clinical relapse and severe endoscopic recurrence ranged in the placebo groups from 0% to 78%, and from 30% to 79%, respectively. Therefore, an interpretation of the results of the trials is problematic, with conflicting data, and at present there is no consensus on which is the better drug. This meta-analysis analyzes the variability in placebo response by looking at the heterogeneity among the studies, and aims at interpreting this variability. Materials and Methods Selection of Randomized Trials This analysis was performed in accordance with the QUOROM statement. 29 Retrieval of randomized controlled trials (RCTs) was based on the Cochrane Controlled Trials Register, the Cochrane Library, MEDLINE, and ENBASE, using the following medical subject headings: postoperative recurrence, Crohn s disease, 5-aminosalicylic acid, azathioprine, 6-MP, metronidazole, ornidazole, fish oil, probiotics, IL-10, budesonide, randomized or randomised trial, Abbreviations used in this paper: CI, confidence interval; RCT, randomized controlled trial; OR, odds ratio by the AGA Institute /08/$34.00 doi: /j.gastro

2 November 2008 RATES OF CLINICAL RELAPSE AND SEVERE ENDOSCOPIC RECURRENCE 1501 and clinical trial. The search included literature published through December The computer search was supplemented with manual searches for reference lists of all available review articles, primary studies, meeting abstracts, and bibliographies of books to identify other studies not found in the computer search. When the results of a single study were reported in 1 publication, only the most recent and complete data were included in the meta-analysis. Studies were included in the analysis if they were RCTs comparing all the above-mentioned drugs with placebo, if they had been published or accepted for publication as full-length papers or as abstracts, and if clinical relapse or endoscopic recurrence was assessed as an outcome measure of the effect of the treatment. We also included 1 trial in which the comparison with no treatment was made. 13 Quasi-randomized trials and observational studies were excluded. Decisions on which RCTs to include were made unblindly by 2 reviewers (G.C. and M.C.). Disagreements were resolved by discussion. Excluded RCTs were identified together with the reason for exclusion. Clinical relapse was defined as either a CDAI of 150 or as an increase from 60 to 100 points of baseline value. Severe endoscopic recurrence was defined as a Rutgeerts score of 2 on the colonoscopy performed during follow-up. Of the 198 studies reviewed, 16 RCTs met the inclusion criteria Studies were excluded if the crude rates of clinical relapse were not reported, 16,21 24 if the Rutgeerts score was not reported, and if they compared mesalamine with 4-aminosalicylic acid 25 or different drugs without placebo Review of the Trials The trials were first reviewed using a list of predefined, pertinent issues that concerned characteristics of patients and treatments. A set of quality criteria was also assessed for trials published as full papers. We used the quality criteria suggested by Nicolucci et al (Appendix 1). 30 Each RCT was evaluated and scored by 3 independent investigators who had not participated in the trials they evaluated. Discrepancies among reviewers were infrequent (overall interobserver variations 10%) and were resolved by discussion. For each trial, the total score was expressed as a percentage of the maximum achievable score. Statistical Analyses Pooled estimates of the placebo clinical relapse and severe endoscopic recurrence rates were calculated using random-effects logistic regression analysis after applying sample weights according to the placebo sample size, as implemented using SAS version 8.1 (SAS Institute, Cary, NC) software, PROC NLMIXED command. The assumption of heterogeneity implied by the utilization of random-effect models is justified by the differences in patients features and study characteristics. Heterogeneity among studies was assessed with the Pearson 2 test. A recommended approach to dealing with heterogeneity is sorting the heterogeneous group of studies into subgroups according to a stratifying variable suspected of having caused the inconsistency. Therefore, stratum-specific rates of the placebo clinical relapse and severe endoscopic recurrence rates for different patientand study-level covariates were calculated. We used 9 stratifying variables: mean patient age, proportion of males, study sample sizes, length of follow-up, disease duration, number of participating centers, year of study publication, proportion of patients with ileal disease, and proportion of patients who underwent first resection. Only univariate logistic regression models were used to examine the association between study design and the placebo relapse or recurrence rates owing to the limited number of studies using the same criteria for response. The correlation between continuous measures of study characteristics was assessed with the Pearson correlation coefficient. We did not consider multivariable analysis because of the wide heterogeneity and lack of complete data for identification of possible variables that could explain heterogeneity. Publication bias was assessed by the Egger regression symmetry test for publication bias. 31 For all analyses, P.05 was considered statistically significant. All analyses were completed with SAS version 8.1 (SAS Institute) software. Results Description of the Studies After review of the titles and abstracts, 16 articles 5 20 (15 full-length papers 5 13,15 20 and 1 abstract 14 ), fulfilled the inclusion criteria and were selected for review. All the studies included in the meta-analysis were double-blind, placebo-controlled trials, except one. 13 All but 2 6,8 were multicenter, with the number of centers ranging from 2 to 31. A total of 1664 patients were included in these RCTs, 799 of whom received nonactive treatment. The sample size of the placebo group of each RCT varied greatly, ranging from 22 to 166 patients. All but 1 18 of the RCTs considered only patients who were in remission after intestinal resection; in 1 trial, 18 data on patients with both surgically and medically induced remission were reported, but it was possible to extract some data on patients who were operated on. The distribution of the main characteristics of patients in the placebo arm of the 16 RCTs 5 20 considered in the present analysis is reported in Table 1. Data on dropout patients are reported in ,14 17,19,20 of the 16 RCTs considered in this meta-analysis; 114 of 799 placebo-arm patients were withdrawn from the RCTs, with a percentage of dropout among all the trials ranging from 3% to 67%.

3 1502 RENNA ET AL GASTROENTEROLOGY Vol. 135, No. 5 Table 1. Characteristics of Patients Treated With Placebo in Trials for Prevention of Postoperative Relapse and Severe Endoscopic Recurrence in CD Study Comparison drug Year Follow-Up (mos) Patients (placebo) Mean age Male (%) Clin R End R Severe end R Disease duration (mos) Ileo, n (%) Caprilli et al 13 Mesalazine *** Yes Yes Yes 28 NA Brignola et al 7 Mesalazine Yes Yes Yes (55) McLeod et al 19 Mesalazine Yes Yes No NA 16 (21) Rutgeerts et al 6 Antibiotics NA Yes Yes Yes 120 NA Florent et al 16 Mesalazine No Yes Yes 68 NA Belluzzi et al 14-3 Fatty acids NA NA Yes Yes Yes NA NA Sutherland et al 18 Mesalazine # 55 # Yes No No NA 49% # Hellers et al 12 Steroids Yes Yes Yes NA NA Ewe et al 20 Steroids Yes Yes No 81 9 (22) Lochs et al 17 Mesalazine Yes Yes Yes (42) Colombel et al 9 IL Yes Yes Yes (100) Prantera et al 8 Probiotics Yes Yes Yes (86) Hanauer et al 10 Mesalazine/6-MP Yes Yes Yes 127 NA Marteau et al 11 Probiotics Yes Yes Yes (50) Rutgeerts et al 5 Antibiotics Yes Yes Yes 36 NA Van Gossum et al 15 Probiotics Yes Yes Yes NA 4 (11) NA, not available; NO, not obtainable. *Indication available, although for some patients 1 indication may be present. **Data on perforating pattern only. ***No treatment used as comparison. # Data reported are cumulative (medical and surgical). Of the 16 RCTs, ,17,19,20 reported data on both clinical and endoscopic recurrence, whereas 1 16 reported only data on endoscopic recurrence, and 1 18 only the data on clinical relapse. For the analysis of clinical recurrence we considered 15 RCTs 5 15,17 20 ; in 10 RCTs, 5,7,8,11 13,15,17,18,20 clinical relapse was assessed using the CDAI score (5 of them 7,8,13,15,18 defined clinical relapse as a CDAI of 150, 3 11,12,20 as a CDAI of 200, and 2 5,17 as a CDAI of 250). In 5 RCTs, 6,9,10,14,19 the definition of clinical relapse was the occurrence of symptoms compatible with CD (Table 2). The length of follow-up differed among the trials, ranging from 3 to 72 months. We excluded from the analysis 2 19,20 of the 15 RCTs that reported data on endoscopic recurrence, because 1 of these 19 reported cumulative data on both endoscopic and radiologic rate of recurrence, and 1 20 reported cumulative data on clinical and endoscopic recurrence, making it impossible to extract data on the endoscopic recurrence. For an analysis of severe endoscopic recurrence, 13 RCTs, 5 17 with a total of 648 patients in the placebo arm, were considered (Table 3). In all of these trials, endoscopic recurrence was assessed using the Rutgeerts endoscopic scoring system; in 8 RCTs 5,8,9,11,12,15 17 severe endoscopic recurrence was defined as an endoscopic score of 2; in 5, 6,7,10,13,14 severe endoscopic recurrence was defined as a score 2. The time of the endoscopic assessment of recurrence differed among the trials, ranging from 3 to 36 months after surgery. In the Rutgeerts studies, 5,6 endoscopic recurrence was evaluated after 3, 12, and 36 months. Nine studies 7,8,9,11,15,17 20 reported data on the site of the disease, and these data were lacking in 7 studies. 5,6,10,12 14,16 Only 7 5,7 9,11,15,16 of the 16 RCTs reported data on smoking habit. Data on indication for surgery were lacking in 9 studies. 7,9,10,13,14,16 18,20 Data on disease duration, mean age, and percentage of males in the placebo population were uniform among all the trials. The overall mean age was approximately 33 years (range, 29 37), the mean percentage of males was 48% (range, 32% 68%), and the mean disease duration was 71 months (range, ). Methodologic quality scores ranged from 60% to 95% on a scale of 0% 100% (Appendix 2). With regard to the quality of the studies, all trials reported a statement on randomization and double blinding, although the blinding method was not reported in 1 trial. 13 The randomization method was not described in 7 trials. 5 7,9,13,16,20 The description of dropouts was reported in 14 trials, 5 12,14 17,19,20 but in 4 trials 15% of the patients were lost to follow-up, 10,12,15,20 and in 2 trials 13,18 the information was not reported. In 6 studies, 5 7,9,15,16 information on compliance was not reported. A sample size calculation was reported in all 5,6,9 20 but 2 trials. 7,8 All the trials 5 20 reported detailed information on response evaluation, side effects, and complications. Placebo Relapse Rate The pooled estimate of the placebo relapse rate was 23.7% (95% confidence interval [CI], 13 35; range, 0 78). There was a statistically significant heterogeneity among studies (P.0001; Figure 1). Logistic regression analysis was used to identify potential sources of heterogeneity among the studies. Using the univariate logistic regression, the only factor asso-

4 November 2008 RATES OF CLINICAL RELAPSE AND SEVERE ENDOSCOPIC RECURRENCE 1503 Table 1. Continued Ileocolon, n (%) Colon, n (%) Dropout (%) No. of centers Indications for surgery (%) Steno Fist Intr First vs. 1 resection (%) Smokers (%) NA NA NA 15 NO* NO* NO* NA NA 19 (44) NA 4 (93) 8 NA NA NA 75/ (30) 37 (49) 8 (10) NA NA NA NA 2 (7) /50 NA NA NA 8 (13) 12 NO* NO* NO* 40/60 57 NA NA 4 3 NA NA NA NA NA 50 NA NA 31 NA NA NA NA NA NA NA 18 (27) /74 NA 24 (60) 8 (20) 17 (42) 3 NO* NO* NO* 33/67 NA 89 (54) 8 (5) 5 (3) 29 NO* NO* NO* NA NA NA 3 (13) 1 (4) 7 NO* NO* NO* NA 18 2 (9) 1 (4) 3 (13) /27 27 NA NA 27 (67) 5 ** ** ** NA NA 24 (48) 1 (2) 3 (6) /18 26 NA NA 5 (12) NA 77/ (83) 2 (6) 9 (25) /31 33 ciated with the placebo relapse rate was study duration (odds ratio [OR], 0.28; 95% CI, ). To examine whether there were differences within each stratum of relevant study features, we next calculated pooled estimates of the placebo relapse rate within each stratum and evaluated heterogeneity among strata. Heterogeneity was less evident in studies in which the year of publication was before 2000, in which the percentage of first resection was 50%, and in which the CDAI was 200 (Table 4). Heterogeneity persisted in the 2 strata as far as study duration, disease duration, and number of centers are concerned. Placebo Severe Endoscopic Recurrence Rate The pooled estimate of severe recurrence rate was 50.2% (95% CI, 28 73; range, 30 79). There was statistically significant heterogeneity among the studies (P.00038; Figure 2). At univariate analysis, no variable was identified as a predictor of placebo endoscopic recurrence rate. Table 2. Clinical Data and Clinical Relapse Rate in Placebo Arm of Postoperative Trials in Crohn s Disease Study Follow-Up (mos) Comparison drug Sample size (n) CDAI Clinical relapse (n) Placebo rate (%) Dropout (n) Clinical relapse without dropout (n) Placebo rate (%) with dropout Colombel et al 9 3 IL * Van Gossum et al 15 3 Probiotics Marteau et al 11 6 Probiotics Brignola et al 7 12 Mesalazine Sutherland et al Mesalazine NA NA NA Ewe et al Steroids Hellers et al Steroids Prantera et al 8 12 Probiotics Belluzzi et al Omega-3 fatty acids 24 * Rutgeerts et al 5 12 Antibiotics Lochs et al Mesalazine Hanauer et al Mesalazine/6MP 40 * NA NA Rutgeerts et al 6 36 Antibiotics 30 * Caprilli et al Mesalazine ** NA NA NA McLeod et al Mesalazine 76 * CDAI, Crohn s Disease Activity Index; IL, interleukin; NA, not available. *Clinical relapse was assessed by occurrence of symptoms compatible with CD. **In this study, no treatment was used as comparison.

5 1504 RENNA ET AL GASTROENTEROLOGY Vol. 135, No. 5 Table 3. Clinical Data and Severe Endoscopic Recurrence Rate in Placebo Arm of Postoperative Trials in Crohn s Disease Study (Rutgeerts score) Follow-Up (mos) Comparison drug Sample size (n) No. of patients undergoing endoscopy** Severe endoscopic recurrence (n)/ Severe recurrence placebo rate among all sample sizes (%) Severe endoscopic recurrence rate in those undergoing endoscopy Van Gossum et al ( 2) 15 3 Probiotics Colombel et al ( 2) 9 3 Iinterleukin Rutgeerts et al ( 2) 6 3 Antibiotics Rutgeerts et al ( 2) 5 3 Antibiotics Florent et al ( 2) 16 3 Mesalazine Marteau et al ( 2) 11 6 Probiotics Prantera et al ( 2) 8 12 Probiotics Brignola et al ( 2) 7 12 Mesalazine Hellers et al ( 2) Steroids Belluzzi et al ( 2) Omega-3 fatty acids Rutgeerts et al ( 2) 5 12 Antibiotics Lochs et al ( 2) Mesalazine Hanauer et al ( 2) Mesalazine/azathioprine Caprilli et al ( 2) Mesalazine Rutgeerts* et al ( 2) 6 36 Antibiotics *Results reported at 3 months and at the end of follow-up. Analysis only on data of the last follow-up. **The number of patients undergoing endoscopy was calculated (in the studies that did not report it) by subtracting the dropouts from the total sample size of the placebo arm. Heterogeneity was present in studies in which endoscopy was performed after 12 months; in studies published before and after 2000; in studies with 10 centers participating in the trial; in studies with long disease duration; and in studies in which there was a higher number of patients operated on once (Table 4). Consistent results were observed when assessing data pooled from studies with disease duration 80 months, from studies in which there was a number of patients who underwent first resection 50%, and from studies with 10 centers participating in the trial. Sensitivity analysis excluding the 2 outlier RCTs 5,13 with the highest and lowest placebo endoscopic recurrence rate showed that the pooled estimate of severe recurrence rate was 51.5% (95% CI, ), with no significant heterogeneity (P.64). Publication Bias The Egger publication bias plots for relapse and severe endoscopic recurrence placebo rates are shown in Figures 3 and 4. The Egger test for publication bias for the placebo relapse rate (Figure 3) showed that the risk of having missed or overlooked trials was significant: The P-value was.024 with the Egger test. By contrast, the P value was.44 with the Egger test for the placebo severe endoscopic recurrence rate (Figure 4), indicating absence of publication bias. Figure 1. Placebo rates of relapse in studies using a CDAI score of either 150 or an increase from 60 to 100 points of baseline value. Discussion The placebo rate of relapse and recurrence after surgery in CD may be a reliable measure of the natural course of disease and a basic measure for calculating sample size in RCTs evaluating drugs for prevention of relapse and recurrence. The homogeneity of these rates among trials is a mandatory condition for obtaining reliable data on natural course, and for accurately calculating the sample size. Heterogeneity of these rates among RCTs may reflect both different selections of population, and interobserver variability in the evaluation of the main outcomes. Our analysis demonstrates that the heterogeneity of the clinical outcome after surgical resection of patients with CD who received placebo in RCTs is a feature of

6 November 2008 RATES OF CLINICAL RELAPSE AND SEVERE ENDOSCOPIC RECURRENCE 1505 Table 4. Stratum-Specific Clinical Relapse Rate and Severe Endoscopic Recurrence Rate According to Features of the Study Outcome No. of studies % Pooled estimate (95% CI) P for heterogeneity Range (%) Median (%) Clinical relapse (13 35) Study duration (mo) ( ) ( ) Publication year Before ( ) Since (10 32) Disease duration (mo) a ( ) ( ) First resection (%) b 50% ( ) % ( ) Centers (n) ( ) ( ) CDAI c ( ) ( ) Severe endoscopic recurrence (28 73) Study duration (mo) ( ) ( ) Publication year Before ( ) Since (38 69) Disease duration (mo) d ( ) ( ) First resection (%) e 50% ( ) % ( ) Centers (n) ( ) ( ) a Five missing. 12,14,15,18,19 b Seven missing. 9,10,13,14,17,18,19 c Five missing. 6,9,10,14,19 d Three missing. 12,14,15 e Five missing. 9,10,13,14,17 these studies. This inconsistency among RCTs is not surprising if one considers potential biases in the selection of patients with different demographic and clinical characteristics, the different timing of clinical evaluation, and the interobserver variation of outcome assessment. Heterogeneity in clinical relapse remains if patients are stratified according to the length of follow-up: In patients with a follow-up of 12 months, 5,7,8,12,14,18,20 the rate of clinical relapse ranged was 10% 38%. Among these studies, the only outlier was the study by Prantera et al, 8 in which clinical relapse was not considered a main outcome. In the studies carried out after 12 months, because of high variability in the placebo rate (from 21% to 78%) the heterogeneity was statistically significant. It is important to underline that in the Caprilli et al study, 13 in which the relapse rate was 21%, no treatment was compared to mesalamine. This rate was the lowest among the studies and is difficult to explain if placebo is considered to have some therapeutic effect. Heterogeneity was present in the RCTs carried out 12 and 12 months, in RCTs involving few and many centers, and in the RCTs with a disease duration 80 or 80 months. Heterogeneity was less apparent in the studies published before 2000, and in the studies that included fewer patients operated on once, making these findings difficult to interpret. Patients who had a CDAI of 200 at operation had a homogeneous rate of relapse, whereas those with a CDAI of 200 had a heterogeneous rate of clinical relapse, indicating that a population with a CDAI of 200 is more heterogeneous in terms of severity. By univariate analysis, the only variable associated with the placebo rate was the length of follow-up, which seems reasonable given that clinical relapse increases with time. In contrast, duration of follow-up was not associated with endoscopic recurrence rates in our univariate analysis. In stratified analyses, there was statistically less heterogene-

7 1506 RENNA ET AL GASTROENTEROLOGY Vol. 135, No. 5 Figure 2. Placebo rates of severe endoscopic recurrence in studies using a Rutgeerts score of 2 on the colonoscopy performed during follow-up. ity among the studies of 12 months, or a shorter duration; however, the range of recurrence rates was still wide (from 30% to 79%). Because endoscopic recurrence is used as a surrogate for future clinical relapse, it is not surprising that there were fewer studies assessing this outcome after 1 year of follow-up. The few studies (4 RCTs) in which evaluation of severe endoscopic recurrence was performed after Figure 3. Placebo relapse rate. Egger publication bias plot. The proportion/standard error ratio for placebo relapse rate is plotted against the precision of the ratio along with the regression line of the intercept. Figure 4. Placebo severe endoscopic recurrence rate. Egger s publication bias plot. The proportion/standard error ratio for placebo severe endoscopic recurrence rate is plotted against the precision of the ratio along with the regression line of the intercept. 1 year showed significantly broad heterogeneity. However, some caution should be adopted when interpreting the results of these last 4 studies. In the Hanauer et al study, 10 the evaluation of relapse was made on the basis of symptoms, without using the CDAI score, and the number of dropouts was high (67%), making the evaluation unreliable. In the Caprilli et al study, 13 the severe endoscopic recurrence after 24 months was very low (21%) compared with the studies in which the evaluation was carried out within 1 year, indicating an underestimation of recurrence. It is important to emphasize that in none of these RCTs was there an estimate of the inter- and intraobserver variations in the definition of severe endoscopic recurrence. The rate of severe endoscopic recurrence among studies carried out by 10 centers was more homogeneous, indicating that it is likely that the interobserver variation increases with the increasing number of centers participating in the study. There was heterogeneity in the placebo recurrence rate in patients with only 1 resection and in those with long disease duration. It is difficult to find an explanation for this heterogeneity. Severity of disease, assessed by CDAI, and concomitant therapy at entry, were also not available in most of the RCTs, thus excluding an evaluation of these important predictors in our analysis. The number of overall placebo sample size analyzed was about 800, which is not large enough to allow definitive conclusions. In many studies the relevant clinical variables (smoking, CD pattern, etc) to explain heterogeneity among the studies were lacking, making it impossible to include these variables in the analysis. The analysis was performed on summary data, and more details might have been revealed with an analysis of individual

8 November 2008 RATES OF CLINICAL RELAPSE AND SEVERE ENDOSCOPIC RECURRENCE 1507 patient data. The evaluation of the methodologic quality did not seem to influence the assessment of the treatment efficacy of the different drugs, due owing to the mean high quality of the studies (half of these RCTs were above 70% of the maximum score). Finally, we should be particularly concerned about publication bias in settings in which small studies are being conducted. Publication bias can occur, so that studies showing a significant reduction of relapse, or severe endoscopic recurrence, are more likely to be published than those showing no difference. The risk of having missed or overlooked trials in the setting of studies assessing relapse was substantial, as assessed by tests for publication bias. Therefore, it is likely that small studies with a low rate of relapse and/or small drug effect remain preferentially unpublished. However, publication bias was probably not substantial and considered unlikely to change the magnitude of our pooled estimate of severe endoscopic recurrence. The available evidence is sufficient to conclude that the placebo rates of clinical relapse and severe endoscopic recurrence in postoperative CD are variable and no single design variable seemed to explain this heterogeneity. This analysis suggests that severe endoscopic recurrence may be more efficient for detecting drug efficacy, although possibly less clinically meaningful. Planning an RCT for a long-term outcome is problematic, both in terms of clinical relapse and severe endoscopic recurrence, owing to a high variability among control rates. Furthermore, indirect comparison among trials with different follow-up durations, and different drugs, is to be discouraged because the different estimate of drug efficacy could be entirely related to the different baseline risk of the studied population. References 1. Su C, MD. Outcomes of placebo therapy in inflammatory bowel disease. Inflamm Bowel Dis 2006;12: Bernstein C.N. The placebo effect for gastroenterology: tool or torment. Clin Gastroenterol Hepatol 2006;4: Achkar JP, Hanauer SB. Medical therapy to reduce postoperative Crohn s disease recurrence. Am J Gastroenterol 2000;95: Rutgeerts P, Geboes K, Vantrappen G, et al. Predictability of the postoperative course of Crohn s disease. Gastroenterology 1990;99: Rutgeerts P, Van Assche G, Vermeire S, et al. Ornidazole for prophylaxis of postoperative Crohn s disease recurrence: a randomized, double-blind, placebo-controlled trial. Gastroenterology 2005;128: Rutgeerts P, Hiele M, Geboes K, et al. Controlled trial of metronidazole treatment for prevention of Crohn s recurrence after ileal resection. Gastroenterology 1995;108: Brignola C, Cottone M, Pera A, et al. Mesalamine in the prevention of endoscopic recurrence after intestinal resection for Crohn s disease. Italian Cooperative Study Group. Gastroenterology 1995;108: Prantera C, Scribano ML, Falasco G, et al. Ineffectiveness of probiotics in preventing recurrence after curative resection for Crohn s disease: a randomised controlled trial with Lactobacillus GG. Gut 2002;51: Colombel JF, Rutgeerts P, Malchow H, et al. Interleukin 10 (Tenovil) in the prevention of postoperative recurrence of Crohn s disease. Gut 2001;49: Hanauer SB, Korelitz BI, Rutgeerts P, et al. Postoperative maintenance of Crohn s disease remission with 6-mercaptopurine, mesalamine, or placebo: a 2-year trial. Gastroenterology 2004; 127: Marteau P, Lemann M, Seksik P, et al. Ineffectiveness of Lactobacillus johnsonii LA1 for prophylaxis of postoperative recurrence in Crohn s disease: a randomised, double blind, placebo controlled GETAID trial. Gut 2006;55: Hellers G, Cortot A, Jewell D, et al. Oral budesonide for prevention of postsurgical recurrence in Crohn s disease. The IOIBD Budesonide Study Group. Gastroenterology 1999;116: Caprilli R, Andreoli A, Capurso L, et al. Oral mesalazine (5-aminosalicylic acid; Asacol) for the prevention of post-operative recurrence of Crohn s disease. Gruppo Italiano per lo Studio del Colon e del Retto (GISC). Aliment Pharmacol Ther 1994;8: Belluzzi A, Campieri M, Belloli C, et al. A new enteric preparation of -3 fatty acids for preventing post-surgical recurrence in Crohn s disease. Gastroenterology 1997;112: Van Gossum A, Dewit O, Geboes K, et al. Multicenter randomized-controlled clinical trial of probiotics (Lactobacillus jonhsonii, La1) on early endoscopic recurrence of Crohn s disease after ileo-caecal resection. Inflamm Bowel Dis 2007;13: Florent C, Cortot A, Quandale P, et al. Placebo-controlled clinical trial of mesalazine in the prevention of early endoscopic recurrences after resection for Crohn s disease. Groupe d Etudes Therapeutiques des Affections Inflammatoires Digestives (GETAID). Eur J Gastroenterol Hepatol 1996;8: Lochs H, Mayer M, Fleig WE, et al. Prophylaxis of postoperative relapse in Crohn s disease with mesalamine: European Cooperative Crohn s Disease Study VI. Gastroenterology 2000;118: Sutherland LR, Martin F, Bailey RJ, et al. A randomized, placebocontrolled, double-blind trial of mesalamine in the maintenance of remission of Crohn s disease. The Canadian Mesalamine for Remission of Crohn s Disease Study Group. Gastroenterology 1997;112: McLeod RS, Wolff BG, Steinhart AH, et al. Prophylactic mesalamine treatment decreases postoperative recurrence of Crohn s disease. Gastroenterology 1995;109: Ewe K, Bottger T, Buhr HJ, et al. Low-dose budesonide treatment for prevention of postoperative recurrence of Crohn s disease: a multicentre randomized placebo-controlled trial. German Budesonide Study Group. Eur J Gastroenterol Hepatol 1999;11: Bresci G, Petrucci A, Banti S. 5-aminosalicylic acid in the prevention of relapses of Crohn s disease in remission: a long-term study. Int J Clin Pharmacol Res 1991;11: Fiasse R, Fontaine F, Vanheuverzwyn R. Prevention of Crohn s disease recurrence after intestinal resection with Eudragit-Lcoated 5-amynosalicylic acid preliminary results of a 1-year double-blind, placebo controlled study (abstr). Gastroenterology 1991;100:A Bondesen S, the Danish 5-ASA group. Mesalazine (Pentasa) as prophylaxis in Crohn s disease: a multicenter, controlled trial (abstr). Scand J Gastroenterol 1991;26: Thomson ABR, International Mesalazine Study Group. Coated oral 5-aminosalicylic acid versus placebo in maintaining remission of inactive Crohn s disease. Aliment Pharmacol Ther 1990; 4: Schreiber S, Howaldt S, Raedler A. Oral 4-aminosalicylic acid versus 5-aminosalicylic acid slow release tablets: double-blind, controlled pilot study in the maintenance treatment of Crohn s ileocolitis. Gut 1994;35: Herfarth H, Obermeier F, Tjaden C, et al. Double-blind, double dummy, randomised, multicentre, comparative study on the efficacy and safety of azathioprine (AZA) versus mesalazine

9 1508 RENNA ET AL GASTROENTEROLOGY Vol. 135, No. 5 (5-ASA) for prevention of postoperative endoscopic recurrence in Crohn s disease. Abstract. DDW 2006 AGA A-480 T Nos P, Hinojosa J, Aguilera V, et al. Azathioprine and 5-ASA in the prevention of postoperative recurrence of Crohn s disease [in Spanish]. Gastroenterol Hepatol 2000;23: Caprilli R, Cottone M, Tonelli F, et al. Two mesalazine regimens in the prevention of the post-operative recurrence of Crohn s disease: a pragmatic, double-blind, randomized controlled trial. Aliment Pharmacol Ther ;17: Moher D, Cook DJ, Eastwood S, et al. Improving the quality of reports of meta-analyses of randomised controlled trials: the QUOROM statement. Quality of Reporting of Meta-analyses. Lancet ;354(9193): Nicolucci A, Grilli R, Alexanian AA, et al. Quality evolution and clinical implications of randomized, controlled trials on the treatment of lung cancer: a lost opportunity for meta-analysis. JAMA 1989;262: Egger M, Smith DG, Schneider M, et al. Bias in meta-analysis detected by a simple, graphical test. BMJ 1997;315: Received February 12, Accepted July 19, Address requests for reprints to: Mario Cottone, MD, Professor of Internal Medicine, Dipartimento di Medicina, Pneumologia e Fisiologia della Nutrizione Umana, Università di Palermo, Palermo, Italy; dickens@tin.it The authors thank Warren Blumberg for his help in editing this paper. The authors disclose no conflicts. Appendix 1. Quality Scoring System 1. Randomization A (6). By central or independent source B (4). By sealed opaque envelopes C (0). Other or not reported 2. Efficacy of Randomization A (6). When a table showing the distribution of potential prognostic factors (disease duration, previous intestinal resection, remission duration, site of disease) among treated and controls was presented and no statistically significant differences were found B (3). When authors said that a balancing of prognostic factors was obtained, but their distribution was not reported in tabular form C (0). Information not reported 3. Blinding A (6). Double-blinded B (3). Single-blinded C (0). Unblinded or information not reported 4. Compliance A (6). Assessed by objective methods, such as a percentage of scheduled visits attended or pill count B (3). Assessed only in patients on active treatment; assessed by non-objective methods (information from patients and relatives) C (0). Information not reported 5. Withdrawals A (6). When the article stated that no patients were lost to follow-up, or reported the precise number of patients withdrawn from the analysis B (0). When either 0.15% of the patients were lost to follow-up or no information was reported 6. Analysis of Withdrawals A (6). When the principle of intention-to-treat was applied in handling withdrawals B (0). When withdrawals were excluded from analysis 7. Sample Size A (3). A priori estimate of sample size reported B (0). A priori assessment of sample size not reported or not done 8. A posteriori Estimate of Study Power A (3). When authors considered the possibility of b error when no statistically significant difference between treatments was observed B (0). When this possibility was not explicitly considered by authors C (0). Not assessable: when studies reported a statistically significant difference (item excluded from the denominator) 9. Response Evaluation A (6). When criteria used in evaluation of responses were explicitly reported B (0). Information not reported 10. Side Effects and Complications A (6). Number, type, and outcome of side effects assessed in a blinded fashion in both treated and control groups B (3). Unblinded assessment, or information reported only in treated patients; incomplete information C (0). No information reported 11. External Validity A (6). Inclusion and exclusion criteria clearly stated: the study reported information on source population from which patients were selected (two ways of reporting were acceptable: either the authors said that all consecutive patients were enrolled or they reported the actual number of patients potentially eligible but not enrolled) B (3). Only one of the above conditions fulfilled C (0). No information reported

10 November 2008 RATES OF CLINICAL RELAPSE AND SEVERE ENDOSCOPIC RECURRENCE 1509 Appendix 2. Quality Scoring Components per Each Randomized Controlled Trial Scoring on each component of the quality score Study Total score % of maximum score McLeod et al Lochs et al Caprilli et al Sutherland et al Brignola et al Rutgeerts et al Hellers et al Prantera et al Marteau et al Hanawer et al Colombel et al Rutgeers et al Ewe et al Van Gossum et al Florent et al Belluzzi et al 14 n.e. a a n.e., not evaluable.