The use of placebo in randomized clinical trials is necessary in most diseases to establish drug efficacy. The use of placebo allows blinding of patients and investigators, which helps to prevent bias in determining endpoints. This is particularly important in inflammatory bowel disease (IBD) studies, which often utilize symptoms reported by the patients and assessments made by the investigators, to determine response and remission to medical therapy. In addition, the natural course of CD is marked by variable periods of relapse and remission. Therefore, understanding the outcomes of placebo-treated subjects is important in designing future clinical trials.
Maintenance of clinical remission
In this study, we identified significant heterogeneity among the studies in terms of rates of clinical remission. Interestingly, few features of the studies were associated with placebo remission rates. Prior steroid therapy was associated with an increased clinical remission rate. Otherwise, no features were statistically significant and stratified analyses provided limited insight into observed heterogeneity. The remission rates in the active treatment arm were found to be positively associated with remission rates in the placebo arm. This finding is not surprising because, in a single RCT, similar subject factors are enforced across both arms, making the sample relatively homogeneous.
The clinical outcomes were independent of duration of follow-up, a study feature previously shown to influence placebo remission rates in active CD.2 For practical reasons, most studies of post-operative therapy follow patients for 1 year to assess clinical outcomes. Given that most patients who undergo surgery will ultimately have a clinical relapse if followed long enough, the lack of an association of follow-up duration with remission rates in our analysis probably reflects the lack of variability in follow-up duration among the studies rather than a true biological effect. We performed additional analysis for study duration in terms of placebo remission or endoscopic recurrence rates on the basis of 4-week increases in duration of follow-up. Similarly, we did not find a significant association with placebo rates associated with increasing duration of follow-up (OR 1.0, 95% CI 0.9–1.02 for placebo remission rates, OR 0.98, 95% CI 0.95–1.01%).
Visit frequency, which was also found to be a significantly associated with placebo response rates in RCTs involving active CD, was found to be inversely related to maintaining clinical remission in the placebo arm, but this difference did not achieve statistical significance (OR 0.43, 95% CI 0.18–1.01). This may be because of detection bias as more visits or opportunities for probing by investigators may result in an increased likelihood for the patient to report symptoms and prompt intervention to evaluate for recurrence, such as through endoscopy or radiography.
The evaluation of endoscopic recurrence is a frequently employed outcome measurement in clinical trials as it has been shown to provide prognosis for disease course and thus may aid in directing further therapy at an earlier time period. Endoscopic recurrence has been observed in up to 60% of patients who underwent ileocolonic resection at 6 months.1 The severity of the endoscopic lesions seen post-operatively in the absence of symptoms has been shown to be predictive of clinical recurrence.1 In our study, the endoscopic outcome was significantly associated with disease distribution and phenotype, which also explained some of the between-study heterogeneity. Interestingly, these subject characteristics did not significantly impact the clinical remission rates in the placebo treated group. As the median follow-up duration was the same for both outcomes, it is likely that clinical recurrence was delayed compared to endoscopic recurrence and with longer follow-up, the same associations may have been seen for clinical remission. Alternatively, the lack of significance seen in the clinical remission rates with respect to disease distribution and phenotype may also be as a result of misclassification bias given the less objective nature of measuring clinical outcomes. Finally, the associations observed in our analyses of clinical outcomes or endoscopic outcomes could have resulted from type-2 or type-1 statistical errors, respectively.
Lower placebo endoscopic recurrence rates were more likely to be associated with higher percentages of patients with only small bowel involvement. Inversely, it was found that placebo-treated patients with both small and large bowel involvement were more likely to have higher endoscopic recurrence rates. This is expected as patients with more extensive disease, in particular, concomitant small bowel and colonic disease, are more likely to relapse post-operatively than those patients who had isolated small bowel or colonic disease.18, 19 Fistulizing type of CD was also found to be associated with higher placebo endoscopic recurrence rates. This is consistent with previous observational studies, which have reported earlier recurrence and need for subsequent surgery in patients with perforating disease as opposed to nonperforating disease.20, 21
We also found in our study, higher rates of endoscopic recurrence in those placebo-treated patients who had been on prior immunomodulator therapy and/or had prior surgery. Withdrawal of immunomodulator therapy, primarily azathioprine, even after prolonged remission of CD has been shown to increase the risk of relapse of CD.22 While prior surgery may be indicative of preoperative disease severity and thus could contribute to higher endoscopic recurrence rates, the finding in our study of higher endoscopic rates with prior surgery is actually contrary to what has been shown in prospective and retrospective observational studies evaluating the natural course of CD post-operatively.1, 19
While our data suggest important insights for interpreting the results of prior studies, they may also contribute to the design of future clinical trials in post-operative maintenance of CD. While no single factor appeared to account for higher placebo clinical remission rates, there appeared to be several patient characteristics, which were associated with higher placebo endoscopic recurrence rates as mentioned previously. Therefore, modifying inclusion criteria on the basis of these characteristics may identify possible ‘high-risk’ groups which could potentially reduce the required sample size by establishing a larger treatment effect.
Theoretically, over a sufficiently long follow-up time, one should observe an increase in the cumulative relapse rate with increasing follow-up time. Surprisingly, our study did not identify a significant association between the duration of follow-up and placebo clinical remission or recurrence rates for the durations of follow-up which were used in the included studies (4 months – 3 years). Although not designed specifically to address this question, these data suggest that there is no substantial increase in the placebo recurrence rate with increasing duration within this time frame. As such, it may be more efficient in future studies to increase the number of patients with 1–2 years of follow-up rather than attempting to study fewer patients with longer follow-up, which would potentially lead to a higher attrition rate.
While there was no single predictor found to be associated with recurrence in all of the clinical trials, several of the individual studies found at least one of the significant predictors from our study to be a notable patient characteristic when evaluating clinical or endoscopic recurrence rates. These predictors included prior medical therapy, prior surgery for CD, disease distribution, disease duration and disease type.5, 6, 9, 15 The study by Lochs et al. contributed substantially to the pooled estimate of the placebo remission rate (Figure 2) and identified that duration of disease and steroid intake prior to surgery influenced their recurrence rates.9
Certain studies can be considered outliers in terms of the remission and recurrence rates. Possible hypotheses may be posed for the outliers of remission rates found among the clinical trials. For instance, the study by Hanauer et al., reported a placebo clinical remission rate of 23% at 2 years. The investigators postulated that their use of a clinical grading scale rather than CDAI for defining recurrence may have contributed to this.6 When directly comparing the high rate of placebo clinical remission (91% at 1 year) in the trial by Prantera et al.,7 with the low clinical remission rate (34% at 4 months) in the trial from Florent et al.,12 we can see that the baseline characteristics of the study groups differed with respect to disease distribution/location and percentage of subjects with previous operations for CD, which were identified as characteristics significantly influential on rate of placebo recurrence from this meta-analysis.
There are several limitations to this study. First, based on the study design, the collection and analyses of group-level data may not accurately reflect the responses of individual subjects and thus cautious application of results to individual patients must be applied. However, the results achieved are still potentially useful in the interpretation of study results in PC-RCTs in IBD, particularly with reference to factors that may differ between the study groups. Furthermore, it is important to note those characteristics, particularly disease distribution and phenotype, which may influence the outcome and should thus be accounted for in the study design or analyses.
Publication bias is a potential limitation of most meta-analyses. Studies which demonstrate a significant benefit in the active treatment arm may be more likely than negative studies to be published. This would be more likely to occur with small studies than with large studies given the large amount of time and money typically devoted to such studies. Conceivably, exclusion of a number of small, unpublished studies could influence our overall study results if those small studies consistently had higher or lower rates of recurrence in the placebo arm than the published studies. If lower recurrence rates in the placebo arm resulted in a lower likelihood of a statistical significance within small unpublished studies, our estimate of the placebo clinical remission rate may be slightly too low.
In conclusion, the placebo outcomes in PC-RCTs for postsurgical maintenance therapy in CD were highly variable. While there is no single design feature that explains the heterogeneity demonstrated in the estimate of clinical remission rates, endoscopic outcomes were associated with disease distribution and phenotype, and accounting for these factors removed the heterogeneity demonstrated within our study results. Knowledge of these factors can be helpful when comparing results of different studies and when planning new studies of post-operative therapy.