SEARCH

SEARCH BY CITATION

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgement
  8. References
  9. Appendices

Background  Remission and response are the main outcomes to evaluate the efficacy of new treatments for Crohn’s disease (CD).

Aim  To explain variation of remission and response rates in active luminal CD.

Methods  We studied control patients from trials of biological therapies through articles retrieved by MEDLINE search (from 1997 to 2007) and by bibliography review. Thousand nine hundred and thirteen control patients from 28 trials were identified; data were extracted by three independent observers and pooled by DerSimonian and Laird random effect model; factors influencing remission and clinical response were explored by metaregression for aggregated data.

Results  The pooled control rates of remission and response were 17% and 33%, respectively, both with significant heterogeneity among studies (P < 0.0001). At metaregression, the time of primary outcome evaluation was associated with remission, whereas the trial’s criteria for defining response and publication year were predictors of response. CDAI score, CRP levels or other clinical variables related with disease activity or concomitant medications were not significant factors.

Conclusions  Populations used as ‘add-on’ treatment comparator in trials of biological therapies for active luminal CD are poorly characterized and outcomes are heterogeneous. Planning of future trials will require better description of patients and concomitant therapies, blinding of outcome assessors and homogeneous criteria of outcome definition.


Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgement
  8. References
  9. Appendices

Recent insights into the pathogenesis of Crohn’s disease (CD) have led to the development of new treatment options, represented by biologically targeted treatments.1, 2 Such treatments are aimed at neutralizing proinflammatory mediators or consist of administration of anti-inflammatory molecules.3 Among these molecules are TNF-α inhibitors, other inflammatory cytokines inhibitors (IL-6, IL-12), anti-inflammatory cytokines (IL-10, IL-11), adhesion molecules inhibitors, cellular signals transduction inhibitors or natural immunity stimulators. Several randomized controlled trials (RCTs) of biological molecules in active inflammatory CD have been conducted in recent years.4 The design of these trials has been an ‘add-on’ study design, in which patients with active CD5 were given standard therapy such as aminosalicylates, corticosteroids and immunosuppressants and were randomly assigned to receive also either a new agent or a placebo.

In active CD, the percentage of placebo response changes with factors such as the site of disease, the duration of disease and its activity, and also with concomitant therapies and a smoking habit.6 In a meta-analysis of placebo outcomes in active CD patients, including a large set of trials assaying different drugs from 1976 to 2000, Su et al.7 identified factors that influenced the placebo remission rates, such as study duration, number of study visits and disease severity at entry. On the other hand, no significant predictor of response was found.

An accurate estimation of outcome rates in control populations and the identification of trial and patient characteristics that are associated with these rates are of help in interpreting results and for future planning of trials of medical therapy in active CD.3–5, 7 We carried out a systematic review of RCTs of biological therapy in active, luminal CD and applied meta-analysis techniques to describe control populations, to assess their outcome rates and, finally, to explore the reasons for possible heterogeneity of estimates.

Methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgement
  8. References
  9. Appendices

A literature search of RCTs of therapies for CD was carried out starting from 1997, i.e. the year of the well-known first RCTs of biologicals.14–16 We used the MEDLINE–PubMed service (clinical study category search: clinical queries: therapy) and limited our search to English papers and adult patients. Comments, editorials and letters were excluded (see Appendix 1 for the combination of MESH terms). EMBASE, Cochrane Library and Cochrane Controlled Trials Register were also consulted. The references from the identified articles and previously published meta-analyses were manually searched to identify other potential studies.

Eligible studies were RCTs comparing an active treatment represented by a biological therapy with a control group including a placebo in adult patients with luminal CD in acute phase at baseline [CD Activity Index (CDAI) > 150]. Endpoints of interest from the studies were clinical response and remission measured through the CDAI score, which consists of eight subjective and objective criteria of disease activity, ranging from 0 to 600.8 A higher score generally represents a more severe disease activity. Remission is defined as a score that measures ≤150 points, while response is defined as a reduction in the score exceeding a predefined value, usually 70 or 100. We considered as response the reduction of 100 points of the CDAI score. Only when this measure was not available, a reduction of 70 points was alternatively considered.

The main reason for exclusion was the absence of active disease or a not inflammatory pattern of disease (short bowel syndrome, draining fistulas, stricture with obstructive symptoms, abdominal abscess). Trials that considered ‘steroid-sparing’ (defined as discontinuation of steroid therapy without a disease flare represented by CDAI score ≥220 points) as primary outcome were excluded from the present meta-analysis, as well as reports in which the number of patients in the experimental and in the control groups was not reported or in which the outcome was the maintenance of remission and response.

Decisions about which trials to include were blindly taken by two reviewers (FR, MC) and disagreements were solved by discussion. Excluded trials were identified with the reason for exclusion. Data concerning trials and patients characteristics as well as treatment outcome were abstracted by each study by three independent reviewers (FT, AS, FR) and discrepancies were settled by consensus. A quality analysis of each study was conducted independently by two authors (FT, MC) by computing the Jadad score.9 This is a well-established and validated scale applying seven criteria (five for good and two for poor study quality) to obtain a numerical score between 0 and 5 with 0 being the poorest and 5 being the highest design and reporting quality.

The recorded trial characteristics were: publication year, journal, nationality of the first author, number of participating centres, sample-size calculation, number of arms, number of total patients randomized, number of patients in control group, class of biological therapy tested in treatment arm, dose and route of administration of the experimental drug, duration of treatment and follow-up, definition of response and remission, primary and secondary outcomes with relative time of evaluation (weeks) and quality of trial’s score. The recorded patient characteristics were: age, gender, weight, duration of the disease, smoking history, disease location, previous intestinal resection, previous and concomitant medications, C-reactive protein (CRP), CDAI score and minimum CDAI score at entry.

The endpoints from the trials were combined using the DerSimonian and Laird10 random effect model to estimate the rate and corresponding 95% confidence interval (CI) for each clinical outcome. The random effects model was chosen to provide estimates that consider variance both between and within studies. Heterogeneity testing applied the Q-statistic with P-values <0.10 being considered significant. Studies that did not mention a specific outcome were excluded from the analysis for that endpoint. Because of the aggregated nature of data on outcomes and covariates, study-level metaregression analyses were performed to search for reasons of heterogeneity found among the estimates.11 The dependent variable was the log-odd from each arm for the outcome of interest. Weights were assigned according to the estimated variance from the log-odd. The residual between trial heterogeneity was expressed as τ2, estimated by a restricted maximum likelihood method using an iterative procedure.12, 13 To establish which factors and how many factors would have been included in the metaregression, we considered either previous knowledge of factors shown to be associated with remission and response in patients with active CD7 and the number of retrieved trials. CDAI score at entry, time of outcome evaluation (weeks), publication year, number of study visits, definition of response and activity class of luminal CD were selected for the analysis. Stratum-specific pooled estimates of the outcomes of interest within each stratum of those covariates that resulted statistically significant were calculated to explore whether associations of these covariates with the outcome of interest would account for the heterogeneity between studies. All analyses were carried out using Stats Direct statistical software version 2.6.1 (19.1.2007) and Stata 8.0 (©1999 Stata Corporation, College Station, TX, USA).

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgement
  8. References
  9. Appendices

A total of 237 potentially relevant articles were identified (Figure 1). After a review of the abstracts, 166 papers (137 RCTs and 29 prospective studies) were excluded because active treatment was not a biological therapy. The remaining 71 articles on biologicals were analysed for inclusion: 28 fulfilled the selection criteria14–41 and 43 were excluded. The causes of exclusion were outcome different from induction of remission, steroid-free remission as outcome, trials on maintenance therapy, trials on fistulizing disease, lack of placebo arm, studies on ulcerative colitis and rheumatoid arthritis (Appendix 2).

image

Figure 1.  Flow-diagram of study selection.

Download figure to PowerPoint

The degree of agreement between the reviewers exceeded 95%. The distribution of the main characteristics of the 28 RCTs considered in this study is reported in Table 1. The studies included a total of 1913 patients in the control groups, ranging from 5 to 325 per arm. Only seven trials (25%) included more than 100 patients among controls. Most of the RCTs (75%) were phase II trials. About 70% of trials were published by gastroenterology journals after the year 2000. TNF antagonists were evaluated in 12 (43%) studies, adhesion molecule inhibitors in six (21%) studies, anti-inflammatory interleukins in five (18%) studies, inhibitors of Th polarization in four (14%) studies and growth factors in only one study. Each trial was sponsored by the pharmaceutical company manufacturing the experimental drug.

Table 1.   Distribution of main characteristics of 28 selected RCTs
VariablesNumber of trials
  1. CDAI, Crohn’s Disease Activity Index.

PC-RCTs (number of studies)28
Total patients randomized6732
Total patients under placebo ‘add on’1913
Arms
 211 (39%)
 36 (21%)
 47 (25%)
 54 (14%)
 Total number of arms68
Trial phase
 221 (75%)
 37 (25%)
Publication year
 ≥200023 (82%)
 <20005 (18%)
Journal
 General8 (29%)
 Specialistic20 (71%)
Prime author nationality
 Europe9 (32%)
 US14 (50%)
 Canada4 (14%)
 Japan1 (4%)
Biologic therapy
 TNF antagonists12 (43%)
 Adhesion molecule inhibitors6 (21%)
 Anti-inflammatory interleukins4 (14%)
 Inhibitors of Th1 polarization4 (14%)
 Granulokines1 (4%)
 Inhibitors of proinflammatory cytokine receptors1 (4%)
Patient selection
 Mild–moderate5 (18%)
 Moderate4 (14%)
 Moderate–severe19 (68%)
Primary outcome
 Remission9 (32%)
 Response9 (32%)
 Remission and response4 (14%)
 Safety6 (21%)
Remission definition
 CDAI ≤15022 (78%)
 CDAI ≤150 and other criteria6 (21%)
Response definition
 CDAI ≥704 (16%)
 CDAI ≥70 and other criteria2 (8%)
 CDAI ≥1009 (37%)
 CDAI ≥100 and ≥709 (37%)
Trial analysis
 Per protocol4 (14%)
 Intention to treat24 (86%)
Primary outcome
 Remission9 (32%)
 Response9 (32%)
 Remission and response4 (14%)
 Safety6 (21%)
Concurrent steroid therapy
 Yes25 (89%)
 No3 (11%)
Concurrent immunosuppressors
 Yes24 (86%)
 No4 (14%)
Drug route
 Intravenous15 (53%)
 Subcutaneous12 (43%)
 Oral1 (4%)
Number of placebo administrations (i.v., s.c. or oral) (s.d.)11 (15)
Number of study visits (s.d.)6.5 (1.9)
Sample-size calculation
 Yes20 (71%)
 No8 (29%)
Blinding description
 Yes15 (54%)
 No13 (46%)
Randomization description
 Yes21 (75%)
 No7 (25%)
Withdrawals and dropouts description
 Yes23 (82%)
 No5 (18%)
Jadad score
 ≥323 (82%)
 <35 (18%)

With regard to the definition of disease activity, five (18%) trials included patients who were considered to have a mild-to-moderate disease (CDAI at entry between 150 and 220), four (14%) included patients with a moderate disease (CDAI at entry between 221 and 450) and 19 (68%) included patients with a moderate-to-severe disease (CDAI at entry between at least 220 and also more than 450).

About 80% of the RCTs considered remission and/or response as the primary outcome. All the 28 included articles reported remission data, whereas response data were available in only 24 papers (Table 2). In 22 studies (79%), remission was defined as having a CDAI score of <150 or ≤150, while in six RCTs (21%) additional criteria were used in combination with the definition of remission, including the decrease in the CDAI score of ≥100 points or ≥70 points from baseline, the steroid-free remission or the corticosteroid dosage ≤ the level at baseline.

Table 2.   Characteristics of 28 studies containing remission and/or response data
Author (year)Biological therapyTime of outcome (weeks) Placebo size (n) CDAI score at baseline*Definition of remission Remission rate (%)Definition of response Response rate (%)
  1. CDAI, Crohn’s Disease Activity Index; IBDQ, inflammatory bowel disease quality score.

  2. * Mean or median.

Stack et al. (1997)14CDP571210253CDAI ≤1500
Targan et al. (1997)15Infliximab425288CDAI ≤150 + IBDQ score 170–1908≥70 + no change in any concomitant medications16
Van Deventer et al. (1997)16IL-10413292CDAI ≤150 + reduction ≥10023.1≥10023.1
Yacyshyn et al. (1998)17ISIS-230245291CDAI ≤15020
Sands et al. (1999)18IL-11315308.5CDAI ≤150 + reduction ≥700≥10013.3
Fedorak et al. (2000)19IL-10423261CDAI ≤1500
Schreiber et al. (2000)20IL-10466271CDAI ≤150 + reduction ≥10018.2>10027.3
Gordon et al. (2001)21Natalizumab212273CDAI <1508.3
Sandborn et al. (2001)22CDP571258332CDAI ≤1506.9≥70 ≥10027 13.8
Sandborn et al. (2001)23Etanercept420265CDAI <15020≥70 or CDAI <15045
Sands et al. (2002)24IL-11649310CDAI <150 + reduction ≥10016.3≥70 ≥10039 32.6
Ghosh et al. (2003)25Natalizumab663300CDAI ≤15027≥7038.1
Ito et al. (2004)26Anti-IL-61213295CDAI ≤1500≥7030.8
Mannon et al. (2004)27Anti-IL-12188279CDAI ≤15012.5≥10025
Sandborn et al. (2004)28CDP57128132301CDAI ≤15020.4≥70 ≥10027 23.5
Winter et al. (2004)29Certolizumab pegol425310CDAI ≤15020.≥10056
Sandborn et al. (2005)30Natalizumab10181303CDAI ≤15030.4≥7048.6
Schreiber et al. (2005)31Certolizumab pegol1273CDAI ≤15023.3≥10035.6
Korzenik et al. (2005)32Sargramostin843300CDAI ≤15019≥70 ≥10044 26
Hanauer et al. (2006)33Adalimumab474296CDAI ≤15012.2≥70 ≥10037 24.3
Hommes et al. (2006)34Fontolizumab443303CDAI ≤15011.6≥10032.5
Reinisch et al. (2006)35Fontolizumab410304CDAI ≤15030≥70 ≥10060 60
Rutgeerts et al. (2006)36Onercept838306CDAI ≤15023.7≥10036.8
Schreiber et al. (2006)37p38862306CDAI ≤15030.6≥7056.4
Targan et al. (2007)38Natalizumab12250299CDAI ≤15016≥70 ≥10032 22
Sandborn et al. (2007)39Adalimumab4166313CDAI ≤1507≥70 ≥10034 25
Yacyshyn et al. (2007)40ISIS-230212110CDAI ≤150 + reduction ≥7034.5≥10054
Sandborn et al. (2007)41Certulizumab pegol6325297CDAI ≤15017≥100 ≥7027 38

In eight trials, response was defined as having a decrease in the CDAI score of ≥70; in 10 trials, it was defined as a decrease in the CDAI score of ≥100 points. In six trials, the data of response were reported for a reduction of both 70 and 100 points (in this case, we considered only the results according to the more restrictive criteria). Of the eight trials in which response was defined as having a decrease in the CDAI score of ≥70, two used additional criteria in combination with the definition of response (e.g. response not accompanied by a change in any concomitant medications, or response defined as having a decrease in the CDAI score of ≥70 or, also, a CDAI score of <150 points).

The evaluation of the outcomes was performed within 6 weeks in 64% of the trials. The time at which outcomes were evaluated for pooling was that reported when remission or response were the primary outcomes (Table 2). Most of the drugs (54%) were administered by intravenous injection.

With regard to their quality, although all trials reported a statement on randomization and all trials except one32 were declared double-blinding, the randomization method was not described in 11 trials (39%) and the blinding method was not described in 13 trials (46%). No study reported a statement on blinding of outcome assessment. The description of withdrawals and dropouts was reported in 23 trials (82%). A sample-size calculation was reported in 20 (71%) of trials. Anyway, a Jadad score of at least 3 was attributed to 22 (79%) studies.

The distribution of main characteristics of patients per arm is reported in Table 3. In most of the trials, the proportion of patients assuming steroid and immunosuppressor therapy simultaneously was not available. All studies but one reported mean or median CDAI at entry (mean 294, s.d. 18). Data on CRP levels at entry were available from 14 studies.15, 21, 26, 28, 30–33, 35–39, 41 In nine trials, data were given through mean change of CRP levels with time between treated and controls. Very few studies reported a subgroup analysis on remission and response based on patients with CRP levels lower or higher than 10 mg/L. A correlation between entry CDAI and CRP levels was absent in the 13 studies in which both data were available.

Table 3.   Distribution of main characteristics of patients per arm in 28 selected randomized controlled trials
 Number of placebo armsMean (s.d.)Range of means
  1. CDAI, Crohn’s Disease Activity Index; CRP, C-reactive protein.

Age (years)2836.3 (3.7)26–42
Male (%)2845.9 (11.4)24–77
Weight (kg)1769.4 (6.3)53–75
Duration of disease (years)238.7 (1.6)5.7–11
Disease location
 Ileum2239.3 (23.9)0–80
 Colon2030.1 (16.2)8–41
 Ileum–colon1846.4 (20.6)9–92
 Perianal722.8 (12.4)0–33
Previous intestinal resection1245.4 (8.7)33–61
Percentage of patients with steroids therapy2545.9 (24.6)0–100
Percentage of patients with immunosuppressive therapy2530.9 (12.9)0–47
Percentage of patients with aminosalycilate therapy2655.6 (19.8)23–100
Percentage of patients with prior anti-TNF therapy1140.4 (23.9)12–100
CDAI score at baseline27294 (18)253–332
CRP (mg/L) at baseline1314.1 (10.9)0.9–35
Number of study visits2714.1 (10.9)3–10
Number of placebo administrations (i.v., oral or s.c.)2814.1 (10.9)1–56

Figure 2 shows remission data: rates ranged from 0% to 34% and the pooled estimate was 17% (95% CI: 13–21%). Statistically significant heterogeneity among the studies (Q-test: 103.4, P < 0.0001) was found, with residual between study heterogeneity (τ2 = 0.043). Univariate metaregression identified time of primary outcome evaluation as unique factor positively associated with remission (Table 4). The final multivariate metaregression model, adjusted for the other variables, included only trial’s time of primary outcome evaluation as a study level factor associated with remission (Table 4).

image

Figure 2.  Rates of remission in controls of biological trials using placebo as ‘add-on’ treatment.

Download figure to PowerPoint

Table 4.   Metaregression analysis for the arm-placebo effect on remission (28 trials) and on response (24 trials)
VariablesCoefficientS.E.P-valueτ2CoefficientS.E.P-valueτ2
  1. The dependent variable is the observed log-odd from each trial for remission. Weights have been assigned according to the estimated variance from the log-odd.

  2. S.E., standard error of the coefficient.

  3. τ2 is a measure of residual between trial heterogeneity, estimated by a restricted maximum likelihood method using an iterative procedure.

  4. * Continuous.

  5. † In multivariate analysis, τ2 refers to the final model including only two variables.

Univariate metaregression for remissionMultivariate metaregression for remission
Time of primary outcome evaluation (>6 vs. ≤6 weeks)0.5960.2050.0040.1130.4870.2120.0220.102
Publication year*0.0620.0440.1590.202Not included
Mild-to-moderate CD (yes vs. no)−1.4550.7850.0640.191
CDAI score at entry*−0.0020.0080.7590.162
Number of study visits0.0760.0590.2020.149
Univariate metaregression for responseMultivariate metaregression for response
Definition of response (≥70 vs. ≥100 points)0.4530.2540.0740.1990.5860.2400.0150.158†
Publication year (≥2002 vs. <2002)0.0800.0450.0760.2400.6310.2610.016 
Time of primary outcome evaluation (>6 vs. ≤6 weeks)0.0010.2180.9590.264Not included
Mild-to-moderate CD (yes vs. no)−0.2300.4270.5900.262
Number of study visits−0.0020.0690.9730.254
CDAI score at entry−0.0030.0080.7060.222

Response data were reported in 24 out of 28 studies. The pooled estimate of response rate was 33% (95% CI: 28–38%), with evidence of heterogeneity among the studies and residual between study heterogeneity (Q-test: 115.4, P < 0.0001; τ2 = 0.0535) (Figure 3). At univariate metaregression, trial’s criteria for defining response (CDAI reduction of at least 70 points vs. of at least 100 points) and trial’s publication year (after 2002 vs. previous to 2002) were the factors marginally associated with response (Table 4). These two factors remained significant at multivariate metaregression. After stratifying across various categories of these variables (Table 5), heterogeneity remained statistically significant in all the strata.

image

Figure 3.  Rates of response, defined as a reduction of 100 points of Crohn’s Disease Activity Index (CDAI) score, in controls of biological trials using placebo as ‘add-on’ treatment. * Rates of response defined exclusively as a reduction of 70 points of CDAI score.

Download figure to PowerPoint

Table 5.   Stratum-specific remission and response rates in the placebo group for statistically significant predictors
VariablesnPooled estimate 95% CIP-value for test of heterogeneity
Remission
Time of primary outcome evaluation (weeks)
 >610137–180.0003
 ≤6182318–260.0003
Response
Definition of response
 ≥7064131–500.016
 ≥100183025–35<0.0001
Publication year
 ≥2002173630–42<0.0001
 <200272517–330.069

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgement
  8. References
  9. Appendices

In this study, we looked at outcomes of placebo ‘add-on treatment’ controls from trials of biological therapies for CD published in the last 10 years. Efficacy was not of interest because we considered it not appropriate to combine data from molecules characterized by different mechanisms of action. We found large heterogeneity in both remission and response rates assessed by conventional CDAI’s criteria in patients with active CD. A few design characteristics of trials, but not patient-related variables were found to be factors explaining such heterogeneity. A poor characterization of patients in terms of disease activity and an inadequate description of concomitant therapies was a typical aspect of these trials.

It should be noted that the quality assessment of included trials revealed methodological limitations. Among these limitations, those regarding the report of blinding are of particular concern by considering the known variation among researchers in the administration and implementation of the CDAI score.42 Any conclusion about the factors that drive the placebo outcome when this instrument is used under circumstances similar to those described in the trials included in our study should be arrived at with caution because of the risk of assessment bias. A limitation of the trials we included in our study was also represented by an insufficient characterization of patients regarding response to previous therapies and regarding concomitant treatment during the study. The criteria for definition of efficacy of conventional therapy were often vague and variable between the studies. In fact, in many cases, patients were defined as resistant to steroids, despite the fact that the dose was below 40 mg daily. In some studies, immunosuppressive therapy was started 3 months before the beginning of the study, although the action of these drugs requires at least 4 months of treatment. Moreover, the rate at entry of patients taking concomitant steroids, immunosuppressive therapy and aminosalycilates varied between studies and often the real proportion of patients assuming more than one of these concomitant therapies was not defined. This wide variability produces a heterogeneous population within studies and it makes it difficult to interpret any possible influence of concomitant therapies on ‘placebo-add-on’ outcomes. Although only 11 trials reported data on the proportion of control patients at entry with prior anti-TNF therapy (range 12–100%), a subgroup analysis based on this characteristic was possible in four trials alone and it did not show significant differences in the pooled response rates.

We report a pooled estimate of remission rate of 17%. Heterogeneity of rates among trials was evident, ranging from 0% to 34%; it was statistically explained by time of primary outcome evaluation within 6 weeks as the most important factor. Because patients are more likely to enter spontaneous remission with time, a positive correlation between the time of primary outcome evaluation and control outcome rate could be expected even if these patients take other drugs. Among factors that may influence such correlation there are, apart from a true placebo effect, the natural history of CD itself, the patient’s conviction that participating in the trial is better than usual care (more frequent contacts) and regression to the mean of measures of disease activity. Also, in the Su et al.’s study7 that reported a very similar pooled remission rate, a similar time-related variable, the study duration (per month increment), was shown to be a factor positively associated with remission. However, in that study as well in our study, there was a significant heterogeneity within strata. Such kind of heterogeneity that weakens our results is shown in Figure 4 in which the range of reported remission rates at 4 weeks as well as at 12 weeks is from 0% to more than 30%. It should be noted that in our study, there was a trend towards increase in placebo remission rates with longer periods of follow-up,27, 28 as observed in other studies.7, 45 Moreover, it should be noted that only one40 of the five trials showing rates of more than 27%25, 30, 35, 37, 40 did use as outcome criteria a CDAI ≤150 points plus a score reduction of at least 70 points. Yet, almost all the trials included in our study were multinational trials, and in only one trial25 was the small size a possible cause of false positive result. It is possible that other factors that are not available from published papers may be associated with the placebo remission rates and thus influence the heterogeneity. For example, the prior response to corticosteroid or other therapy or the smoking status of patients receiving placebo may be related to spontaneous remission in the short term.

image

Figure 4.  Rates of remission in controls of 28 biological trials using placebo as ‘add-on treatment’, according to time of primary outcome evaluation. Circle’s dimension is proportional to placebo arm size. bsl00033 Trial first author when remission rate outweighs 25% or time is higher than 12 weeks.

Download figure to PowerPoint

The pooled estimate of response rate in our study was 33% by considering the 24 trials reporting data on this outcome. In the 18 trials, using the more stringent criterion of ≥100 points, the response was 30%. In Su’s analysis, the response rate was 19% in the eight studies containing data with response defined as a ≥100 points decrease in the CDAI score. We speculate that the higher response rate in placebo group in our study could depend on concomitant therapies to which placebo was added in recent trials of biological therapies. When a 70-point reduction in CDAI score was considered a response criterion, patients responded better to the placebo (average response rate was around 40%), indicating that less restrictive criteria allow a more favourable response, but could be prone to phenomena such as regression to mean. The variability in the definition of response complicates the interpretation of results of our meta-analysis. The at least 100-point reduction in CDAI score definition is preferable in evaluating response rates,43 considering that in a majority of patients with moderate-severe disease (CDAI from 220 to 400), a reduction of only 70 points in CDAI score would not correspond to a real clinical benefit. Anyway, the response rate of at least 30% in control arms, mainly in the large and more recent trials, suggests that the sample of future trials needs to be sized according to this baseline expected rate. Consider as an example that if the expected placebo response rate (≥70 points) is 20%, the required sample size (power 90%, delta + 15%) is 181 patients per group. In a large trial of natalizumab,38 the sample size increased up to 231 patients per group under similar circumstances (power 90%, delta + 15%) when the placebo rate was considered 40%.

We found high heterogeneity (Q-test: P < 0.0001) both in response and in remission rates. This heterogeneity remained statistically significant also after stratifying across various categories of variables. It has been suggested that use of metaregression should be restricted to investigation of differences between trials that relate to trial features and patient characteristics that vary substantially across trials and not within trials, when these features have been prespecified and many trials are available.44 Moreover, to provide a way of investigating patient characteristics, it is necessary to move to inspection of relations within trials and to compare subgroups of patients within every trial and then combine these results over trials. Because of lack of subgroup data from each trial, it was impossible to perform this kind of analysis in our study. In addition, even the suggestions derived from metaregression of aggregated trial level data should be considered exploratory. In fact, aggregation bias has the potential to cause spurious results when covariates relating to average patient characteristics, such as age or weight, are used. Hence, caution is required. Our results are not directly comparable with Su et al.’s meta-analysis7 that looked at all trials on active disease starting from 1966, including various drugs such as mesalamine and immunosuppressive therapies. In that meta-analysis, the trials included a larger spectrum of disease activity than in our analysis: the median CDAI score at baseline was 256 (range 190–287) in the trials testing therapies other than biological and 278 (range 253–311) in the subgroup of the biological trials. The Su’s study showed an inverse relationship between CDAI score at entry and remission rate, and a positive association between CDAI score at entry and response rate. These opposite findings were explained suggesting that patients with more severe disease may be more likely to have spontaneously a decrease in CDAI score of 70–100 points with placebo alone than patients with less severe disease. We did not find an association between CDAI at baseline and response or remission rates in spite of the fact that median CDAI score at baseline was typical of severe disease. A meta-analysis based on individual patient data would have been more appropriate, although not feasible, to manage the large variability underlying the use of the CDAI score both as a tool to characterize patients at entry and as a measure of outcome.

In summary, our study shows large heterogeneity both in response and in remission rate in the placebo used as ‘add-on’ treatment comparator in trials of biological therapy for active CD. The response rate of at least 30% in these control arms, mainly in the large and more recent trials, suggests that the sample of future trials needs to be sized according to this baseline expected rate. A better characterization of patients in terms of disease activity maybe combining CDAI with objective markers of inflammation such as CRP levels or with endoscopic data on mucosal appearance, and in terms of adequate description of concomitant therapies could be helpful from a clinical point of view. Moreover, other aspect of design should be improved, particularly those regarding all the blinding aspects of the trial and the agreement among outcomes assessors. The prevention of these sources of variability is a future effort for both clinicians and study sponsors.

Acknowledgement

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgement
  8. References
  9. Appendices

Declaration of personal and funding interests: None.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgement
  8. References
  9. Appendices

Appendices

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgement
  8. References
  9. Appendices

Appendix 1

((“crohn disease”[TIAB] NOT Medline[SB]) OR “crohn disease”[MeSH Terms] OR crohn’s disease[Text Word]) AND (randomized controlled trial [Publication Type] OR (randomized [Title/Abstract] AND controlled [Title/Abstract] AND trial [Title/Abstract]) AND (“humans”[MeSH Terms] AND English [la] AND “adult”[MeSH Terms]) NOT (“letter”[pt] OR editorial [pt] OR comment [pt])) AND (“1997/01/01”[EDAT] : “2007/12/31”[EDAT])

Appendix 2

Excluded papers (number)

Maintenance therapy (nine).

  • Rutgeerts P, D’Haens G, Targan S, Vasiliauskas E, Hanauer SB, Present DH, Mayer L, Van Hogezand RA, Braakman T, DeWoody KL, Schaible TF, Van Deventer SJ. Efficacy and safety of retreatment with anti-tumor necrosis factor antibody (infliximab) to maintain remission in Crohn’s disease. Gastroenterology 1999; 117: 761–9.

  • Hanauer SB, Feagan BG, Lichtenstein GR, Mayer LF, Schreiber S, Colombel JF, Rachmilewitz D, Wolf DC, Olson A, Bao W, Rutgeerts P, ACCENT I Study Group. Maintenance infliximab for Crohn’s disease: the ACCENT I randomised trial. Lancet 2002; 359: 1541–9.

  • Feagan BG, Yan S, Bala M, Bao W, Lichtenstein GR. The effects of infliximab maintenance therapy on health-related quality of life. Am J Gastroenterol 2003; 98: 2232–8.

  • Sands BE, Anderson FH, Bernstein CN, Chey WY, Feagan BG, Fedorak RN, Kamm MA, Korzenik JR, Lashner BA, Onken JE, Rachmilewitz D, Rutgeerts P, Wild G, Wolf DC, Marsters PA, Travers SB, Blank MA, van Deventer SJ. Infliximab maintenance therapy for fistulizing Crohn’s disease. N Engl J Med 2004; 350: 876–85.

  • Rutgeerts P, Diamond RH, Bala M, Olson A, Lichtenstein GR, Bao W, Patel K, Wolf DC, Safdi M, Colombel JF, Lashner B, Hanauer SB. Scheduled maintenance treatment with infliximab is superior to episodic treatment for the healing of mucosal ulceration associated with Crohn’s disease. Gastrointest Endosc 2006; 63: 433–42.

  • Sands BE, Blank MA, Diamond RH, Barrett JP, Van Deventer SJ. Maintenance infliximab does not result in increased abscess development in fistulizing Crohn’s disease: results from the ACCENT II study. Aliment Pharmacol Ther 2006; 23: 1127–36.

  • Colombel JF, Sandborn WJ, Rutgeerts P, Enns R, Hanauer SB, Panaccione R, Schreiber S, Byczkowski D, Li J, Kent JD, Pollack PF. Adalimumab for maintenance of clinical response and remission in patients with Crohn’s disease: the CHARM trial. Gastroenterology 2007; 132: 52–65.

  • Sandborn WJ, Hanauer SB, Rutgeerts P, Fedorak RN, Lukas M, MacIntosh DG, Panaccione R, Wolf D, Kent JD, Bittle B, Li J, Pollack PF. Adalimumab for maintenance treatment of Crohn’s disease: results of the CLASSIC II trial. Gut 2007; 56: 1232–9.

  • Schreiber S, Khaliq-Kareemi M, Lawrance IC, Thomsen OØ, Hanauer SB, McColm J, Bloomfield R, Sandborn WJ; PRECISE 2 Study Investigators. Maintenance therapy with certolizumab pegol for Crohn’s disease. N Engl J Med 2007; 357: 239–50.

Fistulizing disease (five).

  • Present DH, Rutgeerts P, Targan S, Hanauer SB, Mayer L, van Hogezand RA, Podolsky DK, Sands BE, Braakman T, DeWoody KL, Schaible TF, van Deventer SJ. Infliximab for the treatment of fistulas in patients with Crohn’s disease. N Engl J Med 1999; 340: 1398–405.

  • Gao Q, Hogezand RA, Lamers CB, Verspaget HW. Basic fibroblast growth factor as a response parameter to infliximab in fistulizing Crohn’s disease. Aliment Pharmacol Ther 2004; 20: 585–92.

  • Sands BE, Blank MA, Patel K, van Deventer SJ, ACCENT II Study. Long-term treatment of rectovaginal fistulas in Crohn’s disease: response to infliximab in the ACCENT II Study. Clin Gastroenterol Hepatol 2004; 2: 912–20.

  • West RL, van der Woude CJ, Hansen BE, Felt-Bersma RJ, van Tilburg AJ, Drapers JA, Kuipers EJ. Clinical and endosonographic effect of ciprofloxacin on the treatment of perianal fistulae in Crohn’s disease with infliximab: a double-blind placebo-controlled study. Aliment Pharmacol Ther 2004; 20: 1329–36.

  • Lichtenstein GR, Yan S, Bala M, Blank M, Sands BE. Infliximab maintenance treatment reduces hospitalizations, surgeries, and procedures in fistulizing Crohn’s disease. Gastroenterology 2005; 128: 862–9.

Other outcome (14).

  • Baert FJ, D’Haens GR, Peeters M, Hiele MI, Schaible TF, Shealy D, Geboes K, Rutgeerts PJ. Tumor necrosis factor alpha antibody (infliximab) therapy profoundly down-regulates the inflammation in Crohn’s ileocolitis. Gastroenterology 1999; 116: 22–8.

  • D’haens G, Van Deventer S, Van Hogezand R, Chalmers D, Kothe C, Baert F, Braakman T, Schaible T, Geboes K, Rutgeerts P. Endoscopic and histological healing with infliximab anti-tumor necrosis factor antibodies in Crohn’s disease: a European multicenter trial. Gastroenterology 1999; 116: 1029–34.

  • Dejaco C, Reinisch W, Lichtenberger C, Waldhoer T, Kuhn I, Tilg H, Gasche C. In vivo effects of recombinant human interleukin-10 on lymphocyte phenotypes and leukocyte activation markers in inflammatory bowel disease. J Investig Med 2000; 48: 449–56.

  • Colombel JF, Rutgeerts P, Malchow H, Jacyna M, Nielsen OH, Rask-Madsen J, Van Deventer S, Ferguson A, Desreumaux P, Forbes A, Geboes K, Melani L, Cohard M. Interleukin 10 (Tenovil) in the prevention of postoperative recurrence of Crohn’s disease. Gut 2001; 49: 42–6.

  • Hommes D, van den Blink B, Plasse T, Bartelsman J, Xu C, Macpherson B, Tytgat G, Peppelenbosch M, Van Deventer S. Inhibition of stress-activated MAP kinases induces clinical improvement in moderate to severe Crohn’s disease. Gastroenterology 2002; 122: 7–14.

  • Lichtenstein GR, Bala M, Han C, DeWoody K, Schaible T. Infliximab improves quality of life in patients with Crohn’s disease. Inflamm Bowel Dis. 2002; 8: 237–43.

  • Tilg H, Ulmer H, Kaser A, Weiss G. Role of IL-10 for induction of anemia during inflammation. J Immunol 2002; 169: 2204–9.

  • Lichtenstein GR, Yan S, Bala M, Hanauer S. Remission in patients with Crohn’s disease is associated with improvement in employment and quality of life and a decrease in hospitalizations and surgeries. Am J Gastroenterol 2004; 99: 91–6.

  • Feagan BG, Bala M, Yan S, Olson A, Hanauer S. Unemployment and disability in patients with moderately to severely active Crohn’s disease. J Clin Gastroenterol 2005; 39: 390–5.

  • Yacyshyn BR, Schievella A, Sewell KL, Tami JA. Gene polymorphisms and serological markers of patients with active Crohn’s disease in a clinical trial of antisense to ICAM-1. Clin Exp Immunol 2005; 141: 141–7.

  • Geboes K, Rutgeerts P, Opdenakker G, Olson A, Patel K, Wagner CL, Marano CW. Endoscopic and histologic evidence of persistent mucosal healing and correlation with clinical improvement following sustained infliximab treatment for Crohn’s disease. Curr Med Res Opin 2005; 21: 1741–54.

  • Louis EJ, Watier HE, Schreiber S, Hampe J, Taillard F, Olson A, Thorne N, Zhang H, Colombel JF. Polymorphism in IgG Fc receptor gene FCGR3A and response to infliximab in Crohn’s disease: a subanalysis of the ACCENT I study. Pharmacogenet Genomics 2006; 16: 911–4.

  • Gao Q, Meijer MJ, Schlüter UG, van Hogezand RA, van der Zon JM, van den Berg M, van Duijn W, Lamers CB, Verspaget HW. Infliximab treatment influences the serological expression of matrix metalloproteinase (MMP)-2 and -9 in Crohn’s disease. Inflamm Bowel Dis 2007; 13: 693–704.

  • Slonim AE, Bulone L, Damore MB, Goldberg T, Wingertzahn MA, McKinley MJ. A preliminary study of growth hormone therapy for Crohn’s disease. N Engl J Med 2000; 342: 1633–7.

  • Schreiber S, Feagan B, D’Haens G, Colombel JF, Geboes K, Yurcov M, Isakov V, Golovenko O, Bernstein CN, Ludwig D, Winter T, Meier U, Yong C, Steffgen J, BIRB 796 Study Group. Oral p38 mitogen-activated protein kinase inhibition with BIRB 796 for active Crohn’s disease: a randomized, double-blind, placebo-controlled trial. Clin Gastroenterol Hepatol 2006; 4: 325–34.

Steroid-free remission (four).

  • Schreiber S, Nikolaus S, Malchow H, Kruis W, Lochs H, Raedler A, Hahn EG, Krummenerl T, Steinmann G, German ICAM-1 Study Group. Absence of efficacy of subcutaneous antisense ICAM-1 treatment of chronic active Crohn’s disease. Gastroenterology 2001; 120: 1339–46.

  • Yacyshyn BR, Chey WY, Goff J, Salzberg B, Baerg R, Buchman AL, Tami J, Yu R, Gibiansky E, Shanahan WR, ISIS 2302-CS9 Investigators. Double blind, placebo controlled trial of the remission inducing and steroid sparing properties of an ICAM-1 antisense oligodeoxynucleotide, alicaforsen (ISIS 2302), in active steroid dependent Crohn’s disease. Gut 2002; 51: 30–6.

  • Feagan BG, Sandborn WJ, Baker JP, Cominelli F, Sutherland LR, Elson CO, Salzberg BA, Archambault A, Bernstein CN, Lichtenstein GR, Heath PK, Cameron S, Hanauer SB. A randomized, double-blind, placebo-controlled trial of CDP571, a humanized monoclonal antibody to tumour necrosis factor-alpha, in patients with corticosteroid-dependent Crohn’s disease. Aliment Pharmacol Ther 2005; 21: 373–84.

  • Feagan BG, Sandborn WJ, Lichtenstein G, Radford-Smith G, Patel J, Innes A. CDP571, a humanized monoclonal antibody to tumour necrosis factor-alpha, for steroid-dependent Crohn’s disease: a randomized, double-blind, placebo-controlled trial. Aliment Pharmacol Ther 2006; 23: 617–28.

Lack of a placebo arm (five).

  • Yacyshyn BR, Barish C, Goff J, Dalke D, Gaspari M, Yu R, Tami J, Dorr FA, Sewell KL. Dose ranging pharmacokinetic trial of high-dose alicaforsen (intercellular adhesion molecule-1 antisense oligodeoxynucleotide) (ISIS 2302) in active Crohn’s disease. Aliment Pharmacol Ther 2002; 16: 1761–70.

  • Rutgeerts P, Lemmens L, Van Assche G, Noman M, Borghini-Fuhrer I, Goedkoop R. Treatment of active Crohn’s disease with onercept (recombinant human soluble p55 tumour necrosis factor receptor): results of a randomized, open-label, pilot study. Aliment Pharmacol Ther 2003; 17: 185–92.

  • Schröder O, Blumenstein I, Stein J. Combining infliximab with methotrexate for the induction and maintenance of remission in refractory Crohn’s disease: a controlled pilot study. Eur J Gastroenterol Hepatol 2006; 18: 11–6.

  • Herrlinger KR, Witthoeft T, Raedler A, Bokemeyer B, Krummenerl T, Schulzke JD, Boerner N, Kueppers B, Emmrich J, Mescheder A, Schwertschlag U, Shapiro M, Stange EF. Randomized, double blind controlled trial of subcutaneous recombinant human interleukin-11 vs. prednisolone in active Crohn’s disease. Am J Gastroenterol 2006; 101: 793–7.

  • Sands BE, Kozarek R, Spainhour J, Barish CF, Becker S, Goldberg L, Katz S, Goldblum R, Harrigan R, Hilton D, Hanauer SB. Safety and tolerability of concurrent natalizumab treatment for patients with Crohn’s disease not in remission while receiving infliximab. Inflamm Bowel Dis 2007; 13: 2–11.

Ulcerative colitis or RA.

  • Maksymowych WP, Blackburn WD Jr, Tami JA, Shanahan WR Jr. A randomized, placebo controlled trial of an antisense oligodeoxynucleotide to intercellular adhesion molecule-1 in the treatment of severe rheumatoid arthritis. J Rheumatol 2002; 29: 447–53.

  • Su C, Salzberg BA, Lewis JD, Deren JJ, Kornbluth A, Katzka DA, Stein RB, Adler DR, Lichtenstein GR. Efficacy of anti-tumor necrosis factor therapy in patients with ulcerative colitis. Am J Gastroenterol 2002; 97: 2577–84.

  • Probert CS, Hearing SD, Schreiber S, Kühbacher T, Ghosh S, Arnott ID, Forbes A. Infliximab in moderately severe glucocorticoid resistant ulcerative colitis: a randomised controlled trial. Gut 2003; 52: 998–1002.

  • Armuzzi A, De Pascalis B, Lupascu A, Fedeli P, Leo D, Mentella MC, Vincenti F, Melina D, Gasbarrini G, Pola P, Gasbarrini A. Infliximab in the treatment of steroid-dependent ulcerative colitis. Eur Rev Med Pharmacol Sci 2004; 8: 231–3.

  • Rutgeerts P, Sandborn WJ, Feagan BG, Reinisch W, Olson A, Johanns J, Travers S, Rachmilewitz D, Hanauer SB, Lichtenstein GR, de Villiers WJ, Present D, Sands BE, Colombel JF. Infliximab for induction and maintenance therapy for ulcerative colitis. N Engl J Med 2005; 353: 2462–76.