If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
UMR 5558 CNRS, Equipe EMET, Université Claude Bernard Lyon 1, Lyon, FranceCentre de ressources et de compétences de la mucoviscidose, Hospices Civils de Lyon, Lyon, France
UMR 5558 CNRS, Equipe EMET, Université Claude Bernard Lyon 1, Lyon, FranceCentre de ressources et de compétences de la mucoviscidose, Hospices Civils de Lyon, Lyon, France
We assessed placebo response (PR) in patients with CF participating in randomised clinical trials (RCTs).
•
PR was assessed on three relevant outcome measures used in RCTs in cystic fibrosis: FEV1, CFQR and BMI.
•
No evidence was found of PR on FEV1 and CFQR; however, a small but significant PR was found on BMI.
Abstract
Background
Patients' and families' expectation that a cure for cystic fibrosis (CF) will be found is high. In other debilitating conditions, high expectation has been shown to drive a strong placebo response (PR). Therefore, our goal was to evaluate PR on objective continuous outcomes (FEV1, BMI) and the CF Questionnaire Revised-Respiratory Domain (CFQR-RD) monitored during randomised clinical trials (RCTs) for CF.
Methods
We conducted a meta-analysis after a systematic review of the literature carried out to identify RCTs with FEV1, CFQR-RD and BMI as outcome measures. The standardised mean difference (SMD) was calculated to estimate the PR. A meta-regression analysis was conducted to assess other contributing factors on PR such as study design, trial duration, patient age and disease severity.
Results
Out of 289 RCTs found in the search, we identified 61 articles (published from 1987 to 2017) with respectively 59, 17 and 9 reporting FEV1, CFQR-RD and BMI at the start and at the end of the RCTs. No significant PR was found on FEV1 or CFQR-RD. However, a small but significant PR was found on BMI SMD, 0.09 (95% CI (0.01; 0.17); p = 0.03).
Conclusion
The PR seems higher when measuring BMI. However, it is not clear whether this improvement can be explained by a PR alone.
], with possible benefit and improvement of symptoms. While some evidence illustrates that a true placebo effect is biologically modulated by neurotransmitters [
Since 1955 with Beecher's statement on the “true” placebo effect, this term has regularly been misinterpreted and confounded with the “perceived placebo effect” or the “placebo response” [
In CF, patients and family's expectations may interfere with the PR on several of the above listed factors. However, a systematic evaluation of PR in CF has never been addressed [
]. This may be of importance for clinicians to better determine the “true” magnitude of the clinical benefit they may expect for their patients. This may also be important for CF researchers for methodological purposes (power calculation, study design, outcome measure selection).
The aim of this study was therefore to determine the PR based on three continuous outcomes considered as particularly relevant in CF: respiratory function measured with forced expiratory volume in one second (FEV1), quality of life with the respiratory domain of the Cystic-Fibrosis Questionnaire Revised (CFQR-RD) and nutritional status with body mass index (BMI).
2. Material and methods
2.1 Literature search
We performed a literature search using PubMed (US National Library of Medicine, Bethesda, MD, USA) and the Cochrane Library (John Wiley and Sons, Chichester, UK) focusing on placebo-controlled RCTs in patients with CF. The last bibliographic search was done on December 12th, 2018. We used the following terms: “placebo AND cystic fibrosis AND randomised controlled trial” as well as “cystic fibrosis AND placebo” and filtered the type of study (“clinical trial” for PubMed and “trial” for the Cochrane Library).
2.2 Selection of meta-analyses
Criteria for inclusion were randomised double-blind placebo-controlled trials in patients with CF of any age and without a lower limit for the date of publication. The age limit between adults and children was set at 18 years old. Eligible interventions were all pharmacological treatments excluding homeopathic treatments, specific diets and vitamin supplementation.
Our research was restricted to studies published in English or French.
2.3 Data extraction
For each study included, the following information was extracted and entered in the database: [
] change from baseline to the end of the study for three continuous outcomes in the placebo and treatment arms: FEV1, BMI,health-related quality of life outcomes with the respiratory domain of the CF questionnaire revised (CFQR-RD), [
We classified pharmacological interventions during RCTs into one of the five drug categories (the first three being the most frequently explored in RCTs in CF): pulmonary (P), nutrition (N), microbiology/anti-infective (M), basic defect (BD) and other (O).
2.6 Outcome measures
We extracted the change from the start (participant characteristics at study entry) to the end of the trial (even if it did not correspond to the time point evaluation of the study's primary endpoint) for the three continuous outcomes most commonly used in CF RCTs: FEV1, BMI and CFQR-RD. FEV1 and BMI were considered as “objective” outcome measures and the CFQR-RD as a continuous “subjective” outcome measure.
2.7 Dealing with missing data
Since we considered continuous outcomes, when the standard deviation (SD) was missing, we estimated it from the standard error (SE) or confidence interval (CI) [
The PR was defined as the difference in the outcome measured in the patients of the placebo arm between baseline and the “end-of-study” time points. To anticipate heterogeneity in the continuous data reporting (FEV1, BMI and CFQR-RD), we calculated the standardised mean difference (SMD) for each outcome instead of the MD. A positive SMD value indicates an improvement under placebo and inversely for a negative SMD value.
Since heterogeneity was expected, a meta-analytic random effects model (inverse variance method) was used, rather than a fixed-effects model [
]. The heterogeneity of the SMD across the studies was assessed using the I2 statistical test (which can be interpreted as the proportion of the observed discrepancy in the estimation of the effect, within a group of trials, which cannot be accounted for by random variation) [
] design of the study (cross-over design coded 0 and parallel design coded 1). A QE-test was performed to assess residual heterogeneity when moderators were included. QM was the statistical test for omnibus test coefficients. The coefficients were expressed using the β letter. All analyses were performed with R (R-studio Inc.; Version 3.4.4; https://www.r-project.org/).
3. Results
3.1 Description of studies
We identified 1417 reviews. After screening the titles and abstracts, and the exclusion of irrelevant and duplicate studies, 250 reviews were screened (Fig. 1). Sixty-one RCTs (from 1987 to 2017) were finally analysed (Table S1 supplemental material). Respectively 59, 17 and 9 RCTs reported results for FEV1, CFQR-RD and BMI.
There were 58 trials with a parallel design and three with a cross-over design. When the literature search was conducted, there were 29, 14, 12, 3 and 3 RCTs categorised into the pulmonary, microbiology/anti-infective, basic defect, nutrition and “other” categories, respectively. There was a low risk of bias (Cochrane assessment: 1 and Jadad score between 4 and 5)for 29 RCTs (47.5%). It remained undetermined for the others (Cochrane assessment: 2).
Concomitant treatments were specified in 46 RCTs (75.4%). Adverse effects (of any type) were reported in 32 studies (52.5%) with no significant difference between the placebo and treatment arms (p > 0.05). Placebo arms contained 4648 patients (2242 males) and the treatment arms included 4917 patients (53.9% males). The mean age in the placebo arm was 19.3 (range, 2.3–32.7) years. The mean trial duration was 207.8 days.
3.2 PR evaluated on FEV1
PR SMD was estimated at −0.16 in a random effect model (95% CI (−0.24; −0.08); p < 0.0001) (Fig. 2), indicating a trend toward deterioration of FEV1 in the placebo group. A significant heterogeneity across studies was identified (I2 = 81.9%, Q(df = 58) = 319.16, p < 0.0001). The funnel plot was not asymmetrical (Fig. S1-A; supplemental material).
Fig. 2Forest plot of placebo responses evaluated on FEV1.
Individual standardised mean differences (SMD) were calculated for each study and are indicated separately on each line. A positive SMD value indicates an improvement under placebo and a negative SMD value a deterioration under placebo. Overall perceived placebo effect SMD was estimated at −0.16 in a random effect model (95% CI, 0.24; −0.08); p = 0.0002, indicating a trend toward deterioration of FEV1 under placebo arm.
Univariate meta-regression was then performed to assess the influence of disease or study-related factors on PR assessed on FEV1(Table 1). Year of publication did not affect FEV1 in the placebo group (QM (df = 1) = 2.58, β = 0.01, p = 0.1), nor did age of the participants at inclusion (QM (df = 1) = 0.23, β = −0.003, p = 0.63). The PR on FEV1 did not differ between adults and children (QM (df = 1) = 0.23; p = 0.63), nor did trial duration (QM (df = 1) = 0.02, β = 0 p = 0.88) as well as the type of the intervention (QM (df = 4) = 1.63, p = 0.80) influence PR. Given that the number of studies varied between parallel group (n = 56) and cross-over studies (n = 3), it was not possible to evaluate the influence of study design on PR on FEV1. Finally, patients' FEV1 baseline value did not influence PR (QM (df = 2) = 2.68, p = 0.26).
Table 1Univariate meta-regression of the potential influence of trial- and patient-related factors on perceived placebo effect assessed through FEV1, CFQR-RD and BMI in RCTs conducted in patients with CF.
Variables (k = number of studies)
Categorical data QM (df) (p-value)
Continuous data β (p-value)
FEV1
Year of publication (k = 59)
–
0.01 (0.1)
Trial duration (k = 59)
–
0 (0.88)
Age (k = 57)
–
−0.003 (0.63)
CF lung disease severity
2.68 (df = 2) (0.26)
–
Classification of drug
1.63 (df = 4) (0.8)
–
Trial design
NP
–
CFQR-RD
Year of publication (k = 17)
–
0.007 (0.93)
Trial duration (k = 17)
–
−0.002(<0.0001)
Age (k = 17)
–
0.04 (0.0001)
CF lung disease severity
2.76 (df = 2) (0.25)
–
Classification of drug
1.13 (df = 2) (0.57)
–
Trial design
NP
BMI
NP
NP
Univariate meta-regression analysis was used to evaluate the influence of the above factors on PPE through FEV1 and CFQR-RD: year of publication, trial duration, age, lung disease severity and classification. Trial design could not be integrated into the meta-regression. The meta-regression could not be performed for BMI because the number of available studies was under 10 (k = 9).
Abbreviations: k corresponds to the number of available trials for the outcome of interest. β corresponds to the coefficient of meta-regression for each continuous variable tested. NP: not performed.
The overall SMD for CFQR-RD was estimated at −0.11 (95% CI (−0.34; 0.11); p = 0.32) (Fig. 3). Wide heterogeneity across studies was found (I2 = 93.6%, p < 0.0001). The funnel plot was not asymmetrical (Fig. S1 B, supplemental material).
Fig. 3Forest plot of placebo responses evaluated on CFQR-RD.
Perceived placebo effect standardised mean difference (SMD) was estimated to −0.11 (95% CI, (−0.34; 0.11); p = 0.32). It was statistically non-significant, indicating an absence of PPE on this outcome measure.
Using univariate meta-regression(Table 1), a greater PR was observed on CFQR-RD in older patients (QM (df = 1) = 16.9, β = 0.04, p-value <0.0001) with one outlier which appeared to drive the effect. Once removed, the effect of age was no longer significant (QM (df = 1) = 0.97, β = −0.009, p-value = 0.32). PR assessed on CFQR-RD did not differ between adults and children ((QM (df = 1) = 0.89, p = 0.34)). Year of publication (QM (df = 1) = 0.007, β = 0.003, p = 0.93), the type of intervention (QM (df = 2) = 1.13, p-value = 0.57), patients' baseline FEV1 (QM (df = 2) = 2.76, p-value = 0.25) did not influence PR assessed on CFQR-RD. Trial duration was found to influence PR assessed on CFQR-RD as well (QM (df = 1) = 79.7, β = −0.002, p-value <0.0001). The longer the trial duration, the more the CFQR-RD deteriorated in the placebo group. As observed with age, an outlier drives this effect, since after removal the result was no longer significant (QM (df = 1) = 0.80, β = 0.0006, p-value = 0.37). Finally, assessing study design on PR using the CFQR-RD was not possible given the low number of trials in each group.
3.4 PR evaluated on BMI
The SMD assessed on BMI was estimated at 0.09 in a random effect model (95% CI (0.01; 0.17); p = 0.03), indicating a trend toward improvement of BMI in the placebo group (Fig. 4).
Fig. 4Forest plot of placebo responses evaluated on BMI.
The funnel plot was not asymmetrical (Fig. S1 C supplemental material). Because of the small number of RCTs reporting BMI, we were unable to perform meta-regression to explore the contribution of other factors such as age at inclusion, study design or the type of intervention. Moreover, we were unable to analyse data form children and adults apart, because BMI results were not given separately.
4. Discussion
To our knowledge, this is the first meta-analysis to assess PR in patients with CF investigated in RCTs. The research question behind this work may have implications on the interpretation of the therapeutic effect of past, ongoing and future RCTs for both clinicians and CF researchers.
PR is the combination of the true placebo effect and other factors that may alter the response measured on certain outcomes in patients under the placebo arm of a RCT [
]. In the present study, no PR difference was found in patients with CF when assessing continuous outcomes such as FEV1 and CFQR-RD. However, a weak but statistically significant PR was found on BMI.
We conducted this meta-analysis on PR by choosing outcomes that were commonly reported and the most relevant regarding CF. CFQR-RD, FEV1 and BMI are three continuous variables largely used in RCTs and the two latter outcomes in CF clinics. They explore the three most important dimensions of CF disease (i.e. CF-related lung disease with FEV1; the patient's quality of life with CFQR-RD and nutritional status with BMI) [
]. Despite the limitations of both FEV1 and BMI in properly tracking a therapeutic effect in some patients, particularly the youngest patients whose FEV1 and BMI may be within normal ranges, they remain the outcomes on which clinicians, the FDA and the European Medicine Agency base their decisions to assess the therapeutic effect of an intervention.
We found that there was no evidence of a PR in patients with CF when looking at FEV1 or CFQR-RD. Both tended to deteriorate between the start and the end of the trials. We found that FEV1 decreased in the placebo group during RCTs independently of patient- or trial- related factors. With the CFQR-RD the deterioration in the placebo group was influenced by the patient's age and the trial duration mainly because of an outlier trial. These results likely reflect both the progressively deteriorating nature of the CF but also a possible regression to the mean. The genetic origin of the disease and the current standard of care, which mainly treats symptoms, explain that CF remains a slowly progressive medical condition without potential for remission [
]. If the deterioration of FEV1 and CFQR-RD had been mainly driven by the disease progression, a “time-dependent” deterioration would have been found. This was not the case, and the meta-regression analysis showed these two outcomes were not impacted by trial duration. We therefore believe that it reflects the regression to the mean of FEV1 and CFQR-RD. At the start of the trial, it is likely that patients are selected at their best clinical condition and “regress” to their usual (mean) outcome measures. Regression to the mean is a well-known factor explaining PR and needs to be considered in order to properly interpret the results observed in placebo arms [
However, a small (SMD 0.09; 95% CI 0.01–0.17) but statistically significant PR was detected on BMI. By comparison, but in a very different pathological condition, a PR was observed in young patients with intellectual deficiency with a SMD of 0.468 (SE: 0.150; p = 0.002) [
]. Patients with CF are more at risk of stunted growth with low BMI. Our results seem to indicate that patients with CF tended to improve their BMI (i.e. nutritional status) in the placebo group during RCTs. It is not clear whether this improvement can be explained by a PR alone. There are several other reasons explaining that patients improve their BMI during RCTs. Firstly, a 0.09 standardised mean difference on BMI between the two arms of an RCT indicates a very small absolute change in weight between the two groups of patients. Secondly, the improvement of BMI in the placebo group may also reflect [
Classical conditioning and expectancy in placebo hypoalgesia: a randomized controlled study in patients with atopic dermatitis and persons with healthy skin.
]. Retrospectively, it was not possible to distil out true placebo effect from these other factors. Because of the low number of trials included in the meta-regression analysis, we were also unable to explore a number of important contributing factors, particularly age at study entry and the class of the investigational drug tested. Regarding age, several groups have reported that the PR was more pronounced in children suffering neurological or neurodevelopmental conditions [
]. However, it seems from our group that the magnitude of the PR of children is essentially based on disease, age, study design and the outcome studied [
Improving our knowledge on placebo responses in patients with CF may have pragmatic implications for both clinicians and CF trialists: for clinicians, when looking at the results of RCTs and in the perspective of using the tested drug in their patients, to better determine the magnitude of the therapeutic effect they may expect in real life; for CF researchers, this may be of importance for outcome selection, power calculations and study design when using outcomes potentially submitted to placebo responses. Using a “placebo-run-in-period” during RCTs could be useful but it may overestimate the therapeutic benefit [
The potential influence of PR on BMI in patients with CF may deserve attention even if BMI is not usually used as a primary endpoint in CF RCTs. For 10 years, the basic defect of CFTR can be partially restored using CFTR potentiators, amplifiers and activators, alone or in combination [
]. More than a dozen RCTs using CFTR modulators have been experimented in patients with CF to date. The results of these RCTs have consistently shown an improvement in patients' BMI while sometimes showing a less convincing functional respiratory benefit when looking at FEV1 changes. The higher “nutritional” benefit can be questioned, and several possible explanations have already been discussed above. Moreover, it is uncertain whether the observed PR for BMI would be consistent across different age groups and how it can be translated in clinic to accurately evaluate clinical benefit.
The main strengths of this study are the originality of the research question and the rigorous method of meta-analysis and meta-regression conducted after an exhaustive literature search. There are a number of limitations, however. Firstly, despite being exhaustive, the number of RCTs available for analysis was relatively low despite the high number of RCTs conducted in patients with CF to date. Indeed, a significant number of RCTs (65 RCTs with missing data at the start and/or at the end of the study and an additional 63 RCTs that reported data as abstracts only) could not be included in the final analysis. Secondly, other respiratory outcome such as pulmonary exacerbation, which is an important patient-related outcome measure, should be explored because FEV1 alone does not capture the entire spectrum of CF lung disease. Unfortunately, this analysis was not possible because there were no data available at baseline, making the evaluation of the PR between the start and the end impossible.
In conclusion, this work indicates that patients undergoing RCTs may be submitted to a small but significant PR on BMI. It is not clear whether this improvement can be explained by a PR alone. This study emphasizes the importance of having appropriate control groups in clinical trials.
Funding
This research did not receive any specific grant from funding agencies in the public, commercial or not-for-profit sectors.
Perceived placebo effect standardised mean difference (SMD) was estimated at 0.09 in a random effects model (95% CI, 0.01; 0.17); p = 0.03, indicating a small but statistically significant improvement of BMI under placebo arm.
Figure S1: Funnel plot of standardised mean difference (SMD) for FEV1 (A), CFQR-RD (B) and BMI (C). Funnel plots were not asymmetrical, indicating no publication bias.
References
Schlangen M.
Reimann A.L.
Medical needs of cystic fibrosis patients and policies for fair co-operation between small and middle-sized companies and patient organizations.
Classical conditioning and expectancy in placebo hypoalgesia: a randomized controlled study in patients with atopic dermatitis and persons with healthy skin.
30833-6&id=gr1.jpg){kind=link}
30833-6&id=gr2.jpg){kind=link}
30833-6&id=gr3.jpg){kind=link}
30833-6&id=gr4.jpg){kind=link}