Abstract
BACKGROUND Multiple System Atrophy (MSA) is a rapidly progressive and fatal neurodegenerative disease, with no effective treatment. Estimating the placebo and nocebo responses will help better design and interpret clinical trials.
Objective To estimate the placebo and nocebo responses in MSA and explore their determinants.
METHODS Electronic databases were searched up to November 2020. Randomized, blinded, placebo- or sham-controlled trials of patients with MSA were included if quantitative data were extractable on the placebo arm. The primary outcomes were: placebo response, defined as the within-group change from baseline, using any scale measuring motor outcomes; and nocebo response, defined as the proportion of patients experiencing adverse effects in the placebo arm. Random-effects meta-analyses were used to pool data. Several predetermined subgroup analyses and metaregressions were performed. PROSPERO registration number: CRD42021222915.
RESULTS We included 21 randomized controlled trials (614 participants). Pooled placebo response was an increase in the Unified MSA Rating Scale (UMSARS) parts I and II of 9.09 points (95% CI 7.78 to 10.31, I2=94.00%, 9 studies, 304 participants). Pooled nocebo response was 63,88% (CI 95% 41.15 to 84.05, I2 =93.03%, 13 studies, 331 participants). Both placebo and nocebo responses were greater in trials with longer duration, whereas nocebo response was also higher in studies testing pharmacological interventions when compared with non-pharmacological interventions.
CONCLUSIONS There may be a favorable response associated with the placebo, but this data needs to be compared with a “no treatment group” in order to validate its real impact. The nocebo response is high and should be considered in future clinical trial design and interpretation.
Introduction
Multiple System Atrophy (MSA) is a progressive and fatal neurodegenerative disease secondary to oligodendroglial cytoplasmatic inclusions of misfolded a-synuclein protein.1 Its hallmark manifestations are autonomic dysfunction with different degrees and combinations of parkinsonism, pyramidal signs or cerebellar ataxia.2 There is an unmet need for efficacious symptomatic and disease-modifying interventions for people with MSA; however clinical trials thus far have revealed unsatisfactory results.
The placebo effect is described as the impact of the expectation of receiving a therapeutic intervention and this effect may hinder signal detection in clinical trials3. However, no studies systematically evaluate its magnitude and characteristics in MSA. Here we describe the placebo and nocebo responses in randomized placebo-controlled trials of patients with MSA with the aim of informing future trial design and interpretation.
Methods
This report follows the PRISMA guidelines4 and the protocol was prospectively registered with PROSPERO (CRD42021222915). Full methodological details can be found in Supplementary Materials.
Eligibility criteria
Randomized controlled trials (RCTs), of parallel or crossover design, double- or single-blinded, studying the effects of therapeutic interventions in people with clinical diagnosis of MSA; trials had to be published or registered with results; all studies had to be placebo or sham controlled; participants had to be adults (≥ 18 years of age). Due to small study bias, studies including less than five participants with MSA were excluded. There were no restrictions on disease state, number of centers, language, setting, duration, or year of publication. Studies had to report quantitative data on at least one of the following outcomes within the placebo arm:
Primary Efficacy Outcome
“placebo effect response”, defined as the within-group change from baseline, using any rating scale measuring the motor domain.
Primary Safety Outcome
“nocebo effect response”, defined as the proportion of patients experiencing adverse effects (AE) in the placebo arm.
Secondary Efficacy Outcomes
within-group change from baseline, using any rating scale measuring the following domains: functional ability, autonomic features, quality of life, and patient or clinical subjective impression of change.
Secondary Safety Outcomes
proportion of withdrawals and of patients experiencing serious adverse effects (SAE) in the placebo arm.
Information Sources and Search
MEDLINE, Embaseand the Cochrane Register of Controlled Trials (CENTRAL) and NIH’s ClinicalTrials.gov from inception to November 2020. Search strategies are presented in Supplementary Materials.
Selection Process
Two independent reviewers screened references and assessed full-texts for eligibility. Disagreements were resolved with discussion or by a third author.
Data collection process
Two independent reviewers extracted data to a piloted electronic data extraction form that was cross-checked for accuracy.
Risk of bias in individual studies
We used the Cochrane Risk of Bias tool5. The following items were rated low, unclear or high risk of bias : random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, and selective reporting. Selective reporting was classified as high risk of bias if the study did not report on all our predefined outcomes. The study overall was classified as moderate if at least one item was at moderate risk, and as high risk if any item was deemed as high risk. Two authors independently assessed each domain and disagreements were solved by discussion or by a third author.
Statistical methods
Only placebo arm data were collected and analyzed. Placebo and nocebo data were derived from the last measured within-group difference or proportion in the placebo arm, respectively. Data was combined using random-effects meta-analyses techniques, namely the Dersimonian-Laird inverse-variance weighted model (Dersimonian and Laird 1986) for continuous and categorical variables. Permutational meta-regressions with 1,000 repetitions were used to explore association of the studied outcomes with continuous and categorical variables (i.e. meta-regression and subgroup analyses, respectively). Heterogeneity between results was assessed using the I2. Statistical package Stata 16.0 (Houston, Texas) was used and results are presented with a 95% confidence interval (95% CI).
Results
The electronic search returned 952 records, 45 references were subject to full text review, and 21 studies6–25 were included. Full search results, characteristics of included and excluded studies and risk of bias can be found in the Supplementary Materials.
Placebo responses
Primary outcome
Seventeen studies (574 patients) reported on MSA severity assessment scales, nine (304 patients) using Unified MSA Rating Scale (UMSARS) part I and II, and 8 (270 patients) using other tools. The overall pooled placebo response standardized mean difference was 1.04 (95% CI 0.59 to 1.49, n = 574, 17 studies; Supplementary Materials) and 9.09 UMSARS part I and II points (95% CI 7.78 to 10.31; 9 studies, n=304; Figure 1, Table 1) Subgroup analysis revealed a positive association with trial duration (Table 1). Statistical heterogeneity was not explained by the predefined subgroup and sensitivity analysis.
Forest plots for placebo (upper panel, UMSARS I & II) and nocebo (lower panel,) responses. ES, Effect size; 95% CI, 95% confidence interval.
Secondary outcomes
Four studies (250 participants) reported on Functional Ability, 3 (49 participants) reporting on the Unified Parkinson’s Disease Rating Scale (UPDRS) part II. The pooled placebo response was 2.52 UPDRS part II points (95% CI 0.45 to 4.61, n = 49, 3 studies; Supplementary Materials). Six studies (187 participants) reported on autonomic evaluations, however using different assessment techniques not possible to compare. Three studies (146 participants) used the Composite Autonomic Symptom Score (COMPASS), placebo response of 16.65 (95% CI 2.58 to 30.73; n = 146, 3 studies; Supplementary Materials). Not enough information was available to evaluate other domains. For the evaluated domains, planned subsequent analyses were not performed due to data sparsity.
Nocebo responses
Primary outcome
13 studies (331 patients) reported the proportion of patients experiencing AE in the placebo arm. Overall, 63,88% (95% CI 41.15 to 84.05 n=331, 13 studies; Figure 1 and Table 1) of patients experienced AE. This proportion was higher when comparing pharmacological to non-pharmacological interventions and positively associated with trial duration (Table 1). Statistical heterogeneity was not explained by the predefined subgroup and sensitivity analysis.
Secondary outcomes
13 studies (366 participants) reported on withdrawals and 10 studies (291 participants) on SAE. 13,32% (95% CI 7.54 to 20.13, n=366, 13 studies; Supplementary Materials) of the trial participants withdrew study participation and 13,80% (95% CI 5.77 to 24.02, n=291, 10 studies; Supplementary Materials) reported SAE. Meta-regression and subgroup analyses can be found in the Supplementary Materials.
Discussion
To our knowledge this work is the first systematic review evaluating the placebo and the nocebo responses in MSA. Our analyses showed a mean placebo response in MSA clinical trials of an increase (worsening) of 9.09 points in UMSARS part I and II. This effect is higher in studies with a longer duration. We also documented that 63.88% placebo-treated participants report AEs, with higher proportions in longer studies and in pharmacology studies.
This response in the placebo arm is expected, given the rapidly progressive nature of MSA. However, our results are limited by the available evidence, both in terms of the amount of eligible data and the absence of a “no treatment group”26. Using meta-regression we estimate an average annual increase of 11.44 points in UMSARS part I and II. Wenning et al. natural history cohort27 observed an annual worsening of 14.7 points, which among other explanations, could support the existence of a placebo response in our data (i.e., clinical trial participants on average progress lower than natural history study participants). If we assume this hypothesis to be true, the difference between the annualized rate of change in placebo arms and observational studies is larger than the estimated minimal clinical important decline for UMSARS part I and II of 3.0 points.28
The observed proportion of AE is aligned with previous studies and can also be confounded by the disease’s natural history. Expectedly, longer trials have a higher perceived dose exposure motivating larger nocebo responses.29 The differential effect noted in trials studying pharmacological interventions might be related with the follow-up duration as on average they are longer than non-pharmacological intervention trials.
There are limitations to our work. The reduced number of eligible studies limits our statistical power, weakening our ability to detect signals and the precision of the estimates. We refrained from including small studies and analyzing outcome domains with scarce data. Although our outcomes and analyses were predefined, we runned multiple analyses, increasing the risk of false positives. For these two reasons, secondary endpoints and subgroup or meta-regressions should be interpreted with caution. Cross-over studies were included, where a potential carry-over effect can overestimate the nocebo response. However we did not observe this effect on subgroup-analysis. Finally, the absence of data from prospectively and contemporaneous no-treatment-group arms precludes us from making assertive statements about placebo and nocebo effects, as these effects can be confounded by the Hawthorne effect (i.e. the effect of changing behaviors due to the awareness of being observed.30
We believe these observations could be of interest for planning and interpreting future clinical trials of people with MSA.
Data Availability
All data produced in the present work are contained in the manuscript
Bibliography
