Fax: (713) 792-6094
Placebo and nocebo effects in randomized double-blind clinical trials of agents for the therapy for fatigue in patients with advanced cancer
Article first published online: 13 NOV 2009
Copyright © 2009 American Cancer Society
Volume 116, Issue 3, pages 766–774, 1 February 2010
How to Cite
de la Cruz, M., Hui, D., Parsons, H. A. and Bruera, E. (2010), Placebo and nocebo effects in randomized double-blind clinical trials of agents for the therapy for fatigue in patients with advanced cancer. Cancer, 116: 766–774. doi: 10.1002/cncr.24751
- Issue published online: 25 JAN 2010
- Article first published online: 13 NOV 2009
- Manuscript Accepted: 13 MAY 2009
- Manuscript Revised: 1 MAY 2009
- Manuscript Received: 9 APR 2009
A significant response to placebo in randomized controlled trials of treatments for cancer-related fatigue (CRF) had been reported. A retrospective study was conducted to determine the frequency and predictors of response to placebo effect and nocebo effects in patients with CRF treated in those trials.
The records of 105 patients who received placebo in 2 previous randomized clinical trials conducted by this group were reviewed. The proportion of patients who demonstrated clinical response to fatigue, defined as an increase in Functional Assessment of Chronic Illness Therapy-Fatigue score of ≥7 from baseline to Day 8, and the proportion of patients with a nocebo effect, defined as those reporting >2 side effects, were determined. Baseline patient characteristics and symptoms recorded using the Edmonton Symptom Assessment Scale (ESAS) were analyzed to determine their association with placebo and nocebo effects.
Fifty-nine (56%) patients had a placebo response. Worse baseline anxiety and well-being subscale score (univariate) and well-being (multivariate) were significantly associated with placebo response. Commonly reported side effects were insomnia (79%), anorexia (53%), nausea (38%), and restlessness (34%). Multivariate analysis indicated that worse baseline (ESAS) sleep, appetite, and nausea were associated with increased reporting of the corresponding side effects.
Greater than half of advanced cancer patients enrolled in CRF trials had a placebo response. Worse baseline physical well-being score was associated with placebo response. Patients experiencing specific symptoms at baseline were more likely to report these as side effects of the medication. These findings should be considered in the design of future CRF trials. Cancer 2010. © 2009 American Cancer Society.
Cancer-related fatigue (CRF) is defined as a persistent, subjective sense of tiredness or exhaustion related to cancer or its treatment and is not proportional to recent activity, interferes with usual functioning, and does not usually resolve with rest.1 CRF is the most common symptom reported in patients with advanced cancer2, 3 and is estimated to be present in approximately 60% to 90% of patients receiving active treatment, and in 30% to 75% of cancer survivors.4-6 Multiple therapeutic approaches have been proposed to treat this condition. However, to our knowledge, to date, there is no single drug intervention considered standard therapy for CRF.
In 2 randomized controlled trials, our group investigated the psychostimulant methylphenidate and the anticholinesterase inhibitor donepezil in the treatment of CRF. These trials failed to demonstrate a significant difference between the drugs and placebo. In both trials, we noted considerable response to placebo.7, 8 This so-called placebo effect has previously been extensively described in the literature for pain, Parkinson disease, the immune system, asthma, and depression.9-12
Placebo is described as a biologically inert substance, or any other form of therapy or intervention, that when given as an intervention is not expected to produce favorable outcomes. A placebo effect is any favorable psychobiologic effect after the administration of a placebo.13 To distinguish the therapeutic effect of an inert substance from the harmful effects that it may cause, the term nocebo, which in Latin means I will harm, is used. A nocebo effect is defined as any distressing effect of a placebo, and is less studied in the literature.14
As with randomized controlled trials of treatments for other symptoms, such as pain, randomized controlled trials of therapy for fatigue may be influenced by significant confounding effect of the placebo and nocebo effects, which may prevent accurate estimation of the power needed to determine efficacy. To our knowledge, no studies have been published to date that indicate the placebo and nocebo effects on fatigue. The purpose of the current study was to determine the frequency and predictors of placebo and nocebo effects in patients with CRF, which could allow for better design of future fatigue treatment trials and also aid future researchers in their interpretation of results, particularly with regard to reported side effects.
MATERIALS AND METHODS
We conducted new analyses of data already collected from 254 patients with CRF who participated in 2 clinical trials previously conducted by our team between July 1, 2003, and July 6, 2006. The current study was approved by the institutional review board of The University of Texas M. D. Anderson Cancer Center. Two trials have already been reported.7, 8 In 1 trial, patients were randomly assigned to receive either methylphenidate or placebo, and in the other trial, patients were randomly assigned to receive either donepezil or placebo. The patients took their medications for 7 days. All patients had advanced cancer and reported a fatigue score of at least 4 on the Edmonton Symptom Assessment Scale (ESAS)15 during the last 24 hours on at least 4 consecutive days. Medications, including chemotherapy that the patients were already taking before the trial were not restricted or discontinued. Patients taking antidepressants were maintained on stable doses during the study period. The patients were included in the current retrospective study if they received placebo as an intervention for fatigue. There were 22 patients randomized to placebo who were not evaluable due to dropping out or missing data. A total of 105 patients who received placebo in the 2 previous trials were pooled for analysis. Data collection is described in the flowchart in Figure 1.
The following demographic information was collected: age, gender, race, marital status, educational level, and primary cancer diagnosis. Functional Assessment of Chronic Illness Therapy-Fatigue (FACIT-F) scores, ESAS scores, Mini-Mental State Examination (MMSE) scores, and reported side effects were also collected.15-17 FACIT-F is an assessment tool that has been validated for use in CRF. It is comprised of 4 domains of well-being (physical, social, emotional, and functional) and an additional 13-point fatigue subscore. The patient rates the intensity of fatigue and other symptoms on a scale of 0 to 4. The scores range from 0 to 54, and higher scores correspond to less fatigue.16 The ESAS is a validated tool that is used to assess the intensity of 9 common symptoms in patients with cancer or chronic illness (pain, fatigue, nausea, depression, anxiety, anorexia, drowsiness, shortness of breath, and sleep) as well as feeling of well-being. The patient rates each symptom on a scale of 0 to 10, with 0 being no symptom and 10 being the worst possible symptom.15, 18 The MMSE is a 30-point assessment tool that is used to determine cognitive status, with the higher number denoting more intact cognition.17
In both studies, the patients were asked if they developed side effects from the drug. The side effects listed were slurred speech, restlessness, behavioral change, dizziness, vertigo, tachycardia, insomnia, and anorexia.
Our study had 2 parts, the placebo part and the nocebo part. In this study, the term placebo will be used to designate the positive and therapeutic effects of the inert substance, and nocebo will be used to denote the effects of the inert substance that were considered harmful to the patient. For the placebo part, patients who were included in the study were classified as either responders or nonresponders to placebo. A patient was considered to be a responder if there was an improvement (increase) in FACIT-F score of at least 7 points between baseline and end of study (Day 8). This definition of responder was taken from our previous studies.7, 8 For the nocebo part of our study, patients who reported ≥2 side effects from the inert drug were considered to have a nocebo effect. This cutoff point was based on the fact that the median number of side effects reported by patients was 2.
To determine the association between baseline characteristics, FACIT-F scores, ESAS scores, MMSE scores, and response to placebo, the responders and nonresponders were compared using the chi-square test for categoric variables and a Wilcoxon rank sum test for continuous variables. Subsequent multivariate regression analysis was done using variables that were noted to be significant to determine the best model for predictors. Significance levels less than .05 were considered statistically significant. The same statistical analyses were performed to determine the association between various factors and nocebo effect.
Patient demographic characteristics are summarized in Table 1. Among the 105 patients included in this analysis, 63 (60%) were women, 79 (75%) were Caucasians, 69 (66%) were married, and 52 (51%) had a college degree or higher. The most common primary malignancy was breast cancer, which was the primary malignancy in 35 (33%) patients.
|Patient Characteristics||No. of Patients (%) n = 105|
|Female gender||63 (60)|
|College or higher||52 (51)|
|High school||28 (27)|
|Less than high school||22 (22)|
|Primary cancer diagnosis|
|Head and neck||6 (6)|
Fifty-nine (56%) patients of reported a response to the placebo. Patient characteristics of responders and nonresponders to placebo are summarized in Table 2. On univariate analysis, factors found to be significantly associated with response to placebo were worse baseline ESAS anxiety and worse baseline FACIT-F physical well-being subscale score, fatigue subscale score, and total FACIT-F score (Table 2). Other variables tested did not demonstrate any significant association with response to placebo. On multivariate regression analysis, only worse baseline FACIT-F physical well-being subscale score was found to be a significant predictor of placebo response (odds ratio [OR], 0.86; P = .001).
|Characteristic||Responders (n = 59; 56%)||Nonresponders (n = 46; 44%)||P|
|Female||37 (63)||26 (57)||.52|
|Asian||1 (2)||1 (2)||.53|
|Black||7 (12)||3 (7)|
|Caucasian||43 (73)||36 (78)|
|Hispanic||7 (12)||6 (13)|
|Other||1 (2)||0 (0)|
|Median age (range), y||59 (37-84)||58 (37-78)||.63|
|Married||40 (70)||29 (66)||.75|
|Divorced||9 (16)||7 (16)|
|Single||8 (14)||8 (18)|
|College or higher||28 (49)||24 (53)||.16|
|High school||13 (23)||15 (33)|
|Less than high school||16 (28)||6 (14)|
|Primary cancer diagnosis|
|Breast||18 (31)||17 (37)||.49|
|Gastrointestinal||5 (8)||2 (4)|
|Genitourinary||4 (7)||8 (17)|
|Lung||13 (22)||7 (15)|
|Gynecologic||2 (3)||2 (4)|
|Head and neck||5 (8)||1 (2)|
|Hematologic||9 (15)||5 (11)|
|Other||3 (5)||4 (9)|
|Median Mini-Mental State Examination score (range)||30 (24-30)||30 (24-30)||.70|
|Median ECOG performance status score (range)||2 (0-4)||2 (0-3)||.169|
|Median baseline FACIT-F score (range)|
|Physical Well-being||13 (3-25)||17 (7-24)||.001|
|Social/Family||23 (7-28)||24 (13-28)||.17|
|Emotional||18 (6-24)||18 (9-24)||.72|
|Functional||12 (4-25)||14 (6-28)||.19|
|Fatigue subscale||16 (1-47)||19 (4-43)||.045|
|Total FACIT-F score||80 (38-141)||94 (61-134)||.004|
|Median baseline ESAS (range)|
|Pain||4 (0-10)||3 (0-10)||.112|
|Fatigue||7 (4-10)||7 (4-10)||.44|
|Nausea||0 (0-9)||0 (0-6)||.24|
|Depression||3 (0-8)||2 (0-9)||.11|
|Anxiety||3 (0-8)||2 (0-8)||.01|
|Drowsiness||5 (0-10)||4 (0-10)||.88|
|Shortness of breath||2 (0-10)||2 (0-10)||.42|
|Appetite||3 (0-10)||3 (0-10)||.89|
|Sleep||5 (0-10)||5 (0-10)||.34|
|Well-being||5 (0-9)||5 (0-9)||.44|
Pooled patients reported 10 different side effects. These side effects and the proportions of patients who reported each side effect are listed in Table 3. Eight side effects reported by patients in both studies. In the methylphenidate study, nausea was reported in addition to the other 8 side effects. In the donepezil study, skin changes were reported in addition to the other 8 side effects. The most common side effects reported were insomnia (79%), followed by anorexia (53%), nausea (33%), and restlessness (34%).
|Side Effectsa||No. of Patients (%)|
|Slurred speech||5/105 (5)|
|Behavioral change||14/105 (13)|
|Skin problems||10/50 (20)|
There was no association noted between the reporting of side effects and response or lack of response to placebo. Multivariate analysis for each of the symptoms indicated that worse baseline ESAS sleep, appetite, and nausea scores were associated with increased reporting of these side effects. With each unit increase in the FACIT-F functional well-being score, patients were 15% less likely to report insomnia. Each unit increase in the FACIT-F fatigue subscale score was associated with patients reporting less anorexia. Patients with baseline nausea were twice as likely to report nausea as a side effect and those with baseline anxiety were more likely to report restlessness. Other variables that demonstrated significant association in our multivariate logistic regression model are shown in Table 4.
|Side Effecta||Variables of Nocebo Effect||OR||P|
|Restlessness||Baseline FACIT-F Total Score||1||0.94|
|Baseline FACT-G Physical Well-being score||0.99||0.87|
|Baseline performance status||1.39||0.2|
|Baseline ESAS Anxiety||1.24||0.02|
|Baseline ESAS Pain||1.07||0.41|
|Baseline ESAS Drowsiness||1.05||0.53|
|Baseline ESAS Shortness of breath||1||0.96|
|Baseline ESAS Well-being||1.05||0.65|
|Baseline ESAS Fatigue||0.67||0.05|
|Baseline ESAS Nausea||0.84||0.41|
|Primary cancer diagnosis||3.29||0.08|
|Mini-Mental State Examination score||0.87||0.63|
|Baseline FACT-G Social/Family Well-being||0.84||0.11|
|Dizziness||Baseline FACIT-F total score||1.01||0.46|
|Baseline FACT-G Physical Well-being||0.99||0.95|
|Baseline performance status||1.87||0.03|
|Baseline ESAS Sleep||1.06||0.5|
|Baseline ESAS Well-being||1.2||0.06|
|Baseline FACT-G Fatigue subscale||1.23||0.001|
|Baseline performance status||3.97||0.03|
|Baseline ESAS Depression||1.54||0.02|
|Baseline ESAS Sleep||1.4||0.04|
|Baseline ESAS Drowsiness||0.85||0.46|
|Baseline ESAS Nausea||1.04||0.83|
|Baseline ESAS Well-being||1.04||0.84|
|Insomnia||Baseline Mini-Mental State Examination score||1.8||0.01|
|Baseline FACIT-F total score||1.01||0.77|
|Baseline FACT-G Functional Well-being||0.85||0.01|
|Baseline FACT-G Physical Well-being||0.9||0.17|
|Baseline FACT-G Emotional Well-being||0.86||0.27|
|Baseline performance status||1.32||0.49|
|Baseline ESAS Nausea||0.76||0.04|
|Baseline ESAS Sleep||1.26||0.04|
|Baseline ESAS Anxiety||1.11||0.51|
|Baseline ESAS Drowsiness||1.02||0.87|
|Baseline ESAS Pain||1.02||0.91|
|Baseline ESAS Appetite||1.23||0.01|
|Baseline FACIT-F total score||1||0.91|
|Baseline FACT-G Fatigue subscale||0.92||0.01|
|Baseline FACT-G Functional Well-being||1.01||0.91|
|Baseline FACT-G Physical Well-being||0.98||0.72|
|Baseline ESAS Fatigue||0.9||0.53|
|Baseline ESAS Drowsiness||1.04||0.66|
|Baseline shortness of breath||1.08||0.33|
|Baseline ESAS Sleep||1.14||0.12|
|Baseline ESAS Well-being||1.07||0.51|
|Mini-Mental State Examination score||1.72||0.19|
|Baseline FACIT-F total score||1||0.8|
|Baseline FACT-G Physical Well-being||1.11||0.5|
|Baseline ESAS Nausea||2.1||0.004|
|Baseline ESAS Anxiety||0.97||0.85|
|Baseline ESAS Shortness of breath||1.02||0.84|
A total of 96 (91%) patients reported at least 1 side effect to the placebo drug. Using the cutoff of 2 as a basis of grouping those with nocebo effect and those without it, we found that 30 (29%) patients were considered to have no nocebo effect and 75 (71%) patients reported nocebo effects. Worse baseline pain (P = .05), drowsiness (P = .05), and sleep (P = .03) were associated with nocebo effect. Multivariate analysis demonstrated that patients with worse baseline sleep scores were more likely to report more side effects (OR,1.20; P = .021). The frequency of reported side effects is illustrated in Figure 2.
Greater than half of the patients in this study responded to placebo. Earlier studies by Beecher et al demonstrated placebo effects in approximately 35% of patients, and Brown reported that placebo effect occurred in approximately 30% to 40% of patients.13, 19 It has been suggested that the nature of the disorder or symptom influences whether a placebo effect occurs and that subjective symptoms such as pain, anxiety, and depression are more amenable to placebo effects than are more objective measures such as blood pressure.20 It is not surprising, therefore, that a subjective symptom such as fatigue can be influenced by the placebo effect.
To the best of our knowledge, the mechanisms underlying the placebo effect is not fully understood. One proposed mechanism involves the brain's reward circuitry. On the basis of findings from studies on pain and Parkinson disease, it has been hypothesized that expectations of reward or clinical improvement and great desire for effect play a critical role in the placebo effect.10, 21, 22 This may in part explain why there was a significant association in our current study between worse baseline Functional Assessment of Cancer Therapy-General (FACT-G) physical well-being score and placebo response. Patients with worse baseline fatigue may have had greater desire for effect and might also have increased expectations of improvement given their higher symptom burden. Unfortunately, we did not measure patients' expectation and this should be done in future research. Another possible explanation for the observed placebo effect in this group of patients may be related to the concept of regression to the mean, in which the measured change could be secondary to nonsystematic variations.
Although some patients were being treated with medications that can affect fatigue such as chemotherapy or antidepressants, they were receiving stable doses during the duration of the study. Therefore, we believe that their potential to confound our findings are not significant, particularly in these short-term 7-day studies.
The concept of a direct relation between expectation of a desired effect and the actual occurrence of such an effect has been supported by other studies. Linde et al reported that, in 864 patients who received acupuncture for therapy for pain, improvement was significantly associated with high expectations of treatment effect.23 In a study of 26 premenopausal women with irritable bowel disease, investigators concluded that placebo analgesia was associated with increased expectations of pain relief and decrease in negative emotions such as anxiety.9 Other authors have reported this same relation for other conditions.11, 24, 25
Kaptchuk et al reported that there were specific personalities that respond to placebo interventions.26 In earlier studies by the pain placebo pioneer team of Lasagna and Beecher, the investigators reported that only 13 of 93 patients who received placebo twice and reported therapeutic response were reliable placebo responders and that these 13 patients were more anxious, more self-centered, and had more baseline somatic complaints than patients who were nonresponders or inconsistent placebo responders.27 The findings of the current study that patients with higher baseline ESAS anxiety were more likely to respond to placebo are consistent with those reports.
Other characteristics that were previously shown to contribute to placebo response such as gender and age were not observed in this study.28-30 As pointed out by previous investigators, the variables reported in different disease conditions are very difficult to replicate, not generalizable, and inconsistent.31, 32 It could be that these factors would only be contributory to other conditions and not for fatigue. Hyland et al and Whalley et al reported that most variables are related to symptoms and disease characteristics and not patient characteristics.33, 34
The nocebo effect is even less understood than the placebo effect but is believed to be related, similar to the placebo effect, to some neurobiologic mechanism involving negative expectations regarding treatment outcome, expectations of harm, worsening or vulnerability,14 prior conditioning as by previous untoward experiences such as adverse reactions to drugs or interventions,35 and certain psychologic characteristics.
Nearly 25% of patients taking a placebo experience side effects.36 Rosenzweig et al reported that 19% of healthy volunteers taking placebo reported side effects.37 Pogge reported in a review of 67 placebo-controlled trials that at least 23% of patients who received a placebo reported at least 1 side effect.38 In the current study, 91% of the patients reported at least 1 side effect from a placebo drug, and 71% of patients were considered to have nocebo effect. It is likely that the higher frequency reported in the current study compared with other studies can be explained by negative expectations and symptoms that were already reported to be present even before treatment. Providing a list of all the potential side effects of the active drug may have created a negative expectation of treatment outcome and harm. It has been shown for example, that negative expectations result in the amplification of the pain being reported.10 The list of side effects may have conditioned the patient to expect these to develop over the course of the trial.
To our knowledge, there are no reports in the literature of specific patient characteristics associated with the nocebo effect. Some investigators reported that people who tend to be anxious or depressed are more at risk of developing nocebo effect in response to attempts at treatment.35 The results of the current study indicated that patients with anxiety were more likely to report restlessness and that those with depression were more likely to report more tachycardia. These findings suggest, as has been reported in previous studies, that the nocebo effect is related to somatization, whereby the patient expresses emotional distress as physical symptoms.
We found interesting results for those patients who reported nocebo effect. Patients who had insomnia at baseline were more likely to report more nocebo effects. We speculate that if patients did not feel particularly well at baseline, they were much more likely to express their symptoms as a side effect.
Many commonly reported nocebo effects in patients receiving placebo are more generalized such as nausea, fatigue, drowsiness, and insomnia.31 These were observed in the current study patients as well, with insomnia being the most frequently reported, followed by anorexia, nausea, and restlessness. What was of interest was that patients had reported some of these as symptoms (insomnia, anorexia, and nausea) in their baseline ESAS assessment. It would appear that patients were misattributing these symptoms as nocebo effects rather than as resulting from the underlying disease. This has been described previously for pain by Turner et al.31 More research is needed to test this hypothesis.
Our hypothesis that patient's expectations of outcome, whether positive or negative, are associated with either the placebo and nocebo effect needs to be interpreted cautiously due to the retrospective nature of the study. Future research is needed to test this hypothesis. In addition, the current study consisted of a very small cohort of patients, and future studies with a larger patient population are needed to confirm our findings.
Previous authors have described placebo side effects, which are side effects attributed to taking the placebo drug and are not considered to be nocebo effects.14 In the case of the nocebo effect, there is a negative expectation associated with the reporting of side effects. In the context of clinical trials, as was in the 2 trials that we have analyzed, when potential side effects to the active drug are enumerated to the patient, negative expectations can occur.
To our knowledge, this is the first study to specifically examine placebo and nocebo effects in patients with CRF. We found that a good proportion of patients experienced placebo effect and an even greater proportion reported nocebo effects. The implications of these findings are important in the context of research and treatment. Clinical trials on CRF should take into consideration placebo and nocebo effects when designing clinical trials. When patients randomized to placebo demonstrate substantial clinical response, it would be difficult for patients that received the treatment drug to demonstrate an even greater response.39, 40 Failure to consider the placebo effect may explain why previous trials have failed to demonstrate significant therapeutic effects of drugs over placebo. Strategies in research design aimed at perhaps minimizing the placebo effect, such as a bigger sample size or longer trial period, are needed to more effectively evaluate the real effect of treatment. Our observation of a nocebo effect further justifies use of a placebo in clinical trials because it permits better appraisal of side effects of active drug. Researchers need to recognize that the nocebo phenomenon exists. Our findings on the nocebo effect also raise the question regarding the disclosure of potential side effects in the daily clinical setting. If negative expectations can influence the reporting of adverse events, how then should we inform our patients regarding the possibility of side effects without causing harm or suffering?
With increasing interest in and better understanding of the placebo and nocebo phenomena, supportive therapy along with pharmacologic intervention aimed at maximizing treatment benefit may be more incorporated in routine patient care.
Greater than half of patients with advanced cancer enrolled in the fatigue trials responded to placebo. A worse baseline physical well-being score was associated with placebo response. Patients experiencing specific symptoms at baseline were more likely to report these as side effects of the medication. These findings should be considered in the design of future clinical trials of the treatment of CRF.
CONFLICT OF INTEREST DISCLOSURES
Dr. Bruera is supported in part by National Institutes of Health grant numbers R01NR010162-01A1, R01CA1222292-01, and R01CA124481-01.
Dr. Hui is funded by a fellowship from the Clinician Investigator Program, Royal College of Physicians and Surgeons of Canada.
- 1National Comprehensive Cancer Network. Cancer Related Fatigue. Clinical Practice Guidelines in Oncology. Fort Washington, Pa: National Comprehensive Cancer Network; 2007.