Comparative “Nocebo Effects” in Older Patients Enrolled in Cancer Therapeutic Trials: Observations From a 446-Patient Cohort

Abstract

BACKGROUND

A nocebo is an inert substance associated with adverse events. Although previous studies have examined the positive (placebo) effects of such inert substances, few have examined negative (nocebo) adverse event profiles, particularly in older patients who have higher morbidity and can experience frequent and severe adverse events from cancer therapy.

METHODS

This study focused on placebo/nocebo-exposed patients who participated in 2 double-blind, placebo-controlled, cancer therapeutic studies, namely, North Central Cancer Therapy Group trial NCCTG 97-24-51 and American College of Surgeons Oncology Group trial Z9001, with the goal of reporting the comparative, age-based adverse event rates, as reported during the conduct of these trials.

RESULTS

Among the 446 patients who received only placebo/nocebo and who were the focus of the current report, 161 were aged ≥65 years at the time of respective trial entry, and 5234 adverse events occurred. Unadjusted adverse event rates did not differ significantly between patients aged ≥65 years and younger patients (rate ratio, 1.01; 99% confidence interval, 0.47–2.02), and the findings were similar findings for grade 2 or worse adverse events and for all symptom-driven adverse events (for example, pain, loss of appetite, anxiety). Adjustment for sex, ethnicity, baseline performance score, and individual trial resulted in no significant age-based differences in adverse event rates. Similar findings were observed with an age threshold of 70 years.

CONCLUSIONS

Adverse events are equally common in older and younger cancer patients who are exposed to nocebo and thus require the same degree of clinical consideration regardless of age.

INTRODUCTION

Inert substances can be associated with both positive (placebo) and negative (nocebo) adverse event profiles in controlled clinical trials.¹ In contrast to a placebo, few clinical studies have examined the adverse event profiles of a nocebo, herein referred to as the “nocebo effect”; and some researchers have gone so far as to describe the investigation of this nocebo effect as “unethical” because of the potential to induce harm, particularly in vulnerable patients, such as older patients or those with cancer.² Yet it remains important to study the nocebo effect for at least 4 reasons. First, understanding the adverse profiles associated with an inert substance enables us to better gauge the adverse event profiles of cancer drugs themselves. Such information allows us to gain a more accurate sense of drug tolerability in a cancer setting, in which the expectations of adverse events from antineoplastic agents are especially high and in which such expectations may drive the reporting of adverse events.

Second, previous studies suggest that the negative consequences of a nocebo may be clinically detrimental.^3,4 These negative effects translate into poorer adherence to cancer treatment and poorer cancer outcomes. Understanding the nocebo effect in both older and younger patients with cancer is warranted because of its potential to directly and negatively impact patient care and outcomes.

Third, some investigators have gone so far as to suggest that inert substances have biologic consequences that go beyond their role of benchmarking therapeutic outcomes against the natural history of a disease or of enabling adjustments to the phenomenon of regression toward the mean. Ben-Shaanan and colleagues used an animal model to demonstrate that the favorable psychological effects of a placebo (the opposite of a nocebo) drive innate and adaptive immunity, both of which lead to activation of monocytes and macrophages, a reduction in bacterial load, and an enhanced T-cell response.⁵ The activation of such immune pathways raises the possibility that subgroups of patients manifest different responses to inert substances, possibly as a result of the variable dynamic potency of innate and adaptive immune mechanisms or perhaps because of other biologic mechanisms. Assessing such effects is particularly important in older patients, because senescence is associated with well described changes in immune mechanisms.^6,7 To our knowledge, no previous studies have examined clinical responses to inert substances in older patients with cancer, thus making the current study novel. If indeed placebo/nocebo effects have biologic underpinnings derived from demographic factors, such as age, then it is important to ascertain their comparative, potentially divergent prevalence and severity.

Fourth, understanding the nocebo effect is particularly important in older cancer patients, who appear to suffer more frequent and more severe adverse events from chemotherapy compared with their younger counterparts.^8–12 Along similar lines, older patients often have more comorbidities, which can give rise to the reporting of more adverse events that could be inaccurately attributed to cancer treatment.^8–12 To our knowledge, no previous studies have examined the nocebo effect specifically in older patients with cancer, although demographics indicate an ongoing increase in the number of these older patients. This last point makes the current study particularly relevant and important.

The current study relied on 2 prospectively conducted placebo-controlled, therapeutic trials in patients with cancer to determine whether older patients with cancer report more pronounced adverse event profiles than younger patients.^13,14 Both of these trials used a single-agent, inert substance with no chemotherapy in the placebo/nocebo arm. Comparatively benchmarking such presumed nocebo effects might contribute to a better understanding of drug-induced or morbidity-induced adverse event profiles in older patients with cancer, enhance the design and interpretation of cancer clinical trials, and carry relevant clinical implications.

MATERIALS AND METHODS

Overview

This study focused on 2 double-blind, placebo-controlled, cancer therapeutic studies, namely, North Central Cancer Treatment Group (NCCTG) 97-24-51 and American College of Surgeons Oncology Group (ACO-SOG) Z9001, both of which were conducted within the National Cancer Institute’s cooperative group mechanism.^13,14 Because cancer therapeutic trials rarely include a placebo/nocebo arm with no concomitant chemotherapy, these trials were selected because they included such an arm. In addition, these trials were chosen because they required the reporting of not only severe but also grade 1 and 2 adverse events. NCCTG 97-24-51 demonstrated no role for maintenance oral carboxyaminotriazole in patients who had achieved a favorable response after chemotherapy for metastatic nonsmall cell lung cancer; ACOSOG Z9001 demonstrated that adjuvant imatinib improved recurrence-free survival in patients who had undergone resection for primary gastrointestinal stromal tumors. In the current study, all patients who had received only the inert substance—that is, only those who had received placebo/nocebo with no chemotherapy—were included.

This report describes an age-based analysis of adverse events within the placebo/nocebo arms of both multi-institutional trials. The Mayo Clinic Institutional Review Board provided approval before data analyses.

Endpoints and Definitions of Adverse Events

The primary objective of this study was to report and compare adverse event rates in older versus younger cancer patients, all of whom had been exposed to only placebo/nocebo in a double-blind manner. Age was defined in a binary fashion as ≥65 years versus <65 years at the time of study entry based on extensive precedent. Similar analyses were performed with an age-defined threshold of ≥70 years versus younger patients.

Adverse events were recorded during the conduct of each trial based on Common Terminology Criteria for Adverse Events, versions 2.0 and 3.0 (National Cancer Institute/National Institutes of Health, Bethesda, MD) in NCCTG 97-24-51 and ACOSOG Z9001, respectively. Adverse events were characterized as grade 1 (mild), grade 2 (moderate), grade 3 (severe), grade 4 (life-threatening), and grade 5 (lethal), thus encompassing the full spectrum of severity. In the current study, adverse events were characterized based on whether they were symptom-driven (based on either direct or indirect patient report) or nonsymptom-driven (based on laboratory or other objective clinical evidence). Examples of symptom-driven adverse events include pain, loss of appetite, anxiety, cough, depression, diarrhea, and fatigue. Examples of nonsymptom-driven adverse events include hematologic toxicity, other abnormal laboratory values, and physical examination findings. Attribution of adverse events was not included, because prior studies question its value.^15,16

Analyses

Adverse events were sorted and reviewed in aggregate, based on whether they were grade 2 or worse and whether they were symptom-driven. Adverse event rates were modeled as a function of age with Poisson regression models, which included offset terms for time on study and were adjusted for sex, ethnicity, patient baseline performance score, and trial. These adjustment factors were selected for their potential to influence nocebo effects as a result of a review of the published literature and age-related differences within the study cohort. Because it was uncertain whether adverse events were independent for a given patient and because independence is a key assumption in Poisson models, observed adverse event rates (defined as the number of adverse events divided by time on study) were also assessed with a linear regression model. To assess the extent to which baseline covariates might have impacted the age-based group differences, unadjusted versions of the Poisson (with an offset for time on study) and linear models were also analyzed. Poisson models allowed for the mean and variance to differ (quasi-Poisson modeling). Statistical tests were 2-tailed; a P value < .004, which was derived from constructing 12 different models in our comparisons (that is, .05 of 12), was considered statistically significant and accounted for multiple testing. Confidence intervals (CIs) were calculated and are reported as 99.6% but are noted in the text and tables as 99%. All statistical analyses were performed using SAS version 9.3 (SAS Institute Inc, Cary, NC) and R version 3.2.3 (R Foundation for Statistical Computing, Austria, Vienna).

RESULTS

Demographics

In total, 446 placebo/nocebo-exposed patients were the focus of this report. One hundred sixty-one patients were aged ≥65 years at the time of entry into their respective trial, and 105 were aged ≥70 years. Demographics are notable for more favorable baseline performance scores in younger patients (Table 1).

TABLE 1.

Demographics

Characteristic	Age-Based Groups
	No. (%)a		Pb	No. (%)a		Pb
	Aged ≥65 y, N = 161	Aged <65 y, N = 285		Aged ≥70 y, N = 105	Aged <70 y, N = 341
Age at clinical trial entry: Median [range], y	71 [65–91]	54 [18–64]	—	74 [70–91]	56 [18–69]	—
Sex
Women	73 (45)	127 (45)	.87	51 (49)	149 (44)	.3797
Men	88 (55)	158 (55)		54 (51)	192 (56)
Performance status
0	96 (60)	207 (73)	.010	51 (49)	252 (74)	< .0001
1	61 (38)	76 (27)		50 (48)	87 (26)
2	4 (2)	2 (1)		4 (4)	2 (1)
Trial
NCCTG 97-24-51	45 (28)	47 (17)	.004	35 (33)	57 (17)	.0002
Z9001	116 (72)	238 (84)		70 (67)	284 (83)
Ethnicity
Hispanic	1 (0.6)	16 (6)	.018	0 (0)	17 (5)	.0645
Non-Hispanic	154 (96)	253 (89)		100 (95)	307 (90)
Unknown/not reported	6 (4)	16 (6)		5 (5)	17 (5)

Abbreviation: NCCTG, North Central Cancer Treatment Group; Z9001, an American College of Surgeons Oncology Group trial.

^aNumbers in parentheses refer to percentages unless otherwise specified.

^bChi-square or Kruskal-Wallis tests were used, as appropriate.

Overall Adverse Events

In total, 5234 adverse events occurred. Adverse events were grade 1, 2, 3, 4, or 5 with a frequency of 4283, 718,141, 11, and 3 episodes, respectively. In 78 episodes, the adverse event was not graded. On average, 12 adverse events per patient (standard deviation, 16) occurred, and patients remained on study for 0.03 to 56 months. To better characterize the nature of these adverse events, those of any grade that occurred in >10% of all patients who reported a specific event at least once during the course of either trial included the following: alkaline phosphatase elevation, diarrhea, alopecia, anemia, anorexia, hyperglycemia, constipation, cough, dizziness, renal dysfunction, indigestion, shortness of breath, arthralgias, myalgias, edema, fatigue, ataxia, flatulence, headache, leukopenia, nausea, neuropathy, pain, rash, vomiting, and weight gain.

Age-Based Comparisons of Adverse Event Rates

Including all grades of adverse events, we observed that approximately 2 adverse events per month occurred in patients aged ≥65 years, and 4 per month occurred in younger patients. An average of 0.5 grade 2 or worse adverse events per month occurred in patients aged ≥65 years, and 0.3 per month occurred in younger patients. With respect to symptom-driven adverse events, an approximate average of 1 per month occurred both in patients aged ≥65 years and in younger patients. In patients aged ≥70 years and in younger patients, the approximate rate of all adverse events was 2.0 and 3.0, respectively, and the approximate rate of grade 2 or worse was 0.5 and 0.3, respectively. Rates of symptom-driven adverse events also were estimated as 1 in both age groups with an age threshold of 70 years.

Unadjusted adverse event rates did not differ significantly between patients aged ≥65 years and younger patients (rate ratio, 1.01; 99% CI, 0.47–2.02) (Table 2). Similarly, for grade 2 or worse adverse events and for all symptom-driven adverse events, patients aged ≥65 years did not manifest significantly different unadjusted adverse event rates compared with younger patients (rate ratio, 1.33 [99% CI, 0.51–3.28] and 1.23 [99% CI, 0.59–2.45], respectively). Analogous findings were observed using an age threshold of 70 years (Table 2).

TABLE 2.

Analyses of Age-Based Adverse Events^a

Model	Age, y	Mean Observed Rate per Month of All Adverse Events in Older (Younger) Patients	Rate Ratio [99% CI]	P	Mean Observed Rate per Month of Grade 2 or Worse Adverse Events in Older (Younger) Patients	Rate Ratio [99% CI]	P	Mean Observed Rate per Month of Symptom-Related Adverse Events in Older (Younger) Patients	Rate Ratio [99% CI]	P
Rate of adverse events	≥65 vs younger	1.78 (3.49)	1.01 [0.47–2.02]	.98	0.48 (0.30)	1.33 [0.51–3.28]	0.37	0.72 (1.28)	1.23 [0.59–2.45]	.41
	≥70 vs younger	2.04 (3.13)	1.28 [0.54–2.71]	.37	0.55 (0.31)	1.56 [0.52–3.99]	.20	0.82 (1.16)	1.60 [0.71–3.29]	.07
Rate of adverse events + sex + performance score + ethnicity + trial	≥ 65 vs younger	—	0.78 [0.36–1.64]	.35	—	1.02 [0.43–2.31]	.94	—	0.97 [0.46–1.98]	.89
	≥70 vs younger	—	0.85 [0.34–1.94]	.59	—	1.01 [0.37–2.45]	.99	—	1.12 [0.48–2.44]	.69

Abbreviation: CI, confidence interval.

^aPoisson regression was used and reported, but similar results were generated with linear regression models.

Multivariate Age-Based Comparisons

After adjusting for sex, ethnicity, baseline performance score, and trial, no significant differences were observed in the rate of all adverse events (risk ratio, 0.78 [99% CI, 0.36–1.64] and 0.85 [99% CI, 0.34–1.94] for age thresholds of 65 and 70 years, respectively). Similarly, no significant age group differences were observed for all grade 2 or worse adverse events or for all symptom-driven adverse events (Table 2).

DISCUSSION

In the current study, we examined age-based adverse event rates in cancer patients who had received an inert substance within the context of a cancer therapeutic trial. We observed what is perceived as a “nocebo effect” among patients, regardless of their age—with greater than 5000 adverse events reported in patients who were receiving an inert substance. The incorporation of a “nocebo” into the design of these trials served to capture patients’ perceived, anticipated, or very real adverse events that were not related to the investigational agent. This study underscores the importance of including an inert substance in a clinical trial to capture the prevalence and severity of adverse events that might otherwise have been attributed to an investigational agent. It is noteworthy, however, that we did not observe age-based, statistically significant differences in nocebo adverse event rates, regardless of whether we compared groups based on total, grade 2 or worse, or symptom-driven adverse event rates. These findings suggest that, at least in adults, a notable nocebo effect appears to occur irrespective of patient age.

An age-independent nocebo effect has important implications. First, previous studies have reported higher rates of adverse events in older patients with cancer. For example, Schild and colleagues described higher rates of myelosuppression and pneumonitis in older patients with cancer compared with their younger counterparts.¹² The current study demonstrates comparable age-based nocebo effects, thereby suggesting that such previously reported differential age-based adverse event rates are not a result of differences in perception of toxicity on the part of patients or health care providers or of differences in how adverse events are reported. The current study underscores that reported adverse event rates should be taken equally seriously in older and younger patients with cancer.

Second, regardless of whether or not a particular agent is believed to be well tolerated, educational efforts might be of value in helping to manage the negative expectations associated with a specific drug. Schenk described how health care providers’ word choices can give rise to so-called “language traps,” which, in turn, can spawn worse symptoms; for example, he suggests that a commonly used statement, “Here’s your pain medication,” might be more favorably and less suggestively phrased as, “Here’s some medication to help you get comfortable,” with a diminishment of the nocebo effect.⁴ Educating health care providers on how to educate patients might make cancer treatment more tolerable for all adult patients of any age. Indeed, in view of the high rates of nocebo-related adverse events observed in the current study, future studies might choose to test educational interventions to determine whether they lead to better tolerance of cancer treatment and to reduced rates of adverse event reports in nocebo-exposed patients.

The current study has limitations. First, this study was undertaken within the context of 2 clinical trials, which were testing agents that required patients to sign a consent form and, in the case of carboxyaminotriazole, to receive a noncommercially available drug with only a preliminary track record of safety.^13,14 It is possible that enrollment in a clinical trial heightened the nocebo effect and that, outside a clinical trial setting, a much lower nocebo effect would have occurred. Hence, some caution should be exercised when applying the conclusions from this study to a nonclinical trial setting. Second, to explain the adverse events captured in these trials, morbidity—rather than a “nocebo effect”—might be invoked. Admittedly, some of these adverse events are likely attributable to cancer or to other concurrent morbidity, but it seems unlikely that morbidity is the sole explanation for the numerous adverse events reported in the inert substance arms of these 2 trials for at least 2 reasons. Quality-of-life data suggest that baseline symptomatology was minimal among patients for whom such data were available, thus suggesting that at least some of the subsequent adverse events might have been related to patients’ anticipation of adverse events from the investigational agents.¹³ Moreover, most patients in the current study were participating in a postoperative adjuvant trial, in which the majority (even among the placebo/nocebo-exposed patients) remained cancer-free well beyond 1 year—circumstances that make it impossible to evoke cancer consistently in the attribution of adverse events. Thus, although not all the adverse events described in the current study were incontestably derived from anticipated adverse events related to the potential of having received an investigational agent, it seems possible that at least some were.

In summary, an inert substance, or a nocebo, serves an important role in understanding the adverse events reported in cancer clinical trials. This so-called “nocebo effect” appears to be nonage-dependent. Understanding this nocebo effect is important not only for interpreting clinical trial data but perhaps also for managing patients’ expectations of adverse events with the goal of improving the tolerability of cancer therapy.