Fax: (585) 461-5601
Patient expectation is a strong predictor of severe nausea after chemotherapy
A University of Rochester Community Clinical Oncology Program study of patients with breast carcinoma
Article first published online: 29 OCT 2004
Copyright © 2004 American Cancer Society
Volume 101, Issue 11, pages 2701–2708, 1 December 2004
How to Cite
Roscoe, J. A., Bushunow, P., Morrow, G. R., Hickok, J. T., Kuebler, P. J., Jacobs, A. and Banerjee, T. K. (2004), Patient expectation is a strong predictor of severe nausea after chemotherapy. Cancer, 101: 2701–2708. doi: 10.1002/cncr.20718
- Issue published online: 16 NOV 2004
- Article first published online: 29 OCT 2004
- Manuscript Accepted: 8 SEP 2004
- Manuscript Revised: 20 AUG 2004
- Manuscript Received: 1 JUL 2004
- National Cancer Institute. Grant Number: U10CA 37420
- response expectancy;
Patients may use their past experiences with nausea, as well as information about the incidence of nausea from chemotherapy that other patients have experienced, to form a prediction, or response expectancy, of nausea from their own upcoming chemotherapy. Mounting evidence suggests that these expectancies relating to nausea are significant predictors, and, likely, contributing factors to the development of treatment-related nausea.
The patients in the current study were participants in the control arm of a multicenter clinical trial conducted between November 1999 and July 2001 by the University of Rochester Community Clinical Oncology Program. All patients in the current report were age ≥ 18 years and were about to begin a first cancer treatment regimen containing doxorubicin.
Expectancy of nausea assessed before patients received their first doxorubicin-based chemotherapy treatment was found to be a strong predictor of subsequent nausea and in fact was stronger than previously reported predictive factors, including age, nausea during pregnancy, and susceptibility to motion sickness. Women who believed it was “very likely” that they would have severe nausea from chemotherapy were five times more likely to experience severe nausea than fellow patients who thought its occurrence would be “very unlikely.”
Further studies confirming an expectancy of nausea as a risk factor are warranted as are studies examining the benefit to a patient's quality of life from modifying antiemetic treatment guidelines to take into account symptom expectancies. Finally, ethically acceptable interventions that are designed to reduce patients' nausea expectancies or increase their expectancies of nausea control should be developed and studied. Cancer 2004. © 2004 American Cancer Society.
Despite a great deal of progress in preventing and controlling chemotherapy-induced nausea and emesis, they continue to be commonly reported. Control of emesis by the 5-HT3 class of serotonin antagonists has resulted in a few patients with cancer who experience emesis during chemotherapy.1 Nausea, however, continues to be controlled inadequately as it is commonly reported by a majority of patients at some time during a course of chemotherapy.1 In addition to causing intrinsic misery, nausea can lead to a poor quality of life (QOL), decreased nutrition, an inability to work, and it can affect compliance with chemotherapy programs.2–4
Not all patients have the same risk of experiencing nausea after chemotherapy. Previously reported risk factors include the chemotherapy regimen used, younger age, and female gender. Previous experience with motion sickness and nausea with pregnancy are less important predictors. Patients may use their past experiences with nausea, as well as information regarding the incidence of nausea after chemotherapy that other patients have experienced, to form a prediction or response expectancy, of nausea from their own upcoming chemotherapy. Mounting evidence suggests that these expectancies relating to nausea are significant predictors and likely are contributing factors to the development of treatment-related nausea. Two investigations, both conducted before the widespread use of the 5-HT3 antiemetics, have found a relation between expectations for the development of nausea and emesis after chemotherapy and their subsequent occurrence.5, 6 Other studies conducted both before7, 8 and after8–10 the extensive use of 5-HT3 antiemetics have found a statistically significant relation between expectations for nausea and nausea development, but not between expectations for emesis and subsequent emesis. A relation between expectations for nausea and subsequent anticipatory nausea also has been reported.11, 12 The association between expectations and chemotherapy-induced nausea does not appear to be dependent on characteristics of the chemotherapy treatment.9
On a theoretic level, Kirsch13 suggests that an anticipation or response expectancy for the occurrence of a physiologic sensation (e.g., nausea) can generate corresponding subjective experiences, and, to the extent that they do so, are self-confirming. He distinguishes expectancies for nonvolitional responses, such as nausea and pain from both stimulus expectancies (i.e., anticipation of external consequences, such as food, money, praise, or punishment) and intentions (i.e., anticipation of voluntary response), and hypothesizes that response expectancies directly affect both physiologic and psychologic outcomes. Similarly, the nocebo hypothesis, discussed by Hahn,14 proposes that expectations of sickness cause symptoms in patients who expect them. The role of expectancy in the development of nausea can be explained partially by schema theory,15, 16 which posits that information (e.g., sensory data) will be interpreted through relevant schema, meaning that an individual expecting a symptom (e.g., nausea) will be more likely to interpret sensations as nauseous than an individual not expecting the symptom.
Patients receive information regarding the possibility of nausea from a variety of sources, including their health care providers, family and friends, and media. Even though a given piece of information may not have been scientifically derived or validated for that particular patient, it may, nonetheless, have been a strong factor in the formation or his or her response expectancy regarding nausea. We report on a study that examined the contribution of various sources of information to the formation of patients' nausea expectancies and examined their relative predictive ability in regards to subsequent chemotherapy-induced nausea.
MATERIALS AND METHODS
The patients were participants in a multicenter clinical trial conducted between November 1999 and July 2001 at 17 geographically diverse member sites of the University of Rochester Community Clinical Oncology Program. All patients were age ≥ 18 years and were about to begin a first cancer treatment regimen containing cisplatin or doxorubicin. Patients receiving concurrent radiotherapy or interferon, or patients with clinical evidence of bowel obstruction or symptomatic brain metastases, were excluded. All received a 5-HT3 receptor antagonist before chemotherapy. Dexamethasone or other corticosteroids were allowed, as were all ancillary treatments as appropriate for the control of symptoms caused by the cancer or its treatment. The type and amount of antiemetic medication received during treatment on Days 2–5 were not regulated. Patients randomized to the experimental treatment groups were assigned to wear acupressure or acustimulation wrist bands. The patients in the control group received standard care only. To have data from a homogeneous sample of patients not confounded by an intervention, the patients in the experimental group were not considered in the current analysis, nor were men, or women receiving cisplatin, or women not being treated for breast carcinoma. The complete study is reported elsewhere.17 The institutional review board of the University of Rochester and of each participating site approved the protocol. Written informed consent was obtained from each subject.
Demographic data and detailed information regarding chemotherapy agents, antiemetics, and other medications administered at the time of each chemotherapy treatment were abstracted from medical records by study personnel. Patients also were asked if they had nausea during pregnancy (coded 1 = yes, 0 = no or not applicable) and whether they were susceptible to motion sickness (coded 1 = yes, 0 = no).
Expectation of developing nausea as a result of treatment was assessed before treatment using a patient-completed, paper-and-pencil questionnaire administered in the treatment center by a member of the research staff. It began with a written explanation, “We are interested in what you have learned about some possible side effects of chemotherapy and where this information came from” and contained the question, “Before you spoke to your doctor about possible side effects of chemotherapy, what did you think the chances were that you would have severe nausea from your treatment?” A second identically worded question substituted the word “vomiting” for the words “severe nausea.” The possible responses were “very unlikely,” “unlikely,” “about even chance,” “likely,” and “very likely,” which were coded 1–5, respectively, and treated as a continuous variable for analysis purposes. The same responses were solicited for the questions, “What did your doctor say your chances of having the severe nausea were?” and “What did your doctor say your chances of having the vomiting were?” These latter two questions were then repeated substituting “nurse” for “doctor.”
A second questionnaire, also completed by patients before treatment, directly assessed their expectations of developing nausea on a 5-point Likert-scale, anchored at one end by 1 (“I am certain I will not have this”) and at the other end by 5 (“ I am certain I will have this”). Patients were also asked a multiple-choice question, “What do you think your level of nausea will be at its worst after this treatment?” The possible responses were “very mild or none at all,” “mild,” “moderate,” “severe,” “very severe,” and “intolerable,” which were coded 1–6, respectively. Two additional questions relating to nausea susceptibility/expectancy were also included. The first was, “In general, do you think you are more susceptible to nausea than your friends and family are?” The second was, “Based upon what you know of yourself, how much chemotherapy-related nausea do you think you will have compared to other patients receiving the same treatments?” Possible responses for each question were “more” (scored as 3), “about the same” (scored as 2), and “less” (scored as 1).
Nausea and emesis were measured by a 5-day patient report diary developed for this purpose by Burish et al.18 and Carey and Burish19 and used by us in previous large clinical trials. Each day is divided into four segments (morning, afternoon, evening, and night) and patients reported the severity of nausea and the number of emesis episodes for each period of the day. Reporting of nausea and emesis started with the evening period on the day of treatment and continued throughout the following 4 days for a total of 18 assessment times. Severity of nausea was assessed on a 7-point rating scale, anchored at one end by 1 (“Not at all nauseated”) and at the other end by 7 (“Extremely nauseated”). The description, “Moderately nauseated,” was centered on the scale below the 4. Patients were given this questionnaire to complete at home over the several days immediately after treatment and were instructed to return it to the clinic site, using the stamped, preaddressed envelope that was provided.
The analyses described are secondary analyses of an existing database and were limited to female patients with breast carcinoma receiving doxorubicin-based chemotherapy in the control group because these criteria provided the largest homogeneous sample of patients receiving standard care.
The 5-day patient report diary was used to calculate an average level of nausea severity for each patient and to determine whether patients experienced emesis at any point after treatment was received. In addition, we coded a patient as having experienced severe nausea if she reported a nausea severity score of 6 or 7 at any of the 18 assessment times. The three measures of nausea and emesis (i.e., average nausea, any severe nausea, and any vomiting) are used as outcome measures throughout these analyses.
The initial analyses were correlational to determine which patient variables were significantly correlated with the three outcomes of nausea and emesis. Expectancy variables that had a significant correlation with any of the nausea and emesis outcome measures were examined further using regression analyses controlling for appropriate patient variables.
Two hundred and one female patients with breast carcinoma treated with doxorubicin as outpatients were eligible. Of these 201 patients, 194 (97%) patients provided evaluable data. The majority (85%) of these women were white; 64% had received some college education and 94% had completed high school. Of the 194 patients, 186 (96%) received some type of antiemetic after treatment.
Prevalence of nausea and emesis
Forty-two (22%) of the 194 women reported having severe nausea (a rating of 6 or 7 on the 7-point scale) on Day 1 of chemotherapy (acute nausea). Fifty-eight (30%) of the patients had severe nausea at some point during Days 2–5 (delayed nausea). The proportion of patients with severe nausea on each of these 4 days was 16%, 17%, 12%, and 10%, respectively. Overall, 73 (38%) of the women experienced severe nausea at some point during the 5 days. Average nausea scores for the 5 days were 2.3, 2.3, 2.6, 2.3, and 2.0, respectively, with the overall average nausea score being 2.3. The proportion of women who experienced emesis on each of the 5 days was 17%, 15%, 14%, 14%, and 10%, respectively, with 37% of the women having emesis at least once (Fig. 1). Spearman rho correlations between the occurrence of emesis and the report of severe nausea on the 5 assessment days ranged from a high of 0.66 on the day of treatment to a low of 0.37 on Day 3 (all P values were < 0.001).
Correlations and comparisons among study variables
Pearson and Spearman correlation analyses, as appropriate, demonstrated a significant relation between several of the study variables and the three nausea and emesis outcome measures. As expected, age was found to be related significantly to all three measures with correlations ranging from −0.18 to −0.26 (all Ps < 0.05). Nausea during pregnancy (1 = yes, 0 = no or not applicable) was found to be correlated significantly with any vomiting, (rho = 0.30) and with average nausea (rho = 0.24), both Ps ≥ 0.001. Susceptibility to motion sickness (yes vs. no), however, was not found related significantly to any of the nausea/emesis measures.
Patients' expectations for the worst level of nausea they would experience after treatment were significantly correlated with subsequent severe nausea (rho = 0.16, P = 0.03) and with average nausea (rho = 0.20, P = 0.006), but not with the occurrence of emesis. Patients' levels of certainty that they would experience nausea after treatment were related significantly to average nausea (r = 0.14, P = 0.05), but not to severe nausea or the occurrence of emesis. Patients' beliefs that they were more susceptible to nausea than friends and family as well as their expectations concerning how much chemotherapy-related nausea they would have compared with other patients each was correlated significantly with the three outcomes of nausea and emesis. Correlations ranged from 0.16 to 0.19 (all P values were < 0.05).
The strongest correlations between any of our expectancy measures and the nausea and emesis outcomes derived from the question, “Before you spoke to your doctor about possible side effects of chemotherapy, what did you think the chances were that you would have severe nausea from your treatment?” Response to this question was related significantly to all three nausea and emesis measures with correlations ranging from 0.20 to 0.27 (Ps < 0.01). The correlations between the response to the companion question concerning expectations of emesis and the three nausea and emesis outcomes were slightly smaller but still significant and ranged from 0.14 to 0.25 (Ps < 0.01). Contrary to the above findings, responses to the four questions querying what the patient's physician and nurse said (as reported by the patient) concerning the chances of having severe nausea and/or emesis were not correlated with any of the three nausea and emesis outcomes (Ps > 0.1). Table 1 presents the correlations for selected study variables of interest.
|Any severe nausea|
|0.12||0.11||0.14b||−0.16b||Certainty that nausea will occur|
|0.16b||0.11||0.20a||−0.13||0.63a||Expected nausea at its worst|
|0.27a||0.20a||0.21a||−0.14b||0.43a||0.37a||Preclinicc nausea expectancy|
|0.25a||0.19a||0.14a||−0.16b||0.35a||0.35a||0.80c||Preclinicc emesis expectancy|
|0.14||0.30a||0.24b||0.01||0.27a||0.15b||0.14b||0.17b||Nausea during pregnancy|
|0.17b||0.18b||0.19b||−0.19a||0.54a||0.52a||0.37a||0.36a||0.17b||Compared with other patientsd|
|0.17b||0.17b||0.16b||−0.13||0.34a||0.31a||0.26a||0.28a||0.25a||0.50a||Susceptibility to nauseae|
The response to the question concerning the likelihood of experiencing severe nausea that the patient had before speaking to her physician about side effects was compared, using a Student t test for paired data, with the response from the query concerning what the patient's physician said (as reported by the patient) concerning the likelihood of having severe nausea. Patients' expectations that they would have severe nausea were significantly higher than their physicians' expectations (scale range = 1–5; patient mean = 3.3, physician mean = 2.6; P < 0.001). Patients' higher expectancies of symptom likelihood held true for emesis as well (patient mean = 3.3, physician mean = 2.8; P < 0.001) and also as compared with emesis expectancies given by nurses (both P values < 0.001). The mean expectancy values for severe nausea and emesis from nurses (as reported by patients) were 2.7 and 2.9, respectively.
Expectancy as a predictor of nausea and emesis
We further examined the relation between each of the expectancy measures and the subsequent report of severe nausea separately using binary logistic regression analyses, with any severe nausea as the dependent variable. Each of the analyses controlled for the two patient variables (i.e., age and nausea during pregnancy) that were correlated significantly with the occurrence of nausea in our previous analyses by entering them at the first step in the equation. Of these analyses, only the one that examined nausea expectancies held by patients before speaking to the physician concerning possible side effects of chemotherapy was statistically significant (odds ratio = 1.56, P = 0.002). Thirty-one subjects answered the question concerning the likelihood that they would experience severe nausea as very likely and 16 subjects answered very unlikely. This question was a robust predictor of severe nausea, i.e., with 68% of the patients who answered very likely reporting severe nausea on ≥ 1 of the 5 assessment days compared with < 13% of the patients who answered very unlikely—a 5-fold increase. The patients who rated their expectation of nausea as likely (n = 49), about even (n = 54), or unlikely (n = 42) reported severe nausea 39%, 35%, and 29% of the time, respectively (Fig. 2).
Analyses identical to those described in the preceding paragraph, with the exception that any vomiting was substituted for severe nausea as the dependent variable, all failed to reach statistical significance. A similarly structured set of hierarchical regression analyses were used to examine the relation between the expectancy measures and average nausea. None of these reached conventional levels of statistical significance, but the analysis examining expectancies for severe nausea held before speaking to the physician regarding possible side effects of chemotherapy closely approached significance (P = 0.054). Mean nausea across the 5-day assessment period for the 5 expectancy groupings ranked in order of expected nausea, starting with very likely, were 2.9, 2.3, 2.3, 2.3, and 1.5, respectively.
In this sample of 194 patients with breast carcinoma, the expectancy of nausea assessed before receipt of the first doxorubicin-based chemotherapy treatment was found to be a strong predictor of subsequent nausea and in fact was a more robust predictor of severe nausea than previously reported factors, including age, nausea during pregnancy, and susceptibility to motion sickness. The strongest predictor among expectancy measures was the one that assessed response expectancies for severe nausea held by the patient before she spoke with her physician. This remained a significant predictor of nausea even after controlling for age and nausea during pregnancy, the two patient factors found to be significantly correlated with nausea and emesis in this patient sample. Women who believed it was very likely that they would have severe nausea after chemotherapy were five times more likely to experience severe nausea than fellow patients who believed its occurrence would be very unlikely. Patient responses also were analyzed for average nausea during the 5-day interval. Again, a pattern was evident, with patients in the group who reported an expectation of very unlikely having a mean nausea score of only 1.5 on the 7-point scale, and the patients with the very likely expectation reporting a mean nausea of 2.9, almost twice as high.
We also found that despite administration of a 5-HT3 receptor antagonist, greater than one-third of female patients receiving doxorubicin-based chemotherapy had severe nausea at some point after treatment, with approximately 22% of these patients experiencing the symptom on the day of treatment. The current study data must be interpreted with some caution because it was not mandated that patients receive dexamethasone, along with their 5-HT3 receptor antagonist, as is recommended in the current American Society of Clinical Oncology guidelines, which were published 1 month after the study began accrual.14 Nevertheless, considering that most, if not all, of the oncologists recruiting patients to the current study likely followed the guidelines and provided dexamethasone, the high prevalence of severe nausea after treatment is noteworthy.
Surprisingly, patients' nausea expectancies held before speaking to their oncologist were found to be a better predictor of subsequent nausea than expectancy measured after this encounter. This might be accounted for, in part, by the fact that information concerning the likelihood of treatment-induced severe nausea provided to patients by both physicians and nurses is overly optimistic. This statement is subject to a potentially important caveat as it is actually information from these health care professionals as interpreted by the patient that was assessed, because we did not directly query either physicians or nurses and do not know what, if any, written information patients were given or what patients were actually told concerning the likelihood of experiencing nausea. A second factor that most likely contributed to the greater predictive accuracy of the patients' expectancies is their personal knowledge of how susceptible they were, in general, to nausea. Evidence supporting this comes from the two questions we asked (one in relation to friends and family and the other in relation to other patients) to assess patients' beliefs concerning their relative susceptibility to nausea. Each showed significant correlations with expectancies for symptom occurrence held by patients before speaking to their health care team and with subsequent symptom report.
Findings from the current study corroborate previous research from our group, as well as from others, which demonstrated significant correlations between expectations for nausea and subsequent nausea.5–10
It is not possible, however, to ascertain from either the current study or these previous reports whether expectancies for nausea are merely predictive of subsequent nausea or are to some extent causal as well. This latter possibility is supported by a review by the British National Coordinating Centre for Health Technology Assessment, which provides persuasive support that expectancies play the central role in the placebo effect and are associated with the development of medical symptoms.20 The authors concluded that the 93 reviewed studies of placebo and expectancy in the clinical environment (selected from 47,600 references screened for their relevancy and quality) support the hypothesis that expectancies are the principal mechanism by which placebos have their effect and that negative outcome expectancy affects health outcomes.
Additional evidence that expectancies may have a causal role in symptom development can be inferred from experiments that have successfully reduced symptoms by manipulating response expectancies. Experiments designed to enhance positive expectations have shown that enhancing expectations for a positive outcome can increase gastrointestinal motility in patients undergoing abdominal surgery,21 diminish experimentally induced pain,22–24 reduce blood loss during elective spinal surgery,25 modulate cellular immune dysregulation,26 enhance immune reaction to an injection of tuberculin,27 potentiate the release of endogenous dopamine,28 and reduce nausea.29–31
To our knowledge, it remains unknown whether querying about expectancies affects their magnitude in some way or whether talking to patients who have high nausea and emesis expectancies can, in some way, reassure them and perhaps reduce those expectancies and, furthermore, whether expectancies altered by this process would result in reduced symptoms. Further research pertaining to this may be very fruitful with regard to the potential benefit the patient's QOL. It is also unknown at this point whether physicians do, or should, take patients' expectancies for nausea and emesis into account when prescribing antiemetic medications. It is certainly the case that the three major antiemetic treatment guidelines32–34 do not include patients' expectancies for developing nausea and emesis as a risk factor. Further studies confirming this as a risk factor are warranted, as are studies examining the benefit to patient QOL from modifying antiemetic treatment guidelines to take into account symptom expectancies. Finally, ethically acceptable interventions that are designed to reduce patients' nausea expectancies or increase their expectancies of nausea control should be developed and studied.
The authors thank the following Community Clinical Oncology Program (CCOP) Principal Investigators who provided and cared for study patients while they collected data: Tom R Fitch, M.D. (Mayo Clinic Scottsdale CCOP, Scottsdale, AZ); Brian Issel, M.D. (Hawaii MBCCOP, Honolulu, HI); Patrick J. Flynn, M.D. (Metro-Minnesota CCOP, St. Louis Park, MN); Richard J. Rosenbluth, M.D. (Northern New Jersey CCOP, Hackensack, NJ); H. Irving Pierce, M.D. (former Principal Investigator [deceased] Northwest CCOP, Tacoma, WA); and Jeffrey J. Kirshner, M.D. (Syracuse Hematology-Oncology CCOP, Syracuse, NY).
The authors also thank Shonda Ranson for her outstanding work in assuring the accuracy and completeness of the database.
- 2Impact of cancer-related fatigue on the lives of patients: new findings from the fatigue coalition. In: MartyM, PecorelliS, editors. Fatigue and cancer. Amsterdam: Elsevier, 2001: 3–16., , , et al.
- 3Supportive care in cancer. New York: Marcel Deckker, 1999., , .
- 6Postchemotherapy nausea and vomiting in cancer patients receiving outpatient chemotherapy. J Psychosoc Oncol. 1991; 9: 117–130., , , .
- 10Specific, and not generalized, response expectancies predict chemotherapy-induced nausea severity. Proceedings of the Society of Behavioral Medicine's Twenty-Second Annual Meeting. Ann Behav Med. 2001; 23: S074., , .
- 33NCCN antiemesis practice guidelines. Oncology (Huntingt). 1997; 11: 57–89., , , et al.