How do patient expectancies, quality of life, and postchemotherapy nausea interrelate?

Authors

  • Ben Colagiuri BPsych,

    Corresponding author
    1. Medical Psychology Research Unit, School of Psychology, University of Sydney, Sydney, New South Wales, Australia
    • School of Psychology, A18, University of Sydney, 2006 NSW, Australia===

    Search for more papers by this author
    • Mr. Colagiuri is a recipient of an Australian Postgraduate Award.

    • Fax: (011) 61 2 9036 5223

  • Joseph A. Roscoe PhD,

    1. Behavioral Medicine Unit, University of Rochester Community Clinical Oncology Program Research Base, University of Rochester James P. Wilmot Cancer Center, Rochester, New York
    Search for more papers by this author
  • Gary R. Morrow MS, PhD,

    1. Behavioral Medicine Unit, University of Rochester Community Clinical Oncology Program Research Base, University of Rochester James P. Wilmot Cancer Center, Rochester, New York
    Search for more papers by this author
  • James N. Atkins MD,

    1. Southeast Cancer Control Consortium, Community Clinical Oncology Program, Goldsboro, North Carolina
    Search for more papers by this author
  • Jeffrey K. Giguere MD,

    1. Cancer Centers of the Carolinas Community Clinical Oncology Program, Greenville, South Carolina
    Search for more papers by this author
  • Lauren K. Colman MD

    1. Northwest Community Clinical Oncology Program, Tacoma, Washington
    Search for more papers by this author

Abstract

BACKGROUND.

Increasing evidence suggests a relation between patient expectancies and chemotherapy-induced nausea. However, this research has often failed to adequately control for other possible contributing factors. In the current study, the contribution of patient expectancies to the occurrence and severity of postchemotherapy nausea was examined using more stringent statistical techniques (namely hierarchical regression) than other similar studies that have relied on bivariate correlations, chi-square tests, and stepwise regression, and further extended upon previous research by including quality of life (QoL) in the analysis.

METHODS.

In all, 671 first-time chemotherapy patients taking part in a trial comparing antiemetic regimens answered questions regarding their expectancies for experiencing nausea. Patients then completed a diary assessing both the occurrence and severity of their nausea in the 4 days after their first infusion.

RESULTS.

Stronger expectancies for nausea corresponded with greater average and peak nausea after chemotherapy and this was after controlling for age, sex, susceptibility to motion sickness, diagnosis, and QoL. Interestingly, patients classified as highly expectant (first quartile) experienced significantly greater average and peak nausea than those classified as somewhat expectant, slightly expectant, and not expectant (second, third, and fourth quartiles, respectively), whereas there were no significant differences between these lower levels of expectancy. Furthermore, increases in average nausea led to a significant reduction in QoL after chemotherapy.

CONCLUSIONS.

Patient expectancies appear to contribute to postchemotherapy nausea and patients that are highly expectant of experiencing nausea appear to be at particular risk. Interventions that target these patients should reduce the burden of nausea and may also improve QoL. Cancer 2008. © 2008 American Cancer Society.

Although methods for the prevention and control of emesis have improved greatly, nausea continues to be a significant burden to patients undergoing chemotherapy. The vast majority of cancer patients still report experiencing nausea at some point during their chemotherapy treatment1 and it is particularly troublesome for patients receiving highly emetogenic chemotherapies such as cisplatin and doxorubicin.2, 3 Nausea is often rated as 1 of the most severe and debilitating side effects of the treatment4–6 and can negatively affect patients' nutritional habits, ability to work, and motivation to follow recommended treatment regimens.6, 7 It is also inherently unpleasant and contributes significantly to poorer quality of life (QoL).7–11

Given the continued high prevalence of chemotherapy-related nausea and its detrimental impact on QoL, determining what causes this nausea and how it might be reduced remains an important goal for clinicians and researchers alike. Patient expectancies represent 1 possible causal mechanism that is open to manipulation. Indeed, a growing body of evidence has demonstrated a link between patient expectancies for nausea and their actual experience of both anticipatory12, 13 and postchemotherapy nausea.1, 14–18 In 1 study, patients who believed that they were “very likely” to experience severe nausea from their chemotherapy treatment were 5 times more likely to subsequently experience severe nausea than those who believed that they were “very unlikely” to do so.18

This type of evidence for expectancy-induced nausea appears to fit in well with current research on the placebo effect, in which expectancy is believed to play a key role.19–22 Typically, this research shows that positive expectancies can lead to beneficial health outcomes (eg, pain relief).23, 24 However, there is also evidence to suggest that negative expectancies can lead to adverse outcomes, often labeled the nocebo effect.25, 26 One study found that when patients participating in a clinical trial of sulfinpyrazone or aspirin for unstable angina were warned about gastrointestinal discomfort, they were much more likely to report subsequent gastrointestinal discomfort during treatment and had a 6-fold drop-out rate compared with patients who were not warned about this possibility.27

However, 1 limitation to previous research on expectancies and chemotherapy-related nausea is that it has often failed to adequately control for other possible contributing factors, with the majority of these studies relying on bivariate correlations,18 chi-square tests,17 and/or stepwise regression.15, 16, 18 Although this undoubtedly stems from the inherent impossibility of randomly assigning patients to different levels of expectancy, it nonetheless limits the certainty with which we can attribute a causal role to expectancy. Furthermore, it is unclear whether there is a linear relation between patient expectancies and nausea or whether there is a certain level of expectancy required to produce an effect. If the relation is nonlinear, it could be the case that those with particularly high levels of expectancy experience more nausea than those with lower levels of expectancy. Conversely, perhaps having very low expectancies for nausea has a protective effect and results in less nausea in these patients compared with those who have higher levels of expectancy.

The current study aimed to overcome these limitations to determine, with more confidence, whether expectancy contributes to postchemotherapy nausea. To do this, we adopted more conservative statistical techniques, namely, hierarchical regression, to evaluate the independent contribution of expectancy to postchemotherapy nausea over and above other potential contributing factors such as age, sex, susceptibility to motion sickness, and baseline QoL. Although 2 previous studies1, 14 investigating the relation between expectancy and postchemotherapy nausea have employed hierarchical regression, both involved only a small number of patients relative to the number of variables included in their models. We also categorized patients into groups according to their level of expectancy to investigate whether a particular level of expectancy either heightened or protected against the occurrence of nausea in patients undergoing chemotherapy. In addition, we conducted analyses examining the relation between pretreatment QoL and nausea expectancies as well as between postchemotherapy nausea and posttreatment QoL.

MATERIALS AND METHODS

Participants

The patients were participants in a multicenter trial comparing antiemetic regimens for the treatment of delayed nausea. They were enrolled from 18 private practice oncology groups in the U.S. between June 12, 2001, and June 11, 2004. Eligible patients were aged ≥ 18 years with any cancer diagnosis and were about to receive their first chemotherapy treatment containing doxorubicin and antiemetic prophylaxis with a 5-HT3 receptor antagonist, ondansetron, granisetron, or dolasetron plus dexamethasone or the equivalent dose of intravenous methylprednisolone on the day of treatment. Patients were randomized to receive 1 of 3 regimens for the control of delayed nausea on Days 2 and 3 after chemotherapy: Arm 1, which was comprised of prochlorperazine administered at a dose of 10 mg orally every 8 hours; Arm 2, which was comprised of any first-generation 5-HT3 receptor antagonist using standard dosage; or Arm 3, which was comprised of prochlorperazine administered at a dose of 10 mg orally as needed. Written informed consent was obtained from each participant. Only minimal differences were observed among study arms (see Hickok et al2 for additional details). In the following report, the study arm was statistically controlled for as appropriate.

Measures

At the time of enrolment, patients completed a pretreatment assessment comprising an on-study interview, a QoL questionnaire, and an expectancy questionnaire. The on-study interview contained questions regarding demographics, previous treatment, and susceptibility to motion sickness. QoL was assessed using the Functional Assessment of Cancer Therapy Scale-General (FACT-G; version 4). It is a widely used and well-validated measure for assessing QoL in cancer patients undergoing chemotherapy treatment.28 The expectancy questionnaire has been previously used by our group29 and contained questions assessing patient expectancies for post-treatment nausea, vomiting, fatigue, and hair loss. Four of these questions assessed expectancies for nausea. One required the patients to rate the likelihood that they would experience nausea after their chemotherapy treatment on a 5-point scale ranging from 1 (“I am certain I will not have this”) to 5 (“I am certain I will have this”). A second question required the patients to rate the expected severity of their postchemotherapy nausea as “very mild or none at all,” “mild,” “moderate,” “severe,” “very severe,” or “intolerable.” A third question asked the patients to rate their perceived susceptibility to nausea compared with their friends and family as either “more,” “less,” or “the same.” A final question required patients to rate the likelihood of experiencing chemotherapy-related nausea compared with other cancer patients with the same diagnosis and undergoing the same treatment, again as “more,” “less,” or “the same.” These 4 expectancy questions were then combined by averaging z-scores to produce a single expectancy measure.

Postchemotherapy nausea was assessed via a 4-day patient diary developed by Burish et al.30 and Carey and Burish31 specifically for this purpose. The diary required the patients to rate the severity of their nausea on a 7-point scale ranging from 1 (“not at all nauseated”) to 7 (“extremely nauseated”) for the morning, afternoon, evening, and night separately for the 4 days after their first chemotherapy treatment. The diary was then used to calculate average nausea over this period as well as peak nausea, classified as the highest severity rating for nausea at any time in the 4 days after treatment. At the end of the fourth day, patients were again given the FACT-G to complete.

Statistical Analyses

Hierarchical regression was used to determine the predictors of postchemotherapy nausea. Age, sex, and susceptibility to motion sickness comprised the first step, followed by diagnosis, study arm, baseline QoL, and, finally, expectancy. Diagnosis was dummy coded using breast cancer as the reference group and combining myeloma, endometrial, sarcoma, and bladder cancer patients into a single group because of their low numbers. To assess the impact of level of expectancy on nausea, we created 4 approximately equal groups using quartiles based on the combined expectancy measure. We classified the groups as not expectant (0–25th percentile), slightly expectant (26–50th percentile), somewhat expectant (50–75th percentile), and highly expectant (76–100th percentile). Analysis of covariance (ANCOVA) was then used to compare these groups with follow-up pairwise comparisons using the Bonferroni procedure to control the Type I error rate. Hierarchical regression was used to assess the impact of postchemotherapy nausea on QoL. Here, age and sex were entered as the first step, followed by study arm, then baseline QoL, and then nausea in the final step. Finally, partial correlation was used to assess the relation between QoL before treatment and expectancy, controlling for age, sex, susceptibility to motion sickness, and diagnosis. The analyses were conducted separately for average nausea and peak nausea. All analyses were conducted using SPSS software (version 12.0; SPSS Inc, Chicago, Ill) and the results were considered significant when P < .05.

RESULTS

In all, 691 patients enrolled in the study, of which 671 provided evaluable data with an average age of 53 years (range, 25–90 years). The majority of these patients were female (94%) and white (88%), and had received some college education (59%). Approximately 90% had breast cancer, 9% had lymphoma, and the remaining 1% had a mix of myeloma, endometrial, sarcoma, and bladder cancers.

Overview of Nausea and QoL

In all, 562 patients (84%) reported at least some nausea in the 4 days after treatment and 165 (25%) reported severe nausea (rating of 6 or 7 on the 7-point scale). Overall, mean nausea during the 4 days after chemotherapy was 2.2 (standard deviation [SD] = 1.2) and the mean peak nausea was 4.0 (SD = 2.1). Before their first chemotherapy treatment, patients had a mean QoL of 86.3 (SD = 13.0), which decreased to 76.3 (SD = 15.8) after the treatment.

Expectancy and Nausea

For average nausea, the overall model in which age, sex, diagnosis, susceptibility to motion sickness, baseline QoL, and expectancy were included was significant and accounted for 16% of the variance (R2 = 0.16; F8,659 = 15.3 [P < .001]) (Table 1). Expectancy had a significant impact on average nausea. Specifically, an increase of 1 unit on the expectancy measure was associated with a 0.27 increase in average nausea after controlling for all other variables in the model (R2 change = 0.02; b = −0.27; t659 = 4.18 [P < .001]). Age and baseline QoL were the only other significant predictors in the model. An increase in age of 10 years corresponded to a decrease of 0.3 points on average nausea (b = −0.03; t659 = 7.12 [P < .001]), and a 10-point increase in baseline QoL corresponded to a 0.1 decrease in average nausea (b = −0.008; t659 = 2.26 [P = .02]).

Table 1. Final Step in the Hierarchical Regression Analysis of Predictors of Postchemotherapy Nausea
Average nauseabSEβP
  • SE indicates standard error; QoL, quality of life.

  • *

    Significant at P < .001.

  • Dummy coded with breast cancer as the reference group.

  • Significant at P < .05.

Age*−0.030.01−.27<.001
Gender0.130.28.02.65
Motion sickness0.020.10.01.84
Lymphoma−0.400.21−.10.06
Other diagnosis0.420.60.03.48
Study group0.030.05.02.53
Baseline QoL−0.010.00−.09.024
Expectancy*0.270.06.17<.001
Peak nauseabSEβP
Age*−0.050.01−.26<.001
Gender−0.100.49−.01.83
Motion sickness0.110.17.02.52
Lymphoma−0.840.37−.12.023
Other diagnosis0.201.03.01.85
Study group0.130.09.05.17
Baseline QoL,*−0.010.01−.08.048
Expectancy*0.430.11.15<.001

For peak nausea, the overall model was also significant and accounted for 15% of the variance (R2 = 0.15; F8,659 = 14.3 [P < .001]) (Table 1). Expectancy was again significant after controlling for all other variables. Here, an increase of 1 unit on the expectancy measure corresponded to a 0.43 increase in peak nausea (R2 change = 0.02; b = −0.43; t659 = 3.86 [P < .001]). Similar to average nausea, age and baseline QoL were also found to be significant predictors. A 10-year increase in age was associated with a 0.51 decrease in peak nausea (b = −0.05; t659 = 6.97 [P < .001]). A 10-point increase in baseline QoL corresponded to a 0.1 decrease in peak nausea (b = −0.01; t659 = 1.98 [P < .048]). In addition, lymphoma patients reported 0.84 points less peak nausea than breast cancer patients (b = −0.84; t659 = 2.28 [P < .02]). Overall, expectancy accounted for 2% of the variance in both average nausea and peak nausea after controlling for all other variables.

Level of Expectancy and Nausea

Figure 1A shows average nausea across the different levels of expectancy. The ANCOVA, with age, sex, susceptibility to motion sickness, diagnosis, baseline QoL, and study group as covariates, revealed that average nausea differed significantly as a function of level of expectancy (F3,658 = 8.9; P < .001). Using pairwise comparisons, highly expectant patients reported significantly higher levels of average nausea than all other levels of expectancy. Specifically, highly expectant patients reported average nausea as 0.62 points higher than not expectant patients (t658 = 4.49; P < .001), 0.56 points higher than slightly expectant patients (t658 = 4.12; P < .001), and 0.47 points higher than somewhat expectant patients (t658 = 3.61; P = .002). There were no significant differences in average nausea noted between somewhat expectant, slightly expectant, and not expectant patients (highest t659 = 1.18; P = 1).

Figure 1.

Covariate-adjusted mean (± standard error of the mean) (A) average and (B) peak nausea by level of expectancy. Highly expectant individuals reported both more average nausea and higher peak nausea than all other expectancy levels. Asterisk indicates P < .05; double asterisk, P < .01; triple asterisk, P < .001. No other differences were statistically significant.

Figure 1B shows peak nausea for each level of expectancy. As with average nausea, the ANCOVA revealed significant differences in peak nausea as a function of level of expectancy (F3658 = 6.5; P < .001). Highly expectant patients reported peak nausea 0.94 points higher than not expectant patients (t658 = 3.88; P = .001), 0.91 points higher than slightly expectant patients (t658 = 3.76; P = .001), and 0.63 higher than somewhat expectant patients (t658 = 2.81; P = .03). Again, there were no differences between somewhat expectant, slightly expectant, and not expectant patients in terms of peak nausea (highest t658 = 1.41; P = .96).

Expectancy and Pretreatment QoL

The partial correlation between expectancy and pretreatment QoL was significant (r = −0.29; t660 = 7.79 [P < .001]). This means that, after controlling for age, gender, susceptibility to motion sickness, and diagnosis, stronger expectancy for postchemotherapy nausea was associated with lower QoL before treatment.

Impact of Nausea on Posttreatment QoL

The overall model, with age, sex, baseline QoL, average nausea, and peak nausea included, was significant and accounted for 55% of the variability in postchemotherapy QoL (R2 = 0.55; F8657 = 100 [P < .001]). The only 2 significant predictors in the final model (Table 2) were baseline QoL (b = 0.58; t657 = 17.6 [P < .001]) and average nausea (b = −5.1; t657 = 8.80 [P < .001]). This means that even after controlling for age, sex, diagnosis, baseline QoL, study group, and peak nausea, a 1-point increase in average nausea corresponded to a 5-point decrease in postchemotherapy QoL. Peak nausea was found to be marginally nonsignificant (b = −0.65; t657 = 1.92 [P = .056]). Together, average and peak nausea accounted for 20% of the variability in postchemotherapy QoL (R2 change = 0.20; F2657 = 143 [P < .001]).

Table 2. Final Step in the Hierarchical Regression Analysis of Predictors of Quality of Life After Chemotherapy
Quality of LifebSEβP
  • SE indicates standard error; QoL, quality of life.

  • *

    Dummy coded with breast cancer as the reference group.

  • Significant at P < .001.

Age0.050.04.04.20
Gender1.092.61.02.68
Lymphoma*−0.161.99.00.94
Other diagnosis*−5.715.53−.03.30
Baseline QoL0.580.03.47<.001
Study group0.190.51.01.71
Average nausea−5.110.58−.40<.001
Peak nausea−0.650.34−.09.056

DISCUSSION

As with previous studies,1, 14–18 patient expectancies for nausea before chemotherapy were a significant predictor of the actual occurrence of nausea after treatment. Although expectancies accounted for only 2% of the variance in both average and peak nausea, this was after controlling for age, sex, susceptibility to motion sickness, and QoL. As such, these results demonstrate an independent contribution of expectancy to postchemotherapy nausea and thereby provide strong support for expectancy as a causal factor in chemotherapy-related nausea.

A novel and interesting finding was that patients classified as highly expectant of nausea experienced higher levels of both average and peak nausea than all other levels of expectancy, whereas none of the other lower levels of expectancy (somewhat expectant, slightly expectant, and not expectant) were found to differ significantly from one another. This suggests that patients who are highly expectant of experiencing nausea are particularly at risk of actually experiencing nausea after their chemotherapy treatment, whereas those with lower levels of expectancy share a similar decreased risk. This was also after controlling for age, sex, susceptibility to motion sickness, and QoL, and again points to a causal influence of expectancy on postchemotherapy nausea.

A second novel finding was that stronger expectancies for nausea were associated with lower QoL before treatment. Here, the direction of causality is uncertain. It is possible that having strong expectancies for nausea after chemotherapy could detract from QoL before treatment by creating apprehension and stress regarding the treatment. Conversely, it could be that those with lower QoL before treatment are in a poorer state of health and therefore expect to be affected more by the chemotherapy treatment.

Increases in average nausea resulted in poorer QoL posttreatment, a result also consistent with previous research.7–11 Here, a 1-point increase in average nausea corresponded to a 5-point decrease in QoL. Interestingly, peak nausea did not have a significant impact on QoL after controlling for average nausea. This tends to suggest that the continued presence of nausea has a more debilitating impact on chemotherapy patients' QoL compared with a severe bout of nausea. Overall, average nausea and peak nausea independently of other variables accounted for 20% of the variability in postchemotherapy QoL, a very large effect. Clearly, not only does nausea remain prevalent, it also continues to be a significant burden to cancer patients' QoL.

Given the findings of the current study, patient expectancies reflect a possible point of intervention to reduce chemotherapy-induced nausea as well as to improve cancer patients' QoL. A variety of studies in areas other than oncology have already shown that expectancy manipulations can have a beneficial impact on health outcomes. Perhaps most relevant is evidence that enhancing positive expectancies can lead to decreased postoperative nausea in patients undergoing major gynecologic surgery32 and protect against seasickness in naval cadets.33 However, these types of manipulations may be more difficult to implement for cancer patients. First, it would be unethical to provide patients with unrealistically low expectancies regarding the likelihood of experiencing nausea as a result of their chemotherapy. Second, in addition to what cancer patients are told to expect from their treating health professionals, they will also draw on information from their family, friends, and the media as well as any prior experience of nausea, such as motion sickness, morning sickness, and medication-induced nausea. As such, any successful expectancy manipulation would have to consider the source of the patient's expectancies and the significance attributed to that source before attempting to adjust any maladaptive expectancies.

The finding that highly expectant patients experience more nausea than not expectant, slightly expectant, and somewhat expectant patients, and that these latter 3 groups did not differ significantly from each other, may provide an answer. It suggests that completely eradicating expectancies for nausea is not necessary to gain a significant clinical improvement. For example, if a patient was initially highly expectant but through discussion or some other manipulation became somewhat expectant, a significant decrease in both average nausea and peak nausea would be expected with a corresponding improvement in QoL. Further reducing these expectancies to the level of slightly expectant, however, would produce only minimal additional benefit. Thus, a successful intervention might focus on patients that are highly expectant of experiencing nausea with the aim reducing these maladaptive expectancies rather than removing all expectancy. In doing so, the patient would remain aware that there is a possibility of experiencing nausea, thereby avoiding any ethical concerns regarding giving patients unrealistically low expectations of experiencing nausea. Furthermore, a small reduction in expectancy is likely to be much easier to achieve than convincing a highly expectant patient that he/she is very unlikely to experience nausea after chemotherapy.

One possibility along these lines is to include a discussion of the patient's expectancies regarding side effects as part of the chemotherapy education. This would allow for the identification of highly expectant patients and provide an excellent opportunity to discuss and challenge these potentially maladaptive expectancies. Given the correlation between expectancy and pretreatment QoL, reducing strong expectancies for nausea during chemotherapy education may also improve patients' pretreatment QoL. To the best of our knowledge, patient expectancies are rarely, if ever, addressed in chemotherapy education.

Some caution is required when interpreting the relation between expectancy and postchemotherapy nausea. Although we included susceptibility to motion sickness in our analysis, which did not appear to influence postchemotherapy nausea, it is possible that other previous experiences of nausea, such as morning sickness or medication-induced nausea, did contribute to the patients' postchemotherapy nausea. That is, the relation between expectancy and postchemotherapy nausea could arise from the patients' other prior experiences of nausea, either because of a higher susceptibility to nausea in general or as a result of conditioning. If this were the case, expectancy may simply have acted as a marker for history of nausea rather than having a direct causal impact on postchemotherapy nausea. Montgomery and Bovbjerg,34 however, failed to find a correlation between chemotherapy patients' lifetime history of nausea, including susceptibility to motion sickness and nausea during pregnancy, and expectancies for experiencing postchemotherapy nausea. This suggests that patients' expectancies for postchemotherapy nausea may be independent of their previous experiences of nausea in other settings. However, because their analysis included a very small sample (only 31 patients), this finding must be considered preliminary at best and warrants further investigation.

In conclusion, the current study findings provide further evidence that expectancy contributes to the development of postchemotherapy nausea, which, in addition to being inherently unpleasant, also detracts from patients' QoL. Given that highly expectant patients appear at particular risk of experiencing postchemotherapy nausea, an effective preventive strategy might be to target these patients. Including a discussion of expectancies for nausea during chemotherapy education reflects a possibility that would have the added benefit of allowing direct exploration of the sources of these expectancies.

Acknowledgements

We thank Phyllis Butow, PhD, MClinPsych, MPH (Medical Psychology Research Unit, School of Pscyhology, University of Sydney); Robert Boakes, PhD (School of Psychology, University of Sydney); and Luana Colloca, MD, PhD (University of Turin Medical School, University of Turin) for helpful comments on drafts.

Ancillary