The effectiveness and cost-effectiveness of acupressure for the control and management of chemotherapy-related acute and delayed nausea: Assessment of Nausea in Chemotherapy Research (ANCHoR), a randomised controlled trial

Article history paragraph text

The research reported in this issue of the journal was funded by the HTA programme as project number 07/31/02. The contractual start date was in November 20008. The draft report began editorial review in October 2011 and was accepted for publication in May 2012. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The HTA editors and publisher have tried to ensure the accuracy of the authors' report and would like to thank the reviewers for their constructive comments on the draft document. However, they do not accept liability for damages or losses arising from material published in this report.

Copyright statement

© Queen's Printer and Controller of HMSO 2013. This work was produced by Molassiotis et al. under the terms of a commissioning contract issued by the Secretary of State for Health. This issue may be freely reproduced for the purposes of private research and study and extracts (or indeed, the full report) may be included in professional journals provided that suitable acknowledgement is made and the reproduction is not associated with any form of advertising. Applications for commercial reproduction should be addressed to: NIHR Journals Library, National Institute for Health Research, Evaluation, Trials and Studies Coordinating Centre, Alpha House, University of Southampton Science Park, Southampton SO16 7NS, UK.

Existing research

Significant developments in antiemetic therapy over the past two decades have improved the control of chemotherapy-related vomiting. By contrast, chemotherapy-related nausea, both acute and delayed, is still a significant problem in clinical practice, with 42–52% of patients experiencing nausea on any one day in routine practice. 1 Surprisingly, despite improvements in the management of vomiting, postchemotherapy nausea seems to have increased. 2 Furthermore, clinicians often underestimate the experience of nausea, especially with regards to delayed nausea. 3,4

Chemotherapy-induced nausea and vomiting can have a profound effect on the cancer treatment experience5 and is associated with negative effects on daily life and overall quality of life, including in relation to food intake, weight loss, social interactions, dehydration, difficulty with sleeping and anxiety. 5,6 In a qualitative study of patients' experiences, unmanaged nausea was constant in some patients and made them exhausted for long periods after chemotherapy, making recovery between cycles longer. 5 The impact of nausea is greater than that of vomiting,7 and nausea has proven to be more difficult to control. The direct and indirect costs of the experience of nausea and vomiting, especially of delayed symptoms, are considerable. 8 Antiemetic trials have traditionally focused primarily on vomiting and emetic episodes, on which the effectiveness of many antiemetic drugs is judged. Little attention has been directed to the concept of chemotherapy-induced nausea despite the fact that it is increasingly recognised that nausea and vomiting are related but separate entities. 9,10 The need for these two symptoms to be treated as two separate entities is strongly advocated. 10

The reasons behind this incomplete management of chemotherapy-induced nausea and vomiting are multifaceted. They include health professionals' limited understanding of the complex concept of chemotherapy-induced nausea and vomiting and its different phases; limited assessment in clinical practice of chemotherapy-induced nausea and vomiting and its risk factors; using more emetogenic chemotherapy protocols than in the past; not understanding clearly all the pathways involved in the development of chemotherapy-induced nausea and vomiting; and more focus given to the vomiting experience than nausea in clinical trials. 11

As antiemetic medications do not fully control nausea during chemotherapy, non-pharmacological interventions in addition to antiemetics have been tested over the years, especially in the 1980s, including relaxation techniques, coping preparation, imagery and distraction techniques, with positive results in most studies. 12 Acupuncture and its non-invasive form of acupressure have been tested several times after the classic early work of Dundee et al. 13,14 In a literature search of MEDLINE, PubMed and Cumulative Index to Nursing and Allied Health Literature (CINAHL) using the key words ‘acupressure’, ‘nausea’, ‘vomiting’, ‘emesis’, ‘chemotherapy’, ‘cancer’ and combinations, between 1990 and May 2005, we have identified 10 studies specific to oncology, reported elsewhere,15 with 7 out of 10 studies showing positive results and a further two approaching statistical significance. These studies have used a variety of acupressure methods, such as the ReliefBand® [a small battery-operated transcutaneous electrical nerve stimulation (TENS) device designed to stimulate the P6 acupoint];16–18 an acupressure wristband (a small elastic band with a round plastic button applying constant mild pressure on the P6 acupoint);19–21 and direct pressure on the P6 acupoint22 or P6 and ST36 acupoints together. 23 Most studies had small sample sizes of 18–50 patients. The largest study to date (n = 739) testing acupressure and acustimulation showed improvements in nausea and vomiting in men and a similar trend in women to reduce acute symptoms only, although the latter did not reach statistical significance. 24 No improvement in nausea/vomiting was shown in a small study by Roscoe et al. 25 in women with breast cancer using acustimulation (ReliefBand) wristbands. These two studies are suggestive of a possible gender effect. However, most past studies are hampered by small sample sizes, the wide variety of (non-standardised) antiemetics used, differences in the risk factors for nausea and vomiting in these samples, the range of emetogenicity of chemotherapy regimens used and sampling issues. A recent Cochrane systematic review26 of the literature highlights that acupressure reduces acute nausea but not delayed nausea, and has no benefit for vomiting. However, the review was primarily focused on acupuncture rather than acupressure, all different methods of acupressure were examined together and the results regarding specifically vomiting are questionable (as many of the studies included in the review had samples with little, if any, vomiting across experimental and control groups).

Our own work

Over the past 8 years the lead applicant has developed a programme of research in the management of chemotherapy-induced nausea and vomiting that feeds into the current application. This has involved the assessment of the effectiveness of non-pharmacological interventions for the management of chemotherapy-induced nausea and vomiting including progressive muscle relaxation training and imagery techniques;27 pilot testing of acupressure;15 identification of risk factors for chemotherapy-induced nausea and vomiting development such as age, gender and anxiety;28,29 the management of anticipatory nausea and vomiting;30,31 the development of international clinical guidelines for managing chemotherapy-induced nausea and vomiting32,33 and radiation-induced nausea and vomiting;34,35 exploration and further clarification of the concept of chemotherapy-induced nausea as a separate entity from vomiting;36 assessment of chemotherapy-induced nausea and vomiting levels in current clinical practice in the UK;37,38 and development of a chemotherapy-induced nausea and vomiting-relevant clinical scale for the assessment of acute and delayed symptoms. 39 This last is the only chemotherapy-specific scale available to date. In our qualitative study of the experience of chemotherapy-related nausea in 17 patients with cancer in the UK and USA,36 nausea was described as a distressing and complex symptom. Preliminary evidence indicates that nausea is part of a cluster of symptoms. Self-management techniques, such as dietary strategies and distraction techniques, were rooted in participants' understanding of nausea and their beliefs about what caused nausea. Although self-management was common in almost all patients, acupressure was not one of the approaches used. In an observational prospective evaluation using patient self-reports, 102 patients with cancer receiving their first chemotherapy treatment participated. 37 Participants were followed up for four cycles of chemotherapy, providing a total of 272 assessments of nausea and vomiting. The results indicated that acute vomiting was experienced by 15.7% of the patients in cycle 1 and delayed vomiting by 14.7%, whereas acute nausea was present in 37.3% of the patients and delayed nausea in 47.1%, which increased over the four cycles. Moderately emetogenic chemotherapy had the highest incidence of chemotherapy-induced nausea and vomiting, and acute symptoms were more controlled than delayed symptoms. The data suggested that, although vomiting is relatively well controlled, nausea is a significant problem in practice; it also highlighted the high cost of inappropriate use of antiemetics, which was £17,524 for every 100 patients treated over four cycles. 37

Primary objective

1. 1. To assess the clinical effectiveness of self-acupressure using wristbands in addition to standard care compared with standard care with sham acupressure wristbands and standard care alone in the management of chemotherapy-induced (acute and delayed) nausea.

Secondary objectives

1. 2. To assess the cost-effectiveness and extent of use of usual care in patients using acupressure wristbands in addition to standard care compared with that in patients receiving standard care with sham acupressure wristbands and standard care alone for the management of chemotherapy-induced nausea.

2. 3. To assess the quality of life in patients using acupressure wristbands in addition to standard care compared with that in patients receiving standard care with sham acupressure wristbands and standard care alone in the management of chemotherapy-induced nausea and vomiting.

3. 4. To assess the clinical effectiveness of self-acupressure using wristbands in addition to standard care compared with that in patients receiving standard care with sham acupressure wristbands and standard care alone in the management of chemotherapy-induced (acute and delayed) vomiting.

4. 5. To ascertain for which emetogenic level of chemotherapy regimen (i.e. high, moderate or low) self-acupressure using wristbands in addition to standard care is more or less effective in terms of nausea compared with patients receiving standard care with sham acupressure wristbands and standard care alone.

5. 6. To ascertain whether or not any improvement in chemotherapy-induced nausea and vomiting from using acupressure wristbands is different between men and women.

6. 7. To ascertain whether or not there is an age effect from the use of acupressure wristbands in relation to chemotherapy-induced nausea and vomiting.

Design of the study

The study was a randomised controlled trial with three arms (Assessment of Nausea in Chemotherapy Research or ANCHoR). The three arms consisted of usual care plus (1) self-administered acupressure wristbands, (2) sham acupressure wristbands and (3) no additional treatment. The duration of the patients' involvement was for four cycles of chemotherapy, as after four cycles patients not responding to the given chemotherapy may discontinue it, may be offered a different chemotherapy regimen or a different treatment plan, or may be offered supportive care only.

Subjects were allocated to the trial groups through computer-generated randomisation carried out remotely by the trials unit of the Christie NHS Foundation Trust, Manchester. The randomisation method used consisted of minimisation with a random element (stochastic minimisation), balancing for gender,29,40 age (16–24, > 24–50, > 50 years)29,41 and three levels of emetogenic chemotherapy [low, moderate and high according to international American Society of Clinical Oncology (ASCO) and Multinational Association of Supportive Care in Cancer (MASCC) classifications]. 32,42

Biases were minimised through (1) exclusion criteria that leave out some of the factors and sources of nausea and vomiting in cancer patients other than chemotherapy (i.e. intestinal obstruction); (2) the use of covariates for variables that are closely linked with nausea and cannot be excluded as they are present in a large proportion of the population (i.e. anxiety),29,43 to be incorporated during the data analysis as a covariate in analysis of covariance (ANCOVA) models; and (3) the use of stratification for other key risk factors for nausea development during chemotherapy (i.e. age, gender) at the randomisation stage. Stratification, prior to randomisation, is important to ensure that known prognostic factors are equally distributed before measuring the treatment-related variables.

Pilot study using this design

We had carried out a two-arm pilot study of 36 breast cancer patients comparing acupressure wristbands (plus antiemetics) with standard antiemetics only. 15 The present trial has been based on methods tested in this pilot study. Although it is acknowledged that this study was limited, key findings suggested that acupressure improved the nausea experience as well as nausea and vomiting occurrence and distress across the first 5 days of chemotherapy. Improvements were higher in relation to nausea than in relation to vomiting. Mean overall percentage of improvement (pre to post assessment) in the experimental subjects was 44.5%. The study showed that an acupressure trial is feasible, with high levels of compliance (only one patient stopped using the wristband, because of arm swelling), although one-third of the patients did not return completed assessments. The lack of follow-up techniques (i.e. reminder letters), which was due to time constraints, is partly responsible for this and is acknowledged as a limitation of the pilot study. However, missing data in the returned assessments were almost non-existent, and patient logs for acupressure usage were fully completed.

Sham acupressure and acupressure have also been used in another pilot trial that we have carried out recently for the management of cancer-related fatigue;44 patients in the sham group who were informed that they were receiving one of two combinations of (acu)points were blinded until the end of the trial and this group had little improvement compared with the real acupressure group, suggesting that this technique was a credible placebo and thus capable of minimising the likely effect of placebo on the study's findings.

Experimental and control interventions

The study was a Phase III pragmatic randomised trial.

The target population was a heterogeneous group of cancer patients meeting the inclusion criteria and about to receive chemotherapy of high, moderate and low emetogenic potential. Heterogeneity is important to address issues of response to different types of emetogenic chemotherapy, as well as by gender and age, as past literature highlights that these are important in assessing the effectiveness of treatments for chemotherapy-related nausea and vomiting. Minimally emetogenic chemotherapy was not included, as clinical guidelines recommend no antiemetic treatment and the nausea/vomiting level is < 10%.

In the acupressure group, in addition to standard antiemetics, patients were provided with a pair of widely available acupressure wristbands. These bands are elastic wristbands with a 1-cm protruding round plastic button (stud). They are available in two sizes, standard and large. Patients wore the wristband with the stud pressing the P6 acupoint, which is located on the anterior surface of the forearm, approximately a three-finger width up from the crease of the wrist between the tendons of the palmaris longus and flexor carpi radialis. An instruction sheet with a picture of acupoint P6 and how to locate the point was also provided to patients. Patients were provided with a pair of acupressure wristbands and instructed to wear them on both arms and take them off only when showering/bathing. Patients were instructed to wear the wristbands from the morning before chemotherapy administration and for the subsequent 6 days (total 7 days). No other complementary therapy use was recommended during the course of acupressure (although any such use was documented).

In the sham acupressure group, in addition to standard antiemetics, patients were provided with a pair of identical-appearing wristbands, with the only difference being that the sham wristband had the button on the exterior of the wristband and patients were instructed to wear the wristband with the button away from what is the P6 point. There has been an ongoing scientific debate on what constitutes an appropriate sham treatment, and it has been acknowledged that there is no sham method in acupuncture and acupressure studies that can be widely accepted as the optimal method. It is increasingly believed that sham acupuncture/acupressure designs cannot detect the whole placebo effect and may generate false-negative results,45–48 depending on the method used. We had debated the appropriateness of other sham methods but either they were not blinded enough for the purposes of the trial (i.e. they were slightly dissimilar to real acupressure wristbands) or they could be perceived as treatments themselves (i.e. acupressure at other points in the forearm or elsewhere where we had no information as to an effect on the experience of nausea). Patients in the clinics could also talk to each other and realise that they have different interventions or check the P6 point on the internet. Hence, we resolved to use an acupressure technique that appeared to be exactly the same as the active treatment with the only exception being the place of the stud on the wristband (interior or exterior to the band) used. This was also agreed by practitioners who had been consulted about their views on the most appropriate sham method. Furthermore, although it was acknowledged that many patients may have heard of the use of such wristbands, the results of our pilot study suggested that their knowledge of acupressure wristbands would be limited. 15 In addition, the results of our qualitative study on self-management of chemotherapy-induced nausea and vomiting suggested that acupressure was not commonly used by patients. 36 An assessment of blinding at the end of the trial was not conducted as patients had not been informed of the use of both sham and real acupressure bands during the trial, but had instead been informed that two different types of wristbands were being evaluated, with the approval of the ethics committee. Clinicians did not know the patients' group allocation.

The control group received standard antiemetics alone. Standard antiemetics for all three groups were based on ASCO and MASCC international antiemetic guidelines with the exception of neurokinin 1 (NK1) receptor antagonists [e.g. aprepitant (Emend®, Merck)] recommended in highly emetic chemotherapy, which were not available in the NHS. Hence, for highly emetic chemotherapy, patients received a 5-hydroxytryptamine type 3 (5-HT3) receptor antagonist [i.e. ondansetron 8 mg (Zofran®, GlaxoSmithKline)] and dexamethasone 8 mg intravenously before chemotherapy and the same orally for 3 days post chemotherapy; for moderately emetogenic chemotherapy a 5-HT3 receptor antagonist (ondansetron 8 mg) and dexamethasone 8 mg intravenously before chemotherapy and a 5-HT3 receptor antagonist or dexamethasone for 2 days post chemotherapy; and for low emetogenic chemotherapy dexamethasone 8 mg before chemotherapy and no other treatment post chemotherapy]. 32,42 All patients received rescue antiemetics if nausea and/or vomiting was persistent and they failed to respond to the antiemetic treatment (i.e. severe nausea or more than five vomiting episodes), based on the experience of each clinician (as agreed guidelines for rescue antiemetics had not been developed to date).

Inclusion/exclusion criteria

Proposed sample size

In our pilot study15 the mean score for nausea experience averaged over 5 days was 2.79 [weighted average standard deviation (SD) 3.15] in the control group and 1.45 (weighted average SD 2.76) in the intervention group. At least 135 participants per arm would be required to detect this pair-wise difference between arms using a t-test with a conservative Bonferroni-adjusted significance level of 0.05/3 = 0.017 at a power of 90%. The pilot study suggested an attrition rate of 33% and so, initially, at least 202 participants would be required per arm. As the SDs are much larger than the means in the pilot data, they are suggestive of highly skewed distributions; hence, the equivalent non-parametric test (the Mann–Whitney test) will be used. As the asymptotic relative efficiency of the Mann–Whitney test is at worst 0.864, the sample size for a Mann–Whitney test is equal, in the worst case, to the sample size for the t-test divided by 0.864. This would increase the required sample size to 156 per arm before attrition and 233 after attrition, equalling 699 across the three arms.

Statistical analysis

Descriptive statistics have been estimated for all baseline sociodemographic and clinical variables by arm and for outcome variables (scores on nausea and vomiting subscales) by arm. The association between baseline sociodemographic or clinical variables and outcome variables has been assessed using between-group tests or correlations depending on skewness. Primary outcome variables have been compared between the arms using t-tests, one-way analysis of variance (ANOVA), Mann–Whitney tests and Kruskal–Wallis tests, bearing in mind any skewness in the data. Ordinal regression models were employed to permit covariate-adjusted analyses of a grouped version of the primary outcome. An extension of the proportional odds regression model was used for longitudinal analyses over cycles and this was fitted with a generalised estimating equation (GEE) approach. An intention-to-treat analysis model has been followed. As the primary outcome variable was assessed over several days repeatedly, an aggregate score of all assessments in each cycle was calculated before any modelling analysis.

The effect of missing values was assessed by comparing the numbers and percentages of participants with missing values in the three arms of the study; differences in baseline variables between participants with observed and missing outcomes in each arm; and, for participants with observed outcomes, differences in baseline variables between the three arms. There were no clear associations between known predictors of nausea and cases missing the nausea primary outcome. This fact along with the highly non-normal distribution of the primary response (for which imputation methods are not so well developed) informed our decision not to apply multiple imputation analyses.

Randomisation method

The trial arm allocation method was minimisation with a random element over the margins of three factors: gender, age (16–24, > 24–50, > 50 years) and emetogenic risk (low, moderate, high). The first 20 cases were allocated completely at random and thereafter the allocation probability vector was (0.6, 0.3, 0.1) to the arms that would result in the least to the most imbalance respectively. Researchers telephoned the randomisation office staff with the patient details and the staff used an in-house program to obtain the allocated trial arm.

Outcome measures

Wristband compliance and audit

Patients who had been randomised to the acupressure and sham acupressure groups were also asked to provide information about the length of time that they had worn their wristbands on the day of chemotherapy and the 6 subsequent days. A wristband compliance questionnaire (see Appendix 7) was given to the patients per cycle to complete and return in the prepaid envelope together with the other completed scales.

In addition to compliance with wearing wristbands, it was important to determine whether or not patients were wearing the wristbands correctly according to the instructions they had been given. A wristband audit was conducted at two Manchester sites (Christie NHS Foundation Trust and Macclesfield District Hospital) and two Liverpool sites (Clatterbridge Centre for Oncology and the Royal Liverpool Hospital). Research nurses who were involved with patient recruitment to the trial assessed how correctly patients were wearing their wristbands when they attended for chemotherapy treatment. In total, observations were carried out for 35 pairs of wristbands. An audit proforma is shown in Appendix 8.

The trial did not come under Medicines and Healthcare products Regulatory Agency (MHRA) regulations as it was not a clinical trial of an investigational medicinal product as defined by EU Directive 2001/20/EC; a procedure for the reporting of serious adverse events was therefore not required. However, patients were regularly asked about their experiences with regard to wearing wristbands (e.g. when attending for chemotherapy and during routine telephone communication during the trial) and about any problems associated with the wristbands (e.g. if a larger size was required, if the wristband caused any kind of discomfort, if the wristband became damaged). These were logged by the researchers, with appropriate advice offered to patients.

Measurement of costs

Costs were identified, measured and valued using a microcosting approach (in which each component of resource use was identified, estimated and a unit cost derived from market prices and national estimates53). The cost analysis was performed from the perspective of the health service provider and from a societal perspective. Included in the health-care provider costs were those accrued by the acute trusts and the primary care trusts. Costs to the patients and their families, including social care, were considered as the additional costs for society. Indirect costs in terms of workdays lost were also included.

Data were collected prospectively and retrospectively using multiple sources including patient records and patient self-reported questionnaires (see Appendices 9 and 10). The questionnaires reported health service utilisation subsequent to and as a result of chemotherapy-induced nausea/vomiting (e.g. GP visits), patient out-of-pocket expenses such as over-the-counter medicines or transport and the use of services in the social sector such as home help and support from family and friends. Valuation of resource items including hospital resources (e.g. bed-days and staff time) and community resources (e.g. GP visits, home help) was carried out using national estimates;53 market prices were assigned to medication; non-market items, specifically patient time and informal help provided by family and friends, were valued using market wage rates; and out-of-pocket expenses (e.g. bus fares) were also calculated.

In more detail, direct medical costs were defined as the costs of prophylactic or rescue antiemetic medications, drug administration devices, staff time associated with preparing and administering medication and tending to patients with chemotherapy-induced nausea and vomiting, hospitalisations due to chemotherapy-induced nausea and vomiting, hospital outpatient or GP visits due to chemotherapy-induced nausea and vomiting and over-the-counter medications or other complementary therapies. Direct non-medical costs were those for transportation and need for assistance, such as additional childcare. Indirect costs were based exclusively on the number of workdays lost due to chemotherapy-induced nausea and vomiting. Costs that were not included in this evaluation were costs for chemotherapy agents, preplanned visits or hospitalisations for the purpose of chemotherapy administration, and diagnostic and laboratory tests, and other patient management costs not directly related to chemotherapy-induced nausea and vomiting.

Nested qualitative study

There was an exploratory nested qualitative study within the trial that explored patients' reasons for consenting to take part in the trial and their experiences of participating in a randomised controlled trial for acupressure wristbands.

A number of qualitative studies have explored patients' experiences of receiving treatment with acupuncture, and their findings have suggested that the treatment is associated with eliciting benefits beyond the alleviation of the patients' presenting condition. 54–56 These expanded effects of care include improvements in physical/mental health and emotional well-being and changes in personal identity and lifestyle, and can result in patients ‘feeling normal again’ and ‘regaining their lives’. Despite the burgeoning qualitative research exploring the experiences of patients receiving acupuncture, to date no study has been conducted to explore the experiences of users of acupressure or acupressure wristbands. To address this gap in the evidence base, a nested qualitative study was conducted with patients taking part in the main trial.

Research governance

The sponsor of the study was the University of Manchester. The MHRA has confirmed that this trial does not fall under the Medicines for Human Use (Clinical Trials) Regulations 2004, as described earlier.

Service user involvement

Service users were involved at three levels. The first has been at the development phase of this proposal, with the contribution of the chairperson of the National Cancer Research Institute (NCRI) Consumer Liaison Group, who was a named co-applicant in the study, and reviews by expert patients. The second level was monitoring the trial project and guiding it within its scientific framework through chairing and participating in the Trial Steering Committee and the DMEC. Finally, users have advised us in planning appropriate patient-focused dissemination of the trial results at the end of the study. For reviews, contacts, active involvement and access, the research partners' strategy and mechanisms through the NCRI Cancer Experiences Collaborative have been utilised.

Patient recruitment

Three main study centres were involved in patient recruitment, namely Manchester, Liverpool and Plymouth. Recruitment began with a month-long pilot phase in Manchester at the Christie NHS Foundation Trust to test recruitment processes. The first case was randomised on 23 March 2009. A phased recruitment launch followed in the two remaining centres with patient recruitment starting in Liverpool in May 2009 and in Plymouth in June 2009. The last case was randomised on 15 October 2010. In total, 14 hospital sites were involved in recruitment (Table 2) and 500 cases were randomised (166 standard care, 166 sham acupressure and 168 acupressure).

Monthly recruitment for the trial is given in Figure 1.

FIGURE 1.

Trial recruitment by month. Monthly target accrual was recalculated after sample size adjustment in May 2010, with the agreement of the HTA team. Power was initially calculated to 90% but as recruitment was slower than expected the sample size was recalculated to 80% power.

A CONSORT flow diagram (Figure 2) shows the numbers of participants recruited and randomly assigned to the three trial arms and who received the intended interventions and were analysed for the primary outcome.

FIGURE 2.

Patient recruitment into the ANCHoR trial: Manchester, Liverpool and Plymouth sites, March 2009 to October 2010. a, Partial data indicate data collected from less than the complete dataset (baseline plus four cycles of chemotherapy), that is, when at least one assessment was missing. Data represented in this figure include the pilot phase, March–April 2009.

Descriptive statistics by trial arm

The majority of the participants were female, married and aged > 50 years. The key diagnoses of the sample included breast and colorectal cancer and the majority had received moderately emetogenic chemotherapy (including anthracycline-based chemotherapy). Other sociodemographic and clinical data are provided in Table 3.

Assessment of missing data for the primary outcome

A total of 500 cases were randomised, but there are data for only 361 of these cases for the primary outcome, i.e. the primary outcome is missing in around 28% of cases. Table 4 illustrates, for the remaining 497 cases, the proportion of cases in which the primary outcome is missing by various factors thought to influence nausea propensity (three cases were allocated to ‘No acupressure’ but there are no records in the trial database for these three cases, i.e. no completed screening forms and no returned data form). There are no marked associations between any of these factors and the probability of the primary outcome being missing.

Inevitably there is a further dropout of cases with each cycle, as illustrated in Table 5.

Table 6 shows the mean nausea experience of the patients using the Rhodes Index of Nausea, Vomiting and Retching scale. Scores range from 0 to 12, with higher scores indicating higher levels of nausea. Both the sham and acupressure arms experienced less nausea than the standard care arm, although the difference did not reach statistical significance. It is important to note that these mean values represent very low levels of nausea. Only at cycle 4 did the results become significant (p < 0.05), with the acupressure group reporting no nausea experience.

Tables 7 and 8 show the frequencies of the nausea and vomiting experience, respectively, categorised over five ranges of experience.

Tables 9 and 10 present descriptive data of acute and delayed nausea, respectively, based on the MAT scale, showing similar results to those obtained with the Rhodes Index of Nausea, Vomiting and Retching scale. Tables 11 and 12 report the data for acute and delayed vomiting, respectively, based on the MAT scale.

Primary outcome analysis

The primary outcome is the mean nausea experience (days 0–6) for cycle 1 measured using the Rhodes Index of Nausea, Vomiting and Retching scale. As can be seen from Figure 3 the data are highly skewed. The possible range of scores is from 0 to 12, but, in fact, 111/361 (31%) are exactly zero and around half of all values are < 1. No transformation would be successful in normalising such a distribution.

FIGURE 3.

Primary outcome distribution (n = 361). The intervals are [0,1], (1,2], (2,3], … [11,12]. The first and last intervals are closed at both ends and intervals in between are open at their lower bound and closed at their upper bound.

The distribution by trial arm is shown in Figure 4. Because of the highly skewed distribution, the non-parametric Kruskal–Wallis test has been used for the primary comparison of the trial arms. This overall test is non-significant (p = 0.14).

FIGURE 4.

Box and whisker plot of the primary outcome by trial arm.

Provision for pair-wise comparisons (Mann–Whitney U-tests) with a Bonferroni adjustment was made in the trial design: none compared with acupressure (p = 0.23), none compared with sham acupressure (p = 0.05), sham acupressure compared with acupressure (p = 0.40). It should be noted that the reference value for statistical significance is 0.017 and so none of these pair-wise comparisons is statistically significant.

Regression analyses for the nausea primary outcome data

As noted previously, the data are highly skewed and no transformation will be successful in normalising the distribution. The approach adopted here is to group the values into five ordered categories and to utilise regression methods for ordinal data. The categories were chosen after inspection of Figure 3. The first category, ‘Zero’, was chosen as it represented no nausea at all and a large fraction of cases fell into this category (31%). The choice of the other categories was somewhat arbitrary. Five categories is fairly typical for ordinal regression models in the literature and it is desirable that no category has a very small frequency. Category 5 has a broad range but includes only 8% of cases; subdividing it further would be counterproductive. Having chosen the lowest and highest categories the remaining three were simply selected to have equal width. Table 13 shows the categories and the frequencies by trial arm (n = 361).

The main tool used was the proportional odds regression model:60

where k = 1, 2, 3, 4 denotes category; i = 1, 2, …, n denotes case; Yi is the response category for the ith case; xi is a vector of covariate values for the ith case (including the trial arm); αk is an intercept term for the kth category; β is a vector of parameters to be estimated; and Logit(p) = log[p/(1 – p)].

Simple illustration of primary outcome by trial arm and gender

The regression analyses above suggest that there may be a trial arm by gender interaction. To illustrate this, we may return to the box and whisker plots by trial arm further broken down by gender as shown in Figures 5 and 6. Given that we have seen a statistically significant interaction in the ordinal regression analyses, there is some justification to repeat the Kruskal–Wallis tests separately within the two gender strata. This yields significant results for men (p = 0.04; n = 79) and for women (p = 0.01; n = 282). Descriptively, the difference appears to be with the sham arm having a beneficial effect for women but a detrimental effect for men. This is a post hoc analysis and should be interpreted with due caution especially as there are a limited number of men.

FIGURE 5.

Box and whisker plot of the primary outcome by trial arm for men.

FIGURE 6.

Box and whisker plot of the primary outcome by trial arm for women.

Longitudinal regression analyses of mean Rhodes Index of Nausea, Vomiting and Retching nausea experience scores

The mean Rhodes Index of Nausea, Vomiting and Retching nausea experience scores (days 0–6) have also been calculated for cycles 2, 3 and 4. Once again, the scores for each cycle have been grouped in the same manner as previously to a 5-point ordinal scale. These repeated ordinal data have been analysed using an extension of the proportional odds regression model described previously, this time fitted using a GEE approach. 61 A working autoregressive correlation structure has been assumed, but the regression parameters are robust to this assumption. All reported Wald tests have used the robust covariance estimates.

Longitudinal analyses of secondary outcomes

Results for the Hospital Anxiety and Depression Scale and the Functional Assessment of Cancer Therapy – General scale

These two scales were assessed at baseline and at cycles 1–4. Longitudinal linear models were fitted (GEE approach using unstructured covariance matrices) to the cycles 1–4 data using the relevant baseline variable as a covariate along with factors representing cycle and arm. There was no evidence of any trial arm effects on mean values, as can be seen from Table 33.

Scale reliability

A descriptive measure of scale reliability (Cronbach's alpha) has been calculated at baseline. FACT scales may be calculated with four or more completed items; however, for Cronbach's alpha analysis all items must be completed. As can be seen in Table 34, all scales used in the study had very good reliability values.

Wristband audit

In total, 35 ‘wrist pairs’ were observed during the months August–November 2010 in the outpatient departments at four trial sites (Royal Oldham Hospital, Macclesfield District General Hospital, Liverpool Royal Hospital and Clatterbridge Centre for Oncology).

The vast majority of observations indicated that both wristbands (i.e. on both left and right wrists) were being worn correctly. In two instances a patient was observed to be wearing only one of a pair of wristbands on arrival at the outpatient department, with one patient having a swollen left arm and the other patient having removed one wristband in advance of chemotherapy administration with the intention of wearing the wristband after the chemotherapy had been given. Eight patients were not wearing their wristbands, with one patient having swollen hands and the remaining seven patients stating that they intended to wear their wristbands after chemotherapy had been administered. Three patients who were using spare wristbands offered to them by the research nurse correctly demonstrated their use.

Only four of the 68 wristbands observed (i.e. for left and right wrists) were positioned incorrectly. Three of these were worn incorrectly on the right wrist and one was worn incorrectly on the left wrist. Figure 7 shows the distribution of wristband observations.

FIGURE 7.

Wristband observations during the audit (n = 70).

The condition of the wristbands was also noted by the research nurses for a total of 25 pairs. Six wristbands were in poor condition: two pairs of wristbands (one pair being worn for the fourth cycle of chemotherapy and one pair being worn for the second cycle of chemotherapy) were in generally poor condition and one pair of wristbands being worn for the third cycle of chemotherapy had loose acupressure buttons on both bands. All of these bands were replaced. In total, 22 pairs of wristbands were recorded as being in good condition.

Context of the data

Outcome measurement and valuation

Participant health-related quality of life was assessed using the European Quality of Life-5 Dimensions (EQ-5D),67 which was included along with the patient resource-use questionnaires. Differences between the randomised groups at follow-up with respect to EQ-5D scores were investigated using non-parametric Mann–Whitney U-tests as the data were not normally distributed. Change scores over time were evaluated using independent t-tests.

Participant responses to the EQ-5D questionnaire were converted to health-state utility values using the UK tariff values68 and an area under the curve approach. These values were then multiplied by duration (21 days) in each health state and divided by 365 to estimate QALYs. QALYs represent a quality-weighted survival value in which 1 QALY is the equivalent of 1 year of full health. Average QALYs between adjacent time points were calculated to generate smoothed estimates between time points. QALYs were calculated as:

where T1, T2m, T3m, T4m and T5m are time points of assessment in the trial, that is, baseline and end of cycles 1, 2, 3 and 4 respectively.

Missing data

Respondents who fail to complete individual items of the EQ-5D are not allocated a utility index score.

The last observation carried forward (LOCF) method was employed to deal with missing time point data. In this trial there were four cycles of chemotherapy after the baseline observation. If a patient dropped out of the study after the third week, this value was then ‘carried forward’ and assumed to be the score for the fourth cycle missing data point. Similarly, if a patient dropped out after baseline this value was carried forward to the first cycle of chemotherapy. Scores were carried forward to one consecutive time point only. Patients in whom more than one consecutive score was missing were omitted from the analyses. Scores were not carried backwards – that is, those without baseline scores did not have their first follow-up score carried backwards. The advantage of the LOCF approach is that it minimises the number of subjects who are eliminated from the analysis and it allows the analysis to examine the trends over time, rather than focusing simply on the end point. Assuming that scores are expected to improve over time, LOCF may be a conservative way of dealing with missing data. A full case analysis was conducted in the sensitivity analyses in which participants were excluded if they had missing scores from any time point.

We used the same method for missing resource-use data, carrying forward the last observation only once if there were missing data for the next time point. Patients were omitted from the analyses if they had more than one missing time point.

Economic evaluation

Descriptive statistics of costs and EQ-5D scores were calculated by subgroup. Parametric tests (Student's t-tests and ANOVA tests) and non-parametric tests (Wilcoxon signed-rank tests, Mann–Whitney tests and Kruskal–Wallis ANOVA tests) were conducted to evaluate differences in individual characteristics and health-related quality-of-life scores between groups. Mann–Whitney U-tests were used to evaluate the significance of differences in costs between groups as this test is less likely to be influenced by outliers. For the economic evaluation only costs of patients who had a QALY value were included.

The outcome of the cost-effectiveness analysis was an incremental cost per QALY. 61,69 We present incremental cost-effectiveness ratios (ICERs)61,69 representing the ratios of the incremental cost and incremental benefits (QALYs) between acupressure and sham acupressure, between acupressure and standard care and between sham acupressure and standard care. The ICER represents the additional cost per one unit of outcome gained, in this case per QALY gained for each intervention compared with its next best alternative. As a guideline rule, the National Institute for Health and Clinical Excellence (NICE) accepts as cost-effective those interventions with an ICER of < £20,000 per QALY. NICE states that, in general, if a treatment costs > £30,000 per QALY it would not be considered cost-effective. We also present cost-effectiveness plane scatter plots illustrating the uncertainty surrounding the cost-effectiveness estimates.

The cost-effectiveness planes were derived using bootstrapping with replacement. This stochastic uncertainty analysis involved running 10,000 bootstrapped estimates of the incremental costs and QALYs. The bootstrap approach is a non-parametric method that treats the original sample as though it were the population and draws multiple random samples from the original sample. Cost-effectiveness acceptability curves were also generated to illustrate the probability that each treatment would be cost-effective given a range of acceptable threshold values. 70

Sensitivity analyses were carried out to account for uncertainty in the cost and EQ-5D values. We performed the sensitivity analysis by adding and subtracting 20% of the costs and assessing the subsequent impact on the ICERs. The value of 20% is essentially arbitrary but it was considered likely to represent any uncertainty that might exist in parameter values. We performed a sensitivity analysis on the QALYs by conducting a full-case analysis on EQ-5D scores. In contrast to the LOCF, no missing data are imputed and patients with any missing data are omitted from the analyses. Sensitivity analyses were also carried out based on using an average group cost for those outliers with high costs and by adopting a societal perspective for costs, including productivity losses and lost earnings.

Net benefit values71 were generated to enable more traditional analysis of therapy efficacy. Net benefit is a composite representation of cost-effectiveness and willingness to pay that is derived by rearranging the ICER calculation and incorporating the willingness-to-pay threshold value (in this case the £20,000 per incremental QALY threshold of NICE). Net monetary benefit (NMB) is derived for each patient as:

The decision rule becomes to adopt an intervention if the NMB > 0. Because values are transformed from ratio to linear we can subsequently employ standard regression models on the data. Net benefit regression allows us to control for covariates and baseline between-group differences. 72 We applied the variables used for the primary end point regression analyses to NMB data with treatment arm included as an independent variable. Univariate ANOVA and linear regression models were generated.

Results

Data from 450 patients (157 acupressure, 146 standard care, 147 sham acupressure) were included in the base-case analysis.

Resource use and costs

There was very little reported resource use within the trial; thus, caution should be exercised in interpreting the results as a small number of high-resource-use consumers or a few instances of high-cost resource use may be unduly influencing the overall results. Table 35 shows the number of times each of the health service resources was used by trial arm. The acupressure group used the fewest resources and the standard care group used the most.

TABLE 35 - Resource use by the trial participants

Services	Acupressure	Standard care	Sham acupressure
GP surgery visit	5	6	5
GP home visit	3	4	3
District nurse	4	11	17
Contact with oncology hotline for advice	6	11	9
Contact with hospital oncology nurse clinician for advice	9	9	6
Contact with hospital oncology clinic for advice	5	8	7
Hospital inpatient stay	10	14	9
Hospital A&E department	3	1	2
Hospital general outpatient clinic	1	0	2
Other services	3	1	1
Total instances of resource use	49	65	61

In terms of the total NHS resource-use cost, the acupressure group used the least number of resources and also had the lowest cost. Mean total drug costs (including antiemetics and other drugs prescribed for chemotherapy-induced nausea and vomiting and antiemetic use outside of the routine pathway in each arm of the trial) for the groups were £37.07 (SD £101.72) for the acupressure group, £51.66 (SD £150.02) for the standard care group and £24.03 (SD £18.70) for the sham acupressure group. Mean total costs (drug costs plus all other NHS costs plus band costs) for the acupressure group were lower than those for the standard care group or the sham acupressure group (Table 36). ‘All other NHS costs’ include the resource use from Table 35. The NHS costs for the sham group appear higher than those for the other groups and this seems to be driven by a higher number of district nurse visits and a few patients in the sham arm of the trial having longer inpatient stays. There were no significant differences in costs between any of the treatment arms according to the t-tests; this was true for the whole sample as well as for the subsample who also had QALY data. Mean values were distorted by a small number of high-cost outlying patients leading to high levels of uncertainty around the mean cost estimates. Median total costs were £30.92, £22.72 and £30.92 for the acupressure, standard care and sham acupressure groups respectively. Mann–Whitney tests indicated that the standard care group had significantly lower total costs than the acupressure and sham acupressure groups (both p < 0.001); there was no significant difference in costs between the acupressure and sham acupressure groups (p = 0.585). No adjustments for multiple testing were considered necessary. Figure 8 shows the cost distribution for the standard care group and illustrates the presence of a small number of high-cost outliers. A similar pattern was observed for the other two treatment arms.

TABLE 36 - Costs per arm of the trial

	Acupressure (£), mean (SD)	Standard care (£), mean (SD)	Sham acupressure (£), mean (SD)
n	72	62	60
Drug cost – cycle 1	5.47 (1.40)	5.60 (0.65)	5.17 (1.55)
Drug cost – cycle 2	11.71 (36.35)	15.36 (50.43)	5.15 (1.55)
Drug cost – cycle 3	9.92 (33.34)	15.39 (49.77)	6.86 (8.86)
Drug cost – cycle 4	9.96 (33.34)	15.31 (49.80)	6.86 (8.86)
Total drug cost	37.07 (101.72)	51.66 (150.02)	24.03 (18.70)
All other NHS costs – baseline	2.92 (17.56)	0.05 (0.0)	17.29 (124.90)
All other NHS costs – cycle 1	8.52 (34.91)	21.73 (76.90)	9.35 (38.74)
All other NHS costs – cycle 2	9.78 (44.44)	9.18 (44.48)	13.27 (86.59)
All other NHS costs – cycle 3	4.17 (25.80)	16.06 (73.89)	52.13 (303.28)
All other NHS costs – cycle 4	0.0 (0.0)	12.50 (80.47)	37.66 (283.54)
All other NHS costs total	25.39 (81.07)	59.52 (171.78)	129.69 (604.28)
Band cost	8.20 (0.00)	0.00 (0.00)	8.20 (0.00)
Total cost	70.66 (129.75)	111.18 (262.68)	161.92 (604.57)

FIGURE 8.

Total cost distribution for the standard care group.

Table 37 includes costs from the societal perspective. Lost earnings and productivity losses to employers through sickness absence appeared higher in the standard care group than in either of the band groups. Overall societal costs are similar for the band groups and significantly higher (over double) for the standard care group. However, as with the health-care costs, this result is partly driven by a small number of high-cost outlying individuals, making robust conclusions difficult.

TABLE 37 - Societal costs

	Acupressure (£), mean (SD)	Standard care (£), mean (SD)	Sham acupressure (£), mean (SD)
n	72	62	60
Total days out of work	1.18 (6.77)	2.82 (9.89)	0.75 (3.45)
Total lost earnings	54.17 (391.44)	241.42 (1068.43)	31.50 (244.00)
Total expenditure	0.18 (1.15)	0.10 (0.53)	0.00 (0.00)
Social cost – baseline	0.00 (0.00)	19.82 (104.46)	0.00 (0.00)
Social cost – cycle 1	48.95 (255.68)	226.73 (922.55)	19.89 (101.05)
Social cost – cycle 2	52.70 (329.40)	97.17 (347.41)	25.21 (112.07)
Social cost – cycle 3	46.95 (374.77)	69.77 (295.67)	27.80 (159.23)
Social cost – cycle 4	0.00 (0.00)	53.37 (286.99)	18.48 (143.18)
Total social cost	148.60 (925.90)	466.87 (1490.27)	91.38 (472.80)
Total societal costs (social cost + health-care costs)	219.26 (983.56)	578.00 (1557.79)	253.30 (760.40)

Economic evaluation results regarding quality-adjusted life-year gains

Table 41 shows the total costs and QALY gains for each of the three treatment arms. Differences in QALY gains between groups were minimal and suggest only negligible health benefits of acupressure over standard care. The sham acupressure group had the highest QALY gains over the trial period with the standard care group having the lowest. The mean total cost was highest for the sham acupressure group and lowest for the acupressure group. The high SDs for the cost estimates indicate the presence of a few outlying individuals who incurred significant health service costs.

TABLE 41 - Total costs and QALY gains by treatment arm

	Acupressure, mean (SD)	Standard care, mean (SD)	Sham acupressure, mean (SD)
n	72	62	60
Total QALY gain	0.270 (0.062)	0.265 (0.072)	0.283 (0.058)
Total cost (£)	70.66 (129.75)	111.13 (262.68)	161.92 (604.57)

Table 42 provides the cost-effectiveness results, showing the incremental costs and benefits as well as the ICERs for each arm of the trial. Interpretation should be tempered by considering the high level of uncertainty surrounding the cost estimates and the negligible between-group differences observed in QALY gains. The standard care group was dominated by acupressure as it had higher costs and lower QALY gains. The results suggest that acupressure might be cost saving compared with standard care alone. The acupressure group had lower costs than the sham acupressure group; however, the sham group had higher, albeit negligible, QALY gains.

TABLE 42 - Cost-effectiveness results

Analysis	Incremental cost (£)	Incremental QALY gain	ICER (£)
Acupressure vs standard care	−40.47	0.005	−7494.44 (acupressure dominates)
Acupressure vs sham acupressure	−91.26	−0.012	7359.68
Sham acupressure vs standard care	50.79	0.018	2853.37

Table 43 shows the sensitivity analyses that were carried out as well as the results from the non-parametric bootstrapping. The analysis adding 20% to the costs supports the base-case analysis as acupressure still dominates the standard care group. Subgroup analyses by emetogenic risk group were not possible as a very small proportion of patients was rated as high or low emetogenic risk and a large majority was rated as moderate.

TABLE 43 - Sensitivity analyses

Analysis	Incremental cost (£)	Incremental QALY gain	ICER (£)
+20% of costs
Acupressure vs standard care	−48.56	0.005	−8993.33 (acupressure dominates)
Acupressure vs sham acupressure	−109.51	−0.012	8831.61
Sham acupressure vs standard care	60.95	0.018	3424.04
−20% of costs
Acupressure vs standard care	−32.38	0.005	−5995.56 (acupressure dominates)
Acupressure vs sham acupressure	−73.01	−0.012	5887.74
Sham acupressure vs standard care	40.63	0.018	2282.70
EQ-5D full-case scenario
Acupressure vs standard care	−111.51	0.009	−12,319.31 (acupressure dominates)
Acupressure vs sham acupressure	−153.92	−0.018	8538.87
Sham acupressure vs standard care	42.41	0.027	1566.25
Excluding cost outliers
Acupressure vs standard care	−7.58	0.005	−1403.70
Acupressure vs sham acupressure	−9.68	−0.012	780.65
Sham acupressure vs standard care	2.10	0.018	117.98
Societal cost perspective
Acupressure vs standard care	−358.74	0.005	−66,433.33 (acupressure dominates)
Acupressure vs sham acupressure	−34.04	−0.012	2745.16
Sham acupressure vs standard care	−324.70	0.018	−18,241.57 (sham acupressure dominates)
Bootstrapping
Acupressure vs standard care	−40.64	0.005	−9025.53 (acupressure dominates)
Acupressure vs sham acupressure	−90.26	−0.013	7163.27

Including only patients who had completed EQ-5D scores at every time point did not significantly change the ICER. The mean cost and QALY gain estimates from the bootstrapping yield similar ICER results to those of the deterministic base-case scenario. Acupressure still dominates standard care only and is cheaper but marginally less effective than sham acupressure. Excluding cost outliers or including societal costs does not alter the outcome of the key comparisons.

Figures 9 and 10 are cost-effectiveness planes for acupressure compared with standard care and acupressure compared with sham acupressure respectively. For acupressure compared with standard care (see Figure 9), the 1000 sample estimates are spread mainly in the south-east and south-west quadrants, suggesting that acupressure is likely to be cost saving. A high proportion of iteration results are in the south-east quadrant suggesting that acupressure is also likely to lead to a higher quality of life; however, QALY gains are minimal. For acupressure compared with sham acupressure (see Figure 10), most of the sample iterations lie in the south-west quadrant, suggesting that acupressure is less effective (leads to reduced quality of life) than sham acupressure but leads to cost savings.

FIGURE 9.

Cost-effectiveness plane for acupressure compared with standard care.

FIGURE 10.

Cost-effectiveness plane for acupressure compared with sham acupressure.

Figures 11 and 12 are the cost-effectiveness acceptability curves showing the probability that each treatment is a cost-effective choice given a range of cost-effectiveness willingness-to-pay thresholds. Figure 11 compares standard care with acupressure. Using the NICE threshold of £20,000 the probability that acupressure is cost-effective is 0.70 compared with standard care. Figure 12 compares all three treatments. At the NICE threshold of £20,000, sham acupressure appears more likely to be the cost-effective option (p = 0.71) than acupressure (p = 0.20). Only at low willingness-to-pay thresholds (< £7000) does acupressure become the most likely option to be cost-effective. However, considering the negligible QALY gains observed in the study, the low levels of resource use, the low median health-care costs and the leverage of a few high-cost outliers, any results must be treated with caution. It is difficult to make robust claims about the comparative cost-effectiveness of any of the therapy arms.

FIGURE 11.

Cost-effectiveness acceptability curves for acupressure compared with standard care.

FIGURE 12.

Cost-effectiveness acceptability curves for all three treatments.

Objectives

• To outline patients' experiences of using acupressure wristbands.

• To outline the reasons why patients consent to take part in a clinical trial of acupressure wristbands.

• To outline patients' experiences of taking part in a randomised controlled trial of acupressure wristbands.

Findings

Key quantitative findings

Despite the higher proportion of patients showing no nausea in the acupressure group, the results of the trial show that there were no statistically significant differences between the three trial arms in relation to nausea experience. Patients in both wristbands arms had a higher OR for improving their nausea experience than those in the standard care arm, with the sham arm having a higher OR than the acupressure arm. There was a significant gender effect with women in both wristband groups showing significant improvements compared with men. Health-care utilisation was lower in the wristband arms than in the standard care only arm.

Other trials in the past have also shown no significant effects from the use of acupressure in relation to nausea and vomiting management during chemotherapy administration. A review by Lee and Frazier74 examined the results of seven trials of acupressure, with four having positive results and three negative, highlighting that the overall effect of acupressure is strongly suggestive but not conclusive. No significant differences were reported in another trial of 160 women with regards to acute nausea and vomiting, although significant differences were reported with regards to delayed nausea and vomiting. 75 In the largest trial of its kind (n = 739), Roscoe et al. 24 showed that patients experienced less nausea in the first day of chemotherapy in the acupressure arms but there were no significant differences in relation to delayed symptoms. Also, the authors identified a strong gender effect, with men in an acustimulation arm improving but not women, which is in contrast to the results of the current trial. Roscoe et al. 17 also showed, in a small sample of 27 patients (25 women and two men), no statistically significant differences in average severity of nausea between acustimulation of the P6 point, sham acustimulation and standard care; however, the data showed a difference close to statistical significance in the severity of delayed nausea between active acustimulation and no acustimulation (p = 0.06). In addition, patients took fewer antinausea pills during the active acustimulation cycle of this experiment than during the no acustimulation phase (p < 0.05). Negative results have been shown in relation to acupressure and nausea/vomiting symptoms in a large trial of 340 women during labour and delivery. 76 In a trial of acupuncture compared with sham acupuncture during radiotherapy,77 the authors showed less nausea and vomiting in both the real and the sham acupuncture arms than in the standard care arm and justified this as due to non-specific effects of general care or high patient expectancy, which may partly explain our results too.

Key issues in most of the past studies showing positive results include the lack of standardised antiemetic use in the trial participants and inclusion of only or mostly female subjects. If our trial had included only the female subsample, the results would have also been positive. Also, it seems that the vast majority of positive studies in the literature include small sample sizes (< 100 participants), whereas the negative studies (or partly negative) have much larger sample sizes; this suggests that effects observed in methodologically weaker studies cannot always be sustained when larger and more robust trials are carried out. Furthermore, other studies in the past have shown that expectancy,24,78 age and anxiety28,29 together with the antiemetic potential of the chemotherapy are important predictors of, and can affect, the outcome of acupressure, but in our trial, although unidimensionally these were also important, in a multivariate model only gender showed a significant effect.

Our findings suggest a placebo or non-specific effect of the intervention arms. Placebo effects are viewed as a form of interpersonal healing, distinct from spontaneous natural healing or technological healing that depends on physiologically active pharmacological products or procedures. 79 Alkaissi et al. 80 have suggested that acupressure does indeed have a placebo effect in relation to nausea after 24 hours, although correct stimulation of the P6 point is needed to observe decreased rescue antiemetic use and decreased vomiting. Research also suggests that there are different placebo responses, each of which may be influenced by different psychological and neurobiological mechanisms depending on the context in which the placebo is given. 81 The literature also shows that placebos have actual biological effects on the brain and body and are more than response biases. The review by Price et al. 81 concludes that placebo effects reflect mind–brain–body relationships and as such we should not ‘resort to eliminative materialism or forms of dualism that completely divide the mind from the body’ (p. 586).

Trials of acupressure pose a specific problem with regards to the blinding and the choice of placebo, particularly when outcome measures are subjective. We have chosen to use the same wristbands in both the real and the sham groups so that they look identical, with the real acupressure group instructed to have the button pressing the P6 point and the sham group instructed to have the button pointing away from the P6 point on the other side of the arm. We observed during the interviews that some patients in the sham acupressure group (two out of nine) used the wristbands as in the acupressure group because they had looked on the internet or saw others wearing them properly. This may have contaminated our results. It was not possible to create a wristband that would look identical to the real one but which would not have a button or which would exert no pressure. These bands were elastic bands and, as reported by Sinha et al. 76 (through observations from their colleagues in the Department of Industrial and Manufacturing Engineering, Penn State University, PA, USA), elastic bands result in some pressure. This suggests that the pressure of the band in the area proximal to the P6 point, irrespective of the presence of a button pressing the P6 point, may have produced some positive results.

Our sample had generally low levels of nausea and/or vomiting. This may be due to the fact that we have standardised antiemetic use in our study and an inclusion criterion was receiving antiemetics as per MASCC antiemetic guidelines. This low level of experienced symptoms may be a reason for not showing significant differences in the current trial, as we have shown in another observational study of nearly 1000 patients that use of antiemetics during chemotherapy according to MASCC guidelines is associated with significantly improved nausea/vomiting symptoms. 52

A limitation of the trial may be the missing data for the primary outcome. However, the proportion of cases missing the primary outcome (28%) is of a similar order to that anticipated at the design stage (33%). Originally 90% power had been planned for a standardised difference in means of around 0.48. For pragmatic reasons the study power was reduced to 80% so that it could complete in a reasonable time frame. The attained power that the final sample size with complete data for the primary outcome (n = 361) delivered was 80% for a standardised difference in means of 0.46. Also, another limitation that needs to be carefully considered in future trials is the choice of sham wristbands, which in our case may not have been the most optimal design.

Health-care utilisation data

Although there was no statistically significant effect of acupressure over sham acupressure and no acupressure, the health economics part of the trial was run concurrently with the effectiveness part of the trial. Through health economic analyses in negative trials it is possible that intervention benefits that are not apparent in traditional efficacy end points may register in QALYs, as this is a different composite outcome. Current thinking in respect of economic analyses in cases in which there is no clear differential effect/effectiveness of the intervention being assessed is that a full cost-effectiveness analysis is undertaken and uncertainty accounted for. Even given that the EQ-5D changes were below those levels that are minimally important and were not significant, and the QALY gains were minimal (for acupressure vs standard care), one cannot evaluate the cost-effectiveness without considering costs. 82,83

Mean costs resulting from NHS resource use were consistently higher for the standard care only group than for the acupressure group; this finding was relatively robust to sensitivity analyses. However, because very little resource use was recorded in the study and the results may have been unduly influenced by outliers with high costs, this finding is relatively uncertain and must be treated with caution. The results from the EQ-5D score analysis revealed no significant differences in utility changes over time between treatment arms and QALY gains throughout the study were negligible.

The acupressure bands group appears to have reduced health-care resource use while realising negligible improvements in quality of life compared with standard care alone. A rapid review of the literature found no studies including costs of chemotherapy-induced nausea and vomiting to the NHS, precluding comparison with previous research. Neither did the review yield any previous studies reporting EQ-5D scores for this patient group. In the present study the EQ-5D analysis showed no differences in utility between groups. There also appeared to be little overall impact of chemotherapy as no utility score changes between any cycle exceeded (or came close to exceeding) the minimal important difference identified for the measure. EQ-5D scores for older patients at the final chemotherapy cycle were similar to the population norm,84 but younger patients appeared to experience a greater utility decrement relative to norms. The mean utility for both the 25–50 and ≥ 50 age groups was 0.77 after cycle 4 compared with population norms of 0.90 for the 25–54 age group and 0.79 for the 55–74 age group. It is possible that the EQ-5D is not a sensitive enough measure to capture quality-of-life benefits that may be experienced as a result of reduced chemotherapy-induced nausea and vomiting, although a cancer-specific quality-of-life measure (FACT-G) also failed to detect any between-group differences. There is also a non-significant baseline difference in utility scores between groups, which may explain differential QALY gains over the trial period.

There were several limitations of this economic evaluation. We made assumptions regarding drug doses and antiemetics drug type because of missing data and also made assumptions regarding length of hospital stay. Assumptions were also necessary regarding the standard care of each of the patients, relying on expert opinion and assuming that care was the same across centres. It is possible that more expensive (branded) antiemetic treatments may have been used but not captured as not all prescribing information was available. Different sites may have different protocols relating to antiemetic prescription as a standard therapy. However, the ICERs and the resulting conclusions regarding the cost-effectiveness of acupressure were robust to sensitivity analyses and stochastic bootstrapping.

When calculating the EQ-5D missing data we employed the LOCF approach; this method assumes that the participant's EQ-5D response would have been stable between each cycle, rather than declining or improving. It also assumes that missing values are ‘missing completely at random’ (i.e. that the probability of dropout is not related to variables such as disease severity, symptoms or drug side effects). As such, the LOCF method may lead to bias in the results. However, assumptions about both costs and utilities were tested in the sensitivity analyses. Finally, as stated above, it is possible that the EQ-5D is an inappropriate measure to capture health benefits incurred as a result of reduced nausea and that a cancer-specific measure may be more suitable. 85 However, cost–utility analyses and the EQ-5D are compliant with the NICE reference case. Future research should consider cancer-specific utility measurement and comparisons with generic utility values. As the occurrence and impact of chemotherapy-induced nausea and vomiting varies over time in relation to chemotherapy receipt, quality-of-life impact depends on the time of measure completion during the cycle. Thus, research exploring patients' health status on a daily basis throughout the course of a chemotherapy cycle may be warranted. Given the influence of high-cost outliers in the analyses, greater exploration of what is driving these costs and indeed whether or not they relate to chemotherapy-induced nausea and vomiting may be worthwhile; this may involve qualitative follow-up of high-resource-use individuals.

Qualitative data

The nested qualitative study aimed to explore patients' reasons for taking part in a clinical trial of acupressure wristbands and their experiences of using acupressure wristbands and taking part in the trial of acupressure wristbands. The key findings from these data suggest that patients in the wristband arms perceived the bands to be effective and felt more in control of their nausea and vomiting experience.

The motivations that patients identified for participating in the trial are congruent with the findings of previous research. Paterson et al. 86 conducted a nested qualitative study exploring the experiences of patients with migraines receiving acupuncture within a randomised controlled trial. Similar to the present study, they found that patients felt that acupuncture was ‘worth a try’, with patients being eager for symptom relief and having a desire to help with research for altruistic motives, although the theme ‘giving something back’ appears to be much more prevalent within the present study. This may be a consequence of the greater morbidity associated with cancer, although further research would need to be conducted to confirm whether or not this is the case.

Patients were largely satisfied with the organisation and running of the trial. The process of recruitment was generally perceived as straightforward and to have been adequately described. Interestingly, the completion of trial paperwork was seen as being of benefit by some patients, a finding again consistent with previous research. 86 Where the current study findings appear to differ from previous research is in the explicit feeling of ‘well-being’ that many patients experienced as a direct result of knowing that their participation in the trial might lead to the improvement of care for future cancer patients.

Previous qualitative research exploring the experiences of patients receiving treatment with acupuncture suggests that patients experience a range of benefits beyond the alleviation of their presenting condition, including improvements in overall well-being, sleep pattern and energy levels. 54–56 A number of patients who received wristbands in the present study indicated that they had experienced only minor levels of nausea and vomiting during their chemotherapy. Some patients – those receiving both true and sham wristbands – attributed the wristbands as having had a positive impact on the level of nausea and vomiting experienced. However, none of the patients who participated in interviews associated the acupressure wristbands with eliciting benefits beyond the relief of chemotherapy-related nausea and vomiting. This may suggest that a greater stimulation of acupuncture points is required to elicit these effects or that the contextual factors within the acupuncture consultation, such as the interaction between the practitioner and the patient or consultation setting, may work in conjunction with the stimulation of acupuncture points to elicit these expanded effects of care.

The wristbands were not associated with any significant adverse effects and patients found them easy to wear and to be associated only with limited social inhibitions. Importantly, patients appear to have worn the wristbands in the correct location and for the period of time instructed. However, the data revealed that two of the nine sham acupressure group patients had worn their wristbands inappropriately (based on their trial arm allocation), either wearing them inside out or applying pressure manually to the acupoint. Also of importance is the apparent delay in some patients completing the trial paperwork, relying on memory to fill in the forms. Both of these may represent confounding variables, which if generalisable to other patients in the trial could have serious implications for the results of the trial.

Conclusions and research recommendations

Despite several acupressure antiemetic trials suggesting a beneficial effect, the trial heterogeneity and inconsistent findings prevented any definitive conclusions being drawn. Our study, using a strong methodological design and standardisation of antiemetics, showed no significant differences between acupressure and sham acupressure wristbands for the management of nausea and vomiting during chemotherapy. Nevertheless, the somewhat improved, albeit not statistically significant, levels of nausea in both wristband arms needs some attention, although, as minimally important differences in relation to chemotherapy-related nausea and vomiting are currently not established, some caution is necessary here. The use of wristbands led to lower health-care utilisation (although this did not reach statistical significance). Bands are well accepted and are low cost and safe additions to antiemetic drugs, but the ethical aspects of suggesting the use of potentially non-effective interventions that lead to lower health-care costs and health-care utilisation need some careful consideration. There is a sufficiently encouraging signal and a suggestion of potential health resource-use benefits to justify exploration of acupressure in further trials using both no intervention and sham acupressure controls. Questions that need to be answered in the future include whether or not other forms of acupressure such as regular finger acupressure or Korean hand acupressure could be more effective than wristband acupressure. A meta-analysis of existing data on acupressure wristbands may be an appropriate way to provide a more concrete answer to the question of whether or not acupressure wristbands are effective in managing nausea and/or vomiting during chemotherapy.

Trial manager

Wanda Russell, Christie NHS Foundation Trust

Liverpool lead researcher

John Hughes, Royal Liverpool Hospital

Plymouth lead researcher

Matthew Breckons, Derriford Hospital

Research nurses (National Cancer Resarch Institute recruiters to the Assessment of Nausea in Chemotherapy Research trial)

Tina Pritchard, Christie NHS Foundation Trust

Joanne Allen, Christie NHS Foundation Trust

Wendy Cook, Royal Oldham Hospital

Ruth Halford, Royal Oldham Hospital

Stephanie Ridgway, Tameside General Hospital

Elly Anscombe, Stepping Hill Hospital

Sam Corcoran, Stepping Hill Hospital

Sarah McKenna, Stepping Hill Hospital

Lisa Hardstaff, Macclesfield District General Hospital

Barbara Townley, Macclesfield District General Hospital

Claire Gaskell, Macclesfield District General Hospital

Sarah Entwistle, Macclesfield District General Hospital

Sam Jole, North Manchester General Hospital

Deborah Fleetwood, Royal Liverpool Hospital

Alison Hassall, Clatterbridge Centre for Oncology

Lynne Dickinson, Clatterbridge Centre for Oncology

Holly Donaghy, St Helens Hospital

Frances Lawton, St Helens Hospital

Julie Griffiths, Southport General Infirmary

Ann Wearing, Southport General Infirmary

Ingrid Koehler, Torbay District General Hospital

Julie Pascoe, Derriford Hospital

Nicola Donlin, Derriford Hospital

Caroline Snelgrove, Derriford Hospital

Laura Evenden, Derriford Hospital

Elizabeth Whitby , Derriford Hospital

Helen Smith, Derriford Hospital

Rachel Keast, Royal Cornwall Hospital (Treliske)

Sophie Mepham, Royal Cornwall Hospital (Treliske)

Darren Beech, Royal Cornwall Hospital (Treliske)

Research and Development and administrative staff

Charlotte Lever, Christie NHS Foundation Trust

Joanne Johnson, Royal Oldham Hospital

Philippa Hill, Macclesfield District General Hospital

Joy Bailey, Macclesfield District General Hospital

Marilyn McCurrie, Macclesfield District General Hospital

Edith Curran, North Manchester General Hospital

Tracie Cocks, Stepping Hill Hospital

Cancer Research network managers and support staff

Sue Dyde, Christie NHS Foundation Trust

Zoe Coombe, Christie NHS Foundation Trust

Richard Jones, Christie NHS Foundation Trust

Trial Steering Committee for the trial

Professor Kinta Beaver (independent chairperson, University of Central Lancashire)

Olive Craven (independent member)

Dai Roberts (independent member)

Anne Urmston (patient representative)

Data Management and Ethics Committee for the trial

Simon Rogers

David Ryder

John Belcher (patient representative)

April Matthews (patient representative)

Baio Tang

Health economics support (lead Dr Claire Hulme)

Chantelle Brown, University of Leeds

David Meads, University of Leeds

Contribution of authors

Professor Alex Molassiotis (Professor of Cancer and Supportive Care, University of Manchester) conceived and designed the initial plan of the study, had overall responsibility for the study, analysed the data together with David Ryder (Senior Statistician, Christie NHS Foundation Trust, who led on the statistical analysis) and drafted the final report. Dr Wanda Russell (Clinical Trials Manager, University of Manchester) was responsible for the day-to-day operationalisation and management of the study, liaison with the study sites, recruitment at the Manchester sites, quality assurance, producing reports and preparing for meetings and liaising with the National Cancer Research Network (NCRN) database. Dr John Hughes (Research Associate, University of Liverpool) and Dr Matt Breckons (Research Associate, University of Plymouth) were responsible for recruitment and trial management at the Liverpool and Plymouth sites respectively. Dr John Hughes also carried out the qualitative interviews (together with Dr Russell and Dr Breckons) and led the analysis of the qualitative element of the trial. The co-applicants Professor Mari Lloyd-Williams (Professor of Palliative Medicine, University of Liverpool), Professor Janet Richardson (Professor of Health Services Research, University of Plymouth), Dr Sarah Brearley (Lecturer, Lancaster University) and Dr Adam Garrow (Research Fellow, University of Salford) were involved in all stages of the work – design, analysis and commenting on and drafting sections of the final report. Professor Lloyd-Williams and Professor Richardson were also leads for the trial in Liverpool and Plymouth respectively. Dr Malcolm Campbell (Senior Lecture in Statistics, University of Manchester) was responsible for the initial statistical design of the trial and oversaw the statistical elements throughout the trial. Dr Claire Hulme (Director of the Academic Unit of Health Economics, University of Leeds) was the lead for the cost-effectiveness part of the trial, was involved with the design and conduct of this part of the trial and led on the analysis of the health economics data.

The views and opinions expressed herein are those of the authors and do not necessarily reflect those of the Department of Health.

Publication

Molassiotis A, Russell W, Hughes J, Breckons M, Lloyd-Williams M, Richardson J, et al. The effectiveness of acupressure for the control and management of chemotherapy-related acute and delayed nausea: a randomized controlled trial [published online ahead of print April 19 2013]. J Pain Symptom Manage 2013; in press. http://dx.doi.org/10.1016/j.jpainsymman.2013.03.007

Disclaimer

This report presents independent research funded by the National Institute for Health Research (NIHR). The views and opinions expressed by authors in this publication are those of the authors and do not necessarily reflect those of the NHS, the NIHR, NETSCC, the HTA programme or the Department of Health.