Massage with or without aromatherapy for symptom relief in people with cancer
Abstract
Background
Massage and aromatherapy massage are used to relieve cancer‐related symptoms. A number of claims have been made for these treatments including reduction of pain, anxiety, depression, and stress. Other studies have not shown these benefits.
Objectives
To evaluate the effects of massage with or without aromatherapy on pain and other symptoms associated with cancer.
Search methods
We searched the following databases and trials registries up to August 2015: the Cochrane Central Register of Controlled Trials (CENTRAL, 2015, Issue 7), MEDLINE (Ovid), EMBASE (Ovid), PsycINFO (Ovid), CINAHL (EBSCO), PubMed Cancer Subset, SADCCT, and the World Health Organization (WHO) ICTRP. We also searched clinical trial registries for ongoing studies.
Selection criteria
Randomised controlled studies (RCTs) reporting the effects of aromatherapy or massage therapy, or both, in people with cancer of any age. We applied no language restrictions. Comparators were massage (using carrier oil only) versus no massage, massage with aromatherapy (using carrier oil plus essential oils) versus no massage, and massage with aromatherapy (using carrier oil plus essential oils) versus massage without aromatherapy (using carrier oil only).
Data collection and analysis
At least two review authors selected studies, assessed the risk of bias, and extracted data relating to pain and other symptoms associated with cancer, using standardised forms. We assessed the evidence using GRADE (Grading of Recommendations Assessment, Development and Evaluation) and created two 'Summary of findings' tables.
Main results
We included 19 studies (21 reports) of very low quality evidence with a total of 1274 participants. We included 14 studies (16 reports) in a qualitative synthesis and five studies in a quantitative synthesis (meta‐analysis). Thirteen studies (14 reports, 596 participants) compared massage with no massage. Six studies (seven reports, 561 participants) compared aromatherapy massage with no massage. Two studies (117 participants) compared massage with aromatherapy and massage without aromatherapy. Fourteen studies had a high risk of bias related to sample size and 15 studies had a low risk of bias for blinding the outcome assessment. We judged the studies to be at unclear risk of bias overall. Our primary outcomes were pain and psychological symptoms. Two studies reported physical distress, rash, and general malaise as adverse events. The remaining 17 studies did not report adverse events. We downgraded the GRADE quality of evidence for all outcomes to very low because of observed imprecision, indirectness, imbalance between groups in many studies, and limitations of study design.
Massage versus no‐massage groups
We analysed results for pain and anxiety but the quality of evidence was very low as most studies were small and considered at an unclear or high risk of bias due to poor reporting. Short‐term pain (Present Pain Intensity‐Visual Analogue Scale) was greater for the massage group compared with the no‐massage group (one RCT, n = 72, mean difference (MD) ‐1.60, 95% confidence interval (CI) ‐2.67 to ‐0.53). Data for anxiety (State‐Trait Anxiety Inventory‐state) relief showed no significant difference in anxiety between the groups (three RCTs, n = 98, combined MD ‐5.36, 95% CI ‐16.06 to 5.34). The subgroup analysis for anxiety revealed that the anxiety relief for children was greater for the massage group compared with the no‐massage group (one RCT, n = 30, MD ‐14.70, 95% CI ‐19.33 to ‐10.07), but the size of this effect was considered not clinically significant. Furthermore, this review demonstrated no differences in effects of massage on depression, mood disturbance, psychological distress, nausea, fatigue, physical symptom distress, or quality of life when compared with no massage.
Massage with aromatherapy versus no‐massage groups
We analysed results for pain, anxiety, symptoms relating to the breast, and quality of life but the quality of evidence was very low as studies were generally at a high risk of bias. There was some indication of benefit in the aromatherapy‐massage group but this benefit is unlikely to translate into clinical benefit. The relief of medium‐ and long‐term pain (medium‐term: one RCT, n = 86, MD 5.30, 95% CI 1.52 to 9.08; long‐term: one RCT, n = 86, MD 3.80, 95% CI 0.19 to 7.41), anxiety (two RCTs, n = 253, combined MD ‐4.50, 95% CI ‐7.70 to ‐1.30), and long‐term symptoms relating to the breast in people with breast cancer (one RCT, n = 86, MD ‐9.80, 95% CI ‐19.13 to ‐0.47) was greater for the aromatherapy‐massage group, but the results were considered not clinically significant. The medium‐term quality of life score was lower (better) for the aromatherapy‐massage group compared with the no‐massage group (one RCT, n = 30, MD ‐2.00, 95% CI ‐3.46 to ‐0.54).
Massage with aromatherapy versus massage without aromatherapy groups
From the limited evidence available, we were unable to assess the effect of adding aromatherapy to massage on the relief of pain, psychological symptoms including anxiety and depression, physical symptom distress, or quality of life.
Authors' conclusions
There was a lack of evidence on the clinical effectiveness of massage for symptom relief in people with cancer. Most studies were too small to be reliable and key outcomes were not reported. Any further studies of aromatherapy and massage will need to address these concerns.
PICOs
Plain language summary
Aromatherapy and massage for symptom relief in people with cancer
Background
People with cancer may experience symptoms such as pain, anxiety, or distress. Massage with or without aromatherapy (using essential oils, which are natural oils that may have the odour of the plant from which it was extracted) may help relieve these symptoms. Massage involves working and acting on the body with pressure. Massage is given using a carrier oil (base oil or vegetable oil) with or without essential oils. Massage with essential oils such as rose or lavender oil is known as aromatherapy massage.
Key results and quality of evidence
In August 2015, we searched for clinical trials looking at massage with or without aromatherapy for symptom relief in people with cancer. We found 19 small studies (1274 participants) of very low quality. Some small studies suggested that massage without aromatherapy may help relieve short‐term pain and anxiety in people with cancer. Other small studies suggested that aromatherapy massage may provide medium‐ or long‐term relief for these symptoms. However, the quality of evidence was very low and the results were not consistent. We cannot be sure that these treatments will bring any benefit.
Authors' conclusions
Summary of findings
| Massage versus no massage for symptom relief in people with cancer | ||||||
| Patient or population: people with cancer Settings: oncology unit, cancer centre, hospice | ||||||
| Outcomes | Illustrative comparative risks* (95% CI) | Relative effect | No of participants | Quality of the evidence (GRADE) | Comments | |
| Assumed risk | Corresponding risk (95% CI) | |||||
| No massage | Massage | |||||
| Pain (PPI‐VAS) | The mean pain (PPI‐VAS) in the control group was 4.2 points | The mean pain (PPI‐VAS) in the intervention group was 1.6 lower | Continuous data | 72 | ⊕⊝⊝⊝ | Lower score indicates less pain |
| Anxiety (STAI‐state) | The mean anxiety (STAI‐state) ranged across control groups from 30.0 to 37.7 points | The mean anxiety (STAI‐state) in the intervention groups was 5.36 lower | Continuous data | 98 | ⊕⊝⊝⊝ | Not statistically significant by random‐effects model |
| Anxiety (STAI‐state) subgroup 1: children vs. adults ‐ children | The mean anxiety (STAI‐state) for children in the control group was 37.7 points | The mean anxiety (STAI‐state) for children in the intervention group was 14.70 lower | Continuous data | 30 | ⊕⊝⊝⊝ | Lower score indicates less severity in anxiety |
| Anxiety (STAI‐state) subgroup 1: children vs. adults ‐ adults | The mean anxiety (STAI‐state) for adults ranged across control groups from 30.0 to 30.3 points | The mean anxiety (STAI‐state) for adults in the intervention groups was 0.74 lower | Continuous data | 68 | ⊕⊝⊝⊝ | Not statistically significant |
| Anxiety (STAI‐state) subgroup 2: short‐term vs. medium‐term ‐ short‐term (≤ 4 weeks) | The short‐term mean anxiety (STAI‐state) ranged across control groups from 30.3 to 37.7 points | The short‐term mean anxiety (STAI‐state) in the intervention groups was 10.66 lower | Continuous data | 64 | ⊕⊝⊝⊝ | Lower score indicates less severity in anxiety |
| Anxiety (STAI‐state) subgroup 2: short‐term vs. medium‐term ‐ medium‐term (> 4 weeks to < 8 weeks) | The medium‐term mean anxiety (STAI‐state) in the control group was 30.0 points | The medium‐term mean anxiety (STAI‐state) in the intervention group was 3.00 lower | Continuous data | 34 | ⊕⊝⊝⊝ | Not statistically significant |
| * The assumed risk (e.g. the mean control group risk across studies) is provided. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group. | ||||||
| GRADE Working Group grades of evidence | ||||||
| 1 Study with high risk of bias. All downgraded by three levels due to very serious imprecision. | ||||||
| Aromatherapy massage versus no massage for symptom relief in people with cancer | ||||||
| Patient or population: people with cancer Settings: oncology unit, cancer centre, hospice | ||||||
| Outcomes | Illustrative comparative risks* (95% CI) | Relative effect | No of participants | Quality of the evidence | Comments | |
| Assumed risk | Corresponding risk | |||||
| No massage | Aromatherapy massage | |||||
| Bodily pain (SF‐8) ‐ medium‐term (> 4 weeks to < 8 weeks) | The medium‐term mean bodily pain (SF‐8) in the control group was 44.4 points | The medium‐term mean bodily pain (SF‐8) in the intervention group was 5.30 higher | Continuous data | 86 | ⊕⊝⊝⊝ | Higher score indicates less pain |
| Bodily pain (SF‐8) ‐ long‐term (≥ 8 weeks) | The long‐term mean bodily pain (SF‐8) in the control group was 45.4 points | The long‐term mean bodily pain (SF‐8) in the intervention group was 3.80 higher | Continuous data | 86 | ⊕⊝⊝⊝ | Higher score indicates less pain |
| Anxiety (STAI‐state) | The mean anxiety (STAI‐state) ranged across control groups from 24.7 to 47.3 points | The mean anxiety (STAI‐state) in the intervention groups was 4.50 lower | Continuous data | 253 | ⊕⊝⊝⊝ | Lower score indicates less severity in anxiety |
| Anxiety (STAI‐state) subgroup: short‐term vs. medium‐term ‐ short‐term (≤ 4 weeks) | The short‐term mean anxiety (STAI‐state) in the control groups was 24.7 points | The short‐term mean anxiety (STAI‐state) in the intervention groups was 1.1 lower | Continuous data | 32 | ⊕⊝⊝⊝ | Not statistically significant |
| Anxiety (STAI‐state) subgroup: short‐term vs. medium‐term ‐ medium‐term (> 4 weeks to < 8 weeks) | The medium‐term mean anxiety (STAI‐state) in the control group was 47.3 points | The medium‐term mean anxiety (STAI‐state) in the intervention group was 5.1 lower | Continuous data | 221 | ⊕⊕⊕⊝ | Lower score indicates less severity in anxiety |
| Symptoms relating to the breast (EORTC QLQ‐BR23): long‐term (≥ 8 weeks) | The long‐term mean symptoms relating to the breast (EORTC QLQ‐BR23) in the control group was 31.9 points | The long‐term mean symptoms relating to the breast (EORTC QLQ‐BR23) in the intervention group was 9.80 lower | Continuous data | 86 | ⊕⊝⊝⊝ | Lower scores indicate fewer symptoms |
| Quality of life (MYMOP): medium‐term (> 4 weeks to < 8 weeks) | The medium‐term mean quality of life (MYMOP) in the control group was 3.9 points | The medium‐term mean quality of life (MYMOP) in the intervention group was 2.00 lower | Continuous data | 29 | ⊕⊝⊝⊝ | Lower score indicates greater quality of life |
| * The assumed risk (e.g. the mean control group risk across studies) is provided. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group. | ||||||
| GRADE Working Group grades of evidence | ||||||
| 1 Study with high risk of bias. Most except anxiety (four to eight weeks) downgraded by three levels due to very serious imprecision. | ||||||
Background
Description of the condition
Cancer significantly affects a person's quality of life (Alacacioglu 2010), and is associated with a variety of psychological symptoms, such as anxiety (Jackson 2004) and depression (Akechi 2008), and physiological symptoms, such as pain (Jackson 2007), fatigue (Bennett 2009), constipation (Yu 2010), and nausea and vomiting (Hines 2009). Some palliative care reports state that between 14% and 25% of all people with cancer show signs of anxiety disorders (Mantovan 2009), and that 50% of people with cancer are diagnosed with a psychiatric disorder; with the most common diagnosis being depressive disorders (Derogatis 1983).
Description of the intervention
This review focused on massage interventions using a blended carrier oil with essential oils (aromatherapy) and a carrier oil without essential oils.
Massage intervention
Massage is defined as the manipulation of the soft tissues of the body, performed by the hands, for the purpose of producing effects on the vascular, muscular, and nervous systems (Fellowes 2004a). Massage interventions are increasingly being considered as a means of achieving cancer‐related symptom relief (Ernst 2009; Wilkinson 2008), and are often used to address people with cancer's need for human contact (Russell 2008). The main reported effects of massage treatment, in studies with or without randomisation, include pain relief (Gorman 2008); improved immune function (Hernandez‐Reif 2004); reduced levels of anxiety (Campeau 2007) and depression (Krohn 2010); reduced fatigue (Listing 2009), nausea (Billhult 2007), and stress (Listing 2010); and an improved quality of life (Keir 2010). In massage therapy, the pressure should not be applied on the affected area and massage should not be given to people with contraindications such as acute thrombosis, or inflamed skin in the area of therapy (Listing 2009; Listing 2010). Toth 2013 and Hernandez‐Reif 2004 reported that no adverse effects were associated with massage. However, some people reported experiencing physical distress and feeling stressed at follow‐up (Jane 2011). Ernst 2003 also concluded that massage therapies are not totally devoid of risk, though the incidence is rare.
Aromatherapy massage intervention
Aromatherapy massage involves the use of essential oils that are combined with a carrier oil or cream to manipulate the soft tissues of the body (Fellowes 2004a). Lavender essential oil is used as a traditional therapy for pain and relaxation (Denner 2009). Bergamot is a well‐known essential oil used to minimise the symptoms of stress‐induced anxiety, mild mood disorders, and cancer pain (Bagetta 2010). Of all the uses of essential oils in aromatherapy, massage for 30 minutes (Listing 2009;Wilcock 2004), by trained therapists (Listing 2009; Wilcock 2004; Wilkinson 2007), is the most common (Holt 2009), followed by aromatic baths. Vapourisation is also a very effective way of using essential oils, whereby a small amount of oil is vaporised into the air (Oh 2000). Essential oils such as lavender (Soden 2004; Wilcock 2004), rose (Listing 2009), and chamomile (Wilcock 2004) were used for aromatherapy‐massage intervention for people with cancer. The effects of aromatherapy interventions in relieving cancer‐related symptoms have been reported (Wilkinson 2007). However, the evidence supporting a clear benefit of aromatherapy for people with cancer has yet to be established. There are a few studies that appear to exhibit a benefit. For example, Wilkinson 1999 performed a randomised controlled trial (RCT) that suggested that massage with or without essential oils was an effective therapy for reducing anxiety levels. It is believed that the addition of an essential oil can enhance the effect of massage and improve the psychological symptoms and overall quality of life among people with cancer. Soden 2004 compared the effects of massage with an essential oil (aromatherapy massage) to massage without an essential oil, and noted the changes in physical and psychological symptoms in people with advanced cancer. The results appeared to show that the addition of lavender essential oil to the massage did not increase the beneficial effects of massage. Wilcock 2004 reported there were no serious adverse events except a rash following aromatherapy massage. Miller 2012 reported that topical application of limonene containing massage oil to the breast was possibly or probably related to adverse events such as itching, rash, dry skin, lightening skin colour, burning sensation, acne, and headaches.
How the intervention might work
Massage is thought to have an effect by stimulating the skin, blood, and lymphatic system, which boosts blood circulation, aids muscle relaxation, and soothes nerves (McGilvery 1994). However, the mechanisms underlying these effects remain unknown. Aromatherapy massage using essential oils is considered a therapeutic treatment for both the mind and body, and it works mainly on the nervous system, but may also stimulate the immune system and affect emotions. Essential oils are composed of small organic molecules that penetrate the outer skin, work their way into the body and accumulate in the fatty tissue. In addition, the highly volatile oils evaporate and can also be inhaled through the nasal passages. These olfactory cells send messages straight to higher centres of the brain, including the limbic system, which controls the arousal functions of the body and emotional states. Thus, during an aromatherapy treatment, essential oils may enhance both physical and psychological well‐being at the same time (McGilvery 1994).
Why it is important to do this review
This review is an update of a Cochrane review first published in 2004 (Fellowes 2004a). The earlier review found some indication of the usefulness of this alternative therapy, yet with little evidence. As this former review was significantly out of date, Cochrane withdrew it. Nevertheless, this topic needs to be updated due to more recently published RCTs and methodological updates. Physicians and people with cancer need access to the best available and up‐to‐date evidence to make informed treatment decisions. Alternative therapies continue to generate substantial interest.
Objectives
To evaluate the effects of massage with or without aromatherapy on pain and other symptoms associated with cancer.
Methods
Criteria for considering studies for this review
Types of studies
We included randomised controlled trials (RCTs) that evaluated the effect of massage with or without aromatherapy. We excluded studies if they were quasi‐randomised trials, non‐randomised trials, case reports, abstracts, or letters. We applied no language restrictions.
Types of participants
We included adults and children diagnosed with cancer. We included both inpatients and outpatients who received care in any healthcare setting (e.g. hospital, hospice, oncology centre, or community).
Types of interventions
We compared the following interventions:
-
massage (using carrier oil only) versus no massage;
-
massage with aromatherapy (using carrier oil plus essential oils) versus no massage;
-
massage with aromatherapy (using carrier oil plus essential oils) versus massage without aromatherapy (using carrier oil only).
Massage was required to include tissue manipulation using a carrier oil, thereby excluding touch therapies such as therapeutic touch, acupressure, and reflexology. We defined aromatherapy as the use of a blended carrier oil with essential oils and will include only aromatherapy administered with massage. Thus, we excluded inhalations and humidification methods.
Types of outcome measures
Primary outcomes
-
Pain (using validated standard subjective scales (numerical rating scale (NRS), verbal rating scale (VRS), or visual analogue scale (VAS)) for pain intensity or pain relief, or both).
-
Psychological symptoms (including anxiety and depression assessed using validated scales).
Secondary outcomes
-
Other physical symptoms (including fatigue and nausea, etc.).
-
Quality of life (assessed by a valid and reliable assessment instrument).
-
Adverse events.
Search methods for identification of studies
Electronic searches
We searched the following databases.
-
The Cochrane Central Register of Controlled Trials (CENTRAL), 2015, Issue 7 of 12.
-
MEDLINE (Ovid) 1946 to week one of August 2015.
-
EMBASE (Ovid) 1974 to week one of August 2015.
-
PsycINFO (Ovid) 1806 to week one of August 2015.
-
CINAHL (EBSCO) to week one of August 2015.
-
PubMed Cancer Subset to week one of August 2015.
-
South Asian Database of Controlled Clinical Trials (SADCCT) to week one of August 2015.
-
World Health Organization (WHO) ICTRP to week one of August 2015.
See Appendix 1; Appendix 2; Appendix 3; Appendix 4; Appendix 5; and Appendix 6 for the search strategies used. We applied no date or language restrictions.
Searching other resources
We expanded the search strategy to include regional databases such as the Korean databases (KISS, KMbase, KoreaMed, and RISS).
Data collection and analysis
Selection of studies
One review author (SHL) screened the titles and abstracts of the studies identified from the search to eliminate those studies that were not relevant to this review. When the title and abstract did not have sufficient information for screening purposes, we retrieved a full‐text copy to review. Two review authors (ESS, MJK) independently examined each full‐text report for the remaining studies to determine their eligibility for inclusion using a pre‐developed checklist. We excluded articles if they were not randomised. We did not consider for inclusion participants who were not diagnosed with cancer, did not receive a massage or aromatherapy‐massage intervention, or were not evaluated for any of the primary and secondary outcomes. An expert in massage and aromatherapy (ESS) and the research methodologist (SHL) made final decisions about which studies should be included in this review. We resolved disagreements about inclusions by discussion or via a formal consensus method.
Data extraction and management
Two review authors (YMJ, JEJ) independently extracted data from the included studies using a data collection form, which included the following information: source, eligibility, methods, participants, interventions, outcomes, results, and miscellaneous data. We completed a data extraction sheet for every study included in the review. We did not consider extracting individual participant data. Krohn 2010 only presented the results graphically and did not provide mean and standard deviation (SD) values for depression and mood. Soden 2004 reported the mean change value instead of mean and SD for the outcomes including pain, psychological distress such as anxiety and depression, other physical symptoms, and quality of life. We contacted five study authors to request missing data, but they did not respond (Batalha 2013; Khiewkhern 2013; Krohn 2010; Soden 2004; Wang 2015).
Assessment of risk of bias in included studies
Two review authors (KHS, JYY) independently assessed the risk of bias for each study, using the criteria outlined in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011), and adapted from those used by the Cochrane Pregnancy and Childbirth Group, with any disagreements resolved by discussion. We assessed the following for each study.
-
Random sequence generation (checking for possible selection bias). We assessed the method used to generate the allocation sequence as: low risk of bias (any truly random process, e.g. random number table; computer random number generator); unclear risk of bias (method used to generate sequence not clearly stated). We excluded studies using a non‐random process (e.g. odd or even date of birth; hospital or clinic record number).
-
Allocation concealment (checking for possible selection bias). The method used to conceal allocation to interventions prior to assignment determines whether intervention allocation could have been foreseen in advance of or during recruitment, or changed after assignment. We assessed the methods as: low risk of bias (e.g. telephone or central randomisation; consecutively numbered sealed opaque envelopes); unclear risk of bias (method not clearly stated). We excluded studies that did not conceal allocation (e.g. open list).
-
Blinding of outcome assessment (checking for possible detection bias). We assessed the methods used to blind study outcome assessors from knowledge of which intervention a participant received. We assessed the methods as: low risk of bias when this was clearly described, unclear risk of bias when not stated, and high risk of bias if an inappropriate method was described.
-
Incomplete outcome data (checking for possible attrition bias due to the amount, nature, and handling of incomplete outcome data). We assessed the methods used to deal with incomplete data as: low risk (less than 10% of participants did not complete the study or used 'baseline observation carried forward' analysis, or both); unclear risk of bias (used 'last observation carried forward' analysis); high risk of bias (used 'completer' analysis).
-
Selective outcome reporting. We considered studies at low risk of bias if all adverse events were reported. Where there was clear evidence of partial reporting (e.g. most common or more than a given rate), then we considered these studies at high risk of bias. Anything else was unclear risk of bias.
-
Size of study (checking for possible biases confounded by size). We assessed studies as being at low risk of bias (200 participants or greater); unclear risk of bias (50 to 199 participants); high risk of bias (fewer than 50 participants).
-
Other bias. Any other bias noted at the data extraction phase.
We used the Review Manager 5 'Risk of bias' tool (RevMan 2014).
Measures of treatment effect
For dichotomous data, we intended to describe the treatment effect as risk ratios (RRs) with 95% confidence intervals (CIs). For continuous data, we established the mean difference (MD) or the standardised mean difference (SMD) and calculated the 95% CI. If studies did not report SDs, we calculated them using software developed by the UK Cochrane Group (UK Cochrane Centre 2010) based on the methods described in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). We planned to calculate outputs such as the number needed to treat for an additional beneficial outcome (NNTB), number needed to treat for an additional harmful outcome (NNTH), or number needed to treat to prevent an event (NNTp), with 95% CIs from dichotomous data (McQuay 1997); however, this was not possible because no outcomes were reported in this way.
Unit of analysis issues
The analysis aimed to take into account the level at which randomisation occurred in the clinical trials. Review authors considered the unit of analysis issues in each study as follows.
-
Groups of participants randomised together with the same intervention, such as cluster‐randomised trials.
-
Participants receiving more than one intervention, such as a cross‐over trial, or undergoing simultaneous treatment of multiple study sites.
-
Multiple observations for the same outcome, such as repeated measurements, recurring events, and measurements of different body parts.
Dealing with missing data
We contacted the relevant study authors to provide the appropriate data for the meta‐analysis, but they did not respond to our requests. Where necessary, we imputed SDs from the standard error (SE) (Ahles 1999, Wilkinson 2007).
Assessment of heterogeneity
Where there was substantial heterogeneity, we checked whether the data were incorrectly extracted or entered into Review Manager 5 (RevMan 2014). We reviewed inconsistencies in the data that could cause misleading effects. We performed analyses both with and without outlying studies as part of a sensitivity analysis. We conducted subgroup analyses to explore heterogeneity. We assessed statistical heterogeneity visually and by using the I2 statistic. We considered studies with an I2 greater than 60% as having substantial heterogeneity.
Assessment of reporting biases
For assessing publication bias, we intended to assess how much data (e.g. studies and participants) would be required both to be unpublished and to have no treatment effect; however, this was not possible due to the lack of reliable data.
Data synthesis
Where data were available, we combined data from studies in a meta‐analysis. We used a fixed‐effect model if the studies were homogenous; otherwise, we used a random‐effects model. We combined continuous data only where means and SDs were available or calculable and there was no clear evidence of a skewed distribution.
Subgroup analysis and investigation of heterogeneity
When data were available, we undertook subgroup analyses for comparing the magnitudes of effect within Review Manager 5 (RevMan 2014). We compared effect estimates in different subgroups by considering the meta‐analysis results from each subgroup separately.
-
Children versus adults.
-
Short‐term (four weeks or less) versus medium‐term (greater than weeks and less than eight weeks) versus long‐term (eight weeks or greater).
We were unable to perform subgroup analysis for full‐body massage versus partial massage (hand, foot, shoulder, neck, back, abdomen, and scalp etc.) as planned due to lack of data.
Sensitivity analysis
We undertook a sensitivity analysis to explore the effects of risk of bias and other sources of heterogeneity, where data were available. In subgroup analyses, we estimated the effect of the intervention for each subgroup. We created two 'Summary of findings' tables with pre‐specified outcomes (Higgins 2011).
'Summary of findings' tables
We assessed the overall quality of the evidence using the GRADE system (GRADEpro GDT 2015), and presented in the 'Summary of findings' tables, to present the main findings of the review in a transparent and simple tabular format. In particular, we included key information concerning the quality of evidence, the magnitude of effect of the interventions examined, and the sum of available data on the main outcomes of pain, anxiety, symptoms relating to the breast, and quality of life.
The GRADE system uses the following criteria for assigning grade of evidence.
-
High quality: further research is very unlikely to change our confidence in the estimate of effect.
-
Moderate quality: further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate.
-
Low quality: further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate.
-
Very low quality: we are very uncertain about the estimate.
We decreased grade if:
-
serious (‐1) or very serious (‐2) limitation to study quality;
-
important inconsistency (‐1);
-
some (‐1) or major (‐2) uncertainty about directness;
-
imprecise or sparse data (‐1);
-
high probability of reporting bias (‐1).
Results
Description of studies
See: Characteristics of included studies and Characteristics of excluded studies tables.
Results of the search
The literature search identified 21,376 studies (to week 1 of August 2015), which was reduced to 17,512 after we removed duplicates. We assessed 55 full‐text articles as potentially eligible and excluded 34 of them. A total of 19 studies (21 reports) met the inclusion criteria for this review; we included 14 studies (16 reports) in a qualitative synthesis and five studies in a quantitative synthesis (meta‐analysis). There were two reports each for two studies (Fernandez‐Lao 2012; Listing 2009). The results of the search and screening processes are illustrated in Figure 1.
Study flow diagram.
Included studies
We included 19 studies (21 reports) with 1274 participants (700 interventions and 574 controls). Thirteen studies (14 reports) compared massage with no massage; these studies included 596 participants (302 interventions and 294 controls). Six studies (seven reports) compared aromatherapy massage with no massage; these studies included 561 participants (281 interventions and 280 controls). Two studies compared massage with aromatherapy with massage without aromatherapy; these studies included 117 participants (54 interventions and 63 controls). See Characteristics of included studies table. Table 1 provides additional details of the evaluated trials.
| Study ID | Primary outcome | Secondary outcome | |||
| Pain | Psychological symptoms | Other physical symptoms | Quality of life | Adverse events | |
| Comparison 1. Massage vs. no massage | |||||
| ‐ | Anxiety‐STAI‐state Depression‐BDI | ‐ | ‐ | ‐ | |
| Pain‐VAS | ‐ | ‐ | ‐ | ‐ | |
| ‐ | Anxiety‐VAS | Nausea‐VAS | ‐ | ‐ | |
| ‐ | Anxiety‐VAS Anxiety‐STAI‐state | ‐ | ‐ | ‐ | |
| ‐ | ‐ | Fatigue‐POMS | ‐ | ‐ | |
| ‐ | Anxiety‐STAI‐state | ‐ | ‐ | ‐ | |
| ‐ | Anxiety‐STAI‐state Anxiety‐SCL‐90‐R | ‐ | ‐ | ‐ | |
| Pain‐PPI‐VAS | Mood‐VAS | ‐ | ‐ | Having physical distress Progress of disease | |
| ‐ | Depression‐PHQ Mood‐BSF | ‐ | ‐ | ‐ | |
| Pain‐VAS | Anxiety‐HAD Depression‐HAD Psychological distress‐RSCL | Physical symptom distress‐RSCL | Quality of Life‐RSCL | ‐ | |
| Pain‐VAS | Anxiety‐VAS | ‐ | Quality of Life‐McGill | ‐ | |
| Pain‐ESAS:AM | Anxiety‐ESAS:AM Depression‐ESAS:AM | Nausea‐ESAS:AM | ‐ | ‐ | |
| Pain‐PAT (or SNVR) | ‐ | ‐ | Quality of Life‐Graham | ‐ | |
| Comparison 2. Aromatherapy with massage vs. no massage | |||||
| Pain‐VAS | Anxiety‐VAS | Fatigue‐VAS | ‐ | ‐ | |
| Limb pain‐GBB | ‐ | Arm symptoms‐EORTC QLQ‐BR23 | ‐ | ‐ | |
| Pain‐VAS | Anxiety‐HAD Depression‐HAD Psychological distress‐RSCL | Physical symptom distress‐RSCL | Quality of Life‐RSCL | ‐ | |
| Pain‐VAS | Anxiety‐STAI‐state | ‐ | ‐ | ‐ | |
| ‐ | Mood disturbance scale‐POMS | ‐ | Quality of life‐MYMOP | Rash | |
| Pain‐EORTC | Anxiety‐STAI‐state | Fatigue‐EORTC | Quality of life‐EORTC | ‐ | |
| Comparison 3. Aromatherapy with massage vs. massage without aromatherapy | |||||
| Pain‐VAS | Anxiety‐HAD Depression‐HAD Psychological distress‐RSCL | Physical symptom distress‐RSCL | Quality of Life‐RSCL | ‐ | |
| ‐ | Anxiety‐STAI‐state (SAI) Anxiety‐STAI‐trait (TAI) Psychological distress‐RSCL | Physical symptom distress‐RSCL | Quality of life‐RSCL | ‐ | |
| Abbreviations: BDI: Beck Depression Inventory; BSF: Berlin Mood Questionnaire; CES‐D: Center for Epidemiological Studies Depression; EORTC QLQ‐BR23: European Organization of Research and Treatment of Cancer Quality of Life Questionnaire Breast Module (lower scores of the arm and breast symptoms indicate fewer symptoms); ESAS:AM: Edmonton Symptom Assessment System‐Ascites Modification; GBB: Giessen Complaints Inventory; HAD: Hospital Anxiety Depression Scale; MYMOP: Measure Yourself Medical Outcome Profile (lower score indicates greater quality of life, 7 = 'as bad as it could be' and 0 = 'as good as it could be'); PAT: Pain Assessment Tool; PHQ: Patient Health Questionnaire; POMS: Brief Profile of Mood States; PPI‐VAS: Present Pain Intensity‐Visual Analogue Scale (lower score indicates less pain, 10 = 'pain as bad as it could be' and 0 = 'no pain'); RSCL: Rotterdam Symptom Checklist; SCL‐90‐R: Symptom Checklist‐90‐R; SF‐8: Short‐Form Health Survey‐8 (higher score on bodily pain indicates less pain); SNVR: Skilled Nursing Visit Report form; STAI: State‐Trait Anxiety Inventory (lower score indicates less severity in anxiety); VAS: Visual Analogue Scale. | |||||
Excluded studies
We excluded 34 studies because they were not RCTs (21 studies); they were review articles (seven articles); or assessed multiple therapies (one study), different comparison (two studies), or different outcomes (three studies). See Characteristics of excluded studies table.
Risk of bias in included studies
We created a risk of bias table and graph to summarise our judgements on random sequence generation, allocation concealment, blinding of outcome assessment, incomplete outcome data, selective reporting, size, and other potential sources of bias in the analysed studies (Figure 2; Figure 3). We assessed and classified the methodological components of the trials as low, high, or unclear according to the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011).
Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.
Risk of bias summary: review authors' judgements about each risk of bias item for each included study.
Allocation
Three of the 19 studies were potentially at high risk of bias due to the inadequate random sequence generation (Haun 2009; Listing 2009; Toth 2013). Only six studies were at low risk of bias for selection bias (Fernandez‐Lao 2012; Hernandez‐Reif 2004; Jane 2011; Wilcock 2004; Wilkinson 1999; Wilkinson 2007). One of the 19 studies had a high risk of bias in allocation concealment (Campeau 2007), and 14 studies showed an unclear risk of bias in allocation concealment. Numbered, sealed, opaque envelopes, and central allocation by computer were used for the proper allocation of concealment.
Blinding
Fifteen of the included studies had a low risk of bias for blinding of outcome assessment, whereas four studies showed an unclear risk of bias for this assessment (Ahles 1999; Campeau 2007; Hernandez‐Reif 2004; Wilkie 2000).
Incomplete outcome data
Of the 19 studies, seven were at low risk of attrition bias having allocated the same number of participants to both groups (Ahles 1999; Billhult 2007; Fernandez‐Lao 2012; Haun 2009; Jane 2011; Sohn 2005; Wang 2015). Ten studies were at unclear risk of attrition bias, and the remaining two studies showed a high risk of bias (Wilcock 2004; Wilkie 2000).
Selective reporting
Of the 19 studies, eight were at low risk of bias due to selective reporting (Batalha 2013; Billhult 2007; Campeau 2007; Fernandez‐Lao 2012; Khiewkhern 2013; Krohn 2010; Wang 2015; Wilcock 2004). Nine studies were at unclear risk of reporting bias, and the remaining two studies showed a high risk of bias (Haun 2009; Wilkie 2000).
Size
Eleven studies had fewer than 50 participants in total and we judged them at high risk of bias. One study was at low risk of bias with more than 200 participants (Wilkinson 2007). We judged the remaining studies to be at unclear risk of bias (see: Figure 3).
Other potential sources of bias
We found some imbalanced baseline measurements between the intervention and control groups, which caused potential bias. However, some trials reported that there were no significant between‐group differences in terms of participant characteristics (Hernandez‐Reif 2004), or measurement of symptoms relating to the breast in people with breast cancer or the use of medication (Ahles 1999), or sociodemographic variables at baseline (Krohn 2010; Listing 2009) between the massage and no‐massage groups. There were no imbalances in the baseline measurement of quality of life (Wilkinson 1999), pain intensity, and anxiety (Sohn 2005) between the aromatherapy‐massage and no‐massage groups. Of the 19 studies, five had a high risk of other bias, and five studies had an unclear risk of other bias. We judged the remaining studies at low risk of bias.
Effects of interventions
See: Summary of findings for the main comparison Massage versus no massage for symptom relief in people with cancer; Summary of findings 2 Aromatherapy massage versus no massage for symptom relief in people with cancer
See summary of findings Table for the main comparison and summary of findings Table 2 for the main comparisons.
Studies used a variety of scales to measure symptoms. These included:
-
BDI: Beck Depression Inventory (measurement of depression);
-
BSF: Berlin Mood Questionnaire (measurement of mood);
-
CES‐D: Center for Epidemiological Studies Depression (measurement of depression);
-
EORTC QLQ: European Organization of Research and Treatment of Cancer Quality of Life Questionnaire (measurement of pain, fatigue, and quality of life);
-
EORTC QLQ‐BR23: European Organization of Research and Treatment of Cancer Quality of Life Questionnaire Breast Module (measurement of symptoms relating to the breast and arm; low scale scores of the arm and breast symptoms indicate fewer symptoms);
-
ESAS:AM: Edmonton Symptom Assessment System‐Ascites Modification (measurement of pain, anxiety, depression, and nausea);
-
GBB: Giessen Complaints Inventory (measurement of limb pain);
-
the Graham scale (measurement of quality of life);
-
HAD: Hospital Anxiety Depression Scale (measurement of anxiety and depression);
-
the McGill scale (measurement of quality of life);
-
MYMOP: Measure Yourself Medical Outcome Profile (measurement of quality of life; lower score indicates greater quality of life, 7 = 'as bad as it could be' and 0 = 'as good as it could be');
-
PAT: Pain Assessment Tool (measurement of pain);
-
PHQ: Patient Health Questionnaire (measurement of depression);
-
POMS: Brief Profile of Mood States (measurement of depression, mood disturbance, and fatigue);
-
PPI‐VAS: Present Pain Intensity‐Visual Analogue Scale (measurement of pain; lower score indicates less pain, 10 = 'pain as bad as it could be' and 0 = 'no pain');
-
RSCL: Rotterdam Symptom Checklist (measurement of psychological distress, physical symptom distress, and quality of life);
-
SCL‐90‐R: Symptom Checklist‐90‐R (measurement of anxiety and depression);
-
SF‐8: Short‐Form Health Survey‐8 (measurement of bodily pain; higher score on bodily pain indicates less pain);
-
SNVR: Skilled Nursing Visit Report form (measurement of pain);
-
STAI: State‐Trait Anxiety Inventory; low score indicating less severity in anxiety (measurement of anxiety);
-
VAS: Visual Analogue Scale (measurement of pain, anxiety, mood, nausea and fatigue).
Comparison 1: massage versus no massage
For the comparison of massage versus no massage, we found primary outcome data for pain and psychological symptoms and secondary outcome data for other physical symptoms and quality of life.
Primary outcomes
Pain
One trial with a short‐term follow‐up provided data on pain using PPI‐VAS (n = 72, Jane 2011). Both the massage and no‐massage groups exhibited improvements over the baseline values and short‐term pain relief was greater for the massage group compared with the no‐massage group (one RCT, n = 72, MD ‐1.60, 95% CI ‐2.67 to ‐0.53). One trial reported on pain using PAT or SNVR at short‐term follow‐up and there was no statistically significant difference between groups (n = 56, Wilkie 2000). Four trials with no appropriate data for meta‐analysis reported no effect on pain using the PPI‐VAS (Batalha 2013; Soden 2004; Toth 2013) and ESAS:AM (Wang 2015).
We downgraded the GRADE quality of the evidence for this outcome to very low because of observed imprecision, indirectness, and limitation of study design.
Psychological symptoms
There were data on psychological symptoms including anxiety, depression, mood, and psychological distress.
Anxiety
Three included studies assessed anxiety using STAI‐state (Ahles 1999; Haun 2009; Hernandez‐Reif 2004). There was no significant difference in anxiety between the groups (three RCTs, n = 98, combined MD ‐5.36, 95% CI ‐16.06 to 5.34) (Analysis 1.1; Figure 4). The pooled results showed very high heterogeneity (I2 = 88%). Subgroup analysis revealed that anxiety relief for children was greater for the massage group compared with the no‐massage group (one RCT, n = 30, MD ‐14.70, 95% CI ‐19.33 to ‐10.07; lower score indicated less severity in anxiety) (Haun 2009). However, we considered the size of the effect to be not clinically significant. Data from adults showed no significant difference in anxiety relief between the groups (Ahles 1999; Hernandez‐Reif 2004) (Analysis 2.1). The subgroup analysis revealed that short‐term anxiety relief was greater for the massage group compared with the no‐massage group (two RCTs, n = 64, combined MD ‐10.66, 95% CI ‐14.72 to ‐6.60; lower score indicated less severity in anxiety) (Ahles 1999; Haun 2009). However, we considered the size of the effect to be not clinically significant. Data from medium‐term anxiety relief showed no significant difference between the groups (Hernandez‐Reif 2004) (Analysis 2.2). Three trials assessed anxiety using a VAS. Two trials with no usable data claimed massage therapy had no major impact on intermediate‐term anxiety (n = 100, Campeau 2007; n = 30, Toth 2013). Another trial reported that differences in anxiety between the two treatment regimens were not statistically significant (n = 39, Billhult 2007). Two further trials reported anxiety using a HAD tool and there were no statistically significant differences between the groups (n = 39, Billhult 2007; n = 42, Soden 2004). One trial reported reduced anxiety for an immediate massage therapy effect using the SCL‐90‐R scale (n = 34, Hernandez‐Reif 2004). One trial reported on anxiety at short‐term follow‐up using the ESAS:AM scale and there were significant reductions in anxiety scores in the massage group, but it did not provide evaluable data (n = 80, Wang 2015).
Forest plot of comparison: massage versus no massage, outcome anxiety (State‐Trait Anxiety Inventory (STAI)‐state).
We downgraded the GRADE quality of the evidence for anxiety to very low because of observed imprecision, indirectness, and limitation of study design.
Depression
One trial measured depression using the BDI scale and reported no effects of massage on depression (n = 34, Ahles 1999). Two trials reported no effect of massage therapy on depression using the HAD scale (n = 39, Billhult 2007; n = 42, Soden 2004). Soden 2004 reported statistically significant reductions in depression scores in the massage group; however, this trial did not provide evaluable data. One trial reported that the long‐term massage effects reduced depression using the POMS scale, and also that the immediate massage therapy effects reduced depression using the SCL‐90‐R scale (n = 34, Hernandez‐Reif 2004). One trial reported depression using the PHQ scale, which was significantly reduced immediately after massage compared to the control group (n = 34, Krohn 2010). One trial measured depression using the ESAS:AM scale at short‐term follow‐up and reported significant reductions in depression scores in the massage group; however, this trial did not provide evaluable data (n = 80, Wang 2015).
Mood
One trial reported that massage therapy showed beneficial effects on mood using a VAS (n = 72, Jane 2011). Another trial reported no significant alterations in mood after massage therapy using BSF (n = 34, Krohn 2010).
Psychological symptoms
One trial reported no significant differences between massage and no‐massage groups, and this trial provided no appropriate data for meta‐analysis (n = 42, Soden 2004).
Secondary outcomes
Other physical symptoms
Data were available for nausea, fatigue, and physical symptom distress.
Nausea
Two trials reported no significant differences in short‐term nausea between groups (VAS, ESAS:AM) (Billhult 2007; Wang 2015).
Fatigue
One trial reported no significant differences in short‐term fatigue between groups (POMS) (Fernandez‐Lao 2012).
Physical symptom distress
One trial reported no significant differences between groups (n = 42, Soden 2004).
Quality of life
One trial reported on quality of life at short‐term follow‐up using the Graham scale, and there were no statistically significant differences between groups (n = 29, Wilkie 2000). One trial reported that they were unable to demonstrate any significant long‐term benefit on quality of life using the RSCL scale but this trial provided no usable data (n = 42, Soden 2004). In contrast, one small trial reported that providing massage improved the quality of life at the end of life using the McGill scale (n = 30, Toth 2013). We considered these data to be too unreliable to include in the 'Summary of findings' table.
Adverse events
One trial reported physical distress in one participant who received the massage intervention (n = 36, Jane 2011).
Comparison 2: massage with aromatherapy versus no massage
For the comparison of massage with aromatherapy versus no massage, we found primary outcome data for pain and psychological symptoms and secondary outcome data for other physical symptoms, quality of life, and adverse events.
Primary outcomes
Pain
Three trials reported pain intensity using a VAS. Soden 2004 reported that the effects of aromatherapy demonstrated no significant long‐term benefit on pain intensity (n = 42). In contrast, Sohn 2005 reported that pain intensity was significantly decreased in the aromatherapy group compared with control group (n = 32). Khiewkhern 2013 reported pain at short‐term follow‐up was significantly decreased in the aromatherapy‐massage group than control group (n = 66). One study measured pain using EORTC at short‐, medium‐, and long‐term follow‐ups and reported no significant difference in pain relief between the aromatherapy‐massage and no‐massage groups (n = 221, Wilkinson 2007). One trial reported no significant difference in limb pain relief between groups at medium‐ and long‐term follow‐ups (n = 86, Listing 2009). The same study reported bodily pain using SF‐8 at medium‐ and long‐term follow‐ups. The relief of bodily pain was greater for the aromatherapy‐massage group compared with the no‐massage group for medium‐term (one RCT, n = 86, MD 5.30, 95% CI 1.52 to 9.08) and for long‐term (one RCT, n = 86, MD 3.80, 95% CI 0.19 to 7.41).
We downgraded the GRADE quality of the evidence for this outcome to very low because of observed imprecision, indirectness, and limitation of study design.
Psychological symptoms
We found data for anxiety, depression, mood disturbances, and psychological distress.
Anxiety
Two studies reported anxiety using STAI‐state at short‐, medium‐, and long‐term follow‐ups (Sohn 2005; Wilkinson 2007). The relief of anxiety was greater for the aromatherapy with massage group compared with the no‐massage group (two RCTs, n = 253, combined MD ‐4.50, 95% CI ‐7.70 to ‐1.30) (Analysis 3.1). Subgroup analysis revealed that medium‐term anxiety relief was greater for the aromatherapy‐massage group compared with the no‐massage group (one RCT, n = 221, MD ‐5.10, 95% CI ‐8.57 to ‐1.63) (Wilkinson 2007) (see: Analysis 4.1). Data from short‐term anxiety relief showed no significant difference between the groups (Sohn 2005). One trial using HAD reported no significant long‐term benefit of aromatherapy massage, but did not provide evaluable data (n = 42, Soden 2004). Khiewkhern 2013 (n = 66) reported that anxiety at short‐term follow‐up was significantly decreased in the aromatherapy‐massage group than no‐massage group using VAS, but did not provide appropriate data for meta‐analysis.
We downgraded the GRADE quality of the evidence for this outcome to very low because of observed imprecision, indirectness, and limitation of study design.
Depression
There were no significant differences using BDI and VAS in short‐term depression between the aromatherapy‐massage group and the no‐massage group (n = 32, Sohn 2005; n = 66, Khiewkhern 2013). One trial reported on depression using HAD and found no significant long‐term benefit of aromatherapy massage, but this trial did not provide evaluable data (n = 42, Soden 2004). There were no significant differences using CES‐D in short‐, medium‐, or long‐term depression between the aromatherapy‐massage group and the no‐massage group (n = 221, Wilkinson 2007).
Mood
One trial reported no significant differences in medium‐term mood disturbances between the aromatherapy‐massage group and the no‐massage group using POMS (n = 29, Wilcock 2004).
Psychological symptoms
One trial reported that there were no significant differences in psychological distress between the aromatherapy‐massage group and the no‐massage group using RSCL (n = 42, Soden 2004). This trial did not provide any appropriate data for meta‐analysis.
Secondary outcomes
Other physical symptoms
We found data for fatigue, physical symptom distress, and symptoms relating to the arm in people with cancer and to the breast in people with breast cancer.
Fatigue
One trial reported no significant difference between the aromatherapy‐massage group and the usual care only (no massage) group at six or 10 weeks post‐randomisation using EORTC (n = 221, Wilkinson 2007). Khiewkhern 2013 reported fatigue using VAS at short‐term follow‐up was significantly decreased in the aromatherapy‐massage group than the no‐massage group, but this trial did not provide appropriate data for meta‐analysis (n = 66).
Physical symptom distress
One trial reported that there was no significant difference between the aromatherapy‐massage group and no‐massage group using RSCL, but this trial did not provide any appropriate data for meta‐analysis (n = 42, Soden 2004).
Symptoms relating to the arm and breast
One study measured symptoms relating to the arm in people with cancer using EORTC QLQ‐BR23 (n = 86, Listing 2009). There were no significant differences between the groups in "arm symptoms" at medium‐ or long‐term follow‐ups. The same study measured "breast symptoms" in people with breast cancer at medium‐ and long‐term follow‐ups using the EORTC QLQ‐BR23 (Listing 2009). Long‐term relief of symptoms relating to the breast in people with breast cancer was greater for the aromatherapy‐massage group compared with the no‐massage group (one RCT, n = 86, MD ‐9.80, 95% CI ‐19.13 to ‐0.47).
Quality of life
One study used MYMOP to measure quality of life at medium‐term follow‐up (40 participants randomised but only 29 reported, Wilcock 2004). The medium‐term quality of life score was lower for the aromatherapy‐massage group compared with no‐massage group (one RCT, n = 29, MD ‐2.00, 95% CI ‐3.46 to ‐0.54; lower score indicated good quality of life). In contrast, in two other studies using RSCL and EORTC, the level of quality of life showed no significant difference between the aromatherapy‐massage group and the no‐massage group (n = 42, Soden 2004; n = 221, Wilkinson 2007).
Adverse events
One study reported rash (n = 1) and general malaise (n = 5) among the participants who received aromatherapy massage (n = 29, Wilcock 2004).
We downgraded the GRADE quality of the evidence for the secondary outcomes to very low because of observed imprecision, indirectness, and limitation of study design.
Comparison 3: massage with aromatherapy versus massage without aromatherapy
For the comparison of massage with aromatherapy versus massage without aromatherapy, we found primary outcome data for pain and psychological symptoms, and secondary outcome data for physical symptoms and quality of life.
Primary outcomes
Pain
One trial reported that there was no long‐term benefit of aromatherapy massage on improving pain control using VAS, but this trial did not provide any evaluable data for meta‐analysis (n = 42, Soden 2004).
Psychological symptoms
We found data for anxiety, depression, and psychological distress.
Anxiety
There were no significant differences in anxiety using STAI‐state or HAD at medium‐term follow‐up between the massage with aromatherapy group and the massage without aromatherapy group (n = 42, Soden 2004; n = 103, Wilkinson 1999). One trial did not provide any evaluable data for meta‐analysis (Soden 2004).
Depression
One trial reported that there were statistically significant reductions in depression scores using HAD in the massage without aromatherapy group, but this trial did not provide any evaluable data for meta‐analysis (n = 42, Soden 2004).
Psychological symptoms
Two trials reported no significant differences in psychological symptoms using RSCL at medium‐term follow‐up between the massage with aromatherapy group and the massage without aromatherapy group (n = 42, Soden 2004; n = 103, Wilkinson 1999). One trial provided no evaluable data for meta‐analysis (Soden 2004).
Secondary outcomes
Other physical symptoms
One trial reported on physical symptom distress using RSCL, and found no significant difference between the massage with aromatherapy group and the massage without aromatherapy group (Soden 2004). However, this trial did not provide any evaluable data for meta‐analysis.
Quality of life
Two trials reported no significant difference in quality of life using RSCL at medium‐term follow‐up between the massage with aromatherapy group and the massage without aromatherapy group (n = 42, Soden 2004; n = 103, Wilkinson 1999). We considered these data to be too unreliable to include in the 'Summary of findings' table.
Adverse events
We found no trials reporting adverse events of massage with aromatherapy or massage without aromatherapy.
Discussion
Summary of main results
We included 19 studies (21 reports) with 1274 participants. Thirteen studies (14 reports, 596 participants) compared massage with no massage. Six studies (seven reports 561 participants) compared aromatherapy massage with no massage. Two studies (117 participants) compared massage with aromatherapy with massage without aromatherapy. There was a lack of clear evidence to either support or not support the use of massage for symptom relief in people with cancer. Massage with or without aromatherapy may reduce pain or anxiety, or both. However, most studies were too small to be reliable. One study reported physical distress as an adverse event, another study reported one rash and five general malaises; the remaining 17 studies did not report adverse events.
Overall completeness and applicability of evidence
The objective of this review was to evaluate the effects of massage with or without aromatherapy on symptom relief in people with cancer. We included 19 studies (21 reports) with 1274 participants. The qualitative report included 14 studies (16 reports) and the quantitative synthesis (meta‐analysis) included five studies (Ahles 1999; Haun 2009; Hernandez‐Reif 2004; Sohn 2005; Wilkinson 2007). We included adults and children diagnosed with cancer. Two trials investigated the effectiveness of massage therapy for children with cancer with limited outcome measurement (Batalha 2013; Haun 2009). Batalha 2013 reported pain relief and Haun 2009 only demonstrated the effect on anxiety at short‐term, although both studies had small sample sizes. We included both inpatients and outpatients who received care in any healthcare setting (e.g. hospital, hospice, oncology centre, or community). However, we were unable to examine the differences between inpatients and outpatients because some of the studies did not provide information about the settings. There were just three comparisons: massage (using carrier oil only) versus no massage; massage with aromatherapy (using carrier oil plus essential oils) versus no massage; and massage with aromatherapy (using carrier oil plus essential oils) versus massage without aromatherapy (using carrier oil only). We excluded touch therapies such as therapeutic touch, acupressure, and reflexology, and inhalations and humidification methods since they were not administered using the massage technique. Five small studies reported quality of life (Soden 2004; Toth 2013; Wilcock 2004; Wilkie 2000; Wilkinson 1999). The primary outcomes for this review were pain and psychological symptoms including anxiety, depression, and mood disturbance assessed using validated scales. Ten studies reported pain and 13 studies reported psychological symptoms. We considered other physical symptoms including fatigue and nausea, quality of life, and adverse events as secondary outcomes. Eight studies reported other physical symptoms and six studies reported quality of life. Only two trials reported adverse events: having physical distress due to the progression of the disease during massage treatment (Jane 2011), and rash and general malaise with aromatherapy massage (Wilcock 2004).
Quality of the evidence
Ten studies (52.6%) had an unclear risk of bias for random sequence generation, and three studies (15.8%) had a high risk of bias for random sequence generation. Fourteen studies (73.7%) had an unclear risk of bias for allocation concealment. Fourteen studies (73.7%) had a high risk of bias related to sample size and only one study met our criteria of low risk of bias for size. However, 15 studies (78.9%) had a low risk of bias for blinding of outcome assessment. Two studies (10.55) had a high risk of bias related to incomplete outcome data, and nine studies (47.4%) had an unclear risk of bias related to selective reporting. We evaluated overall quality of the evidence using GRADE (see: summary of findings Table for the main comparison; summary of findings Table 2). Domains of the quality of evidence assessment included study design limitations, inconsistency, indirectness, imprecision, and publication bias. We downgraded the GRADE quality of the evidence for all outcomes to very low because of observed imprecision, indirectness, imbalance between groups in many studies, and limitations of study design.
Potential biases in the review process
We needed to make some variations from the original protocol, which are reported in the Differences between protocol and review section. In this review, we applied no language restrictions; however, only one trial included in the quantitative synthesis (meta‐analysis) was published in languages other than English (Korean) (Sohn 2005). Overlooking some published trials in languages other than English can be the source of potential bias. We found imbalance in the baseline measurement for the following: nausea and distress between the massage and no‐massage groups (Ahles 1999); anxiety (STAI) between the massage and no‐massage groups (Hernandez‐Reif 2004); symptoms relating to the breast between the massage and no‐massage groups (Listing 2009); and quality of life (RSCL) between the massage with aromatherapy and massage without aromatherapy groups (Wilkinson 1999). Batalha 2013 reported differences between the massage and no‐massage groups at the beginning of the trial. One trial had a potential problem due to carryover effects in crossover design (Fernandez‐Lao 2012). One trial applied a different intervention: inpatients received daily session and outpatients received weekly sessions (Haun 2009), which could have presented a source of potential bias. Due to high heterogeneity (I2 = 88%), we performed analyses using both the fixed‐effect model and random‐effects model (see Figure 4), and conducted subgroup analyses by duration (short‐term versus medium‐term) and population group (children versus adults). There were statistically different results according to the effect models. When applying the fixed‐effect model, anxiety relief was greater for the massage group compared with the no‐massage group (three RCTs, n = 98, combined MD ‐8.60, 95% CI ‐12.07 to ‐5.13) (Analysis 1.1; Figure 4). However, the overall size was small so we accepted the finding from analyses applying a random‐effects model that showed no significant difference in anxiety relief between the massage and no‐massage groups (three RCTs, n = 98, combined MD ‐5.36, 95% CI ‐16.06 to 5.34) (Analysis 1.1; Figure 4).
Agreements and disagreements with other studies or reviews
The results of this review of 19 RCTs (1274 participants) were generally in agreement with the original version, which included eight RCTs (357 participants) (Fellowes 2004a), and a previously updated version, which included 10 trials (Wilkinson 2008). While our findings showed that massage without aromatherapy may help relieve short‐term pain and anxiety in people with cancer and aromatherapy massage may provide medium‐ or long‐term relief for pain, anxiety, symptoms relating to the breast, and quality of life, the quality of the studies was poor and not reliable. There remains a lack of reliable evidence for the use of aromatherapy and massage to improve clinical outcomes.
Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.
Risk of bias summary: review authors' judgements about each risk of bias item for each included study.
Forest plot of comparison: massage versus no massage, outcome anxiety (State‐Trait Anxiety Inventory (STAI)‐state).
Comparison 1 Massage versus no massage, Outcome 1 Anxiety (STAI‐state).
Comparison 2 Massage versus no massage: subgroup analysis, Outcome 1 Anxiety (STAI‐state).
Comparison 2 Massage versus no massage: subgroup analysis, Outcome 2 Anxiety (STAI‐state).
Comparison 3 Aromatherapy with massage versus no massage, Outcome 1 Anxiety (STAI‐state).
Comparison 4 Aromatherapy with massage versus no massage: subgroup analysis, Outcome 1 Anxiety (STAI‐state).
| Massage versus no massage for symptom relief in people with cancer | ||||||
| Patient or population: people with cancer Settings: oncology unit, cancer centre, hospice | ||||||
| Outcomes | Illustrative comparative risks* (95% CI) | Relative effect | No of participants | Quality of the evidence (GRADE) | Comments | |
| Assumed risk | Corresponding risk (95% CI) | |||||
| No massage | Massage | |||||
| Pain (PPI‐VAS) | The mean pain (PPI‐VAS) in the control group was 4.2 points | The mean pain (PPI‐VAS) in the intervention group was 1.6 lower | Continuous data | 72 | ⊕⊝⊝⊝ | Lower score indicates less pain |
| Anxiety (STAI‐state) | The mean anxiety (STAI‐state) ranged across control groups from 30.0 to 37.7 points | The mean anxiety (STAI‐state) in the intervention groups was 5.36 lower | Continuous data | 98 | ⊕⊝⊝⊝ | Not statistically significant by random‐effects model |
| Anxiety (STAI‐state) subgroup 1: children vs. adults ‐ children | The mean anxiety (STAI‐state) for children in the control group was 37.7 points | The mean anxiety (STAI‐state) for children in the intervention group was 14.70 lower | Continuous data | 30 | ⊕⊝⊝⊝ | Lower score indicates less severity in anxiety |
| Anxiety (STAI‐state) subgroup 1: children vs. adults ‐ adults | The mean anxiety (STAI‐state) for adults ranged across control groups from 30.0 to 30.3 points | The mean anxiety (STAI‐state) for adults in the intervention groups was 0.74 lower | Continuous data | 68 | ⊕⊝⊝⊝ | Not statistically significant |
| Anxiety (STAI‐state) subgroup 2: short‐term vs. medium‐term ‐ short‐term (≤ 4 weeks) | The short‐term mean anxiety (STAI‐state) ranged across control groups from 30.3 to 37.7 points | The short‐term mean anxiety (STAI‐state) in the intervention groups was 10.66 lower | Continuous data | 64 | ⊕⊝⊝⊝ | Lower score indicates less severity in anxiety |
| Anxiety (STAI‐state) subgroup 2: short‐term vs. medium‐term ‐ medium‐term (> 4 weeks to < 8 weeks) | The medium‐term mean anxiety (STAI‐state) in the control group was 30.0 points | The medium‐term mean anxiety (STAI‐state) in the intervention group was 3.00 lower | Continuous data | 34 | ⊕⊝⊝⊝ | Not statistically significant |
| * The assumed risk (e.g. the mean control group risk across studies) is provided. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group. | ||||||
| GRADE Working Group grades of evidence | ||||||
| 1 Study with high risk of bias. All downgraded by three levels due to very serious imprecision. | ||||||
| Aromatherapy massage versus no massage for symptom relief in people with cancer | ||||||
| Patient or population: people with cancer Settings: oncology unit, cancer centre, hospice | ||||||
| Outcomes | Illustrative comparative risks* (95% CI) | Relative effect | No of participants | Quality of the evidence | Comments | |
| Assumed risk | Corresponding risk | |||||
| No massage | Aromatherapy massage | |||||
| Bodily pain (SF‐8) ‐ medium‐term (> 4 weeks to < 8 weeks) | The medium‐term mean bodily pain (SF‐8) in the control group was 44.4 points | The medium‐term mean bodily pain (SF‐8) in the intervention group was 5.30 higher | Continuous data | 86 | ⊕⊝⊝⊝ | Higher score indicates less pain |
| Bodily pain (SF‐8) ‐ long‐term (≥ 8 weeks) | The long‐term mean bodily pain (SF‐8) in the control group was 45.4 points | The long‐term mean bodily pain (SF‐8) in the intervention group was 3.80 higher | Continuous data | 86 | ⊕⊝⊝⊝ | Higher score indicates less pain |
| Anxiety (STAI‐state) | The mean anxiety (STAI‐state) ranged across control groups from 24.7 to 47.3 points | The mean anxiety (STAI‐state) in the intervention groups was 4.50 lower | Continuous data | 253 | ⊕⊝⊝⊝ | Lower score indicates less severity in anxiety |
| Anxiety (STAI‐state) subgroup: short‐term vs. medium‐term ‐ short‐term (≤ 4 weeks) | The short‐term mean anxiety (STAI‐state) in the control groups was 24.7 points | The short‐term mean anxiety (STAI‐state) in the intervention groups was 1.1 lower | Continuous data | 32 | ⊕⊝⊝⊝ | Not statistically significant |
| Anxiety (STAI‐state) subgroup: short‐term vs. medium‐term ‐ medium‐term (> 4 weeks to < 8 weeks) | The medium‐term mean anxiety (STAI‐state) in the control group was 47.3 points | The medium‐term mean anxiety (STAI‐state) in the intervention group was 5.1 lower | Continuous data | 221 | ⊕⊕⊕⊝ | Lower score indicates less severity in anxiety |
| Symptoms relating to the breast (EORTC QLQ‐BR23): long‐term (≥ 8 weeks) | The long‐term mean symptoms relating to the breast (EORTC QLQ‐BR23) in the control group was 31.9 points | The long‐term mean symptoms relating to the breast (EORTC QLQ‐BR23) in the intervention group was 9.80 lower | Continuous data | 86 | ⊕⊝⊝⊝ | Lower scores indicate fewer symptoms |
| Quality of life (MYMOP): medium‐term (> 4 weeks to < 8 weeks) | The medium‐term mean quality of life (MYMOP) in the control group was 3.9 points | The medium‐term mean quality of life (MYMOP) in the intervention group was 2.00 lower | Continuous data | 29 | ⊕⊝⊝⊝ | Lower score indicates greater quality of life |
| * The assumed risk (e.g. the mean control group risk across studies) is provided. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group. | ||||||
| GRADE Working Group grades of evidence | ||||||
| 1 Study with high risk of bias. Most except anxiety (four to eight weeks) downgraded by three levels due to very serious imprecision. | ||||||
| Study ID | Primary outcome | Secondary outcome | |||
| Pain | Psychological symptoms | Other physical symptoms | Quality of life | Adverse events | |
| Comparison 1. Massage vs. no massage | |||||
| ‐ | Anxiety‐STAI‐state Depression‐BDI | ‐ | ‐ | ‐ | |
| Pain‐VAS | ‐ | ‐ | ‐ | ‐ | |
| ‐ | Anxiety‐VAS | Nausea‐VAS | ‐ | ‐ | |
| ‐ | Anxiety‐VAS Anxiety‐STAI‐state | ‐ | ‐ | ‐ | |
| ‐ | ‐ | Fatigue‐POMS | ‐ | ‐ | |
| ‐ | Anxiety‐STAI‐state | ‐ | ‐ | ‐ | |
| ‐ | Anxiety‐STAI‐state Anxiety‐SCL‐90‐R | ‐ | ‐ | ‐ | |
| Pain‐PPI‐VAS | Mood‐VAS | ‐ | ‐ | Having physical distress Progress of disease | |
| ‐ | Depression‐PHQ Mood‐BSF | ‐ | ‐ | ‐ | |
| Pain‐VAS | Anxiety‐HAD Depression‐HAD Psychological distress‐RSCL | Physical symptom distress‐RSCL | Quality of Life‐RSCL | ‐ | |
| Pain‐VAS | Anxiety‐VAS | ‐ | Quality of Life‐McGill | ‐ | |
| Pain‐ESAS:AM | Anxiety‐ESAS:AM Depression‐ESAS:AM | Nausea‐ESAS:AM | ‐ | ‐ | |
| Pain‐PAT (or SNVR) | ‐ | ‐ | Quality of Life‐Graham | ‐ | |
| Comparison 2. Aromatherapy with massage vs. no massage | |||||
| Pain‐VAS | Anxiety‐VAS | Fatigue‐VAS | ‐ | ‐ | |
| Limb pain‐GBB | ‐ | Arm symptoms‐EORTC QLQ‐BR23 | ‐ | ‐ | |
| Pain‐VAS | Anxiety‐HAD Depression‐HAD Psychological distress‐RSCL | Physical symptom distress‐RSCL | Quality of Life‐RSCL | ‐ | |
| Pain‐VAS | Anxiety‐STAI‐state | ‐ | ‐ | ‐ | |
| ‐ | Mood disturbance scale‐POMS | ‐ | Quality of life‐MYMOP | Rash | |
| Pain‐EORTC | Anxiety‐STAI‐state | Fatigue‐EORTC | Quality of life‐EORTC | ‐ | |
| Comparison 3. Aromatherapy with massage vs. massage without aromatherapy | |||||
| Pain‐VAS | Anxiety‐HAD Depression‐HAD Psychological distress‐RSCL | Physical symptom distress‐RSCL | Quality of Life‐RSCL | ‐ | |
| ‐ | Anxiety‐STAI‐state (SAI) Anxiety‐STAI‐trait (TAI) Psychological distress‐RSCL | Physical symptom distress‐RSCL | Quality of life‐RSCL | ‐ | |
| Abbreviations: BDI: Beck Depression Inventory; BSF: Berlin Mood Questionnaire; CES‐D: Center for Epidemiological Studies Depression; EORTC QLQ‐BR23: European Organization of Research and Treatment of Cancer Quality of Life Questionnaire Breast Module (lower scores of the arm and breast symptoms indicate fewer symptoms); ESAS:AM: Edmonton Symptom Assessment System‐Ascites Modification; GBB: Giessen Complaints Inventory; HAD: Hospital Anxiety Depression Scale; MYMOP: Measure Yourself Medical Outcome Profile (lower score indicates greater quality of life, 7 = 'as bad as it could be' and 0 = 'as good as it could be'); PAT: Pain Assessment Tool; PHQ: Patient Health Questionnaire; POMS: Brief Profile of Mood States; PPI‐VAS: Present Pain Intensity‐Visual Analogue Scale (lower score indicates less pain, 10 = 'pain as bad as it could be' and 0 = 'no pain'); RSCL: Rotterdam Symptom Checklist; SCL‐90‐R: Symptom Checklist‐90‐R; SF‐8: Short‐Form Health Survey‐8 (higher score on bodily pain indicates less pain); SNVR: Skilled Nursing Visit Report form; STAI: State‐Trait Anxiety Inventory (lower score indicates less severity in anxiety); VAS: Visual Analogue Scale. | |||||
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Anxiety (STAI‐state) Show forest plot | 3 | 98 | Mean Difference (IV, Random, 95% CI) | ‐5.36 [‐16.06, 5.34] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Anxiety (STAI‐state) Show forest plot | 3 | Mean Difference (IV, Fixed, 95% CI) | Subtotals only | |
| 1.1 Children | 1 | 30 | Mean Difference (IV, Fixed, 95% CI) | ‐14.70 [‐19.33, ‐10.07] |
| 1.2 Adults | 2 | 68 | Mean Difference (IV, Fixed, 95% CI) | ‐0.74 [‐5.99, 4.51] |
| 2 Anxiety (STAI‐state) Show forest plot | 3 | Mean Difference (IV, Fixed, 95% CI) | Subtotals only | |
| 2.1 Short‐term (≤ 4 weeks) | 2 | 64 | Mean Difference (IV, Fixed, 95% CI) | ‐10.66 [‐14.72, ‐6.60] |
| 2.2 Medium‐term (> 4 weeks to < 8 weeks) | 1 | 34 | Mean Difference (IV, Fixed, 95% CI) | ‐3.0 [‐9.69, 3.69] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Anxiety (STAI‐state) Show forest plot | 2 | 253 | Mean Difference (IV, Fixed, 95% CI) | ‐4.50 [‐7.70, ‐1.30] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Anxiety (STAI‐state) Show forest plot | 2 | Mean Difference (IV, Fixed, 95% CI) | Totals not selected | |
| 1.1 Short‐term (≤ 4 weeks) | 1 | Mean Difference (IV, Fixed, 95% CI) | 0.0 [0.0, 0.0] | |
| 1.2 Medium‐term (> 4 weeks to < 8 weeks) | 1 | Mean Difference (IV, Fixed, 95% CI) | 0.0 [0.0, 0.0] | |
