Summary of main results
See Summary of findings for the main comparison.
Five trials measured agitation on three scales. Participants were less agitated in the aromatherapy group in two studies, Ballard 2002 and Lin 2007; the latter was a cross-over study only reporting overall data. Three other studies (Burns 2011; Cameron 2011; Fu 2013) found no difference in participants' levels of agitation, although Cameron 2011 did not report any actual data and Fu 2013 did not report data separately for each treatment group. The results from two studies that measured behavioural symptoms were highly heterogeneous, with Ballard 2002 showing an effect in favour of aromatherapy and Burns 2011 finding no treatment effect. O'Connor 2013 also found no difference in observed behaviour between aromatherapy and placebo. Two studies (Burns 2011; Fu 2013) showed no difference in adverse effects, and a single study (Burns 2011) showed no difference in quality of life and activities of daily living of participants treated with aromatherapy compared to those treated with placebo.
Overall completeness and applicability of evidence
Although seven studies are included in this review, we were very limited in the data we could use in analyses. Only two studies (Ballard 2002; Burns 2011) reported data on agitation and behaviour that could be used. Furthermore, in Ballard 2002 the participants that were included were taking a range of medication, including antipsychotics and neuroleptics; any of these could have been altered during the trial, which may have had a confounding effect on the results. In Burns 2011 participants were free of psychotropic medication (antipsychotics or cholinesterase inhibitors, or both) for at least two weeks before the study began, so any confounding effects of medication were less likely.
The length of follow-up was short in these two studies, only four to 12 weeks, and both were conducted in nursing homes in the UK, limiting the applicability of the results to other countries and settings.
It is unfortunate that we were unable to include data from the other five trials in the analysis. Cameron 2011 and Smallwood 2001 did not report any useable data, Fu 2013 only reported useable data for adverse events, and Lin 2007 and O'Connor 2013 were cross-over trials that only reported results from the two phases of the trial combined. The results of a cross-over trial are not usually considered suitable when studying an intervention for a progressive disease. Treatment effect is estimated from the within patient changes for the two periods and assumes that the patient's baseline assessment is the same at the start of each of the two phases. This assumption is difficult to justify for a progressive disease. Although Lin 2007 states there was no deterioration between the two phases, the data for each group are not presented separately. We would prefer to rely on the data from the first phase only, but these were not reported and have not been forthcoming from the author. Therefore this study did not contribute to the meta-analyses.
There are several essential oils currently being used that may not be comparable. Furthermore, the mode of administration differed between trials, although the two trials that had useable data both applied melissa oil topically. There is also great variation in assessment procedures and outcomes reported, and some have not controlled for other variables such as medication use. There is also the possibility that, as with antidementia drugs, aromatherapy might have different effects on people with different types or severity of dementia.
Quality of the evidence
The overall quality of the evidence, based on GRADE, is very low. Four out of the seven studies did not report any data that we could use in the analysis, and the three studies that did report data were small with short follow-ups. Ballard 2002 was a cluster-randomised trial and there might have been variation between the eight nursing homes included in the study that was not accounted for in the adjusted analysis, which could have confounded the results from this trial.
Potential biases in the review process
We tried to identify all relevant trials through our search, however it is possible that we may have failed to identify some studies.
Agreements and disagreements with other studies or reviews
Fung 2012 is a systematic review on the use of aromatherapy in treating behavioural problems in dementia. Fung 2012 reports that there is some evidence that aromatherapy has a positive effect on cognitive functioning and reducing BPSDs. However, although the review stated that they included only RCTs, six of the 11 included studies were not randomised and one was not testing aromatherapy, and so we have not included these studies in our review; accounting for the differences in our results.
The great majority of reported research on aromatherapy for people with dementia is of scientifically inadequate quality. In view of the possibility of a host of biases, no conclusions can be drawn from these studies. The reports can, however, give indications to inform researchers in the design of studies investigating aromatherapy. It is beyond the scope of this systematic review to give an account of all these studies, however, a representative selection that includes different study methodologies follows below.
Burleigh 1997 used an ABAB design to assess the effects of lavender, Roman chamomile, rosemary, and marjoram on the Behaviour Assessment Scale of Later Life (Brooker 1993) of seven participants with dementia. They found a significant reduction of challenging behaviour for four of the five female participants, but an increase of challenging behaviour for the two male participants. Six of the participants additionally showed a decreased need for assistance with activities of daily living. Curiously this is one of very few trials that adhere to one of the main principles of aromatherapy, the selection of oils based upon characteristics of each patient. Via this process different oils are often given for the same problems in different patients. West 1994 reported a single-case study of the effects of 'aromatic oils' on the sleeping pattern of a person with dementia. Their results showed an improvement of sleep patterns and a decrease of agitation. Wolfe 1996 assessed the effects of lavender and Roman chamomile on sleep patterns in two people with severe dementia who acted as their own controls. There was a mean increase of peaceful sleep for one participant, but a mean decrease of peaceful sleep for the other participant.
Kilstoff 1998 used the action research method, which is a qualitative method encouraging participants to design, implement, and evaluate an intervention, to assess the effects of lavender, mandarin, and geranium with hand massage on 16 recipients of day-care who had dementia. The findings indicated a perceived strengthening of the relationship between the people with dementia and their carers, and an improvement in feelings of health and well-being for both recipients and carers.
MacMahon 1998 reported a single-participant AB design study of the effects of an aromatherapy intervention, 'Zeal', on the motivational behaviour of one person with dementia. The results show a significant improvement on the rating scale employed. Vetrivanathan (reference unavailable) used The Brief Agitation Scale (Finkel 1993) and The Relaxation Checklist (Luiselli 1982) to evaluate the effects of lavender and massage on seven participants with dementia on an acute assessment ward. The results from this study showed some short term decrease of agitation, an increase in relaxation one hour after the intervention, but a decrease in relaxation before and immediately after the intervention.
Henry 1993 used a cross-over design with nine patients with dementia in a hospital ward to investigate the effects of lavender, using sleep charts as outcome measure. The results showed a significant increase in duration of sleep (P < 0.05).
Gray 2002 studied 13 older people with dementia living in residential care. All participants were described as being consistently resistive to medication administration, and had displayed an ability to perceive aromas. Each participant was exposed to three different aromas by means of a cotton ball taped to their clothing 20 minutes before medication was administered. The essential oils used were lavender, sweet orange, and tea tree, with no aroma as the control condition. Duration of medication administration and frequency and duration of resistive behaviour during this was used as the outcome, but no significant differences were found.
Bowles 2002 used a cross-over design to investigate the effects of lavender, sweet marjoram, patchouli, and vetiver on resistance to nursing care procedures and the frequency and duration of 'dementia-related behaviours' (aggression etc.). The participants were 56 (36 after attrition) aged care facility residents with moderate to severe dementia. The essential oils were blended into an aqueous cream and massaged onto the bodies and limbs of the residents five times a day, and behaviour was recorded throughout the eight weeks of the trial. The control condition was cream only. They found a significant decrease in 'dementia-related behaviours' occurring at times other than during nursing care, while resistance to nursing care increased for half of the participants. A significant improvement was also found on the Mini-Mental State Evaluation (Folstein 1975) for some participants.
Brooker 1997 reported on the effects of massage and lavender, separately and in conjunction with each other, on the disturbed behaviour of four people with dementia on a continuing care ward. The researchers developed individualized disturbed behaviour scales, which they tested for inter-rater reliability (P < 0.001). The results from this study showed a significant difference following the aromatherapy for one participant. For two participants, the massage and aromatherapy was associated with an increase of agitated behaviour.
Holmes 2002 investigated the effects of lavender on agitated behaviour in 15 patients with severe dementia on a long-stay psychogeriatric ward. This was a placebo-controlled trial with blinded ratings, with the participants acting as their own controls. The outcome measure used was the Pittsburgh Agitation Scale (Rosen 1994). The results show a significant mean improvement of the aromatherapy group, but five of the participants showed no change and one a worsening of agitated behaviour.
In finding some support for a beneficial effect of aromatherapy for people with dementia, the literature indicates that further, more adequate research is needed. Of equal importance are the findings of some adverse effects following aromatherapy, questioning the widespread assumption that it at least does no harm.