Dogs that were aggressive to strangers or too anxious to handle safely were excluded from the trial.

The two interventions were:

1. Placebo control
2. Pet Remedy

The Pet Remedy product was a commercially available preparation that includes a 5.37% essential oil blend (Valerian, Vetiver, Basil and Sage) in a volatile base, delivered via a diffuser.

The placebo used the same diffuser type containing only the volatile base.

Dogs were randomly allocated in which they experienced the two treatments to control for order effects (1. Placebo first, followed by Pet Remedy: n = 15; 2. Pet Remedy first, followed by Placebo: n = 13), with approximately 7 days between the dog receiving each treatment.

A power calculation was used to determine appropriate sample size. The statistical analysis was a multivariate regression analysis, with treatment, individual dog and order of treatment as fixed effects.

Experimental set up:

The test pen was a room novel to the dog (at the start of the study), that measured 3 x 3 m, and contained a bed, bowl, and diffuser. A camera allowed remote recording of the dog. The room had washable walls, floor and internal furniture and a built-in fan, with all being cleaned between subjects/tests and a two day period s to allow any diffuser residual smells to dissipate.

Testing procedure:

The on 30 minutes before the test subject (dog) entered. The dog was placed into the room alone for 30 minutes (the details of this process are missing). A camera filmed the dog’s behavior over the thirty minutes. The dog was then removed. Thus, the length of each treatment was 30 minutes, with approximately 7 days between treatments.

Data analysis:

The video footage was renamed by a naïve assistant so that the researcher was blind to the treatment the dog was receiving at the time. The researcher recorded the behaviours according to a pre-prepared ethogram of behaviours (see outcome measures).

The University of Exeter’s ethical review group approved the study.

1. Duration of time that the dogs spent performing each of the following behaviours:

• Autogrooming
• Digging
• Drinking
• Lying down
• Nosing
• Locomotion
• Hind legs (standing on hind legs only)
• Sitting
• Standing
• Circling
• Chewing
• Stretching
• (“Time spent standing on hind legs with front legs resting or digging against exit” Taylor and Madden, 2016)
• Wall bounce

2. Frequency with which each of the following behaviours was performed:

• Barking
• Howling
• Nose licking
• Paw lifting
• Sighing
• Whining
• Yawning
• Urination/defecation
• Panting

1. Duration of time behaviours performed:

• There was no effect of treatment on the duration of time with which each of the 14 behaviours recorded were performed. Although order of treatment exposure was built into the statistical model, the authors do not report this finding in the results.

2. Frequency with which behaviours performed:

• There was no effect of treatment on the frequency with which the nine behaviours recorded were performed. Dogs to sigh less (P = 0.052).

• The study was industry funded (by Unex Designs Ltd.)
• The inclusion criteria were based on subjective assessment by the owners, with no attempt to standardise level of ‘stress’ between dogs (or to control for it statistically).
• The treatment groups were not evenly allocated to control for order effects.
• The authors do not report whether the dogs were currently taking any medications or other supplements that might have an anti-stress/anxiolytic effect.
• Insufficient detail is provided with regards to how the dog was handled in the lead up to entering the room e.g. was the owner present immediately prior to testing commencing, was this standardised between dogs, etc.?
• Some of the behaviours measured are not clear measures of stress.

Each dog was used as its own control, with each dog exposed to four different essential oil scented cloths in order to investigate the effect of each odour on canine behavior within a kennel.

The experimental and control conditions that each dog was exposed to included:

1. Unscented cloth (control 1)
2. Coconut
3. Vanilla
4. Valerian^±
5. Ginger
6. No cloth (control 2)

^±A key herbal component of the Pet Remedy diffuser. Only the controls and the Valerian treatment findings will be reported in the main findings section.

Dogs were all simultaneously exposed to the same treatment condition. Each dog was observed for 2 hours per day (11:00–13:00, with shelter visitors/potential adopters viewing dogs for the second hour) for 3 consecutive days per treatment. Each dog was then given 2 days “wash out period” (no treatment provided) before being exposed to the next treatment condition.

A standardised procedure (5 drops were applied to each cloth – 1 per corner and one in the middle), with and handlers wearing latex gloves to prevent scent contamination during handling. Cloths were scented 60 mins before being placed in the centre of the dog’s run, and dogs were provided with the cloth 30 minutes before observations started.

The order in which treatments were applied was randomly determined (how is not defined), but the same order was used for each dog (because they were housed relatively close to each other and tested on the same days). The order of exposure was in the same order as the treatments (1–6) listed above.

Each dog’s behaviour was recorded using instantaneous scan sampling at 10 minutes intervals by an unblinded observer.

The study was conducted in accordance with the International Society for Applied Ethology’s ethical guidelines for the use of animals in applied animal behaviour research.

1. Standing
2. Sitting
3. Moving
4. Resting
5. Sleeping
6. Stereotyping
7. Vocalising
8. Located in the front half of the kennel

1. Standing
   • No significant effect of Valerian was found on the frequency of standing behaviour observed.

2. Sitting
   • No significant effect of Valerian was found on the frequency of sitting behaviour observed.

3. Moving
   • Dogs moved significantly less when exposed to the Valerian scent cloth, than when exposed to either the unscented cloth (z = -4.05, P < 0.001) or no cloth control (z = -8.34, P < 0.001).
   • The mean (± standard deviation) of observational points in which the dog was vocalising was:

Valerian: 1.73 (±2.25)

Unscented cloth: 5.80 (±4.41)

No cloth: 12.07 (±5.74)

4. Resting
   • Dogs rested significantly more when exposed to the Valerian scent cloth, than when exposed to the no cloth control (z = 4.20, P <0.001). No other pairwise combination was significantly different.
   • The mean (± standard deviation) of observational points in which the dog was resting was:

Valerian: 9.60 (±6.15)

Unscented cloth: 8.13 (±7.07)

No cloth: 4.47 (±4.17)

5. Sleeping
   • No significant effect of Valerian was found on the frequency of sleeping behaviour observed.

6. Stereotyping
   • Stereotypical behaviours were performed at such a low frequency that this behavioural parameter was not statistically analysed.

7. Vocalising
   • Dogs vocalised significantly less when exposed to the Valerian scent cloth, than when exposed to either the unscented cloth (z = -8.34, P < 0.001) or no cloth (Z = -4.66, P < 0.001) controls.
   • The mean (± standard deviation) of observational points in which the dog was vocalising was:

Valerian: 2.53 (±2.70)

Unscented cloth: 8.67 (±6.42)

No cloth: 13.87 (±8.84)

8. Located in the front half of the kennel
   • No significant effect of Valerian was found on the frequency of times the dog was observed to be in the front half of the kennel.

• The observer was not blinded to treatment.
• As the study aim was not specifically to look at the effect of Valerian on stress, the pre-existing stress levels of the dogs was not assessed.
• The dogs considered (by rescue centre employees) to be most susceptible to stress were excluded.
• All dogs received the same treatment at the same time and in the same order so the study is particularly sensitive to order effects and stochastic events that might influence dog behaviour during a particular treatment condition.
• None of the behaviours measured in the ethogram are indicators unique to stress.
• All of the behaviours that were significantly affected by Valerian could be used as measures of arousal more generally without reference to the emotional valence (positive or negative) per se.
• Limited range of behaviours observed.

1. Behavioural therapy + placebo
2. Behavioural therapy + Pet Remedy

On arrival dogs were given a behavioural assessment, the behaviour issue identified and a behavioural programme that was partly tailored to the individual dog was devised.

Dogs were allocated to the treatment group by the order in which they presented at the clinic (alternated between groups). The handler and trainer was blinded to which treatment group dogs were allocated to.

Each dog was assessed at four time points:

1. January (baseline measurement)
2. February
3. March
4. April

With data collected at time points 2–4 being compared to the baseline measurement for each dog.

Each dog, and handler, came to the Animal Behaviour Centre on several occasions each month to work with the trainer on a combination of set exercises, plus exercises specific to the individual dog. The owner also undertook training at home. It is not stated whether the number of centre visits, trainer(s) or at home training sessions were standardised between dogs. The set exercises are not fully defined, with only examples (hand feeding, door manners) of the type of exercises provided.

It is not clear when or how the dog was exposed to the intervention (Placebo or Pet Remedy, dependent upon the group the dog was in), but it appears to have been within the training centre environment.

It is not clear exactly when and how the dogs were scored (excitement/behavior) at each of the four time points.

1. Behaviour score – ordinal scale ranging from 0 (poor/unacceptable) to 5 (normal/good behaviour), with half-points allowed.
2. Excitement level – 0 (not excited) to 5 (very excited), with half-points allowed.

For both measures, the data was handled as follows:

• The difference between the score/level awarded in January (baseline) and each subsequent experimental month (individually) was calculated for each dog.
• This ‘differences’ data for each experimental month was compared statistically inter-treatment groups.

Other outcome measures:

1. Licking of paws – yes/no
2. Licking of bedding – yes/no
3. Suckling – yes/no
4. Paw over lead – yes/no
5. Jumping up – yes/no
6. Turning in circles – yes/no

1. Behaviour score

• The median (range) behaviour score for the placebo group was:

January: 1 (1–1)

February: 2 (1–3)

March: 3 (2–4)

April: 3.5 (2.5–4.5)

• The median (range) behaviour score for the Pet Remedy group was:

January: 1 ( 1–1)

February: 2.5 ( 1–3)

March: 3 (1.5–4)

April: 4 (2–5)

• A statistically significant difference was observed between the placebo and Pet Remedy treatment groups in February (P = 0.0011), March (P = 0.0070) and April (P = 0.0047)
• However, the 95% confidence interval associated with the change (compared to baseline) in behaviour level seen by each month was 0.0– -1.0

2. Excitement level

• The median (range) score for excitement for the placebo group was:

January: 4 (4–4)

February: 3 (1.5–4)

March: 2 (1–3)

April: 1.5 (0.5–3)

• The median (range) score for excitement for the Pet Remedy group was:

January: 4 (4–4)

February: 2.5 (1.5–4)

March: 2 (1–3)

April: 1 (0–2.5)

• A statistically significant difference was observed between the placebo and Pet Remedy treatment groups in February (P = 0.0036), March (P 0.0069) and April (P = 0.0009)
• However, the 95% confidence interval associated with the change (compared to baseline) in behaviour level seen by each month was 0.0– -1.0 for the difference between baseline and February and 0.0–0.5 for the January to March, and January to April time points

Other measures:

1. Licking of paws

• Findings not reported

2. Licking of bedding

• Findings not reported

3. Suckling

• Findings not reported

4. Paw over lead

• Findings not reported

5. Turning in circles

• Findings not reported

• The study was funded by the producers of Pet Remedy. The level of other involvement is not stated but a commercial statistical services company was hired by them to help design and analyse the study, and the disseminated report is produced by the statisticians hired to do so. There is no statement to outline the limit of their involvement in this study.
• The study has not been published in the scientific press and so was not subject to the accepted peer review process prior to its findings being reported publicly.
• There is insufficient information provided to allow this study to be replicated by independent researchers.
• There is insufficient information provided to allow the presence of, for example, confounding variables that might also explain the findings to be identified.
• The authors’ do not state the aim of the study until midway through the methods section. The aim is very broad; to see if behavioural therapy plus Pet Remedy is more effective than just behavioural therapy alone, but combined with the inclusion criteria this is very broad.
• The inclusion criteria are inadequately defined. There is no measure of severity of the behaviour problem, and, from the limited information provided, the behavioural problems were diverse. It is not clear from the description of the sample how many of these dogs were stressed (i.e. if the paper addresses the PICO).
• A negative control that did not include behavioural therapy would have been useful here.
• It is reported that sample size calculations were performed but not what the test was, what parameters were used to estimate minimum necessary sample size (e.g. power, effect size), or what that minimum size was. Thus, it is not known whether the eventual sample size was based upon this calculation.
• Dogs were assessed at the start of the study to establish what the problem was and what behavioural therapy was needed. A protocol was then devised that was partially tailored to the individual dog in terms of the training administered. However, insufficient detail is available here in terms of the nature of the exercises/therapy or the frequency with which they were carried out.
• It is not clear whether these dogs were involved in 1-2-1 training or in group training sessions. If the later, this may have reduced independence of each observation and raised questions about the true sampling unit.
• There is no information available on what the placebo was or how it was administered.
• There is no information on what Pet Remedy is or what formulation or method of dissemination was used (e.g. diffuser, or spray/topical or environmental application?).
• There is insufficient information on where the dogs were exposed to Pet Remedy/Placebo (training venue, home, both, etc.?) and whether this was standardised between dogs.
• Anecdotally, Pet Remedy has a distinctive aroma so it seems reasonable to assume that, despite blinding, the presence of the product would be detectable by the dog, owner, and behaviourist/trainer exposed to the product.
• The manufacturers claim that the product works on all animals. If this product had any effect on the individuals rating the dogs’ behaviour, was the perception affected by exposure to Pet Remedy? It is impossible to disentangle any effects of Pet Remedy on the rater with any effects on the dog’s behaviour as they are confounded.
• The authors report that data was collected for each dog at four time points (January/February/March/April) at monthly intervals, but it is not clear how these time points were separated temporally. Was there an equal number of days between time points, and was this the same for each dog?
• The primary outcome measures were both subjective measures. No attempt to reduce subjectivity (e.g. through detailed descriptors for each category) appears to have been undertaken.
• It is unclear who undertook the measurement of the primary outcome measures, or when in the session this occurred.
• There does not appear to be any attempt to assess inter-observer reliability (e.g. would two people award the same score to the same dog measured at the same time point), or intra-observer reliability (the latter would be more problematic to achieve though given the design).
• The primary outcome measures are very limited in their scope (e.g. what behaviours contribute to a global assessment of the dog’s excitement or behavioural levels?).
• The study reports also collecting proxy measures for stress but does not then report any analyses for these measures. It refers to these as objective, but in the same sentence refers to them as incidences of “excessive behavior”. “Excessive” suggests a subjective assessment, not an objective one.
• No detail is available as to how the proxy measures for stress were measured. For example, did the researchers use continuous sampling, instantaneous sampling or fixed time intervals (if the latter, at what time points), all occurrences or one–zero measures, etc. When were these measured? During centre behaviour sessions, at home, etc.?
• Neither of the primary outcome measures specifically measure stress, and the authors do not justify by reference to the wider literature the choice of outcome measures selected.
• It is difficult to comment fully on the suitability of the secondary measures either without more information on how/when these measurements were taken and how they addressed the study aim. However, the authors mention stress so if the aim was to measure stress it is not clear why the authors did not pick more universal or subtle measures of potential stress like yawning.
• It is not stated whether this study was reviewed by an ethics committee or whether informed consent was obtained from the clients.
• The researchers do not report whether the data was normally distributed or not, and present both parametric and non-parametric measures for the raw data. The inter-quartile range is not reported. The median is reported in this Knowledge Summary as this reflects the type of analytical statistics (non-parametric) that the authors undertook.
• The lower boundary for the 95% confidence interval for all differences data parameters reported was 0.0, which could mean that the true median value for the difference between the two groups was zero (or no difference). Nb. It is assumed median due to the values reported and the distribution around the median difference value, but the authors do not state this, and mean and standard deviation values are also reported. The researchers do not acknowledge this possibility in the summary of findings.