Anxious dogs: n = 16 (later reduced to 14, possibly 13¹), split between two groups (experimental diet versus control diet).

Non-anxious dogs: n = 16, split between two groups. 1 group (n = 8) received the experimental diet; 1 group received the control diet (n = 8).

¹ The authors reporting of this aspect is weak and they report eliminating 3 anxious dogs, but then claim later in the same section to have reduced the anxious treatment group from 16 dogs to 14 dogs. They do not report how this affected numbers of dogs in each group, therefore, the minimum sample size for one of the diet groups could be as low as 6 (if 2 are missing) or 5 (if 3 are missing).

1. Anxious dogs fed the control diet
2. Anxious dogs fed the experimental diet
3. Non-anxious dogs fed the control diet
4. Non-anxious dogs fed the experimental diet

The two diet treatments compared were:

1. Experimental diet: a complete dry dog food that was coated with caseinate hydrolysate (CH) (the milk component containing alpha-casozepine). The authors do not report the inclusion rate of CH.
2. Control diet: the same diet as the experimental diet but without the CH.

Researchers were blinded to which dogs were receiving which treatment.

Experimental timeline:

Forty dogs were assessed for anxiety and either included in the study (n = 32) or rejected (n = 8).

Anxiety assessment used two methods:

1. The Evaluation Scale of Anxiety, and
2. The Reactivity Evaluation Form (REF) (more details below).

Based on the combined scores, dogs were ranked for level of anxiety. The top 16 dogs were then assigned to the anxious group and the bottom 16 dogs were assigned to the non-anxious group. The intermediary 8 dogs were not used.

Within group, the dogs were paired-housed such that dogs with the most similar scores were housed together (e.g. the most anxious dog was housed with the second most anxious dog, etc).

A randomised block design was then used to allocate diet treatments to the dogs. Dogs were blocked according to whether they were anxious or non-anxious (based on the anxiety assessment methods outlined above), and then the 8 pairs of dogs within each block were randomly allocated to receive one of the two diet treatments. Both dogs within the pair received the same diet.

The dogs included were allocated as described above and then pair-housed. Dogs were introduced to their new kennel mate and given 3 days to adjust to this. Then, the study commenced.

The study lasted either 68 or 69 days (not clear enough to say for certain).

1. Baseline period (6 days): baseline parameters were measured for serum cortisol, serum lysozyme, the neutrophil:lymphocyte ratio, and mean heart rate. Heart rate was measured telemetrically using a heart rate monitor strapped to the dog’s chest and recorded at 5 second intervals. Dogs were fed their normal diet during this period.

2. Initial evaluation phase (T1) (2 or 3 days, this is unclear from the paper^*): on the first day, the REF was completed and a blood sample taken. On day two, the mean heart rate was determined for ten minutes while the dogs were video recorded for 10 minutes. This video was later viewed by an observer blinded to treatment and group and the frequency or proportion of time spent performing certain behaviours recorded.

^*In the experimental phase section of the paper, the authors report “after an initial 3 day evaluation phase (T1)” and “each evaluation (T1, T2, and T3) lasted 2 days”.

3. Feeding phase 1 (28 days). Dogs were given ad libitum access to the food with food replenished once a day (PM), with both dogs in the pen fed the same diet.

4. Second evaluation phase (T2, lasted 2 days). Data collection as documented in T1.

5. Feed phase 2 (28 days). Dogs fed as documented in feeding phase 1.

6. Third evaluation phase (T3, lasted 2 days). Data collection as documented in T1.

Initial assessment of anxiety (to select dogs):

1. Evaluation scale of anxiety: a subjective assessment of anxiety in which dogs were observed and scored from 1 (low) – 6 (high).
2. A two part objective behavioural assessment, comprising an ordinal scale of anxiety level (score 1 – 4) and the presence of various clinical signs of anxiety (e.g. lick lip: yes / no, score 1 point if yes) based on the dogs response to handler presence outside and inside the test pen.

Both assessment tools were developed by the researchers of this study.

Two experienced handlers familiar with the dogs independently assessed the dogs using these two assessment methods and then agreed the final score per dog.

1. REF
2. Proportion of time spent performing various behaviours in the test pen over 10 minutes (exploration; locomotion, passive behaviour, orientated to the environment, scratching, oral behaviour, vocalisation, play, panting, grooming)
3. Frequency of behaviours performed while in the test pen for 10 minutes (yawning, lip – licking, elimination, drinking, eating)

Physiological parameters:

1. Heart rate
2. Cortisol
3. Neutrophil: lymphocyte ratio
4. Lysozyme

It is not clear whether plasma or serum cortisol and lysozyme was measured as the authors report the terms interchangeably.

No significant differences were found between the baseline period data and the initial evaluation phase (T1) data, so the authors used T1 as their baseline data period.

Behavioural outcome measures:

1. REF:

• At the start (T1), anxious dogs had significantly higher REF scores than non-anxious dogs (Mann-Whitney U Test, P < 0.001).
• At the start (T1), there was no significant difference between anxious dogs fed the control diet and anxious dogs fed the experimental diet.
• Dogs on the experimental diet had significantly lower REF scores by T3 (end of the study) (Friedman Test, P < 0.01).
• No further formal statistical comparisons are reported. The figure associated with this outcome measure suggests that anxious dogs fed both the control and experiment diets always scored higher than non-anxious dogs fed either diet. It also suggests that there is no divergence in anxiety scores of non-anxious dogs fed either diet.

2. Behaviour:

• Anxious dogs and non-anxious dogs show no significant differences in behaviour at the start of the study (T1).
• Anxious dogs showed less exploratory behaviour at the end of the study (T3) compared to the beginning (T1) (Friedman, P < 0.05). There was no change in level of exploratory behaviour between time points in anxious dogs fed the control diet.
• Non-anxious dogs fed the control diet showed less scratching behaviour over time (Friedman, P < 0.05)
• Anxious dogs fed the control diet showed increased lip licking over time (Friedman, P <0.05).
• Anxious dogs showed a significant increase in orientation to the environment behaviour at T3, compared with T1 (Friedman, P < 0.05).

Physiological outcome measures:

1. Heart rate:

• No significant effects were observed.

2. Cortisol:

• Remained within the normal physiological range for anxious and non-anxious dogs during the whole study
• Cortisol level of anxious and non-anxious dogs did not differ at the start (T1).
• Anxious dogs fed the experimental diet showed a significant decrease in cortisol over time (Friedman, P < 0.05). Post hoc testing indicated that a significant reduction was observed between T1 and T2 (Wilcoxon, P < 0.05).
• Non-anxious dogs showed a tendency to have a reduction in cortisol over time (Friedman, P < 0.10).

3. Neutrophil:lymphocyte ratio:

• A significant reduction in neutrophils (and increase in lymphocytes) was observed over time in anxious dogs (General Linear Model, P < 0.05) but not non-anxious dogs. Diet treatment had no effect on the ratio for either group of dogs.

4. Lysozyme:

 No PICO relevant significant effects were observed.

The authors fail to report the inclusion rate for caseinate hydrolysate (containing alpha-casozepine) in the diet (g/kg/Dry matter).

The dogs were fed ad libitum during the study so the dose rate would have varied between dogs.

Laboratory population in which the breeding life experiences of the dogs were homogenised and tightly controlled. This contrasts to a pet population and may limit extrapolation of the findings to the wider population of dogs.

The authors undertook a discriminant analysis to retrospectively (after all the data was collected) eliminate dogs whose REF score did not accurately predict which dogs were identified as anxious through the Evaluation Scale of Anxiety (or vice versa, it is not clear which). As a consequence, 2 or 3 anxious dogs were removed. The authors fail to report how this affected the sample size of anxious dogs in the control diet and experimental diet groups respectively. Given the low sample sizes within each group (n = 8), this has implications for the interpretation of any findings or extrapolation to the wider population. This is compounded by the wide degree of variation observed (where standard deviations are reported). Retrospective removal of such a large proportion of the sample size for a group should also be questioned.

The authors fail to report the measure of central tendency used to report the REF and cortisol data, and do not provide any measure of variation around this central tendency.

The authors do not provide enough information to interpret the cortisol findings. They report that anxious and non-anxious dogs’ cortisol levels did not differ at T1, but do not report the findings for the subgroups relevant to the treatments (especially anxious dogs fed the experimental diet versus anxious dogs fed the control diet). Eyeballing the figure presented suggests that these two sub-groups might vary at the start and combining this data would have increased the variation associated with the anxious group data and made it more difficult to pick out significant differences. However, the lack of error bars makes it difficult to draw many inferences here.

The authors report appears to use the convention of using P < 0.01 (tendency) P < 0.05 (significant), P < 0.01 and P < 0.001 (highly significant). It would have been useful to see the exact P value associated with each finding as most appear to be P < 0.05. The way that they have analysed the behavioural data that they have reported suggests that there may have been many statistical tests performed and this increases the likelihood of a type 1 error (finding a significant finding where one does not exist).

The ethogram derived behavioural data is very problematic to interpret:

• The authors report the findings using mean (± standard deviation) which implies the data is normally distributed. However, they analyse this data using non parametric tests (Friedman test, with post-hoc testing via the Wilcoxon test) which implies the data is not normally distributed. The error bars suggest that there was a large amount of variation associated with the performance of some behaviours.
• The authors do not provide enough information for readers to be able to draw their own conclusions about what the behavioural data may indicate. It would have been useful to see the descriptive statistics reported for all the ethogram data, broken down by both treatment diet and anxiety group.
• The authors appear to have cherry picked which behavioural outcome comparisons to report and key comparisons (e.g. between anxious dogs fed the control diet and anxious dogs fed the experimental diet) are largely missing.

 It is not immediately clear that changes between some of the behaviours in the ethogram would represent a change in the level of anxiety.

Pet dogs with anxiety-related disorders

The inclusion criteria was as follows:

• At least 3 months old
• Weigh between 1.5 – 42kg
• Diagnosed with a behavioural complaint related to anxiety
• At the first visit, the dog scored above 19 (out of a maximum of 45) on the Emotional Disorders Evaluation in dogs scale (EDED) (as used by Pageat, 1995).

Dogs were excluded if:

• The problem had been present for less than four weeks
• There was any evidence that the problem was caused by disease, illness or injury.
• Any psychotropic medications had been administered in the previous two weeks
• If the dog scored less than 20 on the EDED scale.

Two treatments (19 dogs per treatment) were compared:

1. Behavioural modification therapy (BMP) plus oral alpha-casozepine (15mg/kg/24hrs)
2. BMP plus oral selegiline hydrochloride (0.5mg/kg/24hrs)

Study design:

• Multi-centre trial that utilised 7 certified veterinary behaviourists.
• Double or triple blind (not clear which) methodology in which the owner, veterinary behaviourist and the supervisors (undefined, researchers?) were unaware which dog received which drug while the study was ongoing.
• Dogs were allocated to treatments using a randomised block design (blocked according to a veterinary behaviourist treating the dog, then randomly allocated to receive a treatment based on a pre-defined randomisation list e.g. dog 1 was given drug 1, dog 2 was given drug 2, and so on).
• The BMP was tailored to the dog, diagnosis given and how severe this problem was. Thus, it varied between dogs but BMP approach did not vary systematically between the two treatment groups.
• The trial lasted 57 days.

Experimental timeline:

• Day 0: Inclusion visit, owners completed the EDED questionnaire. Dogs that scored greater than 19 were included.
• Day 14: Follow up phone call.
• Day 28: Follow up physical visit.
• Day 42: Follow up phone call.
• Day 56: Follow up physical visit.

At each contact, the owner was asked to complete the EDED scale again AND provide a subjective assessment of improvement on a scale of -10 to +10, with 0 indicating no change, and values higher and lower than this indicating, respectively, an improvement or worsening of the dog’s behavior.

The EDED score was previously developed by one of the authors for use in investigating the ability of seligiline to reduce anxiety (Pageat, 1995, cited  by the current authors), and uses a range of questions to assess the dog’s emotional state on a scale from 9-45. A score of 9-13 is considered to be normal, 14-18 indicative of a dog with a phobia, with dogs that have a score that sits between 18

-30 considered to have an anxiety-related disorder. The authors do not define what a score above 30 means in relation to the dog’s emotional state. They do not state whether this scale is validated for this use.

• EDED score
• Owner subjective assessment of improvement

The authors also introduced a binomial outcome (success / failure of treatment) “success” measurement. This was based upon the dog achieving two things:

1. An EDED score of < 20, and
2. An owner subjective assessment of improvement score of ≥ +6.

EDED scores:

• The average EDED score for both treatment groups decreased between visits 1 (day 0) and visit 3 (day 56) (alpha-casozepine group + BMP: F(4, 18) = 22.20, P < 0.0001; selegiline + BMP: F(4,18) = 26.33, P < 0.0001).
• At the start of the study (visit one), there was no significant difference between the two treatment groups. The average EDED scores on visit 1 were: alpha-casozepine + BMP: 25.0; selegiline + BMP: 24.2
• At the end of the study (visit 3, day 56), there was no significant difference between the two treatment groups. The average EDED scores on visit 3 were: alpha-casozepine + BMP: 26.3; selegiline + BMP: 16.6.

Owner subjective assessment of improvement:

• The data analysed using an ANOVA statistical test are not reported in a clear enough way to precisely define what the researchers think they found.
• The average improvement score of the apha-casozepine + BMP group increased from 2.7 on day 14 to 5.3 on day 56.
• The average improvement score of the selegiline + BMP group increased from 3.0 on day 14 to 5.4 on day 56.
• At the end of the trial (day 56) the improvement score did not significantly differ between the two treatment groups (Mann-whitney U test, P = 0.73)

Success / failure measurement:

• There was no significant association between treatment received and proportion of dogs that responded successfully to the treatment (χ² test, d.f. = 1, P = 0.74).
• The number of successes (failures) in the two groups was: alpha-casozepine + BMP: 10 (9); selegiline + BMP: 9 (10).

There was no BMP only treatment group and / or BMP + placebo group.

The paper cited by the authors (Pageat, 1995) is in French so it is not possible to use this paper to better understand the usage of the EDED scale in relation to assessment of selegiline.

The authors do not make clear what they mean by average. Were they presenting medians or means?  They do not report any measures of variation (e.g. standard deviation, inter-quartile range, standard error of the mean).

The authors switch between using parametric and non-parametric tests to study aspects of the same data. They do not report any information to indicate that they considered the underlying distribution of the data.

Statistical output reported for the between treatment group pairwise comparisons at visit 1 (day 0) and visit 3 (day 56) is unclear as only one P value is reported (P = 0.79) for both results.

Technically, the authors cannot make the claim reported in the EDED scores section (that scores decreased between visits one and three for each treatment group) because they only appear to have carried out an ANOVA. The degrees of freedom indicate that they fitted 5 time points (presumably physical visits 1-3, and telephone calls 1-2). However, no post hoc test has been carried out to identify between which two (or more) pairs of time points the significant difference(s) arises between. Eyeballing the graphical data suggests that a significant difference was most likely between visit 1 and 3, but this is complicated by the lack of variation reported. This criticism also applies the ANOVA analysed aspect of the owner subjective assessment of improvement scores.

The reporting of the ANOVA analysis for the owner subjective assessment of improvement scores is not clear enough. It is assumed that they are comparing four time points (physical visits 2-3, telephone visits 1 -2 because of the degrees of freedom reported) but they do not include enough information.

The claims that are made throughout the research paper are misleading and introduce a potential interpretation bias on the part of the reader with limited research methodology knowledge. The authors consistently refer to the two groups as alpha-casozepine and selegiline, but make limited mention of the fact that these dogs all also received a behaviour modification programme, with no mention at all in the results or conclusion (there is no discussion section). Despite this major confounding variable, the authors make statements like:

“both compounds were equally efficacious” (results), “both products were efficient to decrease the EDED score” (abstract), “due to this efficacy…. Alpha-casozepine (Zylkene) should be considered an option by the veterinary surgeon for the biological management of anxiety” (abstract), etc.

These claims cannot be made given the lack of an adequate control group as each compound could be equally efficacious at having no impact on anxiety (i.e. the BMP was the successful component).

Dogs were excluded if they:

• Were pregnant
• Less than one year of age
• Had had a medical condition (i.e. not necessarily current)
• Were a recipient of behavioural therapy or medications aimed at treating behavioural problems (i.e. psycho-active drugs)

• 4 dogs were lost to follow up during the study; the others withdrew for reasons unconnected to the trial (health issues of the dogs or personal circumstances of the owner).

1. Control diet (Royal Canin Select Skin Care Vets Plan)
2. Experimental diet containing elevated levels of the amino acid tryptophan and alpha-casozepine (Royal Canin Calm Canine)

• The alpha-casozepine inclusion rate was 1.35g/kg dry matter (DM), to provide dogs with a dose of approximately 20mg/kg/day (the recommended dose for dogs under 15kg).
• The tryptophan content of the experimental diet was 3.04g/kg DM (the control diet contained 2.83g/kg DM).
• The diets were not identical in all other respects. All the nutritional components reported differed in inclusion rate.
• The dogs were retained in their home environment throughout the study and fed by their owner (who was blinded to the treatment they were feeding their dog).

The experimental timeline is as follows:

• The owners of potential canine participants completed a C-BARQ questionnaire. Dogs that scored as anxious (n = 44) were included in the trial.
• Dogs received the control diet for 8 weeks. They were fed twice a day. Data collection took place in week 7 (see below for outcome measures) and comprised completion of the C - BARQ questionnaire to assess behavioural indices of anxiety and two urine samples collected to assess urinary cortisol levels. One sample was taken in the home environment (baseline, unstressed measurement) and one sample was taken 2hr post vet clinic visit (the ‘stressed’ measurement).
• Dogs then received a 1 week wash out period. The diet fed during this period is not stated and is assumed to be the dog’s normal diet.
• Dogs then received the experimental diet for 8 weeks. All other details are the same as for the control diet phase.

Further details:

• The wash out period length was determined by a prior preliminary experiment using high-pressure liquid chromatography which demonstrated that the dietary additives would be entirely removed from the dogs systems within one week.
• Cortisol analysis was carried out in duplicate.
• Urine collection was carried out at the same point of day to reduce the risk of time of day effects on cortisol production.

The outcome measures studied were all potentially relevant to answering this PICO. The outcome measures included:

• Urinary cortisol:creatinine ratio (UCCR)
  • Baseline measurement: taken in the dog’s home environment
  • Post-stressor measurement: taken 2hrs post vet clinic visit
• Severity of stranger – directed aggression
• Severity of owner – directed aggression
• Severity of stranger – directed fear
• Severity of non-social fear
• Level of touch sensitivity

All, bar UCCR, were assessed using specific questions within the C-BARQ questionnaire. This questionnaire asks owners a series of questions to assess each parameter, and owners are asked to score their dog on a scale from 0 (no signs of the behavior) to 4 (signs are severe). The scores for each parameter’s questions (e.g. for stranger-directed aggression there are 10 related questions) are then averaged, to provide a mean score for that parameter.

The C-BARQ questionnaire is a widely used, validated method of assessing canine behavior in a range of ways (not specifically anxiety).

UCCR:

• Baseline UCCR of the dogs was not significantly different between treatments.
• Post stressor UCCR was significantly different to baseline measurement, irrespective of treatment (i.e. the dogs were stressed by the veterinary visit) (P < 0.01).
• The Mean (±standard error of the mean, S.E.M.) UCCR of the dogs post-stressor was significantly lower (P = 0.04) when they were being fed the experimental diet (32.17 ± 0.21 x 10-6) than when they were being fed the control diet (39.61 ± 0.3-6).

Questionnaire assessment of the dogs’ anxiety-related behaviours:

• There was a significant reduction in stranger-directed aggression when the dogs were being fed the experimental diet (t = 2.44, d.f. = 15, P = 0.014). The mean (± S.E.M.) scores were: control: 1.25 (± 0.22); experimental: 0.85 (± 0.20).
• There was no effect of diet on owner-directed aggression. The mean (± S.E.M.) scores were: control: 0.47 (± 0.07); experimental: 0.43 (± 0.07).
• There was a significant reduction in stranger-directed fear when the dogs were being fed the experimental diet (t = 3.08, d.f. = 22, P = 0.014). The mean (± S.E.M.) scores were: control: 1.51 (± 0.22); experimental: 1.15 (± 0.14).
• There was a significant reduction in non-social fear when the dogs were being fed the experimental diet (t = 2.00, d.f. = 17, P = 0.031). The mean (± S.E.M.) scores were: control: 1.44 (± 0.22); experimental: 1.20 (± 0.23).
• There was a significant reduction in touch sensitivity when the dogs were fed the experimental diet (t = 3.56, d.f. = 25, P < 0.001). The mean (± S.E.M.) scores were: control: 1.34 (± 0.14); experimental: 1.01 (± 0.12).

The study was partially funded by Royal Canin. The authors do not state whether Royal Canin had any input into the experimental design.

The study was not specifically designed to study the effects of alpha-casozepine on anxiety. Rather, it was set up to study the effects of a commercial diet that contained alpha-casozepine and increased levels of tryptophan. Thus, there was a complete confounding variable when using this study to address the PICO question.

The diets used were not identical in all other respects, other than the dietary additives. This was sufficient for a commercial aim (to show the diet was beneficial) but limited interpretation of the effects of any specific additive.

The study did not randomise the order in which the dogs received the treatments. Therefore, the study is at risk of order effect biases. It is not clear why this was not done as the order that the diets was presented to the dogs was under the control of the experimenters and the experimenters provide no explanation for why this approach was taken.

The authors refer to preliminary data to establish that dietary additives to the diet would be fully eliminated from the body within one week, in order to justify the one week washout period. However, it is unclear what dietary additives they are referring to as the experimental diet containing the additional dietary additive (alpha-casozepine) or increased levels of the dietary additive tryptophan was presented to all dogs in the second (post wash out period) phase of the study. 

Cortisol is a non-specific measure of arousal, rather than a specific biomarker for stress. Therefore, it is possible that the dogs were excited or stimulated by a trip to the vets, rather than fearful. The authors fail to determine what the dogs’ normal behavioural response is to a trip to the vets.

The C-BARQ questionnaire assessment of the dogs to determine which dogs were included in the study provides insufficient detail.  

1. The C-BARQ questionnaire used is not a questionnaire that has been validated to anxiety per se. Several of the categories are specifically designed to assess fear, but some are not. Or particular relevance here is the aggression-related categories. Not all aggression is derived from anxiety.
2. The definition of an ‘anxious dog’ (i.e. that meets the inclusion criteria) was very loosely defined. Dogs were only excluded if, on all the questions asked (n = 32¹), the owner scored the dog as a 0 (no sign) to 1 (mild signs observed). Thus, a dog could be included if the owner scored it as a 2 (somewhere between mild and moderate) on just one of the questions.

¹Assumes mutual exclusivity of the questions. The authors do not clarify whether one question assesses more than one parameter simultaneously.