Clinical Nutrition xxx (2017) 1e10

Contents lists available at ScienceDirect

Clinical Nutrition
journal homepage: http://www.elsevier.com/locate/clnu

Original article

The association between adherence to a Mediterranean style diet and

cognition in older people: The impact of medication
Roy J. Hardman a, *, Denny Meyer a, Greg Kennedy a, Helen Macpherson b,
Andrew B. Scholey a, Andrew Pipingas a
a
Centre for Human Psychopharmacology, Swinburne University of Technology, Melbourne, Australia
b
Institute for Physical Activity and Nutrition, Deakin University, Geelong, Australia

a r t i c l e i n f o s u m m a r y

Article history: Background: Recent reviews indicate that adherence to a Mediterranean diet may be associated with
Received 12 April 2017 better cognitive functioning. In assessing these relationships in older individuals, previous studies have
Accepted 22 October 2017 not taken into account medication usage that may support or compromise cognitive functioning.
Objective: To investigate the association between adherence to a Mediterranean style diet, cognition and
Keywords: medication usage in cognitively healthy older individuals.
Health of ageing population
Design: Data were assessed from individuals aged 60e90 years (mean ¼ 77.8 years, SD ¼ 6.7) from 15
Impacts of diet in the elderly
independent living aged care villages around Melbourne, Australia. Participants' diets were assessed
using a food frequency questionnaire (FFQ). Cognition was assessed using reaction times from the
Swinburne University Computerised Cognitive Assessment Battery (SUCCAB). Prescribed medications
were recorded and analysed using binary measures. Cluster analyses were used to group participants in
terms of cognitive measures and medications taken. Analyses controlled for age, gender, average daily
kilojoule (kJ) intake and medication cluster.
Results: The relationship between cognitive speed clusters and medication clusters was significant (Chi-
squared ¼ 10.63, df ¼ 3, p ¼ 0.014). The odds ratio of 1.533 for average daily food intake suggested that
for each additional kilojoule of average daily intake, the odds of belonging to the slower reaction time
cluster increased by 53% and odds ratio of 0.573 for Mediterranean diet score suggested that for every
additional unit, the odds of belonging to the slower reaction time cluster declined by 43%. The rela-
tionship between Mediterranean diet score and cognition was only significant when medication use was
taken into account.
Conclusion: These data demonstrate that when medications are considered, a higher Mediterranean diet
score is associated with a faster response on cognitive function tests. The present findings also indicate
that it is pertinent to take into account medication use when investigating relationships between dietary
status and cognitive performance.
© 2017 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND
license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction contributed to the obesity epidemic in these countries [1]. Obesity

has been linked with impaired cognitive function and an elevated
Diet is considered to be one of the greatest contributors to risk of late-onset dementia, such as Alzheimer's disease [2,3].
overall health. The term ‘Western diet’ is used to describe a diet A Western diet and a sedentary lifestyle impacts general health,
containing large amounts of red meat, refined sugars, grains, and including increasing the rate of obesity, high blood pressure, high
high fat foods that is common in Western countries. The high levels blood triglycerides, high levels of LDL and low levels of HDL
of saturated fat and trans-fatty acids consumed in this diet have cholesterol and insulin resistance. These comorbidities have
resulted in an increased incidence of coronary artery disease, dia-
betes, chronic pain syndrome, inflammatory disease, cardiovascu-
lar disease, respiratory disease and end organ damage [4]. The
* Corresponding author. Faculty of Health, Arts and Design, Swinburne University comorbidities including high blood pressure, high cholesterol, in-
of Technology, Mail H98, Hawthorn, Victoria 3122, Australia.
E-mail address: rhardman@swin.edu.au (R.J. Hardman).
sulin resistance and obesity are classified as the metabolic

https://doi.org/10.1016/j.clnu.2017.10.015
0261-5614/© 2017 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/
).

Please cite this article in press as: Hardman RJ, et al., The association between adherence to a Mediterranean style diet and cognition in older
people: The impact of medication, Clinical Nutrition (2017), https://doi.org/10.1016/j.clnu.2017.10.015
2 R.J. Hardman et al. / Clinical Nutrition xxx (2017) 1e10

2. Methods
Abbreviations
2.1. Recruitment and approval
ANOVA analysis of variance
c2 test chi-squared test Participants were aged 60e90 years and living independently in
r coefficient of correlation samplme 15 aged care and retirement villages in and around Melbourne,
R coefficient of multiple regression Australia. Recruitment took place between 1 April 2014 and 30 June
CI confidence intervals 2015.
DF degrees of freedom
FFQ food frequency questionnaire 2.2. Eligibility criteria
kJ kilojoules
MedDiet Mediterranean diet Participants were fluent in written and spoken English. Partici-
MedDietS Mediterranean Diet Score pants had to obtain the approval of their medical practitioner to be
NS not significant involved in the trial.
n number of observations
SD standard deviation
2.3. Ineligibility criteria
SUCCAB Swinburne University Computerised Cognitive
Assessment Battery
Participants were unable to participate if they had a significant
F variance ratio
visual impairment, had a neurological or uncontrolled psychiatric
disorder, were unable to walk independently and safely, or used
illicit drugs or cognitive enhancing medications. Finally, those who
had suspected cognitive impairment (defined as a score <24 on the
syndrome, and can lead to cognitive impairment [5]. Metabolic Mini Mental State Examination) or depression (a score >9 on the
syndrome may impair cognition in the elderly, especially in those Geriatric Depression Scale) were also excluded.
with inflammation. Metabolic syndrome, particularly in ageing
populations, is treated pharmacologically, to reduce blood pressure 2.4. Ethical clearance
and cholesterol levels [6].
The Mediterranean diet (MedDiet) is considered a healthier This study was approved by the Swinburne University Human
alternative to the Western diet. The MedDiet is a diet with an Research Ethics Committee (project number 2013/057).
abundance of plant foods in the form of fruits, vegetables, breads, Selection criteria were followed in accordance with the pub-
other forms of cereals, beans, nuts, seeds, fish and olive oil as its lished protocol [13].
main source of monounsaturated fats; while dairy foods, red meat Figure 1 outlines the recruitment profile of the 105 participants
and chicken are consumed in lower quantities, and red wine is who entered the trial. Five participants left prior to initiation of any
consumed moderately with meals. These foods provide a high assessment. The 100 remaining participants, 28 males and 72 fe-
intake of b-carotene, vitamin C, tocopherols, omega-3 fatty acids, males, were assessed, of whom 93 responded to the question
various minerals and other beneficial substances such as poly- regarding their use of prescription medications. Among these 93
phenols and anthocyanins [7,8]. Adherence to the MedDiet was participants, 75% were prescribed more than two medications, 24%
initially considered relevant to health when it was demonstrated more than five medications, and 9% more than seven medications
that mortality was reduced within the populations located in the (mean 4.16, SD ¼ 2.17).
areas of Southern Europe [8,9].
Two recent systematic reviews have indicated that the MedDiet
may be neuroprotective as well as cognition-enhancing in the 2.5. Diet assessment
shorter term [10,11]. Additionally this diet is likely to be protective
against accelerated cognitive decline and the transition to mild Diet was assessed using the Cancer Council of Victoria Dietary
cognitive impairment and dementia. Medications taken by older Questionnaire (Food Frequency Questionnaire [FFQ] for Epidemio-
participants, and how medications may impact the protective ef- logical Studies Version 2 [DQES v2], November 2014) [14]. The
fects of a MedDiet had not been assessed. output was utilised to produce a Mediterranean diet score (Med-
Within a Western society the use of medications is part of the DietS) in accordance with Trichopoulou et al., 2003 [15]. The sex-
primary armamentarium to reduce the impact of comorbidities, specific median allows for a comparative cut-off to be made be-
such as the metabolic syndrome. To this end, in Australia, as in the tween genders on food consumption [15]. Beneficial foods, such as
rest of the developed world, there has been a substantial increase in vegetables, legumes, fruits, nuts, cereals, and fish, were assigned a
the use of blood pressure medications and statins to alleviate the value of 0 if a person's consumption was below the median, and a
effects of the metabolic syndrome. The use of medications and the score of 1 if it was equal to, or above, the median. Food components
impact of comorbidities may provide further understanding of the detrimental to health, such as meat, poultry, and dairy, consump-
effects of cognitive changes within an ageing population [12]. tion above the median was scored as 0, and intake below the me-
The primary aim of this study was to investigate the association dian was scored as 1. For alcohol, a score of 1 was given provided
of adherence to a Mediterranean style of diet with respect to consumption was within a specified range. When considering fat
cognition, while taking into consideration the use of medications. intake, the ratio of monounsaturated lipids to polyunsaturated
This study utilised the baseline data from a randomised controlled lipids was evaluated, with a higher ratio being more acceptable and
trial investigating the effects of the MedDiet on cognition in a score of 1 allocated accordingly. Kilojoules consumed was also
cognitively healthy older people living independently within aged reported within the FFQ. Thus, the total MedDietS ranged from
care facilities: the Lifestyle Intervention in Independent Living 0 (minimal adherence to the traditional MedDiet) to 9 (maximal
Aged Care (LIILAC) study [ACTRN12614001133628] [13]. adherence) [15].

Please cite this article in press as: Hardman RJ, et al., The association between adherence to a Mediterranean style diet and cognition in older
people: The impact of medication, Clinical Nutrition (2017), https://doi.org/10.1016/j.clnu.2017.10.015
 R.J. Hardman et al. / Clinical Nutrition xxx (2017) 1e10 3

Fig. 1. Profile of participants.

2.6. Cognitive assessment 2.8. Demographic and morphometric measures

Assessment of cognitive performance utilised the Swinburne Age, gender, education and smoking status were recorded as
University Computerised Cognitive Assessment Battery (SUCCAB). part of the study requirements. Height, weight, hip and waist
The SUCCAB is a validated computer-based cognitive battery con- circumference were measured and recorded.
sisting of eight measures that focus on the cognitive domains that
decline with increasing age and includes simple and choice reac- 2.9. Covariates and preliminary analysis
tion times, immediate and delayed recognition, congruent and
incongruent Stroop colour-words, spatial working memory and All food groups were calculated in grams per day and the total
contextual memory [16]. Speed of response to correctly performed daily kilojoules consumed per day was also calculated. Age (in
trials was recorded for each of these measures. The analysis was years) was calculated from self-reported birth date. Speed of
focussed on response time to correctly performed trials; we have response to each cognitive task was converted into a Z-score. Age,
argued previously that response time is a particularly sensitive sex (gender), and daily energy intake were the control variables.
measure of cognitive ageing [16]. Figure 2 shows the variable assessment process.

2.7. Medication assessment 2.10. Statistical analysis

All participants assessed at baseline had been evaluated as Cluster analyses were used to group 93 participants in terms of
cognitively healthy, and during screening, usage of prescription and their cognitive measures and medications taken. Clustering allows
non-prescription medications was recorded. The medication data the assignment of participants to groups with similar data patterns
were recorded using a binary (Yes/No) measure. for the variables considered. Clustering enables structures to be

Please cite this article in press as: Hardman RJ, et al., The association between adherence to a Mediterranean style diet and cognition in older
people: The impact of medication, Clinical Nutrition (2017), https://doi.org/10.1016/j.clnu.2017.10.015
4 R.J. Hardman et al. / Clinical Nutrition xxx (2017) 1e10

cognitive clusters, which were compared in terms of average daily

energy intake (kilojoules), MedDietS and age using ANOVA tests,
while a comparison in terms of gender was carried out using a
crosstab test.
Similarly, participants were clustered in terms of their pre-
scribed medications, again using Ward's method but allowing for
the binary nature of these data. The dendrogram suggested four
clear clusters (see Fig. 3). The relationship between the medication
and SUCCAB clusters was investigated using a crosstab test. Finally,
hierarchical binary logistic regression analysis was used to deter-
mine the importance of MedDiet as a determinant of the cognition
clusters, when controlling for medication cluster, age, gender, and
average daily energy intake. All analyses were conducted using
SPSS Statistics Version 23.

3. Results

This study investigated the cognitive capabilities of an Austra-

lian cohort living in independent living aged care in relation to their
adherence to a MedDiet and medication use.
Table 1 shows that the mean age of the combined group of
participants was 78 years, with a relatively high mean BMI of 28.6, a
high mean hip-to-waist ratio of 1.06, and a moderate level of
Fig. 2. The variable assessment stages. consumption of alcohol and fats [18].
The cluster analysis for the SUCCAB measures suggested two
clusters; the first containing 76 people with faster reaction times,
identified within the data and relies on discriminant analysis to
while the second contained 17 people with relatively slow reaction
check if the group's differences are statistically significant and if all
times (see Appendix).
variables significantly discriminate between groups [17].
The cognitive assessments for the two SUCCAB clusters are
Participants were clustered in terms of their cognitive outcomes,
depicted in Table 2. A significant difference was evident between
as measured by the SUCCAB reaction time Z-scores for correctly
the clusters for all the SUCCAB tasks. Mean performance accuracy is
performed trials, utilising Ward's method for hierarchical clus-
included in Table 2 for illustrative purposes.
tering. The resulting dendrogram suggested two clear SUCCAB
Table 3 shows the comparison of the SUCCAB clusters in terms of
age, total MedDietS, and average daily energy intake. These vari-
ables were used in subsequent analyses. Both clusters had a similar
average age and there was no significant difference in MedDietS
and daily energy intake (kJ) between the clusters. A cross tabulation
test (Chi-squared ¼ 4.65, df ¼ 1, p ¼ 0.031) showed a significantly
higher percentage of females in the slower reaction time cluster
(91%) than the faster reaction time cluster (67%).
Cluster analysis of medications prescribed, using binary data
suggested four clusters (see Fig. 3). Fewer medications were taken
by those in Cluster M1 than those in the other clusters (t(89) ¼ 7.82,
p < 0.001). Seventy-five percent of participants were using more
than two medications, and 24% were taking more than five
medications.
Details of the four clusters are explained in Table 4. The major
medication groups within each group are shown as percentages.
Cluster M1 was a relatively healthy group, of whom 41.4% took
blood pressure medication and 27.6% took omega-3 long chain fatty
acids. With 13.8% on blood thinners and 3.4% taking cholesterol
lowering drugs, it may suggest that general practitioners are taking
precautionary measures for this group.

Table 1
Characteristics of the sample.

Participants Mean (SD)

Age (years) 77.8 (6.73)

BMI (kg/cm) 28.6 (4.45)
Years of education 12.8 (3.52)
Blood pressure (sys/dias) 138/72 (16/10)
Alcohol g/day 116.78 (163)
Fig. 3. Medication clustering. *M1 (Relatively healthy), M2 (Reflux concerns) M3 (Bone
Mono/saturated fats (g/day) 1.01 (0.13)
concerns) and M4 (Cardio-compromised) are the respective clusters.

Please cite this article in press as: Hardman RJ, et al., The association between adherence to a Mediterranean style diet and cognition in older
people: The impact of medication, Clinical Nutrition (2017), https://doi.org/10.1016/j.clnu.2017.10.015
 R.J. Hardman et al. / Clinical Nutrition xxx (2017) 1e10 5

Table 2
Nonparametric (ManneWhitney U) test comparison of scores for cognition clusters.

SUCCAB task Mean reaction time Z-score (SD) Mean performance accuracy, % (SD)

Faster reaction Slower reaction Z-test Faster reaction Slower reaction Z-test
time (N ¼ 76) time (N ¼ 17) time (N ¼ 76) time (N ¼ 17)

Simple reaction time 0.24 (0.61) 1.09 (1.59) 4.80*** 100 (0) 100 (0) 0
Choice reaction time 0.25 (0.59) 1.26 (1.42) 5.01*** 95.1 (5.3) 97.6 (4.1) 2.10*
Immediate recognition 0.36 (0.62) 1.12 (1.13) 5.83*** 76.1 (9.9) 75.7 (10.4) 0.01
Delayed recognition 0.23 (0.82) 0.84 (0.93) 4.35*** 71.3 (10.6) 73.1 (12.8) 0.39
Congruent stroop 0.23 (0.86) 1.00 (1.07) 4.02*** 97.3 (4.3) 98.2 (2.6) 0.53
Incongruent stroop 0.12 (0.77) 0.57 (1.71) 2.321* 96.4 (5.6) 94.3 (11.4) 0.56
Spatial working memory 0.34 (0.71) 1.39 (0.99) 5.73*** 66.8 (9.6) 64.8 (7.9) 0.92
Contextual memory 0.19 (0.62) 0.74 (1.67) 3.08** 79.9 (14.4) 75.0 (19.2) 1.04

Note: *p < 0.05, **p < 0.01, ***p < 0.001.

Table 3
Comparison of SUCCAB clusters.

F stat p-Value Mean (SD) Faster reaction time Mean (SD) Slower reaction time
(N ¼ 76) (N ¼ 17)

Age 1.94 0.167 77.26 (7.25) 78.94 (6.02)

Total Mediterranean diet score 1.13 0.286 4.8 (1.53) 4.24 (1.79)
Energy intake, kJ (1000)/day 1.64 0.203 6.357 (1.876) 7.14 (2.91)

Table 4
Percentage of participants on each medication.

Medications Medication clusters

Relatively healthy, n ¼ 29 Reflux concerns, n ¼ 18 Bone concerns, n ¼ 18 Cardio-compromised, n ¼ 28 Overall, N ¼ 93

Angina 0.0 22.2 0.0 36.0 15

Antidepressants 10.3 0.0 5.6 25.0 11.8
Arrhythmias 0.0 27.8 22.2 10.7 12.9
Blood pressure 41.4 66.7 77.8 89.3 67.7
Blood thinners 13.8 27.8 33.3 53.6 32.2
Cholesterol drugs 3.4 38.9 33.3 96.4 44.1
Pain 6.9 72.2 11.1 2.1 19.4
Sleeping drugs 0.0 22.2 0.0 3.6 5.4
Ulcerations/reflux PPIs 13.8 83.3 16.7 17.9 29.0
Vitamin D 13.8 33.3 100.0 32.1 39.8
Vitamin B 6.9 38.9 16.7 0.0 12.9
Omega-3 fatty acids 27.6 16.7 44.4 0.0 20.4
Calcium 6.9 22.2 55.6 3.6 18.3
Glucosamine 6.9 5.6 33.3 3.6 10.8
CoQ10 0.0 0.0 16.7 3.6 4.3
Vitamin E 0.0 0.0 16.7 0.0 3.2
Magnesium 0.0 27.8 22.2 10.7 12.9
Women per group 19 15 16 15 65

Cluster M2, the group with reflux concerns had a moderate-to- indicating that they are potentially being treated for coronary heart
high number of blood pressure medications, at 66.7%, pain medi- disease and/or metabolic syndrome. Although the medications
cations, at 72.2%, and ulcer/reflux/protein pump inhibitors (83.3%). taken by this cluster suggest that this group is cardio-
The high prescription of anti-ulcer drugs (H2 agonists) and protein compromised, Fig. 4 shows that they demonstrate a much faster
pump inhibitors (PPIs) suggests that these patients are being cognitive reaction speed, which may be due in part to their medi-
treated for gastro-oesophageal reflux disease. This condition may cations potentially being cognitively protective.
lead to physical complications or symptoms that impact on well- The relationship between the cognitive speed clusters and
being and quality of life; hence the substantial pain medication medication clusters was found to be significant (Chi-
usage of 72.2%. squared ¼ 10.63, df ¼ 3, p ¼ 0.014). As shown in Fig. 4, overall 18.3%
Cluster M3, the group with bone concerns. This group of participants were assigned to the slow reaction time cluster, with
comprised mainly women, and had a moderate-to-high number of the lowest percentage (3.6%) of these from the cardio-
blood pressure medication (77.8%), omega-3 supplementation compromised medication cluster, a relatively low percentage
(44.4%), calcium supplements (55.6%), and vitamin D (100%). This (13.8%) from the relatively healthy medication cluster, rising to
was a very clear indication that they are being treated for osteo- 27.8% for the reflux concern cluster and 38.9% for the bone concern
penia or osteoporosis. medication cluster.
Cluster M4, the cardio-compromised group, who had the It is notable that Cluster M4, the group with the highest pro-
highest number of blood pressure medication, at 89.3%, blood portional use of blood pressure medications (angiotensin blocking
thinners, at 53.6%, and cholesterol lowering medication, at 96.4%, drugs), and highest use of statins and blood thinners is well

Please cite this article in press as: Hardman RJ, et al., The association between adherence to a Mediterranean style diet and cognition in older
people: The impact of medication, Clinical Nutrition (2017), https://doi.org/10.1016/j.clnu.2017.10.015
6 R.J. Hardman et al. / Clinical Nutrition xxx (2017) 1e10

Fig. 4. Medication Cross tabulation clusters demonstrating reaction times.

represented in the fastest reaction time cluster (96.4%), while the the MedDietS is trending to significance, average daily energy intake
relatively healthy cluster is less well represented in this cluster. is the only significant predictor of SUCCAB cluster at Stage 2.
Stage 3 analysis included the medication clusters. Table 7 shows
3.1. Binary logistic regression analysis for the SUCCAB clusters that when the medication clusters are included in the regression
with the Cardio-Compromised (M4) cluster as the reference cluster,
The use of binary logistic regression is appropriate for predicting average daily energy intake and MedDietS are significant predictors
cognitive clusters as only two SUCCAB clusters were used, enabling for the SUCCAB clusters. In Table 7, Cluster M4 (the cardio-
us to explain the relationship between the SUCCAB clusters and the compromised cluster) is regarded as the reference category as
other variables of interest. this was the fastest reaction time group and it allowed for ease of
The assessment is staged. The control variables entered at Stage interpretation. The odds ratio of 1.533 for average daily intake
1. The results given in Table 5 represent the control variables suggests that for every additional kilojoule, the odds of belonging to
entered at Stage 1 and demonstrate no significant difference be- the slower reaction time cluster increase by 53% on average when
tween the SUCCAB clusters in daily energy consumption, age and age, gender, MedDietS and medication cluster are controlled for.
gender. The odds ratio of 0.573 for the MedDietS suggests that for every
Stage 2 analyses are shown in Table 6, indicating that when the additional unit on the MedDietS the odds of belonging to the slower
MedDietS is included, the average daily energy intake is significant reaction time cluster decline by 43% on average when age, gender,
with respect to the cognition clusters. The odds ratio of 1.33 suggests average daily intake and medication cluster are controlled for. The
that for every additional kilojoule of average daily intake, the odds of odds ratio of 11.87 for the reflux concern cluster (M2) suggests that
belonging to the slower reaction time cluster increase by 33% on participants in this cluster are on average 11.87 times more likely
average when age, gender and MedDietS are controlled for. Although than participants in the cardio-compromised cluster to fall into the

Table 5
Stage 1 binary logistic regression analysis.

Step 1 variables B SE Wald (df ¼ 1) p-Value Odds ratio for slower 95% CI for odds ratio
cluster, Exp(B)
Lower limit Upper limit
3
Daily energy intake (kJ  10 ) 0.191 0.127 2.271 0.132 1.21 0.944 1.552
Age (years) 0.049 0.043 1.309 0.253 1.05 0.966 1.143
Male 1.497 0.809 3.422 0.064 0.224 0.046 1.093
Constant 6.297 3.562 3.126 0.077

Note: no control variables are significant at stage 1.

Table 6
Stage 2 binary logistic regression analysis.

Step 2 variables B SE Wald (df ¼ 1) p-Value Odds ratio for 95% CI for odds ratio
slower cluster
Lower limit Upper limit

Average daily energy intake (kJ  103) 0.285 0.14 4.168 0.041 1.33 1.011 1.749
Age 0.061 0.046 1.781 0.182 1.063 0.972 1.163
Male 1.4 0.82 2.914 0.088 0.247 0.049 1.231
MedDiet score 0.371 0.208 3.191 0.074 0.69 0.459 1.037
Constant 6.227 3.715 2.808 0.094

Please cite this article in press as: Hardman RJ, et al., The association between adherence to a Mediterranean style diet and cognition in older
people: The impact of medication, Clinical Nutrition (2017), https://doi.org/10.1016/j.clnu.2017.10.015
 R.J. Hardman et al. / Clinical Nutrition xxx (2017) 1e10 7

Table 7
Stage 3 binary logistics regression analysis.

Step 3 variables B SE Wald (df ¼ 1) p-Value Odds ratio for 95% CI for odds ratio
slower cluster
Lower limit Upper limit
3
Average daily energy intake (kJ  10 ) 0.427 0.166 6.626 0.010 1.533 1.107 2.121
Age 0.051 0.049 1.1 0.294 1.053 0.957 1.158
Male 0.837 0.893 0.879 0.349 0.433 0.075 2.492
Total Med score 0.557 0.244 5.204 0.023 0.573 0.355 0.925
Relatively healthy (M1) 1.527 1.22 1.566 0.211 4.605 0.421 50.351
Reflux concerns (M2) 2.474 1.253 3.896 0.048 11.865 1.017 138.378
Bone concerns (M3) 3.504 1.255 7.801 0.005 33.248 2.844 388.743
Constant 7.719 4.291 3.236 0.072 0.000

slower reaction time cluster when age, gender, average daily intake older individuals may also act to enhance, and preserve cognitive
and MedDietS are controlled for. Similarly, the odds ratio of 33.25 function in the longer term. Conversely medications may under-
for the bone concern cluster (M3) suggests that participants in this mine cognitive function and any benefits afforded by adherence to
cluster are on average 33.25 times more likely than participants in a healthier diet. Previous studies have not considered these po-
the cardio-compromised cluster (M4), to fall in the slower reaction tential confounding effects of medications on MedDietecognition
time cluster when these variables are controlled for. relationships.
The summary of these results is shown in Table 8, suggesting The aim of this study was to investigate the association between
that the effect of age and gender are not significant, while it appears adherence to a Mediterranean style diet and cognition in cognitive
that the effect of the MedDiet was suppressed until Stage 3 when intact older adults living independently, and to determine the
medications were controlled for. Interestingly if the medication impact of medication use on cognitive outcomes. Based on previous
cluster is not taken into consideration, we do not see a significant findings it was expected that those with higher adherence to a
effect of the MedDietS on cognitive outcomes. MedDiet would show faster reaction times on the cognitive tasks
The summary of results in Table 8 demonstrates that when the performed [23]. As the MedDietS values only showed a trend to
medication use is taken into consideration, a higher MedDietS is significance, we further investigated, potential confounding in-
associated with a faster response during the performance of fluences of medication use. It was postulated that the medications
cognitive tasks, particularly with respect to immediate recognition taken by the participants may influence their MedDietS, and sub-
and spatial working memory, which are the major discriminating sequently, their cognitive reaction time assessments. In fact,
variables for the two cognitive clusters. In addition, it was found without taking into consideration the impact of medications, we
that the cardio-compromised participants in Cluster 4, who were were unable to demonstrate the association between adherence to
taking higher levels of blood pressure and statin medications, are a MedDiet and cognitive outcomes.
more likely to have faster response times than participants in All the participants in the present study had been assessed as
medication clusters M2 and M3, the reflux concerns and bone cognitively healthy and were living independently in aged care and
concerns clusters, but similar response times to participants in retirement villages. At the time of assessment their potential un-
cluster M1, the relatively healthy cluster. The Hosmer Lemeshow derlying medical conditions had been managed by their medical
test indicates that the final model describes the data well (Chi- practitioner and where required, prescribed particular medications
square ¼ 4.44, df ¼ 8, p ¼ 0.816), confirming that medication cluster to normalise their conditions in order to control the underlying
needs to be accounted for when considering the relationship be- identified comorbidities.
tween MedDietS and cognition. A large proportion of the participants in our trial had been
prescribed various antihypertensive drugs such as, angiotensin
4. Discussion econverting enzyme inhibitors (ACE), diuretics (D), beta-blockers
(BB), and angiotensin receptor blockers (ARB). In the “cardio-
Previous research has suggested that adherence to a Mediter- compromised group,” 89% of participants were using an ACE, ARB,
ranean diet has the potential to improve cognition, reduce cogni- D or BB medications. Interestingly, this cardio-compromised group
tive impairment and cognitive decline and reduce the incidence of had the fastest response times as compared with the other medi-
dementia [19e23]. However, a high prevalence of medication use in cation clusters and were comparatively similar to the “Relatively
healthy” cluster. A recent network meta-analysis compared anti-
hypertensive drugs with respect to the incidence of dementia and
Table 8 cognitive function, demonstrating that all antihypertensive drugs
Summary of binary logistic regression analysis for the SUCCAB clusters. may to some extent be cognitively protective; in particular ARB's
Predictors Odds ratios for slower reaction showed significant benefits to cognitive functioning compared with
time cluster placebo [24]. Further studies need to be conducted to substantiate
Stage 1 Stage 2 Stage 3 our current results with regard to antihypertensive drugs and
3 cognitive performance.
Average daily energy intake (kJ  10 ) 1.21 1.33** 1.53*
Age 1.05 1.06 1.05 High levels of midlife total cholesterol have been associated
Gender 0.22 0.25 0.43 with compromised late life episodic memory, slower psychomotor
Mediterranean diet score 0.69 0.57* speed [25e27] and a higher risk of Alzheimer's disease (AD) [28]. It
Cardio-compromised cluster (M4) 1.00 would therefore be expected that the maintenance of lower
Relatively healthy cluster (M1) 4.61
Reflux concern cluster (M2) 11.87*
cholesterol levels through statin use (44% overall and 96% in the
Bone concern cluster (M3) 33.25** cardio-compromised group), would also confer ongoing benefits to
cognitive health [29,30]. However, a number of studies also indicate
Note: *p < 0.05, **p < 0.01, ***p < 0.001.

Please cite this article in press as: Hardman RJ, et al., The association between adherence to a Mediterranean style diet and cognition in older
people: The impact of medication, Clinical Nutrition (2017), https://doi.org/10.1016/j.clnu.2017.10.015
8 R.J. Hardman et al. / Clinical Nutrition xxx (2017) 1e10

that statin medications may themselves impact adversely on duration was not assessed; only type of medication was recorded
cognition, possibly through side effects including increased phys- given that this was a secondary outcome of the study. Future studies
ical and mental fatigue [31,32]. While prescribed medications are may consider more specific aspects of medication usage, including
critical for ongoing patient medical care, these other factors may act the class of drug, dosage and length of medication use.
to confound the understanding of dietecognition relationships.
Clearly more research is needed to better understand positive and 5. Summary and conclusion
negative impacts on cognitive health of both antihypertensive and
statins medications; dosage, duration and different drug classes. This study of a random selection of participants living inde-
A relatively large proportion (25%) of participants in the cardio- pendently within aged care facilities has demonstrated that those
compromised cluster were also taking antidepressant medications. who have a higher MedDietS perform faster on cognitive assess-
Research suggests that these medications have a significant posi- ments only when medication use has been taken into account.
tive effect on psychomotor speed and delayed memory recall [33] Previous research has suggested that medications act to normalise
and no adverse impact on the risk of cognitive decline leading to health and potentially stabilize cognitive function, however, no
dementia [34]. Pain medication usage was high within the reflux previous studies have taken medications into account when
group (72%) and may explain the slower reaction times observed considering the association between adherence to a MedDiet and
in this cluster. Previous studies that have evaluated pain experi- cognitive performance. Larger cross sectional and longitudinal
ences, such as radicular or neuropathic pain, reported a basic studies need to be conducted to properly evaluate this proposition.
slowing of reaction times. For example, slowing was observed in
relation to incongruent and congruent Stroop stimuli trials, irre-
spective of other comorbidities and pain medication [35,36]. Authors and contributors
Research within this area is complex and further focused cognitive
clinical research needs to be conducted to gain a clearer under- All authors made a substantial contribution to the conception
standing of pain and related pain therapy and relationships with and design of this research, and all authors have been involved in
cognition [36]. critically revising the work for important intellectual content. The
Overall this study has found that when medication use was final approval of the version to be published has been agreed by all
taken into consideration, that a higher adherence to a MedDiet is authors and an agreement to be accountable for all aspects of the
associated with faster cognitive reaction times. Interestingly the work in ensuring that questions related to the accuracy or integrity
cardio-compromised group was the cluster with the fastest of any part of the work have been appropriately investigated and
response times; suggesting that the medications also acted to resolved.
normalise cognitive function. However, there may be other in-
stances where medication use may have adversely affected cogni- Conflict of interest
tion; this premise requires further investigation.
The participants in this study were screened as cognitively The authors declare that the research was conducted in the
healthy e a key inclusion criterion for entry into the present study. absence of any commercial or financial relationships that could be
Through analysis of medication clusters it has become apparent construed as a potential conflict of interest.
that cognitive status may also depend on medication use and that
this is an important factor to take into account when considering Ethical approval and registration
the impact of diet on cognitive status. The alternative is to exclude
all individuals that are taking medications from similar studies. As Ethical approval was granted by Swinburne University of Tech-
observed in our sample, there is a very high prevalence of medi- nology Human Research Ethics Committee-SUHREC 2013/057. The
cation use in older, independently living Australians. Excluding trial was registered with the Australia and New Zealand Clinical
individuals taking medications would result in a sample that is not Trials Registry [ACTRN12614001133628]. The Universal trial num-
representative of the general population. Instead, we suggest that ber is U1111-1161-5364.
medication use be recorded and included in statistical analysis.
Moreover, we suggest that previous studies in older cohorts that
were not able to establish a relationship between MedDiet adher- Source of support/funding
ence and cognition may have been confounded by medication use.
There was no financial support for this research. The internal
4.1. Strengths and limitations resources and support was made through the Faculty of Health, Arts
and Design, School of Health and Centre for Human Psychophar-
A key strength of this study is that the population of participants macology, Swinburne University of Technology.
recruited were cognitively healthy and living in a homogenous
environment, yet the profile of medications was diverse. All par- Acknowledgements
ticipants were tested within their own independent living facilities
reducing potential confounding effects of travel time and assess- The Centre for Human Psychopharmacology, Swinburne Uni-
ment within an unfamiliar University setting. Age-sensitive com- versity of Technology, facilitated this research. We greatly appre-
puterised cognitive assessments were used to investigate dietary ciate the support of centre staff and their resources.
influences of a MedDiet, taking medications into account. Blood collection, analysis and storage were provided by Dor-
A limitation of the study is the disproportionate sampling of evitch Pathology in Melbourne. Access to participants was provided
women (70%) compared with men, a characteristic that is typical of by Australian Unity Ltd, BaptCare Ltd, Churches of Christ in
independent living and aged care facilities in Australia [37]. Gender Queensland Ltd, Villa Maria Catholic Homes Ltd. Health Care 2 U
was considered as a control variable in our statistical analysis but provided nutritional guidance and the preparation of a Mediterra-
did not significantly impact on relationships. Nevertheless, future nean diet cooking menu. Cobram Estate supplied extra virgin olive
studies should aim to achieve a gender balance in their cohorts. A oil to support the clinical trial. The LIILAC study team greatly
further limitation of the current study is that medication dosage and appreciate the support of these organisations.

Please cite this article in press as: Hardman RJ, et al., The association between adherence to a Mediterranean style diet and cognition in older
people: The impact of medication, Clinical Nutrition (2017), https://doi.org/10.1016/j.clnu.2017.10.015
 R.J. Hardman et al. / Clinical Nutrition xxx (2017) 1e10 9

Appendix. Reaction time dendrogram for the SUCCAB longitudinal and prospective trials. Front Nutr 2016;3. https://doi.org/
10.3389/fnut.2016.00022. Article 22.
showing two clear clusters
[11] Petersson SD, Philippou E. Mediterranean diet, cognitive function, and de-
mentia: a systematic review of the evidence. Adv Nutr 2016;7(5):889e904.
https://doi.org/10.3945/an.116.012138.
[12] Crichton GE, Bryan J, Hodgson JM, Murphy KJ. Mediterranean diet adherence
and self-reported psychological functioning in an Australian sample. Appetite
2013;70:53e9. https://doi.org/10.1016/j.appet.2013.06.088.
[13] Hardman RJ, Kennedy G, Macpherson H, Scholey AB, Pipingas A. A randomised
controlled trial investigating the effects of Mediterranean diet and aerobic
exercise on cognition in cognitively healthy older people living independently
within aged care facilities: the Lifestyle Intervention in Independent Living
Aged Care (LIILAC) study protocol [ACTRN12614001133628]. Nutr J 2015:14.
https://doi.org/10.1186/s12937-015-0042-z.
[14] Victoria CCo. Internet: http://www.cancervic.org.au/research/epidemiology/
nutritional_assessment_services1.
[15] Trichopoulou A, Costacou T, Bamia C, Trichopoulos D. Adherence to a Medi-
terranean diet and survival in a Greek population. N Engl J Med 2003;348(26):
2599e608. https://doi.org/10.1056/NEJMoa025039.
[16] Pipingas A, Harris E, Tournier E, King R, Kras M, Stough CK. Assessing the
efficacy of nutraceutical interventions on cognitive functioning in the elderly.
Curr Top Nutraceutical Res 2010;8(2e3):79e87.
[17] Murtagh F, Legendre P. Ward's hierarchical agglomerative clustering method:
which algorithms implement ward's criterion? J Classif 2014;31(3):274e95.
https://doi.org/10.1007/s00357-014-9161-z.
[18] Magarey AM, Daniels LA, Boulton TJC. Prevalence of overweight and obesity in
Australian children and adolescents: reassessment of 1985 and 1995 data
against new standard international definitions. Med J Aust 2001;174(11):
561e4.
[19] Tsivgoulis G, Judd S, Letter AJ, Alexandrov AV, Howard G, Nahab F, et al.
Adherence to a Mediterranean diet and risk of incident cognitive impairment.
Neurology 2013;80(18):1684e92.
[20] Ye X, Scott T, Gao X, Maras JE, Bakun PJ, Tucker KL. Mediterranean diet,
healthy eating index 2005, and cognitive function in middle-aged and older
Puerto Rican adults. J Acad Nutr Dietetics 2013;113(2):276e81.
[21] Gardener SL, Rainey-Smith SR, Barnes MB, Sohrabi HR, Weinborn M, Lim YY,
et al. Dietary patterns and cognitive decline in an Australian study of ageing.
Mol Psychiatry 2015;20(7):860e6. https://doi.org/10.1038/mp.2014.79.
[22] Valls-Pedret C, Sala-Vila A, Serra-Mir M, Corella D, de la Torre R, Martinez-
Gonzalez MA, et al. Mediterranean diet and age-related cognitive decline: a
randomized clinical trial. JAMA Intern Med 2015;175(7):1094e103. https://
doi.org/10.1001/jamainternmed.2015.1668.
[23] Hardman RJ, Kennedy G, Macpherson H, Scholey AB, Pipingas A. Adherence to
a Mediterranean-style diet and effects on cognition in adults: a qualitative
evaluation and systematic review of longitudinal and prospective trials. Front
Nutr 2016;3(22). https://doi.org/10.3389/fnut.2016.00022.
[24] Marpillat NL, Macquin-Mavier I, Tropeano AI, Bachoud-Levi AC, Maison P.
Antihypertensive classes, cognitive decline and incidence of dementia: a
network meta-analysis. J Hypertens 2013;31(6):1073e82. https://doi.org/
References
10.1097/HJH.0b013e3283603f53.
[25] Solomon A, Kareholt I, Ngandu T, Wolozin B, MacDonald SWS, Winblad B, et al.
[1] Zobel EH, Hansen TW, Rossing P, von Scholten BJ. Global changes in food Serum total cholesterol, statins and cognition in non-demented elderly. Neurobiol
supply and the obesity epidemic. Curr Obes Rep 2016;5(4):449e55. https:// Aging 2009;30(6):1006e9. https://doi.org/10.1016/j.neurobiolaging.2007.09.012.
doi.org/10.1007/s13679-016-0233-8. [26] Fratiglioni L, Mangialasche F, Qiu C. Brain aging: lessons from community
[2] Gotsis E, Anagnostis P, Mariolis A, Vlachou A, Katsiki N, Karagiannis A. Health studies. Nutr Rev 2010;68(12):S119e27. https://doi.org/10.1111/j.1753-
benefits of the Mediterranean diet: an update of research over the last 5 years. 4887.2010.00353.x.
Angiology 2015;66(4):304e18. https://doi.org/10.1177/0003319714532169. [27] Ma CR, Yin ZX, Zhu PF, Luo JS, Shi XM, Gao X. Blood cholesterol in late-life and
[3] Fitzpatrick AL, Kuller LH, Lopez OL, Diehr P, O'Meara ES, Longstreth WT, et al. cognitive decline: a longitudinal study of the Chinese elderly. Mol Neuro-
Midlife and late-life obesity and the risk of dementia cardiovascular health degener 2017:12. https://doi.org/10.1186/s13024-017-0167-y.
study. Arch Neurol 2009;66(3):336e42. [28] Rantanen KK, Strandberg AY, Pitkala K, Tilvis R, Salomaa V, Strandberg TE.
[4] Mattson MP, Maudsley S, Martin B. A neural signaling triumvirate that in- Cholesterol in midlife increases the risk of Alzheimer's disease during an up to
fluences ageing and age-related disease: insulin/IGF-1, BDNF and serotonin. 43-year follow-up. Eur Geriatr Med 2014;5(6):390e3. https://doi.org/
Ageing Res Rev 2004;3(4):445e64. https://doi.org/10.1016/j.arr.2004.08.001. 10.1016/j.eurger.2014.05.002.
[5] Babio N, Toledo E, Estruch R, Ros E, Martinez-Gonzalez MA, Castaner O, et al. [29] Barone E, Di Domenico F, Butterfield DA. Statins more than cholesterol
Mediterranean diets and metabolic syndrome status in the PREDIMED ran- lowering agents in Alzheimer disease: their pleiotropic functions as potential
domized trial. Can Med Assoc J 2014;186(17):E649e57. https://doi.org/ therapeutic targets. Biochem Pharmacol 2014;88(4):605e16. https://doi.org/
10.1503/cmaj.140764. 10.1016/j.bcp.2013.10.030.
[6] Starr JM, McGurn B, Whiteman M, Pattie A, Whalley LJ, Deary IJ. Life long [30] Samaras K, Brodaty H, Sachdev PS. Does statin use cause memory decline in
changes in cognitive ability are associated with prescribed medications in old the elderly? Trends Cardiovasc Med 2016;26(6):550e65. https://doi.org/
age. Int J Geriatr Psychiatry 2004;19(4):327e32. https://doi.org/10.1002/ 10.1016/j.tcm.2016.03.009.
gps.1093. [31] Gnjidic D, Naganathan V, Ben Freedman S, Beer CE, McLachlan AJ,
[7] Trichopoulou A, Kyrozis A, Rossi M, Katsoulis M, Trichopoulos D, La Vecchia C, Figtree GA, et al. Statin therapy and cognition in older people: what is the
et al. Mediterranean diet and cognitive decline over time in an elderly Med- evidence? Curr Clin Pharmacol 2015;10(3):185e93. https://doi.org/
iterranean population. Eur J Nutr 2015;54(8):1311e21. https://doi.org/ 10.2174/157488471003150820152249.
10.1007/s00394-014-0811-z. [32] Chatterjee S, Krishnamoorthy P, Ranjan P, Roy A, Chakraborty A,
[8] Feart C, Samieri C, Alles B, Barberger-Gateau P. Potential benefits of adherence Sabharwal MS, et al. Statins and cognitive function: an updated review. Curr
to the Mediterranean diet on cognitive health. Proc Nutr Soc 2013;72(1): Cardiol Rep 2015;17(2). https://doi.org/10.1007/s11886-014-0559-3.
140e52. https://doi.org/10.1017/s0029665112002959. [33] Rosenblat JD, Kakar R, McIntyre RS. The cognitive effects of antidepressants in
[9] Keys A, Menotti A, Karvonen MJ, Aravanis C, Blackburn H, Buzina R, et al. The major depressive disorder: a systematic review and meta-analysis of ran-
diet and 15-year death rate in the 7 countries study. Am J Epidemiol domized clinical trials. Int J Neuropsychopharmacol 2016;19(2). https://
1986;124(6):903e15. doi.org/10.1093/ijnp/pyv082.
[10] Hardman Roy J, Kennedy G, Macpherson Helen, Scholey Andrew B, [34] Bali V, Holmes HM, Johnson ML, Chen H, Fleming ML, Aparasu RR. Compar-
Pipingas Andrew. Adherence to a Mediterranean-style diet and effects on ative effectiveness of second-generation antidepressants in reducing the risk
cognition in adults: a qualitative evaluation and systematic review of

Please cite this article in press as: Hardman RJ, et al., The association between adherence to a Mediterranean style diet and cognition in older
people: The impact of medication, Clinical Nutrition (2017), https://doi.org/10.1016/j.clnu.2017.10.015
10 R.J. Hardman et al. / Clinical Nutrition xxx (2017) 1e10

of dementia in elderly nursing home residents with depression. Pharmaco- [36] Ferreira KD, Oliver GZ, Thomaz DC, Teixeira CT, Foss MP. Cognitive deficits in
therapy 2016;36(1):38e48. https://doi.org/10.1002/phar.1680. chronic pain patients, in a brief screening test, are independent of comor-
[35] Hu KS, Fan ZW, He SC. Uncovering the interaction between empathetic pain bidities and medication use. Arq Neuropsiquiatr 2016;74(5):361e6. https://
and cognition. Psychol Res Psychol Forsch 2015;79(6):1054e63. https:// doi.org/10.1590/0004-282x20160071.
doi.org/10.1007/s00426-014-0634-9. [37] Australian Bureau of Statistics AG.

Please cite this article in press as: Hardman RJ, et al., The association between adherence to a Mediterranean style diet and cognition in older
people: The impact of medication, Clinical Nutrition (2017), https://doi.org/10.1016/j.clnu.2017.10.015