Neurocognitive effects a combined polyphenolic-rich herbal extract in healthy middle-aged adults – a randomised, double-blind, placebo-controlled study

ABSTRACT

Objective: This study assessed whether polyphenolic rich supplement containing Bacopa monnieri (BM: 300 mg), Panax quinquefolius ginseng (PQ: 100 mg) and whole coffee fruit extract (WCFE: 100 mg) could enhance cognitive performance, affect and cerebral-cortical activation over 28-days of intervention.

Method: A randomised, double-blind, placebo-controlled, between-group study of 52 healthy adults between 35 and 65 years (M = 50.20, SD = 9.37) was conducted. Measures of cognition, affect and brain activity were measured at three time points: baseline, 28 days post intervention and 14 days post washout. At each time point, haemodynamic response in the prefrontal cortex (PFC) was measured using functional near-infrared spectroscopy (fNIRS), and serum brain-derived neurotrophic factor (BDNF).

Results: The polyphenolic-rich supplement reliably improved positive affect and delayed recall compared to placebo following 28 days of supplementation. For the brain, those in the active condition showed greater PFC activation on performance of the 2-back tasks post supplementation compared to placebo (p < .05, d = 0.6).

Discussion: This is the first report of a 28-day supplement intervention and 2-week follow-up study to assess changes in affect, cognition, cerebral haemodynamic response and BDNF in healthy middle-aged adults. The potential synergistic effects of polyphenolic compounds on neurocognitive function in middle-aged adults through emotional-cognitive processing and cognitive reserve are important for promoting brain and cognitive health.

KEYWORDS:

• Polyphenols
• brain
• fNIRS
• affect
• Bacopa
• Ginseng
• coffee-fruit
• compassion

1. Introduction

There is increasing focus on the role of nutritional, herbal and nutraceutical compounds on mental and cognitive functioning. Recent acute studies demonstrate promising effects of bioactive polyphenolic compounds, such as cocoa, citrus, resveratrol, caffeine, chocolate, green tea, gingko biloba, berries (to name a few), on neurophysiological and cognitive outcomes [1]. However, the potential synergistic effects and interactions of complex pharmacological compounds on biological systems underpinning cognitive performance remain mostly unknown.

Experimental and longitudinal studies demonstrate numerous biological effects of active polyphenolic compounds such as glucoregulation, anti-inflammatory and antioxidant effects, as well as neuronal preservation, neurogenesis and increasing cerebral blood flow [2]. As a group, polyphenol-rich products consistently increase cerebral blood flow, and modulate brain activity after single, acute doses and chronic supplementation [3]. The mechanism through which specific polyphenolic components, for example flavonoids or anthocyanins, are associated with neuromodulatory effects may stem from numerous mechanisms of action related to antioxidant activity, microbiome changes within the gut-brain axis, and potential neuromodulator effects of brain-derived neurotrophic factor (BDNF: [4]).

Whilst there are numerous studies highlighting cognitive task-specific benefits of singular extract or compound supplementation, there are unknown pharmacodynamic effects and complex dose and time effects on cognitive performance. Little is known about how polyphenol-rich compounds can affect cognitive performance in concert. Preliminary research [5] showed a potential synergistic effect of a single administration of a polyphenolic-rich combined formulation of 300 mg Bacopa, 100 mg American Ginseng and 100 mg Whole Coffee Fruit Extract on working memory tasks in healthy adults, mean age 34 years. The results showed improvements in working memory performance concomitant with lower prefrontal cortex activation, ascertained using functional near-infrared spectroscopy (fNIRS) over a 2-hour period compared to placebo, at doses (of the individual compound) lower than previous research to produce effects on working memory However, the effects from longer-term intake in a group of middle-aged adults, an age in which the impact of supplementation may be more pronounced due to cognitive resources, remains unexplored. The challenge to current research is to understand whether compounds may produce synergistic effects to benefit cognitive performance and if specific quantities are needed for these effects to occur.

Consequently, this study aimed to determine the neurocognitive effects of the combined formulation across mood and cognitively demanding working memory tasks following 4 weeks of supplementation compared to placebo. We examined brain activation related to cognitive task performance to understand cerebral activation of the PFC via fNIRS during task performance and explored the effects of supplementation on BDNF concentrations.

2. Materials and methods

2.1. Design

This study was a between-subjects double-blind, randomised controlled study operating at a single centre. Participants were given either an active supplement (commercially available as CopaPrime+™) or a placebo for 4 weeks, followed by a 2-week washout period. Participants attended the University on three separate occasions, at weeks one, four and six.

2.2. Participant and recruitment

Sixty healthy adults (46 females, 14 males) aged between 35 and 65 years (M = 50.20, SD = 9.37) enrolled in the study, with 52 participants included in the final analysis (see consort statement in Figure 1). Inclusion criteria included age, attendance at three specified time-points, no history or diagnosis of a heart disease, diabetes, head injury, stroke or neurological conditions. If participants reported taking medications, these were low-dose and preventative medication-use for typical population ailments (e.g. every day and consistent medication for blood pressure, thyroid, cholesterol, hormone replacement therapy, or antidepressants). Participants were recruited via university newsletters, flyers, paid Facebook advertisements and other social media outlets, and word of mouth. Participants contacted researchers via email, phone, Facebook Messenger or an online expression of interest form. At the completion of testing sessions, participants were given a $50 gift card as an honorarium towards travel and parking costs.

Figure 1. CONSORT flow diagram of the participant progress through a randomised trial of two groups.

This study was conducted according to the guidelines laid down in the Declaration of Helsinki and all procedures involving participants were approved by the Human Research Ethics Committee of Central Queensland University (HREC: 21386, date 24/3/2021). Participants provided written consent. The study was registered with the Australian New Zealand Clinical Trials Registry (ACTRN12620001331921). Figure 1 shows the CONSORT participant flow diagram dictating the participant numbers from recruitment and enrolment through to analysis.

2.3. Procedure

One day prior to the first appointment, participants completed an online baseline questionnaire (demographic and dietary measure detailed below). On testing days, 2 h before their visit, participants were instructed to fast and abstain from consuming any stimulants (e.g. tea, coffee or caffeine-containing products). Compliance was assessed through self-report before the commencement of session. Participants also completed the Positive Affect Negative Affect Scale (PANAS: [6]) to determine mood at the beginning of the session for the day of testing. Prior to commencing the cognitive tasks, participants completed a glucose reading via a finger prick to ensure performance was not influenced by blood glucose levels. Participants were excluded at the time of testing if they had a glucose reading higher than 6.5 mmol/L or lower than 4.8 mmol/L (confirmed by secondary measures after waiting 20–30 min). To minimise the time-of-day effects on performance, participants were scheduled for repeat assessments at the same time of day across all appointment visits. Days between testing sessions were also recorded.

Each study day comprised a practice module for baseline testing session on mood and cognitive measures to ensure familiarisation with the tasks. Participants’ PFC activity was measured during the cognitive tests via fNIRS and participants’ tolerability of the testing was monitored throughout the session. There were no reports of any adverse effects or intolerance to the study procedures or supplements during the study.

2.4. Measures

2.4.1. Demographic measures

Participants were asked to provide self-reported age, height, weight, years in education, occupation and hours worked per week, self-rated health (Likert scale from 1 = poor, to 5 = excellent), hours spent exercising per week, medications and or dietary supplements and vitamins consumed on a regular basis, and alcohol (number of standard drinks per week) and cigarette consumption.

2.4.2. Dietary measure

The Dietary Screening Tool (DST) is a 20-item questionnaire that asks participants to provide an estimate of intake frequency of specific commonly eaten foods based on a seven-day period [7]. The DST has been used to explore the association between diet quality and mood disorders in an Australian population [8]. For each food item a number between 0 and 8 is assigned. The total sum of scores (range 0–105) indicates diet quality, where higher scores indicate higher dietary quality.

2.4.3. Mood measures

2.4.3.1. Depression anxiety and stress scale (DASS)

The 21-item version of the DASS [9] was used to assess three negative affective states of depression, anxiety and stress. Participants indicated on a four-point scale whether each statement, such as ‘I found it hard to wind down’ applied to them from 0 (did not apply to not at all), to 3 (applied to me very much, or most of the time). Scores are summed for each affective state.

2.4.3.2. Positive affect negative affect scale (PANAS)

The PANAS [6] consists of 20 affective descriptors (e.g. ‘enthusiastic’, ‘jittery’) that participants rate on a 5-point scale from 1 (very slightly or not at all) to 5 (extremely) the extent to which they have experienced the described affective state during the past week. Participants completed the measure prior to each testing day reflecting on the week prior to capture any subjective changes in the participants’ experience over the course of the study. On testing days, a researched variation of ‘today’ was used to identify participants’ mood at the start of testing, which could potentially influence cognitive performance on the day and between-group differences.

2.4.3.3. Adult prosocialness scale

his 16-item scale [10] assessed behaviours and feelings linked to sharing, helping, care-taking, and feeling empathy towards others. Participants rate the items on a 5-point scale ranging from 1 (never/almost never true) to 5 (always/almost always true), with a total summed score.

2.4.3.4. Social safeness and pleasure scale

This 11-item scale measured how people experienced pleasure and positive feelings in social situations [11]. Participants were asked to rate how they feel on a 5-point scale ranging, 1 (almost never) to 5 (almost all the time), in various situations. For example, ‘I feel a sense of warmth in my relationships with people’. Total score ranges from 11 to 55, with higher scores indicating feeling more socially safe.

2.4.3.5. Fears of compassion scale

This 30-item scale measures participants’ beliefs and thoughts in regard to kindness and compassion in three areas of their life [12]. Participants were asked to rate on a 5-point scale the degree to which they agree with each statement from 0 (don’t agree at all) to 4 (completely agree). Fear of compassion for others comprises 10-items (e.g. ‘people will take advantage of me if they see me as too compassionate’). Fear of compassion from others comprises 13-items (e.g. ‘wanting others to be kind to oneself is a weakness’). Fear of compassion for self-comprises 15-items (e.g. ‘I feel that I don’t deserve to be kind and forgiving to myself’). Higher scores in all three fear of compassion scales indicate higher fears of compassion. This scale provides an insight into relating to others and self.

2.4.3.6. Tolerability

Tolerability was assessed via follow-up phone call 1-week post commencement of supplementation to support and assess adherence, and for monitoring any experience of adverse events through open questions such as ‘How have you been feeling since starting the supplement?’

2.4.4. Cognitive measures

The cognitive test battery was created and conducted using Eprime3 (Psychology Software Tools, Pittsburgh, PA) to assess working memory (n-back tasks), selective attention and inhibition (Stroop, Go/No-go), and verbal working memory and delayed recall (RAVLT), which have been shown to be sensitive to fNIRS responses in healthy adults. To ensure participants’ familiarity with the n-back tasks, participants were presented the 1-back task first followed by the RAVLT word lists (randomised versions across participants), followed by the remaining working memory tasks, randomised alternate versions to mitigate task order effects, ending with the 30 min delayed verbal recall trial and word recognition task of the RAVLT. To accommodate for fNIRS measurement during cognitive testing, a block design paradigm was used, such that all cognitive tests had four ‘active’ blocks lasting at least 30 s, with rest blocks, lasting 25–30 s, spaced in between. The task battery was approximately 45 min per participant.

2.4.4.1. N-back tasks

Single letters comprised of letters ‘A’ to ‘J’ are presented on the computer screen. Each active work block (total of four work blocks presented) consisted of 15 letters presented on the screen for 500 msec with 1500 msec separating the presentation of each letter. When participants saw a correct match with the letter previously presented before the stimulus letter (i.e. 1-back, 2-back or 3-back, with a correct match appearing 25% of the time), they were instructed to press the spacebar, otherwise, no key was to be pressed. Response time, accuracy (% correct) and error scores were obtained for each work block task and averaged to give an overall score and response time for 1-back, 2-back and 3-back tasks.

2.4.4.2. Stroop colour-word task

For the Stroop task, the task was adapted to present in block design four colour words (blue, yellow, red and green). In the congruent condition, the words yellow, blue, green or red were presented to participants and were written in their correct respective colours. When presented with a stimulus word, participants had to respond by pressing the correct arrow key (keys were labelled with coloured stickers of the four separate colours). For the incongruent condition, the same test parameters were used except the word that appeared was never written in its respective colour and instead was randomly presented as one of the other three respective colours. Participants were instructed to press the arrow key that matched the colour of the word (keys were colour coded accordingly) and not what the word said. The congruent Stroop acted as the ‘rest’ block and the incongruent Stroop as the ‘active’ blocks for fNIRS analysis purposes. Each word stimulus was presented on the screen until participants pressed a key, with 1000 msec between each stimulus. Before the task, a short practice module was provided to familiarise participants with the task. Four work blocks were presented for each condition (congruent and incongruent). The task was scored for accuracy (%) and response times (msec).

2.4.4.3. Go/no-go task

The Go/No-go task or the Stop-signal task presents participants with either ‘Go’ or ‘No-go’ stimuli in a pseudorandom order. The Go-No-Go task was comprised of a block of all ‘Go’ stimuli (which acted as the control block for fNIRS analysis purposes), followed by a block of ‘No- Go’ and ‘Go’ stimuli. The ‘Go’ stimuli were represented by a blue circle, the ‘No-Go’ stimuli by a grey circle and when the ‘Go’ stimulus appeared participants were instructed to press the spacebar as quickly and accurately as possible. In the Go-No-Go blocks, the ‘No-Go’ stimuli were presented 30% of the time, as this has been found to require a higher cognitive demand and neural abilities to attend to information and inhibit a response. Stimuli were presented for a maximum of 500 msec with 500 msec in between each stimulus presented, for a total of 30 stimuli presented in each active work block. A total of four work blocks were presented.

For the Stroop and Go/no-go tasks, an inverse efficiency score (IES: [13]) was calculated to better understand the speed and accuracy trade-off in completing the task. The formula is as follows: IES = RT/PC (RT = reaction time in msec, PC = percentage correct in decimals). The score provides an estimate of the overall speed of attention. A lower score indicates better performance because a quicker response time was provided for each accurate response. Conversely, a higher score means that the response time is longer for each accurate response.

2.4.4.4. Verbal working memory and delayed recall

The Rey Auditory Verbal Learning Task (RAVLT) was used to assess immediate recall, learning, delayed recall and recognition memory. This task was presented through audio recording in order to support fNIRS block design analysis. A list of 15 nouns (list A) was read over five trials and, after each trial, participants were asked to recall, in any order, as many of the 15 words as possible. Correct responses for the five trials were summed to produce a measure of immediate recall. After a trial consisting of 15 different words (list B), participants were again required to recall the words that were presented in list A, and then again after an interval of 45 min (trial 7, delayed recall). After trial 7, participants were presented with a sheet of 50 words containing the words from lists A and B among 20 distracter words. Participants were asked to recognise the words from lists A and B and indicate the list they came from. Each word correctly identified was scored as a measure of recognition.

2.4.4.5. Functional near-infrared spectroscopy fNIRS

FNIRS is based on the concept of neurovascular coupling that suggests an increase in brain activity leads to an increase in oxygen consumption, leading to a change in region/site-specific concentration of oxyhaemoglobin (HbO) and deoxyhaemoglobin (HHb) [14]. The use of fNIRS in this study is a significant theoretical and functional effort to precisely define and operationalise baseline level (resting or control condition), cognitive load and training effects following a nutritional intervention and significantly extends the previous acute study in this area [5].

In this study, an 8-channel portable fNIRS system (Octamon, Artinis Medical Systems, The Netherlands) was used. The 8-channel system (four channels on each hemisphere) was spatially designed onto a headband as a 2 × 4 channel montage and was placed on the pre-frontal cortex (PFC). Each channel transmitter emits NIR light at 2 wavelengths (750 and 850 nm) during the cognitive test battery. All fNIRS signals were sampled at 10 Hz and individual differential path-length factor was used to account for age-related changes in brain structure. Following data collection, all participant data were exported into HOMER3 (a MATLAB-based analysis package) for pre-processing, which included motion artefact detection and correction, converting optical density to concentration and block averaging.

Following pre-processing, all data were exported into Microsoft Excel whereby channels were averaged according to the left PFC (4 channels) and right PFC (4 channels). For this study, only changes in oxyhaemogoblin (HbO) concentration were reported and analysed.

2.4.5. Blood glucose measurement

On testing days, blood glucose tests were undertaken via AccuChek Performa glucometer, test strips and single-use Lancet Safe- T-Pro Plus lancets (Roche, Castle Hill, NSW) as an accurate and reliable standard for glucose testing [15].

2.4.6. BDNF detection and quantification

The neurotrophin Brain-Derived Neurotrophic Factor (BDNF) has a fundamental role in the maintenance and plasticity of neuronal networks during adulthood. Blood samples were collected from participants on testing days at the Sullivan Nicolaides Pathology clinical trial collection centre. Serum samples were prepared with blood drawn after cognitive testing and allowed to clot for 1 h at room temperature. Samples were frozen and stored. BDNF was detected using the Millipore-Milliplex immunoassay (insengMAP Human Neurodegenerative Disease Magnetic Bead Panel (HNDG3MAG-36 K)). Preparation of the cytokine BDNF measures were conducted in line with the manufacturer’s guidelines. Total serum BDNF concentrations for each participant were used for comparisons between pre and post supplementation.

2.4.7. Supplements

Supplements were provided by USANA Health Sciences Inc. (Utah, USA), and participants were randomly allocated to consume the supplement from either group A or B (blinded active or placebo) upon enrolment into the study. Randomisation was completed using a random number generator, called random.org.

The active treatment was made up of commercially available extract of Bacopa monnieri (150 mg/tablet) (Bacognize TM), commercially available extract of American ginseng Panax quinquefolieus (50 mg/tablet) (Cereboost TM), and Whole coffee fruit extract (50 mg/tablet). Participants consumed a two-tablet dose to provide a total dose of 300 mg Bacopa (2 × 150 mg), 100 mg Ginseng (2 × 50 mg) and 100 mg of whole coffee fruit (2 × 50 mg), for 4 weeks (28-days). This provided a combined active ingredient intake of 500 mg per day. The placebo treatment was predominately microcrystalline cellulose (581 mg/tablet), to provide a total intake of 1162 mg of cellulose per day.

The manufacturer retained the lot number identity of the active and placebo products in a sealed file until the study was unblinded following data analysis. Both supplements were hard-coated tablets matched for red colour, texture and smell. Compliance was assessed through self-report and return of the supplement bottle at second testing session, with the number of tablets remaining recorded.

2.4.8. Data and statistical analysis

The primary outcome is the collective measure of affect, cognitive performance and brain function (changes in HbO) in response to active or placebo supplement use, taking into account baseline function. Differences between baseline performance scores for supplement conditions were considered, and response time for N-Back tasks were significantly different at baseline. As such, response times were subsequently used as covariates in analysis (ANCOVA) of post supplement and follow-up change in task performance for the 1 back, 2-back and 3-back tasks. Further, as gender was significantly different between supplement conditions, gender was used as a covariate in analysis of supplement effects using ANCOVA.

All cognitive performance and HbO data were converted and analysed as change-from-baseline scores, with post supplementation change between time 1 and 2, and follow-up testing being the change between post-supplement and post-2 week (14 days) washout. This approach was used to better understand change within and between supplement conditions. All statistical analysis was conducted using SPSS Statistical Software with significance set at p < 0.05. Cohens d was used for estimates of effect size. Reliable change index (RCI) scores were considered for each measure [16]. Calculations of the RCI used the standard deviation and reliability (Cronbach’s alpha) from published sources where normative scores from a sufficiently large sample were reported or the reliability score of the scale within this cohort.

3. Results

3.1. Baseline assessment

There were no significant differences between supplement conditions on baseline background demographics measures or dietary quality, shown in Table 1.

Table 1. Demographic and descriptive information for supplement conditions (means, SD).

Variable	Total sample (N = 52)	Active (n = 26)	Placebo (n = 26)	Test statistic
	M (SD)	M (SD)	M (SD)
Sex (n, %)				χ^2 = 3.90, p = 0.04
Female	40(76.9%)	17(32.69%)	23(44.23%)
Male	12(23.1%)	9(17.3%)	3(5.76%)
Age (Years)	50.46(9.46)	50.69(10.11)	50.23(8.98)	F = 0.030, p = .862
Education (Years)	17.31(4.65)	16.50(3.5)	18(5.49)	F = 1.58, p = .214
Dietary Quality (DST)	66.25(10.76)	65.00(14.98)	67.50(10.26)	F = .483, p = .490
Hours work (Weekly)	24.45(18.23)	21.63(18.45)	27.28(17.91)	F = 1.257, p = .268
Sick days in past 3 months	1.71(2.52)	1.65(2.51)	1.77(2.57)	F = 0.027, p = .871
Health rating (compared to others)	1.94(0.64)	2.08(0.63)	1.81(0.63)	F = 2.369, p = .130
Health rating (compared to a perfect state of health)	2.48(0.78)	2.58(0.76)	2.38(0.80)	F = 0.788, p = .379
Body Mass Index (BMI)	26.83(7.40)	27.27(5.83)	26.35(8.91)	F = 0.190, p = .665
Current medical conditions (Yes) (n, %)	23(44.2%)	13(24.98%)	10(19.21%)	χ^2 = .702., p = .402
Current medication use (Yes) (n, %)	23(45.1%)	12(23.53%)	11(21.57%)	χ^2 = 0.167, p = .683
Current supplement use (Yes) (n, %)	27 (51.9%)	13(48.1%)	14(51.82%)	χ^2= 0.07, p = .781
Exercise (days per month)	17.02(10.39)	14.65(9.58)	19.38(10.81)	F = 2.78., p = .101
Alcoholic drinks (in last Week)	2.71(3.84)	2.50(3.06)	2.92(4.55)	F = .155 p = .696
Smoker (Yes) (n, %)	3(5.77%)	1(1.92%)	2(3.85%)	χ^2= 0.354, p = .552

3.2. Adherence to supplementation intake

The number of days of supplementation and follow-up for each supplement group were considered for any effects. There was no statistically significant difference between groups on number of days of supplementation. For the number of days in the follow-up timeframe (time 3), the days between testing sessions were significantly different (p = .048), with those in the active condition having more days between testing sessions than those in the placebo condition. The number of days between testing sessions was subsequently used as a covariate in analysis of the change scores at follow-up. At post-supplementation (time 2), there were no significant differences between supplement conditions on number of tablets remaining.

3.3. One-day testing measures

There were no significant differences between supplement condition for blood glucose levels or affect on testing days.

3.4. Mood and prosociality

As shown in Table 2, the change scores for mood and affect measures show a significant improvement in positive affect for those in the active condition compared to placebo condition post supplementation (p = .007, d = 0.8). For those in the placebo condition they showed a significant reduction in fears of receiving compassion from others post supplement compared to the active condition. There were no statistically significant changes in mood or prosociality at follow-up (time 3).

Table 2. Baseline and change scores for mood and behaviour measures by supplement condition, presented as mean (SD).

	Baseline scores			Post-supplement Change scores			Follow-up change scores
	Active (N = 26)	Placebo (N = 26)	F ratio, p-value	Active	Placebo	F ratio, p-value	Active	Placebo	F ratio, p-value
DASS – Dep	5.61(5.57)	5.15(7.31)	.066, p = .79	−.96(1.61)	−.38(2.41)	1.41, p = .24	.38(2.15)	.08(2.28)	.004, p = .952
DASS – Anx	3.38(3.99)	5.46(5.83)	2.24, p = .141	−.11(1.24)	−.69(2.14)	2.30, p = .135	.19(1.54)	−.57(1.50)	1.64, p = .205
DASS – Stress	10.76(6.83)	10.69(7.69)	.001, p = .970	−.53(2.73)	−.76(2.64)	.049, p = .826	−.34(2.75)	−.76(2.62)	.551, p = .462
Prosociality	61.92(10.56)	60.30(10.56)	.304, p = .584	1.07(5.36)	.03(6.27)	.948, p = .335	−2.34(5.53)	−.65(6.27)	.427, p = .516
Social Safeness	39.42(9.01)	40.73(10.50)	.232, p = .632	2.19(5.26)	1.34(5.21)	.960, p = .332	−1.0(5.65)	.538(4.35)	1.09, p = .302
FOC towards others	14.11(8.25)	16.50(8.29)	1.07, p = .304	−.46(5.42)	−1.92(5.39)	.863, p.35	−.73(4.32)	.57(5.12)	.759, p = .388
FOC – from others	9.65(8.83)	11.11(8.43)	.372, p = .544	.38(6.32)	−2.23(3.96)	4.16, p = .047	.57(5.60)	.23(6.52)	.586, p = .448
FOC – self	8.96(12.46)	9.07(8.68)	.002, p = .969	.26(4.64)	−.03(5.79)	.370, p.546	.65(4.73)	−1.07(6.74)	.025, p = .875
PANAS – positive	25.46(6.15)	28.53(5.49)	3.61, p = .063	2.46(5.13)	−.53(3.73)	8.00, p = .007	−1.26(5.93)	−.07(4.18)	.468, p = .497
PANAS – negative	20.07(4.43)	21.15(4.97)	.678, p = .414	−.42(3.83)	−34(3.68)	.001, p = .974	−.07(3.53)	−1.1(4.15)	.431, p = .515

**Gender used as covariate in all analysis, for follow-up time point analysis, days between testing was also used as covariate. FOC = Fears of compassion.

When considered with a reliable change index (RCI), there was a trend towards the likelihood of 19% of individuals in the active condition reliably improving in positive affect compared to placebo (p = .071), shown in Table 4. For the fears of compassion scales, there was no reliable change for fear of receiving compassion from others. There was a trend on the measure of fears of compassion to self (p = .060), with a likelihood of 11% of those in the placebo condition reliably improving compared to those in the active condition.

Table 4. Post supplementation reliable change score for mood measures for each condition.

Mood measures	Condition N = 26 per condition	Reliable change
		n-n ^a^: (% of n = 26)	X^2	p
DASS-Depression	Active	0	1.02	.313
	Placebo	1 improved (3.85%)
DASS-Anxiety	Active	0	1.02	.313
	Placebo	1 improved (3.85%)
DASS-Stress	Active	0	NA
	Placebo	0
Prosociality	Active	5 improved (2 decline) (19.23%: 7.69%)	.84	.35
	Placebo	4 improved (1 decline) (15.38%: 3.85%)
Social Safeness	Active	9 improved (1 decline) (34%: 3.85%)	.083	.773
	Placebo	6 improved (3 decline) (23%: 11.53%)
Fears of compassion – towards others	Active	2 improved (0 decline) (7.69%)	.754	.385
	Placebo	4 improved (0 decline) (19.23%)
Fears of compassion – from others	Active	3 improved (2 decline) (11.53%: 7.69%)	.591	.442
	Placebo	3 improved (0 decline) (11.53%)
Fears of compassion – self	Active	1 improved (2 decline) (3.85%: 3.85%)	4.35	.035
	Placebo	3 improved (4 decline) (11.53%: 15.38%)
PANAS – positive	Active	5 improved (1 decline) (19.23%:3.85)	3.01	.071
	Placebo	1 improved (2 decline) (3.85%: 7.69%)
PANAS – negative	Active	1 improved (2 decline) (3.85%)	1.08	.298
	Placebo	1 improved (0 decline) (3.85%)

^a The first n represents those that did meet reliable change index and score indicates improvement on measure, the second n represents those who did meet reliable change index and score indicates worse/decline on measure.

3.5. Cognitive performance

The active condition showed a trend towards improved delayed recall post supplementation compared to placebo condition (p = .06, d = 0.54), shown in Table 3. Post supplement (time 2) change shows 2-back response times were faster in placebo condition than the active condition (p = .028, d = 0.6), however both conditions showed a slower response overall at time 2. For those in the placebo condition, there were trends for fewer errors on 1-back (p = .054, d = 0.55) and 3-back (p = .072, d = 0.51) tasks than active condition, and slightly higher accuracy on 3-back post supplement compared to active condition (p = .053, d = 0.56) based on change scores. At follow-up (time point 3), there were no significant differences between conditions.

Table 3. Baseline and change scores (mean, SD) for cognitive performance measures by supplement condition.

	Baseline scores			Post-supplement Change scores			Follow-up Change scores
Cognitive measure	Active	Placebo	F ratio, p-value	Active	Placebo	F ratio, p-value	Active	Placebo	F ratio, p-value
Ravlt total sum (A1-A5)	53.92 (9.93)	55.76 (7.78)	.556, p = .459	5.65 (6.46)	6.15 (5.85)	.194, p = .662	3.11 (8.61)	.96 (6.69)	.407, p = .527
Ravlt B1	8.08 (2.44)	8.15 (1.93)	.016, p = .900	.583 (2.54)	.15 (2.30)	.624, p = .434	.65 (1.95)	00 (2.77)	.048, p = .828
Ravlt A6	11.50 (3.35)	12.08 (2.38)	.512, p = .477	1.30 (2.64)	.769 (1.79)	.176, p = .677	−.03 (2.00)	.30 (1.93)	.246, p = .622
Ravlt A7 (delayed recall)	11.38 (3.65)	12.58 (2.30)	1.98, p = .165	1.307 (1.95)	−.0385 (2.04)	3.70, p = .06	.19 (2.03)	.23 (1.90)	.017, p = .898
Ravlt Recog	23.62 (5.61)	25.08 (3.66)	1.23, p = .272	1.42 (2.83)	.34 (2.92)	1.40, p = .242	−.34 (3.14)	.11 (3.64)	.00, p = .998
Ravlt Distractions	18.12 (1.36)	17.23 (2.37)	2.71, p = .106	.038 (2.39)	.00 (2.36)	.035, p = .853	.23 (2.17)	.07 (2.51)	.09, p = .761
1-Back RT* (msec)	458.43 (95.33)	520.84 (107.18)**	4.92, p = .031	12.80 (81.48)	28.42 (83.56)	3.43, p = .07	18.62 (65.80)	2.64 (76.91)	.014, p = .710
1-Back Accuracy (%)	92.54 (8.48)	94.00 (9.40)	.342, p = .562	3.12 (7.95)	3.72 (7.38)	.851, p = .361	1.20 (9.01)	−1.80 (6.39)	2.43, p = .125
1-Back Error (n)	1.24 (1.74)	2.04 (3.40)	.445, p = .508	−.57 (1.39)	−1.34 (2.39)	3.89, p = .054	−.03 (1.79)	.53 (2.50)	1.43, p = .237
2-Back Response Time (msec)	466.26 (95.91)	569.16 (107.18**)	11.01, p = .002	25.78 (97.58)	19.26 (130.30)	5.12, p = .028	13.16 (121.65)	−10.96 (90.15)	.398, p = .531
2-Back Accuracy (%)	80.96 (17.22)	84.88 (13.69)	.828, p.367	4.69 (9.54)	5.91 (16.49)	1.84, p = .180	1.71 (14.21)	1.13 (9.88)	.074, p = .787
2-Back Error (n)	4.54 (3.42)	3.69 (3.57)	.760, = .387	−1.30 (2.69)	−1.00 (3.09)	.287, p = .594	−.03 (3.30)	−.30 (1.91)	.01, p = .919
3-Back RT (msec)	457.87 (164.533)	460.79 (198.19)	.065, p = .800	−5.08 (162.03)	51.38 (233)	1.57, p.309	−4.96 (138.45)	5.62 (159.37)	.002,p = .962
3-Back Accuracy (%)	63.57 (18.72)	55.99 (17.84)	2.89, p = .142	.34 (17.11)	8.74 (24.14)	3.93, p = .053	2.18 (16.61)	1.37 (21.38)	.575, p = .452
3-Back Error (n)	9.77 (4.54)	10.88 (3.89)	.990, p = .324	−.50 (4.30)	−2.50 (4.98)	3.37, p = .072	−.73 (3.48)	−.80 (4.23)	.360, p = .551
Go-NoGo RT (msec)	314.37 (36.01)	305.83 (32.61)	.802, p = .375	−2.54 (46.07)	−10.61 (53.57)	.215, p = .645	6.84 (32.20)	8.74 (31.82)	.003, p = .959
Go-NoGo Accuracy (%) (inhibited)	88.81 (7.90)	87.30 (8.44)	.441, p = .510	−1.40 (8.76)	.16 (9.37)	.329, p = .569	1.70 (5.74)	2.76 (5.42)	.448,p = .507
Go-NoGo IES^a	354.06 (26.12)	350.62 (22.04)	.263, p = .610	2.44 (30.35)	3.13 (19.67)	.009 p = .998	−1.06 (25.39)	−.40 (24.66)	.013,p = .908
Stroop Congruent RT (msec)	712.88 (121.43)	777.80 (171.71)	2.47, p = .122	−28.136 (55.89)	−36.79 (70.06)	.074, p = .786	−1.77 (49.90)	1.60 (87.53)	.287,p = .595
Stroop Congruent Accuracy (%)	98.42 (1.42)	98.62 (1.77)	.178, p = .675	−.22 (1.44)	.11 (1.58)	2.13, p = .151	.57 (1.46)	.20 (1.35)	2.58, p = .116
Stroop Test – Congruent IES^a	723.52 (117.82)	787.53 (167.66)	2.53, p = .117	−26.73 (57.36)	−37.44 (71.43)	.22, p = .640	−5.59 (51.28)	.04 (87.25)	.534, p = .46
Stroop Test – Incongruent RT (msec)	410.44 (82.83)	442.66 (95.64)	1.68, p = .200	−6.61 (60.37)	−24.73 (40.96)	1.40, p = .241	−19.90 (53.76)	−9.00 (22.72)	1.85, p = .179
Stroop Test – Incongruent Accuracy (%)	98.42 (1.51)	98.66 (1.82)	.259, p = .613	.29 (2.05)	.33 (1.23)	.837, p = .365	−.05 (1.16)	−.38 (1.77)	3.30, p = .075
Stroop Test – Incongruent IES^a	416.95 (83.87)	449.19 (100.56)	1.57, p = .215	−8.04 (59.54)	−27.23 (45.36)	1.74, p = .193	−20.02 (55.68)	−6.97 (28.09)	2.59, p = .114

**Gender used as covariate in all analysis, for follow-up time point, days between testing was also used as covariate. For N-back tasks, baseline RT (response time) was used a covariate also.

The reliable change index shows significant differences between condition change scores with 46% of those in the active condition showing a reliable improvement in delayed recall performance compared to those in the placebo condition, which showed 30% decrease in positive affect (p = .046) – see Table 5.

Table 5. Post supplementation reliable change score for each cognitive measure for each condition.

Cognitive measure	Condition	Reliable change index met: improved (decline)
		(n-n)^a % of N = 26	X^2	p
Ravlt total sum (A1–A5)	Active	11 improved (1 decline) (43.3%: 3.85%)	.00	1
	Placebo	11 improved (1 decline) (43.3%: 3.85%)
Ravlt B1	Active	7 improved (5 decline) (26.9%: 19.23%)	1.23	.266
	Placebo	3 improved (7 decline) (11.53%: 26.9%)
Ravlt A7 (delayed recall)	Active	12 improved (1 decline) (46.1%: 3.85%)	6.1	.046
	Placebo	5 improved (8 decline) (19.23%: 30.76%)
Ravlt Recog	Active	4 improved (1 decline) (15.38%: 3.85%)	.115	.73
	Placebo	4 improved (2 decline) (15.38%: 7.69%)
1-Back Accuracy (%)	Active	2 improved (0 decline) (7.69%)	.221	.638
	Placebo	3 improved (0 decline) (11.53%)
1-Back RT* (msec)	Active	0	1.02	.313
	Placebo	0 improved 1 decline (3.85%)
1-Back Error (n)	Active	1 improved (0 decline) (3.85%)	.354	.552
	Placebo	2 improved (0 decline) (7.69%)
2-Back Accuracy (%)	Active	0 improved (0 decline)	2.08	.09
	Placebo	2 improved (0 decline) (7.69%)
2-Back RT (msec)	Active	0 improved (3 decline) (11.53%)	2.02	.364
	Placebo	0 improved (1 decline) (3.85%)
2-Back Error (n)	Active	1 improved (0 decline) (3.85%)	.354	.552
	Placebo	2 improved (0 decline) (7.69)
3-Back Accuracy (%)	Active	1 improved (0 decline) (3.85%)	3.24	.071
	Placebo	4 improved (1 decline) (15.38%)
3-Back RT (msec)	Active	0 improved (1 decline) (3.85%)	.00	1
	Placebo	0 improved (1 decline) (3.85%)
3-Back Error (n)	Active	2 improved (0 decline) (7.69%)	2.36	.124
	Placebo	4 improved (1 decline) (15.38%:3.85%)
Go-NoGo RT (msec)	Active	4 improved (5 decline) (15.38%: 19.23%)	.087	.768
	Placebo	1 improved (7 decline) (3.85%: 26.9%)
Go-NoGo Acc** (%) (inhibited)	Active	4 improved (3 decline) (15.38%: 11.53%)	.433	.510
	Placebo	5 improved (0 decline) (19.23%)
Stroop – Congruent RT (msec)	Active	2 improved (0 decline) (7.69%)	.754	.385
	Placebo	4 improved (0 decline) (15.38%)
Stroop – Congruent Acc (%)	Active	7 improved (3 decline) (26.9%: 11.53%)	1.44	.229
	Placebo	3 improved (3 decline) (11.53%:11.53%)
Stroop – Incongruent RT (msec)	Active	2 improved (3 decline) (7.69%:11.53%)	.00	1
	Placebo	5 improved (0 decline) (19.23%)
Stroop – Incongruent Acc (%)	Active	8 improved (4 decline) (30.76%: 15.38%)	1.30	.254
	Placebo	5 improved (3 decline) (19.23%: 11.53%)

^a The first n represents those that did meet reliable change index and score indicates improved performance on measure, the second n represents those who did meet reliable change index and score indicates worse/decline in performance on measure. *RT = response time. **Acc = Accuracy.

3.6. Change in HbO by supplement condition

Shown in Table 6, HbO change score post supplementation showed greater increase in activation for those in the active condition compared to those in the placebo condition for performance on the 2-back task (p = .038, d = 0.6). There were no other changes in overall PFC activation between conditions for task performance at baseline, post supplement or follow-up.

Table 6. HbO response at baseline and change scores (mean and SD).

	Baseline scores			Post-supplement change scores			Follow-up change scores
FNIR measure for task	Active	Placebo	F ratio, p-value	Active	Placebo	F ratio, p-value	Active	Placebo	F ratio, p-value
Ravlt A	.793 (1.98)	.925 (1.15)	.087, p = .770	.505 (2.45)	.023 (1.26)	.793, p = .377	−.517 (1.99)	−.052 (1.56)	.873, p = .355
Ravlt B	.462 (2.46)	.780 (1.83)	.277, p = .601	1.01 (4.99)	.23 (3.06)	.458, p = .502	−.460 (4.94)	−.112 (3.30)	.089, p = .766
RavltA7 (delayed recall)	.497 (3.44)	1.69 (3.06)	1.76, p = .190	.691 (4.96)	−.594 (3.81)	1.09, p = .300	−.054 (4.94)	−.744 (3.42)	.341, p = .562
Ravlt Recog	1.49 (2.56)	.587 (2.63)	.786, p = .380	−.910 (4.67)	.071 (2.86)	.835, p = .365	.702 (5.02)	.622 (3.31)	.005, p = .947
1-Back	−.855 (1.12)	−.805 (1.15)	.025, p = .901	.557 (1.80)	.024 (1.57)	1.38, p = .245	−.381 (1.74)	−.023 (1.73)	.549, p = .462
2-Back	−1.40 (1.19)	−.881 (1.24)	2.44, p = .124	.835 (2.01)	−.197 (1.42)	4.53, p = .038**	−.172 (1.69)	.005 (1.68)	.143, p = .707
3-Back	−1.06 (1.44)	−1.25 (1.40)	.228, p = .635	.818 (3.07)	.707 (1.78)	.025, p = .874	−.441 (2.58)	−.183 (1.66)	.182, p = .671
Go-NoGo	−.451 (1.60)	−.589 (1.03)	.137, p = .713	.181 (1.39)	−.113 (2.5)	.275, p = .602	−.011 (1.37)	.444 (2.87)	.532, p = .469
Stroop – Congruent	−.014 (1.91)	−.198 (1.33)	.161, p = .690	−.115 (2.72)	1.29 (4.177)	2.06, p = .157	−.895 (4.96)	.131 (2.87)	.834, p = .366
Stroop – Incongruent	−.288 (1.42)	−.049 (1.49)	.349, p = .557	1.25 (5.84)	.30 (3.12)	.528, p = .471	−.792 (5.73)	−.196 (2.68)	.230, p = .634

** Change noted in left hand side 2 back F = 4.56, p = .038 only, not right hand side. No other left or right hand activation changes note.

3.7. BDNF

There were no significant differences between supplement conditions in BDNF levels at any time point.

4. Discussion

This study demonstrated that a combined formulated dose of Bacopa, ginseng and whole coffee fruit reliably improved positive affect and delayed recall compared to placebo following 28 days of supplementation. This is the first report of a 28-day supplement intervention and 2-week follow-up study to assess changes in affect, cognition, cerebral haemodynamic response and BDNF in healthy middle-aged adults. For the brain, the overall battery of tasks indicated that those in the active condition showed greater PFC activation on performance of the 2-back tasks post supplementation compared to placebo, despite improved (faster) response time changes for the 2-back task in the placebo condition. There were no significant changes in BDNF levels across supplement conditions.

As there were baseline differences on the n-back task response time, response time was used as a covariate in change-score analysis, as was gender. The change from baseline response in the placebo condition compared to active condition is likely a practice-re-test effect. Namely, the significantly slower response time at baseline in the n-back task for those in the placebo condition compared to the active condition indicates a greater likelihood of ‘speeding up’ by time 2 through re-test. In contrast, those in the active condition over time and at post supplementation compared to placebo have a consistent response time trajectory. Thus, it is plausible that at baseline, the active condition group were already demonstrating engaged mental effort to complete the tasks compared to the placebo condition group, who showed a significantly slower response time.

Importantly, the significant increase in fNIRS PFC activation for the active condition during 2-back task post supplement compared to placebo suggests that those in the active condition were mentally ‘working’ to address the cognitive working memory demand of the battery in the absence of task-specific improvement. This interpretation becomes more plausible when the flow of the overall task battery is considered. Specifically, the presentation of the 2-back and 3-back tasks was randomised in the middle of the battery, with the 1 back at the start of the battery for familiarisation, along with the word lists for the RAVLT, and the delayed recall and recognition tasks presented at the end to ensure 30-min delayed recall. Consequently, at the post-supplement testing session, those who received the active supplement were engaging with the task (indicated by increased PFC activation), worked slightly slower and maintained accuracy across the n-back tasks in order to retain the information for delayed recall later in the battery, which reliably improved. The ability to better utilise cognitive resources may have been aided by the biological resource in the polyphenol supplement that was not available to those in the placebo condition and draw on cognitive reserve (CR) to retain and recall information for the delayed recall task [17].

There was a significant increase in positive affect for those in the active condition compared to placebo post supplementation, and a significant reduction in fear of expressing compassion from others in the placebo condition. For the Positive affect (PANAS) result, those in the active condition compared to placebo showed a trend for reliable change in up to 19% of those who received the supplement compared to those who did not (p = .071). The overall significant difference in change on this measure compared to placebo post-supplement is important in relation to the reliable improvement in delayed memory recall performance for those in the active condition. Emotion control and cognitive regulation within the brain are integrated processes to either enhance or impair learning and longer-term memory [18], mobilising cognitive resources for working memory and recall [19]. Physiological resources that increase the capacity for emotional and pro-social processing therefore can be a way to stabilise emotional-cognitive processing in the brain that may underpin CR into older adulthood.

Consequently, it is plausible that there are synergistic effects of a combined polyphenol supplementation in healthy middle-aged adults compared to placebo after 28 days (4-weeks), for three reasons. First, 30-min delayed recall tasks are sensitive and predictive of longer-term cognitive function [20] and suggest a potential mechanism of effect not typically observed following 4 weeks compared to 8 or 12-week supplementation periods Second, changes in delayed recall are typically related to changes in the activation of the medial temporal lobe in healthy older adults (aged 52–92), a region highly associated with memory performance, which was not measured in this study and suggests an activation in a region of the brain not typically explored in supplementation studies. Third, the daily dose of each ingredient was consumed at lower amounts than previous longer-terms studies (i.e. daily dose of Bacopa in the current study of 100 mg vs 300mg–450 mg daily dose in studies of 12 weeks duration [21]).

4.1. Implications and limitations

Cognitive changes in middle-age are subtle and there are plausible protective and collaborative neural mechanisms and brain processes working in concert that can impact reliable and meaningful change for healthy individuals. The findings suggest the additional bioavailable polyphenolic substrates contributed to a cognitive reserve that facilitated better delayed recall memory, possibly through promoting neural efficiency in concert with overlapping processes of the pre-frontal brain regions involved in emotional regulation and increased positive affect in middle-aged adults. Whilst there were no detectable changes in serum BDNF levels in this study, this is likely due to a range of environmental constraints, including and not limited to lockdowns for COVID and exposure to the virus, inadequate overall dose in this age group and overall health status as baseline. Unfortunately, recruitment and subsequent sample size for this study were negatively impacted by the arduous procedural and negative public sentiment during the pandemic. Whilst challenging, this context establishes a fitting ecological validity and significance for the finding. That is, individuals that received supplementation felt significantly more positive compared to placebo, during a time when many were feeling low in Australia [22].

5. Conclusion

This study demonstrates novel improvements of positive affect and reliable change in delayed recall following a cognitively demanding working memory battery post 28 days of supplementation with the active polyphenol-rich supplement compared to placebo. This study also replicated the feasibility and sensitivity of fNIRS in determining pre-frontal cortex activation related to task performance following a nutritional intervention. Further research would benefit from an integrated approach to explore the effects of nutritional interventions on cognitive-emotional processes in middle-aged adults, and study designs to decipher the effects of an optimal dose and time frame for supplementation to promote brain and mind health into older adulthood.

Author contributions

Conceptualisation, T.B and W.T; methodology, T.B; software, T.B, J.M and W.T; formal analysis and original draft preparation, T.B, J.M and W.T; writing – review and editing, T.B, J.M, W.T; supervision, T.B; project administration, T.B, J.M; funding acquisition, T.B. All authors have read and agreed to the published version of the manuscript.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The authors confirm that the data supporting the findings of this study are available within the article and are available upon request to corresponding author.

Additional information

Funding

This research was funded by USANA Health Sciences, Inc. [grant number RSH/5406].

Notes on contributors

Talitha Best

Talitha Best is a Professor in Psychology and Clinical Psychologist who specialises in mental health, nutrition, cognitive neuroscience and compassion-based approaches to support brain and mind health.

Jessica Miller

Jessica MiIller is a research coordinator and Masters graduate in magnetic resonance technology, with experience in clinical research.

Wei-Peng Teo

Dr Wei-Peng Teo is an exercise neuroscientist who specializes in the use of non-invasive brain stimulation and neuroimaging techniques to measure neural adaptations and cognitive functioning to exercise, lifestyle and nutritional interventions.