Feasibility and tolerability of Musa sp. pseudo-stem core powder as a dietary supplement with gut microbiota-modulating potential: a pilot randomised controlled trial

Abstract

Background

The gut microbiota is increasingly recognised as a therapeutic target for mental disorders, with nutritional supplements providing a direct approach to modulating the gut microbiota. However, existing evidence largely focuses on single nutrients or phytochemicals, highlighting the need for novel multi-component interventions. This pilot study evaluated the feasibility and tolerability of a dietary supplement powder derived from Musa sp. pseudo-stem core that contains a combination of microbiota-modulating nutrients and phytochemicals.

Methods

We recruited 30 healthy adults, aged between 18 and 65 years, from the community to a triple-blinded, 4-week randomised placebo-controlled pilot trial. Participants were randomised to receive either 10 g/day of the Musa sp. pseudo-stem core powder or maltodextrin placebo. Supplement feasibility was defined as consuming at least 80% of the allocated supplement along with habitual diet. Tolerability was defined as no changes in gastrointestinal (GI) symptoms and stool consistency compared with placebo. Exploratory outcomes included psychological symptoms, quality of life, urinary hippurate measured using nuclear magnetic resonance spectroscopy, and stool gut microbiota profiling using 16S rRNA sequencing.

Results

At the end of the study, 86% of the participants in the supplement group met the 80% consumption threshold, demonstrating intervention feasibility. There were no observed differences in the changes in GI symptoms or stool consistency between the supplement and placebo groups over the period of the study, indicating supplement tolerability. There were no observed differences in the changes in psychological symptoms, quality of life, and urine hippurate levels between groups at week 4. The weighted UniFrac distance showed a structural difference of the gut microbiota community between the supplement and placebo groups at week 4, although no observed differences were found in alpha diversity. Differences in the changes of two bacterial taxa each belonging to the family Anaerovoracaceae and Prevotellaceae were observed in the supplement group compared with the placebo group at week 4.

Conclusion

Our pilot study showed that Musa sp. pseudo-stem core powder was feasible and well-tolerated as a dietary supplement, with potential for modulating the gut microbiota. Future adequately powered trials are warranted to evaluate its effect on mental health and to elucidate underlying biological mechanisms.

Trial registration

The trial was registered with the Australian New Zealand Clinical Trials Registry (ACTRN12623000454673) on 2 May 2023.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s40814-026-01877-9.

Key messages

• What uncertainties existed regarding the feasibility?

  It was not known whether Musa sp. pseudo-stem powder would be feasible and well-tolerated as a daily dietary supplement when consumed alongside the habitual diet, without causing adverse gastrointestinal symptoms in healthy adults.

• What are the key feasibility findings?

  The Musa sp. pseudo-stem powder as a supplement was feasible and well-tolerated with no significant changes to gastrointestinal symptoms. Recruitment and retention were successful, and biological sample collection was achievable.

• What are the implications of the feasibility findings for the design of the main study?

  The findings provide a foundation to a larger, longer-duration trial. High adherence and tolerability justify its use in extended interventions, and observed microbiota shifts warrant continued inclusion of microbial outcomes as potential mechanistic pathways.

Background

Mental disorders are a leading contributor to the global burden of disease [1], with a complex pathophysiology involving neurobiological, immune, metabolic, and gastrointestinal pathways [2]. Given the limitations in the current pharmacological and psychological therapies, including partial effectiveness and variability in treatment response [3, 4], there has been a surging interest in adjunctive therapies that target novel biological pathways implicated in mental disorders. One biological system of increasing relevance to mental health is the microbiota–gut–brain axis [5]. The gut microbiota is one of the central elements of the microbiota–gut–brain axis and plays a vital role in regulating this axis through mechanisms, including immune modulation, stimulation of the vagal nerve and hypothalamic-pituitary-adrenalin axis, and production of bacterial metabolites [5, 6]. Alterations in gut microbiota composition and function have been associated with several mental disorders, including depression and anxiety [7, 8].

Nutritional interventions provide a direct and accessible strategy for modulating the gut microbiota, particularly plant-based dietary supplements rich in dietary fibre, prebiotics, and phytochemicals such as polyphenols [9–11]. These compounds are primarily metabolised by the gut microbiota and yield metabolites that possess numerous physiological functions. For example, dietary fibre fermentation by the gut microbiota yields molecules such as short chain fatty acids (SCFAs) [9, 11], which impact the gut–brain axis through various pathways including immune modulation, interacting with G-protein coupled receptors, and also contributing to neurotransmitters and neurotrophic factor production [12–14]. These mechanisms are implicated in the pathophysiology of mental disorders [15]. Similarly, polyphenols and their microbial metabolites modulate microbial composition, immune pathways, and intestinal permeability, thereby exerting additional effects on gut–brain axis signalling [16, 17].

Consistently, a growing body of evidence suggests that supplementation with gut microbiota-targeted plant-derived compounds, including dietary fibre, prebiotics, and polyphenols, may beneficially influence mental and brain health through modulation of the microbiota–gut–brain axis [18–20]. However, most existing studies have investigated isolated nutrients or extracts, and there is limited human research examining whole-food plant-based supplements that combine multiple gut microbiota-targeted nutrients and phytochemicals within a single natural matrix [19]. Such combinations may exert synergistic effects through complementary mechanisms acting on the gut microbiota and gut–brain axis [21]. Therefore, we focused on developing a supplement from Musa sp. pseudo-stem core, a plant material naturally rich in dietary fibre and polyphenols.

Musa sp. (banana) is a herbaceous flowering plant belonging to the family Musaceae with most parts of the plant including the root, pseudo-stem, fruit, and inflorescence used as food or components of traditional medicine [22, 23]. Whilst previous research has predominantly focused on the fruit [24], considerably less attention has been given to the edible pseudo-stem, a vascular stem comprising a tender core surrounded by tightly packed overlapping leaf sheaths [25], which is often considered an agricultural waste. In many traditional societies, the Musa sp. pseudo-stem core is consumed as a vegetable or juice extract, due to its rich nutritional profile [25, 26]. The pseudo-stem core is rich in dietary fibre [27], phytochemicals including polyphenols [27, 28] such as p-hydroxybenzoic and gallic acids, and minerals such as potassium, calcium, and zinc [27]. Apart from its nutritional value, pre-clinical evidence has demonstrated antibacterial and antioxidant properties of Musa sp. pseudo-stem core derivatives [29]. Importantly, the aqueous extracts of the Musa sp. pseudo-stem core have demonstrated anti-depressant [30] and anxiolytic [31] effects in animal studies.

Collectively, these findings suggest that Musa sp. pseudo-stem core may represent a promising gut microbiota-targeted nutritional intervention with potential relevance to mental health. However, despite its nutritional and therapeutic potential, human research investigating Musa sp. pseudo-stem core or its derivatives remains extremely limited, with only one uncontrolled study in people with urinary disorders reporting improvements in urinary calculi, suggesting the therapeutic potential of Musa sp. pseudo-stem core derivatives in tablet form [32].

To address this gap, we conducted a pilot randomised controlled trial (RCT) to investigate the feasibility and tolerability of a Musa sp. pseudo-stem core powder supplement developed through co-design input from food technologists and consumers. We also explored its impact on psychological symptoms, quality of life, gut microbiota composition, and urinary hippurate, a gut microbiota-derived metabolite associated with mental disorders [33, 34], with the aim of elucidating preliminary potential mechanisms of action.

Methods

Study design

Healthy adults from the community (living within the Greater Geelong/Melbourne region) were recruited to a 4-week parallel group, triple-blinded, randomised placebo-controlled trial. This study was approved by the Deakin University Human Research Ethics Committee (DUHREC, 2022-218) and was registered on the Australian New Zealand Clinical Trials Registry (ACTRN12623000454673) prior to commencing the study. A Clinical Trial Notification (CTN) was submitted to the Australian Therapeutic Goods Administration (TGA) through DUHREC.

Eligibility criteria

Participants aged 18–65 years were deemed eligible if they had no major diseases (e.g. diabetes, cancer, bipolar disorder, or eating disorders); no significant health concerns (e.g. awaiting test results, ongoing medical investigations, or a history of gastrointestinal (GI) surgery other than cholecystectomy or haemorrhoidectomy); and no medical conditions that could affect adherence to the trial (e.g. cognitive disorders or intellectual disability). Participants were also required to be willing to consume the investigational product and able to understand study materials and instructions in English.

Participants were excluded if they had any major GI disorders, including inflammatory bowel disease, coeliac disease, or irritable bowel syndrome; perceived or formally diagnosed food intolerances (e.g. lactose or gluten intolerance) or food allergies (e.g. dairy or protein allergies); current use of prebiotic, probiotic, or fibre supplements; adherence to therapeutic diets (e.g. calorie-restricted or low-FODMAP diets); antibiotic use within the previous month; or were pregnant or lactating.

Recruitment and screening

Participant recruitment occurred between May 2023 and November 2023. The study was conducted via both web-based virtual platform (video call) and in-person visits at the Food & Mood Centre (FAMC) in Geelong. Recruitment occurred through multiple settings including a webpage hosted on the FAMC website, social media, newsletters, and leaflet/pamphlets. Individuals interested in participating were directed to complete an online pre-screening questionnaire administered through Research Electronic Data Capture (REDCap) to determine preliminary eligibility. Pre-eligible participants were subsequently contacted by the study researcher to discuss the study in further detail. Those who appeared eligible and keen to participate were scheduled for an enrolment and baseline appointment conducted via a Zoom video call.

Enrolment, baseline assessment, and allocation of intervention

During the scheduled Zoom video call (day 1), pre-eligible participants underwent re-screening to confirm eligibility prior to enrolment. Participants who met the eligibility criteria were enrolled and provided informed consent before commencing baseline assessments. This appointment also marked the beginning of the 10-day baseline assessment period.

During the call, participants completed demographic and medication-use questionnaires, and the study procedures were explained in detail. Following the video call, participants completed the remaining baseline assessment questionnaires online, including measures of GI symptoms, stool outputs, psychological distress, quality of life, physical activity, and habitual dietary intake over the subsequent 9 days (Additional file 1: Table 11).

On day 10, participants attended an in-person study appointment (visit 1) to return stool and urine samples collected at home, marking the completion of the baseline assessment period. During visit 1, participants were randomised to receive either 10 g/day of Musa sp. pseudo-stem powder (hereafter supplement) or maltodextrin (placebo) for a period of 4 weeks. Both the supplement and placebo were provided in 5 g opaque sachets masking the product enclosed (in total each participant received 56 sachets). Participants were instructed to consume two sachets per day by mixing the powders with a variety of usually consumed food products across main meals and snacks and advised not to change their habitual diet. Participants were provided a booklet containing information on the type of usually consumed foods with which the study products could be mixed (e.g. yogurt, soup, pasta, and cereal) and the recommended quantity of the study product to mix with such foods (e.g. 2–3 g in 1 serve/200 g of yogurt).

Post-baseline data collection

Participants commenced the intervention the day following visit 1 and consumed the allocated study product for 4 weeks. During this period, participants completed weekly online questionnaires concerning GI symptoms, stool outputs, and study product adherence throughout the 4-week study duration. During the final study week, participants also completed questionnaires concerning psychological distress, quality of life, physical activity, and habitual dietary intake (Additional file 1: Table 1). At the end of the study, participants attended a final in-person appointment (visit 2) to return stool and urine samples collected at home. Participants received a $15 e-gift voucher at each in-person visit in compensation of their parking cost.

Study product development

The prototype development of the supplement was conducted at the Consumer Analytic Safety Sensory (CASS) Food Research Centre at Deakin University (Additional file 1: Section 2). The supplement used in the trial was subsequently contract manufactured at an FDA-certified facility in Brisbane (refer to Additional file 1: Section 2.1 for manufacturing details). Proximate analyses, fibre characterisation, and polyphenol analyses were performed on the supplement powder (Additional file 1: Tables 2 and 3). A moderate dose of 10 g/day was selected for this study, despite previous trials of plant-based dietary supplements (e.g. chia seed) commonly using doses of 25–30 g/day [35, 36]. This choice was made because it was the first study of its kind, and we wanted to minimise the risk of any potential GI discomfort in participants, whilst still ensuring that enough active constituents were present in the tested dose (Additional file 1: Tables 2 and 3). Maltodextrin, used as the placebo, was purchased from an online retail seller (Bulk Nutrients). The participant booklet on study product usage was developed based on product development trials and a sensory evaluation study that were conducted using the supplement powder in 48 healthy participants (approved by the Deakin University Human Ethics Advisory Group-Health, HEAG-H155_2021). Further details regarding the sensory evaluation methods and findings are provided in Additional file 1 (Section 2.4 for sensory evaluation methods and Table 4 and Figs. 1−7 for results of this phase).

Table 2.

Gastrointestinal symptoms and stool consistency scores at baseline and post-baseline timepoints in the supplement and placebo groups, and the between-group differences in change over time using generalised estimating equations models (supplement group n = 14; placebo group n = 16)

Gastrointestinal symptoms scores (11-point NRS)
	Baseline M (SD)	Week 1 M (SD)	Week 2 M (SD)	Week 3 M (SD)	Week 4 M (SD)	Baseline vs week 1		Baseline vs week 2		Baseline vs week 3		Baseline vs week 4
						Estimate	95% CI	Estimate	95% CI	Estimate	95% CI	Estimate	95% CI
Abdominal pain
Supplement	0.79 (0.97)	1.29 (1.49)	1.14 (1.10)	0.62 (1.12)	1.08 (1.32)	0.25	−0.92, 1.42	0.75	−0.01, 1.51	0.15	−0.72, 1.02	0.61	−0.18, 1.41
Placebo	0.75 (0.93)	1.00 (1.67)	0.36 (0.84)	0.43 (0.51)	0.43 (0.65)
Bloating
Supplement	1.64 (1.60)	1.79 (1.53)	1.43 (1.55)	1.38 (1.76)	1.31 (1.60)	−0.04	−1.39, 1.30	0.51	−0.41, 1.43	0.32	−0.66, 1.30	0.46	−0.70, 1.62
Placebo	1.44 (2.00)	1.62 (1.89)	0.71 (1.44)	0.86 (1.35)	0.64 (1.28)
Wind
Supplement	2.07 (1.49)	2.50 (1.56)	2.07 (1.77)	1.85 (1.91)	1.85 (1.95)	0.37	−0.88, 1.61	0.54	−0.47, 1.56	0.68	−0.65, 2.00	0.46	−0.87, 1.80
Placebo	1.69 (1.45)	1.75 (1.34)	1.14 (0.95)	0.79 (0.58)	1.00 (0.68)
Nausea
Supplement	0.43 (0.85)	1.07 (1.27)	0.86 (1.03)	0.62 (1.19)	0.62 (1.04)	0.39	−0.56, 1.34	0.61	−0.04, 1.26	0.15	−0.74, 1.04	0.44	−0.35, 1.22
Placebo	0.25 (0.77)	0.50 (1.21)	0.07 (0.27)	0.29 (0.83)	0.00 (0.00)

Stool output scores (BSFS)
Stool type
Supplement	3.57 (1.22)	4.36 (1.15)	3.43 (1.09)	3.79 (1.12)	3.62 (1.33)	1.35	0.47, 2.22	−0.13	−1.13, 0.88	0.30	−0.73, 1.34	0.35	−0.80, 1.49
Placebo	3.88 (1.15)	3.31 (1.14)	3.86 (1.35)	3.79 (1.12)	3.57 (1.45)

Stool frequency (per day)
Supplement	1.35 (0.53)	1.89 (0.44)	1.58 (0.91)	1.73 (1.44)	1.62 (0.68)	0.85	0.38, 1.33	0.41	−0.16, 0.98	0.51	−0.36, 1.38	0.32	−0.31, 0.94
Placebo	1.45 (0.65)	1.13 (0.39)	1.26 (0.51)	1.32 (0.62)	1.41 (0.66)

M mean, SD standard deviation, NRS Numeric Rating Scale (range 0–10); BSFS Bristol Stool Form Scale (range 1–7). Stool frequency is presented as mean bowel actions per day. The estimate reflects the between-group differences in change in scores for gastrointestinal symptoms and stool output

Table 4.

Between-group differences in changes in centred log-transformed abundance of bacterial taxa from baseline to week 4 between the supplement and placebo groups using generalised estimating equations (supplement group n = 14 and placebo group n = 16)

ASV	Estimate	Standard error	95% CI	q value	Class	Family	Genus
ASV_1070	1.90	0.62	0.68, 3.13	0.234	Clostridia	Anaerovoracaceae	Family_XIII_UCG-001
ASV_4647	1.67	0.55	0.60, 2.74	0.234	Bacteroidia	Prevotellaceae	Paraprevotella
ASV_2195	1.57	0.66	0.27, 2.87	0.565	Clostridia	Ruminococcaceae	CAG-352
ASV_3341	−0.62	0.27	−1.16, −0.09	0.565	Clostridia	Lachnospiraceae	[Eubacterium]_hallii_group
ASV_4638	1.47	0.65	0.20, 2.75	0.565	Bacteroidia	Prevotellaceae	Alloprevotella
ASV_507	−0.37	0.16	−0.69, −0.40	0.565	Bacilli	Izemoplasmatales	Izemoplasmatales
ASV_2979	1.60	0.75	0.14, 3.08	0.565	Clostridia	Lachnospiraceae	Lachnospiraceae_NK4A136_group
ASV_2350	1.33	0.62	0.10, 2.55	0.565	Clostridia	Ruminococcaceae	Faecalibacterium
ASV_440	−0.63	0.30	−1.21, 0.05	0.565	Clostridia	Christensenellaceae	Uncultured
ASV_3131	−0.40	0.19	−0.77, −0.03	0.565	Clostridia	Lachnospiraceae	Dorea
ASV_2510	−1.21	0.58	−2.35, −0.07	0.565	Clostridia	Butyricicoccaceae	UCG-008
ASV_5276	−0.61	0.30	−1.19, −0.04	0.565	Clostridia	Ruminococcaceae	Paludicola
ASV_3437	−0.88	0.42	−1.71, −0.05	0.565	Clostridia	Lachnospiraceae	Shuttleworthia

The estimate reflects the between-group differences in change

Fig. 1.

Consort flow diagram

Randomisation and blinding

A stratified permuted block randomisation was used to equally assign participants to the intervention and placebo group on a 1:1 ratio. A third-party statistician independent of the research team developed a computer-generated randomisation table utilising random block sizes and partitioned by four strata based on age and sex: (1) age: 18–44, sex: female; (2) age: 45–65, sex: female; (3) age: 18–44, sex: male; (4) age: 45–65, sex: male. Four blinded “study kit” allocation lists, one for each stratum, containing a random unique three-character alphanumeric study kit codes (e.g. KL5, ZA1, CJ8) were prepared and added in place of a group (intervention or placebo) by the same statistician. An unblinded research assistant (not involved in any other study procedures) printed the kit labels and labelled each study products with the appropriate kit code, which ensured the study investigators, statistician, and participants were blinded to the group allocation (triple-blind).

Primary outcomes

Intervention feasibility, intervention tolerability, and study feasibility were the primary outcomes.

Intervention feasibility

Intervention feasibility was assessed through adherence to the study products. Adherence was defined as consuming at least 80% of the allocated supplement sachets (45 out of 56 sachets), a cut-off threshold commonly used to measure adherence in trials [37, 38]. Participants recorded the number of sachets consumed each day to the nearest half-sachet using a daily diary and transposed the total number of sachets consumed during the week to a weekly online adherence questionnaire. This allowed researchers to monitor adherence on a weekly basis and contact participants if any issues were noted. Participants returned any unconsumed sachets and the daily diaries at the end of the study (visit 2).

Intervention tolerability

Intervention tolerability was defined as no difference in the change of GI symptoms and stools with normal consistency compared with the placebo group over a 4-week period.

Individual GI symptoms, including abdominal pain/discomfort, bloating, flatulence, and nausea, were assessed over three consecutive days at baseline and post-baseline every week using an 11-point simple Numeric Rating Scale (NRS) questionnaire (score range: 0–10), where 0 indicated “no symptoms at all” and 10 indicated “extremely worst/awful symptoms”. The NRS is a clinical rating scale, which has previously been validated for measuring pain [39].

Stool consistency and frequency were assessed over three consecutive days at baseline and post-baseline every week using the Bristol Stool Formation Scale (BSFS). The BSFS is a validated 7-point rating scale for use in healthy individuals, ranging from type 1 (hard stools) to type 7 (watery stools), with types 3–5 considered within the normal range [40]. For all analyses, mean values for the weekly 11-point NRS symptoms scores, BSFS stool consistency scores, and stool frequency counts were calculated from the 3-day recordings.

Study feasibility

Study feasibility was assessed by recruitment (percentage screened participant that were eligible) and attrition (percentage drop-out post-randomisation) rates.

Secondary outcomes

Psychological symptoms, quality of life, perceived well-being, urinary hippurate, and gut microbiota were the secondary outcomes. These were measured at baseline and week 4.

Psychological symptoms

The psychological parameters of depression, anxiety, and stress were assessed using the short-form version of the Depression Anxiety Stress Scale (DASS-21) questionnaire [41] at baseline and week 4.

Quality of life and perceived well-being

Quality of life was assessed using the Quality of Life-Eight Dimensions (AQoL-8D) questionnaire [42] and perceived well-being was assessed using the World Health Organisation Well-Being Index (WHO-5) [43] at baseline and week 4.

Urinary hippurate

Participants collected their first morning void urine at home on the day of their in-person appointment (visits 1 and 2), using a urine collection pot, and refrigerated it until returned to the study team. Upon receipt, the urine samples were aliquoted and stored at −80 °C until analysed for hippurate levels using nuclear magnetic resonance spectroscopy. Details on sample preparation and analysis are included in Additional file 1 (refer to Section 4.6.1).

Gut microbiota

Participants collected stool samples at home in a stool collection pot the day before their in-person appointments (visits 1 and 2) and stored them immediately in the freezer until returned to the study team. Upon receipt, the stool samples were aliquoted and stored at −80 °C until shipped to the Australian Genomic Research Facility for genomic DNA extraction and 16S ribosomal RNA (rRNA) gene sequencing and read processing. The QIAamp DNA kit (Qiagen) was used to extract bacterial genomic DNA. The V3–V4 hypervariable regions of the 16S rRNA genes were amplified by PCR using the 341-forward and 806-reverse primers. The resulting paired end reads were processed using a divpro pipeline and further trimmed and quality filtered using a Cutadapt followed by denoising with DADA2 [44] via the denoise-paired method. Diversity profiling analysis was performed with QIIME2 [45] by assigning taxonomic classification for each Amplicon Sequence Variant (ASV) using a reference database.

Other data

Habitual dietary intake was assessed using a 3-day food record both at baseline and week 4 using a research-specific food recording app (Research Food Diary) and nutritional composition was analysed using the Xyris FoodWorks 10 software. These data were collected to assess participants’ dietary intake during the intervention period. We also collected information on physical activity level using the International Physical Activity Questionnaire-Short Form (IPAQ-SF) [46] and anthropometry (weight and height) at baseline and week 4 (Additional file 1: Table 1).

Adverse events

Adverse events were monitored weekly during the intervention period using a weekly online adverse events questionnaire. All types of adverse events were assessed according to the National Health and Medical Research Council’s Safety Monitoring and Reporting in Clinical Trials Involving Therapeutic Goods.

Sample size

Given the pilot and feasibility nature of this study, a formal sample size or power calculation for treatment effect estimates was not undertaken. The study was designed to evaluate the feasibility and tolerability of the intervention, assess the feasibility of the study, and generate preliminary estimates for a future adequately powered trial [47]. A target sample size of 46 participants (23 per group), allowing for an anticipated attrition rate of 15%, was selected.

Statistical analyses

Baseline characteristics were summarised using descriptive statistics and reported as mean (M) and standard deviation (SD), median and interquartile range, or frequency and percentage, as appropriate. Study outcomes relating to intervention feasibility and study feasibility were reported using mean (SD) or frequencies and percentage.

Generalised estimating equations models were used to estimate the between-group differences over the study period for outcomes with repeated measurements, including intervention tolerability (GI symptoms and stool consistency). These models included the study group (intervention vs control), timepoint (baseline vs follow-up), and the study group–by–timepoint interaction as covariates. The study group–by–timepoint interaction was used to estimate the between-group difference in change from baseline to post-baseline time points. Analyses were conducted according to the intention-to-treat principle.

Additionally, for objective outcomes (i.e. gut microbiota and urinary hippurate), a separate per-protocol analysis was performed by including only participants who provided biological samples at both baseline and week 4. All statistical analyses were performed using R version 4.4.1.

Statistical analyses of the gut microbiota

Pre-processing of ASV data: A series of pre-processing steps was performed as per the recommendations from Callahan et al. [48] to exclude taxa with nonbacterial lineage (i.e. archaea, mitochondria, and chloroplast) and taxa unassigned at both phylum and kingdom levels. Alpha diversity metrics (i.e. observed species, Shannon, ACE, Simpson, inverse Simpson, and Fisher) were calculated using this pre-processed data (phyloseq package version 1.48 [49]). Prior to computing beta diversity metrics (i.e. weighted and unweighted UniFrac distances and Bray-Curtis distance: vegan package version 2.6–6.1 [50]) and conducting relative abundance testings, an additional prevalence filtering step was followed to exclude low-prevalence taxa (present in less than 1% of the samples) from the pre-processed data [48]. Further, the prevalence-filtered data was agglomerated to the genus level (due to limitations of 16S sequencing) and centred-log ratio transformed, before conducting differential abundance testing.

The difference in the change in alpha diversity from baseline to week 4 between the supplement and placebo groups was examined using a generalised estimating equations model. The difference in the change in beta diversity from baseline to week 4 between the supplement and placebo groups was computed using permutational multivariate analysis of variance (PERMANOVA; 999 permutations). Principal coordinates analysis was used to visualise the results.

Differential abundance analyses at the genus level were conducted using generalised estimating equations models, with adjustment for multiple comparisons using the Benjamini-Hochberg false discovery rate (FDR) procedure. Given the large number of taxa examined, genera demonstrating evidence of differential abundance at a nominal p value threshold of <0.05 were selected. For these taxa, effect estimates with 95% confidence intervals and FDR-adjusted p values (q values) were presented. A q value represents the expected proportion of false-positive findings among results deemed significant after correction for multiple testing. Consistent with approaches used in some microbiome analysis packages [51], and given the exploratory nature of these analyses, q values < 0.25 were considered indicative of a meaningful signal after adjustment for multiple comparisons.

Results

Participant characteristics

A total of 30 individuals were recruited between May and November 2023 and were either randomised to the supplement group (n = 14) or placebo group (n = 16) (Fig. 1). Three participants withdrew post-randomisation (one in the supplement group and two in the placebo group). Baseline characteristics, including demographics, GI symptoms, stool outputs, dietary intake, psychological status, and physical activity levels, were comparable between the supplement and placebo groups (Table 1).

Table 1.

Baseline characteristics

	Supplement group	Placebo group
Agea	49.4 ± 12.0	48.9 ± 10.3
Sex n (%)b
Male	0 (0)	1 (6)
Female	14 (100)	15 (94)
Education n (%)b
University education	11 (79)	15 (94)
High school	3 (21)	1 (6)
Employment n (%)b
Employed	11 (79)	13 (81)
Unemployed	3 (21)	3 (19)
Sleep hours n (%)b
≥8 h	8 (57)	6 (38)
<8 h	6 (43)	10 (62)
Smoking n (%)b
Non-smoking	13 (93)	16 (100)
Smoking	1 (7)	0 (0)
BMI (kg/m2)a	27.5 ± 4.7	25.6 ± 3.0
Physical activity levels n (%)b
Low	8 (57)	11 (69)
Moderate	4 (29)	1 (6)
High	2 (14)	4 (25)
GI symptomsc
Abdominal pain	0.5 (0.0–1.0)	0.5 (0.0–1.0)
Bloating	1.5 (0.0–3.0)	1.0 (0.0–2.0)
Wind	2.0 (1.0–3.0)	1.5 (0.8–2.3)
Nausea	0.0 (0.0–0.8)	0.0 (0.0–0.0)
Stool consistency (type)c	4.0 (3.0–4.0)	4.0 (3.0–4.3)
Stool frequency (per day)c	1.3 (1.0–1.7)	1.3 (1.0–1.8)
DASS-21 item total scorec	59.0 (53.0–69.5)	54.0 (50.0–60.5)
DASS-21–depression scores	18.0 (18.0–21.5)	16.0 (14.0–18.0)
DASS-21–anxiety score	18.0 (16.0–21.5)	16.0 (14.0–18.0)
DASS-21–stress score	26.0 (19.5–28.0)	21.0 (16.0–26.0)
WHO-5 indexc	17.0 (15.3–18.0)	18.0 (14.8–19.0)
AQoL-8Da	64.5 (60.5–70.8)	60.0 (54.8–65.0)
Composition of dietc
Energy (kJ/day)	7249 (5577–8393)	8826 (6448–11,093)
Carbohydrate (g/day)	177 (140–234)	190 (144–239)
Starch (g/day)	107 (78–117)	131 (80–157)
Fibre (g/day)	20 (16–24)	24 (20–37)
Protein (g/day)	66 (56–87)	101 (80–110)
Fat (g/day)	71 (52–85)	89 (58–115)

Data presented as ^a: mean ± SD, ^b: percentage, or ^c: median (IQR)

BMI body mass index, DASS-21 Depression, Anxiety and Stress Scale, WHO-5 World Health Organisation Index (min = 0, max = 25: higher scores representing best quality of life); AQoL-8D Assessment of Quality of Life-Eight Dimensions (presented in unweighted scores, min = 35, max = 176: lower scores represent better health). Physical activity level assessed using the International Physical Activity Questionnaire-Short Form (IPAQ-SF)

Intervention feasibility

At the end of the study (week 4), 86% (n = 12/14) of the participants randomised to the supplement group consumed over 80% of the allocated sachets (i.e. consumed >45 of the 56 sachets provided per participant), demonstrating feasibility of the intervention. Similarly, 88% (n = 14/16) of those randomised to the placebo group consumed over 80% of the allocated sachets. Moreover, the sachet consumption was similar between the supplement (M ± SD: 49 ± 9 sachets) and placebo groups (M ± SD: 49 ± 15 sachets) at the end of the study.

Intervention tolerability

There were no observed differences in the changes in GI symptoms (i.e. abdominal pain, bloating, wind, and nausea) between the supplement and placebo groups over the 4-week period (Table 2 and Fig. 2). There was no observed difference in the change in stool type between the supplement and placebo groups (Table 2 and Fig. 3), except at week 1, when the supplement group demonstrated greater increase in stool type compared with the placebo group.

Fig. 2.

Estimated marginal means and 95% confidence intervals of gastrointestinal (GI) symptoms from baseline to week 4 in the supplement and placebo groups. The y-axis represents the GI symptoms severity, scored on a 11-point simple Numeric Rating Scale ranging from 0 (no symptoms) to 10 (extremely worst/awful)

Fig. 3.

Estimated marginal means and 95% confidence intervals of A stool consistency, with y-axis representing stool types classified according to the Bristol Stool Formation Scale (type 1: hard stools, type 2: sausage-shaped and lumpy, type 3: sausage-shaped with cracks, type 4: sausage or snake-like, type 5: soft blobs with clear-cut edges, type 6: mushy with ragged edges, type 7: completely liquid; B stool frequency of participants

Similarly, there was no observed difference in the change in stool frequency between the supplement and placebo groups (Table 2 and Fig. 3), except at week 1, when the supplement group demonstrated a greater increase in stool frequency compared with the placebo group.

Study feasibility

Study feasibility data included recruitment and attrition rate. Of the 161 participants who were eligible after completion of the online pre-screener, 24% (n = 39/161) were enrolled and underwent baseline assessments. Of these, 76% (n = 30/39) were randomised and 90% of them completed the study (n = 27/30), demonstrating feasibility of the recruitment strategies and study protocol.

Psychological symptoms, quality of life, and perceived well-being

There were no observed differences in the changes in depression, anxiety, stress, quality of life, and perceived well-being between groups at week 4 (Table 3; Additional file 1: Figs. 8–10).

Table 3.

Psychological symptoms, quality of life, urinary hippurate, physical activity, and body weight in the supplement and placebo groups at baseline and week 4, and the between-group differences in change over time using generalised estimating equations models (supplement group n = 14; placebo group n = 16)

	Baseline M (SD)	Week 4 M (SD)	Baseline vs week 4
			Estimate	95% CI
Psychological symptoms
DASS-21 total scores
Supplement	64.14 (16.22)	57.54 (11.02)	−3.18	−11.37, 5.02
Placebo	55.00 (7.83)	51.57 (9.42)
DASS-21 depression score
Supplement	19.43 (5.40)	18.31 (4.89)	0.02	−2.97, 3.01
Placebo	17.00 (3.01)	15.86 (2.77)
DASS-21 anxiety score
Supplement	18.86 (3.66)	16.62 (2.36)	−1.46	−3.45, 0.54
Placebo	16.50 (2.25)	15.71 (1.90)
DASS-21 stress score
Supplement	25.86 (8.47)	22.62 (6.60)	−1.74	−7.07, 3.59
Placebo	21.50 (5.44)	20.00 (6.37)
Quality of life
AqoL-8D score
Supplement	66.64 (11.07)	65.23 (11.34)	−0.66	−5.64, 4.31
Placebo	60.75 (8.41)	60.00 (9.93)
WHO-5
Supplement	16.57 (2.74)	16.46 (3.33)	−0.38	−1.85, 1.10
Placebo	16.88 (2.78)	17.14 (3.66)
Urinary hippurate
Supplement	3.36 (2.84)	2.40 (2.28)	−1.36	−3.51, 0.79
Placebo	2.55 (1.73)	2.96 (1.31)
Physical activity level
Supplement	2.14 (0.66)	2.00 (0.71)	−0.26	−0.80, 0.28
Placebo	1.81 (0.54)	1.93 (0.62)
Body weight (kg)
Supplement	73.61 (13.73)	75.39 (12.46)	0.27	−3.63, 4.17
Placebo	74.56 (9.29)	76.07 (8.62)

DASS-21 Depression Anxiety Stress Scale (total DASS scores represent the sum of the individual DASS depression, anxiety, and stress scores); AQoL-8D Assessment of Quality of Life-Eight Dimensions, WHO-5 World Health Organisation Index-5, M mean, SD standard deviation. The estimate reflects the between-group differences in change

Urinary hippurate

There was no observed difference in the change in urinary hippurate level between the supplement and placebo groups at week 4 (Table 3; Additional file 1: Fig. 14) and these results were unchanged in the per-protocol analysis (Additional file 1: Table 7).

Gut microbiota

Of the 30 participants, 26 provided stool samples at baseline and week 4 (13 in each supplement and placebo group). Three participants provided a stool sample only at baseline due to withdrawal from the study (n = 1 supplement group and n = 2 placebo group) and one participant in the placebo group provided a stool sample only at end of the intervention (due to a nationwide telecommunication outage that hindered sample delivery at baseline), resulting in a total of 56 samples for analysis, in which around 1.4 million reads were assigned to 5772 ASVs. After pre-processing and prevalence filtering, 5744 ASVs remained, belonging to nine phyla, 186 genera, and 199 species.

There were no observed differences in the changes in any of the alpha diversity metrics (i.e. observed species, Shannon, ACE, Simpson, inverse Simpson, and Fisher) between the supplement and placebo groups at week 4 (Additional file 1: Table 8 and Fig. 13). When considering beta diversity, there were no observed group-related changes in unweighted UniFrac (R² = 0.04) or Bray-Curtis (R² = 0.04) distances between groups (Additional file 1: Figs. 16 and 17), whilst a difference in weighted UniFrac distance (R² = 0.12) was observed between groups from baseline to week 4 (Fig. 4; Additional file 1: Table 10). These alpha and beta diversity metrics were similar in the per-protocol analysis (Additional file 1: Tables 9 and 11).

Fig. 4.

The principal coordinates analysis (PCoA) plot of beta diversity assessed by weighted UniFrac distances. The arrow headed lines represents the structural change of the gut microbiota of an individual participant. The base of an arrow corresponds to the baseline gut microbiota, and the tip of the arrow corresponds to the gut microbiota at week 4

Differences in the changes in the abundance of 13 bacterial taxa were observed between the supplement and placebo groups at week 4 (Table 4). The supplement group showed a greater increase in the abundance of six taxa compared with the placebo group (Fig. 5). After FDR correction, two taxa each belonging to the families Anaerovoracaceae and Prevotellaceae remained among the strongest observed signals. The supplement group also showed a greater reduction in the abundance of the seven taxa compared with the placebo group (Fig. 6); however, these signals were not retained after FDR correction (Table 4). In the per-protocol analysis, differences in the changes were observed for nine bacterial taxa between the supplement and placebo groups; however, these signals were not retained after FDR correction (Additional file 1: Table 11).

Fig. 5.

Estimated marginal means and 95% confidence intervals of the relative abundance of taxa demonstrating a differential increase in the supplement group from baseline to week 4 compared with the placebo group before FDR correction (p < 0.05). The Y-axis represents centred log-transformed abundance of bacterial taxa

Fig. 6.

Estimated marginal means and 95% confidence intervals of the relative abundance of taxa demonstrating a differential reduction in the supplement group from baseline to week 4 compared with the placebo group before FDR correction (p < 0.05). The Y-axis represents centred log-transformed abundance of bacterial taxa

Other data

There were no observed differences in the changes in physical activity level, body weight, and dietary composition between groups at week 4 (Tables 3 and 5; Additional file 1: Figs. 11–13).

Table 5.

Dietary composition in the supplement and placebo groups at baseline and week 4, and the between-group differences in change over time using generalised estimating equations models (supplement group n = 14; placebo group n = 16)

Dietary composition	Baseline M (SD)	Week 4 M (SD)	Baseline vs week 4
			Estimate	95% CI
Total energy (kJ/day)
Supplement	9226 (9302)	8321 (2494)	1858	−3850, 7567
Placebo	9555 (5831)	6791 (2162)
Carbohydrate (g/day)
Supplement	248 (293)	212 (70)	40	−147, 227
Placebo	238 (218)	163 (64)
Starch (g/day)
Supplement	148 (190)	111 (39)	−0.90	−108, 106
Placebo	130 (73)	94 (54)
Fibre (g/day)
Supplement	28 (36)	28 (13)	10	−15, 35
Placebo	36 (33)	25 (12)
Protein (g/day)
Supplement	101 (122)	101 (56)	22	−53, 97
Placebo	103 (53)	81 (34)
Fat (g/day)
Supplement	82 (59)	75 (23)	22	−19, 63
Placebo	93 (47)	65 (26)

M mean, SD standard deviation. The estimate reflects the between-group differences in change

Adverse events

In the supplement group, 29% (n = 4/14) participants and in the placebo group, 31% (n = 5/16) participants experienced adverse events. Among these, seven adverse events were related to worsened GI symptoms in the first 2 weeks (supplement group = 3, placebo group = 4). Of these, four adverse events (two in each group) were assessed to be probably related to the intervention or placebo. Out of the total nine adverse events reported, two were deemed unrelated to the study (supplement group = 1, placebo group = 1): consulting a physician for shoulder pain and hospitalisation for diverticulitis, which was ruled by the treating physician as unrelated to the study.

Discussion

We conducted the first pilot RCT to assess the feasibility and tolerability of Musa sp. pseudo-stem core powder as a dietary supplement. Intervention feasibility was achieved with 86% of participants meeting the adherence threshold of 80%. Intervention tolerability was demonstrated by the absence of changes in GI symptoms and normal stool consistency compared with placebo over the 4-week intervention period. Exploratory analysis of the gut microbiota showed some evidence for community level structural differential-changes between the supplement and placebo groups. In the supplement group, a greater increase in the abundance of two bacterial taxa each from the families Anaerovoracaceae and Prevotellaceae was observed at week 4 compared with placebo. These taxa are associated with higher fibre intake [52–54] and butyrate production [55]. This RCT suggests the feasibility and tolerability of Musa sp. pseudo-stem core powder as a dietary supplement with potential of gut microbiota modulation.

Feasibility and tolerability are crucial factors that determine the successful integration and sustained use of functional dietary supplements by individuals. If overlooked, these factors may hinder their application, regardless of their nutritional and therapeutic benefits. The 86% adherence rate observed in our study suggests the feasibility of the daily intake of Musa sp. pseudo-stem core powder as a supplement. This rate aligns with adherence levels reported in previous RCTs utilising microbiome-modulating therapies, such as prebiotic and probiotic supplements [56–58]. Regarding GI symptoms, although no overall differences in the change in GI symptoms were observed between the supplement and placebo groups, mild to moderate adverse events, such as worsened GI symptoms, were reported in both placebo and supplement groups in the first 2 weeks of the intervention by a small number of individuals but subsided during the last 2 weeks. These findings are consistent with other studies that have demonstrated mild changes in GI symptoms in response to microbiome modulation with prebiotic fibres and are likely due to a mild increase in fermentative gas production by the microbes [56, 58, 59].

The weighted UniFrac distance analysis indicated a differential change in the gut microbiota composition between the supplement and placebo groups. In both the intention-to-treat and per-protocol analyses, the supplement group showed signals of an increased abundance of bacterial taxa belonging to the families Anaerovoracaceae and Prevotellaceae, which are associated with high-fibre diets, the fermentation of plant polysaccharides [52–54], and SCFA production [55]. Similar patterns were observed for the genus Faecalibacterium [60], which has been linked to improved depressive and anxiety symptoms. These microbiota changes may be attributed to the fibre and polyphenol content of the supplement. However, the small sample size limits the ability to meaningfully interpret differences in bacterial taxa between groups, and therefore these findings should be considered exploratory. Notably, the significance of these findings—except for the two taxa in the Anaerovoracaceae and Prevotellaceae families—was attenuated after FDR correction, likely reflecting the exploratory nature of the study and small sample size.

We also analysed hippurate, a marker of bacterial metabolism that correlates with dietary intake of polyphenols [33, 61], and found no evidence of a difference in the change in urinary hippurate concentration between groups. This outcome may indicate a lack of true effect of our supplement on hippurate levels or could be attributed to the small sample size. Additionally, the variability in hippurate concentration due to fluctuations in daily urine volume, which was not normalised to a stable molecule like creatinine (known for its consistent excretion rate [62]), represents a limitation of our analysis.

Strengths and limitations

Our study had strengths. First, we deployed a robust triple-blind, placebo-controlled design, including blinding of the researcher leading the analysis, to mitigate any potential bias that could arise during participant allocation, intervention delivery, outcome assessment, and data analysis. The inclusion of a placebo comparator strengthened the internal validity of the study by enabling differentiation between intervention-specific effects and changes that may have occurred due to expectation or natural variation over time [63]. Second, we demonstrated high adherence rates in both groups. This may be due to the high tolerability and acceptability of the product. Our sensory evaluation component, embedded during the supplement development phase, likely facilitated this high acceptability. This evaluation of the intervention’s sensory properties across various foods, including cereal, pasta, soup, and juice, informed optimal mixing proportions and supported the development of a participant booklet that guided integration of the supplement into habitual dietary practices. This may have facilitated the seamless incorporation of the supplement into participants’ habitual diets, prompting high adherence.

However, several limitations should be considered when interpreting the results of this study. Although the study was designed to include both female and male participants, only one male participated, limiting the generalisability of the findings related to the feasibility and tolerability of the supplement. Of note, we were unable to reach our target numbers for this study (n = 23 in each arm) due to a stringent timeline and budgetary limitations. The dose of the intervention was based with an emphasis on minimising adverse effects on GI symptoms (not focused on instigating physiological changes), since the Musa sp. pseudo-stem core powder was being tested as a supplement for the first time in a RCT in humans. Whilst there was evidence from our study that this dosage instigated some changes in the gut microbiota at both the community and individual taxa levels, these findings should be interpreted cautiously given the small sample size, exploratory nature of the analyses, and the potential for less precise confidence intervals arising from using generalised estimating equations models in a small sample. Future RCTs are needed to test clinical efficacy and, if efficacious, to determine if gut microbiota modulation is central to the underlying mechanisms.

Conclusion

This pilot trial demonstrated the feasibility and tolerability of powder derived from the core of the Musa sp. pseudo-stem as a daily dietary supplement with gut microbiota-modulating potential. Future adequately powered trials are warranted to evaluate its impact on gut and mental health outcomes.

Authors’ contributions

HA: conceptualisation (lead), data curation (lead), formal analysis (lead), methodology (lead), project administration (lead), funding acquisition (lead), resources (lead), product development and sensory evaluation experiments (equal), microbiome analysis (lead), writing the original draft (lead), writing—review and editing (lead); ML: formal analysis (supporting), writing—review and editing (supporting); MOH: microbiome analysis (supporting), writing—review and editing (supporting), visualising (lead); MB: investigation (supporting), trial methodology (supporting), writing—review and editing (supporting), supervision (supporting); RK: product development and sensory evaluation experiments (equal); SL: product development and sensory evaluation experiments (equal); SG: product development and sensory evaluation experiments (equal); HS (co-senior author): investigation (supporting), trial methodology (supporting), writing—review and editing (supporting); FNJ (co-senior author): supervision (supporting), investigation (supporting), trial methodology (supporting), writing—review and editing (supporting). All authors approved the final version of the article, including the authorship list.

Funding

This study was funded by an Executive Deans Postdoctoral Fellowship grant awarded to HA and Deakin University IMPACT seed grants. HS is supported by a National Health and Medical Research Council (NHMRC) Investigator Grant (APP2018118). MB is supported by a NHMRC Leadership 3 Investigator grant (GNT2017131). FNJ is supported by a NHMRC Investigator Grant (#1194982). She has received: (1) competitive Grant/Research support from the Brain and Behaviour Research Institute, the NHMRC, Australian Rotary Health, the Geelong Medical Research Foundation, the Ian Potter Foundation, and The University of Melbourne; (2) industry support for research from Meat and Livestock Australia, Woolworths Limited, the A2 Milk Company, Be Fit Foods, and Bega Cheese; (3) philanthropic support from the Fernwood Foundation, Wilson Foundation, the JTM Foundation, the Serp Hills Foundation, the Roberts Family Foundation, and the Waterloo Foundation; and (4) travel support and speakers honoraria from Sanofi-Synthelabo, Janssen Cilag, Servier, Pfizer, Network Nutrition, Angelini Farmaceutica, Eli Lilly, Metagenics, and The Beauty Chef. She is on the Scientific Advisory Board of the Dauten Family Centre for Bipolar Treatment Innovation and Zoe Limited. FNJ has written two books for commercial publication.

Data availability

Data will be made available upon request to the corresponding author.

Declarations

Competing interests

FNJ has received industry support for research from Meat and Livestock Australia, Woolworths Limited, the A2 Milk Company, Be Fit Foods, and Bega Cheese, and travel support and speakers honoraria from Sanofi-Synthelabo, Janssen Cilag, Servier, Pfizer, Network Nutrition, Angelini Farmaceutica, Eli Lilly, Metagenics, and The Beauty Chef. She is on the Scientific Advisory Board of the Dauten Family Centre for Bipolar Treatment Innovation and Zoe Limited. FNJ has written two books for commercial publication. All other authors declare no competing interests.