Fibre-rich Foods to Treat Obesity and Prevent Colon Cancer trial study protocol: a randomised clinical trial of fibre-rich legumes targeting the gut microbiome, metabolome and gut transit time of overweight and obese patients with a history of noncancerous adenomatous polyps

Abstract

Introduction

Recently published studies support the beneficial effects of consuming fibre-rich legumes, such as cooked dry beans, to improve metabolic health and reduce cancer risk. In participants with overweight/obesity and a history of colorectal polyps, the Fibre-rich Foods to Treat Obesity and Prevent Colon Cancer randomised clinical trial will test whether a high-fibre diet featuring legumes will simultaneously facilitate weight reduction and suppress colonic mucosal biomarkers of colorectal cancer (CRC).

Methods/design

This study is designed to characterise changes in (1) body weight; (2) biomarkers of insulin resistance and systemic inflammation; (3) compositional and functional profiles of the faecal microbiome and metabolome; (4) mucosal biomarkers of CRC risk and (5) gut transit. Approximately 60 overweight or obese adults with a history of noncancerous adenomatous polyps within the previous 3 years will be recruited and randomised to one of two weight-loss diets. Following a 1-week run-in, participants in the intervention arm will receive preportioned high-fibre legume-rich entrées for two meals/day in months 1–3 and one meal/day in months 4–6. In the control arm, entrées will replace legumes with lean protein sources (eg, chicken). Both groups will receive in-person and written guidance to include nutritionally balanced sides with energy intake to lose 1–2 pounds per week.

Ethics and dissemination

The National Institutes of Health fund this ongoing 5-year study through a National Cancer Institute grant (5R01CA245063) awarded to Emory University with a subaward to the University of Pittsburgh. The study protocol was approved by the Emory Institutional Review Board (IRB approval number: 00000563).

Trial registration number

NCT04780477.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• This study offers a comprehensive analysis of the effects of a high-fibre diet on human microbiota, metabolome and colonic mucosal biomarkers of colorectal cancer (CRC) over a 12-month period, while also evaluating the long-term benefits of nutrition education on reducing obesity and CRC risk at ~3 years.

• Analysis of faecal short-chain fatty acids, indicative of fibre consumption and bile acids, representative of fat intake, aids in monitoring dietary adherence and determining if intensified behavioural changes and fibre supplementation are required.

• Dietary compliance in the intervention group is a potential limitation, and the study is not sufficiently powered to evaluate if the anticipated changes in mucosal biomarkers predict polyp recurrence or malignant transformation.

• Given that this cancer prevention study specifically targets healthy individuals, future results may not be generalisable to patients with cancer.

Background

Colorectal cancer (CRC) is the third most common form of cancer in the USA.¹ Obesity increases the risk of at least 13 cancers, including CRC.^{2 3} Burkitt’s original hypothesis⁴ from 1963 highlights that westernised diseases such as CRC and obesity may result from fibre deficiency from the commercial refinement of foods. Many plausible mechanisms explain why high-fibre diets, especially a high-legume diet (HLD), may reduce CRC risk. First, fibre is fermented by the colonic microbiota to produce short-chain fatty acids (SCFAs). The SCFA butyrate has a remarkable array of colonic mucosal health-promoting, anti-inflammatory and antineoplastic properties.^{5 6} Second, microbiota break down plant cell walls releasing phytochemicals, which also have powerful anti-inflammatory and anticarcinogenic effects.^{7 8} Third, colonic transit is accelerated, reducing contact time with luminal carcinogens, such as heterocyclic amines formed from cooked red meat,⁹ and secondary bile acids (BA), induced by a high-fat diet and synthesised by the colonic microbiota.^{10 11}

A recent randomised controlled feeding study incorporated a 2-week food exchange, where African American subjects from Pittsburgh were fed a high-fibre (~50 g/day), low-fat African-style diet and rural Africans were fed a high-fat, low-fibre western-style diet. Results suggested that within weeks, mucosal and faecal biomarkers of cancer risk responded favourably to the high-fibre diet, with proliferative rates and inflammatory biomarkers decreasing and microbiota composition adapting to increase butyrogenesis.¹²

Our prior research suggests that fibre may reduce cancer risk indirectly by promoting weight loss, improving insulin sensitivity and decreasing inflammation.^{5 13–15} On average, individuals consume a similar weight of food daily; thus, replacing energy-dense foods with lower energy density foods, such as legumes, should potentiate weight control.¹⁶ Legumes are high in resistant starch, insoluble fibre and especially soluble fibre. Therefore, legumes absorb water during digestion, increasing viscosity, encouraging stomach distension and inducing satiation.¹⁷

Fibre-rich diets may also affect other physiological mechanisms important for weight control.^{18 19} Trypsin inhibitors and other bioactive compounds found in legumes (eg, lectins) may directly stimulate cholecystokinin secretion in the proximal intestine to increase satiety.^20–22 SCFA production may play a role in appetite regulation through stimulation of anorexic gut hormones, peptide YY (PYY) and glucagon-like peptide (GLP-1), slowing gastric emptying.^{20 23} Moreover, microbial acetate has been shown to suppress appetite through central hypothalamic mechanisms involving changes in transcellular neurotransmitter cycles.²⁴

Emerging human evidence links the gut microbiome, insulin resistance, inflammation and obesity with adenomatous polyps and colon cancer.^{23 25–28} This study will provide an opportunity to characterise gut motility, microbial changes and metabolome composition profiles that may influence weight loss and have a role in the prevention of adenomas and CRC, providing novel and potentially therapeutic information.

Study aims and outcome measures

The overall aim of this study is to perform a randomised controlled trial of an HLD compared with a control diet in 60 highest-risk middle-aged participants to measure its ability to reduce body weight. Biomarkers of insulin resistance, systemic inflammation, gut transit and colon cancer risk will be included. We hypothesise that restoring the diet with natural high-fibre content, principally with legumes, will lead to a more significant weight loss and improvements in biomarkers associated with colon cancer risk compared with a control diet. See table 1 for a detailed timeline of outcome measures.

Table 1.

Intervention and assessment activities timeline

(W=weekly; M=monthly; B=bimonthly, X=once, N=if needed)
Phase	BL	Intervention (entrées, instruction)						Self-direction			Maintenance			Ext
Month	0	1	2	3	4	5	6	7	8	9	10	11	12	~36
Food provision	W*	W	W	W	W	W	W
Nutritionist (in person or Zoom)		W	W	W	W	W	W	M	M	M	M	M	M	N
Nutritionist (phone, email, text)		W	W	W	W	W	W	B	B	B	B	B	B	M
Assessments
Mucosal biopsy	X						X						X	X
Stool sample for microbiome and metabolomics	X						X						X	X
Faecal fibre consumption biomarkers (SCFA/BA)	X	X	X	X	X	X	X	X	X	X	X	X	X	X
Urine, first AM	X						X						X	X
Gastric emptying test	X						X
Bowel function questionnaire	X						X						X
Body weight	X	W	W	W	W	W	W	M	M	M	M	M	M	X
Waist circumference	X						X						X
DXA	X						X						X
Blood—fasting insulin (mIU/L), blood glucose (mg/dL), hs-CRP (mg/L)	X						X						X	X
Diet recalls (2/time point), run-in	X			X			X			X			X	X
Physical activity assessment	X						X						X	X
Pedometers	X						X						X	X
RISE-Q59, DSQ58	X						X						X
VAS60	X	X	X	X	X	X	X	X	X	X	X	X	X
Adherence contacts		W	W	W	W	W	W	B	B	B	B	B	B	M

*A 1-week run-in diet of control diet entrées and sides at weight maintenance energy level is provided to all participants before baseline assessments and randomisation. Ext=extended follow-up (when a participant has a subsequent colonoscopy).

BA, bile acids; DSQ, Diet Satisfaction Questionnaire; DXA, dual-energy X-ray absorptiometry; hsCRP, high sensitive C reactive protein; RISE-Q, Reasons Individuals Stop Eating Questionnaire; SCFA, short-chain fatty acid; VAS, Visual Analogue Scale.

Methods and analysis

This study is a parallel-arm randomised clinical trial in overweight/obese healthy persons with a history of noncancerous adenomatous polyp(s). Investigators will be blinded to the diet treatment; however, participants may be able to discern which diet they are randomised to. Preportioned entrées will be provided with regular nutrition education sessions with a dietitian.

Patient and public involvement

Patients and/or the public will not be involved in the design, conduct, reporting or dissemination plans of this research.

Participant recruitment

We aim to accrue 60 middle-aged adults (50% male, 50% female) using a combination of targeted advertisements in the Emory Gastroenterology (GI) Clinics and mailings sent to individuals who may be eligible because of their colonoscopy results.

Eligibility criteria

Inclusion

(1) Free-living adults 40–75 years old, (2) body mass index (BMI) 25–40 kg/m², (3) colonoscopy within 3 years that found/removed ≥1 adenoma >0.5 cm, (4) English speaking, (5) ambulatory and (6) able to provide informed consent.

Exclusion

(1) Serious medical condition, (2) history of CRC, bowel resection, polyposis syndrome or inflammatory bowel disease, (3) smoked regularly in the past year, (4) dietary restrictions substantially limiting compliance (5) planning on substantially changing usual exercise behaviour, (6) regular use of medication that may interfere with study procedures and (7) women currently pregnant, breast feeding or planning a pregnancy.

Informed consent

Eligible participants will be invited to an in-person screening at the Clinical and Translational Science Alliance at Emory University (CTSA). After signed informed consent is obtained, we will conduct the standard screening tests required for healthy participants. A separate consent will also be obtained for permission to store biological samples for future studies related to obesity and CRC.

Confidentiality

Confidentiality will be assured by using subject codes rather than personal identifiers. Any electronic data will be encrypted and accessible only with a login and protected password by the study staff. A certificate of confidentiality from the National Institutes of Health (NIH) has been attained. After the study is completed, all data and specimens will be kept secure according to NIH and Food and Drug Administration (FDA) regulations.

Study intervention

Run-in phase

Before randomisation, participants will proceed through a 1-week run-in where foods representative of the control diet will be provided. The run-in helps to standardise conditions and provides confidence in the participant’s ability to adhere to the study protocol.

The study statistician will use the default random number generator in the R Software program V.4.1.3. to allocate subjects to each treatment arm. To conceal the randomisation sequence to eligible participants and study investigators, we will use numbered, opaque envelopes that contain the treatment assignment and allocate men and women separately for enrollment by the study coordinator.

Diet and nutrition education

We will use the Mifflin-St. Jeor equation to estimate energy needs for weight maintenance reducing this value to facilitate weight loss of 1–2 lbs/week (minus ~500–1000 kcal/day).²⁹

Prepared portion control entrées

To all participants, we will provide preportioned entrées for two meals/day in months 1–3 and one meal/day in months 4–6. The HLD group will receive entrees from a menu cycle developed with a standard set of legumes primarily from the Phaseolus vulgaris species (eg, navy, pinto, black, kidney beans) to limit nutrient and phytochemical variability. The diet will contain approximately 250 g of legumes per day (~1 ½ cups cooked). This level will add approximately 30 g of dietary fibre/day from the legume dishes, ensuring a total intake of ~45–50 g/day. Previously, this level reduced colonic mucosal biomarkers of cancer risk within 2 weeks and is associated with minimal colon cancer risk in rural Africans^{12 30} and a reduction in large polyp recurrences in the Polyp Prevention Trial.³¹ The control group will also receive preportioned meal replacement entrées with legumes replaced by lean chicken/meat. Please see online supplemental materials 1, for example, meal plans for each group. All entrées will be prepared, preportioned and stored at the CTSA under the supervision of the bionutritionist. A printed sheet will be provided to record the amount of each entrée eaten. Education about the consumption of ad libitum sides tailored for weight loss was provided with the American Diabetes Associated food lists for weight management serving as a general guide.³² The intake of sides contributing to total energy intake was not controlled by the study to enable evaluation of the role of legumes in promoting control of self-selected food and energy intake.

Self-direction and maintenance

Participants will continue on their respective diets in months 7–9 but will assume responsibility for food preparation. Long-term weight control is associated with frequent self-monitoring (eg, weight checks), replacement of high-energy-density foods with lower energy density alternatives and portion control, among other strategies.³³ Skill-building and behavioural strategies to address the aforementioned behaviours will be incorporated during bionutritionist encounters at the time points outlined in table 1. Participants in both arms received comparable nutrition advice at equal time points with a focus on weight management and action-oriented eating behaviour tips. During months 10–12, participants will interact with study staff monthly for follow-up and support.

Extended follow-up

Twelve months is adequate time to assess changes in mucosal biomarkers of CRC risk, but not for assessing polyp recurrence or cancer development. Consequently, to explore the long-term success of our diet behaviour modification training on weight control, mucosal biomarker suppression, polyp recurrence and carcinogenesis, we will extend the follow-up to the participants’ next routine surveillance colonoscopy. This will be exploratory as the numbers will likely be insufficient to show significant reductions in polyps or cancer but will provide essential data for a definitive large-scale population study aimed at increasing the consumption of plant-based foods and reducing the risk of Westernised diseases should our intervention prove positive. We will ask participants to notify us of the scheduling of surveillance colonoscopies and request information on the size, multiplicity, anatomic location and histology of any polyps or cancer.

Data collection

Anthropometry

Bodyweight, height and waist circumference will be measured using the most recent National Health and Nutrition Examination Survey (NHANES) procedures.³⁴ We will also give participants smart scales, such as the Fitbit Aria (https://www.fitbit.com/global/us/products/scales/aria-air), for weekly home self-monitoring. Changes in body fat, composition and distribution will be assessed using dual-energy X-ray absorptiometry.

Inflammation/insulin sensitivity

Fasting blood samples will be collected, aliquoted and stored frozen at −80°C for analysis of biomarkers of insulin resistance and systemic inflammation (see table 1). Please see online supplemental materials 2 for further details on these methods.

Stool and urine sample collection

Procedures developed in our NIH-supported studies will be used to ensure scientific rigour for collecting and analysing stool and urine samples. All samples will be aliquoted and held at –80°C for future DNA extraction and microbiome and metabolome analysis. Please see online supplemental materials 3 for further details on these methods.

Gut microbiome and metabolome

We will use real-time qPCR to analyse the functional microbial genes responsible for synthesising butyrate³⁵ and secondary BA.³⁶ The effect of our intervention on faecal and mucosal-attached microbes associated with inflammation and neoplastic transformation will also be assessed.^37–41

Global microbiota sequencing will be performed using 16S rRNA gene (16S) sequencing. Genomic DNA extractions will be performed using a bead-beating approach (Qiagen DNeasy Powersoil Kit). Reagent blanks will be included as negative controls, and both cells and genomic DNA from a microbial community of known composition (ZymoBiomics Microbial Community Standards; Zymo Research, Irvine, California, USA) will be included as positive controls. The V4 region of the 16S rRNA gene will be amplified with inline barcoded primers⁴² and sequenced on an Illumina MiSeq platform. Sequences will be deconvolved and processed through an in-house sequence quality control pipeline.⁴³ Taxonomic classification will be performed with the Ribosomal Database Project Naive Bayesian Classifier with the Silva reference database^{44 45} for subsequent statistical analyses.⁴³

Faecal samples will be transferred to Imperial College London for metabolomics analysis.¹ H Nuclear Magnetic Resonance spectroscopy-based global profiling,⁴⁶ together with liquid chromatography-mass spectrometry-based targeted assays (eg, SCFAs, BA, amino acids) will be applied according to in-house developed protocols.^47–50

Fibre consumption biomarkers

In our studies in Africa and the USA, we have noted an association between fibre intake and the ratio of faecal SCFA to BA. A ratio >10 is associated with low cancer risk, while <5 is associated with high risk (p<0.0001). We showed that increasing fibre intake in African Americans was associated with an increase in the ratio from < 5 to >10.¹² The SCFA:BA ratio will be measured monthly in stool samples as a marker of compliance and individual response as described elsewhere.^{51 52}

Gastric motility/intestinal transit time

A wireless capsule motility system (SmartPill Corporation, Buffalo, New York, USA) that consists of an indigestible single-use capsule, a receiver, and display software, will be used to examine changes in intestinal transit time. After an overnight fast, the participant will report to the CTSA and swallow the SmartPill capsule with 50 mL of water. The participant is asked to avoid strenuous exercise, alcohol, smoking and medications that may affect GI motility and record bowel movements, food intake, sleep and GI symptoms. The data receiver and diary will be returned after 5 days for analysis. Gastric emptying time, small bowel and colonic transit time are estimated by measuring changes in pH, pressure and temperature. Diet-related changes in motility will provide physiological information about diet response with implications for weight loss success, microbiome changes and mucosal carcinogen contact time.

Mucosal biopsy biomarkers

To assess changes in mucosal biomarkers associated with CRC risk, participants will undergo an unprepped flexible sigmoidoscopy following an overnight fast at various time points indicated in table 1. It is essential to avoid the use of bowel preps as they could affect the microbiota and mucosa. Biopsies will be taken from the sigmoid colon at the furthest region easily accessed (eg, splenic flexure or transverse colon). Participants will be informed that if their sigmoidoscopy procedures reveal any serious health issues, the information gained will be used for research purposes, and participation in the diet intervention will be stopped. A general health assessment by H&E staining and immunohistochemistry to measure epithelial proliferation by Ki67 staining of proliferative cells,⁵³ epithelial apoptosis by cleaved caspase-3 staining,⁵⁴ and inflammation by counting CD3+ intraepithelial lymphocytes and CD68+lamina propria macrophages will be performed as previously reported.¹² Mechanisms of action of the HLD in inducing changes in proliferation will be investigated by measuring the changes in the microbiome and its metabolome and by measuring the changes in the expression of genes known to regulate host defence, inflammation, cell cycling, apoptosis and DNA repair. As funding allows, we will perform supportive investigations into host genome responses to diet by Affymetrix Human Transcriptome Array for expression profiling at the University of Pittsburgh Genomics and Proteomics Laboratories, backed up by in-house RT² qPCR assays (SA biosciences, Qiagen). Also, of interest is the effect of the diets and their relative butyrogenesis on oncogenic miRNAs, which have been associated with increased proliferation in high-meat low-fibre diets, which was reversed by resistant starch fibre supplementation (30 g/day) in human studies.^{9 55 56}

Diet assessment/compliance and behaviour questionnaires

Two telephone 24-hour recall interviews will be conducted, 1 weekday and 1 weekend at each time point in table 1. The interviews will be conducted using a multiple-pass interview with the automated self-administered 24-hour diet recall method.⁵⁷ We will use questionnaires designed to assess aspects of eating behaviour that may be important for weight management, including diet satisfaction, reasons for meal termination, and hunger and satiety reported about meals and the overall day. The 28-item Diet Satisfaction Questionnaire⁵⁸ and the 31-item Reasons Individuals Stop Eating Questionnaire⁵⁹ will be used at BL, 6 and 12 months. Visual Analogue Scales⁶⁰ allow participants to mark their responses to questions related to hunger and satiety on a line anchored at each end. We will use this tool to evaluate hunger and satiety before and after each meal (six times/day) and at the end of the day.

Physical activity

We will monitor changes in activity level at time points indicated in table 1 to evaluate if group differences could influence weight loss results. Activity will be assessed across several domains (eg, leisure, domestic and exercise) using the leisure time activity survey from the Cancer Prevention Study-3, which provides a continuous indicator of overall activity.⁶¹ Additionally, subjects will be provided with pedometers for use throughout the study and will track and report their steps for 7 days at the time points indicated in table 1. We will ask that participants not make significant changes to physical activity; however, given the study’s duration, physical activity changes may occur.

Adherence

Along with regular meal provision, adherence will be encouraged and monitored through regular contact with the study nutritionists, food record sheets, regular weigh-ins to promote self-monitoring, diet recalls and assessment of fibre intake through calculation of the SCFA:BA ratio.

Data management

The study master database and backup procedures will be designed during the initial research phase. Data will be entered into Research Electronic Data Capture (REDCap)⁶² and sent to the password-protected Emory data manager to organise, compile and clean for statistical analysis. Data transfer will use secure internet protocols and will observe all IRB and Health Insurance Portability and Accountability Act (HIPAA) requirements.

Withdrawal of participants

At any time during the run-in or throughout the study, participants may withdraw by providing the principal investigator with a written and dated notice of that decision. If they leave the study before the final planned study visit, the researchers may ask the participants to report their weights at the time of their originally scheduled visit and to provide the results related to a colonoscopy they might undergo during the duration of the study. Should participants withdraw from the research without their consent (due to pregnancy, significant health issues, etc), they would be notified by the principal investigator or study coordinator. We will not institute formal withdrawal based on recurrent adenoma, as these are not life-threatening events.

Potential risk and benefits to participants

Given their increased risk of CRC, those diagnosed with adenomatous polyps are more likely to benefit from health promotion programmes.⁶³ Foodborne illness is a potential risk for participants; however, our strict inclusion and exclusion criteria and food safety protocol will help minimise the likelihood of occurrence. Also, the high-fibre diet may not be well tolerated in some participants (bloating, flatulence), however, we will increase fibre intake gradually and encourage participants to consume adequate amounts of water. Additional risks include venipuncture and gastrointestinal bleeding after mucosal biopsy, although this occurs at rates of less than 1%.⁶⁴

Statistical considerations

Power

For weight loss, power is based on a difference in the trajectory of weight loss between groups over 6 and 12 mos. A 5% decrease in weight is clinically significant due to improvements in blood pressure, lipid profile and insulin sensitivity.⁶⁵ A significant difference in weight loss between the two groups is considered to be 1.0 kg±0.9 kg. We expect to observe a larger difference at six mos. (eg, 1.5 kg) and a 1.0 kg difference maintained at 12 months. With a final sample of 60 participants, we would have >95% power for a weight loss difference of 1.0 kg or even 0.8 kg at both time points. If weight change in men differs from women by 0.8 or 0.9 kg (±1.27 kg), the power to detect the difference is 67% or 77%, respectively. These numbers are more than sufficient to observe clinically meaningful changes in secondary outcomes. They will also facilitate our global microbiome and metabolic analyses where exact power calculations are not feasible. Based on our previous research, we plan to enrol and randomise 70 participants, allowing for a drop-out rate of ~15%. Additionally, we estimated the power for detecting a difference in mucosal proliferation at month six between the legume-intervention and control groups based on mucosal proliferation (measured by Ki-67 protein), using changes observed in a 2-week fibre supplementation study with African Americans.¹² While a 20% difference between fibre and placebo implies clinical significance, we have at least 80% power to detect a reduction of at least 9.2%. Our anticipated reduction is higher, approximately 10%.

Proposed analyses

Primary analyses will be intention to treat. Differences at baseline between treatment groups for key variables will be assessed, and those that differ meaningfully will be included as suspected confounders in multivariable models. For those who have a missing outcome of interest, we will fit regression models using those without the missing outcome, use multiple imputations and retain them in the analyses.^{66 67} Analyses will be performed using the most recent versions of SAS version 9.4 (SAS Institute, Cary, NC) for body size, blood measures, and transit time and R (R Foundation) for the mucosal biomarkers and microbiome. Statistical significance will require p<0.05.

To evaluate the role of the HLD on weight reduction, we will focus on the longitudinal change in measures of body size during the most intensive intervention phases (BL to 6 months). Primary analyses will initially use the net change in weight and contrast the change at 6 months in the intervention with that in the control group. We will also evaluate longitudinal changes in mucosal biomarkers from baseline to month 6. Secondarily, we will evaluate weight maintenance and tissue marker changes by contrasting intervention-control changes in measures between 6 and 12 months. Subsequently, we will model outcome trajectories throughout the study using mixed linear models with fixed group, time, group-by-time terms, and random subject and time (slope) terms.

For mucosal biomarkers, after appropriate transformation, changes in markers will be assessed by generalised linear models. Since the genes of interest have been specified in advance, control of the false discovery rate in the stated analyses is not necessary. Exploratory analyses will use the method of Benjamini and Hochberg.⁶⁸

We will assess intervention-related inflammation and insulin sensitivity changes by comparing the area under the curve.⁶⁹ In secondary analyses, these results will be analysed using mixed linear models. Mechanisms that may influence the response to the high-legume intervention, such as a change in intestinal transit time and microbiota profiles will also be investigated.

To assess the effect of changes in microbiota on 16S RNAs gene-based analysis of gut microbial composition, the taxonomic profiles will be evaluated with three quantitative approaches that each account for the compositional nature of the data⁷⁰: distance based, abundance based and distribution based. Each approach will be used to understand the microbiota descriptively and to hypothesis–test associations with the clinical data. Intersample distances (eg, beta diversity) will be used to identify biome types (‘enterotypes’) through hierarchical clustering and the elucidation of taxa influencing sample differentiation. Models will be fit using the 16S profiles with distance-based, (PERMANOVA)⁷¹ to associate the microbiota as a response, with clinical variables as predictors. Multinomial log-linear modelling will also be used to associate clinical variables with hierarchical biome clusters to identify clinical phenotypes. An abundance-based approach applies the additive log ratio (ALR) transformation to relative abundances (to break the spurious correlation among taxa in compositional data) to analyse each taxon as independent and normally distributed.⁷² ALR transformed taxonomic abundances will be used as either predictors (multiple regression) or responses (multivariate regression) while controlling for appropriate covariates (eg, sex, age, BMI) in linear models. Based on their positive or negative correlation, ALR taxonomic values will be analysed for correlation, to identify taxa with potential for cooperation (eg, facilitation or syntrophy) or competition (eg, displacement or predation). Distribution-based approaches, such as within-sample or alpha-diversity, analyse the sample profiles as probability distributions. As an index, taxonomic diversity will be analysed in separate linear models as either a response or a predictor variable in association with the clinical variables. In analyses where ALRs or a diversity index is used to represent the microbiome, the microbiome ‘as a predictor’ will be compared against the inverted model where the microbiome is considered ‘as a responder’ while controlling for the same covariates. Due to the potentially large number of predictors that may be correlated, principal component (PC) analysis will be applied hierarchically to groups of related variables to identify PCs that can represent the variance for the dataset or identify variables highly correlated with the identified PCs to serve as proxies for the entire dataset. For repeated measures (time), paired difference analyses will be performed to associate predictors with the degree of differences between the two samples from the same subject. Here the difference in taxonomic abundance, overall composition or diversity between the two samples from the same subject is analysed.⁷³

Finally, in secondary analyses to estimate the effect of treatment if everyone complied (no non-compliers or dropouts), the group assignment will be used as an instrument using methods described elsewhere.⁷⁴ Regression diagnostics, including residual analyses and assessment of autocorrelation patterns, will be used.

Permissions and approvals

The study is also registered with clinicaltrials.gov (NCT04780477). In addition, a material transfer agreement was fully executed.

Study oversight

The Winship Cancer Institute Data and Safety Monitoring Committee (DSMC) is responsible for reviewing pertinent aspects of study conduct, including patient safety, protocol compliance and data collection. Due to the low risk associated with this trial, monitoring will be conducted once within the first year of enrolment for consent and eligibility only. During the initial monitoring visit, 10% of total patient enrolment will be monitored. The DSMC will review monitoring report deficiencies and toxicity data provided by the study team and make recommendations for trial continuation, modification or suspension. The committee reserves the right to conduct additional audits if necessary. The principal investigator or designee is responsible for notifying the DSMC once the trial is open to accrual.

Results dissemination

We will provide early reports and presentations of the findings to the NIH in progress reports. Because this study will demonstrate important results relevant to populations who consume a westernised diet, we will inform the scientific community through professional presentations and peer-reviewed journal articles.

Ethics statements

Patient consent for publication

Not applicable.