Protocol for the Nutrition OUtReach In Systems of Healthcare (NOURISH) trial: a phase II multicentre, parallel-group randomised controlled trial of medically tailored groceries to prevent malnutrition in transplant and cellular therapy patients

Abstract

Introduction

Food insecurity (FI) is a health-related social risk associated with malnutrition, morbidity and increased health expenditures in patients undergoing cancer treatment. Among transplant and cellular therapy (TCT) patients, FI is exacerbated by the financial toxicity associated with treatment costs. The Nutrition OUtReach In Systems of Healthcare (NOURISH) trial is a multicentre randomised controlled trial designed to evaluate the impact of in-clinic delivery of medically tailored groceries (MTGs) on malnutrition risk, clinical outcomes and healthcare costs among food insecure TCT patients.

Methods and analysis

We will recruit 210 food-insecure patients from four sites across the USA to be randomised to receive an MTG bag (intervention) or information on food bank resources (control). We hypothesise that the intervention will decrease malnutrition risk (primary endpoint) by mitigating FI, improving clinical outcomes and reducing healthcare costs while patients are receiving TCT. We will assess malnutrition by using the Global Leader in Malnutrition criteria. A χ² test will examine the null hypothesis that 100-day malnutrition rates in the intervention and control arms are equal, using a two-sided 0.05 significance level. Trajectories over time of calorie intake and protein intake will be compared using mixed linear models.

Ethics and dissemination

This study used human participants and was approved through the University of Kansas Medical Center Institutional Review Board (STUDY00161189). All participants provided written informed consent prior to participation in the study. We plan to disseminate results to study participants, participating institutions, peer-reviewed medical journals and national organisations.

Trial registration number

NCT06802406.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• Pilot testing and participant feedback inform site-specific protocol adaptations, improving feasibility and implementation across settings.

• Standardised nutrient analysis of the medically tailored groceries ensures nutritional consistency minimising confounding variables for the primary endpoint and allowing for the flexibility of included foods.

• Stratified randomisation by transplant and cellular therapy type accounts for variability in clinical course; correlative studies will also explore microbiome changes as a potential mechanism for improved outcomes.

• Limitations include variability in clinical courses, potential need for enteral nutrition, difficulty assessing adherence and potential for limited generalisability.

Introduction

In 2023, 13.5% of US households experienced food insecurity (FI), defined by lack of consistent access to adequate foods for every person in a household.¹ Risk of FI is negatively associated with income, employment, race/ethnicity and disability.² FI may force individuals with chronic illness to choose between food and healthcare, contributing to adverse outcomes in chronic disease management, including cancer.^3–6

The prevalence of FI increases to 17%–55% in patients with cancer and is driven by the financial toxicity of treatment.^{7 8} Compared with other countries, the out-of-pocket cost of cancer treatment in the USA is high, with patients and caregivers spending between US$180 and US$2600 per month compared with US$4–US$609 spent in Canada, Australia and Western Europe.⁹ Increased cost for healthcare can further increase the risk of FI. In cancer care, the combination of FI with treatment side effects (eg, nausea, vomiting and anorexia) can lead to unintended weight loss and malnutrition, estimated to contribute to 20% of cancer deaths.^{7 10} Failure to address FI may exacerbate health disparities and worsen outcomes in cancer care.

We have found that FI and malnutrition are concerns in patients with leukaemia, lymphoma, multiple myeloma and other haematological malignancies who undergo procedures such as haematopoietic stem cell transplantation (HCT) and chimeric antigen receptor T-cell therapy (CAR-T) (collectively known as transplant and cell therapy or TCT). HCT can be autologous, using the patient’s own stem cells, or allogeneic, using stem cells from a donor. Before transplantation, patients undergo conditioning therapy consisting of chemotherapy and/or radiation.¹¹ Following conditioning, stem cells are infused and with engraftment occurring within approximately 30 days. CAR-T therapy involves collecting a patient’s T cells and genetically engineering them to recognise specific tumour antigens. The modified cells are then reinfused to target and eliminate cancer cells. Both HCT and CAR-T therapy are associated with treatment-related toxicities and side effects that can increase nutritional challenges such as gastrointestinal complications, anorexia and fatigue.^{11 12} When compounded by FI, these treatment-related challenges may further increase the risk of malnutrition, as demonstrated in our previous studies.¹³

Malnutrition is a significant risk factor for complications such as delayed immune reconstitution, infections and graft-versus-host disease.^14–16 Our previous work showed that patients who reported FI before HCT were more likely to develop malnutrition, require total parenteral nutrition and report financial toxicity compared with those without FI. FI was also associated with impaired psychological and functional outcomes at 1 year post-HCT.¹³ Furthermore, TCT patients may be temporarily relocated closer to the cancer centre due to high risk of infectious complications. This displacement may serve as an additional barrier to food procurement and preparation for a population already experiencing FI. These findings highlight the need for interventions that address FI and reduce malnutrition in this population.

To combat FI and reduce health disparities in patients with cancer, both screening and mitigation programmes are necessary.^{17 18} Existing interventions include Supplemental Nutrition Assistance Programme and meal delivery services which can reduce healthcare utilisation and healthcare costs,^17–24 but participation is limited by stigma, racial discrimination, transportation issues and operational hours.^{25 26} Home delivery services such as Meals on Wheels may reduce some barriers but often cost 2–3 times traditional food bank options.^{23 27 28} As a result, many eligible individuals continue to face food access barriers.

Food is Medicine (FIM) interventions represent a growing area of interest in addressing FI through tailored nutrition provisions. These interventions aim to improve health outcomes by aligning food provision with medical needs.^29–31 One such approach is the use of medically tailored groceries (MTGs), which are treatment-specific grocery items selected by a registered dietitian nutritionist (RDN). MTGs are typically collected at a clinic or community point and prepared at home by patients who are food insecure and managing diet-related conditions.²⁹ Unlike general food assistance, MTGs provide personalised and therapeutic nutrition beyond general food assistance.

There is a need for further research and implementation of evidence-based FIM interventions to address FI and associated health disparities.^{30 32} To ensure sustainability, these interventions must demonstrate clinical significance and cost-effectiveness. Although literature has shown that FIM programmes can reduce healthcare expenditures and improve chronic disease management, evidence in cancer populations is limited.^{23 33 34} Our study focuses on TCT because these therapies are among the most complicated and expensive procedures in cancer care,³⁵ and are often reimbursed through bundled payments. Under these payment models, reductions in complications and overall healthcare costs can improve health system financial outcomes, encouraging the implementation of cost-effective and clinically beneficial interventions.

The Nutrition OUtReach In Systems of Healthcare (NOURISH) is a multicentre, randomised controlled trial that aims to establish partnerships between National Cancer Institute (NCI)-designated Comprehensive Cancer Centers quaternary cancer centres (QCCs) and local FBs to deliver MTGs to patients directly in-clinic and assess its impact on malnutrition, clinical outcomes and healthcare expenditure. QCCs are designated centres that go beyond tertiary services to provide specialised care and run investigative trials such as ours. These centres are recognised as critical leaders for research and innovative patient care. To our knowledge, NOURISH represents one of the first MTG interventions in a cancer care setting. By leveraging health system and community collaboration, we will evaluate if MTGs can reduce malnutrition rates and improve clinical outcomes in food-insecure TCT patients.

Methods and analysis

Study design and objectives

NOURISH is a multicentre, two-arm parallel group, randomised controlled trial taking place at four geographically diverse QCCs: the University of Kansas Cancer Center (KUCC), Duke Cancer Institute (Duke), Stanford Cancer Institute (Stanford) and Memorial Sloan Kettering Cancer Center (MSKCC); their respective FB partners include Harvesters, Food Bank of Central and Eastern North Carolina, Second Harvest and South Asian Food Bank. Our aim is to develop MTGs in partnership with TCT patients, registered dietitians, FB staff and clinical teams. Following the randomised controlled trial, we will evaluate effects on peri-TCT malnutrition, clinical outcomes and dietary quality through 1-year post-transplant, as well as healthcare costs to assess the long-term sustainability of cancer centre and FB partnerships. Enrolment began in May 2025 and is expected to continue until May 2029 (see figure 1 for study aims).

Figure 1. Study aims. NOURISH is an innovative clinical-community partnership programme to reduce food insecurity and malnutrition in patients receiving TCT. Aims include: (1) determine medically tailored grocery contents by partnering with TCT patients, caregivers, clinical teams and food bank programme managers to adapt medically tailored groceries based on patient feedback and local dietary preferences, while ensuring nutritional guidelines for TCT patients are met. (2) Determine the impact of the intervention on peri-TCT malnutrition, clinical outcomes dietary quality through 1-year post-TCT. (3) Evaluate the impact of the intervention on healthcare costs to facilitate long-term sustainability between QCC and FB partnerships. FBs, food banks; NOURISH, Nutrition OUtReach In Systems of Healthcare; QCCs, quaternary cancer centre; RCT, randomised controlled trial; TCT, transplant and cellular therapy.

Eligibility criteria

Inclusion criteria

• Planning to receive TCT.

• Screen positive using Hunger Vital Sign validated screening tool questions³⁶:

  1. ‘Within the past 12 months, you worried that your food would run out before you got money to buy more’.

  2. ‘Within the past 12 months, the food you bought just did not last and you did not have money to get more’.

• Age 8–80.

• Able to read/write English or Spanish (many patient-reported outcome measures lack validated translation in other languages).

• Able to demonstrate understanding of the study with verbal confirmation and willing to sign informed consent.

Exclusion criteria

1. Patients who do not tolerate oral nutrition at the time of study enrolment.

Recruitment, screening, consent

All patients preparing for TCT at participating institutions will be screened for participation in the study. Potential participants will be identified through treatment lists in the electronic medical record (EMR) or internal databases at each study site. Study coordinators will identify patients who have previously screened positive for FI and receive a referral from clinicians such as social workers, nurses or physicians. Potential participants should be screened for FI as a part of standard care throughout clinical course but can be screened again for the purpose of study recruitment. Coordinators will then review patients’ scheduled TCT workup dates and, with approval from clinical teams, approach patients during their clinic appointments. During the visit, coordinators will review consent forms, provide opportunities for participants to ask questions and answer any questions raised. Before providing written informed consent, participants verbally confirm understanding of the study. There will be a separate and optional consent for exploratory outcomes in the study. For sites recruiting paediatric patients, the study team will obtain assent from minors and informed consent from their parents or guardians.

Randomisation

After consenting, patients will be randomised 1:1 using permuted block randomisation to ensure rolling balance between study arms to receive either the intervention (MTGs+information on local FBs) or control (information on local FBs only). Randomisation will be stratified by (1) site (KUCC, MSKCC, Duke, Stanford), (2) type of TCT (autologous HCT, allogeneic HCT, CAR-T) and (3) age (adults vs paediatrics). Randomisation will be conducted for all sites through our centralised Research Electronic Data Capture (REDCap) database. REDCap is a password-protected, secure web-based application for managing surveys and data.^{37 38} Blinding of participants and study coordinators is not feasible due to the nature of the intervention; however, the study principal investigators will remain blinded.

Study timeline

Patients will be enrolled at ‘workup’ (day −30, or 30 days before day 0, the day of TCT infusion). Workup represents the start of the pre-TCT phase and extends through day −7 (sign-off). Day −7 represents the start of the peri-TCT phase and extends through return home (day 100). During the peri-TCT phase, patients come to clinic several times per week for close clinical monitoring. If patients live outside a certain radius, typically 30 min, they are required to temporarily relocate closer to the cancer centre due to the risk of infectious complications. Patients in the intervention group will receive MTGs twice a week in clinic during the peri-TCT phase. The peri-TCT phase ranges from 1 to 3 months depending on type of TCT and patient complications. Once medically cleared, patients can return home which marks the end of the peri-TCT phase. During both the pre- and post-TCT phase, patients can live at home, and both trial groups may receive food from a food pantry; the intervention group will continue to receive MTGs at various clinic visits (figure 2).

Figure 2. Study phases. Patients will be enrolled at ‘workup’ (day −30 or around 30 days before day 0, the day of transplant and cellular therapy (TCT) infusion). Workup represents the start of the pre-TCT phase. This phase extends through sign-off (day −7), which is the final check before patients begin pre-TCT chemotherapy and/or radiation. Sign-off through return home (day 100) represents peri-TCT phase, when patients are coming in daily or several times per week to quaternary cancer centre to receive care. During peri-TCT those patients in the intervention group will receive medically tailored grocery (MTG) twice a week in clinic. This period may vary from 1 to 3 months depending on type of TCT and patient complications. After the patient returns home, the study is in the post-TCT phase. During the pre- and post-TCT phase, patients will live at home and may receive food from a local food bank near home (control and intervention), plus receive MTG when they are seen at peri- and post-TCT clinic visits (intervention only).

Information control group

Patients in the control arm will be provided with information on food pantries they can access near their home and the QCC where they receive treatment. In addition, the control group will also have access to a centralised website where they can go to find nutrition resources and education for people with cancer and study contact information. Per standard clinical care, all patients have access to RDNs for individualised nutrition assessment and counselling.

Intervention group

In addition to receiving MTGs, participants in the intervention group have access to a centralised website that includes information on local food pantry locations, video cooking demonstrations, recipes featuring items from the MTGs and educational materials. These materials cover topics such as using the MTGs while managing comorbid conditions (eg, diabetes, hypertension), food safety, cancer-related symptoms and treatment side effects. Participants will receive MTGs twice a week during the peri-TCT phase. After this period, the frequency of MTG delivery will be reduced and aligned with participants’ clinical visits, once a month or less depending on visit frequency. MTG contents will be tailored to individual dietary needs such as allergies, medical or dietary restrictions, and cultural or religious preferences. While the nutrient and food group composition will be standardised across bags and study sites, flexibility will be allowed based on FB inventory. See the Medically tailored grocery bag design and delivery section for further details.

Participant feedback and retention

Due to close clinical monitoring of patients, it is anticipated that participants will rarely be lost to follow-up. Study visits are aligned with routine clinical appointments to minimise participant burden. Throughout the study, up to four participants per site will be invited to participate in 30 min, semi-structured interviews conducted around day 30 and/or day 100. These timepoints represent two natural breakpoints in treatment allowing for unique perspectives on participant satisfaction throughout the study’s progression. The interviews will include open-ended questions designed to gather feedback on participants’ likes/dislikes related to study participation and related procedures. Interview guides will be specific to group assignments and type of cellular therapy received. The results will be used to inform the study advisory boards about potential modifications to the study protocol, including food bag contents and delivery methods, to promote participant satisfaction.

Medically tailored grocery bag design and delivery

MTG bags were developed by research RDNs with input from clinical RDNs, patient advocates and community partners. Each QCC-FB site pair will tailor food selections to regional and cultural dietary preferences. Using Nutrition Data System for Research³⁹ (NDSR), all MTGs will undergo nutrient analysis to ensure standardisation across bag variations. MTGs will be tested with pilot patients before the full randomised controlled trial implementation, and participants can provide feedback and request changes based on preferences or allergens throughout the study. Pilot testing of MTGs using implementation outcome measures⁴⁰ from volunteer patients and input from participant feedback will enable adaptations to included items.

All foods provided will be shelf-stable and in accordance with neutropenic dietary precautions. While recent evidence suggests that neutropenic diets may not reduce infection risk in transplant patients, the use of shelf-stable foods remains a pragmatic, safety-conscious approach during periods of profound immunosuppression.^{41 42} This approach supports food safety, MTG standardisation and sustainability through FB sourcing.

Food will be checked for expiration dates, damaged packaging and compatibility with participants’ allergies or dietary restrictions. Substitutions will be provided by the on-site study coordinator, a licensed RDN. Each food bag will be covered with an unlabelled black fabric (figure 3) to prevent others from seeing its contents and to reduce potential stigma. Staff will assist participants with transporting MTGs to the car as needed. MTG packages are estimated to cost US$20–US$30 each, depending on product prices at the time of purchase. Costs may vary by location and product availability due to supply chain fluctuations. The number of bags distributed throughout the study will be tracked for cost analysis.

Figure 3. NOURISH medical tailored grocery bag. The intervention medically tailored grocery bags (MTG) bags were developed by research and clinical RDNs with input from patient advocates and community partners. Each cancer centre and food bank site pair will customise food selections to local dietary preferences. All bags will undergo nutrient analysis using NDSR to ensure consistency across variations. During the randomised controlled trial, participants will be invited to participate in semi-structured interventions that aim to gather feedback on the MTG bags. Feedback will inform future protocol modifications. Bags will be covered with black fabric to reduce stigma and transported with staff assistance if needed. Each bag weighs ~13 lbs, and participants may receive more than one. KUCC, University of Kansas Cancer Centre; NOURISH, Nutrition OUtReach In Systems of Healthcare.

Patient and public involvement

Patient advocates, pilot patients and community partners contributed to the design of the MTG bags, study assessments and outcome measures through collaboration with the research team prior to initiation of the trial. Feedback throughout the study will be shared with the study steering committee to inform modifications to the protocol. While patients were not involved in forming the research question or primary outcomes, their input has directly shaped the intervention design and implementation.

Endpoints and data collection

Primary endpoint

The primary endpoint of interest is peri-TCT malnutrition and will be assessed by clinical staff or study coordinators using the Global Leader in Malnutrition (GLIM)^{43 44} and confirmed by a trained clinician. The GLIM⁴³ is a diagnostic measure for malnutrition based on the presence of at least one phenotypic and one aetiological criterion (see table 1). In the case of hospitalisation, malnutrition risk will be assessed using the Patient Generated Subjective Global Assessment Short Form (PG-SGA SF)⁴⁵ rather than GLIM as it is more sensitive for capturing rapid changes in nutritional status during active cancer treatment and does not require a physical examination. For paediatric populations, the STRONGkids nutrition screening tool will be used to assess malnutrition, as it has demonstrated validity as both a screening and diagnostic tool.⁴⁶ Regardless of the screening tool used, any participant who meets malnutrition criteria will be referred to a clinical dietitian for confirmation of the diagnosis.

Table 1. Global Leader in Malnutrition diagnostic criteria⁴³.

Phenotypic criteria
Unintentional weight loss (%)	>5% within past 6 months >10% beyond 6 months
Body mass index (kg/m2)	<20 if <70 years (Asia: <18.5) <22 if ≥70 years (Asia: <20)
Muscle mass	Reduced*
Etiologic criteria
Reduced food intake or assimilation	Ingestion ≤50% of needs from 1 to 2 weeks Any reduction for >2 weeks Any chronic GI condition that adversely impacts food assimilation or absorption
Disease burden/inflammation	Presence of acute disease/injury or chronic disease related

^*Reduced muscle mass will be assessed using handgrip strength and gait speed. When patients are unable to complete these assessments, we will use calf circumference using the European Society for Clinical Nutrition and Metabolism (ESPEN) cut-offs.

BMI, body mass index; GI, gastrointestinal.

Secondary endpoint

Secondary endpoints include incidence of infections, incidence of graft-vs-host disease (GVHD), incidence of relapse, treatment-related mortality, overall survival, quality of life,^{47 48} financial toxicity,^{49 50} physical function,^51–53 cognitive function,^{48 54} mental health,⁵⁵ dietary intake and quality,^{56 57} FI,^{1 36} health-related social needs^{58 59} and frailty.^{60 61} We will also gather data to compare cumulative mean measures of medical resource use (days spent in the inpatient hospital, days spent in the outpatient day hospital, number of hospital readmissions and emergency room visits) and associated costs incurred over the peri-cellular therapy period (start of chemotherapy to discharge home) as well as over the first 3 months between the intervention and control groups. Exploratory endpoints include muscle health, accelerated ageing and impacts on the microbiome (see tables 2 and 3). Full details on outcomes and measures can be found in tables 4–6.

Table 2. Schedule of events (adult).

Assessment	Pre-TCT		Peri-TCT					Post-TCT
	Day −30 (±14)	Day −10 (±7)	Day 7 (±3)	Day 14 (±3)	Day 30 (±7)	Day 60 (±14)	Day 100 (±14)	Day 120* (±30)	Day 180 (±30)	Year 1 (±60)
Confirm eligibility	X
Demographics	X
NCI Multi-factor Screener72	X									X
USDA AFSM73	X									X
CMS AHC HRSN58 59 74	X									X
Function
FACIT-COST49 50	X						X			X
PROMIS-2947	X				X		X			X
EQ-5D-5L75 76	X				X		X			X
6MWD51–53	X				X		X			X
PHQ-977	X				X		X			X
MoCA54	X				X		X			X
Fried Frailty60	X				X		X			X
Short Physical Performance Battery78 79	X				X		X			X
Diet and nutrition
GLIM43 61	X	X	X	X	X	X	X	X	X	X
24-hour Dietary Recall80	X	X	X	X	X	X	X	X	X	X
Survey on food bank utilisation and satisfaction	X	X			X	X	X	X	X	X
Exploratory
Optional plasma and PBMCs	X	X	X	X	X	X	X	X	X	X
Optional muscle ultrasound	X	X	X	X	X	X	X	X	X	X
Optional stool (for microbiome and FoodSeq)	X	X	X	X	X	X	X	X	X	X
Optional semi-structured interview					X		X

^*Timepoint is required for allo-HCT patients only, other transplant and cell therapy types will not be required to complete this timepoint but can for additional compensation.

CMS AHC HRSN, Center for Medicare and Medicaid Services Accountable Health Communities Health-related Social Needs; EQ-5D-5L, EuroQol 5-dimension Level 5; FACIT-COST, Functional Assessment of Chronic Illness Therapy Comprehensive Score for Financial Toxicity; GLIM, Global Leader in Malnutrition; MoCA, Montreal Cognitive Assessment; 6MWD, 6-Minute Walking Distance; NCI, National Cancer Institute; PBMC, Peripheral Blood Mononuclear Cells; PHQ-9, Patient Health Questionaire-9; PROMIS-29, Patient Reported Outcome Measures-29; TCT, transplant and cellular therapy; USDA AFSM, US Department of Agriculture Adult Food Security Module.

Table 3. Schedule of events (paediatrics).

Assessment	Pre-TCT		Peri-TCT					Post-TCT
	Day −30 (±14)	Day −10 (±7)	Day 7 (±3)	Day 14 (±3)	Day 30 (±7)	Day 60 (±14)	Day 100 (±14)	Day 120* (±30)	Day 180 (±30)	Year 1 (±60)
Confirm eligibility	X
Demographics	X
NCI Multi-factor Screener72†	X									X
Hunger Vital Signs36	X									X
CMS AHC HRSN58 59 74†	X									X
FACIT-COST49 50†	X						X			X
PROMIS-2548†	X				X		X			X
EQ-5D-Y-3L81	X				X		X			X
PHQ-9 Modified55†	X				X		X			X
PROMIS 7A Cognitive48†	X				X		X			X
1-Minute Sit to Stand78	X				X		X			X
STRONG Kids Nutrition Screener82	X	X	X	X	X	X	X	X	X	X
24-hour Dietary Recall80†	X	X	X	X	X	X	X	X	X	X
Survey on food bank utilisation and satisfaction†	X	X			X	X	X	X	X	X
Optional plasma and PBMCs	X	X	X	X	X	X	X	X	X	X
Optional muscle ultrasound	X	X	X	X	X	X	X	X	X	X
Optional stool (for microbiome and FoodSeq)	X	X	X	X	X	X	X	X	X	X
Optional semi-structured interview					X		X

^*Timepoint is required for allo-HCT patients only; other transplant and cell therapy types will not be required to complete this timepoint but can for additional compensation.

^†Please note age specifications for this assessment and/or parent/proxy completion requirements.

CMS AHC HRSN, Center for Medicare and Medicaid Services Accountable Health Communities Health-related Social Needs; EQ-5D-Y-3L, EuroQol 5-dimension Youth Level 3; FACIT-COST, Functional Assessment of Chronic Illness Therapy Comprehensive Score for Financial Toxicity; NCI, National Cancer Institute; PBMC, Peripheral Blood Mononuclear Cells; PHQ-9- Modified, Patient Health Questionaire-9 Modified for Teens, Patient Report Outcome Measures 8A Cognitive and 7A Cognitive; PROMIS-29, Patient Reported Outcome Measures-29/25; TCT, transplant and cellular therapy.

Table 4. Primary outcome.

Peri-TCT malnutrition

GLIM is a diagnostic measure for malnutrition based on the presence of at least one phenotypic criterion (unintentional weight loss, low BMI or reduced muscle mass) and at least one aetiological criterion (reduced food intake/assimilation or inflammation/disease burden).43 Participants screening positive for malnutrition will be referred to an RDN, who will do a full nutrition assessment and confirm the malnutrition diagnosis. The PG-SGA SF45 will be used for adult participants completing a timepoint remotely. Additionally, for inpatient participants, the PG-SGA SF can be used in place of the GLIM if necessary. At each assessment timepoint, at least one malnutrition assessment—either the GLIM or the PG-SGA SF—will be completed to ensure that nutrition status is systematically evaluated throughout the study. At sites recruiting paediatrics, this will be evaluated using the STRONG Kids Nutrition Screener.82 All malnutrition diagnoses will be confirmed by RDN

BMI, body mass index; GLIM, Global Leader in Malnutrition; PG-SGA SF, Patient Generated Subjective Global Assessment Short Form; RDN, Registered dietitian nutritionist; TCT, transplant and cellular therapy.

Table 6. Exploratory outcomes.

Muscle health	Muscle health: assessed using muscle ultrasound which quantifies changes in intramuscular adipose tissue, glycogen stores, muscle wasting and sarcopenia at day 100 and year 1. Approximate time frame: 1 year
Accelerated ageing	Accelerated ageing: will collect plasma and peripheral blood mononuclear cells to evaluate biomarkers of accelerated ageing for future analysis. Approximate time frame: 1 year. Measurement tool: biomarker analysis. Home food bank utilisation: percentage of patients reporting use of food bank after intervention. Approximate time frame: 1 year
Healthcare costs	Impact of intervention on healthcare costs: approximate time frame: 1 year. Measurement tools: FACIT-COST, EQ-5D-5L.76 All-cause medical resource use and associated costs of resources to support the NOURISH intervention. Preference-weighted measure of health status using EQ-5D-5L and EQ-5D-Y-3L
Microbiome	Impact on participant microbiome (16S and/or shotgun sequencing).86 Approximate time frame: 1 year

EMR, electronic medical record; EQ-5D-5L, EuroQol 5-dimension Level 5; EQ-5D-Y-3L, EuroQol 5-dimension Youth Level 3; FACIT-COST, Functional Assessment of Chronic Illness Therapy Comprehensive Score for Financial Toxicity; NOURISH, Nutrition OUtReach In Systems of Healthcare.

Table 5. Secondary outcomes.

Secondary outcomes
Incidence of infections	Incidence of infections and subtypes of infections (blood stream, respiratory or other) captured through the medical record through day 100
Incidence of graft-vs-host disease	Incidence of acute graft-vs-host disease (GVHD) through day 180 and chronic GVHD captured through the medical record through year 1
Incidence of relapse	Incidence of relapse through year 1 captured through the medical record through year 1
Treatment-related mortality	Treatment-related mortality (Kaplan-Meier) captured through the medical record
Overall survival	Overall survival (Kaplan-Meier) captured as standard of care through the medical record
Quality of life	Quality of life assessed using PROMIS-2947 score at day 100 and year 1. In patients <15 years old this survey will be replaced by both the paediatric and parent/proxy version of PROMIS-2548
Financial toxicity	Financial toxicity assessed using percentage of patients who are positive on FACIT-COST49 50 at day 100 and year 1. At paediatric sites, this survey will be completed by parent or proxy on behalf of the participant
Physical function	Physical function assessed using the average 6-Minute Walk Test51 52 distance and percentage of patients who are below expected at day 100 and year 1. Short Performance Physical Battery (SPPB) will also be used to assess lower extremity function in older persons. In paediatric participants physical function will be assessed by using the 1 min sit to stand78
Cognitive function	Cognitive function assessed using the average MoCA54 score and percentage of patients who are impaired at day 100 and year 1. In paediatric patients ≥15 this assessment will be replaced by self-reported PROMIS cognitive function short form 8A and in participants <15 this survey will be replaced by paediatric and proxy versions of the self-reported cognitive function short form 7A to assess cognitive function48
Mental health	Mental health assessed using the percentage of patients who are positive on PHQ977 at day 100 and year 1. For paediatric patients ≥12 years old this survey will be replaced by PHQ-9 Modified83 and for participants <12 years old this survey will be skipped
Dietary habits	Dietary quality and habits will be assessed using the USDA multiple pass method for 24-hour dietary recalls80 and evaluated using average overall diet quality scores as assessed using the Healthy Eating Index (HEI)−2020, NCI Multi-factor Screener and the WCRF/AICR cancer prevention score.56 62 63 At year 1. For paediatric participants, both the participant and parent/proxy will be present for the recall. Parents or proxy will be asked to fill out the NCI-Multi-factor Screener on behalf of the participant. FoodSeq57 a novel tool to provide objective data on patient’s food consumption will accompany 24-hour recalls in patients who consent to stool samples
Nutrition support	Need for enteral or parenteral nutrition will be tracked throughout the study. Enteral nutrition will be tracked using EMR and regimens will be documented at 24-hour recalls. Use of PN will be gathered from EMRs and used as part of healthcare utilisation for cost analysis
Food insecurity	Impact on food insecurity: derived from the average number of affirmative responses to a 6-item survey. Higher scores indicate greater food insecurity, as measured by the USDA AFSSM short form survey. At year 1. In paediatric participants, this survey will be replaced by Hunger Vital Signs36
Health-related social needs	Health-related social needs will be assessed CMS AHC HRSN 10-item survey tool.74 84 Higher scores represent increased needs. At workup and year 1. For paediatric participants, a parent or proxy will be asked to fill out this survey on behalf of the participant
Fried frailty	Fried frailty85: percentage of patients meeting 3 or more of the diagnostic criteria for frailty at day 100 and year 1. Approximate time frame: 1 year This assessment will be skipped in paediatric participants

AFSSM, Adult Food Security Survey Module; CIBMTR, Center for International Blood and Marrow Transplant; CMS AHC HRSN, Center for Medicare and Medicaid Services Accountable Health Communities Health-related Social Needs; EQ-5D-5L, EuroQol 5-dimension Level 5; FACIT-COST, Functional Assessment of Chronic Illness Therapy Comprehensive Score for Financial Toxicity; GVHD, graft-vs-host disease; HEI, Healthy Eating Index; MoCA, Montreal Cognitive Assessment; NCI, National Cancer Institute; PHQ, patient health questionnaire; PROMIS, Patient reported outcome measures; SPPB, short performance physical battery; USDA, US Department of Agriculture; WCRF/AICR, World Cancer Research Fund/American Institute for Cancer Research.

Data collection and management

Data on clinical outcomes and demographic characteristics—including age, sex, race, ethnicity, education, marital status, employment status and income—will be captured using institutional EMRs. Additional assessments and surveys will be captured prior to and during study-specific visits. Study surveys will be emailed to participants through REDCap^{37 38} for completion prior to their visit. If patients are unable to complete the surveys in advance, they will be provided with an iPad or a paper version of the surveys at their study visit and the data coordinator will be responsible for entering data from paper into REDCap. Dietary recalls will be collected by a trained interviewer from KUCC for all sites and analysed and stored on NDSR.³⁹ Due to the low-risk nature of this study, adverse events will not be formally assessed. The project will be monitored annually by the data safety and monitoring committee and audited by the quality assurance manager team as needed.

Data from the semi-structured interviews will be captured via password-protected videoconference. The interviews will be audio-recorded so that transcripts may be generated for review. The recordings and transcripts will be stored on the University of Kansas Medical Center secure drive for research and only include participants’ study identification numbers so that all primary investigators (apart from the one conducting the interviews) will remain blinded to study arm.

Statistical analysis and data sharing plan

Participant flow through the trial will be tracked, summarised and reported in accordance with the Consolidated Standards of Reporting Trials (CONSORT) guidelines using a standardised flow diagram. This diagram will visually account for all individuals from initial screening through enrolment, randomisation, group allocation, study retention and final analysis. Specifically, the diagram will document the exact number of participants assessed for eligibility, excluded prior to randomisation (with reasons), allocated to each intervention arm, lost to follow-up at each study time point and ultimately included in the primary and secondary outcome analyses

Primary endpoint and power calculation

A χ² test will be used to test the null hypothesis that 100-day malnutrition rates in the intervention and control arms are equal, using a two-sided 0.05 significance level. Since we will adhere to intent-to-treat principles, we will include all randomised subjects in this test. Those subjects who dropped the study before day 100, or who had a missing malnutrition status at day 100, will be considered to have malnutrition for the purposes of intent-to-treat analyses. To examine the robustness of the results, an additional test will include only subjects who provided a 100-day malnutrition status. Given that we found the malnutrition rate in our TCT patients to be 70% in food insecure patients versus 45% with patients who were food secure 45% with patients who were food secure,¹³ we hypothesise that there will be at least a 20 percentage-point difference against the null hypothesis. At this threshold, we have accounted for the potential impact of residual confounding on the observed 25% difference in malnutrition between food-insecure and food-secure patients. Having 95 evaluable patients in each arm would give approximately 80% power to detect this difference (70% control vs 50% intervention) with a two-sided alpha of 0.05. Expecting about 10% dropout or missing data, we will enrol a total of 210 patients. Sample size calculations were performed using PASS 2008 (NCSS, Kaysville, UT, USA). Statistical analyses will be performed using SAS for Windows (V.9.3 or above: SAS Institute, Cary, NC, USA). Statistical analyses will be performed using SAS for Windows (V.9.3 or above: SAS Institute, Cary, NC, USA).

Secondary endpoints

We will use random effects linear models to compare the trajectories of diet quality between groups by using 24-hour dietary recall data obtained at days 0, 7, 14, 30, 60 and 100. Diet quality will be assessed with the Healthy Eating Index (HEI)−2020 score. To avoid bias, RDNs will conduct and score assessments while blinded to treatment arm. We will also compare the 2018 World Cancer Research Fund/American Institute of Cancer Research scores.^62–64 The models will adjust for relevant biological variables including sex, age and TCT type.

The incidence of other clinical outcomes (depression, infections, GVHD, relapse) will also be compared using χ² tests at 0.05 significance levels, while time-to-event outcomes (treatment-related mortality, overall survival) will be compared using Kaplan-Meier methods. Continuous variables (6-Minute Walking Distance, Montreal Cognitive Assessment) will be compared using two-sided t-tests. To explore the robustness of the relationship between malnutrition risk and the intervention, a multivariate logistic regression analysis of malnutrition will be conducted that will test differences between arms adjusting for baseline variables such as type of TCT, disease, age, sex, race, ethnicity, education, income, utilisation of FB resources, weight, body mass index, presence of malnutrition at baseline and hospitalisation. We will also regress continuous outcomes on the intervention vs control groups with and without adjusting for baseline patient characteristics and adherence. Two-sided t-tests will compare Patient Reported Outcome Measures-29 subscales between the two study arms, using Bonferroni correction for multiple comparisons at a 0.05 nominal significance level. Analogous corrected tests will compare the five EuroQol 5-dimension Level 5 (EQ-5D-5L) dimensions.

To determine the impact of the intervention on healthcare resource use and costs, inclusive of intervention-related costs, we will use generalised linear models specified using negative binomial error distributions and log links for resource use and gamma error distributions and log links for costs adjusting for baseline patient characteristics. We will apply mixed-effects linear regression models to test for interactions between the study intervention and time on EQ-5D-5L-derived preference weights across the 1-year follow-up period. EQ-5D-5L preference weights will also be used to estimate within-trial quality-adjusted life-years at the participant level. In addition, because statistical power may be limited and the accepted type 1 error rate (ie, p<0.05) for economic outcomes is not established, we will apply non-parametric bootstrapping to generate 90% and 80% CIs corresponding to the differences in mean resource use, costs and quality-adjusted life-years (QALYs) between the groups. If mean costs and QALYs are higher in the intervention group compared with control and there are no differences in within-trial mortality, we will compute net health benefits (NHB) at the participant level as NHB_i=(QALY_i–(Cost_i/λ)) where λ represents the maximum willingness-to-pay per additional QALY. Then, we will apply linear regression modelling to evaluate whether the study intervention is associated with greater NHBs controlling for baseline characteristics (sex, age, TCT type and QCC site). In our base-case analysis, λ will be assigned a value of $120 000 per QALY. In sensitivity analysis, the value of λ will be varied to represent a range of willingness-to-pay per QALY thresholds.^{65 66}

Semi-structured interview qualitative analysis

Interview transcript content will be reviewed independently by at least two study team members. Qualitative content analyses will be used to identify and interpret common themes among the data.⁶⁷ These analyses will be conducted iteratively to best inform the team of the need for protocol modifications to enhance the participant experience and facilitate recruitment and retention.

Missing data

Missing data in our population is often due to events such as early death or dropout that are usually not related to malnutrition. Nonetheless, accounting for the possibility that these events may be related to malnutrition in some subjects, all subjects with missing malnutrition status at day 100 will be treated as malnourished in the intent-to-treat analyses of the primary endpoint, which will include all randomised subjects, for a conservative approach. Also, we do not expect a reduction in power in the sensitivity analyses that will exclude subjects with missing nutrition status because we accounted for 10% attrition in sample size calculations. No data imputation will be conducted for the longitudinal analyses because random effects models are known to be robust to missing data.

Data sharing

Deidentified individual participant-level demographic, clinical, nutritional, laboratory, patient-reported outcome and selected biomarker data from the trial will be shared through the Vivli controlled-access repository, along with key study supporting documents. Data will be deposited by the end of the study or at publication and made available for at least 10 years under a managed access process that ensures participant privacy, regulatory compliance and appropriate reuse.

Exploratory analysis

Microbiome analyses of community composition and structure at the functional and strain levels will be conducted using shotgun DNA sequencing using pipeline sequences of DNA sheared to roughly 300 bp as per our previous work.⁶⁸ We can process the resulting data using a metagenomic pipeline that maps reads to a database of bacterial genomes from V.3.5 of the Integrated Microbial Genomes system, which also includes annotations to Clusters of Orthologous Groups (COG), Kyoto Encyclopedia of Genes and Genomes (KEGG) and Enzyme Commission (EC) databases for functional analyses.^{68 69} Our statistical analyses will first test for the impact of the intervention using generalised linear mixed models (GLMMs). GLMMs are flexible linear models that do not assume errors are normally distributed and allow us to include both categorical and continuous predictors.⁷⁰ We will test the impact of specific dietary patterns as obtained from FoodSeq. We will apply principal components analysis (PCA) to our microbiome and FoodSeq datasets to collapse covarying dietary species into a smaller set of orthogonal principal components representing distinct dietary patterns. Next, we will again use a GLMM framework where our predictor values are the principal components from the PCA. Power analysis⁷¹ indicates that our cohorts (n=190 evaluable individuals) are sufficiently large to build models that account for up to six distinct dietary patterns or food groups while accounting for several epidemiological variables (effect size d>0.04, ⍺=0.05, power=0.8). Such variables in our GLMM will include intervention arm, sex, age, TCT type and QCC site, since it is possible that relationships between outcomes and diet will vary by sampling site, for example, due to differences in the overall diversity of foods available to each population. We will analyse associations between predictor variables (intervention arm, total plant or animal dietary DNA levels) and response variables such as malnutrition and other clinical outcomes such as hospitalisations and overall survival. We will have data on the use of antibiotics, immunosuppressive medications and cytotoxic therapies to be included as confounders in our microbiome analysis.

Ethics and dissemination

This study uses human participants and was approved through the University of Kansas Medical Center Institutional Review Board as the Single IRB of record. The trial was prospectively registered on ClinicalTrials.gov (NCT06802406; first posted 31 January 2025), prior to the study start date of 1 May 2025. As adjustments are made to study protocol, KUCC will communicate any changes to subsites at monthly, multisite meetings. We plan to disseminate results to stakeholders, including study participants and caregivers, participating institutions, subaward organisations and finally peer-reviewed journals. All participants provided written informed consent prior to participation in the study (see online supplemental file 1). We also plan to disseminate results to national organisations, policy makers and media to decrease stigma of FI and create targeted interventions.