Assessing impact of nutrition care by registered dietitian nutritionists on patient medical and treatment outcomes in outpatient cancer clinics: a cohort feasibility study

Abstract

More information is needed about the impact of outpatient nutrition care from a registered dietitian nutritionist (RDN) on patient outcomes. This study aimed to assess the feasibility of a cohort study design to evaluate impact of RDN nutrition care on patient outcomes, describe clinic malnutrition screening practices, and estimate statistical parameters for a larger study. Seventy-seven patients with lung, esophageal, colon, rectal, or pancreatic cancer from six facilities were included (41 received RDN care and 36 did not). RDN nutrition care was prospectively documented for six months and documented emergency room visits, unplanned hospitalizations and treatment changes were retrospectively abstracted from medical records. Most facilities used the Malnutrition Screening Tool (MST) to determine malnutrition risk. Patients receiving RDN care had, on average, five, half hour visits and had more severe disease and higher initial malnutrition risk, although this varied across sites. Documented medical and treatment outcomes were relatively rare and similar between groups. Estimated sample size requirements varied from 113 to 5856, depending on tumor type and outcome, and intracluster correlation coefficients (ICCs) ranged from 0 to 0.47. Overall, the methods used in this study are feasible but an interventional or implementation design might be advantageous for a larger study.

Introduction

Cancer care has largely shifted from the inpatient to the outpatient setting over the last several decades, with about 90% of patients now receiving treatment in the outpatient setting.¹ Depending on tumor type and treatments, 31% to 80% of patients are at risk for malnutrition at the time of their cancer diagnosis or during the course of their treatment due to cancer-related metabolic changes, functional or physical limitations, and/or treatment side effects.^2-4 Malnutrition negatively affects a patient’s ability to undergo and tolerate treatments, and can result in inferior outcomes, including treatment delays and extensions and unplanned hospitalizations and poorer quality of life.^5-8 Overall, patients with cancer who experience malnutrition have higher mortality rates.⁹

Most current evidence on the impact of nutrition interventions for adults with cancer is for patients with gastrointestinal and head and neck cancers and interventions delivered in the inpatient setting for brief periods before and/or after surgery, radiation, or chemotherapy.¹⁰ Given the risk of malnutrition for patients with cancer, integrating registered dietitian nutritionists (RDNs) into outpatient cancer treatment teams may enhance patient outcomes. Several randomized controlled trials (RCTs) outside the United States (primarily in Europe and Asia) have examined the impact of outpatient nutrition interventions (e.g., dietary counseling, recommendations of oral nutrition supplements, and enteral nutrition) delivered or managed by RDNs on outcomes for patients with cancer. Findings from these studies are mixed, with some studies noting significant positive effects on key medical outcomes (e.g., survival, chemotherapy compliance)^11-14, while others did not.^15-19 Similarly, several studies noted positive effects on intake of energy and protein, mitigation of weight and muscle loss and treatment-related side effects, and quality of life^12-14,20-23, while others had mixed findings across these outcomes^15,24-27 or did not find any effects.^{11,16-19,28-30} A few studies noted significant effects of nutrition interventions only for subgroups of patients with severe malnutrition or weight loss at baseline^19,31,32 or for patients with less advanced disease.²⁷ There are a few potential explanations for the mixed findings across studies. The studies were conducted across countries with different health systems and included patients with varying degrees of baseline nutritional status and disease severity and many different types of cancer, although gastrointestinal and head and neck cancers were the most common. Across studies, the interventions had variable intensity, duration, and timing relative to surgery or treatment. Most had comparative effectiveness designs, with control groups receiving usual care that often included some level of access to nutrition care from RDNs or other health professionals, because of ethical concerns about completely withholding access to nutrition care.

In the United States, widespread RDN understaffing in outpatient cancer centers creates the conditions for conducting an observational study (i.e., a natural experiment) to examine the impact of RDN nutrition care, in that all patients who screen at nutrition risk do not receive medical nutrition therapy from an RDN.³³ The primary aims of this study were to: 1) test the feasibility of a multi-site, pragmatic cohort study design for examining the impact of RDN nutrition care on nutrition, treatment, and medical outcomes for patients with lung, esophageal, colon, rectal, or pancreatic cancer treated in the outpatient setting; 2) describe current malnutrition screening practices across participating outpatient cancer centers; and 3) calculate sample size requirements by tumor type and intracluster correlation coefficients (ICCs) that could be used to plan a larger study.

Materials and Methods

Study Design

The design of this feasibility cohort study (Figure 1) is described in detail in the published protocol paper.³⁴ In brief, six outpatient cancer centers were identified that screened all patients for nutrition risk, but did not have all patients who screened at nutrition risk receive medical nutrition therapy from an RDN. At each site, an RDN randomly selected patients who screened at nutrition risk and received medical nutrition therapy (“RDN Care” group) and documented aspects of the patients’ initial and follow-up care into the Academy of Nutrition and Dietetics Health Informatics Infrastructure (ANDHII)³⁵ for six months. At the end of the six-month period, each RDN identified additional patients, matched on tumor type, who screened at nutrition risk but did not receive nutrition care from an RDN (“No RDN Care” group). A medical record review was then conducted at each site for all patients (RDN Care and No RDN Care) for that six-month period to extract information on the number of emergency room visits, unplanned hospitalizations, and treatment changes (i.e., treatment reductions, delays, or discontinuations).

Figure 1.

Study design and flow for the six outpatient cancer treatment facilities with malnutrition screening policies in place that were enrolled in the cohort feasibility study. Registered Dietitian Nutritionists (RDNs); Academy of Nutrition and Dietetics Health Informatics Infrastructure (ANDHII); Emergency Room (ER)

Ethical Approval

The University of New Mexico (UNM) Health Sciences Center’s Human Research Protections Office [UNM HRPO] (#18-173) approved the research protocol. Five participating sites relied on the UNM HRPO; one site conducted local institutional review board (IRB) review and approval of the study protocol.

Site Recruitment

Site recruitment efforts were launched at the biannual Academy of Nutrition and Dietetics Oncology Nutrition Dietetic Practice Group’s (ON DPG) Oncology Symposium in April 2018. Study team members also approached individual RDNs to gauge interest in participating in the study. Additionally, general study information was sent out via the ON DPG listserv. RDNs from interested outpatient cancer treatment centers were invited to attend a webinar in May 2018 providing more detailed study information. Site inclusion criteria were having a policy and procedures in place to screen all patients for nutrition risk, using an electronic medical record, signing a site agreement, and obtaining IRB approval to participate in the study.

RDN Training and Survey

Prior to the start of the study, participating site RDNs completed an online, asynchronous study training (1.5 continuing professional education units) on the study protocol and procedures and use of ANDHII. Upon site enrollment, each RDN completed a 17-question survey, which included questions on their professional qualifications and experience and characteristics of the outpatient cancer treatment centers where they were employed. All participating RDNs also received ongoing training and technical assistance from the study coordinator during study implementation.

Patient Selection

For the RDN Care group, eligible patients were adults residing in the United States who: 1) had a diagnosis of lung, esophageal, colon, rectal, or pancreatic cancer; 2) had active or intended cancer treatment comprised of one or a combination of different therapies (such as chemotherapy, radiation, chemoradiotherapy, immunotherapy, or targeted therapy) in the outpatient setting; 3) screened at malnutrition risk; and 4) received nutrition care from the trained RDN within two weeks of being screened. Patients concurrently enrolled in a clinical trial were excluded. Once the RDNs identified eligible patients, they used the random number generator on random.org to determine if each patient’s nutrition care should be documented in ANDHII.³⁶ RDNs entered a range from 0 to 1 in the random number generator; the patient was included in the study if a “1” was generated and the patient was excluded if a “0” was generated. This step was incorporated to minimize RDN bias in patient selection. Each site aimed to have seven patients in the RDN Care group.

To identify patients eligible for the No RDN Care group, the trained RDN reviewed census lists for the same period during which nutrition care documentation was completed at their center. Each RDN selected patients (goal: 7) matched to the initial patients by primary tumor type, who screened positive for nutrition risk but never received nutrition care from an RDN (i.e., had no RDN note in the medical chart). Some outpatient cancer centers (n=2) were unable to identify seven eligible matching patients in this predefined timeframe. To accommodate this, the study protocol was modified to allow RDNs to evaluate their census for up to six months before the start or after the end of their site’s ANDHII documentation period. All other inclusion and exclusion criteria and the random selection process were the same for the matching patients.

ANDHII Documentation

ANDHII is a web-based platform specifically designed to collect de-identified nutrition care data in the Nutrition Care Process framework using standardized nutrition care process terminology.^37,38 For patients in the RDN Care group, RDNs were trained to document their usual patient care delivered according to center policies and practices, per the Office for Human Research Protections Guidance on Research Involving Coded Private Information or Specimens.³⁹ RDNs were encouraged, but not required, to document specific aspects of nutrition assessment (listed in Figure 1) and other details regarding the nutrition care encounter. RDNs documented the initial patient encounter and any follow-up encounters within a 180-day timeframe. ANDHII documentation for the study was completed between February 2019 and August 2021.

ANDHII documentation was reviewed and cleaned monthly. Data cleaning consisted of internal quality control, verifying that values entered were biologically plausible, monitoring the timely entry of initial and follow-up encounters, and verifying that there was no contradictory data within a patient’s encounter or encounters. When obvious discrepancies or errors were identified, the study team requested that the RDN review and modify the documentation, as appropriate.

Medical Chart Review

The medical chart review was conducted by a licensed nurse who is a professionally trained and experienced abstractor. Although the initial study protocol indicated that the abstraction should cover the time from the first oncology visit through 180 days after the start of treatment for all patients,³⁴ the protocol was modified to cover the time from first screen of “at risk for malnutrition” through 180 days after the patient screened at risk for malnutrition. For many patients, these timeframes were the same, but as part of the feasibility assessment for the study, it was identified that at some sites (n=2) where malnutrition screening policies were recently implemented, patients were identified as being at risk of malnutrition far into their cancer diagnosis and treatment trajectory.

The participating sites provided the medical chart abstractor with remote access to the selected patient medical records via a secure web-based portal according to facility-preferred processes. The data abstractor collected de-identified information on for all patients who were included in the study. Variables collected are summarized in Figure 1.

Abstracted data were collected and managed using the Research Electronic Data Capture (REDCap) tool hosted at the UNM Health Sciences Center.⁴¹ REDCap is a secure, web-based application designed to support data capture for research studies. Data validation and skip logic were included in the REDCap version of the abstraction tool to improve data quality. Medical chart abstraction for the study was conducted from October 2020 to December 2021.

Variable Processing

The TNM (tumor site and size, lymph node involvement, and metastatic spread) Classification of Malignant Tumors is a globally recognized standard for classifying the extent of cancer spread.⁴² In consultation with medical oncologists, cut points were created to classify patient disease as more advanced cancer based on TNM information abstracted from the medical record. Patients with a T3 or T4 tumor, lymph nodes at N2 or N3, or presence of metastasis were considered to have more advanced cancer.

For modeling purposes, the number of treatment changes, unplanned hospital stays, and emergency room visits were recoded to bivariate variables indicating that the outcome did not occur or occurred one or more times.

Feasibility Measures

To assess the feasibility of the study methods, the study team used administrative records to track the following measures: time required to onboard (i.e., obtain a fully executed site agreement and central IRB and site IRB approval) six recruited sites; completion of ANDHII documentation for seven patients per site; completion of medical record review for 14 patients per site; and completeness of variables obtained via ANDHII documentation and the medical record review.

Statistical Analyses

Statistical analyses were conducted using Stata SE 16 (College Station, TX, USA) and SAS/STAT version 15.2 (SAS Institute, Cary, NC). We completed descriptive analyses of site and RDN characteristics and patient characteristics, RDN and medical visits, more advanced cancer, initial to final documented Malnutrition Screening Tool [MST] scores⁴⁰, and emergency room visits, unplanned hospital stays, and treatment changes, stratified by study group, site, and/or tumor type. Data are presented descriptively as means and standard deviations (SD), medians and interquartile ranges (IQR), number of observations and respective percentages, or ranges, as appropriate. Graphs were created in Excel 2019 (Microsoft, Redmond, WA, USA) and Stata SE 16.

The Wilcoxon rank-sum test was used to compare the number of treatment changes, unplanned hospital stays, and emergency rooms visits by study group. Separate multilevel logistic regression models (PROC GLIMMIX) were used to explore relationships between study group (RDN Care or No RDN Care) and bivariate medical outcomes controlling for covariates (age, sex, race, ethnicity, primary tumor type, more advanced cancer) and including site as a random effect.

Estimated sample size requirements for a larger study were calculated by tumor type for each binary medical outcome (no ER visits versus ≥1 visit; no unplanned hospitalizations versus ≥1 hospitalizations; and no treatment changes versus ≥1 treatment changes) using G*Power 3.1.9.4 (Universitat Kiel, Germany).⁴³ Input parameters were: z test family; statistical test: logistic regression; type of power analysis: a priori; tails: two; odds ratio as determined from pilot study data; α=0.05; power = 0.80; R² other than X (amount of variability in main predictor that is accounted for by covariates) as determined from pilot study data.

Intracluster correlation coefficients (ICCs) were estimated using methods outlined by Austin et al.⁴⁴ Design effects were calculated as 1 + (12.8 - 1)*ICC where 12.8 = the average cluster size in the pilot study.

Results

Site Onboarding and Characteristics

Six outpatient cancer treatment sites were recruited and onboarded from April 2018 to January 2020. The time required for each site to obtain a fully executed site agreement and IRB approval is detailed in Figure 2 and varied substantially between sites.

Figure 2.

Site recruitment and patient accrual in the oncology cohort feasibility study. Electronic Medical Record (EMR); Institutional Review Board (IRB); Registered Dietitian Nutritionist (RDN).

¹Site was unable to identify a seventh RDN Care patient during the study period, and therefore only identified 6 matching No RDN Care patients.

²Site prioritized nutrition care for patients with esophageal cancer and was therefore unable to identify matching No RDN Care patients for all esophageal patients in the RDN Care group.

³Site was unable to match all patients due to time constraints.

The six sites that participated in the study were located in Indiana, New Hampshire, Oregon, Washington, and Wisconsin. One site included two facilities and two RDNs - a hub site (the primary cancer treatment facility) and a spoke site (a community hospital that managed aspects of oncology care). Table 1 summarizes key characteristics of the participating facilities. Five facilities were associated with a community hospital and two facilities were free-standing clinics. RDN full-time equivalents (FTEs) for oncology patient care at each site ranged from about 0.5 at sites 2, 4, and 6 to around 2 at Sites 1, 3, and 5. The number of new oncology patients seen yearly at each site ranged from 240 to 2,500 patients, and the ratio of RDN FTEs to number of new patients also varied, ranging from one RDN for every 300 new patients at the Site 3 spoke facility to one RDN for every 2,500 new patients at Site 6. About half of the sites included were rural, and Site 5 was in a medically underserved area. Most facilities reported using the MST to screen for malnutrition. The reported timing of malnutrition screening varied across sites, but most facilities indicated that patients are screened for nutrition risk at regular, predetermined intervals during the course of treatment.

Table 1.

Characteristics of participating outpatient cancer treatment facilities in the oncology cohort feasibility study (n = 7).

Item	Facilities
	1	2	3a1	3b1	4	5	6
Facility associated with
Community hospital	X	X		X		X	X
No association (facility is a free-standing clinic)			X		X
RDN full-time equivalents for oncology patient care	1.8	0.5	1.8	2	0.46	2	0.6
Number of new oncology patients seen per year	775	240	950	600	Unsure	2500	1500
RDN:new patient ratio	1:431	1:480	1:528	1:300	-	1:1250	1:2500
Universal malnutrition screening policy at study baseline
Well-established	X		X	X	X		X
Newly established or updated		X				X
Malnutrition risk screening tool(s) or approaches used 2
Malnutrition Screening Tool (MST)	X	X	X	X	X	X
“Modified” MST3						X
Patient-Generated Subjective Global Assessment (PG-SGA)							X
Provider discretion				X
Timing of malnutrition risk screening 3
At diagnosis or shortly thereafter			X		X	X
At regular, predetermined intervals during treatment	X	X	X			X	X
If the patient experiences side effects during treatment				X

¹Site 3 included two facilities and two RDNs - a hub site (3a; the primary cancer treatment facility) and a spoke site (3b a community hospital that managed aspects of oncology care).

²Respondents could choose all that applied.

³The RDN at Site 5 indicated that the site was using a “modified” version of the MST at the time of the baseline study survey (no modification details provided), but that the facility intended to begin using the validated MST.

RDN Characteristics

The seven RDNs that documented care for the study were experienced practitioners (median [IQR] 13 [7, 22] years in dietetics practice and 5 [4, 15] years specifically in oncology nutrition). Over half had master’s degrees (n=4; 57.1%) and nearly three-fourths were Board Certified Specialists in Oncology Nutrition (n=5; 71.4%). They reported providing nutrition care to a median (IQR) of 8 (4, 10) patients in a typical eight-hour workday.

Patient Characteristics and Treatment Visits

Patient accrual by site and group is summarized in Figure 2. For five sites, patient accrual goals were not met because of time constraints or policies dictating that the majority of patients with esophageal cancer receive RDN care. Patient characteristics are presented by group in Table 2. Patients were on average in their mid-to-late 60s and were primarily non-Hispanic white, with lung being the most common primary tumor type. Overall, there were not substantial demographic differences between the groups, although there was a slightly higher proportion of men in the RDN Care group, and the No RDN Care group was, on average, slightly older. A higher proportion of patients in the RDN Care group had severe disease based on tumor staging and metastasis (Figure 3). However, this observation varied across sites, with a higher proportion of the RDN Care group having more severe tumor stage at two sites (sites 1 and 4) and metastasis present at four sites (sites 2, 3, 4, and 5). TNM information was missing (i.e., not located in the chart) for up to one-third of patients, depending on component (missing for: 22 patients [tumor stage], 23 patients [lymph node involvement], and 18 patients [metastatic spread]).

Table 2.

Characteristics of patients included in the oncology cohort feasibility study by group (n = 77 patients)

Patient Characteristic	RDN Care (N = 41)	No RDN Care (N = 36)
	n (%)	n (%)
Sex
Female	18 (43.9)	18 (50.0)
Male	23 (56.1)	18 (50.0)
Race
White	38 (92.7)	34 (94.4)
Asian	1 (2.4)	2 (5.6)
Native Hawaiian or Other Pacific Islander	1 (2.4)	0
Missing	1 (2.4)	0
Ethnicity
Not Hispanic or Latino	37 (90.2)	35 (97.2)
Hispanic or Latino	2 (4.9)	0
Not listed in chart or unknown	2 (4.9)	1 (2.8)
Primary Tumor Type
Lung	18 (43.9)	18 (50.0)
Colon	7 (17.1)	7 (19.4)
Pancreatic1	6 (14.6)	5 (13.9)
Rectal1	4 (9.8)	3 (8.3)
Esophageal2	6 (14.6)	3 (8.3)
	Mean ± SD3	Mean ± SD3
Age, in years4	64.6 ± 10.9	68.7 ± 8.8
Number of encounters with an RDN	5.2 ± 3.7	Not applicable
Length of encounter with an RDN, in minutes (n = 155 encounters among 33 patients)	26.8 ± 14.3	Not applicable

¹Some sites had time constraints that precluded matching all patients with these tumor types.

²Some sites prioritized RDN care for patients with esophageal cancer and therefore were unable to match that did not receive RDN care for all patients.

³SD – standard deviation.

⁴Age at initial appointment.

Figure 3.

Proportion of patients with more advanced cancer stage by study group in the oncology cohort feasibility study. Cancer staging based on the TNM (tumor site and size, lymph node involvement, and metastatic spread). Classification of Malignant Tumors framework, which assesses the amount and spread of cancer in a patient’s body. More advanced cancer stage is defined as having a T3 or T4 tumor, lymph nodes at N2 or N3, or the presence of metastasis. TNM data were missing in up to one third of the medical charts reviewed.

Table 3 summarizes the number of documented RDN and medical visits for patients included in the study by tumor type. Patients in the RDN Care group (n=41) had an average of five visits over the 6-month time frame. The reported length of those visits averaged just under a half hour. The average number of visits was similar across tumor types. Reported RDN visit length was shortest for patients with colon cancer and longest for patients with esophageal cancer.

Table 3.

Documented RDN and medical care visits for patients included in the oncology cohort feasibility study.

		Lung	Esophageal	Colon	Rectal	Pancreatic
RDN Care (n = 41)	# of patients	18	6	7	4	6
	# of total visits	90	37	29	22	33
	# of visits per patient, mean ± SD	5.0 ± 4.0	4.8 ± 2.8	5.3 ± 3.9	5.5 ± 5.1	5.5 ± 3.4
	Length of visit inminutes, mean ± SD	26.6 ± 13.4	35.4 ± 18.0	18.6 ± 7.8	24.4 ± 16.0	30.5 ± 13.0
Medical Care (n = 77)	# of patients	36	9	14	7	11
	# of visits	650	219	206	230	182
	# of visits per patient, mean ± SD	18.1 ± 13.9	24.4 ± 9.6	14.7 ± 9.6	32.7 ± 18.4	16.5 ± 7.9
	Treatments1	399	138	87	147	71
	Surgery	2 (1%)	5 (4%)	8 (9%)	3 (2%)	1 (1%)
	Radiation Therapy	250 (63%)	94 (68%)	2 (2%)	115 (78%)	10 (10%)
	Systemic Therapy	173 (43%)	44 (32%)	77 (89%)	32 (22%)	60 (85%)
	Nutrition risk assessed at visit (Yes)	146 (22%)	23 (11%)	40 (19%)	17 (7%)	43 (24%)

¹Patients may have received multiple treatments at a visit.

On average, patients included in the study (n=77) had about 20 total documented medical visits over the six-month time frame, with the mean number of visits and treatments varying by primary tumor. Patients with rectal cancer had on average twice as many medical visits compared to patients with colon cancer. Nutrition risk was documented in about one in five medical visits for patients with lung, colon, or pancreatic cancers and about one in ten medical visits for patients with esophageal and rectal cancers.

Patient Malnutrition Risk

Initial mean MST scores (available for 61 patients from five sites) indicated slightly higher risk in the RDN Care group (1.8 ± 1.2 [SD]) compared with the group that did not receive RDN Care (1.5 ± 1.1). Forty patients from four of the participating sites had MST scores recorded multiple times during the study period. Patterns of change in MST scores from initial to final documented value by group were different across sites, with final documented average malnutrition risk scores lower in the RDN Care group relative to the group that did not receive care at sites 2 and 4 (Figure 4).

Figure 4.

Change in average Malnutrition Screening Tool (MST) score. Average change was calculated from initial to final documented MST score for the patients that did and did not receive registered dietitian nutritionist (RDN) care (n=40) during the oncology cohort feasibility study by site. MST scores were not available for Site 6. Longitudinal MST scores were not available for Site 3.

Patient Medical Outcomes

Emergency room visits, unplanned hospital stays, and treatment changes were relatively rare within the six-month timeframe examined. For both groups, the median number of ER visits and unplanned hospital stays was zero and the interquartile range (IQR) was 0, 1 (p-values 0.06 and 0.48, respectively). Patients in the RDN Care group had more treatment changes (RDN care - median: 1, IQR: 0, 2; no RDN care – median 0, IQR: 0, 1; p-value 0.18). Within tumor type, there were also no significant differences in the outcomes by group. In multilevel models, outcomes were not significantly different by study group after controlling for covariates and including site as a random effect.

Sample Size Requirements and Intracluster Correlation Coefficients

Total estimated sample size requirements for a future study by tumor type and binary medical outcome are presented in Table 4. Estimated ICCs were 0.47, 0, and 0.21 for the total number of emergency room visits, unplanned hospitalizations, and treatment changes, respectively. These translated to respective design effects of 6.53, 1, and 3.42.

Table 4.

Total estimated sample size requirements for a future oncology study by tumor type and binary medical outcome¹

Medical outcome (binary2)	Lung	Esophageal	Colon	Rectal	Pancreatic
ER visits	341	645	709	284	121
Unplanned hospitalization	3503	264	274	546	5856
Treatment change	3369	113	979	1068	252

¹Calculated using G*Power 3.1.9.4 (Universitat Kiel, Germany). Input parameters were: z test family; statistical test: logistic regression; type of power analysis: a priori; tails: two; odds ratio as determined from pilot study data; α=0.05; power = 0.80; R² other than X (amount of variability in main predictor that is accounted for by covariates) as determined from pilot study data.

²Indicating that the outcome did not occur or occurred one or more times.

Discussion

In this feasibility study, we found that it was possible to complete site recruitment and onboarding, RDN documentation of nutrition care using standardized terminology, and remote abstraction of medical record data. There were some important lessons learned that could inform the design of a future study. The amount of time required for sites to sign a study site agreement and obtain IRB approval was extensive in some cases, exceeding one year for half of the sites. We achieved 92% of our enrollment goal, in part due to limited RDN ability to dedicate time to research activities and primarily because site policies for patients with esophageal cancer precluded identification of patients that did not receive RDN care. There were indications, based on abstracted medical information (i.e., tumor size and metastasis), that at least at some sites, individuals in the RDN Care group had more severe disease and higher malnutrition risk compared with those that did not. Over the six-month timeframe examined in this study, the medical outcomes examined were rare for most patients, resulting in high estimated sample sizes for similar future study, especially for unplanned hospitalizations in patients with lung and pancreatic cancer and treatment changes in patients with lung, colon, and rectal cancer. Substantial design effects were noted for emergency room visits and treatment changes.

We did not find evidence of statistically or clinically significant differences in the number of emergency room visits, unplanned hospitalizations, or treatment changes over a six-month timeframe by study group. Fundamentally, any observations related to differences between groups must be interpreted in the context that this was a feasibility study with no a priori sample size calculation relative to the outcomes of interest. Data collected during the study indicate that larger sample sizes are likely needed to detect differences in the medical outcomes examined in this study by group for specific tumor types. However, a few randomized controlled trials comparing nutrition care from an RDN with usual care for patients with cancer have also noted a lack of effect on treatment and hospitalization-related outcomes, although sometimes they did find significant effects on other important clinical outcomes. In a multi-site study of 341 patients aged 70 or older with multiple tumor types in France, Bourdel-Marchasson et al. compared patients that received six sessions of diet counseling using a standardized nutrition curriculum and oral nutrition supplements (if indicated) during chemotherapy to patients receiving usual care (including nutrition counseling and support as indicated by facility protocol) and noted no significant effects of the intervention on weight changes, chemotherapy management or changes, hospitalization or one and two year mortality.¹⁵ Similarly, in a multi-site study with 134 patients with head and neck cancer in Sweden, Silander et al. compared patients that received individualized nutrition counseling plus a proactive percutaneous endoscopic gastrostomy (PEG) tube (if >1 kg weight loss) with patients who received usual care including nutrition counseling and tube feeding when necessary, and found no effect on total days of hospitalization.²¹ They did observe less weight loss and better quality of life in the intervention versus the comparison group.²¹ In a multi-site study with 107 patients with metastatic colorectal cancer in the Netherlands, van der Werf et al. compared outcomes in patients who received individualized nutrition counseling during chemotherapy, along with progressive oral nutrition support and tube feeding if they were meeting <75% of their intake goals or had significant weight loss, with patients that received usual nutrition care (i.e., who were referred to an RDN by the medical team when indicated).¹¹ They noted no effect on muscle mass or function, quality of life, or treatment outcomes, but did find significant improvements in body weight and progression free and overall survival in the intervention versus the comparison group.¹¹ Sometimes, findings related to hospitalization outcomes were mixed within the same study. In a single site study of 86 patients with laryngeal carcinoma in China, Song et al. compared patients that received individualized, proactive pre- and post-operative nutrition management from a multi-disciplinary team with patient receiving standard of care nutrition management, and noted no effects on hospitalization time or cost, but significantly lower hospital readmissions, along with lower rates of postoperative complications, mortality, and malnutrition rates and better quality of life in the intervention versus comparison group.¹³

In contrast, a few randomized controlled trials did find significant effects on treatment outcomes. In a single site study with 144 patients with gastric cancer undergoing postoperative chemotherapy in China, Xie et al. compared outcomes in patients receiving individualized nutrition education and counseling throughout chemotherapy treatment versus patients that received basic nutrition care and health education only during hospitalization.¹⁴ They found that patients in the intervention group had significantly better chemotherapy compliance than patients in the comparison group.¹⁴ In a multisite study with 307 patients with head and neck cancer in Australia, Britton et al. compared outcomes in patients that received proactive RDN care delivered using cognitive behavioral therapy (CBT) and motivational interviewing (MI) to patients that received usual care (including visits with RDNs).²³ They found that the intervention group had significantly less treatment interruptions, along with less percent weight loss, lower depression scores better PG-SGA and SGA status, and better QOL versus the comparison group.²³

The care provided by the RDNs in our study was similar in intensity (average 5 sessions) to the interventions tested in these studies, but may not have been proactive, depending on when facilities implemented their universal malnutrition screening policies relative to the start of the study activities. In addition, care in this study likely did not include comprehensive, coordinated multidisciplinary care¹³ or consistent use of CBT and MI techniques during counseling sessions across sites²³, as tested in some of the RCTs that noted effects on treatment outcomes. Implementation research examining the best way to use approaches like this in U.S. outpatient cancer settings may be an important future research direction. In particular, we noted variability in malnutrition screening policies and practices across sites, and research to understand best practices for establishing processes to complete initial malnutrition screening, score tools appropriately, order consults, and have the RDN see the patient and communicate findings, such as reduced dietary intake relative to requirements and/or malnutrition diagnosis and implemented interventions, to the rest of the care team in a timely way could be valuable in facilitating proactive, coordinated multidisciplinary care.

In addition to variability in policies and practices across sites, we noted differences in malnutrition risk and cancer advancement across groups in this observational study, although this differed by site in some cases. This is consequential, as some RCTs have noted significant effects of nutrition interventions only for subgroups of patients with poorer nutritional status at baseline^19,31,32 or for patients with less advanced cancer.²⁷ Taken together, the findings of this feasibility study indicate that interpreting the results of a larger, observational study of this design might be challenging, and that the sample sizes and time required to conduct such a study might be prohibitive. It would, however, be possible to use the methods tested in this study (e.g., practicing RDNs implementing a study protocol; longitudinal documentation of RDN nutrition care; remote medical record review) in an individually or cluster-randomized (e.g., stepped wedge²³) trial examining different intensity or timing of RDN care in U.S. outpatient cancer treatment settings. Given the mixed findings from previous studies, careful consideration of the optimal patient population, intervention design, and nutrition and medical outcomes is necessary, and attempts to standardize these parameters across studies are necessary to better pool and interpret evidence in this area.¹⁰ Detailed baseline and longitudinal assessment of parameters such as dietary intake relative to requirements, fat and muscle loss, and malnutrition diagnosis status could provide important context in studies examining impact of nutrition care on medical outcomes. Estimates of additional parameters such as kilocalorie, protein, and fluid requirements can be tracked as well, as these are key nutritional considerations that are often readjusted in response to clinical condition. Assessment of medical outcomes that are particularly relevant by tumor type and treatment modality may also be helpful in understanding the impact of nutrition care; for example, tracking unplanned clinic, emergency room, or hospital visits for dehydration for patients with rectal cancer with an ostomy. To our knowledge, this is the first study of nutrition care in cancer that has examined emergency room visits as an outcome. Emergency department visits cost billions of dollars per year for U.S. payers and could be an important outcome to include in future studies.⁴⁵ Regardless of design, our experience with this study indicates that providing RDNs with intensive support in navigating research processes and systems at their sites could potentially improve the site onboarding timeline. In addition, the abstractor noted that it was essential to receive an orientation from site personnel on how to navigate each site’s medical record.

A strength of this study is the ability to examine and report differences in important parameters across individuals that do and do not receive RDN care and across treatment centers to inform future study design. This study also has several limitations. In this observational design, patients in the No RDN Care group did not have data available for other key nutrition assessment variables, such as dietary intake, that could be compared across groups. The study included a small number of sites and patients with five different tumor types and different stages of cancer. In conducting a pooled analysis, we had a limited ability to consider different trajectories of treatment and outcomes across tumor types and advanced stage of cancer, especially as we were missing aspects of TNM classification for a substantial proportion of patients. Patients included in the study were not racially or ethnically diverse, indicating that future studies should consider recruiting sites in parts of the U.S. that are more diverse (e.g., South, Southwest, and California). We noted many differences in nutrition-related policies and practices across sites, including the use of different malnutrition screening tools (with non-validated tools or approaches used in some sites) and variable timing of screening administration, likely resulting in variability in patients deemed at nutrition risk across sites. We also were not able to assess whether patients in the no RDN care group received nutrition counseling, advice, or management from other health care professionals, and whether this varied across sites. While this variability reflects real world practice, better standardization of these practices through either site inclusion/exclusion criteria or by requiring and supporting all sites to implement the same procedures and practices during the study would improve the interpretability of results for a future multi-site study. Additionally, we acknowledge that we likely only captured ER visits and hospital admissions that occurred within health facilities that had their EMRs connected to the outpatient facility, possibly resulting in an undercount of the outcomes of interest for some patients, and attempts to remedy this limitation should be made in future studies. Finally, much of this study was conducted during the COVID-19 pandemic, which resulted in significant disruptions to healthcare operations across the United States.