Nutritional Status and Quality of Life in Hospitalised Cancer Patients Who Develop Intestinal Failure and Require Parenteral Nutrition: An Observational Study

(1) Background: Malnutrition in cancer patients impacts quality of life (QoL) and performance status (PS). When oral/enteral nutrition is not possible and patients develop intestinal failure, parenteral nutrition (PN) is indicated. Our aim was to assess nutritional status, QoL, and PS in hospitalised cancer patients recently initiated on PN for intestinal failure. (2) Methods: The design was a cross-sectional observational study. The following information was captured: demographic, anthropometric, biochemical and medical information, as well as nutritional screening tool (NST), patient-generated subjective global assessment (PG-SGA), functional assessment of cancer therapy-general (FACT-G), and Karnofsky PS (KPS) data. (3) Results: Among 85 PN referrals, 30 oncology patients (56.2 years, 56.7% male) were identified. Mean weight (60.3 ± 16.6 kg) corresponded to normal body mass index values (21.0 ± 5.1 kg/m2). However, weight loss was significant in patients with gastrointestinal tumours (p < 0.01). A high malnutrition risk was present in 53.3–56.7% of patients, depending on the screening tool. Patients had impaired QoL (FACT-G: 26.6 ± 9.8) but PS indicated above average capability with independent daily activities (KPS: 60 ± 10). (4) Conclusions: Future research should assess the impact of impaired NS and QoL on clinical outcomes such as survival, with a view to encompassing nutritional and QoL assessment in the management pathway of this patient group.


Introduction
Cancer is among the leading causes of morbidity and mortality worldwide [1]. In advanced disease, malnutrition and involuntary weight loss with associated cachexia are observed in up to 80% of patients [2,3]. Half of all cancer deaths worldwide are attributed to malignancies, with a high prevalence of cachexia, such as gastrointestinal and pulmonary malignancies. However, cachexia is also highly prevalent at the end of life regardless of tumour type [4]. Systemic inflammation caused by the underlying malignant process drives cachexia's progression through metabolic disturbances and muscle loss [4,5]. Other factors such as anorexia and cancer-related symptoms that may reduce oral intake, side-effects of medications and therapies, as well as functional impairment and psychological distress also contribute to cachexia [4][5][6]. The prevalence of muscle loss is reported in 20-70% of cancer

Study Setting
This study was conducted at University College London Hospital (UCLH, London, UK) a 665-bed tertiary hospital and a national referral centre for a wide range of cancers, in central London, UK. PN provision services are supported by the multidisciplinary nutrition support team, which reviews all inpatient PN referrals identified by the primary clinical teams.

Study Population
All adult patients, aged 18 years or older, admitted to UCLH with active cancer, referred to the nutrition support team and started on PN between January and June 2019 were screened for inclusion in this study. Patients who were adults with active cancer, had capacity to complete questionnaires, consented for additional anthropometric measures, and were not established on HPN prior to their present admission were included. The design was a cross-sectional observational study. Data were collected through UCLH patient records (paper and electronic) close to the time of referral and during patient interviews (Table S1).

Demographics and Serum Biochemistry
Demographics and medical information: gender, age, primary malignancy, metastases, surgery, chemotherapy and/or radiotherapy before or during PN, and indication for PN (ESPEN) [12].

Anthropometrics, NS, QoLand PS
Anthropometrics and NS: height, weight upon starting PN, habitual body weight, habitual and current BMI, percentage of weight loss upon starting PN from habitual weight, MAC [22], TSF [22], HGS [23], and MUAMC. The latter was calculated with the following equation MUAMC = MAC − (0.314 * TSF) [24]. MUAMC is strongly correlated with whole-body composition assessments (e.g., BIA and dual-energy x-ray absorptiometry) in similar populations [25]. Patients were defined as having cancer cachexia if weight loss > 5% was reported in the past 6 months since diagnosis or BMI ≤ 20 Kg/m 2 and any degree of weight loss > 2% [2].
The UCLH nutritional screening tool (NST) is performed by trained staff (health care assistant, nurse or dietitian) to screen patients within 24 h of admission and weekly thereafter in order to identify those in need of dietary support (Table S2). The NST assesses patients' weight, height, BMI, appetite, dietary intake, weight loss, psychological and neurological status, and physical appearance, while guiding triaging for the appropriate actions required. The NST categorises patients into low, medium and high malnutrition risk groups, with higher scores indicating a greater malnutrition risk (scores 0-2, 3-6, >7, respectively). Medium-and high-risk patients require a dietetic referral.
The patient-generated subjective global assessment-short form (PG-SGA) is a validated nutrition assessment tool for cancer patients designed for self-administration, focusing on weight and food intake changes, symptoms that have persisted for more than 2 weeks, as well as changes in activities and performance. The PG-SGA categorises patients as well nourished, moderately or suspected of being malnourished or severely malnourished, with higher scores indicative of greater malnutrition risk (scores 0-1, 2-8, >9, respectively) [26].
QoL: The functional assessment of cancer therapy-general (FACT-G) is a self-administered 27-item questionnaire. Trained staff (dietitians) calculated the overall (0-108) and subscale scores using specific guidelines (higher scores indicated better QoL) [19]. Permission to use the FACT-G questionnaire was obtained from website [27].
The HPN-QoL questionnaire is specifically designed for oncology patients treated with HPN [28]. This questionnaire was adapted, excluding the HPN-specific questions, and used as a measure of QoL for patients who had started inpatient PN. Permission to use the HPN-QoL questionnaire was sought from the author (Janet Baxter).
PS: The Karnofsky PS (KPS) was assessed by the attending dietitian. Scores range from 0 to 100, with over 50 indicating that the patient is unable to carry out daily tasks, but able to live at home and care for most personal needs with varying amounts of assistance [29,30]. The Eastern Cooperative Oncology Group/World Health Organisation performance status (ECOG/WHO-PS) is a prognostic factor in cancer populations, using a scale from 0 (fully active) to 5 (dead) to indicate the ability of physical activity, movement and self-care [31].

Ethical Considerations
This was a cross-sectional study and the principles of the Declaration of Helsinki were followed during design and analysis. Ethical approval was not required for this study as it was registered as a departmental audit.

Statistical Analysis
Data are presented as the mean (standard deviation [SD]) or frequencies and percentages. Univariate analyses were conducted with the chi-square test, Spearman's rho for correlations, t-tests and ANOVA. The concordance between tools of malnutrition risk assessment, namely the NST, the PG-SGA, BMI and weight loss %, was examined using Cohen's kappa for categorical variables [32]. Exploratory factor analysis based on principal component analysis was used for the HPN-QoL questionnaire for item aggregation and reduction [33], and internal consistency was examined using Cronbach's alpha coefficient [34]. A cluster heatmap with a dendrogram was also produced based on hierarchical clustering. Statistical significance was reported at p < 0.05. All statistical analyses were performed using IBM SPSS Statistics (Release 25.0.0.1 2017, Chicago (IL), USA: SPSS, Inc., an IBM Company) and R 4.0 (R Foundation for Statistical Computing, Vienna, Austria).

Agreement between Malnutrition Risk Assessment Tools
Compatibility between the different nutritional screening tools was assessed with Cohen's kappa and there was no significant compatibility between any tool (p > 0.05) ( Table S3). The NST was moderately correlated with MUAMC (rho = −0.426) and weight on admission (rho = −0.455), and moderately with weight loss at 6 months (rho = 0.502) (Table S4).

Nutritional Status Indicators
Nutritional indexes were examined according to the type of malignancy, indication for PN, presence of metastases and cachexia and are presented in Table 3. There was a trend towards a significant difference in weight upon starting PN according to the primary cancer location, with upper gastrointestinal cancer patients having a lower weight than other types of malignancy (48.1 ± 10.0 kg, p < 0.01). This patient group also had lower baseline BMI values (17.22 ± 4.5 kg/m 2 , p < 0.01). Additionally, patients with cachexia had significantly higher NST scores (cachexia 3.9 (3.5) vs. no cachexia 8.2 (4), p < 0.05).
Nutritional indexes were next examined according to NST, PG-SGA, KPS and WHO-PS scores and are presented in Table 3. There were no significant results observed, except for patients' HGS according to KPS score. Patients in the low KPS scores category (score < 50) had lower HGS measurements (24.6 ± 9.22 kg), while those in the high KPS scores category (score ≥ 50) had higher HGS measurements (44.1 ± 22.6 kg, p < 0.05).   The association between nutritional indexes was examined using Spearman's rho correlation (Table S4). All correlations were significant (p < 0.05). TSF was moderately correlated with BMI on admission (rho = 0.531), weight on admission (rho = 0.469), and weight loss at 6 months (rho = 0.425). MAC was moderately correlated with BMI on admission (rho = 0.553) and strongly with weight on admission (rho = 0.683). MUAMC was very strongly correlated with BMI on admission (rho = 0.805) and weight on admission (rho = 0.830) and moderately with weight loss at 6 months (rho = 0.475). Finally, HGS was moderately correlated with weight on admission (rho = 0.386), MAC (rho = 0.440), and KPS (rho = 0.531).
Presence of cachexia and metastasis, NST and PG-SGA score distributions were examined by PN indication, type of malignancy, metastatic disease and KPS score, and are shown in Tables S5-S8. Only presence of metastatic disease differed by types of primary malignancy, with patients affected by upper gastrointestinal malignancies having a higher prevalence of metastatic disease (p < 0.05).

Exploratory Factor Analysis and Internal Consistency
Factor analysis of the results of the HPN-QoL questionnaire identified five factors, based on the Kaiser criterion [35], also confirmed by the line drop after the fifth component in the Cattell scree plot ( Figure S1) [36]. Table S9 shows the allocation of the items to the respective factors based on their highest factor loading, above the predefined value of 0.4 [37]. Cronbach's alpha for the modified HPN-QoL questionnaire overall was 0.63. When internal consistency was performed separately for the factors, Cronbach's alpha was over 0.80 in all factors. The FACT-G subscales report in a range of poor to good internal consistency, with Cronbach's alpha ranging from 0.567 to 0.840 (Table 4).

Quality of Life Correlations
The QoL questionnaire FACT-G, its subscales, and the factors of the modified HPN-QoL questionnaire were examined according to type of malignancy, presence of metastases and cachexia, and no statistically significant difference was observed (Table S10). The aforementioned were also examined according to classification based on NST, PG-SGA, KPS and WHO-PS scores and no significant differences were detected (Table S10). Only the scores of the energy/independence factor of the modified HPN-QoL questionnaire within the risk of malnutrition groups based on PG-SGA score had a non-statistically significant trend to differ, with severely malnourished patients having lower scores of energy/independence (0.6 ± 1.1 vs. −0.3 ± 0.8, p = 0.053).
The association between the QoL questionnaires and their subscales with NS indexes was examined using Spearman's rho correlation and the significant results appear on Table S4. While neither the FACT-G nor its subscales correlated with any of the nutritional indexes, two of the factors of the HPN-QoL questionnaire did. The pain factor correlated moderately with the HGS on the non-dominant hand (rho = −0.474), and the energy/independence factor correlated moderately with age (rho = 0.525).

Performance Status
Performance status was examined according to the type of malignancy, indication for PN, presence of metastases and cachexia, but there was no significant difference in patients' KPS scores (Table 3). However, when the KPS was examined according to classification based on NST, PG-SGA and WHO-PS scores (Table 3), there was a significant difference in KPS scores between the subgroups within WHO-PS, suggesting that patients in the distinct WHO-PS subgroups had different KPS scores (73.3 ± 5.7 at WHO-PS 1, 64.6 ± 7.8 at WHO-PS 2, 55.4 ± 8.7 at WHO-PS 3, 40.0 at WHO-PS 4, p < 0.001). Finally, the KPS was moderately correlated with length of stay (rho = −0.448) (Table S3).

Cluster Heatmap
A cluster heatmap for the NS, PS and QoL indexes was undertaken and the degree of similarity is presented in a dendrogram, where four clusters were identified (Figure 1).

Performance Status
Performance status was examined according to the type of malignancy, indication for PN, presence of metastases and cachexia, but there was no significant difference in patients' KPS scores (Table 3). However, when the KPS was examined according to classification based on NST, PG-SGA and WHO-PS scores (Table 3), there was a significant difference in KPS scores between the subgroups within WHO-PS, suggesting that patients in the distinct WHO-PS subgroups had different KPS scores (73.3 ± 5.7 at WHO-PS 1, 64.6 ± 7.8 at WHO-PS 2, 55.4 ± 8.7 at WHO-PS 3, 40.0 at WHO-PS 4, p < 0.001). Finally, the KPS was moderately correlated with length of stay (rho = −0.448) (Table S3).

Cluster Heatmap
A cluster heatmap for the NS, PS and QoL indexes was undertaken and the degree of similarity is presented in a dendrogram, where four clusters were identified (Figure 1).

Discussion
This is the first study in the literature examining nutritional status, QoL and PS in an inpatient setting for advanced cancer patients referred for PN due to intestinal failure. The PG-SGA identified 100.0% (24/24) of patients as being at risk of malnutrition and 70.8% (17/24) as having severe malnutrition, the majority of which were gastrointestinal and haematological cancer patients on active oncological treatment. Prevalence from studies with similar groups of patients ranged between 45.1% and 80.4% [38][39][40][41]. Results differed when the NST was used, as 80.0% (20/25) of patients were at risk, for whom close monitoring was required.
The PG-SGA was next compared to other malnutrition screening tools, with differences being noted. Compared to the NST, prevalence was similar for severely malnourished patients, though differences were observed between moderately and well-nourished patients. The PG-SGA is the reference tool for oncology patients' nutritional assessment and is a sensitive, comprehensible, easy and quick tool to assess malnutrition [42,43]. Correcting for short-term improvements in weight and scoring multiple nutrition impact symptoms, it offers high accuracy in distinguishing well-nourished from malnourished patients; however, it ultimately categorises more patients at risk compared to other tools [44,45]. To the best of our knowledge, there have not been other studies assessing nutritional risk in cancer patients using the NST; therefore, comparison of the prevalence of patients identified as at risk of malnutrition using the NST with similar studies is disadvantaged.
Examination of malnutrition risk according to weight loss % and BMI classification failed to identify all patients at risk in our sample, with weight loss % and BMI identifying only 42.3% (11/26) and 34.5% (10/29) of patients, respectively. Previous studies evaluating weight loss and BMI as markers of NS in cancer patients corroborate this discrepancy, attributing such an observation to sudden changes in weight (e.g., ascites, oedema, and fluid retention) or inability to detect differences between fat and fat-free mass [8,46,47].
Patients with upper gastrointestinal malignancies were identified to have lower weight and BMI on admission compared to patients with other types of malignancies. This finding is consistent with other studies and is attributed to the nature and location of oesophageal and gastric cancers, where the risk of malnutrition is highly prevalent by the time of diagnosis, mainly due to bowel obstruction and gastrointestinal symptoms [48][49][50][51]. A borderline difference in MUAMC was noted between different types of malignancy, indicating that severe weight loss due to muscle mass depletion defines this pattern of malnutrition [52,53]. There was a trend for patients with metastatic disease to have a lower BMI compared to patients with no metastatic disease, indicative of cancer cachexia [54].
MAC, HGS, and TSF were significantly correlated with weight and BMI on admission. HGS was moderately correlated with weight on admission and MAC. Muscle mass and strength loss was also noted in our malnourished patients. Previous studies have not found significant results between HGS and weight loss, although lower HGS values have been noted in malnourished patients [55][56][57][58]. Low HGS reflects muscle mass depletion, which is reportedly due to altered protein metabolism during severe weight loss, inflammation, inactivity, anaemia, fatigue and tissue hypoperfusion [55,59]. Next, there was a significant difference in patients' HGS between subgroups based on KPS scores. Patients with a poorer PS had significantly lower HGS measurements compared to patients with a higher PS, a trend noted in other studies as well [55,[60][61][62]. PS is an important indicator of a patient's ability to be functional and perform activities of daily living. The finding that decreased muscle strength could predict functional decline in hospitalised oncology patients is of utmost importance and also carries a prognostic value [14]. PS also reflects several behavioural patterns that are affected by NS.
The energy/independence factor from the HPN-QoL questionnaire was moderately correlated with age. This suggested that older patients felt less independent and more unable to cope with daily life and their illness, accounting for their poor functional status and QoL. Similar results have been reported in other studies [63,64], and a recent meta-analysis reported that 37-55% of older cancer patients required daily assistance [65]. The pain factor from the HPN-QoL questionnaire was moderately correlated with patients' HGS, suggesting that lower HGS was related to a greater sense of pain. Pain is one of the most frequently reported symptoms by cancer patients, as well as one of the most important drivers of diminished appetite [66]. Pain, along with systemic inflammation, contributes to the sense of fatigue, which restricts physical activity and in turn alters protein metabolism and favours muscle wasting [7]. Finally, the dendrogram revealed that QoL clustered closely with PS and HGS, indicating that reduced muscle strength and consequently impaired PS, were the main components that compromised a patient's QoL. Our study replicates this result in line with other studies which have also noted that QoL deteriorates alongside PS reduction in advanced cancer patients [66][67][68].
The main strength of the present study is the thorough assessment of NS, PS and QoL, beyond that of standard clinical assessment, with all measurements performed using validated instruments by experienced health care professionals with adequate training. In terms of limitations, the small study sample and relatively heterogenous cohort of patients with regards to type of malignancy led to reduced power in identifying significant findings. Secondly, the assessment was performed at a single point in time and follow-up information was not available, thus it is not possible to draw adequate conclusions on the causality of relationships. Finally, the lack of whole-body composition measures (e.g., CT/MRI scans) limits the amount of high-quality and reliable data that could be used as the gold standard to compare methods of nutritional assessment.
The present study offers certain implications for practitioners worth discussing. Firstly, since the NST identified patients with malnutrition and proportionately categorised them as having cachexia, further studies or audits could validate the NST in oncology patients, correctly identifying those in need of comprehensive dietetic assessment and detect periodic NS changes after nutritional support with PN. Furthermore, our results strengthen the notion that weight loss and BMI should be assessed assiduously as a sensitive, convenient and non-invasive assessment of cachexia. The next important point is that although PN was consistently used in these patients, muscle mass loss is unlikely to be reversed in refractory cachexia [2]. If loss of muscle mass cannot be prevented during hospitalisation and PS is ultimately compromised, discharge timelines are affected. Finally, since pain and loss of muscle mass have shown to be strongly correlated, multimodal care involving resistance training is strongly advocated in this group of patients [69]. An implication for research would be to investigate the impact of PN on improving the sense of pain, and consequently maintaining patients' muscle strength, as part of the multimodal care.

Conclusions
Malnutrition in cancer patients is still under-recognised and highly prevalent. PN is an intricate therapy that should be judiciously used when indicated. However, it only comprises one part of the multimodal therapeutic strategies of cancer patients including psychological support, symptom management, anticancer therapy and physical activity. Our results highlight the compromised overall status of patients by the time of referral for PN support, hence timely referral as well as concurrent assessment of NS, PS and QoL in this patient group is of paramount importance due to interplays identified among them across the literature. Our aim is to raise awareness on the importance of preventing cachexia and PS and QoL deterioration among hospital staff and change our clinical practice towards appropriate individualised holistic patient-centred care plans, from which patients will derive maximal benefit.

Supplementary Materials:
The following are available online at http://www.mdpi.com/2072-6643/12/8/2357/s1, Figure S1: Scree plot; Table S1: Data collection timeline; Table S2: UCLH nutritional screening tool; Table S3: Agreement among categorical variables of nutritional status with Cohen's kappa; Table S4: Spearman's rho  correlations between variables; Table S5: Cancer cachexia distribution according to type of malignancy, KPS score, metastatic disease, and indication for PN; Table S6: Metastatic disease distribution according to type of malignancy, KPS score, and indication for PN; Table S7: NST score distribution according to type of malignancy, KPS score, cachexia, metastatic disease, and indication for PN; Table S8: PG-SGA score distribution according to type of malignancy, KPS score, cachexia, metastatic disease, and indication for PN; Table S9: Factor analysis and structure; Table S10: Quality of life according to type of malignancy, indication for PN, presence of metastases and cachexia.