Undernutrition in cancer increases morbidity and treatment toxicity, decreases response to treatment, quality of life and almost certainly worsens prognosis(Reference Grant and Rivera1–Reference Ravasco, Monteiro-Grillo and Vidal5), thus increasing healthcare costs(Reference Green6). However, undernutrition continues to be neither screened/diagnosed nor integrated in overall patient care; hence, timely and dynamic adjustments to patients' needs are still theoretical(Reference Holmes7).
Although routine nutritional screening is acknowledged as an essential component of modern cancer care, early risk assessment is still missing presumably due to the lack of a simple, validated and reproducible method(Reference Kondrup, Allisson and Elia8). In order to overcome this drawback, we selected to compare a validated tool used to diagnose undernutrition in oncology, the scored Patient-Generated Subjective Global Assessment (PG-SGA)(Reference Ottery2) with an easy and simple tool to screen patients at nutritional risk. The mandatory integration of PG-SGA can only be effective if antedated by nutritional risk screening. This enables the identification of those patients in need of a full nutritional status assessment by PG-SGA which requires a trained nutrition professional. Those who are not considered to be at risk do not require the full assessment and should be re-screened periodically. In the absence of a ‘gold standard’ for nutritional risk screening, the Malnutrition Universal Screening Tool (MUST)(Reference Elia9, Reference Stratton, Hackston and Longmore10) was chosen because it has content validity (comprehensiveness of the tool), face validity (issues which are relevant to the purpose of the test) and internal consistency. MUST is a screening tool that has shown its strength for application to adult patients across all healthcare settings including oncology(Reference Stratton, Hackston and Longmore10). Despite MUST's excellent agreement with dietitians' assessment of undernutrition(Reference Elia9), so far no studies comparing MUST and PG-SGA have been published.
Therefore, the present study aimed to: (1) classify nutritional risk and status categories in patients with different types of tumours; (2) compare results between nutritional parameters, primarily MUST and PG-SGA; and (3) validate MUST as a feasible screening method for routine use in the radiation oncology setting, by comparison with the validated PG-SGA.
Material and methods
Study design and patient population
This prospective cross-sectional study was approved by the University Hospital Ethics Committee and was conducted in accordance with the Helsinki Declaration, adopted by the World Medical Association in 1964, amended in 1975, revised in 1983 and updated in 2002. Between March 2008 and January 2009, all consecutive patients with diverse cancers referred to the outpatient Radiotherapy Department were considered eligible. Exclusion criteria comprised rare tumours(Reference Longo and Fauci11), and uncooperative patients unable to communicate or to be properly weighed. Before radiotherapy planning, the medical staff registered, for every patient, clinical variables, the location of cancer and burden according to TNM classification(Reference Sobin and Fleming12) as determined by local and whole-body imaging methods. Of the total 450 patients, 165 underwent palliative radiotherapy v. 285 who underwent curative radiotherapy. All patients included in the study provided written informed consent. Data were recorded in individual forms pre-designed for statistical analysis.
The cohort included 450 adult patients (60 %, n 269/450 men; 40 %, n 181/450 women), with a mean age of 62 (sd 13) (range: 18–95) years, with different types of cancers and stages. Tumours in advanced stages (III and IV) were predominant (n 273); the most frequent types of cancers were: breast (n 94), prostate (n 86), lung (n 73) and colorectal (n 61); full details of these are given elsewhere(Reference Ramos Chaves, Boléo-Tomé and Monteiro-Grillo13).
Study measures
Data collection and nutritional assessment were obtained at the onset of radiotherapy by a trained research dietitian (C. B.-T.).
Nutritional status
BMI
Height was measured in the standing position using a stadiometer and weight was determined with a calibrated SECA® floor scale (Seca, Birmingham, UK); both measures were used to calculate BMI (weight (kg)/height (m)2), further classified as undernutrition if < 18·5 kg/m2, normal weight if 18·5–24·9 kg/m2, overweight if 25–29·9 kg/m2 or obese if ≥ 30 kg/m2, in line with age- and sex-adjusted criteria established by the World Health Organization(14).
Percentage of weight loss
Percentage of weight loss was calculated by comparison with the patient's reported usual weight before symptoms, and classified as clinically significant and indicative of undernutrition or high nutritional risk if >5 % over the previous 3–6 months(Reference Stratton, Greene and Elia15).
Patient Generated-Subjective Global Assessment
PG-SGA is a validated nutritional assessment tool, which has been accepted by the Oncology Nutrition Dietetic Practice Group of the American Dietetic Association as the ‘gold standard’ for cancer patients(Reference Ottery2). PG-SGA consists of two sections: (1) weight history, food intake, nutrition impact symptoms and functional capacity; (2) diagnosis, disease stage, age, components of metabolic demand (sepsis, neutropenic or tumour fever, corticosteroids) and physical examination. The patient's nutritional status is then classified into three categories: (A) well-nourished, (B) moderately undernourished or suspected of being undernourished, and (C) severely undernourished. For the present study, and for between-methods comparisons, two categories of the PG-SGA results were created(Reference Stratton, Hackston and Longmore10): well nourished v. moderately+severely undernourished, to enable comparisons with MUST (details are outlined in the following paragraph).
Nutritional risk
Nutritional risk was assessed by the MUST that addresses: (1) current weight status using BMI, which is scored as 0 if >20 kg/m2, 1 if between 18·5 and 20 kg/m2 and 2 if < 18·5 kg/m2; (2) percentage weight loss over the previous 3–6 months is scored as 0 if < 5 %, 1 if between 5 and 10 % and 2 if >10 %; and (3) the acute disease effect is scored with 2 points if there has been, or if it is likely to be, no nutritional intake for >5 d. The scores given to each component are summed up and the total allows the categorisation of patients as in low, moderate or high risk of undernutrition; this score is used to guide patients' reassessment plan as well as the appropriate nutritional care plan(Reference Elia9). For the present study, and for between-methods comparisons with the PG-SGA, two categories of MUST were created: low risk v. medium+high risk of undernutrition(Reference Stratton, Hackston and Longmore10). A Portuguese version of MUST was created with the translation of the original technical terms.
Statistical analysis
Statistical analysis was conducted using SPSS 16.0 (SPSS, Inc., Chicago, IL, USA). Patient's age was expressed as number and percentage, median and standard deviation; cancer location and stage were expressed as number; BMI, percentage weight loss and nutritional risk/status were expressed as number and/or percentage of patients. Correlations were determined using the two-tailed Spearman's test. χ2 Test was used to compare differences in the prevalence of nutritional risk.
Sensitivity, specificity and predictive values were calculated to evaluate whether MUST could be a valid nutritional screening tool in cancer. A receiver operating characteristic (ROC) curve interpreted by relative areas under the curves and Youden value confirmed the consistency of the validation. Sensitivity of a screening tool expresses index sensitiveness to a certain factor; in this study, test sensitivity was the proportion of ‘undernourished’ cases diagnosed by PG-SGA, also found to be at ‘risk of undernutrition’ by the screening tool MUST. A high sensitivity of a screening tool may give false positives, with more patients being classified as at risk of undernutrition. Specificity refers to the proportion of patients without nutritional deficiency by PG-SGA and corroborated by MUST; high specificity may give many false negatives. The positive predictive value is the probability that a patient classified as at risk of undernutrition by a screening tool is effectively found to be undernourished by PG-SGA. The negative predictive value is the probability that a patient classified as not at risk of undernutrition by a screening tool is also defined as well nourished by the reference method. Concordance between MUST and PG-SGA was analysed using the κ coefficient, further classified according to Fleiss(Reference Fleiss16), to assess the correspondence between the results achieved by PG-SGA and MUST – convergent validation test on the result. Statistical significance was determined for P < 0·05.
Results
Nutritional parameters
Overall, BMI revealed that 4 % (n 17/450) of patients were undernourished whereas 63 % (n 282/450) were overweight/obese (BMI ≥ 25 kg/m2). A weight loss >5 % over the previous 3–6 months was found in 33 % (n 101/450) and PG-SGA classified 29 % (n 131/450) as moderately/severely undernourished patients; nutritional risk as evaluated by MUST showed a prevalence of 31 % (n 139/450) of patients at moderate/high risk of undernutrition. Table 1 shows the results of three nutritional parameters, namely BMI, PG-SGA and MUST.
(Total number and percentage of patients)
BMI v. Patient-Generated Subjective Global Assessment
BMI results when compared with PG-SGA showed that sixteen patients (94 %) were correctly classified as undernourished (true positives) and 318 (73 %) patients were correctly classified as well nourished (true negatives), though the latter included overweight/obese (Fig. 1). On the other hand, 115 patients (88 %) were wrongly classified by BMI as well nourished (false negatives), despite being assessed as undernourished by PG-SGA. Thus, BMI alone was not a good predictor of undernutrition risk. Additionally, according to the Youden test, BMI had a poor performance by comparison with the standard, and a weak capacity to effectively detect undernourished patients, with a sensitivity of 0·27, P < 0·08 (95 % CI) and a specificity of 0·23, P < 0·09 (95 % CI). BMI had a positive predictive value of 0·35, P < 0·09 and a negative predictive value of 0·31, P < 0·08 v. PG-SGA.
BMI v. Patient-Generated Subjective Global Assessment (PG-SGA): well nourished (
)/undernourished (
) patients represented in percentages.
Percentage weight loss v. Patient-Generated Subjective Global Assessment
By comparison with PG-SGA, percentage weight loss correctly classified 291 (84 %) patients as well nourished (true positives) and correctly classified seventy-four (73 %) as undernourished patients (true negatives) (Fig. 2). On the other hand, twenty-eight (27 %) patients were falsely classified by percentage weight loss as undernourished, though assessed as well nourished by PG-SGA. Methods were compared by the Youden test. According to the results, percentage weight loss revealed a sensitivity of 0·76, P < 0·002 (95 % CI) and a specificity of 0·85, P < 0·001 (95 % CI). Thus, percentage weight loss had a positive predictive value of 0·79 (P < 0·002) and a negative predictive value of 0·85 (P < 0·001) v. PG-SGA.
Percentage weight loss v. Patient-Generated Subjective Global Assessment (PG-SGA): well nourished ()/undernourished () patients represented in percentages.
Malnutrition Universal Screening Tool v. Patient-Generated Subjective Global Assessment
By comparison with PG-SGA, MUST correctly classified 275 patients (88 %) as without risk of undernutrition (true negatives) and ninety-five (68 %) were correctly classified by MUST as being at risk (true positives). MUST misclassified thirty-six patients (12 %) as without risk (false negatives) who were classified as undernourished by PG-SGA. Conversely, forty-four (32 %) were falsely classified as at risk and were assessed as well nourished by PG-SGA (false positives) (Figs. 3 and 4).
Patients at nutritional risk (Malnutrition Universal Screening Tool (MUST)) and those undernourished by Patient-Generated Subjective Global Assessment (PG-SGA).
Malnutrition Universal Screening Tool (MUST) v. Patient-Generated Subjective Global Assessment (PG-SGA): no risk/at risk patients represented in percentages. , PG-SGA well nourished; , PG-SGA undernourished.
Concordance analysis by the κ coefficient revealed a significant agreement between both methods, κ = 0·86 (P < 0·002). According to the Youden test, MUST had a high sensitivity of 0·80, P < 0·001 (95 % CI) and a specificity of 0·89, P < 0·001 (95 % CI), indicating a very high performance by comparison with the standard, and a strong capacity to effectively detect patients at nutritional risk. MUST when compared with PG-SGA had a 0·87 (P < 0·002) positive predictive value, and a negative predictive value of 1·0 (P < 0·001).
Percentage weight loss v. Malnutrition Universal Screening Tool
Given the previous results, to assess the value of significant percentage weight loss for routine use by comparison with the screening tool MUST turned out to be essential. By comparison with MUST, 304 patients (87 %) were correctly classified by percentage weight loss as without nutritional risk (true positives) and ninety-five (93 %) were correctly classified as being at risk (true negatives). Conversely, just seven (7 %) patients were falsely classified as at risk by percentage weight loss, though not at risk when assessed by MUST (false negatives); likewise, forty-four (13 %) were falsely classified as without risk by percentage weight loss, whereas they were assessed by MUST as at risk (false positives) (Fig. 5). According to the Youden test, percentage weight loss per se had a sensitivity of 0·85, P < 0·001 (95 % CI) and a specificity of 0·91, P < 0·001 (95 % CI) when compared with MUST. Thus, percentage weight loss had a positive predictive value of 0·90 (P < 0·001) and a negative predictive value of 1·0 (P < 0·001) v. MUST, hence showing a strong capacity to primarily identify patients likely to be at risk of undernutrition.
Percentage weight loss v. Malnutrition Universal Screening Tool (MUST): no risk (■)/at risk (□) patients represented in percentages.
By ROC interpreted relative areas under the curves that incorporated the relative scores given to each of the three MUST components, results shared percentage weight loss with a consistently superior statistical performance than the other variables, as well as the ability to detect mild to extreme nutritional changes.
Discussion
This is the first study aiming to validate MUST in the radiation oncology setting. In this prospective study in a large cohort of 450 cancer patients with different tumours, in order to reach a proper validation, single parameters widely acknowledged in the early detection of patients at nutritional risk or likely to become undernourished were tested. Our results demonstrated the validity and effectiveness of MUST in correctly identifying patients at nutritional risk by comparison with PG-SGA validated for nutritional status assessment in oncology, with a high sensitivity (0·80), specificity (0·89), 0·87 positive predictive value and negative predictive value of 1·0. The relevance of our results increases by taking into consideration the acknowledged fact that nutrition remains a mistreated distressing issue for cancer patients despite its impact in quality of life and survival(Reference Rivadeneira, Evoy and Fahey3–Reference Ravasco, Monteiro-Grillo and Vidal5).
PG-SGA remains the only validated and specific tool for a thorough nutritional assessment in oncology, and provides guidance for nutritional intervention though requiring a trained nutrition professional(Reference Ottery2, Reference Bauer, Capra and Ferguson17). In reality, there are not enough trained professionals for every patient; thereby the mandatory integration of PG-SGA in a nutrition protocol can only be effective by an antedated nutritional risk screening, in order to detect patients requiring full assessment and intervention(Reference Ottery2, Reference Kondrup, Allisson and Elia8, Reference Isenring, Capra and Bauer18–Reference Ferguson, Bauer and Gallagher20). The lack of a consensual method for nutritional screening, regardless of their abundance, prompted us to test MUST, a simple though thoroughly validated tool and easy to be applied by any adequately educated health professional(Reference Elia9, Reference Stratton, Hackston and Longmore10).
To identify undernutrition, BMI had a very low sensitivity, specificity, positive and negative predictive values: only 4 % of patients were underweight v. 29 % undernourished by PG-SGA; yet this was the only parameter able to identify the 73 % well-nourished/overweight/obese patients.
Undernutrition is often overlooked in patients with normal or even excessive BMI who have significantly lost weight(Reference Ottery21). Unintentional reported weight loss is a semi-objective criterion, whose meaning relies on cut-offs that reflect the boundaries between normal and abnormal intra-individual weight changes, and a possible underlying condition which when undetected may well result in further weight loss and undernutrition(Reference Stratton, Hackston and Longmore10). Nevertheless, unintentional weight loss is unquestionably a major issue in cancer and accordingly, percentage weight loss >5 % in the last 3–6 months represents high nutritional risk(Reference Kondrup, Allisson and Elia8, Reference Ravasco, Monteiro-Grillo and Vidal19); our results corroborated the consistency of percentage weight loss to detect mild to extreme nutritional changes. By comparison with PG-SGA, percentage weight loss proved to be a valid and reliable nutritional parameter in cancer, revealing high sensitivity, specificity, negative and positive predictive values to detect undernourished patients and those at risk of developing undernutrition. The high performance of percentage weight loss v. MUST is probably due to the fact that the three components of the latter reflect the patient ‘journey’ from the past (weight loss) to the present (current BMI) and into the future (effect of disease). Indeed, by comparison with MUST, the tool under validation in this study, it was confirmed that percentage weight loss per se had a strikingly high sensitivity of 0·85, a specificity of 0·91, a positive predictive value of 0·90 and a negative predictive value of 1·0, thus showing a strong capacity to primarily identify patients very likely to be at risk of undernutrition. The use of weight loss has, however, been often questioned, given the many possible affecting non-nutritional factors(Reference Gibson22).
MUST does not require time-consuming calculations, incorporates objective and subjective/clinical parameters reflecting changes in nutritional risk and status, and most importantly is easy to use and quick to fill in even by any trained professional without nutritional expertise(Reference Elia9, Reference Stratton, Hackston and Longmore10). It is the official tool used by the British Dietetic Association, the Royal College of Nursing, the Registered Nursing Homes Association and the British Association for Parenteral and Enteral Nutrition (BAPEN) and widely used in Europe. In line with recent literature(Reference Isenring, Capra and Bauer18), MUST showed that 31 % of patients were at moderate/high risk of undernutrition, confirmed by the 29 % of patients classified as undernourished by PG-SGA. Up until now, only one study has validated a screening tool in oncology v. PG-SGA; the Malnutrition Screening Tool was tested in a population of fifty patients, and the results showed similar sensitivity, specificity and predictive values(Reference Isenring, Cross and Daniels23). Another study in 144 cancer patients compared nutritional diagnoses obtained with different methods (SGA, MUST, BMI, serum albumin); only MUST and SGA proved to be good diagnostic tests with convergent results(Reference Pereira Borges, D'Alegria Silva and Cohen24).
Early screening for undernutrition eventually preceding a comprehensive nutritional assessment is acknowledged as imperative in the development of quality nutrition intervention protocols, to foster quality of care in oncology(Reference Ottery2, Reference Ravasco, Monteiro-Grillo and Vidal5, Reference Stratton, Hackston and Longmore10). Based on our results, MUST is strongly recommended to be integrated in routine screening in the radiation oncology setting, allowing for an appropriate reassessment schedule and nutrition care plan. Ideally, MUST should be the primary screening tool to refer patients in need of individualised schedules with nutritional professionals, which would then continue individual care with PG-SGA. The subsequent use of PG-SGA exacts every patient nutritional status and requirements. Though MUST is a quick (2–4 min) and easy-to-use tool, health professionals often declare themselves overwhelmed with the multiplicity of procedures. Bearing this in mind, and based on our results, percentage weight loss in the last 3–6 months is a valid and minimum parameter that can be used to predict nutritional risk in routine practice contributing to more effective nutritional care.
As limitations of our study, it is relevant to acknowledge that we included a heterogeneous population of cancer patients in terms of primary site, nutritional goals, radiation fields and prognosis. Furthermore, the study population is restricted to patients receiving radiotherapy and the results should not be generalised to patients who are candidates for systemic anticancer treatments. However, this study should and was designed to be used as an example for future studies in the oncology setting.
Acknowledgements
The authors are indebted to the helpful medical, nursing and technical staff of the Radiotherapy Department of the University Hospital of Santa Maria. The following were the contributions of each author to this study: C. B.-T. was responsible for the collection and assembly of the data; I. M.-G. contributed to the provision of patients, conception and design; M. C. and P. R. were involved in the conception and design, data interpretation and manuscript writing. The authors declare that there were no conflicts of interest with other people or organisations that could inappropriately influence their work. This study was partially supported by a Grant from the ‘Fundação para a Ciência e Tecnologia’ (RUN 437).
