Patient-generated subjective global assessment and classic anthropometry: comparison between the methods in detection of malnutrition among elderly with cancer.

INTRODUCTION A comparative study of the various methods of nutritional assessment currently available in oncology are necessary to identify the most appropriate one, as well as the relationships that exist among the different instruments. OBJECTIVE To compare the nutritional diagnosis obtained by the Patient-Generated Subjective Global Assessment (PG-SGA) with the objective anthropometric measurements in the elderly undergoing oncology treatment and to assess the concordance between the methods used in detecting malnutrition. METHODS A cross-sectional study of the elderly, older than or equal to 60 years in age undergoing oncology treatment. The PG-SGA was performed and the anthropometric parameters including weight, height, Body Mass Index (BMI), arm circumference, arm muscle circumference, corrected arm muscle area, arm fat area, calf circumference, waist circumference, hip circumference, waist-hip ratio and triceps skinfold were evaluated. From a 24-hour recall the energy and macronutrient intakes were estimated. RESULTS A total of 96 elderly patients were evaluated. The PG-SGA identified 29.2% with moderate malnutrition or suspected malnutrition and 14.6% with severe malnutrition. From among the elderly evaluated, 47.9% required critical nutritional intervention. The anthropometric parameters and the consumption of energy and macronutrients revealed significant differences based on the subjective PG-SGA classification. The prevalence of malnutrition ranged from 43.8% to 61.4%, depending upon the instrument used. The method most consistent with the diagnosis of malnutrition provided by the PGSGA was the BMI (kappa = 0.54; CI: 0.347-0.648). CONCLUSIONS The PG-SGA showed a significant correlation with the anthropometric measurements and with food consumption for both the categorical classification, as well as for the scoring system. Diagnosis of malnutrition showed variable prevalence depending upon the method used, and none were found equivalent to the PG-SGA.