Abstract:
Rationale:
In patients with liver cirrhosis, malnutrition is of particular clinical importance. Malnutrition is a state of muscle or total body protein depletion. There is no international consensus on a single gold standard nutritional assessment tool as different tools have been developed based on different populations, age groups, and settings. The aim of this study was to therefore validate a range of approaches for diagnosis of malnutrition against an objective measurement of protein depletion in patients with liver cirrhosis, using nutritional assessment tools such as the Royal Free Hospital Global Assessment scheme (RFH-GA), developed for patients with liver disease, and the recently proposed Global Leadership Initiative on Malnutrition Criteria (GLIM) criteria.
Methods:
A retrospective analysis was undertaken of 231 liver cirrhosis patients (153 M, 78 F) who were referred for nutrition support and underwent: determination of total body protein (TBP) by in vivo neutron activation analysis (IVNAA); fat-free mass (FFM) and appendicular skeletal muscle mass (ASMM) measurements by DXA; FFM and total body water (TBW) measurements by bioelectrical impedance analysis (BIA); body weight (BW); mid-arm muscle circumference (MAMC) by anthropometry; and handgrip strength (HGS) measurements by dynamometry. Patients were assessed as significantly protein depleted (malnourished) if measured TBP was < 82% of that predicted from regression equations based on age, sex, height, and pre-illness weight established in healthy volunteers. Percent BW loss was determined from current BW and pre-illness weight recalled by the patient. TBW was also calculated using a multicompartment approach where total body fat by DXA, TBP, bone mineral content by DXA, and estimated non-bone minerals and glycogen were subtracted from BW. These TBW results were used to correct BW, % weight loss, body mass index (BMI), FFM, and ASMM. FFM and ASMM were also indexed to the square of height as FFMI and ASMI. A HGS index (GSI) was calculated as the ratio of HGS to predicted grip strength based on regression equations developed in healthy volunteers. Malnutrition was assessed using the RFH-GA scheme along with various GLIM criteria and comparisons made with the reference measure. Results:
The application of the reference method (PI) yielded an overall malnutrition prevalence of 51% and prevalence for males was significantly higher than for females (P < 0.0001). The application of the RFH-GA scheme based on body mass index corrected to normal hydration (BMIc) yielded an overall malnutrition prevalence of 52%, in agreement with the reference measure, but did not statistically differ between genders (P = 0.406). The RFH-GA scheme showed fair sensitivity (61%) and specificity (58%), low PLR (positive likelihood ratio) (1.47) and a large NLR (negative likelihood ratio) (0.67), with a diagnostic accuracy of 60%. Among the various GLIM models, the prevalence of malnutrition of the total population ranged from 5% to 73%, compared with 51% identified by the reference. Only GLIM models 2 (corrected BW), 6 (corrected ASMI), 8 (FFMI by BIA), and 10 (corrected FFMI by BIA) were in overall agreement with the reference measure. When separated into sex groups, GLIM models 2, 6, and 10 were in agreement with the reference measure for males (63%). GLIM models 1 (percent BW loss), 2, 6, 8, and 10 were in agreement with the reference measure for females (27%). In terms of validity, GLIM 10 showed the highest overall sensitivity (80%) with poor specificity (35%) and a low PLR (1.22). GLIM 3 (BMI) showed the highest overall specificity (96%) with very poor sensitivity (6%) and a low PLR (1.34). Males showed 100% specificity in GLIM 3 and 7 (FFMI by DXA). The highest PLR (4.27) was observed in GLIM 12 (GSI) for females, and GLIM 12 also showed the highest overall PLR (3.76). GLIM 2 showed the highest overall diagnostic accuracy (70%). AUC was established for the various GLIM criteria by ROC curve analysis. Males in GLIM 6, 9 (corrected FFMI by DXA), 2 criteria all showed good accuracy when validated against PI, AUC of 0.847, 0.844, 0.837, respectively. Females also showed good accuracy when validating against PI, for GLIM 2 criterion (0.800).
Conclusion:
The prevalence of malnutrition using the RFH-GA scheme showed good agreement with the reference measure. However, it could not detect the sex difference in malnutrition prevalence that was clear with the reference measure. In terms of validity, it showed overall fair sensitivity and specificity with a low positive likelihood ratio. Of the 12 GLIM criteria evaluated, the prevalence of malnutrition and validity of each criterion varied according to which phenotypic parameter and cut-offs were used. The importance of adjusting for overhydration is highlighted. Further research based on prospective studies will confirm or refine this interpretation to optimise validity and provide greater
detail on what body composition measures should be used and the specific cut-off points that are optimal for determining reduced muscle mass.