Diagnostic accuracy of PG-SGA, mPG-SGA, and GLIM criteria in malnutrition detection and survival prediction in patients with gastric cancer.

BACKGROUND

This study sought to assess the effectiveness of PG-SGA, mPG-SGA and GLIM criteria for diagnosing malnutrition in individuals with gastric cancer (GC).

METHODS

A total of 906 patients with confirmed GC diagnosis were recruited for this study. Nutritional screening and diagnosis were performed using NRS2002, AIWW, the PG-SGA, modified PG-SGA (mPG-SGA) and the GLIM criteria. The effectiveness of these instruments was evaluated through multiple metrics, encompassing sensitivity, specificity, positive and negative predictive values along with Kappa (κ) values. Kaplan-Meier and multivariable Cox analyses were used to assess survival.

RESULTS

Among 906 patients, the diagnosis of malnutrition was made in 76.7% of the patients using the PG-SGA, 73.2% using the mPG-SGA, 55.5% using the GLIM with NRS2002, and 65.3% using the GLIM with AIWW. In the mPG-SGA, GLIM with NRS2002, and GLIM with AIWW, sensitivity was 94.7%, 68.1%, and 78.0%, while specificity was 97.60%, 85.8%, and 76.30%. When comparing malnutrition identification against PG-SGA, the mPG-SGA demonstrated nearly perfect agreement with PG-SGA (κ = 0.88), while the GLIM criteria displayed fair agreement with PG-SGA, revealing a κ value of 0.41 for PG-SGA versus GLIM with NRS2002 and 0.46 (P < 0.001) for PG-SGA versus GLIM with AIWW (P < 0.001). Adjusted hazard ratios for mortality ranged from 1.49 to 1.73 across the four malnutrition assessment tools (all P < 0.05).

CONCLUSIONS

The mPG-SGA exhibited comparable prognostic performance to PG-SGA and GLIM in individuals with GC and may function as a practical alternative for rapid nutritional assessment in clinical environments.