SD: standard deviation.

Data analysis

The collected data was recorded in a self-developed database and analyzed using the statistical package Stata 11.0 and VCC Stat Beta 3.0 software. For the descriptive analysis of the characterization variables, the relative frequency was calculated for categorical variables, while the mean and standard deviation (SD) were calculated for continuous variables.

The internal validity of the GLIM criteria was assessed through the internal consistency of the items using Cronbach’s alpha. Most authors suggest an acceptable range for reliability coefficients between 0.65 and 0.8, and values below 0.5 are considered unacceptable(24-27).

Concordance between the diagnostic categories obtained through the SGA and GLIM criteria was assessed using the kappa test (κ). The diagnostic categories compared were «without MN», «moderate MN,» and «severe MN.» Additionally, concordance was calculated considering two categories: «without MN» and «with MN.» The strength of agreement is classified as very weak when 𝜅 values are below 0.20, weak between 0.21 and 0.40, moderate between 0.41 and 0.60, good between 0.61 and 0.80, and very good with values above 0.80. The significance level was set at p-values < 0.05(28, 29).

All individuals belonging to the SGA categories “B” and “C” and GLIM categories “moderate MN” and “severe MN” were categorized as “MN”’. The validity of the GLIM criteria in predicting MN was assessed using specificity and sensitivity measures compared to SGA. The classification was based on the following cut-off points: “good” if both values were > 80 %, «fair» if one value was > 80 % and both were > 50 %, and «poor» if at least one value was < 50 %(30).

The safety of GLIM was analyzed using positive predictive value (PPV) and negative predictive value (NPV). Additionally, the positive likelihood ratio (LR+) and negative likelihood ratio (LR-) were calculated. LR+ values of 10 were considered to indicate “highly relevant utility,” while values between 5 and 10 were considered as “moderate utility.” For LR-, values below 0.1 were considered as “highly relevant utility,” and values between 0.1 and 0.2 were considered as “moderate utility”(31-33).

Results

The study included 123 hospitalized patients at risk of MN, with a mean age of 70 years (SD + 16.26), ranging from 20 to 96 years. In terms of gender distribution, 55.28 % were female, while 44.72 % were male. Among the participants, 50.41 % were active oncologic patients, and 30.89 % were surgical patients. The characterization of the study sample is provided in Table 2. Additionally, Table 3 presents the diagnoses obtained using each tool. According to the GLIM tool, the prevalence of MN was 91.05 %, with 39.02 % classified as moderate MN and 52.03 % as severe MN. The SGA tool identified MN in 88.6 % of cases, with 56.10 % categorized as moderate MN and 32.52 % as severe MN.

Table 2. Sample characteristics (n = 123)

*n: number of patients.

CI: confidence interval; GLIM: Global Leadership Initiative on Malnutrition; SGA: Subjective Global Assessment.

The GLIM tool demonstrated acceptable internal consistency, with a Cronbach’s alpha value of 0.6425. Regarding the “total item correlation,” the variable with stronger correlation with the GLIM alpha was “reduced muscle mass” (0.8326), while “reduced intake” (0.4416) showed a lower correlation (Table 4). The item “inflammation” was excluded from the analysis as it was present in 100 % of the sample. The concordance between GLIM and VGS was moderate, with a value of κ: 0.5946 (95 % confidence interval [CI]: 0.46-0.52; p < 0.00001) when considering three categories, and it was good with a value of κ: 0.7777 (95 % CI: 0.68-0.86; p < 0.00001) when considering two categories of the tools.

Table 3. 2 × 2 contingency table to evaluate the diagnostic tools

GLIM: Global Leadership Initiative on Malnutrition; SGA: Subjective Global Assessment.

Table 5 presents the results corresponding to the validity of the GLIM criteria classified as fair. The LR+ estimate demonstrated moderate utility, as the GLIM is 3.47 times more likely to categorize a malnourished patient as malnourished than a patient without MN being categorized as malnourished. On the other hand, the LR- estimate showed highly relevant utility, as it is 0.01 times less likely for a malnourished patient to be categorized as malnourished, compared to a patient without MN being categorized as malnourished.

Table 4. Cronbach’s alpha results

n: number of patients.

BMI: Body Mass Index; GLIM: Global Leadership Initiative on Malnutrition.

Table 5. Validity, diagnostic accuracy, safety, and likelihood ratio results of GLIM criteria using SGA as gold standard

CI: Confidence interval; LR+: Positive likelihood ratio; LR-: Negative likelihood ratio; MN: Malnutrition; NPV: Negative predictive value; PPV: Positive predictive value.

Discussion

The present study is one of the first in Latin America to compare SGA with GLIM in hospitalized patients at risk of MN. The diagnostic tools were exclusively applied, as proposed in clinical practice, in patients at nutritional risk, which would explain the high prevalence of MN found. The SNAQ has demonstrated both validity and reliability in hospitalized patients, exhibiting sensitivity and specificity levels exceeding 75 %. However, it is not exempt from excluding individuals from the study who could potentially be malnourished, which may lead to an overestimation of GLIM’s true performance(34-36).

The prevalence of MN, despite being high due to the aforementioned, was slightly higher using GLIM compared to SGA (91 % versus 88 %). Brito et al., obtained similar results (41.6 % vs. 33.9 %)(10). In contrast, Allard et al., reported a prevalence of 33.29 % versus 45.15 %, highlighting that their study’s retrospective design limited their analysis to the available GLIM data(8). The percentage of severe MN in the present study was much higher using GLIM (52.03 % vs. 32.52 %), as observed in Allard’s study (19.77 % vs. 11.73 %)(8).

On one hand, conducting a factorial analysis of each GLIM criterion reveals that if the item “reduced intake” were to be eliminated, the internal consistency would reach a level considered “good” (close to 0.7). On the other hand, the importance of the variable “reduced muscle mass” becomes evident, since, without it, the internal consistency would be even lower (alpha: 0.4147).

Therefore, it is necessary to establish cut-off points and operative criteria for evaluating muscle mass. Although anthropometry has limitations in terms of lower reliability compared to bioimpedance and dual-energy X-ray absorptiometry, in addition to the possible inter-observer bias(34), this study did not have access to such equipment. It is crucial to standardize a practical method for measuring muscle mass in clinical practice and establish its global use through protocolization(7).

Regarding the reliability of GLIM, the tools demonstrate significant agreement (p < 0.00001) in categorizing individuals into “without MN,” “moderate MN,” and “severe MN” categories, with a moderate level of agreement (0.5946, 95 % CI: 0.46-0.72). Considering the confidence interval, good reliability is achieved(12). The concordance between the tools is good (0.7777, 95 % CI: 0.68-0.86) when categorizing individuals as either “with MN” or “without MN”.

The sensitivity of GLIM in patients at nutritional risk (99.1 %) was in accordance with the recommended values (≥ 80 %), speculating that GLIM easily detects malnourished individuals(34). This was similarly demonstrated by Brito, et al. (86.6 %) and Burgel, et al. (86.8 %)(9, 10). In contrast, Allard et al., found a lower sensitivity (61.30 %)(8). Identifying patients with MN is crucial as it allows for appropriate nutritional treatment, leading to better outcomes, reduced complications, and lower healthcare costs(37).

The specificity of GLIM did not reach acceptable values (71.4 %). Being a diagnostic tool for MN, it should have high specificity to avoid categorizing individuals as malnourished when they are not. GLIM results can vary depending on the different tools used for measuring muscle mass(20). Therefore, as a strategy to achieve a specificity close to 80 %, it is proposed to investigate which of the tools is the most specific and to standardize its use.

The differences in sensitivity and specificity found in other studies could be due both to the disparity in their methodology and to the sample of patients included. Furthermore, the performance of each tool in different populations and the discrepancies in measurement techniques should be taken into account(8-10).

The high prevalence of MN may have influenced the elevated PPV (96.4 %), as it increases under conditions of higher prevalence. The NPV was also high, indicating that only 9.1 % of the patients without MN according to GLIM had MN according to SGA. Both values were higher compared to those reported by Allard et al. The discrepancies observed may be attributed to variations in the prevalence of MN among the studies(8, 38).

The entire study sample exhibited some degree of inflammation or stress, which may be associated with the high sensitivity and lower specificity of the tool. This finding is related to the lack of objective parameters, such as recommended supportive measures (e.g., C-reactive protein [CRP]), for its determination(8, 20).

Among the limitations of the present study, in addition to the previously mentioned use of a screened population with SNAQ, it is worth mentioning that food intake was assessed based on the patient’s or caregiver’s perception. Regarding the measurement of muscle mass, first-line methods recommended by GLIM, such as computed tomography (CT) and bioimpedance, were not available for this study. However, other simple and low-cost alternatives proposed by GLIM were used. Additionally, it would have been interesting to analyze each of the muscle mass measurement tools used separately, observing the performance of GLIM with each of them(7, 34).

During the course of the research, several advantages were observed in using GLIM compared to SGA. The latter involves a thorough physical examination and assessment of functional capacity, which requires more time and prior training to be conducted effectively.

It is important to highlight that the drafting of the research protocol and data collection preceded the validation guidelines of the GLIM criteria published by De van der Schueren and the GLIM Working Group(34). Nevertheless, its use is considered valuable for the development of future research. It is worth noting that, according to the GLIM authors, for concordance validation studies such as this one, the complete objective assessment of nutritional status should be considered as the gold standard method.

Finally, it should be mentioned that by including only those patients who were at risk of MN and not the entire population, the data from this study may only be extrapolated to populations with similar characteristics and not to hospitalized patients in general.

In the absence of a gold standard tool and considering GLIM criteria as a simple, rapid, objective, and inexpensive diagnostic tool to assess nutritional status in clinical practice, it is considered necessary to conduct additional multicenter and cohort follow-up studies to determine its ability to predict adverse clinical outcomes (hospital stays, clinical complications and mortality), and thus establish its predictive validity. For future research, it is suggested to include all hospitalized patients without excluding those without risk of MN, in addition to comparing the results obtained using different methods for measuring muscle mass, in order to establish the most convenient tool for clinical use.

Conclusions

The GLIM tool, applied in patients at nutritional risk, has acceptable internal consistency and moderate to good agreement compared to SGA, based on the utilized categorization. It also demonstrates fair validity, with higher sensitivity, but lower specificity compared to SGA. GLIM criteria achieved moderate diagnostic accuracy compared to SGA. Further studies on the tool’s validity are needed for its use in the hospital setting and to enable its global application.

Key points

• The GLIM criteria were proposed with the aim of reaching a worldwide consensus in defining diagnostic criteria for DRM.
• The objective of this study is to compare the GLIM diagnostic tool with the SGA in clinical practice in order to contribute to its validation in Latin America.
• The GLIM tool showed moderate to good diagnostic accuracy and concordance compared to the SGA when used in patients at nutritional risk.
• The internal consistency of the GLIM was found to be acceptable. The item “reduced muscle mass” was found to be relevant, suggesting the need for future research evaluating the application of different methods for measuring muscle mass in the GLIM criteria.
• In conclusion, the GLIM tool, when applied to patients at nutritional risk, demonstrated fair validity with higher sensitivity and lower specificity compared to the SGA.

Acknowledgments

Our most sincere thanks to:

• María Angélica Nadal, Head of the Dietetics Department at CEMIC University Hospital.
• Staff nutritionists of the Dietetics Department at CEMIC University Hospital.
• Nutrition residents from CEMIC University Institute.
• Dr. Juan Gili and Dr. Hugo Krupitzki, for their methodological advice.

Source of financing

This study received no funding.

Conflict of interest

The authors declare no conflicts of interest.

Statement of authorship

P. Navarro, O. Capelli, J. Adaglio, R. Barritta equally contributed to the conception and design of the acquisition, analysis, and interpretation of the data. P. Navarro and O. Capelli drafted the manuscript. All authors reviewed the manuscript, agreed to be fully responsible for ensuring the integrity and accuracy of the work, and read and approved the final manuscript.

Bibliographic references

1. Correia MITD, Perman MI, Waitzberg DL. Hospital malnutrition in Latin America: A systematic review. Clin Nutr. 2017;36(4):958-67. doi: 10.1016/j.clnu.2016.06.025.
2. Crivelli A, Perman M, Wyszynski D, Alomar F, Bellone M, De Loredo D, et al. Estudio AANEP 99: Prevalencia de desnutrición en hospitales de la Argentina [Internet]. RNC 2001;10:121-34 [cited 2021 May 13]. Available from: https://www.researchgate.net/publication/261287437_99_PREVALENCIA_DE_DESNUTRICION_EN_HOSPITALES_DE_LA_ARGENTINA_METODOLOGIA_DEL_ESTUDIO_Y_ANALISIS_PRELIMINAR_DE_LOS_RESULTADOS.
3. Canicoba M, de Baptista GA, Visconti G. Funciones y competencias del nutricionista clínico. Documento de consenso. Rev Cubana Aliment Nutr. 2013;23(1):146-72. ISSN: 1561-2929.
4. Epp LM, Salonen BR, Hurt RT, Mundi MS. Cross-sectional Evaluation of Home Enteral Nutrition Practice in the United States in the Context of the New Enteral Connectors. JPEN J Parenter Enteral Nutr. 2019;43(8):1020-7. doi: 10.1002/jpen.1510.
5. Van Bokhorst-de van der Schueren MAE, Guaitoli PR, Jansma EP, de Vet HCW. Nutrition screening tools: does one size fit all? A systematic review of screening tools for the hospital setting. Clin Nutr. 2014;33(1):39-58. doi: 10.1016/j.clnu.2013.04.008
6. Barbosa-Silva MCG, Barros AJD. Indications and limitations of the use of subjective global assessment in clinical practice: an update. Curr Opin Clin Nutr Metab Care. 2006;9(3):263-9. doi: 10.1097/01.mco.0000222109.53665.ed.
7. Cederholm T, Jensen GL, Correia MITD, Gonzalez MC, Fukushima R, Higashiguchi T, et al. GLIM criteria for the diagnosis of malnutrition - A consensus report from the global clinical nutrition community. J Cachexia Sarcopenia Muscle. 2019;10(1):207-17. doi: 10.1002/jcsm.12383.
8. Allard JP, Keller H, Gramlich L, Jeejeebhoy KN, Laporte M, Duerksen DR. GLIM criteria has fair sensitivity and specificity for diagnosing malnutrition when using SGA as comparator. Clin Nutr. 2020;39(9):2771-7. doi: 10.1016/j.clnu.2019.12.004.
9. Burgel CF, Eckert I da C, Brito JE, Rodrigues FW, Silva FM. Accuracy of three tools for malnutrition diagnosis in hospitalised patients: Comparison to subjective global assessment. J Hum Nutr Diet. 2021;34(6):935-44. doi: 10.1111/jhn.12907.
10. Brito JE, Burgel CF, Lima J, Chites VS, Saragiotto CB, Rabito EI, et al. GLIM criteria for malnutrition diagnosis of hospitalized patients presents satisfactory criterion validity: A prospective cohort study. Clin Nutr. 2021;40(6):4366-72. doi: 10.1016/j.clnu.2021.01.009.
11. Galindo Martín CA, Aportela Vázquez VA, Becerril Hernández F, Aguilar Medina CR, Ayala Carrillo SL, Chávez Flores A, et al. The GLIM criteria for adult malnutrition and its relation with adverse outcomes, a prospective observational study. Clin Nutr ESPEN. 2020;38:67-73. doi: 10.1016/j.clnesp.2020.06.015.
12. Jiménez Villa J, Argimon Pallás JM. Elaboración del protocolo de estudio. Tamaño de la muestra. In: Jiménez Villa J, Argimon Pallás JM (editors). Métodos de investigación clínica y epidemiológica. 4.th edition. Barcelona: Elsevier; 2013. p. 142-54.
13. Hernández R, Fernández C, Baptista P. Muestreo en la investigación cualitativa. In: Hernández R, Fernández C, Baptista P (editors). Metodología de la investigación. 6.th edition. Mexico F.D.: McGraw-Hill Interamerican; 2014. p. 390-2.
14. Detsky AS, McLaughlin JR, Baker JP, Johnston N, Whittaker S, Mendelson RA, et al. What is subjective global assessment of nutritional status? JPEN J Parenter Enteral Nutr. 1987;11(1):8-13. doi: 10.1177/014860718701100108.
15. da Silva Fink J, Daniel de Mello P, Daniel de Mello E. Subjective global assessment of nutritional status – A systematic review of the literature. Clin Nutr. 2015;34(5):785-92. doi: 10.1016/j.clnu.2014.12.014.
16. Normative grip strength data [Internet]. Patterson Medical Jamar Smart User Manual-Hand dynamometer [cited 2021 July 28]. Available from: https://www.manualslib.com/manual/1272692/Patterson-Medical-Jamar-Smart.html?page=10.
17. Frisancho AR. New norms of upper limb fat and muscle areas for assessment of nutritional status. Am J Clin Nutr. 1981;34(11):2540-5. doi: 10.1093/ajcn/34.11.2540.
18. Guastavino P, Llames L. Principios para la valoración del estado nutricional. In: Canicoba, M. Mauricio S (editors). Valoración del estado nutricional en diversas situaciones clínicas. 1.st edition. Lima, Peru: Private University of the North; 2017. p. 98-110.
19. Lera L, Sánchez H, Ángel B, Albala C. Mini Nutritional Assessment Short-Form: Validation in Five Latin American Cities. SABE Study. J Nutr Health Aging. 2016;20(8):797-805. doi: 10.1007/s12603-016-0696-z.
20. Contreras-Bolívar V, Sánchez-Torralvo FJ, Ruiz-Vico M, González-Almendros I, Barrios M, Padín S, Alba E, et al. GLIM Criteria Using Hand Grip Strength Adequately Predict Six-Month Mortality in Cancer Inpatients. Nutrients. 2019;11(9):2043. doi: 10.3390/nu11092043.
21. Sousa-Santos AR, Amaral TF. Differences in handgrip strength protocols to identify sarcopenia and frailty - a systematic review. BMC Geriatr. 2017;17(1):238. doi: 10.1186/s12877- 017-0625-y.
22. Hogrel JY. Grip strength measured by high precision dynamometry in healthy subjects from 5 to 80 years. BMC Musculoskelet Disord. 2015;16(1):139. doi: 10.1186/s12891-015-0612-4.
23. Belcher HJCR, Smith H. Extended dynamometry: reference values. J Hand Surg Eur Vol. 2019;44(2):196-202. doi:10.1177/1753193418805959.
24. Donis JH. Evaluación de la validez y confiabilidad de una prueba diagnóstica. Avances en Biomedicina. 2012;1(2):73-81.
25. Goforth C. Using and Interpreting Cronbach’s Alpha. Research Data Services and Sciences [Internet]. University of Virginia Library; 2015 [cited 2021 May 13]. Available from: https://data.library.virginia.edu/using-andinterpreting-cronbachs-alpha/.
26. Carvajal A, Centeno C, Watson R, Martínez M, Sanz Rubiales Á. ¿Cómo validar un instrumento de medida de la salud? An Sist Sanit Navar. 2011;34(1): 63-72.
27. Burgos ME, Manterola C. Cómo interpretar un artículo sobre pruebas diagnósticas. Rev Chil Cir. 2010;62(3):301-8. doi: 10.4067/S0718- 40262010000300018.
28. Jiménez Villa J, Argimon Pallás JM. Anexo 3. Sensibilidad y especificidad. In: Jiménez Villa J, Argimon Pallás JM (editors). Métodos de investigación clínica y epidemiológica. 4.th edition. Barcelona: Elsevier; 2013. p. 339-44.
29. Jiménez Villa J, Argimon Pallás JM. Anexo 4. Análisis de la concordancia. In: Jiménez Villa J, Argimon Pallás JM (editors). Métodos de investigación clínica y epidemiológica. 4.th edition. Barcelona: Elsevier; 2013. p. 345-9.
30. Power L, Mullally D, Gibney ER, Clarke M, Visser M, Volkert D, et al. A review of the validity of malnutrition screening tools used in older adults in community and healthcare settings - A MaNuEL study. Clin Nutr ESPEN. 2018;24:1-13. doi: 10.1016/j.clnesp.2018.02.005.
31. Silva Fuente-Alba C, Molina Villagra M. Likelihood ratio (razón de verosimilitud): definición y aplicación en Radiología. Rev Argent Radiol/Argent J Radiol. 2017;81(3):204-8. doi: 10.1016/j-rard.2016.11.002.
32. Otero W, Pineda LF, Beltran LH. Utilidad de la razón de verosimilitud (likelihood ratio) en la práctica clínica. Rev Colomb Gastroenterol. 2001;16(1):33-36.
33. García GJJ. Significado y empleo de la razón de probabilidades en la práctica clínica. Rev Mex Pediatr. 2000;67(4);188-191.
34. de van der Schueren MAE, Keller H, Cederholm T, Barazzoni R, Compher C, Correia MITD, et al. Global Leadership Initiative on Malnutrition (GLIM): Guidance on validation of the operational criteria for the diagnosis of protein-energy malnutrition in adults. Clin Nutr. 2020;39(9):2872-80. doi: 10.1016/j.clnu.2019.12.022.
35. Neelemaat F, Meijers J, Kruizenga H, van Ballegooijen H, van Bokhorst-de van der Schueren MAE. Comparison of five malnutrition screening tools in one hospital inpatient sample. J Clin Nurs. 2011;20(15-16):2144-52. doi: 10.1111/j.1365-2702.2010.03667.x.
36. Kruizenga HM, Seidell JC, de Vet HCW, Wierdsma NJ, van Bokhorst-de van der Schueren MAE. Development and validation of a hospital screening tool for malnutrition: the short nutritional assessment questionnaire (SNAQ©). Clin Nutr. 2005;24(1):75-82. doi: 10.1016/j.clnu.2004.07.015.
37. Palma Milla S, Meneses Gonzalez D, Valero Pérez M, Calso González M, García Vázquez N, Ruiz Garrido M, et al. Costes asociados a la desnutrición relacionada con la enfermedad y su tratamiento: revisión de la literatura. Nutr Hosp. 2018;35(2):442-60. doi: 10.20960/nh.1204.
38. Moriana M, Civera M, Artero A, Real JT, Caro J, Ascaso JF, et al. Validity of subjective global assessment as a screening method for hospital malnutrition. Prevalence of malnutrition in a tertiary hospital. Endocrinol Nutr. 2014;61(4):184-9. doi: 10.1016/j.endonu.2013.10.006.