Acurácia de critérios clínicos e laboratoriais para rastreamento de GCK-MODY a partir de dados clínicos do Brazilian Monogenic Diabetes Study Group (BRASMOD) e Brazilian Type 1 Diabetes Study Group (BrazDiab1SG)
Data
2022
Autores
Peixoto-Barbosa, Renata [UNIFESP]
Orientadores
Giuffrida, Fernando de Mello Almada [UNIFESP]
Tipo
Dissertação de mestrado
Título da Revista
ISSN da Revista
Título de Volume
Resumo
Objetivo: Avaliar a acurácia de parâmetros clínicos e laboratoriais de fácil obtenção na prática clínica no rastreio de GCK-MODY, em comparação a indivíduos portadores de diabetes tipo 1, usando duas coortes brasileiras de pacientes portadores de DM1 e MODY.
Métodos: Este estudo é uma análise retrospectiva de duas coortes brasileiras totalizando 2.687 pacientes portadores de diabetes tipo 1 (Grupo DM1), 202 pacientes com características clínicas sugestivas de MODY, mas sem mutação identificada (Grupo Sem Mutação) e 100 pacientes com mutação no gene GCK (Grupo GCK). O desempenho de preditores isolados e modelos multivariados de regressão logística na discriminação entre Grupo GCK vs. Grupo DM1 e Grupo GCK vs. Grupo Sem Mutação foi avaliado através da área sob a curva ROC (receiver operator characteristic curve), intervalos de confiança de 95% (IC 95%), valor preditivo negativo (VPN) e valor preditivo positivo (VPP).
Resultados: O melhor discriminante entre o grupo GCK e grupo DM1 foi um modelo multivariado composto por HbA1c, glicemia de jejum, idade ao diagnóstico, presença de hipertensão, presença de complicações microvasculares, história de cetoacidose diabética e história familiar de diabetes, alcançando AUC-ROC [95% CI] 0.980 [0.974-0.985], VPP 43.74% e VPN 100%. Comparando grupo GCK e Grupo Sem Mutação, não houve diferença estatisticamente significante entre os modelos. O modelo com acurácia nominalmente superior foi composto por hemoglobina glicada, idade ao diagnóstico, hipertensão e triglicerídeos, com AUC-ROC [95% CI] 0.850 [0.783-0.916]), VPP alcançou 88.36% e VPN 97.7%.
Conclusões: Nós demonstramos a viabilidade de um modelo de alta acurácia (98%) para diferenciar pacientes portadores de DM1 e GCK-MODY. O modelo pode ser útil na seleção mais apropriada dos pacientes que devem ser submetidos ao teste genético, permitindo assim, o tratamento adequado e uma melhor alocação dos recursos.
Objective: To evaluate the accuracy of routinely available clinical and laboratory parameters in the screening of GCK-MODY, leveraging data from the Brazilian Monogenic Diabetes Study Group (BRASMOD) and the Brazilian Type 1 Diabetes Study Group (BrazDiab1SG). Research design and methods: We included 2,687 patients with type 1 diabetes (T1D group), 202 patients with clinical features of MODY but without identified mutations (NoMut), and 100 patients with confirmed GCK-MODY (GCK). Performance of isolated predictors and multivariable logistic regression models in discriminating between GCK vs. T1D and GCK vs. NoMut was assessed by area under receiver operating characteristic curves (ROC-AUC), 95% confidence intervals (95% CI), negative predictive value (NPV), and positive predictive value (PPV). Results: The best discriminator between GCK and T1D groups was a multivariable model comprising HbA1c, fasting plasma glucose, age at diagnosis, presence of hypertension, presence of microvascular complications, previous diabetic ketoacidosis, and family history of diabetes, reaching ROC-AUC [95% CI] 0.980 [0.974-0.985]. PPV reached 43.74% and NPV 100%. Comparing GCK and NoMut groups, there was no statistically significant difference among models. The model with nominally higher accuracy comprehended HbA1c, age at diagnosis, hypertension, and triglycerides (ROC-AUC 0.850 [0.783-0.916]), reaching 88.36% PPV and 97.7% NPV. Conclusions: We have demonstrated the feasibility of a high-accuracy (98%) composite model to differentiate between GCK-MODY and type 1 diabetes, helping clinicians to select more appropriately those patients that should have genetic evaluation for monogenic diabetes and enabling them to establish correct treatment without overusing limited resources.
Objective: To evaluate the accuracy of routinely available clinical and laboratory parameters in the screening of GCK-MODY, leveraging data from the Brazilian Monogenic Diabetes Study Group (BRASMOD) and the Brazilian Type 1 Diabetes Study Group (BrazDiab1SG). Research design and methods: We included 2,687 patients with type 1 diabetes (T1D group), 202 patients with clinical features of MODY but without identified mutations (NoMut), and 100 patients with confirmed GCK-MODY (GCK). Performance of isolated predictors and multivariable logistic regression models in discriminating between GCK vs. T1D and GCK vs. NoMut was assessed by area under receiver operating characteristic curves (ROC-AUC), 95% confidence intervals (95% CI), negative predictive value (NPV), and positive predictive value (PPV). Results: The best discriminator between GCK and T1D groups was a multivariable model comprising HbA1c, fasting plasma glucose, age at diagnosis, presence of hypertension, presence of microvascular complications, previous diabetic ketoacidosis, and family history of diabetes, reaching ROC-AUC [95% CI] 0.980 [0.974-0.985]. PPV reached 43.74% and NPV 100%. Comparing GCK and NoMut groups, there was no statistically significant difference among models. The model with nominally higher accuracy comprehended HbA1c, age at diagnosis, hypertension, and triglycerides (ROC-AUC 0.850 [0.783-0.916]), reaching 88.36% PPV and 97.7% NPV. Conclusions: We have demonstrated the feasibility of a high-accuracy (98%) composite model to differentiate between GCK-MODY and type 1 diabetes, helping clinicians to select more appropriately those patients that should have genetic evaluation for monogenic diabetes and enabling them to establish correct treatment without overusing limited resources.