Pediatric Prediction Model for Low Immunoglobulin G Level Based on Serum Globulin and Illness Status

Abstract

Hypogammaglobulinemia is a condition that requires prompt diagnosis and treatment. Unfortunately, serum immunoglobulin (Ig) measurements are not widely accessible in numerous developing countries. Serum globulin is potentially the best candidate for screening of low IgG level (IgGLo) due to its high availability, low cost, and rapid turnover time. However, multiple factors may influence the probability of prediction. Our study aimed to establish a simple prediction model using serum globulin to predict the likelihood of IgGLo in children. For retrospective data of patients who were suspected of having IgGLo, both serum IgG and globulin were simultaneously collected and measured. Potential factors interfering with serum globulin and IgG levels were investigated for their impact using bivariate binary logistic regression. A multivariate binary logistic regression was used to generate a formula and score to predict IgGLo. We obtained 953 samples from 143 pediatric patients. A strong positive correlation between serum globulin and IgG levels was observed (r=0.83, p < 0.001). A screening test model using serum globulin and illness status was constructed to predict IgGLo. The formula for predicting IgGLo was generated as follows; Predicted score = (2 x globulin (g/dl)) – illness condition score (well=0, sick=1). When the score was <4, the patient has the probability of having IgGLo with a sensitivity of 0.78 (0.71, 0.84), a specificity of 0.71 (0.68, 0.74), PPV of 0.34 (0.29, 0.40) and NPV of 0.94 (0.92, 0.96). This formula will be useful as rapid and inexpensive screening tool for early IgGLo detection, particularly in countries/locations where serum IgG measurement is inaccessible.