Premium
Sensitivity and Specificity of Serum Retinol in Determining Vitamin A Deficiency
Author(s) -
Suri Devika,
Tanumihardjo Jake,
Gan Bryan,
Tanumihardjo Sherry
Publication year - 2015
Publication title -
the faseb journal
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.709
H-Index - 277
eISSN - 1530-6860
pISSN - 0892-6638
DOI - 10.1096/fasebj.29.1_supplement.403.8
Subject(s) - vitamin a deficiency , false positive paradox , medicine , gastroenterology , vitamin , xerophthalmia , retinol , pediatrics , machine learning , computer science
Vitamin A deficiency (VAD) afflicts 190 million children worldwide, increasing the risk of morbidity, mortality and blindness. Serum retinol concentration (SR), a common indicator of VA, is not always associated with liver VA reserves due to homeostatic control and during infection. Sensitivity and specificity are used to evaluate the utility of clinical tests to determine presence of disease. Our objective was to determine the sensitivity and specificity of SR as an indicator of VAD in two rural populations of children from Zambia and Thailand. VA deficiency was based on cutoffs of <0.1 umol/g for liver VA stores (LVA) calculated using stable isotope dilution and <0.7 umol/L for SR. No Zambian children were found deficient using LVA, but 17% were deficient using SR; no children were deficient using SR in the Thai children, but 68% had VAD using LVA. Because there were no true positive cases of VAD in the Zambia children, sensitivity could not be calculated. Specificity was 83%, meaning that 83% of true negative VAD cases were identified, but 27% were false positives. Conversely in the Thai children there were no VAD cases identified by SR, so there were no true or false positives and thus sensitivity was 0%. Specificity was 100%, meaning that all cases identified as not VAD were true negatives. If there was presence of infection as indicated by elevated CRP, AGP or malaria, specificity was reduced; that is, SR was less likely to identify false positives in children without infection in Zambia. We found that specificity of SR as an indicator of VAD was between 83‐100%, but sensitivity could not be calculated because in these populations there were no children with VAD as identified by both SR and LVA. Presence of infection reduced the specificity of SR for determining VA status.