DISTINCTION BETWEEN HOMOZYGOUS AND HETEROZYGOUS SUBJECTS WITH HEREDITARY HEMOCHROMATOSIS USING IRON STATUS MARKERS AND RECEIVER OPERATING CHARACTERISTIC (ROC) ANALYSIS

Citation
N. Milman et Mj. Albeck, DISTINCTION BETWEEN HOMOZYGOUS AND HETEROZYGOUS SUBJECTS WITH HEREDITARY HEMOCHROMATOSIS USING IRON STATUS MARKERS AND RECEIVER OPERATING CHARACTERISTIC (ROC) ANALYSIS, European journal of clinical chemistry and clinical biochemistry, 33(2), 1995, pp. 95-98
Citations number
15
Categorie Soggetti
Biology,"Chemistry Medicinal
ISSN journal
09394974
Volume
33
Issue
2
Year of publication
1995
Pages
95 - 98
Database
ISI
SICI code
0939-4974(1995)33:2<95:DBHAHS>2.0.ZU;2-V
Abstract
The purpose of the present study was: 1) to evaluate which of the four iron status markers (serum iron, serum transferrin, serum transferrin saturation, serum ferritin) displayed the highest discriminatory pote ntial in the distinction between homozygous patients with hereditary h aemochromatosis and heterozygous relatives. 2) to suggest optimum cut- off values for these iron status markers, and 3) to demonstrate how th ese cut-off values change if the expected utility from a correct diagn osis is incorporated into the analysis. The patients and relatives wer e found by a nation-wide epidemiological survey. The study population consisted of 162 patients with clinically overt hereditary haemochroma tosis and 84 asymptomatic heterozygous relatives. The statistical eval uation was performed using receiver operating characteristic (ROC) cur ve analysis. The diagnostic power of the iron markers is expressed as the area under the ROC curve. The optimum cut-off value is at the poin t where the slope of the ROC curve is equal to one. Changes in the opt imum cut-off value at varying expected utility from a correct classifi cation was estimated by changing the scaling of the ROC diagram. Serum iron and serum transferrin had the smallest area under the ROC curve, and were both unsuitable as discriminators. Near complete discriminat ion was obtained with serum transferrin saturation and serum ferritin concentrations, displaying the largest area under the ROC curve (0.991 and 0.998). The optimum threshold value for transferrin saturation wa s 61%, and for serum ferritin concentration 800 mu g/l. The transferri n saturation level reflects the presence of the haemochromatosis allel e, whereas the serum ferritin concentration indicates the degree of ir on overload. Changes in the optimum cut-off value from varying expecte d utility implicated the importance of incorporating this discriminato r.