ABSTRACT
Different laboratory methods are used to measure serum ferritin levels as a marker of iron status in the general population. This study aimed to compare serum ferritin levels using enzyme-linked immunosorbent assay (ELISA) versus immunochemiluminescence (Cobas e411) and immunoturbidimetric (Cobas Integra 400) methods in terms of sensitivity, specificity and accuracy, and whether they can be used interchangeably. A comparative cross-sectional study enrolled one hundred and six adult Yemeni patients (33 males and 73 females) aged 18-55 years, recruited from the dermatology and cosmetic center of Hadhramout Modern Hospital, Mukalla, Yemen. Serum ferritin levels were measured using ELISA, Cobas e411, and Cobas Integra 400 methods. For method comparison, a paired-sample t-test was used. For the consistency between the three methods, they were analyzed with regression and Pearson correlation coefficient. For determining accuracy, a receiver operating curve (ROC) was used. Bias error between the methods was determined through a Bland-Altman plot analysis. Our results did not show any significant statistical difference between ELISA and Cobas e411 (52.55 ± 7.4 µg/L vs. 52.58 ± 7.5 µg/L, p = 0.967), while there were significantly higher values from Cobas Integra 400 results than Cobas e411 (56.31 ± 7.8 µg/L vs. 52.58 ± 7.5 µg/L, p < 0.001) and ELISA (52.55 ± 7.4 µg/L vs. 56.31 ± 7.8 µg/L, p < 0.001). According to the correlation coefficient and linear regression analysis, a strong association between ELISA with Cobas e411 (r = 0.993, p < 0.001) and Cobas Integra 400 results (r = 0.994, p < 0.001) were revealed. For determining accuracy, Cobas e411 and Cobas Integra 400 results showed higher sensitivity (92.0%; 90.0%) and specificity (97.7%; 99.9%) respectively. Additionally, the Bland-Altman plot analysis showed a high agreement between the ELISA and Cobas e411 methods (bias: -0.035). In contrast, there was a low agreement between the ELISA and Cobas Integra 400 methods (bias: -3.75). Similarly, the agreement between Cobas e411 and Cobas Integra 400 methods was low (bias: -3.72). Serum ferritin levels were measured by Cobas e411, and Cobas Integra 400 methods were strongly correlated with the ELISA results, with higher sensitivity, specificity, and accuracy. However, further investigations with larger samples are required for improved accuracy and more precise results, and to determine whether they can be used interchangeably.
