[Meta-analysis of studies on cut-off value of serum ferritin for identifying iron deficiency]

To synthesize evidence to discuss the appropriate cut-off value of serum ferritin (SF) for identifying iron deficiency.

A comprehensive electronic search and manual tracking were performed to collect potential studies on cut-off value of SF for identifying iron deficiency. Most relevant studies were identified according to the inclusion and exclusion criteria and data were extracted. All included studies were divided into three groups according to the cutoff value of SF: 12 -20 microg/L group, 25 microg/ L or 30 microg/L group and 36 - 60 microg/L group. Heterogeneity of the included studies was tested to select proper efficacy model for calculating pooled weighted diagnostic indicators and their 95% CI. Summary receiver operating characteristic (SROC) curve was made and the area under the curve (AUC) and Q * index were calculated respectively for the three groups. Finally, sensitivity analysis were performed.

A total of 14 studies were included and were divided into three groups. After the pooled efficacy analysis, we found that the 12 - 20 microg/L group showed the lowest pooled sensitivity (0.767, 95 CI 0.705 -0.821) but highest specificity (0.959, 95 CI 0.934 -0.976). Compared with the other two groups, the pooled sensitivity of 25, 30 microg/L group (0.877, 95 CI 0.799 - 0.933) was highest, its pooled specificity was (0.944, 95 CI 0.888 - 0.977), the pooled sensitivity and pooled specificity of 36 - 60 microg/L group (0.836, 95 CI 0.797 - 0.870), (0.876, 95 CI 0.846 - 0.901)] were both relatively low. In terms of the integrated diagnostic capabilities, the 25, 30 microg/L group showed the highest diagnostic odds ratio (101.42, 95 CI 36.137 - 284.64), the largest AUC (0.9497 +/- 0.039) and the Q * index nearest 1 (0.8901 +/- 0.052) among the three groups. After we removed the studies with samples less than 50 and the studies from Chinese authors, the 25, 30 microg/L group still showed the best diagnostic efficiency.

25 or 30 microg/L of SF as the cut-off value of identifying iron deficiency shows high accuracy and good integrated diagnostic capability.