Evaluation of bone marrow iron stores in anemia associated with chronic disease: a comparative study of serum and red cell ferritin.

The serum ferritin (SERFER) may be elevated and misleading in the setting of chronic disease (chronic inflammation, liver disease, and neoplasm). The red cell ferritin (RBCFER) may be more stable in clinical situations that affect the SERFER. We compared the ability of SERFER and RBCFER to assess iron stores in these settings. Iron stores were defined by bone marrow aspirate staining. We studied 120 anemic (Hb < 14 g/dl) male patients. Twenty-eight (23%) were iron deficient based on the absence of marrow iron. The SERFER correlation with marrow iron stores (r = 0.58; P < 0.001) was better than the RBCFER (r = 0.36; P < 0.001). Cutoff values for the diagnosis of iron deficiency were determined by chi-square analysis (SERFER < 70 ng/ml; RBCFER < or = 4 ag/RBC). The sensitivity for detecting iron deficiency with SERFER (0.60) was less than RBCFER (0.82). The specificity of SERFER (0.90) was slightly better than RBCFER (0.83). Neither difference reached statistical significance (P > 0.05). The positive predictive value between the two assays was the same (SERFER 0.65 vs. RBCFER 0.59). The combination of SERFER < 70 ng/ml with RBCFER < or = 4 ag/RBC was more specific (0.97) when compared with the SERFER alone (0.90) (P = 0.04). In addition, the potential of this combination to predict iron deficiency (0.82) was higher than that seen with either SERFER (0.65) or the RBCFER (0.59). Our findings show that the RBCFER as a single assay is not anymore accurate than the SERFER. However, we find that the RBCFER can effectively complement the SERFER to either predict iron depletion or confirm the presence of bone marrow iron.