Threshold Ferritin Values to Predict Control of Liver Iron Burden in Thalassemia

Transfusion-dependent (TD) and non-transfusion dependent (NTD) forms of thalassemia are associated with progressive iron overload reflected by increase in serum ferritin and liver iron concentration (LIC). Despite the adoption of non-invasive methods, measurement of LIC is available to only a minority of patients or can only be performed annually. Maintenance of low systemic iron usually depends on frequent determination of ferritin, a less reliable marker whose level can vary among individuals with similar LIC. We evaluated the relationship between LIC and ferritin in TD and NTD (including un-transfused (NT) or intermittently transfused (TI)) individuals to determine the validity of average annual ferritin to identify inadequate control of liver iron (defined as LIC >7 mg/g dry-liver-weight).

After excluding individuals with active hepatitis C virus infection, the LIC (measured by Ferritometer) and serum ferritin values were available from 115 TD (729 observations), 36 TI (139 observations) and 14 NT (28 observations) individual patients.

The median (range) ferritin and LIC were 1,684 (91-14,155) ng/mL and 10.2 (0.14- 59.4) mg/g for TD, 469 (50-4,536) ng/mL and 8.8 (2.4-33.1) mg/g for TI, and 407 (30-1,061) ng/mL and 5.7 (1.3-22.8) mg/g for NT groups, respectively. The slope of the regression line between ferritin and LIC was significantly different between the 3 groups (p=0.019), being lowest in TD, intermediate in TI, and highest in NT. The predicted LIC for ferritin level of 1000 ng/mL was 8.1 mg/g for TD, 12.1 mg/g for TI, and 13.5 mg/g for NT groups. Receiver operating characteristics (ROC) analysis showed that in NTD group (NT+TI), a ferritin threshold of >200, >300 and >500 ng/mL failed to correctly identify 7%, 21% and 38% of individuals with LIC >7 mg/g, respectively. In contrast, in the TM group, these ferritin values were associated with false negative rate of 0%, 0% and 2%, respectively. In a subgroup analysis restricted to TM patients with ferritin <1,000 ng/mL, LIC varied from 0.14 mg/g to 21.1 mg/g. ROC analysis in this subgroup showed no ability of ferritin to discriminate between individuals with LIC <3 mg/g or ≥3 mg/g (AUC 0.54, P=n.s.).

In conclusion, transfusions significantly alter the relationship between serum ferritin and LIC in thalassemia. In TM individuals, ferritin levels do not identify those with very low LIC. NTD patients with seemingly safe serum ferritin levels may have clinically significant hemosiderosis and be at risk for organ injury. These relationships should be factored into making decisions about screening or treatment of iron overload in untransfused or intermittently transfused individuals with thalassemia.