Liver iron concentration quantification by MRI: are recommended protocols accurate enough for clinical practice?

Abstract

Objective: To assess the accuracy of quantification of liver iron concentration (LIC) by MRI using the Rennes University (URennes) algorithm.

Methods: In the overall study period 1999-2006 the LIC in 171 patients was calculated with the URennes model and the results were compared with LIC measured by liver biopsy.

Results: The biopsy showed that 107 patients had no overload, 38 moderate overload and 26 high overload. The correlation between MRI and biopsy was r=0.86. MRI correctly classified 105 patients according to the various levels of LIC. Diagnostic accuracy was 61.4%, with a tendency to overestimate overload: 43% of patients with no overload were diagnosed as having overload, and 44.7% of patients with moderate overload were diagnosed as having high overload. The sensitivity of the URennes method for high overload was 92.3%, and the specificity for the absence of overload was 57.0%. MRI values greater than 170 μmol Fe/g revealed a positive predictive value (PPV) for haemochromatosis of 100% (n=18); concentrations below 60 μmol Fe/g had a negative predictive value (NPV) of 100% for haemochromatosis (n=101). The diagnosis in 44 patients with intermediate values remained uncertain.

Conclusions: The assessment of LIC with the URennes method was useful in 74.3% of the patients to rule out or to diagnose high iron overload. The method has a tendency to overestimate overload, which limits its diagnostic performance.