A predictive model for estimating the number of erythrocytapheresis or phlebotomy treatments for patients with naïve hereditary hemochromatosis

Abstract

Background and aims: Standard treatment for naïve hereditary hemochromatosis patients consists of phlebotomy or a personalized erythrocytapheresis. Erythrocytapheresis is more efficient, but infrequently used because of perceived costs and specialized equipment being needed. The main aim of our study was to develop a model that predicts the number of initial treatment procedures for both treatment methods. This information may help the clinician to select the optimal treatment modality for the individual patient.

Methods: We analyzed retrospective data of 125 newly diagnosed patients (C282Y homozygous), treated either with phlebotomy (n = 54) or erythrocytapheresis (n = 71) until serum ferritin (SF) reached levels ≤100 μg/L. To estimate the required number of treatment procedures multiple linear regression analysis was used for each treatment method separately.

Results: The linear regression model with the best predictive quality (R² = 0.74 and 0.73 for erythrocytapheresis and phlebotomy respectively) included initial SF, initial hemoglobin (Hb) level, age, and BMI, where initial SF was independently related to the total number of treatment procedures for both treatment methods. The prediction error expressed in RMSPE and RMSDR was lower for erythrocytapheresis than for phlebotomy (3.8 and 4.1 vs 7.0 and 8.0 respectively), CONCLUSIONS: Although the prediction error of the developed model was relatively large, the model may help the clinician to choose the most optimal treatment method for an individual patient. Generally erythrocytapheresis halves the number of treatment procedures for all patients, where the largest reduction (between 55% and 64%) is reached in patients with an initial Hb level ≥ 9 mmol/L (14.5 g/dL). ClinicalTrials.gov number NCT00202436.

Keywords: erythrocytapheresis; hereditary hemochromatosis; phlebotomy; prediction rule.

FIGURE 1

A, Observed versus predicted number of treatment procedures using erythrocytapheresis as treatment modality. Predicted numbers are based on model B, using initial SF, initial Hb count, age and BMI as predictors. (R ² = 0.74; RMSPE = 3.8; RMSDR = 4.1). The Y = X line is presented, where points on this line imply that the predicted and observed numbers were the same. B, Observed vs predicted number of treatment procedures using phlebotomy as treatment modality. Predicted numbers are based on model B, using initial SF, initial Hb count, age and BMI as predictors (R ² = 0.73; RMSPE = 7.0; RMSDR = 8.0). The Y = X line is presented, where points on this line imply that the predicted and observed numbers were the same