Prediction of the chance of successful immune tolerance induction in persons with severe hemophilia A and inhibitors: a clinical prediction model

Abstract

Background: Inhibitor eradication to restore factor (F)VIII efficacy is the treatment goal for persons with severe hemophilia A (HA) and inhibitors. Immune tolerance induction (ITI) is demanding and successful in about 70% of people. Until now, it has remained difficult to quantify the probability of ITI success or failure, complicating the decision to initiate or not initiate ITI. Estimating the individual chance of ITI success allows clinicians, patients, and their families to support shared decision-making.

Objectives: We aimed to identify clinical predictors of ITI success and to develop a clinical prediction model to estimate the chance of successful ITI in persons with severe HA.

Methods: This multicenter study included persons with severe HA who received ITI. Clinical data were collected. Successful ITI was defined by a negative inhibitor titer and an adequate response to FVIII concentrates. A multivariable logistic regression model was developed. Model performance and internal validation were performed.

Results: Of 206 participants with a median age of 19.8 months (IQR, 12.1-38.8) at ITI start, 148 (71.8%) achieved ITI success. Our clinical prediction model included 4 predictors of ITI success: cumulative number of FVIII exposure days at inhibitor development, peak inhibitor titer, ethnicity, and F8 mutation type. The C statistic was 0.801 (95% CI, 0.70-0.87).

Conclusion: In our study, including 206 people with severe HA and inhibitors, we developed a clinical prediction model to estimate the chance of successful ITI. After future external validation, this clinical prediction model may be useful for informing clinicians and families.

Keywords: factor VIII; hemophilia A; immune tolerance; probability; treatment outcome.

Figure 1

Calibration plots. These calibration plots can be interpreted as predicted (x-axis) and observed (y-axis) probability of immune tolerance induction success. Uncorrected C statistic represents the calibration of the prediction model when applied to the data set. The corrected C statistic shows the calibration adjusted for overfitting using bootstrap resampling (250 runs). If the predicted probability were equal to the observed probability, the line would have followed the diagonal dotted line.

Figure 2

Nomograms. An illustration of the use of these nomograms. Ethnicity values: 1 = other ethnicities, 2 = Hispanic/Latino ethnicity. F8 mutation type: 1 = other mutation, 2 = large deletion. Model A. A person with a large deletion in the F8 gene (20 points, “points” scale on the upper bar), peak inhibitor titer before immune tolerance induction (ITI) start of 12.4 Bethesda Units (BU)/mL (95 points), Caucasian ethnicity (15 points), and 20 cumulative number of factor (F)VIII exposure days at inhibitor development (8 points) has a total score of (20 + 95 + 15 + 8 =) 138 points (second to lowest bar), corresponding to a predicted probability of 0.78 (lowest bar). Translating this to the clinic, this person has an estimated 78% chance of ITI success before ITI initiation. Model B. After initiating ITI, the peak inhibitor titer increased to 250 BU/mL. We reestimate the chance of ITI success at 6 months on ITI; a large deletion in the F8 gene (39 points, “points” scale on the upper bar), peak inhibitor titer ever measured until 6 months on ITI of 200.0 BU/mL (72 points), Caucasian ethnicity (22 points), and 20 cumulative number of FVIII exposure days at inhibitor development (20 points). The total score is (39 + 72 + 22 + 20 =) 153 points (second to lowest bar), corresponding to a predicted probability of 0.80 (lowest bar). Translating this to the clinic, this person has an estimated 80% chance of ITI success at 6 months on ITI. CED, cumulative number of FVIII exposure days.