Abstract
Background: Hydroxyurea (HU) is recommended to be offered to children with sickle cell anemia (SCA; HbSS and HbSβ0 thalassemia) beginning at 9 months of age (Yawn, JAMA 2014). Intensifying HU to a maximum tolerated dose (MTD) provides fetal hemoglobin (HbF) levels of >20% (Ware, Blood 2010). Yet, children treated with HU at MTD still experience hospitalizations due to SCA-related complications (Nottage, PLoS One 2013). The ability to identify children at high risk of experiencing SCA-related complications while on HU would enhance clinical management and facilitate performance of clinical trials of novel agents. Our objective was to identify risk factors associated with hospitalization and to develop a prediction model for hospitalization of children treated with HU at MTD.
Methods: The Hydroxyurea Study of Long-Term Effects (HUSTLE) is a prospective observational study (NCT00305175) designed to describe the clinical effects of HU in children with SCA. We have analyzed children who initiated HU per protocol and whose dose was escalated to MTD. Laboratory and clinical data were abstracted at baseline and at 9-12 months after enrollment. The first laboratory values of each interval were considered representative of the period. The primary outcome measure was hospitalization. After a multivariate regression logistic model identified risk factors for hospitalization based on a stepwise selection strategy, a smoothed receiver-operating-characteristic (ROC) curve was generated to predict hospitalization for the composite model and for each individual risk factor using the method of maximum-likelihood fitting of univariate distriubutions. Atlhough HbF was not identified in regression analysis among risk factors, because of its known effects as a disease modifier, it was inserted into the composite model. A prediction model was then generated for the identified risk factors utilizing optimal cutoff values defined by the maximum Youden method.
Results: 151 children (mean [SD] age, 8.3 [5.0] years) were analyzed. HU resulted in higher HbF levels (22.6% [9.1] versus 10.1% [7.1]; p<0.0001), lower absolute neutrophil counts (4000/mm3 [2600] versus 7000/mm3 [3800]; p<0.0001), and fewer hospitalizations (7.8% [10/128] versus 28.9% [39/128]; p<0.0001) after 9-12 months compared to baseline.
Children hospitalized after 9-12 months of therapy had similar HbF levels compared to those not hospitalized (OR 0.98 [95% CI: 0.9-1.1]; p=0.5), but higher total bilirubin (BiliT) (OR 1.3 [95% CI: 1.0-1.6]; p=0.03), blood urea nitrogen (BUN) (OR 1.3 [95% CI: 1.1-1.7]; p=0.19), and lactate dehydrogenase (LDH) (OR 1.01 [95% CI: 1.002-1.010]; p=0.006) levels. A ROC curve including insertion of HbF showed an area under the curve (AUC) of 0.90 (Figure 1). Optimal cutoffs for risk factors were: BiliT 1.05 mg/dL, BUN 8.5 mg/dL, and LDH 473 units/L. The probability of hospitalization varied in relation to identified risk factors (Table 1).
Table 1. Predicted probability of hospitalization after 9-12 months of HU at MTD # of risk factors BiliT HbF BUN LDH Probability of ≥1.05(mg/dL) ≤20 (%) ≥8.5(mg/dL) ≥473(units/L) Hospitalization (%) 0 No No No No <1 1 Yes No No No <1 No Yes No No 1.3 No No Yes No 5.8 No No No Yes 19 2 Yes Yes No No <1 Yes No Yes No <1 Yes No No Yes 2.2 No Yes Yes No 10 No Yes No Yes 29 No No Yes Yes 65 3 Yes Yes Yes No 1.1 Yes Yes No Yes 3.8 Yes No Yes Yes 15 No Yes Yes Yes 77 4 Yes Yes Yes Yes 25 LDH, lactate dehydrogenase; BiliT, total bilirubin; BUN, blood urea nitrogen
Discussion: In this pediatric cohort, HU therapy at MTD produced a robust laboratory response with most children having HbF >20% after 9-12 months. At such high levels of HbF, few children experienced a hospitalization due to its protective effects; however, we identified three laboratory parameters (LDH, BiliT, and BUN) that were predictive risk factors for children who continued to be hospitalized despite high levels of HbF. After 9-12 months of HU therapy at MTD the probability of hospitalization ranged from <1% to 65% for children with HbF of less than <20% and <1% to 77% in children with HbF of >20%. This model may help identify children at risk for complications of SCA while being treated with HU at MTD.
Figure 1. Receiver-operating-characteristics curve for the risk of hospitalization as predicted by the composite model and for each identified risk factor. Area under the curve: composite model, 0.90; LDH, 0.74; BUN, 0.65; BiliT, 0.59; HbF, 0.53 Figure 1. Receiver-operating-characteristics curve for the risk of hospitalization as predicted by the composite model and for each identified risk factor. Area under the curve: composite model, 0.90; LDH, 0.74; BUN, 0.65; BiliT, 0.59; HbF, 0.53
Disclosures
Estepp: Eli Lilly and Company: Research Funding; Daiichi Sankyo: Research Funding. Off Label Use: Off label use of hydroxyurea in children with sickle cell anemia.
Determination of Urinary Neopterin/Creatinine Ratio to Distinguish Active Tuberculosis from Latent Mycobacterium tuberculosis Infection
