Intra-Patient Variability in Fetal Hemoglobin Measurements over Time in Sickle Cell Disease Patients Not on Fetal Hemoglobin Inducing Agents

Introduction:

We previously reported results of the placebo-controlled phase II study of the short-chain fatty acid derivative 2,2-dimethylbutyrate in inducing fetal hemoglobin (Hb F) in 76 patients with sickle cell disease (SCD). The primary endpoint was a comparison of Hb F levels in the treatment versus the placebo arms. Week 24 interim analyses revealed no statistically significant difference in change in Hb F levels between the 2 groups. We examined the placebo arm in order to assess untreated, intra-patient variability of Hb F%.

Methods:

Only Hb F values performed by HPLC at the central reference laboratory (Georgia Health Science University) were included. Any Hb F values determined within a 3 month period after a blood transfusion were excluded. Baseline and at least one subsequent Hb F levels were available in 37 of the 38 patients randomized to the placebo arm, which were included in the analysis. Serial determinations were performed at 4 week intervals, resulting in a total of 348 determinations and a median of 9 values per patient (range 2 - 15). Mean and standard deviation of repeat measures of Hb F% in each individual patient were calculated. A mean ± SD of the individual coefficients of variation (SD/mean of repeat measures for each individual) was calculated. Peak-to-trough (maximum - minimum) ranges of repeated measures were also calculated per individual and quartiles for the group determined. To assess the potential for regression to the mean from baseline, we calculated the median peak-to-trough within quartiles of baseline Hb F. We assessed the difference between maximum and minimum values as a percentage of the maximum or percentage deviation from personal best Hb F%, and finally we examined the association of baseline Hb F, age, gender and race with the degree of variability.

Results:

Patient median (range) age was 25.9 years (12 – 46), 24 (63%) were female, the genotype was Hb SS in 30 patients (79%) and Hb S/β⁰ thalassemia in 8 (21%), and no patient was treated with hydroxycarbamide at enrollment. Median (range) values of Hb F percentage at baseline were 7.5% (0.5 – 23.4). The mean coefficient of variation of all Hb F values was 13%, with a peak to trough median (range) of 1.8 (0.1 - 9.8). The mean percentage Hb F % variability over time was 31.8% (S.D +/- 18) and median (range) value of 26.1 (7.2-80) with quartiles depicted in Table 1. Baseline Hb F% negatively correlated with the percentage variability and this association was highly statistically significant (Spearman rho –0.34, p = 0.04).

Discussion:

There is substantial Hb F variability among patients not on any Hb F inducer. This variability has not been previously reported and influences both the standard deviation and standard error, and would likely lower the statistical power of any comparative analysis in a clinical trial. Thus, sample sizes should be larger to be able to detect a sizeable difference between experimental and control groups.

Variability among treated patients is harder to assess because it is difficult to separate out intrinsic variability from treatment effects and medication adherence. We have not attempted to assess the intra-patient variability in Hb F among patients given study drug in this study. However, it seems reasonable to assume that the variability seen in untreated patients would also be seen, to some degree, within treated patients. Thus, using Hb F percentage alone as a measure of medication compliance (with a Hb F inducing agent) may not be reliable.

Possible explanations for this variability over time include artifacts of laboratory technique, increased hemolysis with increased erythropoietic drive, worsening renal function (and subsequent decrease in erythropoietin production), and bone marrow infarction, leading to disrupted hematopoiesis.