Large Scale Analysis of the Real-World Association between Fetal Hemoglobin and Vaso-Occlusive Crises in Sickle Cell Disease

Background Fetal hemoglobin (HbF) attenuates the rate of vaso-occlusive crises (VOCs) in sickle cell disease (SCD), but exact quantification is lacking. Methods We analyzed two large historic datasets with contemporary statistical methods. Data, acquired through BioLINCC, were the Cooperative Study of SCD (CSSCD; USA, 1979-1988) and the Multicenter Study on Hydroxyurea (MSH; USA and Canada, 1992-1995). The CSSCD had 3 planned HbF assessments: baseline, 1st annual visit, 2nd annual visit. We included those aged ≥12 years with at least one HbF measurement at baseline, 1st, or 2nd annual visit (1395 unique individuals, median age 22 years). Follow-up was up to 8 years post baseline. We analyzed the data using one of three approaches: baseline HbF and all follow-up; three HbF readings and one year of follow-up after each visit; three HbF readings and all follow-up up to next visit or follow-up end. For the MSH (HbSS genotype, age ≥18 years, ≥3 VOCs within the pre-trial year, 299 unique individuals, median age 29 years), we used HbF at randomization in the placebo arm and after titration in the hydroxyurea (HU) arm. Placebo participants were offered HU after early trial termination and their follow-up was censored then. HbF was analyzed as a percentage of total hemoglobin. In the MSH data, F-cells (as a percentage of red blood cells [RBC]) were additionally analyzed. F-cells were averaged over follow-up. We used directed acyclic graphs for confounder identification and negative binomial generalized additive models to allow the data to suggest the functional relationship between HbF and the VOC rate. A straight line suggests a linear effect of HbF% on log VOC rate and implies that each percentage point (%pt) increase in HbF or F-cells reduces the VOC rate by a fixed percentage. We regressed VOC rate during follow-up on HbF and adjusted models for age at baseline, sex, ethnicity, and when applicable for β-globin haplotype, assigned treatment (HU/placebo), and received HU dose. Results HbF showed a linear protective effect on log VOC rate in two of the CSSCD analyses and a non-linear progressive effect for the remaining analysis: each additional %pt increase in HbF reduced the VOC rate by a slightly larger percentage. When fitting linear models, each %pt increase in HbF was associated with a 4% (95% CI: 2-6%) to 6% (95% CI: 3-8%) reduction in VOC rate. For MSH data, each %pt increase in HbF was associated with an 8% (95% CI: 4-11%) reduction in VOC rate. The model suggested that the reduction per %pt increase in HbF depended on the level of HbF%, with a larger reduction in VOC rate the higher the HbF%. We found no evidence of change in VOC rate up to 35% F-cells. Above 35% F-cells, there was increasing protection against VOCs. Individuals with ≥70% F-cells had less than half the VOC rate of those with 35% F-cells. Results did not change when the models were adjusted for total hemoglobin. Interpretation The results show that each %pt increase in HbF has a protective effect on VOC rate, which was larger for higher HbF% in two analyses. Since HbF intercalates in HbS polymerization, VOC rate may be driven by HbF/HbS ratio. Hence, increasing HbF% would be progressively more effective through the numerator effect in HbS (like odds). These data further highlight the importance of HbF distribution across RBC, as VOCs occurred at a progressively lower rate for F-cells >35%. HbF content varies also among RBC that fail the F-cell threshold (6pg), and a small number of sickle RBC may suffice to cause a VOC. Translation of these data from its geographic reach to a current, global setting may be limited. However, data from a recent small study on HU in India suggested an upper limit of 9% VOC reduction for every %pt increase in HbF, if all protective effect of HU comes from HbF induction. This corresponds well with the 4-8% reduction in VOC rate found here. The CSSCD remains the largest real-world study on SCD, which contributes pre-treatment data, and the MSH yielded comparable results. Conclusion This large real-world study on prospectively collected data confirms the direct relationship between HbF and VOC occurrence. Further, our results suggest that one %pt increase in HbF% may be more beneficial for patients with a higher starting HbF%. These results highlight the importance of HbF distribution across the RBC population with a focus on HbF as an important SCD therapy target. Repeating the study on data of greater geographic reach would strengthen the interpretation.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.