Re‐assessing the effect of fetal hemoglobin on stroke in the Cooperative Study of Sickle Cell Disease

Fetal hemoglobin (HbF) is the main modifier of sickle cell disease (SCD) severity, with high HbF being associated with a lower risk of death, pain, and acute chest syndrome. The historical Cooperative Study of Sickle Cell Disease (CSSCD), a prospective study led in United States from 1978 to 1988, played an essential role in identifying or confirming the protective effect of HbF on SCD-related complications. However, whether HbF has a protective effect on stroke, a devastating complication of SCD, remains controversial.1, 2 Indeed, the CSSCD did not identify an association with overt stroke,3 nor did the subsequent ancillary study on silent infarct.4 In contrast, other studies have found that higher HbF level was associated with lower risk of overt stroke and silent infarct.1, 2 Several genetic studies have also found that HbF-associated variants influence stroke risk.2, 5 These discrepancies are difficult to reconcile and complicates our understanding of SCD pathophysiology and patient counseling. Moreover, hydroxyurea usage has been generalized but the optimal dosage is unclear, and whether HbF level is a proxy for stroke protection is unknown. Finally, given recent major advances in gene therapy approaches to increase HbF level, whether a protective effect on stroke is expected has major implications. Conclusions from the CSSCD are important as it remains the largest prospective SCD study to date and was conducted before the generalization of disease-modifying therapies (no patients were receiving hydroxyurea and only a few were enrolled in chronic transfusion program). The original CSSCD publication only reported the univariate association of HbF with incident ischemic stroke in SS patients (p = .106).3 Because this analysis was restricted to a subgroup of patients and a subtype of stroke (resulting in limited statistical power because of its small sample size [n = 31 cases]), its findings are difficult to generalize. Thus, we re-analyzed the CSSCD data by performing multivariate analyses that consider: (1) incident and prevalent strokes, (2) all subtypes of stroke (which is relevant given that magnetic resonance imaging was not performed in the CSSCD), and (3) all hemoglobin (Hb) genotypes. We analyzed the complete CSSCD database, with information from 4082 patients and a mean age at inclusion (±SD) of 13.6 ± 12.6 years (Supplemental Table 1). We retrieved HbF values for 2970 patients, including 2252 with the SS genotype. The initial report used a representative HbF value based on all HbF measures recorded for each patient. As this original value was unavailable to us, we used a mean HbF value that had been previously validated in clinical and genetic studies of the CSSCD.5, 6 We acknowledge that this HbF value might slightly differ from the initial analysis. We also note that our database included fewer patients with unknown Hb genotypes (101 vs. 139 in the initial article).3 A total of 57 patients (1% of the cohort) had received, or were suspected of having received, a transfusion. No HbF value was available for these patients, and they were not included in the present analyses. To confirm the validity of our dataset, we repeated several analyses previously done in the CSSCD. We used logistic (for binary variables) or quasi-Poisson (for continuous variables) regression adjusted for age and sex to analyze the effect of HbF on other SCD complications. We confirmed the protective effect of HbF on pain, acute chest syndrome, leg ulcer, priapism, and survival in SS patients (Supplemental Table 2). We retrieved the same prevalence of history of stroke at inclusion as the initial report (n = 153, of which 104 had HbF value, Supplemental Table 1). For these cases, the report form did not indicate the stroke subtype. We found a slightly higher number of patients with incident stroke (n = 94 vs. n = 87 in the original report, considering all stroke subtypes, all Hb genotypes, and only patients with no history of stroke at inclusion). Incident strokes were divided in the database in three subtypes: ischemic, hemorrhagic, and transient ischemic attack. As in the original article, our analyses of incident stroke only considered the 2866 patients with HbF value and without a history of stroke at inclusion. We used Cox proportional hazard regression to compute univariate hazard ratio (HR) for HbF. We also computed multivariate HR by adjusting for sex and age at entry ± Hb genotype. We used logistic regression to compute odds ratio (OR) for the association with history of stroke at inclusion using both univariate and multivariate models. We computed HR and OR for a 1% increase in HbF value. We used the surv_cutpoint function based on maximally selected rank statistics (R software, packages survminer and maxstat) to determine the optimal HbF cutpoint for the association with history of stroke at inclusion and we compared the two groups using the logrank test. First, we reproduced the published lack of association in univariate analysis between HbF and incident ischemic stroke in the SS subgroup (p = .06, Figure 1A and Supplemental Table 3).3 However, in multivariate Cox regression models, the effect of HbF on ischemic stroke was nominally significant when considering the SS subgroup (HR 0.90, 95% confidence interval [CI] 0.80–0.99, p = .038) or all Hb genotypes (HR 0.90, 95% confidence interval [CI] 0.81–0.99, p = .047). We did not identify an association between HbF level and incident hemorrhagic stroke, in either univariate or multivariate models. Considering all subtypes of incident stroke, we found that HbF level was associated with stroke in univariate and multivariate analysis in the SS subgroup (p = .016 and p = .021, respectively), as well as in multivariate analysis in the whole cohort (p = .018). Second, we analyzed the association between HbF and a history of stroke at inclusion. We found a strong protective association in the SS subgroup (OR 0.80, 95% CI 0.74–0.86, p = 2.7 × 10−8, adjusted for age and sex, Figure 1A and Supplemental Table 3) and in the whole CSSCD (OR 0.87, 95% CI 0.81–0.93, p = 5.6 × 10−5, adjusted for age, sex and Hb genotype). We found no association in the SC subgroup but only seven patients had a history of stroke. Given that hematological traits can be correlated, we aimed to assess if HbF had an independent effect on stroke. We fitted a multivariate model with other hematological traits in SS patients. We found that HbF level and white blood cell count were independently associated with stroke whereas hemoglobin level and platelet count were not (Supplemental Table 4). Third, we investigated what HbF threshold can best discriminate between the SS patients with or without a history of stroke at inclusion. We found that patients with a HbF ≤6.725% had almost a three-fold increased risk of stroke (HR 2.88, 95 CI 1.90–4.42, p = 1.2 × 10−6, Figure 1B). We also divided the SS patients in seven subgroups of the same size based on HbF quantiles to study the HbF-stroke dose–response relationship. We found an almost linear relationship between higher HbF level and lower stroke risk (Figure 1C). In conclusion, the expanded analyses of the CSSCD data reported here support the protective effect of HbF on stroke, consistent with other studies.1, 2, 5 This suggests that the beneficial effect of hydroxyurea on stroke is at least partly mediated by HbF and that obtaining higher HbF level should provide increased protection. It also implies that a protection against stroke should be expected from HbF-inducing drugs and gene therapies. Our dose–response analysis did not highlight a plateau (Figure 1C), suggesting that as HbF level increases, so does stroke protection. However, because the highest quantile of our analysis includes all HbF values greater than or equal to 11%, we cannot pronounce on the protection conferred by an increase above this threshold. The low number of hemorrhagic incident stroke limits the statistical power of this analysis. Furthermore, most of the patients had a SS genotype and our results apply mostly to this subgroup. We need larger studies to determine if HbF protects from stroke patients with other Hb genotypes, including SC. Thomas Pincez and Guillaume Lettre designed the study, participated in data interpretation, wrote, and revised the manuscript for critical content and approved the final manuscript. We thank all CSSCD participants for their contribution to this project, as well as the CSSCD investigators for developing and sharing this invaluable resource. We also thank Daniel E. Bauer for comments on an earlier version of this manuscript. T.P. is a recipient of a Charles Bruneau Foundation fellowship award and merit scholarship program for foreign students from the Ministry of Education and Higher Education of Quebec. This work was funded by the Canadian Institutes of Health Research (PJT #186159) and the Canada Research Chair Program (to G.L.). The CSSCD genetic dataset is available on the database of Genotypes and Phenotypes (dbGaP: https://www.ncbi.nlm.nih.gov/gap/), accession phs000366.v1.p1. This work was funded by the Canadian Institutes of Health Research (PJT #186159) and the Canada Research Chair Program (to Guillaume Lettre). The authors declare no competing interests. The CSSCD genetic dataset is available on the database of Genotypes and Phenotypes (dbGaP: https://www.ncbi.nlm. nih.gov/gap/), accession phs000366.v1.p1. Supplemental Table 1. Characteristics of sickle cell disease patients from the CSSCD Supplemental Table 2. Association of fetal hemoglobin level with other clinical complications in the SS patients Supplemental Table 3. Association of fetal hemoglobin (HbF) level with stroke in the CSSCD Supplemental Table 4. Multivariate analysis (using logistic regression) of hematological traits associated with a history of stroke at inclusion in SS patients Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.