Sickle cell disease (SCD) is a group of genetic syndromes characterized by deformation of red blood cells into sickle cells, hemolysis, and downstream vascular insults. SCD carries a high burden of neurological complications including stroke, silent cerebral infarct, and cognitive impairment (CI). Few studies have characterized the pattern of CI in adults and most have been conducted in highly selected populations.

A seminal study by Vichinsky et al.¹ showed that adults with homozygous SCD (HbSS disease or sickle cell anemia) had CI compared with healthy controls, predominantly in psychomotor speed. However, it is difficult to generalize findings from this study to real-world SCD populations that include patients with other genotypes or important clinical factors, such as stroke and treatment with chronic transfusions. Additional studies have found significant differences in information processing/psychomotor speed between adult SCD patients and controls but no differences in other cognitive processes.^{2, 3} We have also previously noted that stroke is a risk factor for slower psychomotor speed, but that CI occurs independent from stroke.⁴

Herein, we report findings on performance in various cognitive processes and domains in a broad cross-section of adult SCD patients with a range of comorbidities, and the effects of stroke history, chronic transfusions, and SCD variants on CI.

SCD patients were recruited from the University of Pittsburgh Medical Center Adult Sickle Cell Program outpatient clinic between 2016 and 2019. All patients with HbSS and HbS/β⁰ thalassemia (severe genotypes) or HbSC and HbS/β⁺ thalassemia (moderate genotypes) 18 years or older (N = 161) were informed about the study during a routine clinic visit while in steady-state (i.e., absence of acute complications). In total, 113 agreed to participate and met the following eligibility criteria: English-speaking, absence of other medical conditions associated with brain or neurocognitive dysfunction, and absence of contraindications to MRI. Healthy controls of Black race aged 18 years and older were recruited from the community and via SCD patient referral; a total of 69 were consented and eligible using similar criteria. Controls were frequency matched to patients based on age and gender.

An extensive neurocognitive test battery was administered to all participants. Ten variables were calculated from the battery, including: a total score and five domain scores from the Repeatable Battery for the Assessment of Neuropsychological Status (RBANS), a total recall t-score and a delayed recall t-score from the Hopkins Verbal Learning Test-Revised (HVLT-R), a scaled score from the Digit Symbol Substitution Test (DSST), and an executive function domain score as a mean of scaled scores from select subsets of the Delis–Kaplan Executive Function System (D-KEFS). See Appendix S1 for details.

In total, 152 participants (86 SCD patients and 66 controls) completed cognitive testing and were included in the analysis. The overall mean age of the cohort was 36.2 years (SD = 12.1), 57% were female, and average length of education was 13.7 years (SD = 2.1). SCD patients and controls had a similar distribution of age and gender, and the SCD participants were representative of the patient population at our institution in terms of relevant demographic and clinical parameters (Table S1).

Across all 10 neurocognitive scores examined, patients had lower average scores than controls (Table S2). Over 50% of the patients scored below the cognitively “normal” range (>1 SD) on the HVLT-R Total Recall Score, RBANS Immediate Memory Index Score, and Language Index Score, and over 60% scored below the cognitively “normal” range on the Visuo-construction index, Attention Index, and Total Index Scores. These percentages are consistent with previous findings.¹

Unadjusted linear regression analysis was conducted for each cognitive score and a multivariable-adjusted linear regression model was created for scores for which there was a significant difference by SCD status, such that β represents the estimated difference in the cognitive outcome score between the designated groups. SCD patients had lower scores on the DSST Scaled Score and Executive Function Domain Score (both P < .001), and in domains assessed by the RBANS, patients had significantly lower scores on the Language Index (P = .014), Attention Index (P < .001), and Total Index Scores (P < .001) (Table S3, Figure S1). The DSST Scaled Score (β = −1.52), Attention Index (β = −10.23), and RBANS Total Index Score (β = −6.85) were significantly lower among patients compared with controls in the multivariable logistic regression adjusting for premorbid verbal IQ, hemoglobin level, and smoking history.

These findings demonstrate the pervasiveness of impairments across heterogeneous areas of cognitive function, independent of genotype. Lower DSST Scaled Score among patients reflects worse performance in attention and information processing/psychomotor speed, while lower RBANS Total Index Score among patients indicates an average of diminished abilities across immediate memory, delayed memory, language, visuo-construction, and attention.

Separate multivariable linear regression models were created to evaluate the additive impact of stroke, chronic transfusion, and genotype among patients for each cognitive score for which significant adjusted differences were found between patients and controls. See Tables S4–S6 for SCD patient characteristics by stroke history, transfusion exposure, and SCD severity, respectively.

In fully adjusted models using controls as the reference group, patients with a history of stroke had significantly lower DSST Scaled Score (β = −3.61), Executive Function Domain Score (β = −1.93), Attention Index Score (β = −18.00), and Total Index Score (β = −14.27), whereas patients without a history of stroke had a significantly lower Attention Index Score (β = −8.94) and Total Index Score (β = −5.63) (Table 1). Also relative to controls, patients who were not receiving transfusions had significantly lower DSST Scaled Score (β = −1.52), Attention Index Score (β = −10.27), and Total Index Score (β = −7.09). The differences for Attention Index Score (β = −9.89) and Total Index Score (β = −4.89) were reduced between patients receiving transfusions and controls, as compared with the differences between patients not receiving transfusions and controls, but remained similar for the DSST Scaled Score (β = −1.54) (Table 1).

Both moderate and severe disease genotypes had lower Attention Index Score (β = −9.93 and β = −11.12, respectively) compared with controls (Table 1). Additionally, the moderate disease group had lower Total Index Score compared with controls (β = −7.42), while the severe disease group had lower DSST Scaled Score relative to controls (β = −2.05). Though we observed no differences between patients with severe and moderate genotypes, both subgroups showed deficits in various cognitive functions relative to controls, indicating the general pathophysiology of SCD may contribute to CI.

Among patients, those with a history of stroke had significantly worse DSST Scaled Score (P = .015), Executive Function Domain Score (P = .012), and Total Index Score (P = .020) than patients without a history of stroke (Table 1). There were no significant differences in cognitive scores between patients receiving and those not receiving transfusions.

The current findings that SCD patients with stroke have CI compared with those without stroke extend previous observations among children⁵ to adults with SCD. We observed that both patients with and without stroke have worse Attention Index Score and Total Index Score compared with controls. As such, our results suggest that other factors may contribute to CI in patients without a history of stroke relative to healthy individuals.

We also show that not receiving chronic transfusions was associated with poorer outcomes on DSST Scaled Score, Attention Index Score, and Total Index Score compared with controls. Chronic blood transfusions may alleviate some of these impairments. However, few patients in our study were receiving chronic transfusions, and hence this finding should be confirmed in studies with a larger sample size. If confirmed, our data suggest that transfusions are protective against CI.

Similar to results from Vichinsky et al.¹ among patients with homozygous SCD, we found that patients with other genotypes performed significantly worse than controls in executive function and attention, as well as in global cognition, as represented by the RBANS Total Index Score. Crawford et al.² also found that SCD patients with severe genotypes have significantly worse information processing speed. However, the authors adjusted for information processing speed among SCD patients in further analyses, which explained the significant deficits in memory, perceptual motor speed, and executive function compared with controls. Though we do not account for information processing speed in SCD patients, we demonstrate lower performance across 10 different measures of diverse cognitive domains and processes in a population of SCD patients with both severe and moderate disease.

A limitation of our study is the relatively small number of confounders we included in the analysis. Arguably, other variables could also impact cognitive function and would be worthy of inclusion in an analysis conducted on a larger sample size. For instance, maternal education impacts cognitive function in children⁶ and may have long lasting effects throughout adulthood. Ultimately, longitudinal studies will be important to better ascertain the role of other factors.

This comprehensive assessment of cognitive function in a representative SCD patient population, including those with multiple genotypes and comorbidities, revealed that patients perform worse across several cognitive processes and domains relative to controls, even after adjusting for known risk factors. Stroke remains an important contributor to CI among SCD patients, but SCD appears to affect cognitive function in the absence of stroke. Understanding the pervasiveness of CI among SCD patients can lead to identifying targeted interventions; the findings from this study can improve clinical care for SCD patients and serve as a basis for measuring the impact of treatments such as chronic transfusions on various domains of cognitive performance.

镰状细胞病（SCD）是一组以红细胞变形为镰状细胞、溶血和下游血管损伤为特征的遗传综合征。SCD 会带来很高的神经系统并发症负担，包括中风、无症状脑梗塞和认知障碍 (CI)。很少有研究描述成人 CI 的模式，而且大多数研究都是在经过精心挑选的人群中进行的。

Vichinsky 等人的一项开创性研究。^{图 1}显示，与健康对照相比，患有纯合性 SCD（HbSS 疾病或镰状细胞性贫血）的成年人存在 CI，主要表现在精神运动速度方面。然而，很难将这项研究的结果推广到现实世界的 SCD 人群，其中包括具有其他基因型或重要临床因素（例如中风和慢性输血治疗）的患者。其他研究发现，成年 SCD 患者和对照组之间的信息处理/精神运动速度存在显着差异，但其他认知过程没有差异。^{2, 3}我们之前也注意到，中风是精神运动速度减慢的一个危险因素，但 CI 的发生与中风无关。⁴

在此，我们报告了患有一系列合并症的成年 SCD 患者的各种认知过程和领域表现的研究结果，以及中风病史、慢性输血和 SCD 变异对 CI 的影响。

SCD 患者于 2016 年至 2019 年间从匹兹堡大学医学中心成人镰状细胞计划门诊招募。所有患者患有 HbSS 和 HbS/β 0 地中海^贫血（重度基因型）或 HbSC 和 HbS/β ⁺地中海贫血（中度基因型）18 年或以上（N  = 161）在稳定状态（即没有急性并发症）的常规门诊就诊期间被告知该研究。总共有 113 人同意参加并符合以下资格标准：会说英语、不存在与大脑或神经认知功能障碍相关的其他医疗状况以及不存在 MRI 禁忌症。通过 SCD 患者转诊从社区招募 18 岁及以上的黑人种族健康对照者；使用类似标准，共有 69 项获得同意并符合资格。对照根据年龄和性别与患者进行频率匹配。

对所有参与者进行了广泛的神经认知测试。从电池中计算出十个变量，包括：神经心理状态评估可重复电池(RBANS)的总分和五个领域分数、霍普金斯语言学习测试的总回忆 t 分数和延迟回忆t分数-修订版(HVLT-R)，数字符号替换测试(DSST) 的标度分数，以及作为Delis-Kaplan 执行功能系统(D-KEFS) 选定子集的标度分数平均值的执行功能域分数。详细信息请参见附录 S1。

总共 152 名参与者（86 名 SCD 患者和 66 名对照者）完成了认知测试并被纳入分析。该队列的总体平均年龄为 36.2 岁（SD = 12.1），其中 57% 为女性，平均受教育年限为 13.7 年（SD = 2.1）。SCD 患者和对照组的年龄和性别分布相似，并且 SCD 参与者在相关人口统计学和临床​​参数方面代表了我们机构的患者群体（表 S1）。

在检查的所有 10 项神经认知评分中，患者的平均评分低于对照组（表 S2）。超过 50% 的患者在 HVLT-R 总回忆评分、RBANS 即时记忆指数评分和语言指数评分上的得分低于认知“正常”范围 (>1 SD)，超过 60% 的患者得分低于认知“正常”范围视觉构造指数、注意力指数和总指数得分的范围。这些百分比与之前的发现一致。¹

对每个认知评分进行未经调整的线性回归分析，并针对 SCD 状态存在显着差异的评分创建多变量调整线性回归模型，使得 β 代表指定组之间认知结果评分的估计差异。SCD 患者的 DSST 量表评分和执行功能领域评分均较低（均P  < .001），并且在 RBANS 评估的领域中，患者的语言指数 ( P  = .014)、注意力指数 ( P  < .001) 和总指数得分 ( P  < .001)（表 S3，图 S1）。 在针对病前言语智商进行调整的多变量逻辑回归中，与对照组相比，患者的DSST 量表评分 ( β  = -1.52)、注意力指数 ( β  = -10.23) 和 RBANS 总指数评分 ( β = -6.85) 显着较低，血红蛋白水平和吸烟史。

这些发现表明，认知功能异质领域的损伤普遍存在，与基因型无关。患者的 DSST 量表得分较低反映了注意力和信息处理/精神运动速度的表现较差，而患者的 RBANS 总指数得分较低则表明即时记忆、延迟记忆、语言、视觉建构和注意力的平均能力下降。

创建单独的多变量线性回归模型来评估患者中中风、慢性输血和基因型对每个认知评分的附加影响，在患者和对照组之间发现显着的调整差异。请参阅表 S4-S6，了解按卒中史、输血暴露和 SCD 严重程度划分的 SCD 患者特征。

在使用对照组作为参考组的完全调整模型中，有中风病史的患者的 DSST 量表评分 ( β  = -3.61)、执行功能域评分 ( β  = -1.93)、注意力指数评分 ( β  = -18.00)显着较低，和总指数得分（β = -14.27），而没有中风病史的患者的注意力指数得分（ β  = -8.94）和总指数得分（β  = -5.63） 显着较低（表1）。同样相对于对照组，未接受输血的患者的 DSST 量表评分 ( β  = -1.52)、注意力指数评分 ( β  = -10.27) 和总指数评分 ( β  = -7.09)显着较低。与未接受输血的患者和对照组之间的差异 相比，接受输血的患者和对照组之间的注意力指数评分 ( β  = -9.89) 和总指数评分 ( β = -4.89) 的差异有所减小，但 DSST 的差异仍然相似等级分数 ( β  = -1.54)（表 1）。

 与对照组相比，中度和重度疾病基因型的注意力指数得分较低（分别为β  = -9.93 和β = -11.12）（表 1）。此外，与对照组相比，中度疾病组的总指数得分较低（β  = -7.42），而与对照组相比，重度疾病组的 DSST 量表得分较低（β  = -2.05）。尽管我们观察到重度和中度基因型患者之间没有差异，但与对照组相比，两个亚组均表现出各种认知功能缺陷，表明 SCD 的一般病理生理学可能导致 CI​​。

在患者中，有中风病史的患者的 DSST 量表评分 ( P  = .015)、执行功能域评分 ( P  = .012) 和总指数评分 ( P  = .020) 明显低于无中风病史的患者（表格1）。接受输血的患者和未接受输血的患者之间的认知评分没有显着差异。

目前的研究结果表明，与没有中风的 SCD 患者相比，患有中风的 SCD 患者患有 CI，这一发现将之前对^{5 岁}儿童的观察结果延伸到了患有 SCD 的成人。我们观察到，与对照组相比，患有和未患有中风的患者的注意力指数得分和总指数得分均较差。因此，我们的结果表明，相对于健康个体，其他因素可能会导致无中风病史的患者发生 CI。

我们还表明，与对照组相比，不接受长期输血与 DSST 量表评分、注意力指数评分和总指数评分结果较差相关。长期输血可以减轻其中一些损伤。然而，我们研究中很少有患者接受长期输血，因此这一发现应在更大样本量的研究中得到证实。如果得到证实，我们的数据表明输血可以预防 CI。

与 Vichinsky 等人的结果类似。¹在纯合子 SCD 患者中，我们发现其他基因型患者的执行功能和注意力以及整体认知（以 RBANS 总指数评分为代表）的表现明显差于对照组。克劳福德等人。²还发现基因型严重的SCD患者信息处理速度明显较差。然而，作者在进一步分析中调整了 SCD 患者的信息处理速度，这解释了与对照组相比，记忆力、知觉运动速度和执行功能的显着缺陷。尽管我们没有考虑 SCD 患者的信息处理速度，但我们在患有严重和中度疾病的 SCD 患者群体中对不同认知领域和过程的 10 种不同测量方法中表现出较低的性能。

我们研究的一个局限性是我们在分析中包含的混杂因素数量相对较少。可以说，其他变量也可能影响认知功能，并且值得纳入对更大样本量进行的分析中。例如，母亲教育会影响儿童的认知功能⁶，并可能对整个成年期产生长期持续的影响。最终，纵向研究对于更好地确定其他因素的作用非常重要。

对代表性 SCD 患者群体（包括具有多种基因型和合并症的患者）的认知功能进行的全面评估表明，即使在调整了已知的危险因素后，患者在多个认知过程和领域的表现也比对照组更差。卒中仍然是 SCD 患者 CI 的一个重要因素，但在没有卒中的情况下，SCD 似乎会影响认知功能。了解 SCD 患者中 CI 的普遍性有助于确定有针对性的干预措施；这项研究的结果可以改善 SCD 患者的临床护理，并作为衡量慢性输血等治疗对认知表现各个领域的影响的基础。