Basic demographic information was obtained at the start of each study (i.e., age, sex, and postcode). The postcode-based Index of Multiple Deprivation was used as an indication of socioeconomic status (SES), as previously described (41). It consists of seven domains: income, employment, education, health, crime, barriers to housing and service, and living environment (42). The data are open-source, and downloadable from the UK Ministry of Housing, Communities & Local Government.¹ The index ranges from rank 1st (most deprived) to 32,844th (least deprived).

General cognitive abilities were assessed using the Wechsler Intelligence Scale for Children (WISC-IV; SAC patients and CYA-TD < 16 years) (11), the Wechsler Adult Intelligence Scale (WAIS-IV; SAC patients and CYA-TD > 16 years) (12) or were estimated using the Wechsler Abbreviated Scale of Intelligence (WASI-II Vocabulary and Matrix Reasoning; POMS patients) (43). Scaled and raw scores for all indices and subtests were calculated to compare age-related changes. Scaled scores were derived from raw scores for each subtest, enabling comparison to the normative group based on the age range. Scaled scores from various subtests were combined together, resulting in indices for various domains of cognitive development: VCI (Verbal Comprehension: Similarities, Vocabulary, Information/Comprehension), PRI (Perceptual Reasoning: Block Design, Matrix Reasoning/Picture Concepts, Visual Puzzles), WMI (Working Memory: Digit Span, Arithmetic/Letter-Number-Sequencing), PSI (Processing Speed: Symbol Search, Coding/Cancellation) and FSIQ (Full-Scale IQ: sum of all raw scores). Raw scores have a mean of 10 and a standard deviation of 3. Scaled scores have a mean of 100 and a standard deviation of 15.

EF was assessed using the examiner-administered Tower subtest from the D-KEFS (44). Tower time was the outcome investigated, which has a mean of 10 and a standard deviation of 3. In addition, for a patient-centric measure, the caregiver-reported Behaviour Rating Inventory of Executive Function (BRIEF) (45) or adult self-reported BRIEF (BRIEF-A) were administered. The Global Executive Composite (GEC), Behaviour Regulation Index (BRI) and Metacognition Index (MI) and all subscale scores were investigated (46). The GEC is a summary index score of the BRI and MI. The BRI contains four different subscales: Inhibit, Shift, Emotional Control and Self-Monitor. The MI contains five subscales: Initiate, Working Memory, Plan/Organize, Task Monitor, and Organization of Materials. The caregiver-reported BRIEF does not contain the subscales of the BRI Self-Monitor and subscale for the MI Task Monitor. It also has an additional subscale the MI Monitor. T scores were used to interpret the BRIEF indices (normal < 60; mildly elevated: 60–64; potentially clinically elevated 65–69; clinically elevated above 70). Scaled and raw scores for all indices and subtests were calculated to compare developmental trajectories.

Participants completed their cognitive assessment on a single day in a quiet and comfortable room to minimise distractions. On the day of cognitive assessment, a fingertip pulse oximeter was used to measure daytime SpO₂ (%). Haemoglobin (Hb in g/L) and Haematocrit (%) were obtained from the most recent laboratory test, partial pressure of oxygen (PaO₂) was assumed to be 100 mmHg and arterial oxygen content (CaO₂ in mL/dL) was calculated as

Trajectories can only be compared if there is a significant linear relationship between mental age and the cognitive variable in the CYA-TD group (32). To explore the importance of any differences in socioeconomic status across different cohorts, all models were corrected for SES. However, SES did not significantly influence the scores on cognitive tests. Hence, it was excluded from the ANCOVA analyses. For the analysis, mental age was calculated as (FSIQ score/100) * chronological age. Hence, the first step of our analysis was to generate simple linear regression models with each scaled cognitive variable as the outcome and mental age as the predictor for the CYA-TD group and the CYA-SCD group separately. Additionally, for raw scores, separate regression lines were calculated for WISC and WAIS data as the total raw scores for each subtest differ (See Supplementary Tables).

For our next step, we adjusted the age such that zero years represented the lowest age at which we collected data, i.e., 8 years. To do this, we deducted the lowest age from each participant’s mental age (30). Following this adjustment for age, we generated ANCOVA models that compared the developmental trajectory for cognitive scores only for those who showed a significant linear relationship with mental age in the CYA-TD group. The developmental trajectory analyses using ANCOVA resulted in two metrics: (1) if there was a delay at onset, i.e., if the two groups differed in performance at the youngest age at which the measurement is taken (also known as overall effect of the group), and (2) if the rate of development differed between the two groups which is the interaction between group performance and mental age (group × MA). It was also possible to calculate the difference between the two scores at onset by subtracting the intercept of the CYA-SCD group from the intercept of the CYA-TD group (delay at onset = intercept of CYA-TD group − intercept of CYA-SCD group). Similarly, difference in the rate of development is calculated by dividing the gradient of CYA-SCD group by that of the CYA-TD group (rate of development = gradient of CYA-SCD group/gradient of CYA-TD group).

For cognitive domains where the CYA-TD group showed a significant relationship between scores and age, but the CYA-SCD group did not, we conducted a rotation analysis (X-Y coordinates) to distinguish between a zero trajectory and no systematic relationship in the CYA-SCD group, since results can be influenced by other factors such as cross-sectional design artifacts, missing disease severity data, and variability and static development in a particular cognitive domain [see detailed discussion in Thomas et al. (32)]. This allows distinction in the CYA-SCD group between plateauing (zero trajectory) or whether, alternatively, there is no systematic relationship, suggesting that there might be other variables that predict the relationship better.

Using the same approach, we also examined if developmental trajectories differed between CYA-SCD with and without SCI.

Cross-sectional developmental trajectory analyses explored cognitive profiles in CYA-SCD in comparison to CYA-TD. SES did not significantly influence any of the cognitive scores and was excluded from further analyses. All simple linear regression results examining the relationships between mental age and performance on cognitive scores are shown in Supplementary Table S2. To compare the groups in the later steps of the analysis, the pre-analysis (simple linear regression) required a statistically significant result for the CYA-TD group (Supplementary Table S2) which was the case for VCI, the VCI Similarities and Vocabulary subtests, PRI, the PRI Block Design subtest and WMI as well as the PSI Coding subtest. For the measures of EF the BRIEF GEC was close to being significantly related to mental age. The BRIEF MI and its subscales Initiate, Working Memory, and Plan/Organize as well as the BRIEF BRI and its subscale Inhibit, all showed a significant relationship with mental age (see Supplementary Table S2).

As our second step, we examined developmental trajectories between cognition and mental age for CYA-SCD and CYA-TD. ANCOVA models were constructed to evaluate the overall effect of the group in terms of delay at onset and the Group × Mental Age (MA) interaction effect to determine the differences in the rate of cognitive development. Significant results from this analysis are summarised in Table 2 with regressions are plotted in Figures 2–7.

For cognitive domains where there was a significant relationship between mental age and cognitive scores only for CYA-TD (Supplementary Table S2), but not for the CYA-SCD, an additional rotation analysis or null trajectory analysis was conducted (Supplementary Table S3) to distinguish between Zero trajectory and no systematic relationship. While there were significant relationships between mental age and cognition for the CYA-TD group none were found for CYA-SCD for PRI Block Design and PSI Coding, BRIEF GEC, BRI and its subscale Inhibition, BRIEF MI and its subscales Initiate, Working Memory and Plan/Organize as well as Tower Time score (Supplementary Table S3).

Similar analyses were conducted by dividing the participants in three groups: CYA-SCD and silent cerebral infarction (CYA-SCD-SCI+), CYA-SCD but no SCI (CYA-SCD-SCI−) and CYA-TD. Simple linear regression analyses examining the relationships between mental age and performance on cognition are shown in Supplementary Table S4. As the statistically significant results for the CYA-TD group were for the same domains, we conducted the analyses comparing the CYA-SCD-SCI+ with the CYA-TD group.

For CYA-SCD we found that haemoglobin significantly predicted the variance in VCI (10%, p = 0.021), VCI Similarities (8%, p = 0.037) and Vocabulary (4% p = 0.041), PSI (4%, p = 0.04) and BRIEF Working Memory subscale (8.2%, p = 0.036). Greater variance was explained by haemoglobin for PRI in the CYA-SCD-SCI+ (29%, p = 0.013).

For CYA-SCD CaO₂ significantly predicted the variance in VCI (12%, p = 0.01) and its subscales Similarities (10%, p = 0.022) and Vocabulary (4%, p = 0.038), PSI (4%, p = 0.033) and its subscale Symbol Search (4%, p = 0.032) and BRIEF Working Memory (7.2%, p = 0.049). Greater variance was explained by CaO₂ for VCI (22%, p = 0.004) and PRI in the CYA-SCD-SCI+ (37%, p = 0.005).

For CYA-SCD SpO₂ only significantly predicted the variance in VCI (9%, p = 0.027).

For CYA-SCD, SES significantly predicted the variance in BRIEF MI subscale Organize Material (9%, p = 0.017). Greater variance was explained by BRIEF BRI Inhibit in the CYA-SCD-SCI+ (22%, p = 0.058), although near-significant.

The main findings of our paper indicate that CYA-SCD tend to have a delay in verbal comprehension skills, but they also appear to have a catch-up period and are no longer delayed as young adults. However, this initial delay could impact on later successful reading skills (50) and successful academic performance in school-aged children (53). Additionally, research in healthy infants between the ages of 7–12 months has shown a significant correlation between language development (Mullen Scales of Early Learning) and grey and white matter volume (e.g., subcortical region hippocampus) (54). Specifically, for CYA-SCD, Steen et al. (55) demonstrated that grey matter volume growth is delayed in children living with SCD aged 3.9–18.5 years. Another factor to consider is that grey matter volume reduces between 2 to 9 years of age in parallel with white matter myelination, after which grey matter volumes stabilise until late adulthood (55). Therefore, the observed developmental delay in verbal comprehension could be related to delayed grey matter volume development in CYA-SCD. Neuroimaging studies have shown associations between frontal lobe volumes and verbal comprehension tasks in CYA-TD (56). While this association remains unexplored in CYA-SCD, it is possible that reduced volumes in the frontal lobe could be contributing factors to the delay observed in verbal comprehension development.

Additional research in CYA-SCD aged 7–17 years found lower verbal comprehension performance (Woodcock-Johnson-III) in those with greater frontal–parietal cortical lesions (57), which was also related to lower cognitive control (Examiner Battery: measuring inhibition and dysregulation of EF). Interestingly, cognitive control is important for successful language comprehension (58). We know that children living with SCD experience EF difficulties (59) affecting the central executive and working memory performance. It is possible that there is a relationship between EF and verbal comprehension in CYA living with SCD as well, as previously shown in a healthy adult population (60). However, these hypotheses need further investigation.

It is important to mention that there may be other factors contributing to the delay observed in verbal comprehension, such as prematurity (61), SES (62), low school attendance and missed schooling due to hospital visits and health related problems. For example, painful crisis is common and is also associated with reduced grey matter volumes in CYA-SCD (63). This evidence suggests that there could be multiple factors affecting verbal comprehension development in CYA-SCD which need to be explored in future research.

Several studies have documented impaired EF skills in CYA-SCD. Our results not only agree with the previous literature (64, 65), but demonstrate that the development of EF skills in CYA-SCD are not only delayed, but show a slower rate of development. Our findings indicate that EF difficulties continue into young adulthood. Most difficulties seem to relate to the child’s ability to control and regulate emotions and behaviour, important for self-regulation (measured on the BRIEF BRI). Our research is the first to show that CYA-SCD plateau prematurely on the development impulsivity and inhibitory control (measured on the BRIEF BRI Inhibition). Recent research in children living with SCD aged 8–15 years found that impulsivity (Conners’ Continuous Performance Test) was negatively associated with health-related quality of life (Paediatric Quality of Life Inventory Sickle Cell Disease Module) (66). There is also a high incidence of ADHD diagnosis (25%) in children living with SCD aged 8–16 years, which may contribute to EF difficulties in these patients (67).

Parents of children living SCD observe behavioural difficulties already at a young age. We found that they have difficulties anticipating an activity or task, develop ideas and problem-solving strategies, which are crucial for efficient time management and planning (measured on the BRIEF MI). Similar findings were observed in 8- to 12-year old children living with SCD compared to typically developing controls (68). Berg et al. (68) found higher scores on the parent and teacher BRIEF for MI and GEC. The authors mention that parents of children living with SCD observe EF difficulties in daily behaviour such as organizing and remembering things (i.e., homework materials, daily chores) and initiating behaviour to start tasks.

Interestingly, older CYA-TD in our sample had more EF difficulties compared to CYA-SCD. There are multiple factors which need to be considered when understanding EF development. In addition to changes in adolescence which may be prolonged into early adult life, the impact of psychosocial factors (i.e., family environment, parental support and engagement) (69) and even school environment (70) could contribute to individual differences in EF-skills development. Studies have shown that family function plays an important role in the development of EF skills early in life (71). Downes et al. (71) noted that a positive family environment and functioning significantly impacted attention control and cognitive flexibility in pre-schoolers living with SCD. It is likely that these functions, especially attention control, may have a cascading effect on the development of other EF skills, suggesting that efforts to support families during the early years could benefit EF development later in life. However, health plays a crucial role in children with SCD, which needs to be considered as well. Current research in people living with SCD in a similar age range to our study (6–31 years) has shown that this group is vulnerable to EF deficits because of cerebral haemodynamic stress related to chronic anaemia and oxygen desaturation (52). For example, cerebral blood flow increases to maintain adequate oxygen delivery to the cortex and is an indicator of cerebral hemodynamic stress, which is associated with worse EF, as measured by the D-KEFS Tower Test (72). One neuroimaging study in the SCD population showed reduced resting state activity in the default mode network (important for cognitive processing) but increased activity in the pain processing regions suggesting allocation of cognitive resources towards pain management which may hinder normal EF development (73).

Besides multiple bio-psychosocial factors that affect EF development, it is also possible that our results could be attributed to recruitment methods. Other cohort participants have attended a neuropsychological evaluation based on a medical referral for academic concerns, whereas we recruited participants from clinics and community settings regardless of any history of cognitive difficulties, so it is probable that our sample has a broader range of EF profiles, making our results applicable in a wider clinical context.

We found that CYA-SCD-SCI+ showed more delay at onset and a slower rate of development for BRIEF MI. These results are in line with previous literature showing an association between poor EF and SCI status (74). Although we did not examine the site of SCI lesions, previous studies looking at SCI presence have noted that frontal lobe SCI lesions could explain poor EF, especially metacognition, such as inhibition and working memory (74, 75). People living with SCD and SCI tend to have poorer white matter microstructural integrity as compared to people living with SCD without SCI which may affect EF as well (76). However, more diffusion tensor-based imaging studies as well as functional connectivity on MRI studies are needed to explore the multidimensional relationship between structural connectivity and EF in detail.

In addition to finding that verbal comprehension was delayed in participants with SCI (although not significantly), we also found that haemoglobin, SpO₂, and CaO₂ are all predictors of verbal comprehension in CYA-SCD. Haemoglobin and CaO₂ were also associated with processing speed. These results are in line with previous studies exploring effects of blood oxygenation measures on cognition in people living with SCD. Using a path-analysis model, Hogan et al. (34) found that reduced SpO₂ is compensated by increased cerebral blood flow to the brain which results in lowered verbal IQ in adolescents living with SCD (mean age 17.4 years). Several studies of blood oxygenation measures and cognition have found that reduced haemoglobin is associated with reduced IQ in adolescents and young adults (51) and also neurodevelopmental delay in 9- to 12-month old infants living with SCD (77). Similarly, reduced processing speed was linked to lower CaO₂, poor white matter microstructural integrity in (35) and reduced white matter structural connectivity (78, 79) in CYA-SCD. Taken together, these data suggest that cerebral haemodynamic stress, central to sickle cell anaemia pathology, may result in poor long-term cognitive outcomes.