Cognitive and behavioral abnormalities in children after hematopoietic stem cell transplantation for severe congenital immunodeficiencies

Hematopoietic stem cell transplantation (HSCT) is a highly successful treatment for severe congenital immunodeficiency syndromes and results in long-term correction of immune function. Since the first transplantations were performed nearly 40 years ago, an increasing number of procedures have been performed with significant improvement in survival rates over time. To date, outcome has been measured in terms of survival and recovery of immune function, but more global measures of health and functioning, such as cognitive and behavioral outcome, have not been formally documented in this patient group.

The impact of HSCT on cognitive outcome has been addressed by a number of reports, but the majority of children studied had undergone transplantation for hematologic malignancies.^1,,,,–6  The findings in these studies are not consistent and vary from minimal evidence for cognitive decline in some studies^2,5  to a significant decrease at 1 year in another.³  However, the 2 largest of these studies suggest that children transplanted at an early age, and especially those transplanted under 3 years of age, are at greater risk of cognitive impairment.^2,3  Some studies have included patients with severe congenital immunodeficiency, but because of the small numbers involved in this group, they could not be analyzed separately. However, a poorer outcome was noted for these children.^3,7 

Severe congenital immunodeficiencies are a diverse group of disorders of the immune system and arise from a number of monogenic defects.⁸  In the majority of diseases, expression of the defective gene is confined to the immune system, whereas in a few specific instances, gene expression is more widespread. In some conditions, such as adenosine deaminase (ADA)–deficient severe combined immunodeficiency (SCID), where ADA is expressed in all body tissues, cognitive defects may occur as an intrinsic consequence of the disease, and this has previously been documented both by ourselves and others.^9,10  However, in other conditions where only immune or hematopoietic cells are adversely affected, no long-term cognitive or behavioral abnormalities would be expected because of the underlying disease process. However, these children have a very abnormal experience in early childhood because of extended periods of illness and social isolation, and this in itself may have an impact on their development.

The impetus for the present study arose from a previous small case-control analysis in which we compared cognitive and behavioral outcome in patients with ADA SCID after HSCT in comparison with a group of non-ADA SCID patients.¹⁰  Patients with ADA SCID were shown to have a significantly lower IQ and a greater incidence of behavioral problems, but surprisingly the non-ADA SCID cohort also had IQ scores that were lower than the population norm and showed abnormalities in areas of behavioral function.

Based on these findings, we initiated a study to assess all patients at our center who had undergone HSCT for severe congenital immunodeficiencies to understand the range and extent of cognitive and behavioral problems and also to identify specific risk factors that may lead to adverse outcome. We looked at a range of possible risk factors, including those identified in previous research (age at transplantation), treatment-related factors (chemotherapy conditioning, length of stay in hospital at time of transplantation, and admission to intensive care), demographic factors (socioeconomic status [SES]), and factors related to the immunodeficiency condition itself (diagnostic group and molecular diagnosis).

All children diagnosed with a severe congenital immunodeficiency and treated by HSCT at one center (Great Ormond Street Hospital, London, United Kingdom) between 1979 and 2003 were contacted to take part in the study. Eligible participants had to be at least 3.5 years old at the time of assessment and at least 1 year after transplantation.

Of 117 potential participants, 105 (90%) completed a detailed psychological assessment; 5 declined to take part and 7 could not be contacted because they no longer attended the hospital, and no current contact details were available. The assessment protocol was agreed by the Research Ethics Committee at Great Ormond Street Hospital, and informed consent was obtained according to the Declaration of Helsinki from parents of children younger than 16 years and from patients older than 16 years. Assent was obtained from children between 8 and 15 years of age.

The patients differed from the United Kingdom population in terms of the high number of families from minority ethnic background and the number of families not speaking English as a first language (Table 1). To control for these factors, a group of siblings was recruited to match for social and ethnic background. Siblings of children from the 2 most common diagnostic groups (SCID or combined immunodeficiency [CID]) were assessed using the same measures of cognitive functioning. There were 62 children in these diagnostic groups, and 37 families had an unaffected sibling, of which 22 (59%) completed the assessment.

The control group were well matched with both the original cohort and the potential sibling sample in terms of English as a first language and ethnic background (Table S2, available on the Blood website; see the Supplemental Materials link at the top of the online article). However, there were more girls in the control group, and they were older. Neither of these factors has been shown to be associated with outcome.

Participants completed a full psychological assessment that included a cognitive assessment and standardized questionnaires assessing emotional and behavioral functioning. SES was rated on a 5-point scale based on the occupation of the main earner.¹¹ 

Cognitive function was assessed using the appropriate Wechsler Intelligence Scale for the child's age: Wechsler Pre School and Primary Scale of Intelligence Third United Kingdom Edition (WPPSI-III^UK),¹²  Wechsler Intelligence Scale for Children Third United Kingdom edition (WISC-III ^UK),¹³  and the Wechsler Adult Intelligence Scale Third United Kingdom edition (WAIS-III^UK).¹⁴  The Wechsler scales give a Full Scale Intelligence Quotient (IQ) based on several subtests assessing different aspects of intelligence, such as verbal reasoning, general knowledge, visuo-spatial, and perceptual skills.

Two deaf children were assessed using the Leiter International Performance Scale-Revised.¹⁵  One child with profound developmental delay was assessed using the Griffiths Mental Development Scales¹⁶  because her level of ability fell below the lowest level of the Wechsler scales. Both of these scales give scores that are compatible with the Full Scale IQ score from the Wechsler Scales.

Emotional and behavioral difficulties were assessed using 3 standardized questionnaires, completed by the child's parent and teacher. The Strengths and Difficulties Questionnaire (SDQ) is a brief screening questionnaire for identifying emotional and behavioral difficulties in children 4 to 16 years of age. It consists of 25 questions covering 4 common problem areas (behavioral. emotional, hyperactivity, and peer relationship difficulties) and one scale measuring prosocial behavior.¹⁷  Normative data are available based on a representative population sample of 10 000 children. The Conners' Rating Scales (short form) is a brief 27-item questionnaire that covers problems with attention, concentration, learning, and hyperactivity in children 3 to 17 years of age. Raw scores are converted into standardized T scores (mean ± SD, 50 ± 10) based on normative data for boys and girls in 5 different age groups.¹⁸  The Adaptive and Behavior Assessment System (2nd edition) is a questionnaire designed to assess day-to-day functioning in children 5 to 16 years of age.¹⁹  It covers 9 different areas (communication skills, community use, functional academics, home living, health and safety, leisure, self-care, self-direction, and social skills). These scaled scores can be combined to give a General Adaptive Composite Score, which is standardized by age and has a mean of 100 and a standard deviation of 15.

Groups were compared using t test or one-way analysis of variance. Stepwise multiple regression was used to determine the main predictors of IQ and SDQ scores. The data were analyzed using SPSS for Windows, version 15 (SPSS, Chicago, IL).

Patient characteristics are shown in Table 1. Because of the large number of X-linked immunodeficiency conditions, there were more boys than girls. The SCID group was defined by either a molecular diagnosis of SCID (including defects of the common γ chain, JAK-3, RAG1/2, IL-7Rα, and major histocompatibility complex [MHC] class II) or those with an unspecified molecular defect but with severe T-cell lymphopenia and absent or abnormal humoral immunity. ADA SCID was separated from other SCID types because of its known neuropsychologic defects. The CID group included those with no molecular diagnosis but with T cells present but low in number and abnormal in function and accompanied by various degrees of humoral abnormalities. Other specific disease entities, such as Wiskott-Aldrich syndrome and chronic granulomatous disease, are also shown. Full intensity conditioning was defined as busulphan (16-20 mg/kg) and cyclophosphamide (200 mg/m²) with or without serotherapy (n = 44). Most patients with reduced intensity transplantations (n = 36) received fludarabine (150 mg/kg) and melphalan (140 mg/m²) with serotherapy (either Campath 1H or antithymocyte globulin), although other reduced intensity transplantations regimens, including fludarabine/cyclophosphamide, were used in a minority of patients. In 23 patients, no chemotherapy conditioning was used.

Average age at assessment was 11 years (range, 3.5-25 years), and average time since transplantation was 7 years, 7 months (range, 13 months to 25 years). Average age at transplantation was 3 years 6 months (median, 13 months). A high proportion of children came from minority ethnic backgrounds compared with the population of the United Kingdom, and a high proportion of the patients had consanguineous parents; 26% of the patient group did not speak English as their first language at home, and this increased to 42% of the patients from an Asian background.

The mean Full Scale IQ score for the whole cohort was 85 (95% confidence interval, 81-90), 15 points lower than the population average of 100 (z = 7.5, P < .001). IQ scores for 19% of the cohort fell within the learning difficulty range of ability (IQ < 70) compared with 2% in the general population. Of the 92 children still in school, 27% had a Statement of Special Educational Needs compared with 2.5% in the general population.²⁰  Nine children were attending a school for children with special needs and/or learning difficulties. Twelve patients were older than 18 years and had left school. Five were studying for higher qualifications or in vocational training, 2 were unemployed, and 5 were in full-time employment.

Mean IQ scores by the 5 main diagnostic groups are shown in Figure 1 and Table S1. All were significantly below the population mean of 100, and ADA-deficient SCID was significantly lower than SCID and CID (P < .01). The Chédiak-Higashi syndrome group also had a very low score.

The SCID group was further analyzed on the basis of underlying molecular diagnosis. Two groups were identified: (1) patients with expression of the defective gene confined to the immune system (γc, JAK3, IL-7Rα, RAG1/2) (n = 27) and (2) no molecular diagnosis or the defective gene also expressed outside the immune system (n = 16) (MHC class II [n = 1], SCID phenotype with dyskeratosis congenita [n = 1], known fibroblast radiosensitivity [n = 3], no diagnosis and radiosensitivity not performed [n = 11]). The mean IQ score for group 1 was near normal and significantly higher than for group 2 (96; 95% confidence interval, 88-104 vs 81; 95% confidence interval, 71-92; P = .03). For those with known cellular radiosensitivity (n = 3), the mean IQ score was 87 (range, 78-94).

Table 2 shows IQ scores for the main factors associated with outcome. In contrast to previous studies, younger age at transplantation (defined as younger than 3 years of age) was not associated with lower IQ score. There was also no effect of any conditioning regimen compared with no conditioning, and no significant correlation between IQ score and length of time spent in hospital during transplantation. Children admitted to a pediatric unit intensive care (PICU) during their transplantation scored significantly lower than those not admitted to PICU.

As might be expected, IQ score was significantly related to SES and was significantly lower in children for whom English was a second language. It was also significantly lower in children from consanguineous parents.

The control group included 22 siblings of children with a diagnosis of SCID or CID. The control group were well matched with both the original cohort and the potential sibling sample in terms of English as a first language and ethnic background (Table S2). However, there were more girls in the control group, and they were older. Neither of these factors has been shown to be associated with outcome.

Figure 2 shows the mean IQ scores for patients and siblings, with the siblings 18 points higher than the patients (P = .006). In addition, siblings with consanguineous parents had lower IQ scores than siblings with unrelated parents (94 vs 113, P = .02), suggesting that consanguinity itself may have an adverse effect on cognitive outcome even in unaffected persons.

Five variables were significantly associated with IQ scores (SES, consanguinity, English as a second language, PICU admission, and diagnosis of ADA-deficient SCID). These and the 2 variables age at transplantation and conditioning were entered into a stepwise multiple regression analysis. Table 3 shows the final model in which the 4 variables SES, consanguinity, diagnosis of ADA, and admission to PICU together accounted for 33% of the variance in IQ score.

The results for parent-rated SDQ are shown in Table 4 for the 5 scales measuring difficulties in emotional and behavioral function and the impact scale based on the 85 children for whom the questionnaire was completed. Higher scores indicate higher levels of difficulties on these scales, except on the prosocial scale where higher scores indicate better social skills. Overall, 25% of the cohort scored above the threshold indicating clinically significant difficulties (total difficulties score ≥ 17) compared with 10% in the general population.^21,22  Scores were significantly higher than the norms for peer relationships and impact for boys, and for emotional, hyperactivity, peer relationships, and total difficulties for girls.

Teacher-rated SDQ scores (Table S3) for 68 children showed that 24% of the cohort scored above the threshold indicating clinically significant difficulties. Scores were significantly higher than the norms for total difficulties and impact for boys, and for peer relationships for girls.

Both parent-rated and teacher-rated total SDQ scores were strongly inversely related to IQ scores (r = −0.44 and r = −0.45, P < .01) indicating that lower cognitive ability was associated with higher levels of emotional and behavioral difficulties.

The parent-rated total SDQ scores were compared for the 5 main diagnostic groups. Patients with ADA SCID had significantly higher scores than those with SCID, CID, or Wiskott-Aldrich syndrome (P < .01). The other main factors associated with parent-rated SDQ scores were IQ score, SES, and consanguinity. However, PICU admission, length of stay in hospital, age, and English as a first language were not associated with outcome. One-way analysis of variance indicated that patients who had received no conditioning had significantly higher scores than those of patients who had reduced intensity conditioning (P = .05) (Table S4), although this is probably related to the fact that a large number of ADA SCID patients received transplantation without conditioning.

The 5 variables significantly correlated with parent rated total SDQ score (IQ, SES, consanguinity, diagnosis of ADA-deficient SCID, no conditioning) were entered into a stepwise multiple regression analysis (Table 5). IQ and SES together accounted for 37% of the variance in total SDQ score, and diagnosis of ADA SCID was of borderline significance. Thus, it appears that IQ and SES are the major determinants of behavioral outcome after HSCT.

On other measures of behavioral outcome, the mean score on the Conners' Rating Scales was significantly higher than the norm (mean ± SD, 56.0 ± 12.1, P < .001), indicating higher levels of concentration, attention, and hyperactivity problems. The mean total score on the Adaptive and Behavior Assessment System was significantly lower than the norm (mean ± SD, 82.9 ± 24.1, P < .001), indicating lower levels of functional skills, consistent with the lower IQ score.

HSCT has transformed the survival of patients with severe congenital immunodeficiency and growing numbers of children can expect a normal life span free of infection. It is therefore important to identify and manage other aspects of health that may otherwise inhibit normal development and function. The results from this large scale study indicate that children who have received HSCT for severe congenital immunodeficiency have lower than average levels of cognitive ability after HSCT. However, because of the large number of patients studied, we were also able to identify specific patient groups which are most severely affected.

The results from previous studies on the long-term impact of HSCT on cognitive outcome have been inconsistent, but this may be partly because of differences in patient groups between the studies and the numbers of persons analyzed.^2,,,–6  Our study included only children with a diagnosis of severe congenital immunodeficiency and is the first large-scale study of its type. Other studies that have included only small numbers of patients with immunodeficiency have also suggested lower cognitive scores for these groups, consistent with the findings from this study.

One of the major risk factors identified is the underlying molecular diagnosis. It is clear that certain specific gene defects while manifesting most profoundly in the immune system can also have significant nonimmunologic sequelae because of their more systemic expression. This study confirms and adds to the findings of cognitive, behavioral, and neurologic defects in ADA SCID.^9,10  ADA is known to be expressed systemically and other nonimmunologic deficits, including audiologic, hepatic, renal, and skeletal abnormalities, are well documented.^23,,–26  We also show that in addition to previously described motor neurologic defects,²⁷  children with Chédiak-Higashi syndrome have severe cognitive problems. The LYST protein defective in Chédiak-Higashi syndrome is ubiquitously expressed and is involved in controlling exocytosis of secretory granules, and its absence may therefore disturb neuronal lysosomal trafficking.²⁸ 

Although mean IQ in the SCID group was abnormal, there was a significant difference between those where the molecular defect was confined to the immune system (γc, JAK3, IL7 Rα, RAG1/2 defects) and those in which no molecular diagnosis was identified or in which the defective gene was more widely expressed. Although in the majority of this latter group the molecular defect is not known, it is possible that a number of patients may harbor defects in genes, such as Cernunnos²⁹  and LIG4,³⁰  both of which are expressed ubiquitously and are associated with developmental abnormalities through defects in double-strand DNA break repair. Our patients did not have overt microcephaly or developmental delay as has been described in these patients, but it is possible that other less obvious problems such as cognitive impairment are manifest. It is also possible that other as yet unidentified genes, which are expressed systemically, are defective in some of these patients. These observations suggest that the underlying genetic defect, regardless of transplantation and transplantation course, has a major influence on neuropsychological outcome.

Also of importance is parental consanguinity, which was strongly related to lower IQ score. The influence of consanguinity is supported by the finding that unaffected siblings from consanguineous pedigrees had a significantly lower IQ than unaffected siblings from nonconsanguineous families. The strong association between parental consanguinity and IQ also suggests that these differences are not related to the transplantation experience itself but that other inherited traits affecting IQ are carried through in such pedigrees. This cohort had a high rate of parental consanguinity as would be expected for children with genetic conditions, such as SCID. Previous studies have also found lower IQ in children of consanguineous parents in Indian families.^31,32  However, these marriages are often perceived within Muslim and Arab cultures to have strong social and economic benefits and are therefore relatively common.³³  Nonetheless, it is important to provide genetic counseling to these communities regarding the raised incidence of such difficulties to promote informed decision making, although this has to be done in a culturally sensitive way.

In contrast to suggestions from previous research, this study showed that age at transplantation is not related to cognitive outcome. Children younger than 3 years of age at the time of transplantation were previously thought to be at greater risk of cognitive decline,²  although the authors were cautious about their conclusion because of the small numbers of patients studied. The present study in which 65 (62%) of the children received transplants at less than 3 years of age convincingly demonstrates no significant risk to IQ as a result of age at the time of HSCT for children with severe congenital immunodeficiency.

Also of importance is the role of conditioning and its effect on cognitive outcome. Unlike other previous studies, we had the opportunity to study a cohort that had been transplanted without conditioning (because of the SCID phenotype and availability of a matched related donor) and to compare outcome with groups who had undergone full or reduced intensity conditioning. The use of reduced intensity conditioning regimens most commonly involving fludarabine/melphalan has improved overall transplantation survival outcome for congenital immunodeficiencies in comparison to busulphan/cyclophosphamide-based regimens, most probably as a result of reduced organ toxicities.^34,35  However, the overall IQ was comparable in all 3 groups and reassuringly suggests no neurotoxic effect as a result of any of these chemotherapeutic agents. In previous reports, total body irradiation has been often cited as the major conditioning risk factor.^36,37  In our cohort, total body irradiation was not a risk factor as this is not a standard HSCT conditioning agent for immunodeficiency patients.

This study was a cross-sectional design, and it is therefore not possible to determine whether these low IQ scores occurred as a result of HSCT or were already present before transplantation. It would have been very difficult to carry out a pre-post assessment of cognitive ability because the median age at HSCT was 13 months and it is not possible to obtain an accurate estimate of cognitive ability at such a young age. The only treatment-related factor associated with outcome was admission to PICU, which may be seen as a marker of the severity of the clinical course during transplantation.

In terms of emotional and behavioral outcome, both SES and lower IQ were found to be strongly associated with SDQ score, and these have consistently been associated with outcome in previous studies.^12,38  Patients also scored lower on a measure of functional or adaptive skills, consistent with the level of cognitive ability observed. This study found that diagnosis of ADA SCID was also associated with higher rates of emotional and behavioral difficulties, reflecting the systemic impact of this condition and consistent with previous research. Most studies of emotional and behavioral outcome for children with other types of chronic illness have found small or no increased risk of behavioral difficulties, but an increased rate of emotional difficulties.³⁹  This study also found increased rates of peer relationship or social difficulties, which is consistent with one other study of children after transplantation.⁴⁰  This may indicate specific difficulties in socialization as a consequence of prolonged hospitalization and isolation during treatment at a young age. However, further research is needed to look at this in more detail because other factors, such as physical appearance and cognitive ability, may also be related to the higher levels of social difficulties observed.

This study has focused specifically on outcome in patients with severe congentital immunodeficiency and given the number of patients studied, is able to demonstrate with confidence specific risk factors for cognitive and behavioral outcome. Genetic factors are the most important determinants for determining cognitive outcome. If the underlying gene defect was lymphoid specific, there were no major long-term cognitive sequelae, but conditions harboring systemically expressed gene defects resulted in lower cognitive outcome. Other undefined genetic determinants of cognitive function are also carried through as a result of parental consanguinity. Behavioral outcome is largely a consequence of IQ and specific genetic diagnoses such as ADA SCID. This study is also important in demonstrating that important procedure-related variables, such as age at HSCT and conditioning regimen, are not significant risk factors. Together, these results emphasize the importance of identifying the underlying genetic defect and will allow more accurate counseling and directed follow-up of transplantation patients.

Contribution: P.T. and H.B.G. designed the research, interpreted the data, and contributed to the writing of the manuscript; P.T., E.P., E.S., and K.O. carried out the research assessments; P.T. analyzed the data and performed statistical tests; T.J.C. provided statistical advice and interpreted the data; J.G. and J.H.X.-B. provided essential data from the BMT database; and A.J., A.J.T., E.G.D., and P.A.V. contributed to research design and writing the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Penny Titman, Department of Psychosocial Services, Great Ormond Street Hospital NHS Trust, Great Ormond Street, London WC1N 3JH, United Kingdom; e-mail: titmap@gosh.nhs.uk.