Neurocognitive deficits in survivors of childhood acute myeloid leukemia

Acute myeloid leukemia (AML) accounts for 25% of all cases of childhood leukemia, and it affects approximately 150–200 patients annually in Japan [1]. Because of improvements in treatment and supportive care over time, event-free survival and overall survival rates for children with AML have approached 60% and 70%, respectively [1, 2]. These improvements have led to an increased focus on late effects in long-term survivors [3, 4]. Recently, late neurocognitive effects among cancer survivors have attracted increasing attention because neurocognitive problems may limit the quality of life and functional outcomes of patients [5].

Although the majority of the literature has focused on evaluating the effects of methotrexate on cognitive outcomes in survivors of childhood acute lymphoblastic leukemia (ALL) [5, 6], many chemotherapeutic agents may cause neurocognitive impairment. Several reports described acute cerebral and cerebellar toxicity following treatment with high-dose cytarabine (HD Ara-C) [7‐10], which is used as a standard regimen for AML.

Because chemotherapy for childhood AML also contains neurotoxic agents and intrathecal therapy (IT) [1, 3], neurocognitive sequelae may occur in AML survivors. Nevertheless, few studies have examined neurocognitive outcomes associated with childhood AML.

This study evaluated the development of late cognitive effects following AML treatment in children with a focus on general intellectual ability and specific neurocognitive domains. In addition, we explored the associations of neurocognitive performance with clinical and demographic factors.

Twenty-eight patients < 20 years of age received a diagnosis of AML (n = 26) or AUL (n = 2) from 2006 to 2016. Among them, 18 patients satisfied the eligibility criteria for this study, and neurocognitive data were obtained from 12 patients. The reasons for exclusion or a lack of participation are noted in Fig. 1. On average, participants were 8.0 years old at diagnosis and 12.9 years old at assessment. Of the 12 participants, five received HSCT, three received total-body irradiation, and one presented with CNS disease at diagnosis. Participant with CNS disease showed ≥ 5 white blood cells with blasts in the CSF but without clinical CNS symptoms. Although brain magnetic resonance imaging performed in all patients before the treatment and an intracranial mass was detected in one patient, this patient was regarded CNS-negative because lack of clinical CNS symptoms. None of the participants developed neurological events or toxicity that affected other nervous systems during chemotherapy. One patient had a history of epilepsy, and none of participants had family medical history, such as psychiatric illnesses and epilepsy. Parental education and household income were used as proxies for family socioeconomic status. The full demographic characteristics of the participants are noted in Table 2.

The participants’ index scores on the Wechsler Intelligence Scales in relation to normative data are summarized in Table 3. The WISC-IV and WAIS-IV were completed by seven and four participants, respectively. Participants who completed the WISC-IV displayed significant impairment on the WMI (t (6) =  − 3.36, p = 0.015, M = 83.3 [95% confidence interval = 71.1–95.4]) but no statistical difference from the normative M for FSIQ and other index scores. Among the participants who underwent the WAIS-IV, FSIQ and the four primary index scores did not differ significantly from the normative M of 100.

As presented in Fig. 2, among the subtest scores, the Letter-Number Sequencing subtest score was the lowest (M = 6.9, SD = 3.2) and the Coding subtest score was the highest (M = 12.0, SD = 2.9) in the WISC-IV cases, whereas the Digit Span subtest score was the lowest (M = 9.5, SD = 2.3) and the Arithmetic subtest score was the highest (M = 12.0, SD = 3.7) in the WAIS-IV cases. The Digit Span subtest score was also low in the WISC-IV cases (M = 7.4, SD = 1.8). However, neither group exhibited any significant differences in the overall effect of the index scores and subtests.

The associations between FSIQ and all demographic and clinical variables listed in Table 2 were examined by univariate analysis. The t-test revealed no significant difference in the FSIQ between the WISC-IV and WAIS-IV cases (M = 93.7, SD = 9.3 and M = 105.0, SD = 15.8, respectively, p = 0.16). Among WISC-IV cases, participants aged 10 years and older showed higher FSIQ score than participants under 10 years of age (M = 100.0, SD = 6.1 and M = 85.3, SD = 4.5, respectively, p = 0.017). As presented in Fig. 3, FSIQ was moderately related to the time since diagnosis (rs = 0.59, p = 0.049) and strongly related to age at assessment (r = 0.69, p = 0.013). A longer time since diagnosis and older age at assessment were associated with high FSIQ. Multiple linear regression was performed to predict FSIQ based on the time since diagnosis and age at assessment. A significant regression equation was found (F (2, 9) = 11.74, R² = 0.66, p = 0.003). The participants’ predicted FSIQ was equal to 76.01 + 2.05 (time since diagnosis) + 0.88 (age at assessment), where time since diagnosis and age at assessment were measured in years. Both time since diagnosis and age at assessment were significant predictors of FSIQ.

The results of this study indicate that following treatment with HD Ara-C for AML, survivors exhibit impairment in working memory compared to population norms despite their comparable FSIQ. In multiple linear regression, we revealed that the time since diagnosis and age at assessment were both associated with FSIQ in survivors.

FSIQ did not differ between our AML survivor sample and Japanese population norms; however, deficits in working memory emerged among AML survivors. This is consistent with the findings of a recent report by the CCSS group, which suggested that survivors had a relative risk of impairment in at least one neurocognitive domain [16]. It was previously reported that deficits in specific cognitive domains may be unrelated to general intelligence [17]. In this study, participants who underwent the WISC-IV demonstrated significant impairment on the WMI. Participants performed poorly on the Digit Span subtest of the WAIS-IV, but impairment in the WMI score disappeared because of a high Arithmetic subtest score. Previous studies suggested that the Arithmetic subtest is a poor predictor of working memory ability, and the Digit Span and Letter-Number Sequencing subtests would be the strongest predictor variables [18, 19]. Because the Digit Span subtest more strongly reflects working memory than the Arithmetic subtest, working memory appears to be the most vulnerable function following childhood AML treatment.

Cerebellar dysfunction may lead to intellectual disorders, such as disturbances of executive function, including deficient planning, set-shifting, abstract reasoning, working memory, and decreased verbal fluency [20]. Cytarabine is known to cause cerebellar toxicity [7‐10]. Autopsies demonstrated the loss of Purkinje cells in the cerebellum and reactive Bergmann glial cell proliferation in adults following HD Ara-C treatment [21]. Diffuse heterogeneous brain hypoperfusion, identified by single-photon emission computed tomography, has been reported in children with AML who received HD Ara-C [22]. Although we cannot state so categorically, it can be speculated that HD Ara-C treatment may cause late neurocognitive effects because of cerebellar toxicity. Prior research supported an association between chemotherapy for childhood ALL and long-term neurocognitive deficiencies [23]. Although IT in AML treatment is less frequent than in ALL treatment [12, 24]., similar neurocognitive sequelae may occur in AML since it contains same agents (methotrexate, cytarabine, and hydrocortisone) used in ALL treatment.

Several studies have described associations of poor neurocognitive outcomes following childhood cancer treatment with lower socioeconomic status, older age at diagnosis, cranial radiation therapy, and a history of seizures [25, 26]. In our study, a longer time since diagnosis and older age at assessment were significantly related to a higher FSIQ, whereas age at diagnosis, gender, parental education, and household income were not significantly associated with neurocognitive outcomes. The length of schooling or some curriculums might explain participants’ improvement of the weak cognitive domain; however, we could not investigate the relationship between neurocognitive performance and Japanese educational system in this study. The lack of a relationship between FSIQ and these variables could reflect our small sample size. Although some participants underwent HSCT with a total-body irradiation conditioning regimen (maximum, 12 Gray), no significant difference in FSIQ was found between participants who did and did not undergo HSCT. This result is consistent with recent findings suggesting that the risk of neurocognitive dysfunction does not differ between HSCT and chemotherapy-only treatment in patients with AML [16, 27]. We could not assess the impact of seizures on neurocognitive function, as only one survivor had a history of seizures in this study. This patient had pre-existing epilepsy but being without medication because of seizure-free for years.

This study had several limitations. First, the generalizability of the study findings may have been compromised by the small sample size, and our results need to be replicated in a larger sample. Second, the cross-sectional design of this study made it impossible to compare neurocognitive outcomes with baseline data. It is possible that participants in this study might have had an above-average FSIQ prior to diagnosis and that the average results observed in the participants might indicate a decline from a higher baseline level. Third, this study lacked a matched comparison group. Although we tried to use sibling controls, few participants had siblings who were close in age. Thus, we were limited to comparing the outcome to the standard Wechsler Intelligence Scale scores. Finally, although some genetic factors may contribute to individual vulnerability to chemotherapy [28‐30], we did not examine any patients for genetic polymorphisms. Future studies, including an assessment of genetic polymorphisms, are required to fully understand the association between genetic factors and vulnerability to neurotoxic agents such as Ara-C.

The neurocognitive sequelae in children with AML have been considered more likely to have been caused by leukemia and its complications at presentation rather than by treatment [31]. By contrast, in our study, only one participant was CNS-positive at diagnosis, and no participants exhibited clinical CNS symptoms during treatment. Therefore, it is conceivable that the influence of the primary disease on cognitive function was small in this study.

Participants displayed no impairment in FSIQ, but they had poor performance in working memory. Working memory is responsible for the temporary storage and manipulation of information. It affects all judgments and activities in everyday life, such as conversation, reading, writing, and calculation. Difficulties in specific cognitive domains, even with good general intellectual ability, may affect the quality of life of survivors of childhood AML. Individually administered intelligence test for survivors of childhood AML might be helpful in detecting domains of cognitive vulnerability. Early detection of neurocognitive late effects and development of interventions for survivors with neurocognitive deficit are required to provide survivors with optimal psychosocial care.