Possible bimodality in the age distribution of pediatric functional neurological disorder

Résumé fourni par la source

Pediatric functional neurological disorder (FND) presents with various symptoms, including limb weakness, gait disturbance, tremor, seizures, and sensory disturbance, while patients' brains show no identifiable organic pathology.1-3 FND can occur as early as 3 or 4 years of age, but is more commonly reported in adolescents (over 10 years of age) who often face greater physical and psychological stressors in family, school, and social life.2, 3 A resting-state fMRI study in pediatric FND showed dysregulation within and across brain networks; an EEG study demonstrated overactivation of networks involved in threat detection; and an inpatient study found a blunted cortisol awakening response, suggesting hypothalamic–pituitary–adrenal (HPA) dysregulation.2 A recent perspective, based on the Ernst Strüngmann Forum on Youth Mental Health, hypothesizes that there are sensitive periods for neuroplasticity—windows when brain connectivity is most malleable—that emerge sequentially with overlap across different neural systems.4 For example, the threat-regulation system (involving cortical-hippocampal-amygdala networks) may be most plastic at earlier ages (3–20 years, peaking around 10–13) than the HPA-axis system (6–23 years, peaking around 13–16). This is followed by other systems such as prefrontal cortex connectivity and social cognition.4 This neurodevelopmental variation could contribute to variability in age at onset. For instance, the age distribution of anxiety disorder exhibits two onset peaks (one at 5.5 years and the other at 15.5 years)—a pattern that may reflect sensitive developmental periods related to threat-regulation and HPA-axis systems.4 However, no formal analysis has tested whether the age distribution of pediatric FND has a single peak or multiple peaks. Therefore, we aimed to identify and analyze secondary datasets that provide patient counts by single-year age category. Notably, most clinical studies of pediatric FND reported age using the mean or median (with range or interquartile range) in their “Table 1,” implicitly suggesting a single peak at 12–14 years (reviews3, 5). We identified three cohorts, among all the 23 pediatric FND cohorts we found, that had suitable publicly available data including patient counts for single-year age categories5-7 (see details in Table S1). First, Grattan-Smith et al. described 52 pediatric FND cases (69% motor, 77% sensory, and 15% seizure) in New South Wales, Australia (1975–1984), with age at admission recorded.6 The diagnosis was made by general pediatricians, pediatric neurologists, or child psychiatrists based on DSM-III; 90% of the patients underwent a formal psychiatric evaluation. Second, Raper et al. reported 124 pediatric FND cases (41% seizure, 18% sensory, and 16% motor) in Newcastle, UK (1997–2017) with age at presentation recorded.7 Twenty-seven of these patients were ≥ 16 years old and were excluded, leaving 97 cases for analysis. The diagnosis was made by pediatric neurologists based on ICD-10. Third, Yong et al. documented 97 pediatric cases (41% motor, 41% sensory, and 20% seizure) in Edinburgh, UK (2018–2020) with age at diagnosis recorded.5 The diagnosis was made by pediatric neurologists or general pediatricians based on DSM-5. Note that differences in diagnostic criteria across the cohorts (e.g., ICD-10 and DSM-III require a preceding psychological stress and a thorough differential diagnosis, whereas DSM-5 does not require a preceding psychological stress but emphasizes an inconsistency between neurological examination findings and symptoms1) might contribute to differing characteristics of the three cohorts. We primarily applied a Gaussian mixture model (GMM), which represents a data distribution as a weighted combination of multiple normal distribution components,8 to the combined data of the three cohorts (n = 246) and to each cohort (Fig. 1). We assumed one or two components (i.e., one and two peaks) and compared the model fits using the Bayesian Information Criterion (BIC) and the Akaike Information Criterion (AIC).8 Results show that in the combined data, BIC is 1177.8 and AIC is 1170.8 when assuming one component, while BIC is 1159.4 and AIC is 1141.9 when assuming two components (Fig. 1d). The reduction in both BIC and AIC when assuming two components suggests that a weighted combination of two components explained the data better. In the combined data, 70% of cases fell into a younger-peak component (mean = 10.4 years, SD = 2.32) and 30% into an older-peak component (mean = 14.0 years, SD = 0.85) (Fig. 1). We also calculated Ashman's D statistic,9 which quantifies the separation between two normal distribution components (values >2 suggest bimodality). Ashman's D value for the combined data was 2.07, which supports the interpretation of the BIC/AIC results. Then, as a robustness check, we applied the Hartigan's dip test, which nonparametrically tests whether a distribution is unimodal (null hypothesis) or not (alternative) without assuming normal distribution. The dip test yielded D statistic of 0.069 (P < 2.2 × 10−16) for the combined data, strongly rejecting unimodality. Please see Fig. 1a–c for the results for each of the three cohorts; the first cohort (the oldest and smallest) exhibits an inconsistent result (the distribution appears bimodal in the histogram Fig. 1a), whereas the second and third cohorts exhibit consistent results supporting bimodality. In summary, these findings suggest that the age distribution of pediatric FND can be bimodal. Each cohort showed a modest peak in late childhood (around 10 years), although the precise peak age varied in the three cohorts (8.76, 11.0, and 9.32 years, respectively). In the combined data, the separation between the younger and older peaks was visually subtle in the histogram, but the overall pattern still indicates a mixture of multiple distributions (Fig. 1d). These findings may suggest the existence of two developmental windows of vulnerability for FND onset. Although it is too early to conclude that childhood-onset FND is associated with dysfunction of the threat-regulation system and that adolescent-onset FND is associated with dysfunction of the HPA-axis system, this potential bimodality may reflect the multifactorial etiology of FND—diverse physical, social, and environmental factors impact different neural systems at various neurodevelopmental stages. It is not known how etiological differences affect treatment approaches in FND, which is an important question for future research. There are other mechanisms to explain the bimodality. First, delayed diagnosis of FND in some cases, combined with prompt diagnosis in the remaining cases, can produce bimodality in the age at presentation/diagnosis of FND. Kozlowska and colleagues reported that the time between symptom onset and formal diagnosis varies from 1 week to 4 years (median of 4 months),10 suggesting the younger peak may correspond to the cases with prompt diagnosis and the older one to those with delayed diagnosis. If this is the case, increased awareness among pediatricians, neurologists, psychiatrists, school nurses, and parents is warranted, since a delayed diagnosis results in a poorer prognosis.2, 3 Indeed, textbooks often state that FND is most frequently seen in adolescents or later1-3, 5; however, this does not necessarily mean that FND cannot occur in children under 10 years of age. Second, certain comorbid or demographic factors (e.g., anxiety levels, sex differences, or different FND subtypes such as movement disorders vs. functional seizures) may explain the bimodal age distribution. Thus, an important limitation of the present analysis is that detailed data on comorbidities, sex, and FND subtypes were not publicly available for all cohorts. Another key limitation is the uncertain representativeness of the analyzed cohorts. We gratefully acknowledge support from the National Institutes of Health (K01AI166347 to AN), and the National Science Fo

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.