The Diagnostic Yield of Investigating Developmental Regression in Children: A Systematic Review and Meta-Analysis

This systematic review was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta- Analyses (PRISMA) guidelines (Page et al., 2021). The protocol was registered in the PROSPERO database on the 12th of July 2023 (Registration number CRD42023416466).

Our search strategy yielded 11,283 published records, of which 347 published records were reviewed in full text and 15 included in the final analysis. Eight full text published records were not retrievable despite attempts to contact authors (Fig. 1).

This is the first systematic review to report the diagnostic yield of investigating children with symptoms of developmental regression. Despite the growing awareness of conditions that feature developmental regression, and frequency of developmental regression as a symptom we only found diagnostic yield information for 596 children, from 15 published records spanning nearly 20 years (from 2005 to 2024).

When childhood conditions were grouped, children with developmental regression and neurological signs (Ma et al., 2013; Muthaffar, 2021), or with suspected DEE (Essajee et al., 2022; Hong et al., 2020), or suspected Rett syndrome (Iwama et al., 2019) were more likely to receive a diagnostic result than children investigated for an inborn error of metabolism (Romao et al., 2017), or autistic children with developmental regression (Byrne et al., 2023; Kosinovsky et al., 2005; Parmeggiani et al., 2010). There was an overall 40% diagnostic yield for children with developmental regression and developmental delay, higher than for children with suspected inborn errors or metabolism (10%) and autistic children (9%).

Developmental regression in autistic children has been studied and reported for decades yet the search for a biological marker for autistic regression remains elusive. As knowledge of phenotype characterisation advances, the heterogeneity of clinical presentation differences are highlighted. Approximately 30 percent of autistic children will regress in speech and social abilities, typically at a mean age of 20 months (Tan et al., 2021), but some autistic children will also lose skills in other domains, or at a later age. Whether later onset, or regression in domains other than speech, language and social skill presentations are phenotypic expansions of autistic regression, or novel conditions requires further scrutiny. Indeed, decisions on whether to investigate autistic children with developmental regression are not uniformly agreed, however in the latest version of the diagnostic statistical manual of mental disorders (5th ed., text rev.) authors recommend that if regression occurs in an autistic child outside the typical age range, then alternative diagnoses should be considered (American PsychiatricAssociation, 2022).

Only two published records reported on neurophysiological findings and both published records reported low diagnostic yields for autistic children with developmental regression (Kosinovsky et al., 2005; Parmeggiani et al., 2010). These results align with a recent meta-analysis on the relationship between electrophysiological (EEG) changes, epilepsy and regressive autism that reported only a weak association (Barger et al., 2017). More broadly inclusive published records are needed to determine the diagnostic utility of EEG for children with symptoms of developmental regression especially considering the significance of EEG investigations in DEE conditions (Scheffer et al., 2017). Greater phenotypic characterisation with EEG correlations may reveal specific features that increase the likelihood of significant EEG findings leading to diagnostic clarification and tailored management, for example with a diagnosis of Landau Kleffner syndrome (Clark et al., 2021).

Metabolic inborn errors of metabolism (IEM) are a heterogeneous group of disorders that can present with symptoms that include developmental regression in children. IEM’s include aminoacidopathies, organic acidaemias, lysosomal storage disorders and urea cycle disorders. Whether genomic testing such as WES, can identify all metabolic conditions that cause symptoms of developmental regression is unclear. WES is purported to be beneficial for diagnostic clarification of some inherited metabolic disorders that present with a broad range of developmental symptoms (Delanne et al., 2021), but data is not specific for children with developmental regression. We report that the pooled diagnostic yield for children with a suspected IEM was 10%, which is higher than the diagnostic yield of metabolic testing reported for children with ID or GDD (Vallance et al., 2021). Early detection of metabolic conditions is essential to ensure access to tailored interventions that may be disease modifying and improve neurodevelopmental outcomes (Eun & Hahn, 2015).

We report that genetic/genomic investigations were most likely to result in an explanatory diagnosis regardless of childhood condition (Table V in supplementary material). Genomic investigations, such as WES, are considered especially important for diagnosing clinically heterogeneous presentations such as are commonly reported for children with developmental regression (Iglesias et al., 2014). WES was reported to be first used as a clinical tool in 2009 and improvements to access and affordability have contributed to the global increase in availability (Phillips et al., 2021), and clinical use over the last decade. However, health inequity exists in accessing genomic testing especially in lower- and middle-income countries (Helmy et al., 2016). In 2020 in Australia the national health insurance scheme, Medicare, introduced a rebate to cover the cost of WES for children under the age of 10 years who met specified criteria including moderate to severe intellectual disability (ID), or global developmental delay (GDD) and or facial dysmorphism and or congenital abnormalities (Sachdev et al., 2021). Currently, children without an ID or GDD or children over the age of 10 years who regress in developmental abilities are not able to access Medicare funding for whole exome sequencing. This review describes the characteristics of included studies which report that the age of children presenting with developmental regression may be older than 10 years and children may not have an ID or GDD. The utility of genomic testing for children with undifferentiated developmental regression is high given diagnostic clarification may result in important changes to management (Tan et al., 2017). For example, in one included study a child with developmental regression and symptoms suggestive of Leigh syndrome had WES testing and a diagnosis of sulfite oxidase deficiency was made, with a direct influence on a management change (Hong et al., 2020). There are health economic benefits to investigating complex presentations with WES over a series of potentially invasive tests and procedures (Schofield et al., 2019). Indeed, WES has been recommended as first line investigation for children with unexplained neurodevelopmental disorders (Srivastava et al., 2019).

The role of neuroimaging for children with developmental regression remains unclear. Both published records that reported MRI-Brain results included in this meta-analysis reported a low diagnostic yield. Similarly, results from a recent systematic review that aimed to report rates of MRI abnormalities in children with developmental delays found that there were widely varying rates of abnormality depending on clinical presentations and insufficient evidence to inform guideline recommendations (Murias et al., 2017). Further published records that consider the symptoms of developmental regression and include potentially differentiating features such as age of onset of regression, types of skills lost and trajectory of loss, and any co-occurring symptoms or clinical findings are essential to determine the diagnostic yield of MRI-Brain specifically for children with developmental regression.

Diagnostic yield varied by year of publication and country/continent of publication (see supplementary material Figure iv). In many countries over the last decade investigations such as genomic testing has become more accessible. We report that published records before 2011 had a diagnostic yield of 13% and neither record included genetic/genomic investigations (Kosinovsky et al., 2005; Parmeggiani et al., 2010). When published records were grouped into continents, the most records were from Asia and were mostly genetic/genomic or metabolic investigations. The use of investigations, such as genomic investigations, and availability of newborn screening tests may vary between countries, and influence choice of investigations. Infants identified through newborn screening may be diagnosed with a condition known to feature developmental regression, changing timing of interventions and improving outcomes. One example is X-linked adrenoleukodystrophy, which is included in newborn screening in many states of the United States of America (Reynolds et al., 2023) that typically presents with developmental regression in later primary school age. Earlier detection and intervention may alter disease progression and enable child and family support network establishment (Moser et al., 2005). Currently in Australia the inclusion of X-linked adrenoleukodystrophy to the panel of conditions screened in newborns is under review with the medical services advisory committee.

Grouping childhood conditions aimed to determine whether developmental abilities, co-occurring conditions like autism or seizures, influence the diagnostic yield for children presenting with developmental regression. However, as described, there are overlaps between condition groups that may have impacted upon reported diagnostic yields. For example, a recent study reported outcomes for 78 children with neurodevelopmental disorders, of whom 7 experienced developmental regression. Of these, all 7 received a diagnostic outcome from whole exome sequencing; 2 children were diagnosed with Phelan McDermid syndrome, 1 with Rett syndrome, 3 with neurogenetic conditions and 1 with DEE. For the child diagnosed with DEE the study did not report clinical details of symptoms prior to or at the time of regression onset. The children with neurogenetic conditions had undifferentiated neurological signs such as diplegia which overlaps with NDD-neuro grouping however whether these signs were present at the onset of regression is unclear (Srivastava et al., 2014). Another study included 187 children with developmental delay with or without regression and seizures. The authors did not report overlapping symptoms, therefore it was not possible to determine how many of the 137 children with regression also presented with seizures (Karimzadeh et al., 2017). Future research needs to report clinical characteristics at the onset of developmental regression to enable clinicians to request investigations based on evidence informed diagnostic yield data tailored to clinical characteristics.

Lastly, there is no agreed approach to define developmental regression and therefore future research must reach towards a consensus to enable consistency across studies that include children experiencing developmental regression.

Publications that report the diagnostic yield of investigating children with symptoms of developmental regression are scarce. Despite the small number of published records and sample size, children with developmental regression and undifferentiated neurological or epileptic symptoms are most likely to receive an informative diagnosis. Diagnostic yield of genetic and genomic investigations were statistically significantly higher when compared with metabolic, neuroimaging and neurophysiology investigations for any child presenting with symptoms of developmental regression.

These results have identified knowledge gaps that need to be addressed to allow immediate translation of study findings to improve care for children with developmental regression. Namely, studies including children with developmental regression should report clinical characteristics at presentation so that the diagnostic yield of investigations completed early are known and this must occur in parallel with an agreed approach to define and measure developmental regression in children.