Background: Aromatic L-amino acid decarboxylase deficiency (AADCd) is an autosomal recessive neurometabolic condition characterized by severe combined neurotransmitter deficiencies leading to hypotonia, oculogyric crisis, developmental delays, movement disorders, and other autonomic symptoms (1). Despite the early onset of AADCd, typically in the first year of life, most patients experience significant delays between symptom onset and diagnosis (1). Given the devastating impact of AADCd on brain development and bodily function, the ability to improve outcomes strongly depends on the ability to identify patients early in the disease course prior to irreversible damage (1). A systematic automated approach to analyzing electronic health record (EHR) data is needed to decrease the AADDd diagnostic delay resulting from phenotypic variability and clinical indicators that overlap with other more common pathologies.
Hypothesis: Individuals at risk for AADCd can be detected using an automated prediction scoring system (APSS) run on EHR data followed by a genetic counselor review.
Methods: An APSS for AADCd based on the initial criteria developed by Himmelreich in 2019 and Wassenberg in 2017 was expanded to include structured and unstructured data points found in EHR. Structured data types included ICD10 codes, procedures, lab results, and imaging results. Unstructured data types included key terms and phrases in notes identified using Natural Language Processing and confirmed by a genetic counselor. The scoring was validated using known cases of AADCd from the medical literature and “real-life” controls. The APSS was run against 5 years of de-identified EHR data from a large children’s healthcare system (1,071,223 patients), a large academic healthcare system (1,700,000 patients), and a community healthcare system in a medically underserved region (247,584 patients). The patients scoring within the “at-risk” or “possible-risk” categories were reviewed by a genetic counselor for accuracy and appropriateness. The results were then returned to the healthcare systems for re-identification and provider notification.
Results: From the diverse sites, 7 patients were identified as “high-risk” for AADCd and 32 patients at “possible risk” for AADCd. Thirty-nine providers were contacted with information about AADCd and symptoms that suggest their patient could have an underlying genetic cause for their symptoms. The vast majority of providers were interested in the information and planned to discuss AADCd with their patients/families as well as possibly refer to genetics or perform testing themselves. Five patients were tested for AADCd at the time of data collection according to site coordinators. Although no patients to date have been diagnosed with AADCd, two patients had positive genetic testing on a cerebral palsy spectrum disorders gene panel. Specifically, testing identified one patient with an ATP-1A3 pathogenic variant possibly related to alternating hemiplegia of childhood (ACH) and one patient with 2 VUSs in CAMTA1 possibly related to Cerebellar Dysfunction With Variable Cognitive And Behavioral Abnormalities. In addition, one patient had a heterozygous pathogenic variant in COQ2, but the tested plasma CoE Q levels were normal and the patient was signed out as a carrier for Primary deficiency of coenzyme Q10.
Data was also collected on the racial distribution of “at-risk” individuals at two sites to determine if this method of identifying at-risk patients decreased unconscious racial bias as compared to available data on regional populations and referrals to genetics. The distribution was as follows: 59% Caucasian, 33.3% African American, 5.1% Asian American, and 2.6% Pacific Islander (see Figure 2).
Twelve patients with overlapping AADCd phenotypes were identified by the APSS as “at-risk” or “possible risk”, but determined by the reviewing genetic counselor to already have a genetically confirmed condition and were removed from the lists. These “positive controls” who met phenotypic criteria included patients with: CHD4 neurodevelopmental disorder, GRIN2B-related neurodevelopmental disorder, CASK-related disorder, SYNGAP1-related intellectual disability, Multiple congenital anomalies-hypotonia-seizures syndrome, Wiedemann–Steiner syndrome, NKX2-1-related disorders, SCN2A-developmental and epileptic encephalopathy, DYRK1A syndrome, Alpha thalassemia X-linked intellectual disability, Allan–Herndon–Dudley syndrome, and Bohring Opitz syndrome. All of these diagnoses would have been appropriate referrals for genetics and testing if they had not already been diagnosed.
Discussions / Conclusions: One of the biggest challenges in identifying patients with rare genetic conditions is ensuring that healthcare providers consider the diseases in their differential diagnosis. This data supports the hypothesis that the development, validation, and implementation of APSS can aid providers in identifying patients at risk for rare genetic conditions like AADCd.
In addition, data review at the two sites also showed that the racial/ethnic background of patients at risk identified by the APSS had a higher percentage of non-white patients than the standard referral population to the large academic center over the past year. These numbers more closely reflect the racial distribution of individuals in Georgia overall (see Figure 2). The use of an APSS may remove a layer of provider referral bias related to racial/ethnic background.
Limitations and Future Directions: All patients presenting at participating facilities during the study time period had equal opportunity to be included in the analysis. However, the data available for review was limited by what is available in the EHR. Some “possibly at-risk” subjects had limited unstructured data for review. Data used for this study is representative of the area covered and may not generalize to other geographic locations. Follow-up studies based on this pilot should explore larger data sets in other geographic regions or population datasets.
Acknowledgments: This work was funded by PTC Therapeutics. We would also like to thank our colleagues at Guardian Research Network, Emory University, the Children’s Hospital of Atlanta, consulting experts in the field, our statistician colleagues, and our dedicated technical development team for their contributions to this project.
- Himmelreich, N., Montioli, R., Bertoldi, M., Carducci, C., Leuzzi, V., Gemperle, C., … & Blau, N. (2019). Aromatic amino acid decarboxylase deficiency: Molecular and metabolic basis and therapeutic outlook. Molecular genetics and metabolism, 127(1), 12-22.
- Wassenberg, T., Molero-Luis, M., Jeltsch, K., Hoffmann, G. F., Assmann, B., Blau, N., … & Opladen, T. (2017). Consensus guideline for the diagnosis and treatment of aromatic l-amino acid decarboxylase (AADC) deficiency. Orphanet journal of rare diseases, 12(1), 1-21.
|Event:||2023 - National Society of Genetic Counseling Annual Conference|
|Author(s):||Dawn Laney1, Jessica Dronen2, Ryan Miller3, Jeff Kopesky3, Nadia Ali1, Ami Rosen1, Levi Thompson1, Andy King1, Jonathan Beus4, Rob Hawthorne1, Rossana Sanchez Russo1, Michael Gambello1|
|Affiliation(s):||1 Emory School of Medicine; 2 ThinkGenetic Inc; 3 PTC Therapeutics; 4 Children’s Healthcare of Atlanta|