Developmental regression is a frequent and severe complication in the Phelan-McDermid Syndrome (PMS). Little is known about the causes and mechanisms of this condition, and no treatment is available to date. Finding medical conditions associated to the apparition of developmental regressions in PMS could help to understand the mechanisms of the regressions. Materials and methods: Using the Phelan-McDermid Data Network allowed us to analyze the Parent Reported Outcomes (PRO) from 233 patients and the full clinical notes history processed by natural language processing (NLP) of 78 patients with PMS, expressed as SNOMED CT codes. We analyzed the PRO with a Phenome-Wide Association Study (PheWAS). After ascertaining the quality of the NLP processing by comparing the codes to a set of manually annotated conditions, clinical notes were analyzed with a PheWAS and several machine learning modelisations: Elastic Net logistic regression, Random Forest, Gradient Boosting Machine. We attempted to aggregate the SNOMED CT by limiting the depth of the tree to reduce the number of features. We identified important variables to predict the presence of developmental regressions. We then manually reviewed the clinical notes and the PRO to fully analyze the available information for these features. Results: The PRO analysis did not yield significant features. In the clinical notes, seizure disorders were close to significance in all the PheWAS (FDR = 0.098). it was also consistently found as the most important factor in the machine learning modelisations. When reading of the clinical notes, 13/30 (43.3%) of the patients with regression also had epilepsy, and 12/48 (25.0%) without regression had epilepsy. Conclusion: This study found a consistent link between epileptic seizures and developmental regressions in PMS. Further studies should confirm it, investigate it’s nature and potential mechanisms. We did not identify other conditions associated to developmental regressions.