With a prevalence of 1.5-2%, intellectual disability (ID) represents a major health problem. It is clinically extremely variable and can result from many different causes. However, it is assumed that genetic alterations are causative in most cases. Exome sequencing is with a diagnostic rate of up to 40% the most efficient test to diagnose patients with neurodevelopmental disorders (NDD). Given the constantly growing number of NDD associated genes, improvements in analysis tools and information on variants, regular re-analysis of negative or unclear cases is recommended. In a long-term study aiming at identifying the genetic basis of NDDs we performed exome sequencing in 204 individuals from 196 families presenting with NDDs who had negative test results including CMA. Filtering against known NDD genes (SysID database) followed by identification of de novo variants in all genes resulted in a diagnostic rate of 44.9% (88/196). Interestingly, the diagnostic yield in singletons was higher than in trios (51.1% (24/47) vs. 42.2% (62/147)). Two of the 88 cases were solved by a duo-analysis. We now re-analyzed 108 unsolved (105 sporadic) cases regarding homozygous, compound-heterozygous, X-linked, de novo and autosomal dominant variants in all genes, considering also an updated list of 1,164 NDD genes and 1,127 candidate genes. Two additional cases could be diagnosed: an affected boy inherited a stop mutation in SETD5 from his healthy mother, which previously was missed due to a unique filtering for de novo variants in a trio setting. Furthermore, a missense variant in GNB1 which was already classified as variant of uncertain significance (VUS) in the initial evaluation could now be classified as likely pathogenic based on new available data in a 43 years old patient. In addition, we identified the probable genetic cause of the NDD in three individuals and identified 20 candidate variants in 17 genes compatible with the clinical presentation, but a firm diagnosis could not be made due to lack of functional validation. The relatively low re-analysis detection rate is probably due to the high initial detection rate based on a combination of diagnostic and research analysis strategy.