The First Shared Task on Extrinsic Parser Evaluation (EPE 2017) compares different dependency representations by evaluating their impact on downstream applications that utilize these parses for other text mining tasks. In the Biomedical Event Extraction downstream task parses are evaluated by using the Turku Event Extraction System (TEES) with the BioNLP'09 Shared Task as the model challenge. The participants parse the BioNLP'09 dataset, after which the TEES system is run, using the parses as features for predicting events on which the parses are compared. Eight teams submitted a total of 44 runs generated with various parsers, and an additional 13 runs were produced with parsers available via the TEES preprocessing system. Although the TEES system has been developed and optimized using the Stanford Dependencies parsing scheme, among the EPE submissions good performance on the TEES system was achieved also with the Universal Dependencies version 1 scheme.