Increasing evidence indicates that patients with obsessive-compulsive disorder (OCD) exhibit alterations in fronto-striatal circuitry. Performance deficits in the antisaccade task would support this model, but results from previous small-scale studies have been inconclusive as either increased error rates, prolonged antisaccade latencies, both or neither have been reported in OCD patients. In order to address this issue, we investigated antisaccade performance in a large sample of OCD patients (n = 169) and matched control subjects (n = 183). As impaired antisaccade performance constitutes a potential endophenotype of OCD, unaffected first-degree relatives of OCD patients (n = 100) were assessed, as well. Furthermore, we conducted a quantitative meta-analysis to integrate our data with previous findings. In the empirical study, OCD patients exhibited significantly increased antisaccade latencies, intra-subject variability (ISV) of antisaccade latencies, and antisaccade error rates. The latter effect was driven by errors with express latency (80–130 ms), as patients did not differ significantly from controls with regards to regular errors (>130 ms). Notably, unaffected relatives of OCD patients showed elevated antisaccade express error rates and increased ISV of antisaccade latencies, as well. Antisaccade performance was not associated with state anxiety within groups. Among relatives, however, we observed a significant correlation between antisaccade error rate and harm avoidance. Medication status of OCD patients, symptom severity, depressive comorbidity, comorbid anxiety disorders and OCD symptom dimensions did not significantly affect antisaccade performance. Meta-analysis of 10 previous and the present empirical study yielded a medium-sized effect (SMD = 0.48, p < 0.001) for higher error rates in OCD patients, while the effect for latencies did not reach significance owing to strong heterogeneity (SMD = 0.51, p = 0.069). Our results support the assumption of impaired antisaccade performance in OCD, although effects sizes were only moderately large. Furthermore, we provide the first evidence that increased antisaccade express error rates and ISV of antisaccade latencies may constitute endophenotypes of OCD. Findings regarding these more detailed antisaccade parameters point to potentially underlying mechanisms, such as early pre-stimulus inhibition of the superior colliculus.