Background and Purpose— The DIAS-2 study was the only large, randomized, intravenous, thrombolytic trial that selected patients based on the presence of ischemic penumbra. However, DIAS-2 did not confirm the positive findings of the smaller DEDAS and DIAS trials, which also used penumbral selection. Therefore, a reevaluation of the penumbra selection strategy is warranted. Methods— In post hoc analyses we assessed the relationships of magnetic resonance imaging–measured lesion volumes with clinical measures in DIAS-2, and the relationships of the presence and size of the diffusion-perfusion mismatch with the clinical effect of desmoteplase in DIAS-2 and in pooled data from DIAS, DEDAS, and DIAS-2. Results— In DIAS-2, lesion volumes correlated with National Institutes of Health Stroke Scale (NIHSS) at both baseline and final time points ( P <0.0001), and lesion growth was inversely related to good clinical outcome ( P =0.004). In the pooled analysis, desmoteplase was associated with 47% clinical response rate (n=143) vs 34% in placebo (n=73; P =0.08). For both the pooled sample and for DIAS-2, increasing the minimum baseline mismatch volume (MMV) for inclusion increased the desmoteplase effect size. The odds ratio for good clinical response between desmoteplase and placebo treatment was 2.83 (95% confidence interval, 1.16–6.94; P =0.023) for MMV >60 mL. Increasing the minimum NIHSS score for inclusion did not affect treatment effect size. Conclusions— Pooled across all desmoteplase trials, desmoteplase appears beneficial in patients with large MMV and ineffective in patients with small MMV. These results support a modified diffusion-perfusion mismatch hypothesis for patient selection in later time-window thrombolytic trials. Clinical Trial Registration— URL: http://www.clinicaltrials.gov . Unique Identifiers: NCT00638781, NCT00638248, NCT00111852.