Background: CT perfusion (CTP) and diffusion or perfusion MRI might assist patient selection for endovascular thrombectomy. We aimed to establish whether imaging assessments of irreversibly injured ischaemic core and potentially salvageable penumbra volumes were associated with functional outcome and whether they interacted with the treatment effect of endovascular thrombectomy on functional outcome. Methods: In this systematic review and meta-analysis, the HERMES collaboration pooled patient-level data from all randomised controlled trials that compared endovascular thrombectomy (predominantly using stent retrievers) with standard medical therapy in patients with anterior circulation ischaemic stroke, published in PubMed from Jan 1, 2010, to May 31, 2017. The primary endpoint was functional outcome, assessed by the modified Rankin Scale (mRS) at 90 days after stroke. Ischaemic core was estimated, before treatment with either endovascular thrombectomy or standard medical therapy, by CTP as relative cerebral blood flow less than 30% of normal brain blood flow or by MRI as an apparent diffusion coefficient less than 620 μm 2 /s. Critically hypoperfused tissue was estimated as the volume of tissue with a CTP time to maximum longer than 6 s. Mismatch volume (ie, the estimated penumbral volume) was calculated as critically hypoperfused tissue volume minus ischaemic core volume. The association of ischaemic core and penumbral volumes with 90-day mRS score was analysed with multivariable logistic regression (functional independence, defined as mRS score 0–2) and ordinal logistic regression (functional improvement by at least one mRS category) in all patients and in a subset of those with more than 50% endovascular reperfusion, adjusted for baseline prognostic variables. The meta-analysis was prospectively designed by the HERMES executive committee, but not registered. Findings: We identified seven studies with 1764 patients, all of which were included in the meta-analysis. CTP was available and assessable for 591 (34%) patients and diffusion MRI for 309 (18%) patients. Functional independence was worse in patients who had CTP versus those who had diffusion MRI, after adjustment for ischaemic core volume (odds ratio [OR] 0·47 [95% CI 0·30–0·72], p=0·0007), so the imaging modalities were not pooled. Increasing ischaemic core volume was associated with reduced likelihood of functional independence (CTP OR 0·77 [0·69–0·86] per 10 mL, p interaction =0·29; diffusion MRI OR 0·87 [0·81–0·94] per 10 mL, p interaction =0·94). Mismatch volume, examined only in the CTP group because of the small numbers of patients who had perfusion MRI, was not associated with either functional independence or functional improvement. In patients with CTP with more than 50% endovascular reperfusion (n=186), age, ischaemic core volume, and imaging-to-reperfusion time were independently associated with functional improvement. Risk of bias between studies was generally low. Interpretation: Estimated ischaemic core volume was independently associated with functional independence and functional improvement but did not modify the treatment benefit of endovascular thrombectomy over standard medical therapy for improved functional outcome. Combining ischaemic core volume with age and expected imaging-to-reperfusion time will improve assessment of prognosis and might inform endovascular thrombectomy treatment decisions. Funding: Medtronic.