Blinded sample size recalculation in adaptive enrichment designs