Objectives: To assess the influence of conventional glass ionomer cement (GIC) vs resin-modified GIC (RMGIC) as a base material for novel, super-closed sandwich restorations (SCSR) and its effect on shrinkage-induced crack propensity and in vitro accelerated fatigue resistance. Methods: A standardized MOD slottype tooth preparation was applied to 30 extracted maxillary molars (5 mm depth/5 mm buccolingual width). A modified sandwich restoration was used, in which the enamel/dentin bonding agent was applied first (Optibond FL, Kerr), followed by a Ketac Molar (3M ESPE)(group KM, n = 15) or Fuji II LC (GC) (group FJ, n = 15) base, leaving 2 mm for composite resin material (Miris 2, Coltène-Whaledent). Shrinkageinduced enamel cracks were tracked with photography and transillumination. Samples were loaded until fracture or to a maximum of 185,000 cycles under isometric chewing (5 H z), starting with a load of 200 N (5,000 X), followed by stages of 400, 600, 800, 1,000, 1,200, and 1,400 N at a maximum of 30,000 X each. Groups were compared using the life table survival analysis (α = .008, Bonferroni method). Results: Group FJ showed the highest survival rate (40% intact specimens) but did not differ from group KM (20%) or traditional direct restorations (13%, previous data). SCSR generated less shrinkage-induced cracks. Most failures were re-restorable (above the cementoenamel junction [CEJ]). Conclusions: Inclusion of GIC/RMGIC bases under large direct SCSRs does not affect their fatigue strength but tends to decrease the shrinkage-induced crack propensity. Clinical significance: The use of GIC/ RMGIC bases and the SCSR is an easy way to minimize polymerization shrinkage stress in large MOD defects without weakening the restoration.