Exposures of low-grade metabasalts and ophicarbonates in the Northern Apennines, and their high- and ultrahigh-pressure metamorphic equivalents in the Western and Ligurian Alps and Tianshan (representing an overall peak P–T range of ~ 0.2–3.0 GPa, 200–610 °C), allow investigation of the effects of prograde metamorphic devolatilization, and other fluid–rock interactions, on degrees of retention and isotopic evolution of C in subducting oceanic crust and associated mantle rocks. Such work can inform models of C cycling at convergent margins, helping to constrain the efficiency of return of initially subducted C via arc volcanism and the fraction of this subducted C entering the deeper mantle beyond arcs. In the metabasaltic rocks, the preservation of finely disseminated carbonate with δ13C overlapping that of seafloor-altered protoliths, and the minimal mineralogical evidence of decarbonation, indicates large degrees of carbonate retention in this suite extending to UHP conditions similar to those beneath modern volcanic fronts. For many of the metabasalts, the δ18O of this carbonate can be explained by closed-system equilibration with silicate phases (e.g., garnet, clinopyroxene) during HP/UHP metamorphism. Larger volumes of carbonate preserved in interpillow regions and as breccia-filling largely escaped decarbonation, showing little or no evidence for reaction with adjacent metabasalt. Calculated devolatilization histories demonstrate that, in a closed-system model, carbonate in metabasaltic rocks can largely be preserved to depths approaching those beneath volcanic fronts (80–90 km). Modeling of open-system behavior indicates that episodic infiltration of such rocks by H2O-rich fluids would have greatly enhanced decarbonation. Trends in O–C isotope composition of carbonate in some metabasaltic suites likely reflect effects of infiltration by externally-derived fluid with or without resulting decarbonation. Most carbonated ultramafic rocks similarly show little mineralogical evidence for decarbonation, consistent with calculated reaction histories, and have δ13C largely overlapping that of seafloor equivalents. However, the high-grade ophicarbonates show more restricted ranges in δ18O consistent with some control by infiltrating fluids, likely during subduction. This combination of field, petrographic, and isotopic evidence, together with calculated decarbonation histories, is consistent with minimal loss of CO2 from these rocks via decarbonation during forearc metamorphism. Combining our results with those of Cook-Kollars et al. (2014; Chemical Geology) for associated W. Alps metasedimentary rocks, we suggest that the majority of the CO2 (perhaps 80–90%, considering the full range of rock types) could be retained through forearcs in more intact volumes of subducting sediment, basalt, and ophicarbonate experiencing closed- or limited open-system conditions. Deep in forearcs and beneath arcs, decarbonation (and also carbonate dissolution) could be enhanced in shear zones and highly fractured volumes experiencing larger fluid flux in part from dehydrating sub-crustal ultramafic rocks in slabs. Degrees of C loss by decarbonation, carbonate dissolution, and partial melting should be particularly significant as the subducting sections experience heating to > 600 °C at depths of 80–120 km (i.e., approximately at depths beneath arcs).