Purpose: The nature of end-of-life (EoL) processes is highly uncertain for constructions built today. This uncertainty is often neglected in life cycle assessments (LCAs) of construction materials. This paper tests how EoL assumptions influence LCA comparisons of two alternative roof construction elements: glue-laminated wooden beams and steel frames. The assumptions tested include the type of technology and the use of attributional or consequential modelling approaches. Methods: The study covers impact categories often considered in the construction industry: total and non-renewable primary energy demand, water depletion, global warming, eutrophication and photo-chemical oxidant creation. The following elements of the EoL processes are tested: energy source used in demolition, fuel type used for transportation to the disposal site, means of disposal and method for handling allocation problems of the EoL modelling. Two assumptions regarding technology development are tested: no development from today's technologies and that today's low-impact technologies have become representative for the average future technologies. For allocating environmental impacts of the waste handling to by-products (heat or recycled material), an attributional cut-off approach is compared with a consequential substitution approach. A scenario excluding all EoL processes is also considered. Results and discussion: In all comparable scenarios, glulam beams have clear environmental benefits compared to steel frames, except for in a scenario in which steel frames are recycled and today's average steel production is substituted, in which impacts are similar. The choice of methodological approach (attributional, consequential or fully disregarding EoL processes) does not seem to influence the relative performance of the compared construction elements. In absolute terms, four factors are shown to be critical for the results: whether EoL phases are considered at all, whether recycling or incineration is assumed in the disposal of glulam beams, whether a consequential or attributional approach is used in modelling the disposal processes and whether today's average technology or a low-impact technology is assumed for the substituted technology. Conclusions: The results suggest that EoL assumptions can be highly important for LCA comparisons of construction materials, particularly in absolute terms. Therefore, we recommend that EoL uncertainties are taken into consideration in any LCA of long-lived products. For the studied product type, LCA practitioners should particularly consider EoL assumptions regarding the means of disposal, the expected technology development of disposal processes and any substituted technology and the choice between attributional and consequential approaches.