We analyze a set of 20 tide gauge records covering the contiguous United States (US) coastline and the period from 1929 to 2013 to identify long-term trends and multidecadal variations in extreme sea levels (ESLs) relative to changes in mean sea level (MSL). Different data sampling and analysis techniques are applied to test the robustness of the results against the selected methodology. Significant but small long-term trends in ESLs above/below MSL are found at individual sites along most coastline stretches, but are mostly confined to the southeast coast and the winter season when storm surges are primarily driven by extratropical cyclones. We identify six regions with broadly coherent and considerable multidecadal ESL variations unrelated to MSL changes. Using a quasi-nonstationary extreme value analysis, we show that the latter would have caused variations in design relevant return water levels (50–200 year return periods) ranging from ∼10 cm to as much as 110 cm across the six regions. The results raise questions as to the applicability of the “MSL offset method,” assuming that ESL changes are primarily driven by changes in MSL without allowing for distinct long-term trends or low-frequency variations. Identifying the coherent multidecadal ESL variability is crucial in order to understand the physical driving factors. Ultimately, this information must be included into coastal design and adaptation processes.