Global archive of dissolved ocean profile data is characterized by a high degree of heterogeneity with data obtained both by the Winkler chemical titration method and by different electrochemical and optical sensors. Respectively, the accurate assessment of the secular change of the dissolved ocean oxygen content critically depends both on the quality control of oxygen data and the estimation of the sensor data bias relative to the reference Winkler method data. Based on the earlier studies on data quality control within the International Quality-Controlled Ocean Database (IQuOD) initiative we developed the quality control procedure for ocean oxygen profile data and applied it to the global archive of oxygen profile data available through the World Ocean Database. Analysis of data outliers reveals significant differences in data quality between the Winkler method and sensor data with data from CTD profiles showing higher outlier percentages compared to the ship-based Winkler method data and data from Argo floats. The co-located quality controlled oxygen profiles were used to estimate offsets between the sensor and Winkler method data. We find significant  time-varying negative overall oxygen offset for Argo oxygen profiles, with mean offset decreasing from about -10-12 mkmol/kg before 2010  to -5 mkmol/kg after about 2015.  Using un-biased Argo oxygen data would result in the apparently stronger deoxygenation of the global ocean.