The constellation of Global Positioning System (GPS) satellites provides a source of continuous, phase-stable electromagnetic signals available for radio occultation observations of our planet. The atmospheric-induced bending of the transmitted rays observed during each occultation can be converted into a refractivity profile using an Abel transform. Since refractivity is related to temperature and humidity, it may potentially be used in global data assimilation for numerical weather prediction (NWP) and for creating climate data sets. We first compare GPS/Meteorology (GPS/MET) 1995 refractivity with various backgrounds and verify that the best expected background presents generally the best fit with the observed refractivity. We implement here an efficient one-dimensional variational (1DVAR) analysis of GPS refractivity that enables retrieving temperature, humidity, and sea-level pressure using the finite volume data assimilation system background. 1DVAR analyses with GPS/MET 1995 data are compared with collocated radiosondes. They show an excellent capacity of the GPS measurements to resolve the tropopause. In the Northern Hemisphere, we demonstrate a net reduction of temperature bias and standard deviation, as compared with the background. The 1DVAR humidity presents reduced standard deviation as compared to the background between 550 and 400 hPa. However, a refractivity bias between the observations and the background in the lower troposphere systematically shifts the 1DVAR humidity downward. A refractivity bias over the whole profile is transformed into a 1DVAR sea-level pressure bias. This study represents a step toward using the GPS radio occultation data in data assimilation systems to improve NWP forecasts and representation of Earth's climate in models.