This article presents a synthetic study of 31 epidemiological panel investigations (published between 1987 and 1998) analyzing how air pollution and health may be related. First, the methodology of these studies is described: choice of population, exposure and health assessment, confounding factors and statistical methods. The main results are presented as follows: for pulmonary performance, respiratory symptoms and medication use per pollutant (PM10 and particles, SO2, O3, NO2). Then the interactions among pollutants or implying other environmental factors are analyzed. Most of the time, an association with particles is found, no matter the health indicator considered. An O3 effect is more often observed with lung function tests than with respiratory symptoms or medication use, the latter being linked first to particles and secondly to SO2. A NO2 effect is more rarely shown. Pulmonary performances are affected by all pollutants. Respiratory symptoms related to air pollution are mostly low respiratory symptoms and cough, particularly with particles and O3. The paper ends by discussing the interest and limits of these studies. It also emphasizes the importance of exposure assessment. Indeed, in order to improve exposure assessment, an essential step in every epidemiological study, new tools need to be developed.