Abstract: | The aim of the present study is to evaluate drawbacks and advantagesof the choice of hospital versus population controls in a casecontrol study on diet and cancer through the analysis of a retrospectivestudy on diet and gastric cancer (GC) conducted in Forli, Italy,involving 232 cases, 430 population controls and 252 hospitalizedcontrols. The present paper reports the comparison of resultson diet and GC risk obtained using the 2 types of controls.Population controls tended, in general, to eat all kinds offoods slightly more frequently (bread, pasta, cold cuts, freshfish, seasoned cheeses, legumes, garlic, onions and preservedfruits), with the exception of cooked vegetables, which werereported less frequently by population than by hospital controls.ORs for specific foods adjusted for confounders and other foodswere consistent in the separate models including populationand hospital controls respectively for all food groups, withthe exception of cooked vegetables which represented a protectivefactor only when hospital controls were considered (high consumers,population controls: adjusted, OR=0.9, trend p value 0.54; highconsumers, hospital controls, adjusted OR=0.5, trend p value<0.01). Hospital controls were slightly less often currentsmokers (22.6 versus 30.0%) and more often regular wine drinkers(57.5 versus 47.8%) compared with population controls, but noneof these variables was associated with GC risk. The main resultsin this study were consistent using both types of controls,nevertheless the distribution of some dietary variables notrelated to the disease under study differed between the 2 controlgroups, suggesting some caution in the use of hospital controlswhen studying diseases other than GC. |