A screening programme for early gastric cancer was introduced in the state of Tachira, Venezuela, in 1980. Screening was performed by photofluorography, using two mobile units. The efficacy of this programme in reducing mortality from stomach cancer was evaluated by means of a case-control study. Cases were 241 individuals who died from stomach cancer in the period 1985-89. Ten live controls per case were drawn from the electoral rolls, matched by sex, age and residence. Exposure to the screening examination of cases and controls was assessed through individual linkage with the programme's centralised database. After the exclusion of examinations occurring within the 6 months preceding the case's diagnosis, the odds ratio (OR) of dying from stomach cancer for those screened was 1.26 (CI 0.83-1.91) and the OR in females was lower than in males: 0.77 (CI 0.33-1.78) and 1.52 (CI 0.94-2.47) respectively. Odds ratios associated with years since last test and number of tests did not differ significantly from 1. These results show the inefficacy of the programme in reducing mortality from gastric cancer in the area. In an attempt to determine whether this result was due to selection bias, an analysis restricted to subjects who had been screened at least once was performed. When examinations occurring after an index date at various intervals before the case's diagnosis were excluded, the screening test appeared to protect from death, although confidence intervals of the odds ratios are large, for example OR = 0.47 (CI 0.24-0.98) when excluding tests within 1 month.