BACKGROUND: Observational studies of diphtheria-tetanus-pertussis (DTP) vaccine from longitudinal study sites have reported divergent effects on child survival, ranging from 10-fold reduction to threefold increased mortality. None of these studies had complete information on DTP vaccinations from both survivors and children who died. We reviewed the data analysis methodology to assess whether methodological differences could explain the divergent results. DESIGN: Studies have used case-control design, survival analysis with interval-fixed vaccination status (landmark approach), and survival analysis with retrospective updating of vaccination status. RESULTS: Seven studies using a case-control design or a landmark approach found a negative effect of DTP on child survival. Eight of nine survival analyses with retrospective updating of vaccination status reported a beneficial effect. This beneficial effect of DTP increased with the length of the interval between data collection visits. Studies with long interval between visits had very high mortality rates among unvaccinated children, low mortality rate ratios for vaccinated compared with unvaccinated children, and strongly beneficial estimates of DTP. CONCLUSION: The divergent results in observational studies of the impact of DTP on child survival are partly because of methodological differences. To assess the impact on mortality of routine vaccinations, observational study designs which minimize the effect of bias are warranted. Randomized trials should be considered.