The unusually rich set of family and child-specific control variables included in the models reduced potential omitted variable bias, providing stronger evidence on whether the estimated effects of low birth weight and childhood socioeconomic disadvantage on disease onset in adulthood were causal.