Estimating marginal causal effects in a secondary analysis of case‐control data