Confidence intervals for the risk ratio in cohort studies under inverse sampling