In bioequivalence, biosimilar and non-inferiority (NI) studies, intent-to-treat (ITT) analysis tends to make the two treatments look similar, thereby is generally considered anti-conservative. The Per Protocol (PP) analysis based on completers and compliers is more able to reflect treatment differences and is usually preferred in equivalence or NI assessments. However, missing data and non-compliance are post-randomization intercurrent events and may introduce selection bias to the study conclusion. Therefore, the usually used naïve PP estimand may not be causal. In this work, we propose a principal stratification causal framework and causal estimands to test equivalence, which was one of the five approaches recommended by the ICH E9 (R1) addendum to address intercurrent events. We also propose a tipping point sensitivity analysis to evaluate the robustness of the current PP estimator in testing equivalence when the sensitivity parameters deviate from the three identified conditions, but stay within a clinically meaningful range. Our work can be applied to comparative biosimilar clinical trials and non-inferiority trials.