OBJECTIVES:  Medication adherence is an important concern for both physicians and payers. Non-adherence can lead to treatment failure and death, as well as an increase in the use of costly resources. Available literature regarding adherence tends to focus on the effect of patient attributes on therapy usage. However, it is likely that attributes specific to the therapy itself (such as adverse events) also have an effect on whether patients are adherent. The ability to predict the adherence rates for a therapy based on its characteristics, particularly in comparison to those of comparator treatments, could add significant value to a product’s profile. The study herein was performed as a feasibility analysis to determine if it is possible to predict patient adherence based on a treatment’s profile. The analysis focuses on type 2 diabetes treatments. METHODS:  Relationships between individual product characteristics and product discontinuation rates were determined. Product characteristics (i.e., adverse event and success rates) were obtained from clinical trials and real-world discontinuation data were obtained from a physician survey. The products included were DPP4s, GLP-1s, basal insulins, and OADs and the attributes examined were nausea/vomiting, other GI effects, hypoglycemia, weight gain, and lack of success/efficacy. From these relationships, a predictive model was derived that would allow the user to estimate a therapy’s adherence based on its profile. RESULTS:  The analysis found a clear correlation between weight change and discontinuation. The piecewise prediction model is as follows: Discontinuation = 0.2453*HbA1c reduction + 0.0144*weight change + 0.5090*proportion of patients with hypoglycemia + 0.0935*proportion of patients with GI side effects + 0.2368*proportion of patients with nausea/vomiting + 0.3943. CONCLUSIONS:  Despite analysis limitations, study trends suggest that it would be feasible to construct a model predicting adherence based on a treatment’s profile. Due to the link between adherence and outcomes, this topic warrants further research.