OBJECTIVES: To compared model fit and impact on key treatment effect estimates from incorporating the assumption of drug class effects in MTCs. METHODS: Two MTC data sets on early and advanced Parkinson’s disease (PD) comprising of several pharmacotherapies belonging to one of the three drug classes: dopamine receptor agonists (DOAs), monoamine-oxidase B inhibitors (MOAB-I), and adenosine A2 antagonists (A2A). We used three models: 1) a conventional random-effects MTC assuming all drugs were independent (model 1); 2) an MTC assuming all drug belonging to the same class exhibited equal effects (model 2); and 3) an MTC assuming some overall drug class effect for each drug class, but allowing for each drug to differ from their class’ overall effect (model 3). We analysed the effect of the included interventions (and drug classes) on improvement on the Unified Parkinson’s Disease Rating Scale (UPDRS) part II and III. We compared model fits (DIC) and the impact on treatment effects and 95% credible intervals from using these three models. RESULTS: For the early PD data set, the model 3 yielded the smallest DIC. Treatment effects from all models were similar, but 95% credible intervals were generally narrower with model 3 and generally the widest with model 1. For the advanced PD data set, model 2 yielded a slightly smaller DIC than model 3, and both smaller DIC than model 1. However, all treatment effects and 95% credible intervals were highly similar. CONCLUSIONS: Incorporating the drug class effect assumption in MTCs may provider better model fits, and thus, more reliable estimates of comparative efficacy and safety.