Instrumental variable (IV) approach is popular in estimating the average treatment effect (ATE) in the presence of unmeasured confounders. While methods for estimating ATE with binary IVs are well-established, challenges arise when the IV is continuous, as is often the case in genetic epidemiology with polygenic risk scores. Current approaches dealing with continuous IV rely on structural equation modeling, yielding biased estimates when the outcomes are binary. In this work, we developed four novel estimators for ATE using continuous IV under the potential outcome framework. Of these, three estimators are consistent under different observed data models and one estimator is triply robust, that is, consistent in the union of these observed data. Simulation results showed that our proposed estimator is unbiased and robust to model misspecification. We further illustrate our approaches to estimate the causal effect of obesity in lung cancer patients’ two-year mortality rate.