The two key challenges of causal inference in observational studies are latent confounding and collider bias. Latent confounding occurs because of the failure to control unmeasured covariates that are common causes of treatments and outcomes, which is usually solved by Instrumental Variable (IV) approaches. Collider bias comes from non-random sample selection caused by both treatments and outcomes, which a different type of instrument, i.e., shadow variables, can address. However, in most scenarios, the two biases simultaneously exist in observational data, and the previous methods focusing on either prove inadequate. To the best of our knowledge, no approach has been developed for causal inference under both biases. In this paper, we propose a novel IV approach, Two-Stage Shadow Inclusion (2SSI), which can address latent confounding and collider bias simultaneously by utilizing the residuals of the treatments as conditional shadow variables. Extensive experimental results on benchmark synthetic datasets and a real-world dataset show that 2SSI achieves noticeable performance improvement under both latent confounding and collider bias compared to existing methods.