Background and purpose — Recurrent dislocation is the commonest cause of early revision of a total hip arthropasty (THA). We examined the effect of femoral head size and surgical approach on revision rate for dislocation, and for other reasons, after total hip arthroplasty (THA). Patients and methods — We analyzed data on 166,231 primary THAs and 3,754 subsequent revision THAs performed between 2007 and 2015, registered in the Dutch Arthroplasty Register (LROI). Revision rate for dislocation, and for all other causes, were calculated by competing-risk analysis at 6-year follow-up. Multivariable Cox proportional hazard regression ratios (HRs) were used for comparisons. Results — Posterolateral approach was associated with higher dislocation revision risk (HR =1) than straight lateral, anterolateral, and anterior approaches (HR =0.5–0.6). However, the risk of revision for all other reasons (especially stem loosening) was higher with anterior and anterolateral approaches (HR =1.2) and lowest with posterolateral approach (HR =1). For all approaches, 32-mm heads reduced the risk of revision for dislocation compared to 22- to 28-mm heads (HR =1 and 1.6, respectively), while the risk of revision for other causes remained unchanged. 36-mm heads increasingly reduced the risk of revision for dislocation but only with the posterolateral approach (HR =0.6), while the risk of revision for other reasons was unchanged. With the anterior approach, 36-mm heads increased the risk of revision for other reasons (HR =1.5). Interpretation — Compared to the posterolateral approach, direct anterior and anterolateral approaches reduce the risk of revision for dislocation, but at the cost of more stem revisions and other revisions. For all approaches, there is benefit in using 32-mm heads instead of 22- to 28-mm heads. For the posterolateral approach, 36-mm heads can safely further reduce the risk of revision for dislocation.