What is the best method for long-term survival analysis?

In the Cox proportional hazards regression model, which is the most commonly used model in survival analysis, the effects of independent variables on survival may not be constant over time and proportionality cannot be achieved, especially when long-term follow-up is required. When this occurs, it would be better to use alternative methods that are more powerful for the evaluation of various effective independent variables, such as milestone survival analysis, restricted mean survival time analysis (RMST), area under the survival curve (AUSC) method, parametric accelerated failure time (AFT), machine learning, nomograms, and offset variable in logistic regression. The aim was to discuss the pros and cons of these methods, especially with respect to long-term follow-up survival studies.