Repeated measures ANOVA is now considered absolute by most statisticians. Look at generalized least squares and mixed effects models. If the missing late time points are missing at random, i.e., conditional on baseline covariates and previous follow-up measurements they are missing completely at random, then you can use a “use all available data” longitudinal model such as the two I listed above. It’s best to use a continuous-time correlation pattern and to use the exact measurement dates (or more accurately the number of days since baseline) in the analysis. Model time with a smooth nonlinear function. The smooth analysis will interpolate to “fill in the blanks”.