You have to analyze the pattern of missing. Missing arranged randomly in a data -missing MCAR (missing values completely at random) are no problematic.
Missing arranged in a non-randomly way are problematic because it affects the generalizability of the results.