It depends on the reason why you have 55% non-response rate. If the reason for non-responsiveness is related to the subject under investigation, then definitely YES. If the missingness happened at random, e.g., software malfunction, one may argue that this will not affect the results much, other than the obvious large power loss. However, even in such a case one should be very cautious, as depending on the specific study design, seemingly random missingness can still heavily bias the results.

Thanks