As limited as Johnsyweb's solution but using a different approach and GNU awk's advanced feature allowing full RegEx record separators resulting in a piece of code that, IMHO, is very awk-ish:
awk -F\\\n '{NF=5}NR-1' RS="PATTERN[^\n]*." OFS=\\\n
So if you want a stable solution that is also readable, go along with Dennis Williamson's answer (+1 for that one) but maybe you too simply enjoy the beauty of awk when looking at lines like this.
