Are the empty lines hard coded? It doesn't look like it in your example but if there is...
Is there a <p><br /></p> or a <p> </p> or something like that??
If there is, you could use a "\s*" between the groups to find any space between. Something along these lines (assuming the blank lines are hard coded as "<p> </p>" :
find: <p> </p>\s*<p>\d+</p>\s*<p> </p>
replace: {nothing - empty}
That will find a blank line before, the line with the number, and a blank line after.
Cheers!