diff --git a/src/calibre/ebooks/html.py b/src/calibre/ebooks/html.py
index 1c15973d3b..1f1e6b94b1 100644
--- a/src/calibre/ebooks/html.py
+++ b/src/calibre/ebooks/html.py
@@ -331,7 +331,8 @@ class PreProcessor(object):
                   # Convert all entities, since lxml doesn't handle them well
                   (re.compile(r'&(\S+?);'), convert_entities),
                   # Remove the <![if/endif tags inserted by everybody's darling, MS Word
-                  (re.compile(r'(?i)<{0,1}!\[(end){0,1}if[^>]*>'), lambda match: ''),
+                  (re.compile(r'</{0,1}!\[(end){0,1}if\]{0,1}>', re.IGNORECASE), 
+                   lambda match: ''),
                   ]
                      
     # Fix pdftohtml markup