Karen mentions

"One of the other substudies (substudy of 015) which was over twice the size of 013 found no such result and this substudy also didnt have the discrepancies in the baseline data."

This is an important concept to keep in mind both for internet posts by the statistically uneducated, as well as basic research published in peer-reviewed journals. Small sample sizes are kind of like political polls - your sample MAY represent the population as a whole, but then again, it may not. How you take your sample, as well as it's size, is critical to the validity of the study to the general population, and in fact its own applied mathematical discipline.

For example, the recent data on matted lymph nodes was pretty scary (or encouraging, if you knew your/your loved one's nodes weren't matted). I'm glad it was posted on this site (although I'd seen it before) - but it's a small sample size. The authors of this study then went on to propose an alternate staging system based on HPV status, and whether or not the lymph nodes are matted. IMO, that is theorizing in advance of the data, but I suppose you have to float these ideas.