You can publish. Not received magic p <0.05, as it does not matter. The options are: well, there is a clear trend (see our group remission rate 64.5%, and in the historical control of 33%) - treatment is clearly effective. 'll Get more patients, and all will turn out. But these patients did not respond to our wonderful new treatment, clearly some are not (too sick, overly healed, etc.) - let's eliminate them from the analysis, and anything that exclusion criteria have appeared after the completion of a detailed study. Still the only option: No, this treatment is inefficient (in fact failed to get statistically significant differences on a group of 18 patients). The most popular option in the thesis: let's make "deeper analysis" (divided into subgroups, we'll see, and determine if it is, this and this). This approach is more welcome, of course, if one of the many additional tests can be obtained cherished p <0,05. A detailed study is published, what next? And then we usually wait for a new international conference. With statistically significant p <0.0000001'm sure many of you are familiar with this pattern (the author has a fairly large sample of own and foreign observations ...  Read the rest of the entry...