New drugs: modern diuretic treatment.

When the humble author submits his paper to a learned journal I suppose one of his main fears is that the editor's fearsome statistician will immediately devour and discard months of hard work on a subject which the author thought he knew as much about as anyone. How thoroughly unfair that someone who knows nothing about the subject can wield such power. The fear is real because the lowly researcher, starved of funds and miles from a friendly statistician, knows that editors have become increasingly strict and look on perfection in this respect as the hallmark of a high scientific standard. I do not quibble with this admirable objective, except to make one point. In my occasional role as a referee I would say it is not difficult to detect statistical weakness in a paper. What is more important, because it is more difficult to detect, is weakness in the underlying physiology, or what ever the science may be, and the trouble is that an impeccable clinical trial can be an extremely effective smokescreen for thoroughly bad physiology. Conversely, the best physiology (with exceptions) does not need refined statistical analysis to convince one of its veracity. No harm will come from good physiology clothed in shaky statistics, but errors may lie hidden for years behind a sound statistical analysis. The moral for the author who suspects his physiology is weak and is determined to pierce the editor's defences is to make sure that his statistics are perfect. How is the poor editor, defending himself against several hundred such authors each week, to detect scientific "codswallop" clothed in beautiful statistics? Only by getting a physiologist to catch a physiologist. That is usually just what the editor does, but I continue to be surprised at the poor scientific quality of many papers. The trouble is that whereas one medical statistician can cover all the clinical sciences, the scientific material can be assessed only by someone working in the same discipline, and even then refereeing error may occur in both directions. Bad physiology may go undetected, and good physiology may end up in the editor's wastepaper basket.