Science in Medical Practice: Probability and Clinical Judgment Instead of Surrogate Certainty

Abstract

The outcome of most therapies in medicine is uncertain. Research based on hypothesis testing with a dichotomous outcome is therefore not very useful in daily practice. An increased probability of recovery of the individual patient is often the best possible outcome.

Attempts to prove that homeopathy is not placebo have shown to be unproductive, both in convincing the medical community because of an implausibility of biological action of high dilutions, and in improving homeopathy's effectiveness. Priority should instead be given to improving the homeopathic method itself. A large amount of long-standing knowledge must be confirmed, but confirmation bias must be avoided. Clinical practice requires probability statistics, Bayesian statistics, clinical judgment and handling of bias. Classification of statistical uncertainty based on cumulative binomial probability seems more appropriate than one cut-off value for statistical significance. Bias should be better understood in order to prevent and correct it. A large amount of clinical expertise is already available to improve our method.

Uncertainty cannot be ruled out in clinical practice and requires probability statistics. Probability and clinical judgment are realities that cannot be ignored. The reduction of bias requires a paradigmatic shift towards critical appraisal of research outcome and searching for independent data that do not necessarily confirm our findings. Such data could reveal bias and help to explain and reduce it.