How to Assess Variation in Homeopathic Prognostic Factor Research?

We need to classify the outcome of prognostic factor research (PFR), especially regarding polar symptoms (PS) - symptoms with opposite values such as amelioration/desire and aggravation/aversion. For instance, in a data collection project 22.9% of the patients responding well to <i>Arsenicum album (Ars)</i> had 'Desire salt' and 5.7% 'Aversion salt'. Can such differences be explained by statistical variation?Frequency distributions of PS were analysed and compared with previous research to reveal differences. Cumulative binomial probability (CBP) and 95% confidence intervals (95% CIs) were calculated to assess the influence of statistical variation on the difference between the medicine population and the remainder of the population and the difference between opposite poles. CBP and 95% CI were compared regarding usefulness for daily practice. 95% CIs were used to calibrate the CBP. Corroboration by comparable symptoms was also used to validate outcomes.In several PS, there was asymmetry between opposite poles and a difference compared with previous research. The most probable cause was using questionnaires, disregarding clinical expertise. This results in asymmetrical frequency distributions when symptoms are common and the criterion 'more than average' was ignored. This, in turn, results in relatively low likelihood ratios (LRs) caused by a 'ceiling effect'.The CBP correlates with 95% CI, indicates the amount of overlap of 95% CIs, and is useful to classify the statistical certainty of PFR outcome. Based on CBP and difference of CBP for opposite symptoms, LR outcome was classified as statistically 'Certain', 'Probable', 'Possible' or 'Questionable'. Cut-offs between classes were based on expert estimates. Part of the outcome could be corroborated by the outcome of similar or opposite symptoms.Asymmetry of symptom frequency distributions in PS can be caused by practitioners not using their expert knowledge while assessing symptoms. A classification of reliability of data based on cumulative binomial chance is more informative and is better understood by experts in homeopathy. Nevertheless, classification of reliability remains partly subjective. Corroboration of outcome and clinical judgment are indispensable for estimating clinical validity of PFR outcomes. Practitioners and researchers participating in PFR need training in statistics and homeopathy respectively.