PT - JOURNAL ARTICLE AU - Dowie, Jack AU - Kaltoft, Mette Kjer AU - Rajput, Vije Kumar TI - 48 Value-based medicine can eliminate over- (and under-) diagnosis AID - 10.1136/EBM-2023-POD.48 DP - 2023 Sep 01 TA - BMJ Evidence-Based Medicine PG - A24--A24 VI - 28 IP - Suppl 1 4099 - http://ebm.bmj.com/content/28/Suppl_1/A24.1.short 4100 - http://ebm.bmj.com/content/28/Suppl_1/A24.1.full SO - BMJ EBM2023 Sep 01; 28 AB - In common with most discussions of medical screening and testing, the Overdiagnosis literature is reluctant to accept the implication of Hammond’s brilliant summary of the situation facing healthcare decision makers at all levels, from individual to global: Irreducible Uncertainty, Inevitable Error, Unavoidable Injustice. This reluctance manifests itself in interpreting uncertainty, error, and injustice as constituting problems to be ‘addressed’, rather than as requiring, at all points in time and whatever the current state of information, decisions to be made. Their formulation as problems reflects the priority ordering of Alternatives-Based Medicine (ABM), with its stress on evidence about options, in contrast to the reverse sequencing in Values-Based Medicine (VBM), with its emphasis on preferences over outcomes. VBM is unattractive because it starts out with the uncomfortable task of establishing what preferences regarding outcome trade-offs are to be applied in the decision - usually reducing to whose preferences – before any consideration of the performance of adoptable alternatives is undertaken. ABM, focusing on the predictive power of informational models of competing options, is not only more scientifically engaging, but means the unrewarding task of processing preferences in a population with heterogeneous values can be downgraded or postponed. A tiny illustration. In a recent exchange we find Horton, et al. disagreeing with Green et al.’s view on the value of the latter’s genetic risk model for prostate cancer: ‘... such genetic risk scores are only weakly predictive and using them prematurely in clinic may create more problems than it solves... With the model set to identify men at 3.7% or higher chance of being diagnosed with prostate cancer within two years (the point at which, from a statistical perspective, the model achieved the optimal balance between sensitivity and specificity), this achieved a detection rate of 71%...’. Only later, ‘We argue that the problem of false positives here deserves attention...’. In their rejoinder, Green, et al. concede ‘The problem of false positives does deserve attention. The specificity of PSA is estimated at around 0.20; only one in five men who do not have prostate cancer will have a negative PSA test’ and ‘careful consideration of the appropriate threshold for [any] risk model is needed’. But this comes afterthey have gone into the Areas Under the Receiver Operating Curves, i.e., the discriminatory power of the competing tests. However, the appropriate threshold Number Needed to Alarm/Number Needed to Reassure ratio, can, and will in VBM, be established without regard to the discriminatory or predictive power of a test. There is no ‘optimal balance’ of Sensitivity and Specificity other than that reflecting the preferences of the ‘optimal’ preferer, which should be established independently of the test characteristics. Many papers in the literature on OverDiagnosis consciously reflect the doxa of population medicine, which conflicts with key principles of individualised and personalised healthcare. But in the many that want to adhere to these, ‘protecting’ the person from having to decide upfront on their personal NNA/NNR threshold – thereby generating the possibility of Over- and Under-Diagnosis - is still endemic.