Net benefit approaches to the evaluation of prediction models, molecular markers, and diagnostic testsBMJ 2016; 352 doi: https://doi.org/10.1136/bmj.i6 (Published 25 January 2016) Cite this as: BMJ 2016;352:i6
All rapid responses
References available on request.
The threshold might be better called an acceptable harm/benefit threshold; since the term : 'net benefit' already rhetorically loads attitudes to the test in favour of its use. It is however a good idea to make the trade off and the possible decision curves explicit in test research reporting. As a 'statistic' it still potentially misleads as it doesn't tell you about (and even implies it doesn't matter) the incremental value/harm of the 'new' test compared to existing care; and it also masks the opportunity cost of doing a new test that may be expensive and involve foregoing other aspects of healthcare, eg where there is a fixed healthcare budget. Clinical utilisation may also be profoundly influenced by the way tests results are presented and the thresholds for outcomes such as probability of outcome are chosen by test owners and researchers.
Some governments already preempt clinical decisions on harm/benefit ratio by the process of approving new technology for public expenditure by agreeing expenditure that has net population harms (measured by QALYs) because the opportunity costs are greater than the gains from the tests. NICE in the UK for example jusifities this by saying they must pay more to support innovative research and because the public likes innovation. The NICE financial threshold for approving new products is greater than the cost to save a QALY with existing healthcare so legitimises a hidden sacrifice of life in order to preserve the market in new technologies.
The major problem with the 'net benefit' statistic is the way it is being measured using expert judgements about population trade offs. There is a danger that experts in a specialist silo, are more likely than not to be 'their chosen pathology' risk averse, innovative technology friendly, and research publication enthusiastic. Experts are also (especially under extreme neoliberal regimes such as the USA) commonly constrained by contracts to follow clinical pathways sold to their employers under contracts with insurance companies who determine and authorise when (expensive) tests will be used.
The 'net benefit' approach may also be shutting the stable door after the horse has bolted since there is also the diagnostic research industry 'ethical' normative standard that demands a sensitivity of 90% (and never mind the specificity) and 'astutely' constructs artificial binary outcomes e.g. tests to predict cancer recurrence, that preempts any clinical attempts to judge harm/benefit ratios, since the test result is presented in terms of e.g. 'eligible' for further treatment and 'directs' decisions positively (ie is normativising to use a biopolitical term).
The arch psychological trap set for the patient is that as an individual you can never know if you might be the one saved or sacrificed which leaves you vulnerable to the fantasy that you are the one 'chosen to be saved' and therefore vulnerable to being objectivised by and subjugated to medicine's neoliberal sovereignty.
Competing interests: No competing interests