Comparing risk prediction models
BMJ 2012; 344 doi: https://doi.org/10.1136/bmj.e3186 (Published 24 May 2012) Cite this as: BMJ 2012;344:e3186
All rapid responses
Rapid responses are electronic comments to the editor. They enable our users to debate issues raised in articles published on bmj.com. A rapid response is first posted online. If you need the URL (web address) of an individual response, simply click on the response headline and copy the URL from the browser window. A proportion of responses will, after editing, be published online and in the print journal as letters, which are indexed in PubMed. Rapid responses are not indexed in PubMed and they are not journal articles. The BMJ reserves the right to remove responses which are being wilfully misrepresented as published articles or when it is brought to our attention that a response spreads misinformation.
From March 2022, the word limit for rapid responses will be 600 words not including references and author details. We will no longer post responses that exceed this limit.
The word limit for letters selected from posted responses remains 300 words.
BMJ’s recent paper and editorial on cardiovascular (CVD) risk prediction models are relevant to patients, clinicians and policy makers. We agree that there is a need to validate the risk models, but we would also suggest to broaden the scope for the risk prediction models, in particular with respect to mortality. Typically, these risk prediction models only consider cause-specific, CVD mortality. While this may make sense from a medical point of view, where medications are considered only to be effective with respect to a specific kind of disease, we are less certain that this carries over to the perspective of patients and policy makers. Rather, we find it likely that their interest would lie in when a patient will die, irrespective of cause. Whether a patient will choose to change life style or take a medication, will likely not only depend on their CVD risk, but also their total mortality risk. It would appear natural to supplement CVD risk information with the corresponding expected lowering in all-cause mortality risk, should they choose to start an intervention. But, as far as we know, such information is not available from current models, for example the European Heart-SCORE model1. It is however possible to amend the model based on a competing risk approach to include non-cardiovascular mortality, and further compute the expected gain based on the current evidence of statin effectiveness, as we have documented recently2. Further, such a coherent model allows computing the expected residual lifetime for the patient, both with and without medication, which in turn allows estimation of the expected prolongation of life due to starting medication. There is evidence that lay people are better able to understand intervention benefits when they are presented in terms of prolongation of life than risk reductions.
It is still unclear which kind of information is most requested and best understood by patients, but regardless we think that risk prediction models should not only be valid, but preferably also be flexible and consistent. Only then will they be able to provide useful information in a format that is accessible to patients.
References
- 1. Conroy RM, Pyörälä K, Fitzgerald AP, Sans S, Menotti A, De Backer G, et al. Estimation of ten-year risk of fatal cardiovascular disease in Europe: the SCORE project. Eur Heart J. 2003 Jun;24(11):987–1003.
- 2. Støvring H, Harmsen CG, Wisløff T, Jarbøl DE, Nexøe J, Nielsen JB, et al. A competing risk approach for the European Heart SCORE model based on cause-specific and all-cause mortality. Eur J Prev Cardiolog [Internet]. 2012 Apr 12 [cited 2012 Apr 16]; Available from: http://www.ncbi.nlm.nih.gov/pubmed/22498473
For further enquiries, please contact Henrik Støvring (stovring@biostat.au.dk).
Competing interests: No competing interests
Re: Comparing risk prediction models
In response to Henrik Stovring's post, the BMJ have already a competing risk version of QRISK (known as QRISK lifetime) which takes account of the competing risk of mortality as well as allowing risk estimation over a variable number of years up to the age of 95 years . The paper can be found at http://www.bmj.com/content/341/bmj.c6624.
The web calculator is at www.qrisk.org/lifetime. The interpretation of competing risk models in a clinical setting with patients however can be challenging. For example there is a differential effect of smoking status on cardiovascular risk and risk of death from other causes, with a stronger effect on the risk of death from other causes. This probably explains the observation that, when competing risks are being accounted for, there may be cases when the lifetime risks of cardiovascular disease in smokers are lower than those for similar patients who are non-smokers (such as with clinical example 3 in the paper). This is likely to occur because the non-smokers have a reduced risk of death from other causes and so are likely to live longer, which increases the length of time over which they may develop cardiovascular disease. Validation of competing risks models also poses significant methodological challenges.
Competing interests: H-C is professor of clinical epidemiology at the University of Nottingham and codirector of QResearch—a not-for-profit organisation that is a joint partnership between the University of Nottingham and EMIS (leading commercial supplier of information technology for 60% of general practices in the UK). JH-C is also director of ClinRisk, which produces open and closed source software to ensure the reliable and updatable implementation of clinical risk algorithms within clinical computer systems to help improve patient care.