European Heart Journal Advance Access originally published online on April 15, 2008
European Heart Journal 2008 29(10):1336-1337; doi:10.1093/eurheartj/ehn151
Why must new cardiovascular risk factors be carefully re-assessed prior to clinical application?
Department of General Dentistry
Boston University School of Dental Medicine
100 East Newton Street, Rm. G-619
Boston MA 02118
USA
Tel: +1 617 414 1060
Fax: +1 617 414 1061.
Email: skjanket{at}bu.edu
Division of Biostatistics
Yale School of Public Health
New Haven, CT
USA
Department of Neurology
SUNY Downstate Medical Center
Brooklyn, NY
USA
We would like to comment on the clinical utility of biomarkers discussed in the article by Smulders et al.1 The authors stated that ‘the issue of causality [of biomarkers] is irrelevant’ but this may not be true if the ultimate goal of medical intervention is to decrease disease occurrence. In epidemiology, if a biomarker is significantly associated with an outcome in a longitudinal analysis, it is presumed that the relationship may be causal. However, not all significant biomarkers are causal factors: rather, some could be risk indicators accompanying a causal factor as in the case of microalbuminuria, troponins, and B-natriuretic protein described by Dr Wang.2 Although these non-causal risk indicators have value in predicting future events, as Dr Smulders stated, only changing the causal factors will confer any changes in disease morbidity and mortality. We agree with Dr Wang in that ‘careful assessment of evidence’ is required and ‘most current biomarkers are not ready for clinical application’.2 If we apply observational study results directly to clinical practice without further testing, the results are extrapolated beyond what the data support and may even be applied in the reverse direction.
We urge careful consideration of the relative risk, c-statistic, and Hosmer–Lemeshow statistics. A significantly high relative risk generated by a biomarker denotes that persons with high levels of the biomarker had a higher disease rate. It does not give evidence that reducing the level of this biomarker will reduce the occurrence of disease. Another test is needed to examine whether reducing the level of this biomarker will reduce the occurrence of disease. Changing the levels of highly significant biomarkers from observational studies has been proved not to change the outcome in many clinical trials.3,4 The discordance between these two study designs should not be interpreted as one being inferior to the other. Each provides valuable information by examining the disease–biomarker relationship from a different direction. Additionally, the relative risk is calculated by dividing the disease-proportion among those with high levels of the biomarker by the disease-proportion among those with low levels, ignoring the distribution of this biomarker among the non-diseased.5 Thus, a significantly high relative risk of a biomarker does not signify its clinical usefulness.6 The c-statistic simultaneously illustrates the distribution of a biomarker among the diseased (true positive) and the non-diseased (false positive); thus, it informs us how well this marker relates to disease status and provides more useful clinical insight than does the relative risk.5 The Hosmer–Lemeshow statistic, contrary to Dr Smulders confidence, will declare ‘no evidence of lack of fit’ if the test statistic falls within the range of 95 percentile when compared with the theoretical probability. This broad range of ‘goodness of fit’ makes this test unsuitable as a mean to assess precise model-fit. Furthermore, it does not offer any information about the model's clinical performance.
Research evaluating new biomarkers is necessary to improve our understanding of pathophysiology. However, clinical application must be preceded by careful re-assessments utilizing other epidemiologic and statistical methods.
This study was supported by the American Heart Association grant (no. 0635351N to S.-J.J.).
References
- Smulders YM, Thijs A, Twisk JW. New cardiovascular risk determinants do exist and are clinically useful. Eur Heart J (2008) 29:436–440.
[Abstract/Free Full Text] - Wang T. New cardiovascular risk factors exist, but are they clinically useful? Eur Heart J (2008) 29:441–444.
[Abstract/Free Full Text] - Hennekens CH, Buring JE, Manson JE, Stampfer M, Rosner B, Cook NR, Belanger C, LaMotte F, Gaziano JM, Ridker PM, Willett W, Peto R. Lack of effect of long-term supplementation with beta carotene on the incidence of malignant neoplasms and cardiovascular disease. [See comment]. N Engl J Med (1996) 334:1145–1149.
[Abstract/Free Full Text] - Howard B, Van Horn L, Hsia J, Manson JE, Stefanick ML, Wassertheil-Smoller S, Kuller LH, LaCroix AZ, Langer R, Lasser N, Lewis CE, Limacher MC. Low-fat dietary pattern and risk of cardiovascular disease: The Women's Health Initiative Randomized Controlled Dietary Modification Trial. JAMA (2006) 295:655–666. (WHI).
[Abstract/Free Full Text] - Janket S-J, Shen Y, Meurman JH. Use and misuse of the receiver operating characteristic curve in risk prediction'. [comment]. Circulation (2007) 116:e133. author reply e134.
[Free Full Text] - Myerburg RJ. Scientific gaps in the prediction and prevention of sudden cardiac death. J Cardiovasc Electrophysiol (2002) 13:709–723.[CrossRef][Web of Science][Medline]
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||