Statistical Evaluation of a Biomarker

Department of Emergency Medicine and Surgery, Groupe Hospitalier Pitié-Salpêtrière, Paris, France.
Anesthesiology (Impact Factor: 5.88). 03/2010; 112(4):1023-40. DOI: 10.1097/ALN.0b013e3181d47604
Source: PubMed


A biomarker may provide a diagnosis, assess disease severity or risk, or guide other clinical interventions such as the use of drugs. Although considerable progress has been made in standardizing the methodology and reporting of randomized trials, less has been accomplished concerning the assessment of biomarkers. Biomarker studies are often presented with poor biostatistics and methodologic flaws that precludes them from providing a reliable and reproducible scientific message. A host of issues are discussed that can improve the statistical evaluation and reporting of biomarker studies. Investigators should be aware of these issues when designing their studies, editors and reviewers when analyzing a manuscript, and readers when interpreting results.

Download full-text


Available from: Yannick Le Manach, Nov 14, 2014
    • "Nevertheless, its performance has not been studied in a population of less severely injured or in a homogenous population such as patients with pulmonary contusions. Furthermore , the only threshold available in the literature is the best statistical threshold determined using the Youden method [27], which is well known to be unsuitable in clinical practice [28]. Conversely, the grey zone approach defines statistical thresholds that can help the clinician to rule out or predict worsening risk of respiratory problems. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Background: Pulmonary contusion is a major risk factor of acute respiratory distress syndrome (ARDS) in trauma patients. As this complication may appear after a free interval of 24-48h, detection of patients at risk is essential. The main objective of this study was to assess the performance of the Thoracic Trauma Severity (TTS) score upon admission in predicting delayed ARDS in blunt trauma patients with pulmonary contusion. Methods: All blunt thoracic trauma patients admitted consecutively to our trauma centre between January 2005 and December 2009 were retrospectively included if they presented a pulmonary contusion on the admission chest computed tomography scan. Main outcome measure was the presence of moderate or severe ARDS (PaO2/FiO2 ratio≤200) for 48h or more. The global ability of the TTS score to predict ARDS was studied by ROC curves with a threshold analysis using a grey zone approach. Results: Of 329 patients studied (75% men, mean age 36.9 years [SD 17.8 years], mean Injury Severity Score 21.7 [SD 16.0]), 82 (25%) presented with ARDS (mean lowest PaO2/FiO2 ratio of 131 [SD 34]). The area under the ROC curves for the TTS score in predicting ARDS was 0.82 (95% CI 0.78-0.86) in the overall population. TTS scores between 8 and 12 belonged to the inconclusive grey zone. A TTS score of 13-25 was found to be independent risk factors of ARDS (OR 25.8 [95% CI 6.7-99.6] P<0.001). Conclusions: An extreme TTS score on admission accurately predicts the occurrence of delayed ARDS in blunt thoracic trauma patients affected by pulmonary contusion. This simple score could guide early decision making and management for a non-negligible proportion of this specific population.
    Injury 09/2015; DOI:10.1016/j.injury.2015.08.031 · 2.14 Impact Factor
  • Source
    • "In certain specific populations, a decision cut-off for ruling in a diagnosis of MI can be calculated via a receiver operator curve (ROC) [15], which helps to determine an optimal hs-cTn concentration in terms of clinical specificity and sensitivity. Reiter et al. found that the optimal hs-cTnT cutoff decision limit for elderly subjects (aged > 70 years) was 54 ng/L, which is about 4-fold the 99th percentile URL of a younger reference population [16]. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Cardiac troponin (cTn) assays have quickly gained in analytical sensitivity to become what are termed 'high-sensitivity cardiac troponin' (hs-cTn) assays, bringing a flurry of dense yet incomplete literature data. The net result is that cTn assays are not yet standardized and there are still no consensus-built data on how to use and interpret cTn assay results. To address these issues, the authors take cues and clues from multiple disciplines to bring responses to frequently asked questions. In brief, the effective use of hs-cTn hinges on knowing: specific assay characteristics, particularly precision at the 99th percentile of a reference population; factors of variation at the 99th percentile value; and the high-individuality of hs-cTn assays, for which the notion of individual kinetics is more informative than straight reference to 'normal' values. The significance of patterns of change between two assay measurements has not yet been documented for every hs-cTn assay. Clinicians need to work hand-in-hand with medical biologists to better understand how to use hs-cTn assays in routine practice. Copyright © 2014 Elsevier Masson SAS. All rights reserved.
    Archives of Cardiovascular Diseases 02/2015; 108(2). DOI:10.1016/j.acvd.2014.11.001 · 1.84 Impact Factor
  • Source
    • "Biomarkers have been increasingly and widely used in clinical practice in recent years for disease diagnosis and prognosis based on their underlying pathological and physiologic mechanisms [1]. In these applications, biomarkers are used either to identify a subgroup of a study population or predict a disease outcome [2]. "
    [Show abstract] [Hide abstract]
    ABSTRACT: The research on biomarkers has been limited in its effectiveness because biomarker levels can only be measured within the thresholds of assays and laboratory instruments, a challenge referred to as a detection limit (DL) problem. In this paper, we propose a Bayesian approach to the Cox proportional hazards model with explanatory variables subject to lower, upper, or interval DLs. We demonstrate that by formulating the time-to-event outcome using the Poisson density with counting process notation, implementing the proposed approach in the OpenBUGS and JAGS is straightforward. We have conducted extensive simulations to compare the proposed Bayesian approach to the other four commonly used methods and to evaluate its robustness with respect to the distribution assumption of the biomarkers. The proposed Bayesian approach and other methods were applied to an acute lung injury study, in which a panel of cytokine biomarkers was studied for the biomarkers' association with ventilation-free survival.
    01/2014; 3(1):32-43. DOI:10.6000/1929-6029.2014.03.01.5
Show more