gms | German Medical Science

Introduction: Biomarkers which provide evidence for patient-treatment interaction are key elements for the development of personalized precision medicine. Statistical strategies to establish this evidence for continuous biomarkers are complex and their results are not easy to communicate. Good graphical representations would help to translate such findings into the clinical community. Although general guidelines on how to present figures in clinical reports are available, there is little guidance for figures explaining the role of continuous biomarker in patient treatment interaction (CBPTI). The aim of the study was to provide recommendations for CBPTI plots, through a systematic review of current practice as well as relevant statistical methodology.

Methods: with a systematic review of parallel-group RCTs published in six major general medical journals between January 2013 and December of 2014 was performed. Additionally, a systematic review of methodological papers in four major biostatistics journals during 2000-2014 was made.

Results: The review of medical journals resulted in four types of graphical CBPTI presentations used in 4 out of 767 papers. The review of biostatistics journals revealed further types of CBPTI plots: the proportion of unfavorable treatment effect plot, marker-by-treatment predictiveness curve, risk curve, selection impact curve, and specific variants of the ROC curve. Seven criteria were derived to assess the quality of a CBPTI plot, including the plot’s axes, the display of the contrast between treatments, the direction of better effect, information on uncertainty, scaling by absolute risk difference, benchmarking between treatments, and information content for medical decision making. A specific R-package is provided.

Discussion: There is considerable scope to improve the graphical representation of CBPTI in clinical reports. The graphical presentation should provide valuable information to aid medical decision-making. The modified ROC is seen as the most informative approach to present and compare the relevance of specific biomarkers in patient-treatment interaction.