Towards prognostic biomarkers in pulmonary arterial hypertension.

T he most reliable way to assess the clinical impact of a therapeutic intervention in pulmonary arterial hypertension (PAH) is through its effect on well-defined clinical end-points, such as mortality. However, this standard is often impractical because of the long period and large number of patients required to reach these clinical end-points, and substitutes, such as the 6-min walking distance, are often used. Biomarkers may also be reliable substitutes for clinical end-points, and are called surrogate end-points when they meet such requirements. According to the US National Institutes of Health Biomarkers Definitions Working Group, a biomarker is ‘‘a characteristic that is objectively measured and evaluated as an indicator of normal biological processes, pathogenic processes, or pharmacologic responses to a therapeutic intervention’’ [1]. However, while this definition is used for drug discovery, another narrower definition is still commonly used: ‘‘disease-associated molecular changes in body tissue and fluids’’, i.e. a biological marker [2]. Among the estimated 150,000 papers documenting thousands of biological markers, less than 100 biomarkers are routinely used in the clinic [2]. In the field of PAH, which is a rare disease, most biomarker studies examined less than 100 patients, and therefore lacked the statistical power needed to demonstrate a strong association between any biomarker and death. Registries such as the Registry to Evaluate Early And Longterm PAH Disease Management (REVEAL) [3] have the advantage of enabling the study of larger populations and, therefore, have much better statistical power. However, in most cases, registries are not designed to evaluate new biomarkers, as they do not reach the necessary quality of biological sampling with storage of samples in a biobank, and are often a mix of incident and prevalent patients. In addition, many studies have focused on just one biomarker, while there is increasing evidence that a single isolated biomarker will not provide a substantial improvement in risk classification and that a multiple-biomarker approach is required. Indeed, recent multiplex assays, such as those used by SOON et al. [4] for inflammatory cytokine measurement, enable a large number of substances to be readily quantified simultaneously, and we need to determine how to handle and interpret this large quantity of biological information. Furthermore, there is a time-lag between preliminary studies showing an independent association with death and the routine use of a biomarker for patients. Before a biomarker can be used routinely, several criteria must be met [5, 6]. First, the measure should add independent information on PAH prognosis, as shown in a survival analysis based on biomarkers with clinical and haemodynamic parameters as covariates. Secondly, the measure should account for a large proportion of risk. This means that the risk factor frequency must be sufficiently high (e.g. a biomarker such as cardiac troponin T (cTnT) will provide prognostic information only for the few patients in which it is detectable, although this can be improved using highsensitivity techniques), and that the relative risk should also be sufficiently elevated. There is no cut-off point to quantify the proportion of risk, but a hazard ratio .2 is frequently suggested [5, 6]. However, risk usually increases progressively with increasing biomarker levels, so various cut-off points must be evaluated. In most cases, these cut-offs will not represent a specific threshold effect, but a compromise between the proportion of the population at risk and the relative risk of morbidity associated with values higher than the cut-off point. As an example, BENZA et al. [3] clearly showed that brain natriuretic peptide levels .180 pg?mL were associated with a poor prognosis, with a hazard ratio close to 2, and the opposite for those with levels ,50 pg?mL. The question remains as to whether such an extreme cut-off would be of any help in the clinical decision making process given the fact that most patients will be between these values. Lastly, the test should be reproducible and widely available. While a large number of biomarkers are initially discovered in research laboratories, they need to be quantified rapidly at a low cost for routine use. To date, only a few biomarkers meet these requirements in PAH.