Cytokines in pulmonary arterial hypertension: consider sensitivity when using multiplex technology

To the Editor:

Cracowski et al. [1] have recently published an interesting substudy in which they analysed the association of 17 baseline plasma cytokine concentrations with mortality in patients suffering from pulmonary arterial hypertension (PAH), using multiplex technology. The authors showed that four cytokines were independently associated with an increase in the adjusted hazard of mortality. Previously, Soon et al. [2] demonstrated that elevated levels of cytokines in PAH also predict survival over a 5-year period. Both these studies reinforce the increasing evidence that inflammation is associated with PAH remodelling.

Nevertheless, it is known that concentrations of circulating cytokines are very low, in the pg·mL⁻¹ range, requiring highly sensitive methods. In the letter by Cracowski et al. [1], despite a well-defined design and a robust statistical analysis, as previously described [3], no information is provided about the analytical performances of the method for cytokine quantification. In order to strengthen such cytokine-based clinical studies, the sensitivity of multiplex technology is herein discussed.

table 1 shows the mean baseline circulating cytokine concentrations from the two predictive PAH-related studies of Cracowski et al. [1] and Soon et al. [2], using different multiplex technologies: the Biochips-Array from Randox (Crumlin, UK) and the plate-based Multi-Array from Meso Scale Discovery (MSD; Rockville, MD, USA), respectively. In addition, the table 1 details, for each cytokine, the sensitivities reported by the manufacturers, and those determined in analytical studies by Fitzgerald et al. [4] and Dabitao et al. [5].

The theoretical sensitivity, also known as the lower limit of detection (LLOD), is defined as the lowest concentration of the analyte that can be distinguished from zero with at least two standard deviations, whereas the functional sensitivity, also known as the lower limit of quantification (LLOQ), is defined as the lowest concentration giving a coefficient of variation <20% on replicates. In hospital practice, for a given parameter measured in biological fluids, a result below the sensitivity of the method should neither be validated nor transmitted to the clinicians, because of its excessive imprecision. As shown in table 1, among the four cytokines independently associated with death in the study by Cracowski et al. [1], three had a mean baseline concentration below the corresponding sensitivities claimed by Randox (interleukin (IL)-1α, IL-13 and tumour necrosis factor-α) and two of these are also below the functional sensitivities as determined by Fitzgerald et al. [4]. Using the same Randox system, Duncan et al. [6] have shown that IL-6 and vascular endothelial growth factor were significantly associated with the occurrence of an adverse event in paediatric PAH, with median concentrations very closed to the sensitivities claimed by the manufacturer. In contrast, considering the PAH patients from the study by Soon et al. [2], the concentrations of the four cytokines independently linked with survival time are much higher than the LLOQs reported by MSD or than those determined by Dabitao et al. [5], making these results absolutely valid, analytically. It should be noted that most of the standard deviations are very high, suggesting a skewed distribution, with a few patients with very high levels and a non-negligible number of patients with cytokine concentrations below the limit of detection [1, 2]. This high number of nondetected data (>50%) has been recently described for cytokines assessed by three commercially available multiplex assays [7]. Consequently, caution should be considered about low cytokine circulating concentrations, because most probably measured with an analytical imprecision >20%, making such promising clinical findings somewhat questionable.

Functional sensitivity should be considered in clinical studies, especially when the main outcome is based on cytokine measurement. This criterion could be determined by repeated measurements of plasma pools, prepared so as to obtain decreasing concentrations of cytokine, allowing determination of the lowest concentration giving an imprecision of 20% using polynomial regression or Horwitz curve, as recommended by the French Accreditation Committee (COFRAC). Such a protocol has been recommended by the National Academy of Clinical Biochemistry for the determination of functional sensitivity of the third-generation thyroid-stimulating hormone test, as recently published [8]. Elsewhere, a similar approach has been used to determine the sensitivity of the high-sensitivity cardiac troponin T test [9].

Recently, using the expensive and time-consuming ELISA method, Heresi et al. [10] have shown that plasma IL-6 concentrations >4.7 pg·mL⁻¹ provide incremental prognostic information in PAH, with valid concentrations above the theoretical sensitivity (LLOD 0.7 pg·mL⁻¹). Similarly, using MSD multiplex technology, Soon et al. [2] showed that serum IL-6 was also associated with survival in PAH patients, with concentrations (19.87±7.45 pg·mL⁻¹) above the sensitivity (LLOQ 1.58 pg·mL⁻¹), making such a cytokine very promising for the prognosis of PAH.

In conclusion, multiplex analysis provides much information from a single biological sample, and is therefore popular and more frequently used in PAH studies. However, cytokine concentrations are low and often close to the sensitivity of the assay, depending on the multiplex assays. This currently prevents the routine clinical use of such biomarkers, and should encourage clinicians to take advantage of the biochemists’ analytic experience.

• Received January 1, 2014.
• Accepted February 7, 2014.

• ©ERS 2014