In this study, results from nephelometry and ELISA were compared in a sample of stable COPD patients. This study provides insight into how assay differences may influence the estimates of inflammation in clinical settings. We are not aware of any other report that has addressed this important clinical and research issue. In fact, the usefulness of systemic biomarkers has been evaluated in several clinical settings (e.g., the emergency room and intensive care unit and during hospitalization) and for a number of diseases; therefore, the standardization of technical assays is necessary for all biomarker studies to provide meaningful data. We were also able to confirm that stable COPD is associated with an increase in CRP and SAA levels compared with healthy controls, which suggests the presence of a systemic inflammatory load, as previously described .
Extensive research has investigated the relevance of different inflammatory markers in COPD. CRP and SAA are major acute-phase reactants in humans that are synthesized in the liver by factors that initiate and maintain the inflammatory response, such as interleukin (IL)-1, IL-6, and tumor necrosis factor α. Although several inflammatory mediators induce and control the secretion of these biomarkers, CRP is predominantly induced by IL-6, whereas SAA is stimulated mainly by IL-1 .
Increased levels of CRP may reflect the severity of airway inflammation  with a strong prognostic value , and CRP has recently become a therapeutic target in cardiovascular diseases . Of note, this molecule can provide prognostic information beyond that which is conveyed by the Framingham risk score . However, the exact pathophysiological relevance of high CRP levels remains controversial. A previous study linked CRP to plasma low-density lipoprotein, which may play a significant role in modulating CRP function under physiological conditions . Taken together, these data suggest that CRP may be involved in lipoprotein metabolism, clearance, and deposition.
Increased SAA concentrations have been reported in a number of chronic inflammatory and neoplastic diseases that have been linked to systemic amyloid deposition . Higher serum SAA levels are also associated with COPD exacerbation and are more predictive than CRP for the severity of these episodes . Growing evidence has suggested that SAA may act as an apoprotein. SAA is involved in the conformational change in the high density lipoprotein (HDL) particle  and causes increased HDL clearance [29, 30]. In light of these findings, it has been hypothesized that increased SAA levels can contribute to accelerated atherosclerosis in stable COPD .
Although we found statistically significant method-dependent differences for CRP and SAA, specific differences for each biomarker were noted. First, the distribution of each molecule is noteworthy. As shown in Figure 1, both of the acute phase reactants displayed a wider distribution when measured using nephelometry in COPD compared with the ELISA assays or measurements obtained from the control group. This finding may at least in part explain our results that suggest a more pronounced difference in COPD patients than in controls. This difference is reflected in the standard deviations, which were higher when the biomarkers were measured using nephelometry in the COPD group. Interestingly, this wide distribution has also been reported by other authors who have used different laboratory techniques (e.g., a high sensitivity chemiluminescent immunoassay) . Additionally, the differences between the two techniques were less marked for SAA than for CRP. When the two methods were compared directly (Tables 2 and 3), the differences in SAA levels did not reach the threshold for significance. However, the analysis of agreement showed a consistent difference between the two methods. As shown by the Bland-Altman test, such differences may be clinically important for COPD patients. It is also noteworthy that the kappa coefficient was low.
Despite the high sensitivity, speed, and accuracy as well as the good correlation of nephelometry with other techniques in humans [31, 32] and animal models , the agreement of this technique with ELISA has not been previously investigated in patients with COPD. Interestingly, a discrepancy between these techniques was reported in an analysis of cerebrospinal fluid IgG . The present study showed that although the concentrations of inflammatory markers obtained using nephelometry and ELISA were moderately correlated, the agreement between the two assays is remarkably low. It is now well-recognized that a good degree of correlation does not necessarily imply a good agreement, and vice versa . Although the two assays correlate well, they may provide different information about COPD patients that could ultimately lead to different conclusions concerning the prognostic significance of the biomarkers.
In this regard, the effect of the assay differences on the selection of appropriate cut-off values for different inflammatory markers is a clinically important issue and therefore must be more precisely characterized. To our knowledge, only one study using nephelometry has attempted to identify a reliable cut-off for increased CRP concentrations (4.21 mg/L) in patients with chronic respiratory disorders . However, other authors have used 3.0 mg/L as a cut-off for high CRP levels that were measured using nephelometry . Although we estimated a cut-off value of 3.1 mg/L for CRP in this study, we found robust evidence of a significant analytical difference between nephelometry and ELISA for the measurement of serum inflammatory markers, with a more marked difference in COPD patients. Although the reasons for the analytical differences between the two assays remain unclear, such a large difference is likely to be clinically significant. It is remarkable in this regard that a difference of 3.9 units for CRP in COPD patients may overcome the threshold value for poor prognostic significance, i.e., 4.21 , 3.0  or 3.1 mg/L (the current study).
Although the differences between the assays were less pronounced for SAA, the same reasoning can be applied to this biomarker. The reference values for SAA in healthy individuals have not yet been thoroughly established. To our knowledge, there is only one study on this issue, which identified a threshold value of 7.0 mg/L using an ELISA assay . Of note, this cut-off agrees with that identified in the present study. Using these thresholds, a method-dependent difference of 1.2 mg/L accompanied by a large standard deviation (Table 3) may have a significant impact on the prognostic significance of this marker in a clinical setting. However, a long-term follow-up of the patients would provide insight as to whether these differences in measurements are relevant from a clinical point of view and would determine the variability and reliability of both markers assessed by both methodologies. The methodology used in the present study did not include this follow-up. Future trials should evaluate this aspect in a prospective cohort.
The relationship between inhaled corticosteroid use and the systemic inflammatory load in COPD is a source of debate. In an initial small study, the use of inhaled corticosteroids at a dose of 1,000 μg of fluticasone per day was shown to be effective in reducing serum CRP levels in patients with COPD . However, several years later, the same group found that this association was not present for CRP in a clinical trial multicenter study with a larger sample size . The present study did not show this association, and our results thus provide further evidence for the lack of association.