Discussion
This prospective study of COPD patients with moderate airflow limitation and heightened CV risk provides two novel findings. First, the systemic levels of the five cytokines (CC-16, sRAGE, SPD, CRP and fibrinogen S), selected because of previous studies suggesting their potential value as predictors of outcomes in COPD, were unrelated to rate of FEV1 decline, exacerbation frequency and hospitalisations in these patients. The systemic levels of CRP and fibrinogen were associated with increased risk of all-cause mortality, confirming the predictive value for this outcome reported by others. Second, compared with placebo, once-daily inhalation of VI, FF and the combination resulted in small but significant decreases in the serum levels of CC-16, but not in the other biomarkers.
Previous studies
There has been a growing interest in the field of biomarkers in COPD as a tool to predict outcomes or as marker of them, thus making them theoretically useful as surrogate markers of response for therapeutic interventions.9 10 21–23 CRP has been the best studied biomarker in COPD. Most studies found that CRP levels are elevated in these patients,6 24 25 as compared with non-smokers and smokers without airflow obstruction, but the relationship between CRP levels and exacerbations as well as mortality remain inconsistent.25 26 Studies have also not consistently shown a relationship between CRP and rate of lung function decline.11 Similarly, several studies have documented a relationship between blood levels of fibrinogen and risk of death in patients in general27 and COPD in particular.5 7 28 Integration of all the data available by the COPD Biomarker Qualification Consortium initiative resulted in US Food and Drug Administration approval of this biomarker to stratify patients for studies of COPD.7 There is less information regarding the relationship between systemic levels of fibrinogen and other important outcomes in COPD, such as rate of lung function decline and hospitalisations.29
Of the many other systemic biomarkers studied in patients with COPD, the circulating levels of CC-16, SPD and sRAGE have shown the strongest association between their levels and COPD outcomes.15 30–34 CC-16 levels have been found to be reduced in patients with COPD and are inversely associated with rate of decline of FEV1 in some studies.13 35 The lung-derived protein SPD, however, is associated with presence of pulmonary inflammation and is elevated in smokers with or without COPD.31 The systemic levels of sRAGE are inversely related to the degree of emphysema as determined by CT scans of the lungs,14 while GWAS have documented the increased prevalence of single nucleotide polymorphisms associated with the RAGE gene.15 All of these proteins have been suggested as potential biomarkers of COPD outcomes, but not all of them have been studied simultaneously and prospectively while evaluating the effect of therapy on their levels.
Although statistically significant associations are important for studies and research, the ideal biomarker for the clinician is useful primarily if it helps a single patient, the one in front of the provider. So if a biomarker is to reach the bedside, it has to have a strong association with an outcome and provide information above and beyond that obtained by good clinical evaluation.
Current findings
The prospective SUMMIT study provided the opportunity to test the predictive ability of the selected circulating proteins on important respiratory outcomes. To our disappointment, none of the systemic levels of the five circulating proteins studied related to rate of FEV1 decline, exacerbations and hospitalisations (table 2, online supplementary figures 1 and 2). Although a true association might have been found if more patients had been recruited, this appears unlikely as several of the studies documenting an association between those biomarkers and the outcomes were smaller or similar in size to ours. In addition, it is known that the rate of FEV1 decline is larger in patients with higher FEV1,36 so if there was a relationship between any of the biomarkers selected and lung function decline, these moderately obstructed patients were the most likely individuals to have demonstrated such an association. Furthermore, a biomarker is most useful when it can help clinicians in their practice usually with a single patient, and certainly, the lack of precision documented in this study renders them of limited use in everyday patient care. However, the association observed between systemic levels of CRP and fibrinogen with mortality in this study (figure 3) supports their validity as predictors of risk of death in general and in patients with heightened CV risk, as were those included in SUMMIT.37 38 We acknowledge that there were no measurements of these biomarkers during exacerbations, where the changes might identify or grade the severity of these events, as has been suggested before.39 Importantly, the significant association observed between baseline serum levels of troponin and subsequent risk of CV events and death in this cohort17 indicates that methodological issues or the study design and completion are unlikely to have influenced the results here reported.
The difference between the negative results in COPD-related outcomes in this study, and the positive ones reported by previous authors may relate primarily to the populations included in the different studies.6–8 11–15 Whereas the majority of the positive results came from cohort studies such as ECLIPSE and COPDgene, which included patients with a wider range of lung function impairment, the population included in SUMMIT consisted of patients with a moderate degree of airflow limitation and heightened CV risk. However, for a biomarker to be useful in clinical practice, it has to help clinicians at any stage of the disease and most of all for patients with milder disease, such as those included in SUMMIT.
One potential use of biomarkers is their response to therapy specifically directed at the disease in question and if this change was associated to a modification in outcomes. In this study, the relative levels of CC-16 were decreased by the three active treatment arms of the study compared with placebo. The effects were small, likely of little clinical meaning, but nevertheless statistically significant (table S2). A decrease in CC-16 was not associated with FEV1 decline in this study, a fact that could have several explanations. First, that CC-16 bears little relation to FEV1 decline, as has been suggested by some studies,12 or second, that the small changes in serum levels of CC-16 carry little biological significance. However, the results do suggest that inhaled therapies may alter systemic levels of cytokines and that if associations with valid outcomes were to be found, one or more cytokines could be used as surrogate markers to evaluate effect of therapy. Interestingly, we found no effect of any of the therapies given in SUMMIT on CRP levels, providing contrast to the findings of an older study that demonstrated reduced CRP levels after the administration of ICSs.40 Conversely, however, our findings are in agreement with those of a further study, in which the administration of inhaled tiotropium did not affect inflammatory markers in patients with COPD.41 Finally, the stability of the selected biomarkers over 3 months is consistent with data previously reported in ECLIPSE.42
Strengths and limitations
The relatively large sample size and multicentre nature of the study, its careful clinical, functional and biological characterisation and its prospective design, follow-up time and clinical adjudicating committee are strengths of this study. However, there were several potential limitations. First, the study had an event-driven design and as such not all patients were followed over a period of several years. However, with an average of 18 months of on-treatment observation and of 27 months for mortality, the determination of two biosamples separated by 3 months and a mean number of 7 spirometric measurements per patient, we believe the observation time is sufficient for all of the outcomes selected. Second, the chosen panel of proteins did not include all possible biomarkers that have been suggested.9 43 However, the ones here measured included those that have the strongest support in the literature as being potentially applicable in clinical practice. Third, it could be argued that there was no derivative and validating cohort. However, this is customary for non-validated biomarkers, whereas in this study, we compared validated clinical and serum biomarkers modelled on studies in the respiratory and CV arena. The main weakness is likely to be the assumption that a few (or possibly only one) biomarkers will be valid for all patients with COPD. Given the heterogeneity of the disease this is unlikely; however, we currently are unable to identify subgroups where specific biomarkers would be of particular value.
Conclusions
In this substudy of the SUMMIT trial of patients with moderate COPD, the serum level of CC-16, sRAGE and SPD or their changes over 3 months were not predictive of rate of lung function decline and risk of exacerbations or hospitalisations. Although systemic levels of CRP and fibrinogen were associated with increased mortality risk, the levels were not associated with rate of FEV1 decline, exacerbations or hospitalisations. These results cast some doubts about the clinical usefulness of the systemic levels of CC-16, sRAGE and SPD as surrogate markers of disease in patients with moderate COPD.