Background

Chronic obstructive pulmonary disease (COPD), characterized by reduced lung function, is a common cause of death; World Health Organization (WHO) reports COPD to cause 5% of deaths globally in 2015 [1]. Clinically, COPD is typically diagnosed using spirometry. The Global Initiative for Chronic Obstructive Lung Disease (GOLD) recommends using the ratio of forced expiratory volume in 1 s (FEV1) to forced vital capacity (FVC) to assess whether the patients suffer of airflow limitation. According to the GOLD criterium, values of FEV1/FVC less than 70% after bronchodilator administration are an indication of airflow obstruction [2]. COPD diagnosis as defined by GOLD also requires respiratory symptoms [2]. Normal ageing causes a natural decline in the lung function which varies with sex and ethnicity [3]. Therefore, a fixed ratio tends to overestimate COPD prevalence in the elderly [4]. To overcome these limitations, the European Respiratory Society (ERS) instead recommends comparing the FEV1/FVC to population-based reference values [5]. According to this definition, individuals with potential airflow obstruction are identified as those with FEV1/FVC below the lower limit of normal (LLN; below the fifth percentile adjusted for age, sex, height, and ethnic group) [4]. In this study, we use the terms pulmonary function decline, lung function decline, and airflow obstruction as an estimate of COPD as we primarily assessed spirometry results and proteomics and not symptoms.

COPD prevalence differs among studies, on the basis of definition criteria, age and regional differences [6, 7]. In an Austrian sample of 1258 Salzburg County inhabitants aged  ≥  40 years the prevalence of irreversible airflow obstruction defined by the GOLD criteria was estimated to 26.1% [8]. A Dutch study presented an LLN-based COPD prevalence of 19% in a population-based sample of adults aged  ≥  40 years [9]. In the Swedish BOLD study, conducted on 548 urban individuals aged 40 years or more [10], COPD prevalence using LLN criteria was calculated to be 10%, compared to 16.2% using GOLD criteria. Backman et al. [11] found an LLN-based COPD prevalence of 4.9% in a Swedish population aged 21–78 years, n  =  1839. The OLIN studies showed a linear increase in COPD prevalence from 45 years of age in smokers in the north of Sweden [12]. Luoto et al. [13] showed a linear effect of age on incidence for airflow limitation for fixed-ratio criteria but not for LLN criteria.

Cardiovascular disease (CVD) is a common comorbidity in COPD and reduced lung function is an important risk factor for CVD and CVD-related death [14]. However, the mechanisms behind the increased risk for CVD in COPD patients are not fully understood. Previous research has identified inflammatory proteins related to systemic inflammation that could distinguish COPD patients from healthy control subjects [15, 16]. Briefly, inflammation is a pathological process mediated by cytologic or chemical reactions [17], and several conditions can lead to increased inflammatory biomarkers in COPD patients. Local inflammation in the airways can cause inflammatory molecules, such as cytokines and chemokines, to be released into the blood stream, leading to systemic inflammation [14]. Repeated exposure of cigarette smoke and air pollution causes chronic inflammation in central and peripheral airways, lung parenchyma and blood vessels in the lungs [18,19,20,21]. Furthermore, parallel processes such as atherosclerosis, heart failure and diabetes can lead to increased systemic inflammation in COPD patients [14].

Proteomics enables simultaneous analysis of the content of several biomarkers in various fluids. It has previously been used to identify potential biomarkers for the pathogenesis of various diseases, including COPD [22, 23]. By identifying CVD and inflammatory biomarkers and investigating their association with lung function, information about airflow obstruction and its association with CVD and systemic inflammation might unveil. This knowledge can provide a deeper understanding of the mechanisms behind CVD, inflammation, and pathogenesis of airflow obstruction. In addition, the presence of biomarkers can potentially improve diagnostic. COPD diagnosis currently requires a subject to be physically and cognitively fit to provide an acceptable spirometry, which hampers diagnosis of the frailest subjects. Serum biomarkers could potentially serve as a diagnostic tool in this group.

In the current study, we investigate the association between CVD, inflammation, and lung function in an elderly population. We analyse the association of 266 serum potential biomarkers related to CVD and inflammation with lung function.

Materials and methods

Study design and participants

The current study is a part of the general longitudinal population study Good Aging in Skåne (Gott Åldrande i Skåne, GÅS) that started in 2001 [24]. The GÅS project comprises three waves recruited 2001–2004 (n  =  2931, Wave 1), 2007–2009 (n  =  1523, Wave 2) and 2013–2015 (n  =  1350, Wave 3), with a participation rate between 60 and 70%. Subjects aged from 60 to 93 years were randomly invited from the Swedish population register [25]. After inclusion, subjects were examined and questioned by a physician, a nurse, and a medical secretary at one out of four research test centres in Skåne. Examination included medical history, electrocardiography, blood samples, spirometry and cognitive tests. Study participants attended follow up visits every third year (subjects aged  ≥  78 years) or every sixth year (subjects aged  <  78 years).

632 blood samples from subjects consecutively examined during 2015–2018 were retrieved from the GÅS study. At that point in time, the prevalence of airflow obstruction in this GÅS population was not fully established. Therefore, to assure that at least some subjects in the study population suffered from airflow obstruction, 72 subjects known to fulfil the COPD criteria according to GOLD were also included. In total, 704 blood samples were analysed.

Proteomic analysis

The blood samples were drawn and immediately stored at -80 degree Celsius. Proteomic analysis and quality control (QC) was performed by Olink, Olink Proteomics AB, Uppsala, Sweden, at Clinical Biomarkers Facility, Science for Life Laboratory, Uppsala University, Sweden [25] in 2019. Briefly, the Olink Proseek Multiplex CVD II, CVD III, and Inflammation kits, measuring 266 CVD- and inflammation-related biomarkers in plasma were used [26]. Ten of the biomarkers overlapped and belonged to more than one panel. The analysis was performed using the proximity extension assay (PEA) [27]. Quality control (QC) was performed by adding four internal controls to each sample and external controls in every analysis. The quality control process was performed in two steps. First each sample plate was evaluated on the standard deviation of the internal controls. This should be below 0.2 NPX. Only data from sample plate that pass this quality control are used in the second step of the quality control process. The quality of each sample was further assessed by evaluating the deviation from the median value of the controls for each individual sample. Samples that deviate less than 0.3 NPX from the median pass the quality control. Samples that did not pass the QC were excluded from the analyses. Values below the limit of detection (LOD) were included in the primary analysis. 654 out the 704 participants’ blood samples passed the Olink QC analysis.

Spirometry

Spirometries were performed using the Vitalograph 2120 electronic flow volume spirometer using Spirotrac IV software (Vitalograph Ltd., Buckingham, UK). The spirometry, including daily calibration, was performed according to American Thoracic Society (ATS) guidelines using bronchodilators by one and the same registered nurse [28]. Up to a maximum of 8 consecutive breathing maneuvers were performed with the goal of completing at least three acceptable curves. Reproducibility criteria was used as an indication of whether more maneuvers were necessary. To avoid exclusion of very frail individuals, only one acceptable manoeuvre was required [29]. Spirometry and blood samples were collected on the same visit.

Definition of airflow obstruction

We defined airflow obstruction using the LLN criterion, as described in the 2012 Global Lung Function Initiative reference equations for the predicted FEV1/FVC ratios [30, 31]. We choose the LLN criterion because it has a higher specificity compared to the GOLD criterium in the elderly population [32].

Statistics

The strength of the association between the biomarkers and subjects with air flow obstruction was estimated using a logistic regression model. A raw model was implemented to assess which protein biomarkers are associated with poor lung capacity. Each protein panel included 93 proteins which were analyzed separately. To acknowledge that samples from the same participants were used to calculate 93 different logistic models (one for each protein) and that protein levels may be correlated, the Yekutieli method [33] was used to keep the false positive discovery (FDR) rate at 5%. This method was preferred over the Benjamini and Hockberg method since the Yekutieli method is expected to control the FDR regardless of the correlation among the p values [34]. Age, sex, and smoking affect the airflow and may also affect the protein concentration levels. To enhance the interpretability of our results, a logistic model including these factors as covariates was applied only to those protein biomarkers that were statistically significant in the crude logistic model. A likelihood ratio test was used to evaluate the difference between the crude and adjusted models.

To assess the impact of the inclusion of values below LOD, the statistical analyses were repeated excluding those values.

The statistical analyses were performed using Stata IC 14.2 software (StataCorp LLC, 4905 Lakeway Drive, College Station, TX 77845, USA). The calculations of the p values were performed using the package q-values [34].

Results

Patient characteristics

The patient flow chart is shown in Fig. 1. In total, 704 subjects were included in the study. Of those 611 subjects had completed spirometry and had blood samples that passed the Olink quality control procedure. Baseline characteristics are presented in Table 1. 10.3% had an FEV1/FVC ratio below LLN, i.e., airflow obstruction according to LLN criteria. Notably, only 60.3% of subjects fulfilling the LLN airflow obstruction-criteria were diagnosed with COPD according to medical records. 6.2% of subjects with normal pulmonary function were diagnosed with COPD. 58.9% of the study population were current or former smokers.

Fig. 1
figure 1

Number of participants

Full size image
Table 1 Demographic characteristics
Full size table

CVD and inflammation biomarker profile

We investigated association in all proteins included in the panels, and no results were discarded. Only one of the 266 analysed biomarkers, plasminogen activator, urokinase receptor (PLAUR), was significantly associated with pulmonary function decline according to LLN (FDR  <  5%, P  <  3.52  ×  10–5). The concentration levels for PLAUR are shown in Fig. 2. It is seen that subjects with poor lung capacity tend to have higher concentration levels of PLAUR. Odds ratio for PLAUR was 4.37 95% CI (2.17–8.79) (see Table 2). This indicates that subjects with decreased lung function were more likely to have increased levels of PLAUR than subjects without lung function decrease. The results for the adjusted logistic model including age, sex and smoking status are shown in Table 2. Also in this model, participants with poor lung capacity have higher concentration levels of PLAUR. The likelihood ratio test suggested that the adjusted model improved the model fitting compared to the crude model (p value 0.0014). The results of the crude logistic model for the other proteins are reported in the Additional file 1 (see Table S1).

Fig. 2
figure 2

Concentration of PLAUR for participants with and without FEV1/FVC below the lower limit of normality. FEV1 force expiratory volume in the first second. FVC forced vital capacity

Full size image
Table 2 Logistic regression to assess the association between the biomarker soluble urokinase plasminogen activator surface receptor (U-PAR) and poor pulmonary function (FEV1/FVC  <  LLN)
Full size table

Sensitivity analyses

The statistical analyses were repeated excluding all values below the limit of detection. There were 2 protein biomarkers from the inflammation panel (IL2 and IFNG) that were completely excluded from the analyses since none or very few values remained above the LOD. The results are shown in the Additional file 1 (Table S2) and were consistent with those presented in the primary analysis.

Discussion

In this study, we investigate how CVD- and inflammation-related serum biomarkers are related to poor lung function in the elderly. We found that subjects with limited airflow tend to have higher concentration of PLAUR. Previous research has shown an association between increased levels of serum PLAUR and COPD in middle-aged patients with moderate COPD [35,36,37,38]. The mechanism behind this association is not fully understood. Upon activation by the serine protease plasminogen activator, urokinase (PLAU), following inflammation, infection or malignancy, PLAUR activates plasminogen into plasmin, promoting inflammatory cell migration and activation, extracellular matrix (ECM) degradation and matrix metallopeptidase (MMP) activation [39,40,41]. MMPs may play a role in the pathogenesis of pulmonary function decline and previous research has shown significantly increased levels of MMPs in both serum and airways [42]. Proteolytic imbalance leads to emphysema development, small airway obstruction and increased sputum levels [43].

In our study, we found no statistically significant association between lung function reduction according to LLN and biomarkers previously associated with lung function decline. Interleukin 6 (IL6) is an acute phase response cytokine released from epithelial cells and macrophages, among other cells, and has in previous researched been positively associated with decreased lung function. Yasuda et al. found a significant increase of IL-6 in serum in both severe and mild/moderate COPD subjects compared to healthy age- and sex-matched controls, and Zeng et al. found a significant increase in stable COPD subjects compared to controls [44, 45]. As these studies compared subjects with diagnosed COPD with controls, increased inflammatory cytokines could be expected. Our study population is based on sample from overall healthy older subjects. Therefore, some participants with pulmonary function decline could be asymptomatic: only 60.3% of subjects fulfilling pulmonary function decline criteria reported being diagnosed with COPD. Thus, we could expect that the overall inflammatory level to be lower in our subjects compared to the above mentioned studies.

Tumor necrosis factor (TNF) is a pro-inflammatory cytokine secreted mainly by macrophages, monocytes, neutrophils, activated T-cells and NK-cells [46]. Increased serum levels of TNF have previously been associated with COPD in a male patient sample (n  =  30) with stable COPD and reduced body weight [46, 47]. The current study could not find a statistically significant association between serum TNF levels and airflow obstruction. Our study population was fit enough to visit the study centre, indicating a physical status without significant weight loss and where increased TNF levels would not be expected.

Surfactant protein D (SFTPD) is produced by pneumocytes and plays a role in the defence against microorganisms. Sin et al. studied 23 patients with advanced COPD (mean age 64.8 years, 12 females) and found an inverse association between FEV1 and serum levels of SFTPD. Ciftci et al. found significant increased serum levels of SFTPD in 63 stable COPD subjects (mean age 63.7 years, 5 females) compared to 25 controls (mean age 53.8 years, 8 females). Our study showed no significant association between serum levels of SFTPD and pulmonary function decline. Our study population was larger and had a more balanced gender distribution, which could partially explain that our results differ from previous research.

Our study has several strengths. The study population consisted of a general geriatric sample of subjects living in urban areas in southern Sweden, which allow us to investigate the association of pulmonary function decline and biomarkers in a representative sample. We defined pulmonary function decline according to LLN criteria as proposed by ERS [5]. The GOLD definition has a tendency to overestimate the prevalence of airflow obstruction in subjects aged  >  50 years [30]. A decreased pulmonary function with age is a physiological change and does not necessarily means that an elderly person has a disease. In addition, sex has been shown to affect LLN significantly, but this is not considered in the GOLD criteria [13]. In the current study, 54.7% were females. Despite the mentioned advantages of the LNN criteria, this definition somewhat underestimates the prevalence of COPD in the elderly [32]. We observed that 6.2% of subjects with FEV1/FVC  ≥  LLN had been diagnosed with COPD. Therefore, we believe that a potential underestimation of clinically diagnosed COPD in our study is negligible. Furthermore, a Belgian study of 567 individuals aged  ≥  80 years did not show a significantly increased risk of missing patients with increased risk of hospitalization or mortality when defining airflow limitation according to LLN instead of GOLD [48]. Similarly, a Finnish study could not show any clear difference in mortality risk between patients with airflow limitation according to GOLD criteria compared to patients with LLN-based airflow limitation definition [49]. We found the airflow obstruction prevalence according to LLN criteria to be 10.3% in our study, which is in accordance with COPD prevalence in the Swedish BOLD study, but higher than the prevalence in the study by Backman et al. [11]. The increased airflow obstruction prevalence could be explained by the fact that the current study only included subjects aged  >  65 years.

Our study has also some limitations. Incomplete data was due to mainly two factors: non-acceptable Olink QC and missing spirometry data. Olink QC was non-acceptable if the blood sample was too small or if the qPCR failed. These factors are not related to the subjects neither to their pulmonary function, and therefore are not expected to influence the presented results. Spirometry data was missing if the subject could not perform an acceptable manoeuvre which could indicate decreased pulmonary function due to the physical status. To minimize the exclusion of frail subjects, only one acceptable spirometry manoeuvre was required.

Conclusion

In conclusion, our study showed a statistically significant association between elevated serum levels of PLAUR and decreased pulmonary function. Further research is needed to investigate the possible role of PLAUR in the pathogenesis in pulmonary function decline, as well as PLAUR as a potential marker for pulmonary decrease.