Abstract
Background
Cognitive impairment (CI) is among the extrapulmonary comorbidities that are increasingly recognised in individuals with COPD and a history of cigarette smoke exposure. While severe hypoxaemia is a well-established risk factor for CI, the role of moderate hypoxaemia (peripheral oxygen saturation (SpO2) 89–93%) is unknown.
Methods
We evaluated the association between moderate hypoxaemia, assessed by pulse oximetry, and CI in participants from the Genetic epidemiology of COPD (COPDGene) study who completed the Phase 3 (10-year follow-up) visit. We performed mediation analysis of moderate hypoxaemia on the association between COPD and CI. We tested for differences in proteomic biomarkers of cerebral hypoxia and performed differential gene expression and pathway enrichment analyses to compare individuals with moderate hypoxaemia with those with CI.
Findings
We found that moderate hypoxaemia at the 5-year follow-up visit, but not at the baseline visit, is associated with CI as assessed at the 10-year follow-up visit (OR 1.49, 95% confidence interval 1.10–1.99). In addition, moderate hypoxaemia significantly mediates the association between COPD and CI (average causal-mediated effect, 27.3%; p<0.001). Out of 38 proteins previously associated with cerebral hypoxia, 13 were significantly associated with moderate hypoxaemia.
Interpretation
Our study establishes moderate hypoxaemia as an important risk factor for CI in individuals with a history of cigarette smoke exposure and as a mediator of the risk of CI in individuals with COPD. We also identify multi-omic biomarkers that better characterise the biological pathways underlying the association between moderate hypoxaemia and CI. Future studies are needed to identify individuals with moderate hypoxaemia who are at the highest risk of CI.
Shareable abstract
Moderate hypoxaemia is an important risk factor for cognitive impairment in individuals with cigarette smoke exposure and mediator of the risk of cognitive impairment in individuals with COPD https://bit.ly/4yfjZzb
Introduction
Despite the respiratory focus of the diagnosis and management of COPD, there is increasing recognition of its systemic effects [1]. Several non-respiratory comorbidities are associated with COPD, including skeletal muscle dysfunction, cardiovascular disease and peripheral vascular disease [2]. In addition, with the overall ageing of the population, COPD has emerged as an important potential contributor to the development of cognitive impairment (CI) [3]. Previous studies have estimated that, individuals with COPD have almost 2.5-times higher odds of developing CI than individuals without COPD [4]. Despite the strength of the association, specific risks factors for CI in individuals with COPD and the underlying pathobiological mechanisms remain unclear.
The development of CI in individuals with COPD is likely to be a result of several factors, potentially in combination [5]. The relationship between severe hypoxaemia (peripheral oxygen saturation (SpO2) ≤88%) and CI is well established, whereby individuals with COPD and severe hypoxaemia have been shown to have over fivefold higher odds of CI than individuals with COPD and normal oxygenation [4]. On the other hand, less is known about moderate hypoxaemia (SpO2 89–93%) as a mediator of CI risk in COPD. Furthermore, the burden of moderate hypoxaemia in individuals with COPD remains unclear. While the prevalence of severe hypoxaemia has been noted to be about 7% in individuals with COPD, the prevalence of moderate hypoxaemia is not established [6].
In the present study, we aimed to determine if pre-existing moderate hypoxaemia is a risk factor for, and potential mediator of, future CI in individuals with COPD and a history of cigarette smoke exposure. To test our hypothesis, we used clinical and multi-omic data from the Genetic epidemiology of COPD (COPDGene) study [7]. We first compared individuals with moderate hypoxaemia and normal oxygenation by demographics, health history and COPD-related characteristics. Next, we compared the proportions of individuals with CI between individuals with moderate hypoxaemia and those with normal oxygenation to establish the association between moderate hypoxaemia and CI. We then compared proteomic biomarkers of cerebral hypoxia between individuals with moderate hypoxaemia and those with normal oxygenation to further demonstrate the biological impact of hypoxaemia. Finally, we conducted transcriptomic analyses using blood RNA sequencing (RNA-Seq) data to identify brain-related gene sets and biological pathways affected by moderate hypoxaemia.
Methods
Study participants
The COPDGene study was a multicentre, prospective, observational cohort study designed to identify epidemiologic and multi-omic factors associated with COPD and related outcomes [7]. The study recruited >10 000 current and former smokers with at least 10 smoking pack-years aged 45–80 years across 21 clinical centres in the USA. Participants were followed up longitudinally across three visits: baseline visit (Phase 1), 5-year follow-up (Phase 2) and 10-year follow-up (Phase 3). Study assessments comprised the collection of demographic characteristics, smoking status, health history, spirometry and chest computed tomography (CT) scans across all three visits. Blood samples for proteomic analysis and RNA-Seq were collected at the Phase 2 visit. The Mini-Cog assessment was performed at the Phase 3 visit. All participants provided written informed consent. The COPDGene study was approved by institutional review boards at all participating centres (Mass General Brigham Institutional Review Board (IRB) #2007P000554).
Definitions
We measured cognitive function at the Phase 3 visit using the Mini-Cog assessment. The Mini-Cog combines a three-item recall task and a clock-drawing test; total scores range from zero to five [8, 9]. We defined CI as a Mini-Cog score of ≤3, in line with multiple prior large studies [10–12]. We used pulse oximetry to measure oxygenation, where SpO2≥93% was categorised as normal oxygenation, SpO2 between 89% and 93% was categorised as moderate hypoxaemia and SpO2≤88% was categorised as severe hypoxaemia, consistent with prior clinical trials of oxygen supplementation [13, 14]. Given the established connection between acute hypoxaemia and impaired cognitive function, our focus was to interrogate the effects of hypoxaemia preceding cognitive assessment, rather than hypoxaemia detected concurrently. We defined COPD as a forced expiratory volume in 1 s (FEV1) to forced vital capacity ratio of <0.70, as per the Global Initiative for Chronic Obstructive Lung Disease (GOLD) guidelines. We defined moderate or more severe COPD as GOLD stage ≥2.
Proteomic measurement and RNA sequencing
The details of proteomic measurements and RNA-Seq in COPDGene have been described previously [15, 16]. Briefly, plasma proteomic profiling was conducted during Phase 2 of the COPDGene study by analysing blood samples from participants using the SomaScan® version 4.0 assay. Standard processing included hybridisation normalisation, within-plate median signal normalisation, calibration, plate scaling and median normalisation to a reference using adaptive normalisation by maximum likelihood [15]. Whole-blood RNA-Seq was also performed at Phase 2 of the COPDGene study using samples collected in PAXgene Blood RNA Tubes. RNA-Seq was performed according to sample quality criteria, including availability of high-quality RNA (RNA integrity number >6) and RNA concentration of ≥10 μg·μL−1. Reads were generated using an Illumina HiSeq 2500. FastQC was used to perform quality control. Alignment to the GRCh38 reference genome was performed using Spliced Transcripts Alignment to a Reference (STAR).
Statistical analysis
Association between moderate hypoxaemia and cognitive impairment
The overview of the study design is illustrated in figure 1. We included only individuals who had completed the Mini-Cog at the Phase 3 visit. We compared demographics, health history, comorbidity count (further described in the supplementary methods) and COPD-related characteristics between individuals with moderate hypoxaemia and those with normal oxygenation at Phase 1, Phase 2 and Phase 3. We used t-tests for continuous variables and chi-squared tests for proportions. We used multivariable logistic regression models to test for an association between hypoxaemia at the baseline, 5-year (Phase 2) and 10-year (Phase 3) visits, measured both continuously and categorised as normal, moderate hypoxaemia or severe hypoxaemia, and the presence or absence of CI, as the dependent variable, at the 10-year (Phase 3) visit. Models were adjusted for age, race, gender, smoking pack-years, smoking status, level of education, use of oxygen supplementation and FEV1 % predicted at Phase 3. We also performed mediation analysis, adjusted for age, race, gender and level of education to estimate the average causal-mediated effects of moderate hypoxaemia on the association of moderate COPD (GOLD stage ≥2) with CI, using the mediation R package. Statistical analyses were performed in R version 4.5.1.
FIGURE 1.
Overview of study design. Panel 1 shows a directed acyclic graph demonstrating the association of COPD and moderate hypoxaemia with cognitive impairment, and the role of moderate hypoxaemia as a causal mediator of the association between COPD with CI. Panel 2 shows representative plots of a volcano plot from differential expression analysis and a heat map from pathway enrichment analysis. Created in BioRender (Ghosh, 2026), https://BioRender.com/b03hjvx.
Proteomic biomarkers of cerebral hypoxia in moderate hypoxaemia and cognitive impairment
In order to further establish hypoxaemia as a potential driver of CI via cerebral hypoxia, we compared the proteomic biomarkers of cerebral hypoxia from individuals with moderate hypoxaemia with those from individuals with normal oxygenation, as well as comparing them between those with and those without CI. We performed a literature review to identify proteomic biomarkers of cerebral hypoxia by searching PubMed and Google Scholar databases for studies published in any language from database inception to 1 July 2025, using a combination of keywords including “cerebral hypoxia”, “hypoxic brain injury”, “anoxic brain injury”, “proteomic”, “protein” and “biomarker” [17–21]. We included proteomic biomarkers of cerebral hypoxia available in the SomaScan 5K assay. We compared log2 transformed means of the proteomic biomarkers of cerebral hypoxia using t-tests. We used a Bonferroni correction for multiple testing, with p≤0.001 considered significant.
Pathway enrichment analyses
We performed gene-set enrichment analyses on both measured blood gene expression and imputed gene expression from each of the 10 imputed brain regions (amygdala, anterior cingulate cortex, caudate, cerebellum, frontal cortex, hippocampus, hypothalamus, nucleus accumbens, putamen, substantia nigra) using the Brain Gene Expression and Network Imputation Engine (BrainGENIE). Further details regarding BrainGENIE and differential expression analyses are presented in the supplementary methods. Weights for pathway enrichment were derived from differential expression analyses using the piano package. We used gene sets curated by the Reactome and Gene Ontology databases. Enrichment p-values were adjusted using the Benjamini–Hochberg false discovery rate procedure to correct for the number of gene sets tested within each module, with a significance threshold set at a false discovery rate of <5%. Distinctly upregulated and distinctly downregulated gene sets were isolated within blood and each brain region. Reactome gene set sub-pathways were compiled according to biological “root pathways”.
Polytranscript risk scores for Alzheimer's disease
Finally, we calculated previously published blood-based and brain region-specific polytranscript risk score (PTRS) profiles for Alzheimer's disease to compare individuals with moderate hypoxaemia with those with normal oxygenation [22]. While our intention was not necessarily to establish moderate hypoxaemia as a risk factor for Alzheimer's disease, we endeavoured to further link moderate hypoxaemia with biomarkers for poor brain health. In the absence of a transcriptome-based risk score for CI, we used the PTRS for Alzheimer's disease as an approximation. The PTRSs are computed as weighted sums of differentially expressed genes, whereby the genes are selected based on the level of significance from differential expression analyses. We complied tissue-specific PTRS analyses at six p-value thresholds of weighted gene expression (p=0.001, 0.01, 0.05, 0.1, 0.5, 1.0). Each analysis results in a single p-value that summarises the association of each tissue-specific PTRS profile with Alzheimer's disease.
Results
Participant characteristics
Among the 3246 study participants who completed the Mini-Cog at the Phase 3 (10-year follow-up) visit, we identified 135 individuals (4.2%) with moderate hypoxaemia at the Phase 1 visit and 153 individuals (5.1%) with moderate hypoxaemia at the Phase 2 visit. Compared with individuals with normal oxygenation, individuals with moderate hypoxaemia were older, were less likely to be African American, had higher smoking pack-years and were less likely to be current smokers at both Phase 1 and Phase 2 (table 1). At Phase 1, individuals with moderate hypoxaemia also had a higher comorbidity count. There was no difference in sex or level of education at either visit between the two groups.
TABLE 1.
Participant characteristics of moderate hypoxaemia versus normal oxygenation at Phase 1 (baseline) and Phase 2 (5-year follow-up) visits
| Characteristic | Phase 1 | Phase 2 | ||||
|---|---|---|---|---|---|---|
| Moderate hypoxaemia | Normal oxygenation | p-value | Moderate hypoxaemia | Normal oxygenation | p-value | |
| Participants, n | 135 | 3111 | 153 | 2820 | ||
| Age, years | 62.65±8.47 | 58.23±8.19 | <0.001 | 67.58±8.66 | 64.17±8.16 | <0.001 |
| Female sex | 63 (46.7) | 1585 (50.9) | 0.375 | 78 (51.0) | 1450 (51.4) | 0.982 |
| African American | 25 (18.5) | 1002 (32.2) | 0.001 | 26 (17.0) | 851 (30.2) | 0.001 |
| Smoking pack-years | 51.80±24.55 | 40.27±22.00 | <0.001 | 51.34±26.44 | 41.72±22.53 | <0.001 |
| Current smoker | 38 (28.1) | 1514 (48.7) | <0.001 | 34 (22.2) | 1048 (37.2) | <0.001 |
| High school completed | 123 (91.1) | 2781 (89.4) | 0.622 | 140 (91.5) | 2548 (90.4) | 0.742 |
| Comorbidity count | 2.91±1.93 | 2.30±1.78 | <0.001 | |||
| GOLD stage | <0.001 | <0.001 | ||||
| PRISm | 10 (7.4) | 367 (11.9) | 11 (7.4) | 340 (12.2) | ||
| 0 | 32 (23.7) | 1689 (54.6) | 34 (22.8) | 1440 (51.6) | ||
| 1 | 10 (7.4) | 292 (9.4) | 11 (7.4) | 271 (9.7) | ||
| 2 | 39 (28.9) | 530 (17.1) | 35 (23.5) | 503 (18.0) | ||
| 3 | 35 (25.9) | 186 (6.0) | 39 (26.2) | 191 (6.8) | ||
| 4 | 9 (6.7) | 27 (0.9) | 19 (12.8) | 48 (1.7) | ||
| Cognitive impairment | 38 (28.1) | 752 (24.2) | 0.341 | 56 (36.6) | 651 (23.1) | <0.001 |
Data are presented as mean±sd or n (%), unless otherwise stated. GOLD: Global Initiative for Chronic Obstructive Lung Disease; PRISm: preserved ratio impaired spirometry.
We also compared COPD-related characteristics between individuals with moderate hypoxaemia and those with normal oxygenation. At both Phase 1 and Phase 2, individuals with moderate hypoxaemia had lower FEV1 % predicted, higher modified Medical Research Council (mMRC) Dyspnoea Scale and St. George's Respiratory Questionnaire total scores and worse radiographic markers of disease on lung CT (supplementary table 1). There was no difference at either visit in decrease in FEV1 over time between the two groups.
Association between moderate hypoxaemia and cognitive impairment
We found that there was no difference in the proportion of individuals with CI at Phase 3 between those with moderate hypoxaemia and those with normal oxygenation at Phase 1 or Phase 3. However, the proportion of individuals with CI at Phase 3 was greater among individuals with moderate hypoxaemia than among those with normal oxygenation at Phase 2 (figure 2). In multivariable logistic regression models, adjusted for age, race, gender, level of education and FEV1 % predicted, moderate hypoxaemia at Phase 2, but not at Phase 1 or Phase 3, was significantly associated with CI (OR 1.49, 95% confidence interval 1.10–1.99). There was no association between SpO2 measured continuously and CI at either visit.
FIGURE 2.
Proportions of individuals with and individuals without cognitive impairment (CI) at Phase 3 (10-year follow-up visit) among individuals with moderate hypoxaemia and those with normal oxygenation at Phase 1 and Phase 2.
In addition, we found that moderate hypoxaemia significantly mediated the direct effect of COPD on CI. After confirming that moderate or more severe COPD (i.e. GOLD stage ≥2) was associated with CI (OR 1.21, 95% confidence interval 1.02–1.45), we calculated the average causal-mediated effect (ACME). The ACME represents the proportion of the total effect of COPD on CI that is attributable to the mediator, which in this case is moderate hypoxaemia. We found the ACME to be 27.3% (p<0.001) (figure 3).
FIGURE 3.
Effect decomposition plot of moderate hypoxaemia as a mediator of COPD on cognitive impairment. Solid lines represent effect with moderate COPD and dotted lines represent effect without moderate COPD.
Proteomic biomarkers of cerebral hypoxia
In our literature review, we identified 38 proteomic biomarkers associated with cerebral hypoxia that were also available on the SomaScan 5K assay. After correction for multiple testing (Bonferroni-adjusted p≤0.001), we found that 13 of the 38 proteins were significantly different between individuals with moderate hypoxaemia and individuals with normal oxygenation at Phase 2, including α-enolase (ENO1), vascular endothelial growth factor C (VEGF-C), brain-derived neurotrophic factor (BDNF), insulin-like growth factor-binding protein 6 (IGFBP-6), heparin-binding EGF-like growth factor (HBEGF), lactate dehydrogenase B subunit (LDHB), glutamic–oxaloacetic transaminase 1 (GOT1), U6 snRNA-associated Sm-like protein (LSM1) and signal recognition particle 14 (SRP14) (table 2). By contrast, we found three proteins that were significantly different between individuals with and individuals without CI (supplementary table 2). At Phase 3, five (ENO1, BDNF, tissue inhibitor of metalloproteinases 1 (TIMP1), LSM1 and LDHB) of the 13 proteins that were significantly different at Phase 2 were significantly different between individuals with moderate hypoxaemia and those with normal oxygenation (supplementary table 3).
TABLE 2.
Log2 fold change of proteomic biomarkers of cerebral hypoxia between individuals with moderate hypoxaemia and those with normal oxygenation at Phase 2
| Protein | Fold change | p-value |
|---|---|---|
| α-enolase (ENO1) | 0.9637 | <0.001 |
| Vascular endothelial growth factor C (VEGF-C) | 0.9795 | <0.001 |
| Brain-derived neurotrophic factor (BDNF) | 0.9604 | <0.001 |
| Insulin-like growth factor-binding protein 6 (IGFBP6) | 1.0055 | <0.001 |
| Heparin-binding EGF-like growth factor (HBEGF) | 0.9670 | <0.001 |
| Lactate dehydrogenase B subunit (LDHB) | 0.9738 | <0.001 |
| Glutamic–oxaloacetic transaminase 1 (GOT1) | 0.9908 | <0.001 |
| U6 snRNA-associated Sm-like protein (LSM1) | 0.9691 | <0.001 |
| Signal recognition particle 14 (SRP14) | 0.9855 | <0.001 |
| Interleukin 6 (IL6) | 0.9741 | 0.001 |
| Tissue inhibitor of metalloproteinases 1 (TIMP1) | 1.0064 | 0.001 |
| Ubiquitin carboxy-terminal hydrolase L1 (UCHL1) | 0.9825 | 0.001 |
| C-reactive protein (CRP) | 1.0232 | 0.001 |
| NADH:ubiquinone oxidoreductase subunit S4 (NDUFS4) | 1.0196 | 0.004 |
| γ-enolase (ENO2) | 0.9876 | 0.005 |
| Insulin-like growth factor-binding protein 5 (IGFBP5) | 0.9899 | 0.012 |
| Erythropoietin (EPO) | 1.0117 | 0.016 |
| Matrix metalloproteinase 9 (MMP9) | 1.0092 | 0.032 |
| Creatine kinase B (CKB) | 0.9901 | 0.085 |
| Microtubule-associated protein τ (MAPT) | 0.9935 | 0.098 |
| Glial fibrillary acidic protein (GFAP) | 1.0086 | 0.106 |
| Defensin α1 (DEFA1) | 1.0044 | 0.11 |
| Cytochrome c oxidase subunit 6C (COX6C) | 1.0032 | 0.176 |
| Oxidised low-density lipoprotein receptor 1 (OLR1) | 1.0072 | 0.218 |
| Arylsulfatase B (ARSB) | 0.9953 | 0.256 |
| Neurofilament heavy chain (NEFH) | 1.0056 | 0.294 |
| Glutathione-disulfide reductase (GSR) | 1.0021 | 0.333 |
| Fas ligand (FASLG) | 1.0042 | 0.337 |
| Interleukin 1β (IL1B) | 0.9971 | 0.386 |
| Erythropoietin receptor (EPOR) | 1.0021 | 0.437 |
| Fibulin 1 (FBLN1) | 1.0013 | 0.446 |
| Intercellular adhesion molecule 1 (ICAM1) | 0.9964 | 0.448 |
| β-enolase (ENO3) | 0.9971 | 0.743 |
| Insulin-link growth factor-binding protein 4 (IGFBP4) | 1.0009 | 0.744 |
| Neuronal cell adhesion molecule (NRCAM) | 0.9988 | 0.753 |
| Neurofilament light chain (NEFL) | 1.0010 | 0.859 |
| Insulin-like growth factor-binding protein 1 (IGFBP1) | 1.0008 | 0.893 |
| Chitinase 3 like 1 (CHI3L1) | 1.0008 | 0.937 |
EGF: epidermal growth factor.
Pathway enrichment analyses
We performed pathway enrichment analyses using gene expression measured from blood and imputed from 10 brain regions in both moderate hypoxaemia and CI, with weights derived from differential expression analyses (see supplementary results) (figure 4). While there were no downregulated pathways in blood gene expression from individuals with moderate hypoxaemia, the immune system pathway was upregulated in blood gene expression in individuals with moderate hypoxaemia. Conversely, there were no upregulated pathways in blood gene expression in individuals with CI, but the cell cycle pathway was downregulated. There were several significantly enriched pathways in the imputed gene expression from the 10 brain regions in both individuals with moderate hypoxaemia and those with CI.
FIGURE 4.
Gene set enrichment heat map results for differential gene expression in cognitive impairment (CI) and moderate hypoxaemia. a) The pathway enrichment results between individuals with and those without CI. b) The pathway enrichment results between individuals with moderate hypoxaemia and those with normal oxygenation. ACC: anterior cingulate cortex; ADME: absorption, distribution, metabolism and excretion; AMY: amygdala; CH: cerebellum; FC: frontal cortex; NAC: nucleus accumbens; PUT: putamen; SN: substantia nigra; CAU: caudate; HIP: hippocampus; HYP: hypothalamus.
Polytranscript risk scores for Alzheimer's disease
We computed PTRSs for Alzheimer's disease using tissue-specific weights by means of blood gene expression and imputed brain gene expression across the 10 brain regions. We computed six PTRSs in each tissue corresponding to the p-value thresholds for genes selected for inclusion in the PTRS. We found that the p<0.001 PTRS for Alzheimer's disease generated using blood gene expression was significantly associated with moderate hypoxaemia (figure 5). All six p-value thresholds for the PTRSs for Alzheimer's disease generated using imputed gene expression in the anterior cingulate cortex were significantly associated with moderate hypoxaemia. Three of the p-value thresholds (p<0.1, p<0.5 and p<1.0) for the PTRSs for Alzheimer's disease generated using imputed gene expression in the hypothalamus were significantly associated with moderate hypoxaemia. No other p-value thresholds for the PTRSs for Alzheimer's disease were significantly associated with moderate hypoxaemia in the remaining brain regions.
FIGURE 5.
Alzheimer's disease polytranscript risk scores (PTRSs) in individuals with moderate hypoxaemia. Each bar represents a different PTRS computed as a weighted sum of differentially expressed genes, whereby the genes are selected based on the level of significance from differential expression analyses. ACC: anterior cingulate cortex; AMY: amygdala; CAU: caudate; CH: cerebellum; FC: frontal cortex; HIP: hippocampus; HYP: hypothalamus; NAC: nucleus accumbens; PUT: putamen; SN: substantia nigra.
Discussion
In this study, we demonstrated the association between moderate hypoxaemia and the presence of CI at a 5-year, but not a 10-year, follow-up interval in individuals with a history of cigarette smoke exposure. In addition, we found that moderate hypoxaemia is an important mediator of the association between COPD and CI. Furthermore, we showed biological changes relevant to cognitive function in peripheral proteomic and transcriptomic biomarkers as well as imputed brain region-specific transcriptomes among individuals with moderate hypoxaemia. Finally, we found that gene expression profiles from individuals with moderate hypoxaemia were significantly associated with polytranscript risk for Alzheimer's disease in blood and two out of 10 brain regions. Taken together, our results highlight moderate hypoxaemia in individuals with COPD and cigarette smoke exposure as a driver of CI and a potentially treatable trait.
The distinction between severe and moderate hypoxaemia traces back to clinical trials for oxygen supplementation. The original trials from the 1970s established the mortality benefit for oxygen supplementation in individuals with severe hypoxaemia. More recent trials have suggested that there is no mortality benefit for individuals with moderate hypoxaemia [13, 14]. However, despite the pre-eminence of mortality as an outcome in oxygen supplementation trials, CI is nonetheless an outcome of great importance to ageing individuals [23]. It remains possible that oxygen supplementation could benefit individuals with moderate hypoxaemia by mitigating risk for CI. However, given the high costs associated with oxygen supplementation and the history of negative trials in individuals with moderate hypoxaemia, further work is needed to better characterise individuals with moderate hypoxaemia at the highest risk of CI [24]. In addition, our finding that medium-term hypoxaemia, but not short-term or longer-term hypoxaemia, is associated with CI will require additional investigation to precisely define the at-risk window and when intervention is most likely to lead to improved outcomes. Our results set the stage for a future study incorporating both biomarkers and potentially neuroimaging to further refine the endotype most likely to benefit from supplemental oxygen.
Accordingly, we identified a number of candidates among both blood-based proteomic and transcriptomic biomarkers. While some of the biomarkers, such as C-reactive protein (CRP), were nonspecific, several biomarkers were specific for neurological pathways [25]. For example, levels of brain BDNF were significantly reduced in individuals with moderate hypoxaemia. BDNF promotes survival of neurons and reduced BDNF levels are associated with Alzheimer's disease, Parkinson's disease and other neurodegenerative diseases [26, 27]. Importantly, however, the magnitude of difference between most of the proteomic and transcriptomic biomarkers was small, and some of the directions of effect were not in the expected direction (i.e., ENO1 and γ-enolase (ENO2) reduced in individuals with moderate hypoxaemia) [28]. We anticipate that future studies that aggregate these findings, in a proteomic or multi-omic risk score, may ameliorate these issues.
Our study has several strengths. First, establishing moderate hypoxaemia as a mediator of the association between COPD and CI moves our understanding of the relationship between the two conditions forward. Second, our analyses are among the first to employ multi-omic biomarkers to better characterise the underlying pathobiology of CI in individuals with COPD and a history of cigarette smoke exposure [29]. Finally, in addition to CI, we demonstrate that moderate hypoxaemia is associated with PTRSs that have been previously derived for Alzheimer's disease in multiple tissue compartments, further suggesting a connection between COPD, cigarette smoke exposure and neurological diseases. Despite these many strengths, our study has limitations. First, we characterised the multi-omic changes in a single study cohort only. The lack of replication is primarily due to limited availability of well-characterised studies that include the required detail of COPD-related phenotypes, cognitive function, and proteomics and transcriptomics. Second, our assessment of cognitive function was available only at Phase 3 (the 10-year follow-up visit), whereas the majority of the associations, including multi-omics, were conducted at Phase 1 and Phase 2. We were therefore unable to determine if CI was incident or existed at previous visits. However, the strength of the association of moderate hypoxaemia at Phase 2 is compelling.
Given the expected increase in both the number of individuals affected by COPD and the size of the ageing population, it is imperative to develop a better understanding of comorbidities associated with both COPD and advanced age, such as CI. Our findings suggest that moderate hypoxaemia, in addition to the well-established association of severe hypoxaemia, accounts for an important proportion of risk of CI associated with COPD. Future studies are needed to identify individuals with moderate hypoxaemia related to COPD and a history of cigarette smoke exposure who are at the highest risk of CI and who, in turn, are most likely to benefit from oxygen supplementation.
Acknowledgements
COPDGene was supported by the US National Heart, Lung, and Blood Institute (NHLBI) (grants U01 HL089897 and U01 HL089856) and by National Institutes of Health (NIH) contract 75N92023D00011. COPDGene is also supported by the COPD Foundation through contributions made to an Industry Advisory Board comprising AstraZeneca, Boehringer Ingelheim, Genentech, GSK, Novartis, Pfizer, Siemens and Sunovion.
Footnotes
Provenance: Submitted article, peer reviewed.
Ethics statement: The COPDGene study was approved by institutional review boards at all participating centres (Mass General Brigham IRB # 2007P000554).
Author contributions: Study conceptualisation: T.M. Mazza, C. Hill, D.J. Burke, K.F. Hoth and A.J. Ghosh; data acquisition: R.P. Bowler, C.P. Hersch and K.F. Hoth; data analysis: T.M. Mazza, C. Hill, D. Pergel, D.J. Burke, K.A. Pratte, C.P. Hersch, S.J. Glatt, J.L. Hess, K.F. Hoth and A.J. Ghosh; statistical support: T.M. Mazza, C. Hill, D. Pergel, K.A. Pratte, C.P. Hersch, S.J. Glatt, J.L. Hess and A.J. Ghosh. All authors were responsible for the critical revision of the manuscript for important intellectual content.
Conflict of interest: K.A. Pratte reports support for the present manuscript from NHLBI (grants U01 HL089897 and U01 HL089856) and NIH contract 75N92023D00011. C.P. Hersh reports support for the present manuscript from NHLBI; grants received from the Alpha-1 Foundation and Bayer; and consultancy fees from Apogee Therapeutics, AstraZeneca, Chiesi, Genentech, Ono Pharma, Sanofi, Takeda and Verona Pharma. J.L. Hess reports support for the present manuscript from the National Institute of Neurological Disorders and Stroke (5R01NS128535-02) and the Central New York (CNY) Community Foundation. K.F. Hoth reports support for the present manuscript from NIH/NHLBI; grants received from NIH/NHLBI, the Centers for Disease Control and Prevention and the Ann Theodore Foundation; support for attending meetings and/or travel from NIH/NHLBI; and a leadership role with the COPD Foundation. A.J. Ghosh reports grants from NIH/NHLBI, and consultancy fees from TDA Research, Inc., Grifols and Casdin Capital. The remaining authors have nothing to disclose.
Support statement: R.P. Bowler is supported by grant R01HL137995. C.P. Hersch is supported by K24HL173667. J.L. Hess is supported by R01NS128535 and the CNY Community Foundation. A.J. Ghosh is supported by K08HL168205. Funding information for this article has been deposited with the Open Funder Registry.
Supplementary material
Please note: supplementary material is not edited by the Editorial Office, and is uploaded as it has been supplied by the author.
Supplementary material
01562-2025.SUPPLEMENT
Supplementary tables
01562-2025.SUPPLEMENT
References
- 1.Barnes PJ, Celli BR. Systemic manifestations and comorbidities of COPD. Eur Respir J 2009; 33: 1165–1185. doi: 10.1183/09031936.00128008 [DOI] [PubMed] [Google Scholar]
- 2.Xiang Y, Luo X. Extrapulmonary comorbidities associated with chronic obstructive pulmonary disease: a review. Int J Chron Obstruct Pulmon Dis 2024; 19: 567–578. doi: 10.2147/COPD.S447739 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Singh B, Parsaik AK, Mielke MM, et al. Chronic obstructive pulmonary disease and association with mild cognitive impairment: the Mayo Clinic study of aging. Mayo Clin Proc 2013; 88: 1222–1230. doi: 10.1016/j.mayocp.2013.08.012 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Thakur N, Blanc PD, Julian LJ, et al. COPD and cognitive impairment: the role of hypoxemia and oxygen therapy. Int J Chron Obstruct Pulmon Dis 2010; 5: 263–269. doi: 10.2147/copd.s10684 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Panda S, Walsh M, Anghel I, et al. Mechanisms and risk factors of cognitive impairment in chronic obstructive pulmonary disease. CHEST Pulmonary 2025; 4: 100196. doi: 10.1016/j.chpulm.2025.100196 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Sundh J, Ekström M. Risk factors for developing hypoxic respiratory failure in COPD. Int J Chron Obstruct Pulmon Dis 2017; 12: 2095–2100. doi: 10.2147/COPD.S140299 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Regan EA, Hokanson JE, Murphy JR, et al. Genetic epidemiology of COPD (COPDGene) study design. COPD 2011; 7: 32–43. doi: 10.3109/15412550903499522 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Borson S, Scanlan J, Brush M, et al. The Mini-Cog: a cognitive “vital signs” measure for dementia screening in multi-lingual elderly. Int J Geriatr Psychiatry 2000; 15: 1021–1027. doi: 10.1002/1099-1166(200011)15:11<1021::AID-GPS234>3.0.CO;2-6 [DOI] [PubMed] [Google Scholar]
- 9.Borson S, Scanlan JM, Chen P, et al. The Mini-Cog as a screen for dementia: validation in a population-based sample. J Am Geriatr Soc 2003; 51: 1451–1454. doi: 10.1046/j.1532-5415.2003.51465.x [DOI] [PubMed] [Google Scholar]
- 10.Cordell CB, Borson S, Boustani M, et al. Alzheimer's Association recommendations for operationalizing the detection of cognitive impairment during the Medicare Annual Wellness Visit in a primary care setting. Alzheimers Dement 2013; 9: 141–150. doi: 10.1016/j.jalz.2012.09.011 [DOI] [PubMed] [Google Scholar]
- 11.McCarten JR, Anderson P, Kuskowski MA, et al. Screening for cognitive impairment in an elderly veteran population: acceptability and results using different versions of the Mini-Cog. J Am Geriatr Soc 2011; 59: 309–313. doi: 10.1111/j.1532-5415.2010.03249.x [DOI] [PubMed] [Google Scholar]
- 12.Rosenbloom M, Borson S, Barclay T, et al. Routine cognitive screening in a neurology practice: effect on physician behavior. Neurol Clin Pract 2016; 6: 16–21. doi: 10.1212/CPJ.0000000000000186 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Continuous or nocturnal oxygen therapy in hypoxemic chronic obstructive lung disease: a clinical trial. Nocturnal Oxygen Therapy Trial Group. Ann Intern Med 1980; 93: 391–398. doi: 10.7326/0003-4819-93-3-391 [DOI] [PubMed] [Google Scholar]
- 14.Long-Term Oxygen Treatment Trial Research Group ; Albert RK, Au DH, Blackford AL, et al. A randomized trial of long-term oxygen for COPD with moderate desaturation. N Engl J Med 2016; 375: 1617–1627. doi: 10.1056/NEJMoa1604344 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Serban KA, Pratte KA, Strange C, et al. Unique and shared systemic biomarkers for emphysema in alpha-1 antitrypsin deficiency and chronic obstructive pulmonary disease. EBioMedicine 2022; 84: 104262. doi: 10.1016/j.ebiom.2022.104262 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Parker MM, Chase RP, Lamb A, et al. RNA sequencing identifies novel non-coding RNA and exon-specific effects associated with cigarette smoking. BMC Med Genomics 2017; 10: 58. doi: 10.1186/s12920-017-0295-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Caramelo I, Coelho M, Rosado M, et al. Biomarkers of hypoxic–ischemic encephalopathy: a systematic review. World J Pediatr 2023; 19: 505–548. doi: 10.1007/s12519-023-00698-737084165 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Stein DM, Lindell AL, Murdock KR, et al. Use of serum biomarkers to predict cerebral hypoxia after severe traumatic brain injury. J Neurotrauma 2012; 29: 1140–1149. doi: 10.1089/neu.2011.2149?download=true22360297 [DOI] [PubMed] [Google Scholar]
- 19.Ahmad Y, Sharma NK, Ahmad MF, et al. The proteome of hypobaric induced hypoxic lung: insights from temporal proteomic profiling for biomarker discovery. Sci Rep 2015; 5: 1–20. doi: 10.1038/srep10681 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Hoiland RL, Rikhraj KJK, Thiara S, et al. Neurologic prognostication after cardiac arrest using brain biomarkers: a systematic review and meta-analysis. JAMA Neurol 2022; 79: 390–398. doi: 10.1001/jamaneurol.2021.5598 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Seo M, Qiu W, Bailey W, et al. Genomics and response to long-term oxygen therapy in chronic obstructive pulmonary disease. J Mol Med 2018; 96: 1375–1385. doi: 10.1007/s00109-018-1708-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Hou J, Hess J, Glatt S. Transcriptomic score: a biomarker index of Alzheimer's risk and resilience. Alzheimer Dement 2021; 17: e050343. doi: 10.1002/alz.050343 [DOI] [Google Scholar]
- 23.Morley JE, Morris JC, Berg-Weger M, et al. Brain health: the importance of recognizing cognitive impairment: an IAGG consensus conference. J Am Med Dir Assoc 2015; 16: 731–739. doi: 10.1016/j.jamda.2015.06.017 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Kahn PA, Worsham CM, Berland G. Characterization of prescription patterns and estimated costs for use of oxygen concentrators for home oxygen therapy in the US. JAMA Netw Open 2021; 4: e2129967. doi: 10.1001/jamanetworkopen.2021.29967 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Dahl M, Vestbo J, Lange P, et al. C-reactive protein as a predictor of prognosis in chronic obstructive pulmonary disease. Am J Respir Crit Care Med 2007; 175: 250–255. doi: 10.1164/rccm.200605-713OC [DOI] [PubMed] [Google Scholar]
- 26.Scalzo P, Kümmer A, Bretas TL, et al. Serum levels of brain-derived neurotrophic factor correlate with motor impairment in Parkinson's disease. J Neurol 2010; 257: 540–545. doi: 10.1007/s00415-009-5357-2 [DOI] [PubMed] [Google Scholar]
- 27.Bathina S, Das UN. Brain-derived neurotrophic factor and its clinical implications. Arch Med Sci 2015; 11: 1164–1178. doi: 10.5114/aoms.2015.56342 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Schoerkhuber W, Kittler H, Sterz F, et al. Time course of serum neuron-specific enolase. Stroke 1999; 30: 1598–1603. doi: 10.1161/01.STR.30.8.1598 [DOI] [PubMed] [Google Scholar]
- 29.Shim TS, Lee JH, Kim SY, et al. Cerebral metabolic abnormalities in COPD patients detected by localized proton magnetic resonance spectroscopy. Chest 2001; 120: 1506–1513. doi: 10.1378/chest.120.5.1506 [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Please note: supplementary material is not edited by the Editorial Office, and is uploaded as it has been supplied by the author.
Supplementary material
01562-2025.SUPPLEMENT
Supplementary tables
01562-2025.SUPPLEMENT