Acute exacerbations of COPD (AECOPD) are episodes of acute worsening of respiratory symptoms that require additional therapy. Exacerbations play a pivotal role in the burden and progressive course of COPD (1). Each event contributes to a progressive decline in lung function (2), reduced health status, low physical activity level (3) and increased health care costs (4). As such, disease management is predominantly based on the prevention of these episodes (1). Yet, in the Netherlands, 30.000 people are admitted to the hospital for an AECOPD every year (5). Although most AECOPD have an infectious origin (6), the underlying mechanisms are heterogeneous and predicting their occurrence in individual patients currently remains unsuccessful (7-9). Furthermore, there is a lack of our understanding in the longitudinal alterations in microbial composition and host-microbiome interactions in the stable state, at AECOPD and during recovery in patients with COPD. This knowledge is essential to improve the early and accurate diagnosis of (the different types of) AECOPD, and for the development of novel antimicrobial and other therapeutic targets and subsequent personalized treatment. These challenges need to be addressed in order to reduce the future impact of these events, avoid unnecessary treatments of individual patients, reduce healthcare utilization and improve overall care for patients with COPD. The current 'Early diagnostic BioMARKers in Exacerbations of COPD' (MARKED) study was designed to investigate several of these gaps in the management of COPD exacerbations.

It is anticipated that complex biomarker panels, rather than a single biomarker, will be identified. Since AECOPD are heterogeneous events in terms of origin, trigger, severity, duration, need for treatment and overall clinical presentation (1, 6, 10-15), we expect to identify different biomarker panels for different subtypes of AECOPD. Furthermore, AECOPD diagnosis relies heavily on the exclusion of differential diagnoses (1), which further rules out the potential of a single predictive AECOPD biomarker.

The MARKED study is an exploratory, prospective, single-center, longitudinal, observational study with eight weeks follow-up. The primary objective is to explore which biomarkers from a panel of frequently measured biomarkers (symptoms, vital signs, lung function parameters and spontaneous sputum, nasopharyngeal swabs, stool and blood samples) predict an exacerbation and/or respiratory infection in patients with COPD. Furthermore, secondary objectives are:

I. to investigate longitudinal alterations in microbial composition and host-microbiome interactions in the stable state, at AECOPD and during recovery.

II. to study the heterogeneity of AECOPD by comprehensive clinical, functional, microbial, proteomic, transcriptomic, genetic, metabolomic, inflammatory and biochemical characterization of these events.

III. to determine the correlation between microbial alterations in the airways/gut and inflammatory biomarkers in blood during longitudinal follow up.

IV. to longitudinally investigate biomarkers of AECOPD in clinically relevant subgroups of patients with COPD (e.g. current versus ex-smokers, high versus low blood eosinophils, frequent vs. infrequent exacerbators).

V. to comprehensively investigate whether host-microbiome interactions, biomarkers and predictive models identify those patients that do not exacerbate despite having a respiratory infection.

All AECOPD evaluated by the physician in Ciro as moderate or severe will be recorded in this study. Moderate AECOPD are defined by a worsening of symptoms >2 consecutive days leading to treatment with systemic glucocorticoids, antibiotics or both. Severe events are characterized by (enhanced) oxygen therapy, non-invasive ventilation (NIV) or hospital admission that lasted >24 hours, in addition to treatment with systemic glucocorticoids, antibiotics or both.

Whilst aiming to include at least 50 patients experiencing an AECOPD and at least 50 patients without an AECOPD, up to 150 consecutive patients with a primary diagnosis of COPD admitted for inpatient pulmonary rehabilitation at Ciro will be recruited for the study. Although sample size rules of thumb exist for multivariable regression modelling and prediction modelling strategies, this study is explorative in nature. Therefore, the amount of biomarkers exceeds the amount when applying the rule of thumb that states that for each potential predictive variable, 10 events should be observed. The study aims to include at least 50 unique individual patients who experience ≥1 AECOPD, and at least 50 individual patients without an AECOPD, to provide accurate benchmark data (i.e. sufficient precision to estimate mean and standard deviation [SD]) for exploratory biomarkers. From previous studies in Ciro it is known that approximately 42% of patients will develop at least one AECOPD during admission (16). Because AECOPD are unpredictable and variable, which is an important rationale for the present study, the study will expectedly need 100-150 patients to be included. The goal of this study is not to develop a multivariate model, but rather to explore the associations between biomarkers and the occurrence of an event.

Baseline characteristics will be reported as mean and SD or as median and interquartile range (IQR) for continuous variables, as appropriate, and as count and percentage for categorical characteristics. Predictors of time-to-first AECOPD during the study period will be modelled using univariate and multivariable Cox proportional hazards regression. Associations will be presented as hazard ratio (HR) and 95% confidence interval (CI). The dependency of the predictive performance of biomarkers will be tested using interaction terms. The concordance-statistic, or c-statistic, will be estimated to assess discriminative ability. Time-dependent AUC-Receiver Operating Characteristics (ROC) plots will furthermore be created. Calibration will be assessed by comparing the predicted probability with the observed probability of an AECOPD, and examined with a calibration plot and calibration slope, assuming no data censoring before the end of follow-up.

Characterization of the microbiome will be determined by alpha- and beta-diversity and relative abundance of bacterial taxa. Alpha-diversity will be treated as a continuous variable and analysed using appropriate statistical tests, whereas ordination of beta-diversity distances will be done using principal component analysis. Multiomic analyses will be used to generate hypotheses about the drivers that promote progression to AECOPD. Differential enrichment analysis for feature selection across all 'omics will be done after accounting for multiple hypothesis testing using a robust model that considers distributional assumptions. The association between genetic variants and (AE)COPD and microbial infections will be studied using whole-exome sequencing data.

Self-organizing maps (SOMs, also referred to as Kohonen maps) will be used to create an ordered representation of the selected attributes at the time of AECOPD by using Viscovery Profiler 7.1 (Viscovery Software GmbH, Vienna, Austria). Based on the identified homogeneous data groups created in the SOM model, clusters will be generated using Viscovery's SOM-Ward Cluster algorithm. Summary variables of clinical characteristics for the total sample, and for clusters, will be presented as mean and SD for quantitative variables, and as percentages for discrete variables. Differences between groups will be assessed using integrated two-sided t-tests. Repeated measure correlations will be used to determine the within-individual association between microbial alterations in the airways/gut and systemic inflammatory biomarkers across patients.

In case of missing data on predictors, stochastic regression imputation with fully conditional specification will be used to impute the dataset to allow the use of all included patients for the analyses. Values will be drawn using predictive mean matching. Statistical significance will be denoted by p<0.05.

The current study is classified as research with negligible risk. There are no serious risks associated with participation in this study. Hematomas might occur from venous blood sampling. Nasal discomfort might be experienced during nasal sampling. Patients might experience breathlessness and fatigue due to additional daily measurements (e.g. questionnaires, lung function) and/or might refuse participation or drop-out of the study because of the burden of frequent sampling and other assessments. Extensive guidelines have been established to monitor scientific research in a structured and protocol-based manner at Ciro. In the context of the NFU report 'Quality assurance for people-related research', the current investigator initiated research was classified as research with negligible risk. The assigned study site monitor will monitor the safety of the participants, and the accuracy of following procedures as described in the protocol by the research staff, on an annual basis. Monitoring will be performed in compliance with Good Clinical Practice (GCP) in order to achieve high quality research and secure patient safety.

In accordance with the Medical Research Involving Human Subjects Act (WMO) the study will be suspended if the health or safety of subjects will be jeopardised. The accredited Medical Ethical Teaching Committee (Medical Research Ethics Committees United [MEC-U], Nieuwegein, the Netherlands) will be notified without undue delay after obtaining knowledge of these events. Adverse and serious adverse events will be recorded. Hospitalized AECOPD will not be considered as serious adverse events. Exacerbations and hospitalized AECOPD will be annually reported to MEC-U. Ciro has established protocols for the management of AECOPD; these protocols will also be followed for patients included in the study. Moderate AECOPD are treated at Ciro, the pulmonary rehabilitation program will be adjusted as needed. Patients with severe AECOPD requiring hospital admission will be referred to a nearby hospital.

Ethical approval for the study has been granted by MEC-U (NL71364.100.19). The study will be conducted according to the principles of the Declaration of Helsinki (64th WMA General Assembly, Fortaleza, Brazil, October 2013 (17)) and in accordance with the WMO. Written informed consent will be obtained from all participants before study participation. Results of the study will be published in peer-reviewed scientific journals and will be presented at (inter)national conferences. If desired, participants will be informed about the outcomes of the study.

Inclusion Criteria:

• ≥40 years old
• ≥10 pack years of smoking
• primary diagnosis of COPD and post-bronchodilator ratio of forced expiratory volume in the first second (FEV1) to forced vital capacity (FVC) of less than 0.70.
• clinical indication for inpatient pulmonary rehabilitation in Ciro
• provided written informed consent

Exclusion Criteria:

• current, i.e. <12 months, (secondary) diagnosis of asthma according to the referring physician
• unstable concurrent cardiovascular, metabolic, renal, gastro-intestinal and musculoskeletal chronic diseases, as judged by the investigator
• chronic use of oral corticosteroids >10 mg prednisolone/day
• initiation of maintenance therapy with macrolides <6 weeks prior to study entry
• anemia, defined as hemoglobin level <8.1 mmol/L in men and <7.5 mmol/L in women
• participation in a study involving investigational or marketed products concomitantly or <8 weeks prior to study entry
• unable to read, speak or understand Dutch

Patients with a primary diagnosis of COPD admitted for inpatient pulmonary rehabilitation at Ciro (Horn, The Netherlands).