The prognostic blood biomarker proadrenomedullin for outcome prediction in patients with chronic obstructive pulmonary disease (COPD): a qualitative clinical review

Introduction

In patients with chronic obstructive pulmonary disease (COPD), as in other medical settings, risk stratification helps caregivers to more appropriately direct diagnostic, monitoring, or therapeutic interventions. More personalized, better-targeted health-care resource application offers opportunities to improve safety, efficacy, and cost-effectiveness of care, as well as quality of life of patients and their loved ones.

COPD’s complexity and heterogeneity have led the Global Initiative for Chronic Obstructive Lung Disease (GOLD) [1] and other groups, e.g., [2–5], to move beyond strictly spirometry-based prognostication to multidimensional risk assessment of patients with this condition. Considerable interest has arisen in using blood biomarkers within this framework [6–9].

One such analyte that may have a role in COPD multidimensional risk assessment is plasma proadrenomedullin (ProADM), a surrogate for the pluripotent regulatory peptide adrenomedullin (ADM) [10]. There exists a substantial observational literature regarding ProADM use as an all-cause mortality predictor in patients with sepsis [11–19], in patients with a variety of underlying diseases presenting to the emergency department (ED) with acute dyspnea [20–26], and especially in patients with community-acquired pneumonia (CAP) [13, 27–38].

Additionally, recent observational studies demonstrated that in patients during or just recovering from COPD exacerbation [39–41] or in patients with stable COPD [40, 42], ProADM is a powerful independent prognosticator of long-term non-survival [39–42]. Some of this work also shows that when combined with demographic and clinical variables, ProADM provides significant incremental mortality prediction accuracy [41, 42]. Additional data [43–45] suggest that ProADM may be a global cardiopulmonary stress marker [45]. As such, this blood biomarker may supply prognostic information when cardiopulmonary exercise testing (CPET) results such as 6-min walk distance (6MWD) are unavailable. Indeed, ProADM may even serve as a simpler, less “invasive” substitute for the 6-min walk test (6MWT) or other CPET in settings where time or other resource constraints or a patient’s advanced disease render such examinations infeasible, albeit this hypothesis awaits confirmation by prospective, randomized, controlled interventional studies.

Aims of the review and methodology

The present review’s goal is to provide clinicians with an overview of ProADM in risk stratification of COPD patients. We begin by briefly describing ADM and explaining why ProADM serves as a surrogate for this regulatory peptide. Next, we summarize observational data regarding ProADM in risk stratification of COPD and other pulmonary diseases/disorders and the analyte’s relationship with cardiopulmonary stress, exercise capacity, and physical activity. We conclude by outlining clinical considerations and future research directions regarding ProADM in patients with COPD. Literature discussed in this article was partly identified through a systematic literature search of English-language publications indexed in PubMed through 12 May 2014 under the terms “adrenomedullin” or “proadrenomedullin” or “ProADM” together with each of the terms “lung”, “pulmonary”, “chronic obstructive pulmonary disease”, “COPD”, “exacerbation”, “dyspnea”, “emphysema”, “bronchitis”, “asthma”, “pneumonia”, “cardiac”, “cardiovascular”, or “exercise”. Notwithstanding this formal methodology, the present paper is a qualitative rather than a systematic review.

ADM and ProADM

First isolated in the early 1990s [46], ADM is a 52-amino acid ringed peptide with C-terminal amidation belonging to the calcitonin superfamily [47, 48]. ADM is ubiquitously expressed in pulmonary, cardiovascular, renal, gastrointestinal, and endocrine tissue and by endothelial cells, vascular smooth muscle cells, cardiomyocytes, fibroblasts, leukocytes, and placental trophoblast cells, among others [49–52]. In pulmonary tissue, ADM expression has been found in endothelial cells including type II pneumocytes, chondrocytes, smooth muscle cells, the columnar epithelium, alveolar macrophages, monocytes, T cells, and neurons of the pulmonary parasympathetic nervous system, as well as small-cell and non-small-cell neoplasia [42, 53].

ADM’s widespread expression throughout the body reflects this molecule’s great variety of biological activities: the peptide seems to act as both a hormone and a cytokine and thus can be seen as a “hormokine”. It acts systemically and in autocrine and paracrine fashion [54, 55], exerting or mediating vasodilatory, natriuretic, diuretic, antioxidative, anti-inflammatory, antimicrobial, and metabolic effects [42, 50, 56–58]. Upregulated by hypoxia, inflammatory cytokines, bacterial products, and shear stress, ADM has in preclinical and animal models been shown to reduce hypoxic pulmonary vascular structural remodeling and fibrosis and to inhibit bronchoconstriction; the molecule also has been shown to stabilize barrier function in the lungs by downregulating pro-inflammatory factors and reactive oxidative species [42, 50, 51, 59–62]. Circulating ADM elevation, e.g., in end-stage pulmonary disease [63], is believed to reflect “high demand” for these compensatory/counter-regulatory effects [28, 42].

Based on the hormone’s biological importance and effects, the utility of measuring ADM in blood seems evident; however, abundant binding to peripheral and local receptors, a short half-life, and ex vivo physical characteristics including instability and “stickiness” make circulating ADM difficult to reliably directly quantify [10, 64].

ADM, however, is only one of five peptides contained on the ADM precursor molecule (Figure 1). During ADM’s processing into mature hormone, it and an adjoining peptide, mid-regional proadrenomedullin (MR-ProADM; referred to here and elsewhere as ProADM), are cleaved off of the precursor in a 1:1 ratio. This stoichiometric secretion, the peptide’s apparently minimal if any biological activity, and hence binding, and considerable chemical stability render ProADM an easily and robustly quantifiable surrogate biomarker for ADM [10, 65].

Figure 1

Schematic representation of the ADM precursor molecule.

The diagram illustrates that the precursor is occupied by five peptides including MR-ProADM, more commonly referred to as ProADM. When the molecule is processed, ProADM and ADM are co-secreted in a 1:1 ratio. The black numerals refer to the amino acids on the precursor molecule. PAMP, proadrenomedullin N-terminal 20 peptide.

ProADM in risk assessment of patients with COPD

Clinical research regarding plasma ProADM in patients with COPD, and particularly, regarding ProADM in risk assessment of such patients, was initially inspired by a study [63] showing significantly elevated ADM concentrations in patients with end-stage pulmonary diseases and a trend toward this finding in the COPD subgroup, relative to controls [39]. As of May, 2014, observational data regarding ProADM in COPD have been published by four groups [39–42] (Table 1). Two of these teams, Stolz et al. at University Hospital Basel and Zuur-Telgen et al. at Medisch Spectrum Twente, Enschede, The Netherlands, reported single-center data; the other two, the Predicting Outcome Using Systemic Markers in Severe Exacerbations of Chronic Obstructive Pulmonary Disease Study (PROMISE-COPD) investigators and the ProHOSP investigators, reported multicenter data. The PROMISE-COPD data derived from 11 centers in 8 European countries, and the ProHOSP data, from 6 Swiss centers. The University Hospital Basel and Enschede study samples were relatively small (n=167 and n=181, respectively); the multicenter samples were substantially larger, 549 (PROMISE-COPD) and 469 (ProHOSP). All cohorts comprised predominantly, or exclusively, patients with moderate to severe COPD. However, unlike the other samples, the ProHOSP patients with COPD lacked dedicated spirometric confirmation of this diagnosis because their baseline data derived from a secondary analysis of an antibiotic stewardship trial in patients with a mixed group of lower respiratory tract infections (LRTIs) [41]. In all four of these studies, the analyses evaluated the risk prediction power of single ProADM measurements in samples obtained when patients were in one or more of stable [40, 42], (severely) exacerbated [39, 40], or “recovery” states [39, 41]. In two of the studies [39, 41], ProADM was measured using a manual luminoimmunoassay [10], while in the other two studies [40, 42], the analyte was quantitated with an automated homogeneous sandwich fluoroimmunoassay [70]. Because these two assays were shown to have a Spearman correlation coefficient of r=0.97, their results may be viewed as interchangeable [70].

Studies focusing on ProADM in COPD risk stratification have had very consistent results and, collectively, have had the following main findings (summarized in Supplemental Data, Panel 1). First, even in the stable state, COPD patients almost always have ProADM concentrations elevated above healthy adult reference values. In the Basel study [39], ProADM levels exceeded a mean reference value [10] in 98% of patients (163/167) at admission for COPD exacerbation resulting in median [interquartile range (IQR) 25th–75th percentile] hospitalization of 9 [1–15] days. The corresponding figures were 90% [140/156] at “recovery” 14–18 days post-admission and 79% [114/144] at “stable state” 6 months post-admission. The Enschede investigators reported that during the stable state, defined as freedom from COPD exacerbation, antibiotics, or prednisolone over the prior >4 weeks, 99% of patients had elevation compared with healthy individuals [40].

Second, as suggested by the three serial measurements in the Basel investigation [39], ProADM values appear to be repeatable. That study’s median [IQR] “recovery” and “stable state” values, respectively, 0.72 [0.55–0.98] and 0.66 [0.49–0.95] nmol/L, did not differ (p=0.350) [39]. Additionally, the Enschede investigators found exacerbated- and stable-state ProADM to be significantly correlated (r=0.73, p<0.001), another sign of internal consistency among values of this biomarker. Further, measurements obtained within the exacerbated, recovery, or stable states seem to be comparable across studies and assays (Table 2).

Third, ProADM values are significantly higher in the exacerbated state vs. the stable state of COPD. For example, in the Basel study [39], median [IQR] ProADM concentration at admission, 0.84 [0.59–1.22] nmol/L, significantly exceeded that at “recovery” or “stable state” (p=0.002).

Fourth, ProADM appears not to be associated with underlying COPD severity. In the Basel study [39], no correlation was seen between GOLD grade and ProADM concentration at admission for severe exacerbation (r=–0.066, p=0.406); indeed, GOLD grade IV patients had the lowest median ProADM levels (Table 2). This observation aligned with findings in another study [72] that in 85 patients with lung diseases (46% of whom had COPD), ProADM values did not differ between patients with FEV₁/forced vital capacity <0.7 vs. ≥0.7: median [IQR] concentration 0.62 [0.46–0.78] nmol/L (n=40) vs. 0.54 [0.47–0.75] nmol/L (n=45), p=0.38.

Also in the Basel study [39], ProADM values did not distinguish among Anthonisen exacerbation types (Table 2).

Fifth, ProADM levels are significantly higher in non-survivors than in survivors (Table 2). This finding pertained to all-cause death, the mortality end point in all published studies, and was consistent across follow-up periods, including the hospital stay [39], 1 year [42], 2 years [39, 42], respective medians of 29 and 35 months post-stable-state measurement and post-exacerbated-state measurement [40], and 5–7 years [41].

Sixth, as a single factor, ProADM is a statistically significant, independent, and accurate long-term all-cause mortality predictor in patients with COPD. In the Basel study [73], multivariate Cox regression showed ProADM above the study exacerbated-state median, 0.84 nmol/L, to be the only independent 2-year mortality predictor, with a 2.368 (1.167–4.803) hazard ratio (HR) (95% confidence interval, CI), p=0.017. None of the other 11 clinical, spirometric, or laboratory variables considered in the multivariate model was independently predictive; these variables included

• Two of four components of the body mass, airflow obstruction, dyspnea, and exercise capacity index (BODE) [2], a frequently used multidimensional assessment tool: body mass index (BMI) and fixed expiratory volume in 1-s percentage of the predicted value (FEV₁% predicted)

• Four other blood biomarkers: white blood cell count (WBC), C-reactive protein (CRP), procalcitonin (PCT), and pro-endothelin 1 (ProET-1)

• Age-adjusted Charlson comorbidity score [74]

• Age

• Partial pressure of oxygen in arterial blood (PaO₂) or partial pressure of carbon dioxide in arterial blood

• Presence of pulmonary arterial hypertension (PAH) (n=123)

Additionally, the Enschede investigators found that after adjustment for age, sex, BMI, and GOLD grade, stable-state ProADM elevated above their 0.71-nmol/L study median remained associated with long-term mortality: corrected HR (95% CI) 2.98 (1.51–5.90).

Further, in PROMISE-COPD, multivariate Cox regression modeling involving ProADM together with the four BODE components, BMI, FEV₁% predicted, modified Medical Research Council (mMRC) dyspnea score, and 6MWD, showed ProADM to be one of three significant independent 2-year all-cause mortality predictors: HR (95% CI) per one-quartile concentration change 1.77 (1.30–2.42), p<0.001; the other two independent predictors were BMI and 6MWD. Based on the likelihood ratio χ² (13.0 vs. 8.5 and 7.5, respectively), ProADM was the strongest of these three prognostic factors.

Studies of potential adverse outcome predictors typically measure predictive accuracy using one or more among a number of variables. These variables, described in Appendix 1, include sensitivity, specificity, positive predictive value, negative predictive value, area under the receiver operating characteristics curve (AUC of the ROC curve), or its equivalent, the c index or c statistic, or net reclassification improvement (NRI) or the integrated discrimination improvement index (IDI) [75]. Findings for ProADM as a single factor to foretell non-survival are summarized in Table 3; observational data thus far suggest that this blood biomarker alone has moderate mortality prediction accuracy, even beyond a half decade of follow-up. Unsurprisingly, ProADM mortality prediction accuracy tends to decrease as follow-up time increases.

Seventh, combined with different groups of demographic/clinical variables, ProADM provides significant incremental mortality prediction power (Table 4). In PROMISE-COPD, combining stable-state ProADM concentration with BODE significantly improved predictive power for 1- or 2-year all-cause mortality compared with using BODE alone [42]. In the ProHOSP study’s COPD subgroup, adding ProADM levels at discharge after severe exacerbation to a demographic/clinical model provided significant incremental predictive power for 1-, 3-, and 5- to 7-year all-cause mortality [41]. The model included age, smoking status, BMI, New York Heart Association (NYHA) dyspnea class, presence/absence of important comorbidities such as chronic renal failure, cancer, coronary artery disease, diabetes mellitus, or chronic heart failure, and exacerbation type (pneumonic vs. non-pneumonic).

Eighth, there are suggestions that ProADM may have a role in predicting adverse outcomes besides death in COPD patients. In the Basel study [39], ProADM levels significantly correlated with hospital length of stay (r=0.274, p<0.0001) (n=167) and trended toward correlation with intensive care unit (ICU) length of stay (r=0.151, p=0.051) (n=16). Additionally, median [IQR] ProADM levels were higher in patients needing vs. not needing ICU admission, a difference approaching significance: 1.23 [0.61–2.14] vs. 0.83 [0.59–1.14] nmol/L, p=0.057.

Lastly, and unsurprisingly, based on literature regarding ADM or ProADM in cardiovascular and malignant disease [48, 76–80], ProADM levels appear to be associated with certain comorbidities in COPD patients. In the Basel study [39], median values were significantly higher in patients with vs. without cancer: 1.225 [0.718–1.770] nmol/L (n=24) vs. 0.810 [0.571–1.100] nmol/L (n=143), p=0.008. Median values also were significantly elevated in those with PAH vs. with normal pulmonary pressures: 1.06 [0.76–1.63] nmol/L (n=28) vs. 0.82 [0.61–1.23] nmol/L (n=95), p=0.031. ProADM concentrations correlated significantly with left ventricular ejection fraction (r=–0.222, p=0.014), but did not differentiate between the subgroups below vs. at or above a cutoff of <40% for this variable. The Enschede investigators reported correlation between ProADM levels and heart failure or myocardial infarction in their COPD patients [40].

ProADM in the risk assessment of patients with other pulmonary diseases/disorders

As noted earlier, there also exists a substantial observational literature regarding ProADM use as an all-cause mortality predictor in patients with a variety of underlying diseases presenting to the ED with acute dyspnea [20–26] and, especially, in patients with CAP [13, 27–38]; key studies are summarized in Table 1. Many of the dyspnea and CAP studies [20–23, 25–27, 29–31] had cohorts comprising >25% of patients with COPD as a cormorbidity. Collectively, the published studies in these settings have had more ethnically and, particularly, geographically diverse samples than did the published ProADM studies focusing on COPD; whereas the latter have taken place overwhelmingly in Caucasian patients and exclusively in Europe, the dyspnea and CAP investigation has in aggregate included more non-white individuals, and in some cases, occurred partly or entirely in North America [21, 23, 24, 28], Asia [25], or Oceania [21, 23]. Vital status follow-up durations in these studies ranged from the hospital stay to 4 years. As with the COPD-focused studies, the CAP- and dyspnea-focused body of work has, albeit with limited exceptions [22, 36], shown ProADM to be a strong, independent mortality predictor in multivariate analysis. Where combining ProADM with clinical/demographic/other laboratory variables has been studied in dyspnea or CAP [23, 24, 32, 35], the biomarker has provided significant incremental prognostic power. Most published ProADM studies in dyspnea or CAP evaluated predictive accuracy of single measurements of this analyte; however, the two published studies to address the issue [23, 34] suggested that serial ProADM measurement may add predictive information.

ProADM and cardiopulmonary reserve/exercise capacity/physical activity

Three papers from PROMISE-COPD [42–44] and one from a single-center study at University Hospital Basel [45] (summarized in Tables 1 and 5) collectively show strong links between ProADM and cardiopulmonary reserve/exercise capacity/physical activity in patients with COPD. This work also suggests that this analyte may be a global cardiopulmonary stress marker that is able to supplement or even substitute for CPET in assessing this health dimension. In the COPD setting, replacing the 6MWT with a simple and almost universally applicable blood test could have two main benefits. First, doing so could increase the availability of exercise capacity-related data, which is hampered by both resource constraints and patient limitations. The resource constraints, particularly relevant to primary care, stem from the relatively complex, time-consuming nature of, e.g., the 6MWT, as well as that test’s requirements for a 30-m track, an examiner certified in cardiopulmonary resuscitation with at least one American Health Association-approved course in basic life support, and supplies and facilities for rapid medical emergency response [42, 81]. The patient limitations, which affect an appreciable portion of the population with more advanced COPD, include frailty, peripheral arterial disease, or musculoskeletal or neuromuscular impairments. Notably, even though these latter impairments were among the PROMISE-COPD exclusion criteria, 8% (51/638) of that study’s sample had unavailable 6MWD. Interestingly, absence of 6MWD data was associated with a statistically significant almost tripling of the 2-year death rate relative to that of patients with such data (n=549): 21.6% vs. 7.8%, p=0.003 [42].

A second potential benefit of replacing the 6MWT with a blood biomarker test might be cost savings: the cost of the former was estimated to approach 40 euros in Switzerland, whereas the cost of a ProADM measurement was estimated at 15–20 euros [42]. Importantly, however, any benefits of using a “ProADM instead of 6MWT or other CPET” strategy must be verified through prospective, randomized, controlled interventional study.

The apparent strong links between ProADM and cardiopulmonary stress/exercise capacity in patients with COPD involve walking distance, exertional hypoxia, and CPET variables. For example, a PROMISE-COPD substudy in. 105 patients with stable COPD from University Hospital Basel examined the association with ProADM levels of daily walking activity measured over 6 consecutive days by accelometry [43]. In two separate stepwise multivariate regression analyses, each including one accelometry variable plus patient age, age-adjusted Charlson comorbidity score, mMRC dyspnea score, and St. George’s Respiratory Questionnaire score, total walking minutes/day or steps walked/day were the sole factors independently associated with ProADM concentration (regression coefficient –0.0004, 95% CI –0.0007 to –0.0002; regression coefficient –0.0004, 95% CI –0.0006 to –0.0001; both p<0.0001). However, fast walk (min/day walking 5 km/h or 81–115 m/min) was not significantly associated with that dependent variable.

More notably, in 549 PROMISE-COPD participants with available ProADM and BODE data, Cox regression modeling demonstrated that ProADM plus “BOD”, i.e., the non-6MWD BODE components, namely, BMI, FEV₁% predicted, and mMRC dyspnea score, had higher prognostic accuracy for 1- or 2-year all-cause mortality than did BODE: c statistics 0.800 vs. 0.745 for 1-year mortality, 0.738 vs. 0.679 for 2-year mortality [42]. These observations held up in post hoc sensitivity analyses adding in the 45 patients with ProADM and only “BOD” data, whether using Hotdeck imputation in this subgroup or assigning those patients the worst possible BODE score, 3 points, for 6MWD. Compared with using “BOD” alone, combining ProADM with “BOD” achieved an NRI of 41.2% (95% CI 15.6%–66.8%) of patients for 1-year mortality risk and of 8.8% (95% CI –10.6%–28.3%) for 2-year mortality prediction. The NRI for 1-year mortality was significant (p=0.0016), although that for 2-year mortality was not (p=0.37).

Additionally, a multivariable linear logistic regression analysis of 1233 6MWTs performed over 2 years by 574 PROMISE-COPD participants with stable COPD found that ProADM as well as post-bronchodilator FEV₁% predicted each independently foretold exertional hypoxemia: respectively, HR 6.85 (95% CI 2.91–16.08) per logarithmic change and HR 0.76 (95% CI 0.72–0.88) per 10% increase, both p<0.001 [44]. Exertional hypoxemia was defined as nadir arterial oxygen saturation <88% on continuous transcutaneous pulse oximetry. As with that of FEV₁% predicted, the significant independent association with exertional hypoxemia of ProADM persisted even after inclusion into multivariate modeling of the annual COPD exacerbation rate plus the age-adjusted Charlson comorbidity score, or of any or all of congestive heart failure, coronary artery disease, or history of acute myocardial infarction. Moreover, adding ProADM to FEV₁% predicted to foretell exertional desaturation provided a significant NRI of 7.4% (95% CI 1.3%–13.6%) (p=0.0184) of patients relative to using the spirometric variable alone.

Another study from University Hospital Basel [45] found pre-exercise ProADM to be significantly associated with impaired peak oxygen consumption, defined as such consumption <14 mL/min/kg – including after multivariable adjustment for factors including age, BMI, and FEV₁ (Table 5). Additionally, ProADM was consistently significantly associated with other variables reflecting impaired cardiac output reserve, ventilatory efficiency, and diffusion capacity (Table 5), and hence, global cardiopulmonary stress.

The investigators found these associations to be at least as strong as those of at-rest B-type natriuretic peptide (BNP), an established cardiac stress blood biomarker that they also measured. According to the authors, these observations suggested that ProADM was a more universal cardiopulmonary stress marker than was the more “cardiac-specific” BNP. Notably, unlike BNP, ProADM did not rise significantly from before to immediately after a maximal exercise test, which the authors noted “may be an advantage of [ProADM] as a robust clinical marker”. The study involved 162 consecutive eligible patients (mean age 56±16 years, 58.6% [95/162] male) with a gamut of cardiac and pulmonary conditions who underwent CPET to assess subjective chronic exercise intolerance. Histories including COPD, other lung conditions, or cardiac disease respectively were present in 39 (24.1%), 70 (43.2%), and 51 (31.5%) patients. The CPET comprised symptom-limited upright cycle exercise tests employing ramp protocols, with continuous electrocardiogram monitoring and non-invasive blood pressure measurement every second minute. Blood biomarkers were quantified in samples obtained at rest and 1 min after peak exercise, defined as peak minute ventilation/carbon dioxide output ratio [45].

Clinical considerations

Confounding factors do not seem to be a major issue with ProADM measurement. Some potential factors have been identified in healthy volunteer, community-based, or general population studies [10, 58, 71, 82, 83], or COPD risk stratification studies, perhaps most notably age [10, 71, 83, 84], which correlates positively with this analyte. However, likely due to disparate study sample characteristics (e.g., age, comorbidity profile) [71], the literature conflicts regarding whether some factors, e.g., gender [58, 71, 83, 84], indeed are confounders. Moreover, multivariate analyses in some of the COPD-focused studies [39–42] have suggested that many of the potential confounders, namely, age, sex, age-adjusted Charlton comorbidity score, BMI, PaO₂, mMRC dyspnea score, and levels of PCT, CRP, or ProET-1, indeed do not affect ProADM’s mortality prediction ability in that setting. Additionally, exclusion of patients with known cardiovascular disease or risk factors from a large population-based sample seemed not to markedly alter ProADM values (Table 2) [71]. Lastly, neither circadian variation nor prandial status appears to affect ProADM measurements [10].

The consensus favoring multidimensional assessment of COPD [1, 2, 85–87], experience with blood biomarkers and risk scoring systems in other pulmonary disease settings [32, 88, 89], and literature focused on ProADM in COPD [42, 90] appear to support three concepts in clinical application of ProADM (Supplemental Data, Panel 2). First, the use of this analyte in conjunction with a limited number of demographic and clinical, and possibly other laboratory, variables may be a practical and effective approach. This approach conforms to the widely accepted adage that biomarker values should be interpreted within the comprehensive context of each particular case rather than in isolation [91].

Second, to increase ease of application and mitigate any effects of confounding factors, risk scoring systems incorporating ProADM should use a very small number of relatively widely spaced cutoff values for this blood biomarker. An example of ProADM use in a multidimensional framework are Stolz et al.’s proposal of “BODE-A” or “BOD-A” risk scoring systems. These classification methods combine ProADM, BMI, FEV₁% predicted, and mMRC dyspnea score, respectively with or without 6MWD [42]. In another example, the ProHOSP investigators evaluated a model combining ProADM with patient age, smoking status, BMI, NYHA dyspnea class, exacerbation type (pneumonic vs. non-pneumonic), and comorbidities (chronic renal failure, neoplastic disease, coronary artery disease, chronic heart failure, and diabetes mellitus) [41]. An embodiment of applying few, widely spaced ProADM cutoffs is the ProHOSP investigators’ proposed CURB65-A score for patients with LRTIs including COPD exacerbation [32] (Table 2). The CURB65-A score combines the five CURB65 criteria, developed for pneumonia severity/risk classification [92], with two ProADM cutoffs, 0.75 and 1.5 nmol/L, to create three risk categories, low, intermediate, and high. The CURB65 criteria comprise new onset confusion, urea >7 mmol/L, respiratory rate ≥30 breaths per minute, systolic or diastolic blood pressure <90 or ≤60 mm Hg, respectively, and age ≥65 years. Tripartite risk stratification is desirable if a biomarker or score seeks to identify both patients who should receive less intense intervention(s) and those who should receive more intense intervention(s) [23] rather than only one of these subgroups.

Third, as should be the case with cutoffs for all multidimensional scoring system components [89], ProADM cutoffs should be calibrated to local conditions, e.g., EDs treating large numbers of COPD patients for acute dyspnea or pneumonic exacerbations probably need higher cutoffs than would outpatient clinics primarily seeing patients in the COPD stable state.

Future research directions

Future research on ProADM in risk stratification of patients with COPD should take a number of different directions (Supplemental Data, Panel 3). First, additional observational data should be gathered regarding non-white patients and never-smokers, two groups mostly absent from studies published to date focusing on ProADM in COPD [39–42].

Second, observational studies should further assess whether serial ProADM measurements offer additional prognostic power. Two published studies in the acute LRTI [34] or acute dyspnea [23] settings have suggested informativeness of moves between “low” and “high” ProADM categories every 2–4 days [34] or from admission to discharge (median [IQR] interval: 7 [3–12] days) [23]. These studies each had approximately 30% or more patients with COPD as a comorbidity and used respective cutoffs of 1.5 [34] or 1.985 nmol/L [23] for the “high” ProADM classification (Table 2). Besides further evaluating this serial measurement strategy, future analyses should assess the incremental prognostic value of absolute or relative, i.e., percentage ProADM changes.

Third, future studies should compare the mortality prediction accuracy of ProADM vs. other blood biomarkers as single factors or in various combinations. Published multivariate or ROC curve analyses have suggested that in the COPD setting, ProADM may provide superior discrimination of non-survival to that of CRP [39, 41], WBC [39, 41], ProET-1 [39], PCT [39–42], copeptin [42], or pro-atrial natriuretic peptide [42] as single factors. Additionally, PROMISE-COPD noted that combinations of ProADM and one or more of the latter three other analytes offered no incremental accuracy over that of ProADM alone [42]. However, these comparisons have been relatively few and studies have not always reported statistical significance data. Further, no comparisons yet have been published of ProADM vs. fibrinogen, interleukins 6 or 8, surfactant protein D, or neutrophil count, blood biomarkers that also have shown accuracy in long-term mortality prediction in patients with COPD [8, 87, 93].

Fourth, interventional studies should be undertaken to demonstrate whether ProADM use in risk-based guidance of monitoring/treatment decisions improves clinical, quality-of-life, or pharmacoeconomic outcomes in patients with COPD. Published research on such topics [68, 69] has been very preliminary. OPTIMA II, a prospective, multicenter, randomized, controlled proof-of-concept study [68] compared hospital length-of-stay in patients with acute LRTIs assigned 1:1 to interprofessional medical and nursing risk assessment with vs. without use of ProADM data. The study involved 313 enrollees, of whom 43 (13.7%) had COPD exacerbation. These patients were treated at any of one acute-care hospital or two post-acute-care centers in Switzerland. The interprofessional risk assessment included CURB65 scoring at admission, medical stability evaluation during hospitalization, and functional and biopsychosocial assessment, respectively using the Self-Care Index and Post-Acute Care Discharge Score at both times. In the ProADM arm, ProADM levels of 0.75 and 1.5 nmol/L helped delimit “low-risk”, “intermediate-risk”, and “high-risk” categories, which in the control arm were determined only by the interprofessional assessment. In both study arms, patient site-of-care assignments were based on an algorithm incorporating the risk category at the given assessment time. However, for any individual case, treating physicians could override algorithm recommendations for a variety of pre-specified medical, psychosocial, or administrative/logistical reasons.

OPTIMA II found that the ProADM group had a shorter mean length of stay did the controls, 6.3 (95% CI 5.4–7.2) days vs. 6.8 (95% CI 5.7–7.9) days. The difference favoring the ProADM group held true after adjustment for age, gender, LRTI type, and CURB65 score. Moreover, the differences consistently favored the ProADM arm in subgroup analyses, including those involving inpatients with a ≥1-day stay, men and women, patients with or without CAP, older or younger patients, patients with greater or lesser comorbidity burden, or CURB65 classes I, II, or III. However, in no case did the difference attain statistical significance; the authors attributed this observation to the relatively small sample size, the great influence of logistical/administrative factors on site-of-care decisions, and the fact that the study centers had for years prior to OPTIMA II strongly emphasized minimizing length of stay [94].

The Biomarkers in Acute Heart Failure study (BACH) investigators conducted separate subanalyses of their American patients (n=831) and European patients (n=726) with acute dyspnea – stemming from COPD in just over 11% of each subgroup – comparing the actual distribution of site-of-care assignments vs. a hypothetical distribution guided by ProADM values [69]. This analysis should be regarded as hypothesis-generating because it used ProADM values in isolation. For the US analysis, sites of care were divided into four levels, discharge, general ward, cardiac care or monitoring unit, and ICU, whereas for the European analysis, sites of care were divided into three levels because the cardiac care unit and ICU were considered as a single level. For both analyses, baseline ProADM ≥5.0 nmol/L would have led to the site of care being stepped up by one level. Values ≤0.50 nmol/L would have led to the site of care being stepped down by one level, except that in the European analysis, stepping down from the cardiac care unit/ICU to the general ward could occur if the ProADM was <1.0 nmol/L. The investigators found that using such ProADM-guided assignment theoretically would have increased the number of discharged patients by 16.7% in the US (384 vs. 329) and 15.9% in Europe (219 vs. 189); notably, the very small number of discharged 90-day decedents (4 for both analyses) would not have increased. Overall, the theoretical NRI was 12.0% (95% CI 5.7%–18.4%) for the US and 16% (95% CI 8.2%–23.9%) for Europe, and 9.7% (81/831) of US patients and 9.9% (72/726) of European patients would have changed site of care.

Conclusions

Observational studies to date show that plasma ProADM, the stable surrogate blood biomarker for the pluripotent regulatory peptide, ADM, is, as a single factor, a strong independent predictor of in-hospital to long-term all-cause mortality in patients with COPD. Additionally, the literature suggests that when combined with demographic and clinical factors, this analyte provides significant incremental prognostic accuracy regarding this end point. This ability may derive from ProADM being a multidimensional marker of cardiopulmonary stress, exercise capacity, and physical activity. ProADM appears to merit further clinical investigation in COPD. This research should take the form of observational studies examining the analyte’s mortality prediction power in never-smokers and non-white patients, its ability to foretell adverse outcomes other than mortality, and its comparison and combination with other biomarkers such as fibrinogen and interleukins 6 and 8, which have shown survival prediction accuracy. In addition, interventional trials should test whether ProADM can improve clinical outcomes by helping guide intervention and monitoring and can save costs by substituting for the 6MWT and other CPET in patients with COPD.

Author contributions: All the authors have accepted responsibility for the entire content of this submitted manuscript and approved submission.

Financial support: P.S. and B.M. have received support to fulfill speaking engagements and travel to medical meetings from Thermo Scientific Biomarkers, Hennigsdorf, Germany, the developer/manufacturer of the ProADM assay; Thermo Scientific Biomarkers also has provided monetary support and reagents to the Kantonsspital Aarau research fund. R.J.M. received support to fulfill speaking engagements and travel to medical meetings from Thermo Scientific Biomarkers.

Employment or leadership: B.M. has served as a consultant to Thermo Scientific Biomarkers. R.J.M. has served/is serving as a paid consultant to Thermo Scientific Biomarkers and holds shares in Thermo Fisher Scientific, Inc., the publicly traded parent company of Thermo Scientific Biomarkers.

Honorarium: None declared.

Competing interests: The funding organization(s) played no role in the study design; in the collection, analysis, and interpretation of data; in the writing of the report; or in the decision to submit the report for publication.