Current Approaches for Phenotyping as a Target for Precision Medicine in COPD Management

ABSTRACT

The study of airway diseases continues to present several challenges for modern medicine. The different disease presentations with variables and overlapping features may result in a real challenge for the clinician. In this context, the concept of precision medicine has started to emerge in order to give answers to some of these challenges from a diagnostic and therapeutic point of view. The main reasons to target for precision medicine in chronic obstructive pulmonary disease (COPD) include that there is variability in the clinical presentation, there is no correlation between the different clinical variables at the patient level, there are a number of relevant clinical variables associated with outcomes, we do have specific therapies for specific patient types, and that there is variability in the clinical response to different therapies. To bring precision medicine into clinical practice several approaches have been used, including the use of independent variables to identify subjects, the use of multidimensional indexes, the so-called clinical phenotypes, and the approximation by the so-called treatable traits. All these approaches have their strengths and weaknesses which are reviewed in the present document. Although there is no universally accepted proposal, the available initiatives provide us with a framework on which to start working and move toward precision medicine in COPD, with the ultimate goal of bringing the best possible medicine to each patient in particular.

KEYWORDS:

• COPD
• clinical phenotypes
• multidimensional evaluation
• precision medicine

Introduction

The study of airway diseases continues to present several challenges for modern medicine. Despite the great advances in the knowledge of the pathogenesis of airway diseases, we still have knowledge gaps that prevent us from having a real vision of airway conditions. Although the majority of patients can be diagnosed with well-defined conditions, in real life, the clinical presentation of these may vary and not completely adjust to our definition and they may also overlap. Consequently, the traditional classification of diseases based on symptoms, sings, and syndromes may not be sufficient to explain diseases. This phenomenon is especially present in chronic obstructive pulmonary disease (COPD) which has been defined as a complex heterogeneous condition in which the majority of patients combine features of the different diseases' expressions (1). Consequently, it has been referred to as a syndrome rather than as a disease (2). In this context, the concept of precision medicine has started to emerge in order to give answers to some of these challenges from a diagnostic and therapeutic point of view (3). In a pathological condition like COPD with a considerable increasing prevalence (4) and notable morbidity and mortality (5, 6), the debate between clinical guidelines and precision medicine is served (7). In the following lines, we will review the reasons that support the need to carry out precision medicine in COPD, as well as the strategies defined to carry it out, showing their strengths and weaknesses.

Should we target for precision medicine in COPD?

After the arguments given above, it would be logical to conclude that, indeed, we must aim for precision medicine in COPD. However, it is important to remember some of the reasons that support this statement. The main reasons to make this statement are: (1) there is variability in the clinical presentation; (2) there is no correlation between the different clinical variables at the patient level; (3) there are a number of relevant clinical variables associated with outcomes; (4) we do have specific therapies for specific patient types; and (5) there is variability in the clinical response. In the next few lines we will expand each one of these points.

There is variability in the clinical presentation

At present, there^{Figure 1} are numerous studies that have shown the variability of the clinical presentation of COPD in various contexts. This variability has been shown in symptoms perception throughout the year, during a month, with a weekly and even daily cadence, 24 hours a day (8). In addition, the variability of symptoms has been shown in both upper and lower airway symptoms (9) and with differences between geographical areas (10).

This variability in symptoms perception can also be shown in multidimensional indexes. Probably, one of the most extended multidimensional indices is the COPD assessment test (CAT), whose validity for the evaluation of patients with COPD has been shown in numerous studies. This instrument has been validated in several different scenarios, including stable disease (11), exacerbations (12), association with certain comorbidities (13), and after interventions, such as pulmonary rehabilitation (14), which make it suitable for use in routine clinical practice (15). Thus, the CAT has become one of the main instruments in the evaluation of health status in COPD. Consequently, it has been incorporated into the GOLD strategy (16). However, when studying the distribution of CAT in the different types of patients in the GOLD document, a wide dispersion of its values is again observed for all patient categories (Figure 1) (17). In addition, the progression of CAT over time associated with interventions such as tele-health has recently been described (18). Consequently, it can be deduced that the CAT could be a good tool to qualify the impact of the disease within each type of patient and help in this way to decide on the treatment in each patient.

Figure 1. Distribution of the CAT score among the different modified GOLD 2011 classification. Reproduced from Lopez-Campos et al. (17). © 2015 Lopez-Campos et al. This figure is published by Dove Medical Press Limited, and licensed under Creative Commons Attribution – Non- Commercial (unported, v3.0) License. The full terms of the License are available at http://creativecommons.org/licenses/by-nc/3.0/. Noncommercial uses of the work are permitted without any further permission from Dove Medical Press Limited, provided the work is properly attributed. Permissions beyond the scope of the License are administered by Dove Medical Press Limited. Information on how to request permission may be found at: http://www.dovepress.com/permissions.php.

Interestingly, this variability in clinical presentation affects also to lung function. The Evaluation of COPD Longitudinally to Identify Predictive Surrogate Endpoints (ECLIPSE) program tracked 2,180 subjects for 3 years with objective of definition of clinically relevant COPD subtypes and the identification of parameters that predict disease progression in these subtypes (19). ECLIPSE clearly showed one important concept in relation to the progression over time of the forced expiratory volume in the first second (FEV₁), this is that FEV₁ was variable in time and that does not always get worse over time. In its follow-up cohort, ECLIPSE observed that there was considerable variability in the decline of FEV₁, with patients who had a striking worsening with declines greater than 40 mL/year, while others did not have such a striking worsening or even improved lung function with time in active treatment (20). This finding broke the preconceived and maintained idea until then that COPD was a disease that presented a constant decline in FEV₁ on which we could not do any intervention to stop it (21). Although the impact of the various drugs on the decline of FEV₁ has not been demonstrated in a convincing manner (22); it seems that in patients under active treatment it is possible to improve lung function in some cases (20).

Finally, the radiological expression of the disease is another aspect that presents considerable variability. Regardless of the degree of spirometric involvement, lesions detected in the airway or parenchyma make the patients completely different (Figure 2).

Figure 2. Exemplification of different COPD cases with a similar functional impairment but different radiological expression. (A) Patient with predominantly airway disease; (B) patient with centrilobular emphysema; and (C) patient with panlobular empysema. Open arrows indicate the most typical and suggestive lesion of the three different diseases. By courtesy of Claudio Tantucci, 2015. Respiratory Medicine Unit, Department of Clinical and Experimental Sciences, University of Brescia, Brescia, Italy.

There is no correlation between the different clinical variables at the patient level

One interesting idea is that the different variables associated with COPD are very bad markers of each other. ECLIPSE evaluated the relationship between FEV₁ and different clinical outcomes and observed an interesting phenomenon (Figure 3). When studying the relationship between FEV₁ and various clinical outcomes, it can be observed that, at the cohort level, there is a significant relationship. In this way, ECLIPSE showed a significant correlation between FEV₁ and dyspnea measured by the modified MRC scale, the exacerbations, the distance walked in the 6-minute walk test, or the quality of life (23). However, if one looks carefully at the graphs (Figure 3), the enormous dispersion of the data can be observed. In such a way that the correlation that is clearly observed at the cohort level is not met at the patient level. This discrepancy between the associations found in a cohort and that found at the patient level is what in epidemiology is called ecological fallacy (24), which is a logical fallacy in the interpretation of statistical data where inferences about the nature of individuals are deduced from inference for the group to which those individuals belong. In short, today we have several works that indicate the dispersion of data when two clinical variables are related to each other (25). Therefore, since they do not correlate at the individual level, it follows that it is necessary to evaluate each of them separately in order to establish the impact of the disease on a specific individual.

Figure 3. Relationship between the severity of airflow limitation and breathlessness as assessed by the mMRC questionnaire (panel A), exercise capacity as assessed by the 6 MWD (panel B), reported exacerbations in the year before inclusion in the study (panel C), and health status as assessed by SGRQ-C (panel D). Reproduced from Agustí et al. (23). © 2010 Agusti et al.; licensee BioMed Central Ltd. This is an Open Access figure distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

There are a number of relevant clinical variables associated with outcomes

From a clinical point it seems logical to keep in mind that the variables that are related to the prognosis of the disease should have a greater weight in the evaluation of patients. Interestingly, the number of these variables is not small. Notably, several works have shown the prognostic impact of many variables. Without intending to be exhaustive, these include diffusing capacity (26), lung hyperinflation (27), exacerbations (28), hyperresponsiveness (29), multidimensional indexes (30), chronic bronchitis (31), physical activity (32), or dyspnea (33). In this regard, the use of a second-level pulmonary functional tests (e.g., volumes, or DLCO) may be recommendable for better staging patients. Interestingly, it has been reported that FEV₁ poorly correlates compared to inspiratory capacity with dyspnea (34), maybe the most important symptom in COPD. Interestingly, all these variables have a variable presentation in individuals with COPD. It follows, therefore, that a precise evaluation is necessary in order to be able to consider all these variables independently when establishing the best therapeutic strategy for each patient.

We do have specific therapies for specific patient types

In recent years, the treatment of COPD has been significantly expanded. The amount of possible treatments for the various aspects of the disease is remarkable. Interestingly, although there are clinical trials that provide data on their efficacy and safety, the core of the treatment is nonpharmacological measures, including tobacco cessation, exercise, and vaccinations, along with the use of long-acting bronchodilators. Specifically, long-acting bronchodilators have shown to improve lung function, by decreasing lung hyperinflation and improving symptoms (35–37), which in turn has been proposed as part of the mechanisms by which bronchodilators decrease exacerbations (38). The rest of treatments should be used in patients with some specific characteristics (39).

One promising therapeutic option in the future is biological therapy (40). Although, there is an active program of investigation of highly selective biologic therapeutic agents in the management of COPD, the available studies have not yielded excellent results so far (41–43). Only mepolizumab has shown a potential role in eosinophilic COPD, though with contradictory results (44). Probably one of the key issues regarding biological therapy in COPD is advance in patient selection.

In the immediate future we will have within our reach the use of three drugs, a long-acting antimuscarinic (LAMA), a long-acting β₂ agonist (LABA), and an inhaled corticosteroid (ICS), in a single inhalation device, which has come to be called triple therapy. Up to now, the possibilities of building this triple therapy were several (45). However, with the arrival of triple therapy in one single inhaler we may need (46, 47). The outcome of clinical trials will have to be evaluated in the context of the therapeutic arsenal and the clinical presentation of patients to select the best candidates (48).

There is variability in the clinical response

Finally, another argument to defend the need to target for precision medicine is that the therapeutic response is variable among patients. Clinical studies that show the efficacy and safety profile of available treatments comparing the benefits of various treatment options in terms of average improvement. However, this average response is a simplification of a much more complex reality that is the individualized response to a given treatment. Just as we mentioned above the relevance of the ecological fallacy in the interpretation of the relationships between variables, here we can apply the same principle and remember that the average improvement obtained by a cohort of patients does not imply that each individual patient will respond with the same magnitude. A recent systematic review made these differences evident by putting in one table the improvements in various clinical outcomes of each of the double bronchodilator therapies (49). In this regard, a recent study evaluated the functional response to double bronchodilation (umeclidinium/vilanterol) as a function of the response to single bronchodilation, using an increase of >12% and >200 mL in the trough FEV₁ as a marker of a positive response (50). They found that approximately one-third of patients responded positively to both bronchodilators, another third responded positively only to one agent, and the remainder did not respond positively to either bronchodilator.

Even for clearly established therapeutic interventions such as maintaining daily physical activity and maintaining exercise as a basis for the treatment of COPD (51–53), there is variability in the response to the different training programs and an initial program and an individualized maintenance strategy must be done (54).

How should we do precision medicine in COPD?

Altogether, it seems obvious that due to the different disease presentation and treatment response that we should target for precision medicine in COPD. The key question, at this point, therefore, is how to bring this precision medicine to the real world. At the moment, there is no universally accepted proposal, but several approaches have been used. Integrated care models which account for the different clinical presentation and considering the doctor, the patient and the relatives perspective are relevant (55). So far, the most relevant and discussed in the literature are: (1) the use of independent variables to identify subjects; (2) the use of multidimensional indexes; (3) the so-called clinical phenotypes; and (4) approximation by treatable traits. Next, we will describe each of these initiatives, their contributions and their limitations.

Independent variables

The^{Table 1} use^{Figure 4} of various variables to identify types of patients has been developed by the Global Initiative for Obstructive Lung Disease (GOLD) document (16). In its first versions, the GOLD initiative opted for the use of FEV₁ as the main variable to identify patients of varying severity (56). However, this categorization of patients according to FEV₁ was considered insufficient to capture the complexity of this syndrome. Some voices started to propose to use dyspnea and the number of exacerbations together with pulmonary function to establish a treatment dartboard based on these three axes (57). Finally, in 2011 GOLD updated its patient categorization proposal, establishing four types of patients according to these three characteristics (58), which has been maintained with small changes until the current proposal of 2017 in which 16 types of patients are identified based on the same variables (16). This new approach provides an interesting advantage. By providing information with two digits, it gives short- and long-term prognostic information, suggesting a use of the concept of control in COPD (59), currently under debate (60). Interestingly, some guidelines have also adopted this approach (61).

The use of independent variables to select patient types has its strengths and weaknesses. Table 1 shows the strengths and weaknesses of both the concept of independent variables and the approximation made by the latest version of the GOLD document. One of the main criticisms of the concept used by GOLD is the selection of more than one variable per axis, together with the identification of arbitrary cut points. Specifically, several studies have highlighted the discrepancy between dyspnea scale and CAT, suggesting that other cut points could be more appropriate (62). Specifically, a recent study has identified 18 points as the ideal cut-off point for the CAT questionnaire, 1.9 points for the COPD Clinical Questionnaire total score, and 46.0 points for St. Georges Respiratory Questionnaire total score (63).

Table 1. Strengths and weaknesses of the use of independent variables for the stratification of patients with COPD, both as a general concept and in the GOLD proposal.

	General concept	Current GOLD proposal
Strengths	• It makes clinical sense• With a reduced number of variables, relevant clinical information is obtained	• The selected variables are clinically relevant• The variables are easy to obtain• Provides information on the short- and long-term impact
Weaknesses	• A small number of variables do not capture the full complexity of the disease• Different types of patients do not have a single treatment• The selection is based on information from cohorts: ecological fallacy	• Cut-off points are arbitrary• More than one variable per axis• Different concepts of risk throughout the document• Frequent exacerbator concept is under debate• Discussion about its prognostic impact• Separation of lung function for treatment selection

Another topic for discussion within the GOLD document is the concept of frequent exacerbator, which requires a profound debate, both for its variability over time, and for its therapeutic implications (64). The current concept of frequent exacerbation started from the cohort ECLIPSE, where they showed a type of patient who continued to have exacerbations despite having an active treatment for 3 years (65). Therefore, 3 years of follow-up were needed to identify these patients. However, current guidelines have simplified this concept by limiting the evaluation of the number of exacerbations to the previous year. Probably, the concept of persistent exacerbator, despite adequate pharmacological treatment (66), is more correct and is worth exploring in future studies and addressed in the recommendations documents, since it poses a real challenge for the clinician. Finally, the prognostic impact of the GOLD classification keeps being a source of debate (67), and the separation of lung function for treatment selection is also controversial.

Multidimensional indexes

The use of multidimensional indexes in COPD began with the creation of the BODE (body-mass index, airflow obstruction, dyspnea, and exercise capacity) index by Celli et al. in 2004 (30). This index has been shown to have an important relationship with various clinical outcomes and aspects of the disease, among which are the prognosis of the disease (30), exacerbations (68), quality of life (69), physical activity (70), or certain comorbidities (71). Since its creation, however, various initiatives have been carried out to create new modified indexes that seek to improve the clinical profile of this index, summarized in Figure 4. In addition, various authors have created new indexes not based on BODE, which, in turn, have also had updates (Figure 4).

Figure 4. Multidimensional indexes available.

The use of multidimensional indexes to select patient types has its strengths and weaknesses. The main strength of this approach is that it includes variables with prognostic importance. However, its objective is to evaluate patients from the prognostic point of view. Specifically, it has been shown that its prognostic capacity is long-term (72). Accordingly, they are not made to see the response to treatments and, therefore, have not been evaluated to choose treatments. To date, only the old version (73) of the Spanish guideline (GesEPOC) recommended the use BODE or BODEx indexes. However, this resulted to be difficult to use in daily clinical practice (74, 75), and it was finally abandoned in the latest version (76).

Clinical phenotypes

The concept of clinical phenotype in airway diseases is not new. Already at the CIBA symposium of 1959 several clinical pictures were established according to clinical presentation, different pathological abnormalities, natural history, and prognosis (77), and even the coincidence in the same patient of COPD and asthma was identified (78). However, the current use of the clinical phenotype in the identification of patients is probably part of the article on the future of COPD that was published in 2010 (79). In this work, the authors define a clinical phenotype as a single or combination of disease attributes that describe differences between individuals with COPD as they relate to clinically meaningful outcomes (symptoms, exacerbations, response to therapy, rate of disease progression, or death). Interestingly, if one follows this definition of clinical phenotype, it would be easy to find a considerable amount of disease variables that describe differences between individuals and that are related to relevant clinical outcomes, e.g., dyspnoea, bronchiectasis, bronchial reversibility, or body weight. Hence, the distribution of patients according to phenotypes began to be a relatively simple scheme that has become considerably more complicated (80). One specific point of controversy is the so-called asthma-COPD overlap syndrome (ACOS). The discussion of this phenotype falls out of the objectives of the present review, but there has been so many definitions proposed and recommended (81), contributing to such confusion that this term should better stand for another confusing obstructive syndrome (Figure 5).

Figure 5. Different definitions of asthma-COPD overlap syndrome according to different guidelines. GesEPOC, Spanish COPD Guidelines; CPPS, Czech Pneumological and Pathophysiological Society; FMS, Finnish Medical Society Duodecim; GOLD/GINA, Global Initiative for Chronic Obstructive Lung Disease/Global Initiative for Asthma; GEMA4.0, Spanish Guideline for Asthma Management 4.0. Adapted from (81).

The Spanish recommendations document was the first to include the concept of clinical phenotypes in 2012 (73). This document has been updated in 2014 (82) and recently in 2017 (76). According to the latest version, the document proposes classifying patients into high or low risk according to the classic variables (dyspnoea, exacerbations, and pulmonary function), recommending the establishment of phenotypes in patients classified as high risk. The selected clinical phenotypes are the nonexacerbator, the asthma-COPD overlap (83), the exacerbator with emphysema and the exacerbator with chronic bronchitis (Figure 6).

Figure 6. GesEPOC 2017 stratification proposal. LAMA: long-acting muscarinic antagonist. LABA: long-acting β₂ agonist; ICS: inhaled corticosteroids. Reproduced with permission from Miravitlles et al.; GesEPOC. Arch Bronconeumol 2017 (76). Copyright material is from the Spanish Society of Pulmonology and Surgery Thoracic (SEPAR).

The use of clinical phenotypes to select patient types has its strengths and weaknesses. Table 2 shows the strengths and weaknesses of both the concept of clinical phenotype and the approximation made by the latest version of the GesEPOC document. One of the main points of controversy is that these documents do not collect all possible phenotypes, e.g., bronchiectasis (84). Therefore, it is not a classification that is mutually exclusive and collectively exhaustive. In addition, some of them are in current debate (85). For example, the asthma-COPD overlap need validation (86, 87), and the frequent exacerbator or the nonexacerbator phenotype may cease to be so with time (88, 89), and the concept itself is controversial as discussed above. These aspects are still to be solved in the use of clinical phenotypes.

Table 2. Strengths and weaknesses of the use of clinical phenotypes for the stratification of patients with COPD, both as a general concept and in the GesEPOC proposal.

	General concept	Current GesEPOC proposal
Strengths	• They easily recognize types of patients• They have a specific therapeutic response	• Includes the most common phenotypes• They are easy to identify• They have some associated treatment strategy
Weaknesses	• They are not mutually exclusive, nor collectively exhaustive• The phenotype must be the natural expression of the disease, with treatment is complex• Not all the characteristics that define them are stable over time• The selection is based on information from cohorts: ecological fallacy	• Variable phenotype concept is controversial• Frequent exacerbator concept is under debate• Definition of ACO, a proposal• Does not include relevant phenotypes, e.g., bronchiectasis• There are patients who are left out with no phenotype

Treatable traits

Finally, the approach based on treatable traits is worth commenting. Recently proposed (90), treatable traits are probably the closest proposals to clinical practice. According to these authors, although the concept of a clinical phenotype is useful for research in order to understand the heterogeneity of the disease, it is of limited use in clinical practice, which deals with individual patients (i.e., precision medicine) and not with groups of patients classified according to a particular clinical phenotype, which corresponds rather to stratified medicine (90). So far, no guidelines or recommendations document has used an approach based on treatable traits.

The use of treatable traits to select patient types has its strengths and weaknesses. On the one hand, the main strengths are that it is the closest thing to clinical medicine, it includes all the characteristics that would have treatment, and it is based on the patient information, not in cohorts. However, on the other hand, it should be considered that a scheme based on treatable traits will never be simple, we need to acknowledge that some characteristics may influence the others, and finally and most importantly, trials on treatment response are required for each one of the traits identified (91).

Conclusions

In conclusion, we clearly know that the clinical expression of COPD is very rich and the clinical behavior of patients varies, to which a different treatment strategy must be established. Consequently, we must aim for precision medicine. The relevant question here is how to make this medicine precise. Although there is no universally accepted proposal, the available initiatives provide us with a framework on which to start working and move toward precision medicine. The future is full of possibilities. Possibly the biomarkers will help us to better define the strategy (92), or maybe systems biology is the answer (93). In any case, the recommendations documents should focus on the identification of types of patients with specific responses to treatments, with patient-based strategies, not based on cohorts, with the ultimate goal of bringing the best possible medicine to each patient in particular.

Declaration of interest

JLLC declares relevant financial activities outside the submitted work during the last 36 months in the form of grants, personal fees, or nonfinancial support from Novartis, Boehringer Ingelheim, Menarini, Grifols, Esteve, Ferrer, Rovi, Gebro Pharma, Bial, and Teva. SC declares relevant financial activities outside the submitted work during the last 36 months in the form of grants, personal fees, or nonfinancial support from GlaxoSmithKline, Novartis Farma SPA, Chiesi SPA, AstraZeneca, Valeas, Boehringer Ingelheim, and Guidotti Malesci.