Multidimensional approach to non-cystic fibrosis bronchiectasis: the FACED score

Introduction

Bronchiectasis is a permanent and, usually, progressive bronchial dilation resulting from the infection and chronic inflammation of the airway, leading to destruction and remodelling of the bronchial wall [1–3]. Bronchiectasis is associated with chronic and frequently purulent expectoration, multiple exacerbations and progressive, potentially disabling dyspnoea. These events gradually worsen the health-related quality of life and lung function of affected patients [4–6]. In recent years, bronchiectasis has become a major health concern for several reasons: a significant increase in the number of diagnoses [7, 8], the increase in the mortality rate [9] and in hospital admissions [10], the considerable health costs involved (US$7827 per hospitalisation) [10], its negative impact on the quality of life [6] and pulmonary function of patients (annual loss of nearly 50 mL of forced expiratory volume in 1 s (FEV₁)) [5] and its deleterious effects on underlying diseases [11].

Several single variables have been used to predict key outcomes of bronchiectasis, such as decline in lung function, various clinical parameters, radiological extension and the presence of chronic colonisation by Pseudomonas aeruginosa [9, 12, 13]. However, as with other airways diseases, the severity and prognosis of bronchiectasis cannot be adequately defined with one single variable. In chronic obstructive pulmonary disease (COPD), for example, some authors have noted a low correlation between symptoms and pulmonary function [14], and this has led to the construction of easy-to-use, multidimensional indices that have provided a more accurate prognosis value than the parameter normally used (FEV₁ in the case of COPD) [15, 16]. With this study, we aimed to create and validate an easy-to-use multidimensional grading system that could classify the severity of non-cystic fibrosis (CF) bronchiectasis more accurately than any single variable, according to its capacity to predict the 5-year risk of all-cause mortality.

Results

Multivariate analysis: construction of the score

table 4 shows the results of the logistic regression analysis after including the 12 variables that were initially selected. The following showed a statistically significant association with 5-year all-cause mortality after diagnosis: age (OR 1.08, 95% CI 1.04–1.11; p=0.001); post-bronchodilator FEV₁ % predicted (OR 0.96, 95% CI 0.94–0.97; p=0.0001); radiological extension of bronchiectasis (OR 1.41, 95% CI: 1.07–1.86; p=0.016); presence of chronic colonisation by P. aeruginosa (OR 1.95, 95% CI 1.01–3.76; p=0.045); and dyspnoea (OR 2.42, 95% CI 1.23–4.80; p=0.01). The optimum cut-off points in the dichotomisation of the variables were as follows. Age: >70 years versus ≤70 years; FEV₁: >50% pred versus ≤50% pred; modified Medical Research Council (mMRC) dyspnoea score: I–II versus III–IV; radiological extension: zero to two lobes with bronchiectasis versus more than two affected lobes.

View this table:

• View inline
• View popup

Table 4– Predictive capacity for mortality of the 12 initial eligible variables for inclusion in the score

table 5 shows the result of the logistic regression analysis, including as independent variables the five dichotomised variables that would comprise the final score called FACED. FEV₁ (F, β=1.65, OR (95% CI) 5.19 (2.76–9.75), p=0.0001); age (A, β=1.61, OR (95% CI) 4.98 (2.67–9.28), p=0.0001); chronic colonisation by P. aeruginosa (C, β=0.86, OR (95% CI) 2.37 (1.28–4.58), p=0.006); extension of bronchiectasis (E, β=0.62, OR (95% CI) 1.87 (1.01–3.46), p=0.04) and dyspnoea (D, β=1.01, OR (95% CI) 2.75 (1.46–5.18), p=0.002). The β-coefficients were rounded in each variable to the nearest whole number to simplify the final score, as shown in table 6.

View this table:

• View inline
• View popup

Table 5– Predictive capacity for mortality of the different dichotomised variables included in the final score

View this table:

• View inline
• View popup

Table 6– Final score, cut-off points of the dichotomised variables and scoring of each variable

Figure 2a shows that the area under the receiver operating characteristic curve (AUC) of the final score to predict 5-year all-cause mortality was 0.87 (the greater the AUC, the better the prognostic value of the FACED score. Any AUC >0.8 is considered excellent). Figure 3a shows how the division into three scoring groups differentiated bronchiectasis into three distinct mortality groups, according to the results of the Kaplan–Meier analysis: mild (4.3% mortality), moderate (24.7% mortality) and severe (55.9% mortality) bronchiectasis, given that the log-rank test marked statistically significant differences in the curves when they were compared two by two. Finally, when the AUC of the cohort of patients was calculated for each centre, it was >0.80 in every case apart from one centre, where it was 0.77 (range 0.77–0.92).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2–

Receiver operating characteristic curves and areas under the curve (AUC) to determine the overall predictive value of mortality of the proposed score, using a) the construction cohort (AUC 0.87, 95% CI 0.82–0.91) and b) the validation cohort (AUC 0.83, 95% CI 0.78–0.89).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3–

Kaplan–Meier curves for all-cause mortality corresponding to the three bronchiectasis scoring groups (mild 0–2 points, moderate 3–4 points and severe 5–7 points) in a) the construction cohort and b) the validation cohort. Log-rank test a) mild bronchiectasis versus moderate bronchiectasis 29.66, p=0.0001; mild bronchiectasis versus severe bronchiectasis 178.65, p=0.0001; moderate bronchiectasis versus severe bronchiectasis 35.067, p=0.0001; and b) mild bronchiectasis versus moderate bronchiectasis 20.02, p=0.0001; mild bronchiectasis versus severe bronchiectasis 126.81, p=0.0001; moderate bronchiectasis versus severe bronchiectasis 28.28, p=0.0001. a) Mild: 10 (4.3%) out of 234 patients died; moderate: 24 (24.7%) out of 98 patients died; severe: 44 (68.8%) out of 64 patients died. b) Mild: 14 (5.6%) out of 249 patients died; moderate: 22 (21.6%) out of 103 patients died; severe: 38 (55.9%) out of 68 patients died.

If we only consider the deaths from respiratory causes, the FACED score presents an AUC of 0.85 (range 0.82–0.89) and is also capable of differentiating, to a statistically significant degree, the three aforementioned groups of patients in relation to their prognosis (figure 4a).

• Download figure
• Open in new tab
• Download powerpoint

Figure 4–

Kaplan–Meier curves for respiratory mortality corresponding to the three bronchiectasis scoring groups (mild 0–2 points, moderate 3–4 points and severe 5–7 points) in a) the construction cohort and b) the validation cohort. Log-rank test a) mild bronchiectasis versus moderate bronchiectasis 32.55, p=0.0001; mild bronchiectasis versus severe bronchiectasis 178.24, p=0.0001; moderate bronchiectasis versus severe bronchiectasis 32.84, p=0.0001; and b) mild bronchiectasis versus moderate bronchiectasis 18.09, p=0.0001; mild bronchiectasis versus severe bronchiectasis 120.78, p=0.0001; moderate bronchiectasis versus severe bronchiectasis 26.84; p=0.0001. a) Mild: 2 (0.9%) out of 234 patients died; moderate: 15 (15.3%) out of 98 patients died; severe: 33 (51.6%) out of 64 patients died. b) Mild: 6 (2.4%) out of 249 patients died; moderate: 14 (13.6%) out of 103 patients died; severe: 31 (45.6%) out of 68 patients died.

Discussion

Due to the multidimensional nature of bronchiectasis, no single isolated parameter has yet proved to have sufficient power for any overall determination of its severity or prognosis. Although the measurement of quality of life could be eligible for this, its application is not generalised and the measuring instruments are not usually validated for individual use [21, 22]. Therefore, as with other airway diseases such as COPD [14, 15] and infectious diseases such as pneumonia [23], it is necessary to construct and validate a score that pinpoints more effectively than any single variable the severity or prognosis of bronchiectasis by embracing the various clinical, functional, radiological and microbiological aspects characteristic of the disease. This study presents the construction and validation of a score that it is easy to calculate, obtain and interpret, while also covering all the aspects mentioned above. This score has been named FACED, an acronym for FEV₁, age, chronic colonisation, extension and dyspnoea.

It should be emphasised that all the variables used for the construction and validation of the score were recorded as soon as possible after the radiological diagnosis of bronchiectasis (within a maximum of 6 months), apart from the hospitalisations, which corresponded to the year prior to a patient’s admission into the study. This is crucial for the avoidance, as far as possible, of any initial interference with the various therapeutic measurements of the prognostic variables studied during the follow-up.

Of the five variables that eventually comprised the score, age and FEV₁ presented the greatest predictive power of mortality. Various studies on patients with non-CF bronchiectasis have demonstrated the relationship of both these variables with an increase in mortality [12, 13]. The high colinearity between FEV₁ and forced vital capacity made it possible to choose FEV₁ to represent the patient’s functional state, as this is the variable most commonly used in the evaluation of airflow obstruction, the most common functional pattern in patients with bronchiectasis [24–26]. Although it is possible that other functional variables, such as those related to exercise tests [27], could have a greater predictive value than FEV₁, they were not taken into account for the construction of this score as they are not generally applied.

The remaining variables, which represent the extension of the disease, the patient’s clinical picture and microbiological aspects, also presented significant predictive power with respect to mortality, albeit to a lesser extent than age and FEV₁. Dyspnoea is usually a variable with great predictive power in most studies of airway diseases, independent of lung function, and this is also the case with bronchiectasis [25], so it should not be surprising that it is one of the components of our score. It was measured using the mMRC scale [26], as this is simple and widely used. Other clinical findings were related to excess mortality in the univariate analysis (for example, the macroscopic appearance of sputum), although this was ultimately excluded from the definitive score in favour of chronic colonisation by P. aeruginosa, with which it is closely associated. Although the measurement of the quantity of sputum produced per day could be an important variable, this was rejected because it is not normally evaluated in general visits. The extension of bronchiectasis, quantified as the number of lobes affected, was incorporated into the score, even though its relative weight was lower than that of other variables. This choice is supported by authors who have found a clear dissociation between the radiological extension of bronchiectasis and its clinical and functional aspects [25], on account of the low correlation between these different variables. Other radiological variables that are more difficult to obtain and interpret, such as the characteristics of the bronchial wall, were not included, even though some studies have found them to be very important in the follow-up of these patients [27, 28]. Lastly, one key characteristic of patients with non-CF bronchiectasis is the microbiological profile. The microbiological profiles, including fungi and atypical mycobacteria, of the patients in our study were meticulously analysed. Of all the different possible combinations, the variable that ultimately demonstrated the greatest capacity to independently predict the probability of 5-year all-cause death was the presence of chronic colonisation by P. aeruginosa. Neither the presence of chronic colonisation by multiresistant Gram-negative bacilli nor the isolation of Staphylococcus aureus, atypical mycobacteria and fungi, presented any significant independent predictive power for mortality. Therefore, it was finally decided to include only chronic colonisation by P. aeruginosa in the score, as this is the only microorganism that has been related in the current literature with an increase in mortality and a poorer functional evolution in patients with non-CF bronchiectasis [5, 29, 30].

It should be noted that the aetiology of bronchiectasis was not considered for the construction of this score, since this usually requires a series of time-consuming complementary tests that prevent the score from being calculated immediately after diagnosis, and therefore limit its application. In any case, the implementation of the score does not obviate the need to investigate the aetiologies of bronchiectasis, especially those that can benefit from specific treatments capable of intervening positively in its evolution [31, 32].

Finally, the predictive capacity of the FACED score was calculated when it was applied only to deaths from respiratory causes 5 years after the diagnosis of bronchiectasis, where it maintained its high predictive power (AUC >0.80 in both the construction and validation cohorts). We also observed a significant discriminatory capacity in the three groups of patients spanning the best and worst prognosis, again in both the construction cohort and the validation cohort.

Among the strengths of the present study, it is important to stress the very large cohort of patients included, who were well characterised as a result of their follow-up via a standardised protocol in specialised outpatient clinics by professionals with great experience in bronchiectasis. Also worthy of note are the score’s multidimensional nature and the ease with which it can be obtained and interpreted, as it covers the main fields typical of patients with bronchiectasis. Lastly, an important strength of our results is their validity in all the participating centres, even though the patients from these centres presented very different characteristics and aetiologies; this indicates the robustness of the constructed score, as well as confirming its validation.

One limitation of the study is the possible influence on mortality of the various treatments administered to the patients over the 5 years of the follow-up. However, although the observational nature of our study implies the presence of limitations intrinsic to this type of methodology, no medical treatments for non-CF bronchiectasis have proved to be related to long-term mortality [31, 32]. Finally, an international validation of the FACED score, including large series of patients with non-CF bronchiectasis, would be desirable.

In conclusion, our results present the construction and validation of the FACED score, which is simple to use and interpret in patients with non-CF bronchiectasis, as well as being multidimensional, on account of its assemblage of different key aspects of the disease. It shows an excellent predictive capacity for mortality and an adequate validation, independent of the varying characteristics of patients from the different participating centres, including different aetiologies, making it a robust and attractive clinical tool. New studies are required to evaluate this score’s sensitivity to change by applying therapeutic variables in the follow-up of these patients.