Clinically Important Deterioration (CID) and Ageing in COPD: A Systematic Review and Meta-Regression Analysis According to PRISMA Statement

Introduction

Chronic obstructive pulmonary disease (COPD) is defined by the latest Global Initiative for Chronic Obstructive Lung Disease (GOLD) report as a heterogeneous lung condition characterized by chronic respiratory symptoms and a persistent, often progressive, airflow obstruction.¹ The need for a multi-dimensional approach, assessing lung function, acute exacerbations, and quality of life of COPD patients, led to the development of the composite clinically important deterioration (CID) endpoint in 2016.² In its original formulation, a CID was defined as any of the following: a decrease of ≥100 mL from baseline in forced expiratory volume in 1 s (FEV₁) and/or a ≥4-unit increase from baseline in St. George’s Respiratory Questionnaire (SGRQ) total score and/or the occurrence of a moderate-to-severe exacerbation of COPD (AECOPD). AECOPD was defined as an acute worsening of COPD symptoms requiring the use of additional treatment.²

The CID concept has proven itself as a reliable tool to quantify the impact of pharmacological treatment in several post-hoc analyses of randomized controlled trials (RCTs),³ overcoming the exclusive effect on FEV₁, which might correlate weakly with patient reported outcomes (PRO).⁴ Throughout the years, the composite CID endpoint also showed a prognostic value⁵ and alternative definitions, including, among the others, the COPD Assessment test (CAT) score⁶ and the Transition Dyspnea Index (TDI),⁷ were developed.

COPD patients can experience a rather fast decline in lung function over the years, ranging from 33 to 66 mL/year in FEV₁ according to different studies.^8,9 However, this annual rate is below the 100 mL decrease considered by the original CID definition, which is indeed intended to be greater than the expected functional decline; for this reason, this value is considered the minimal clinically important difference (MCID) for FEV₁ in COPD.¹⁰

According to the well-known Fletcher–Peto curve¹¹ and to more recent retrospective evidence, the FEV₁ decrease could accelerate with ageing.¹² As a matter of fact, this trend can be identified also in healthy subjects, where the median FEV₁ decline is 22.4 mL/year, ranging from 3 mL/year in the 40–49 years decade to 34 mL/year in the 70–79 years decade, as reported by a recent systematic review.¹³

Likewise, older COPD patients have a higher risk of future AECOPD and a higher mortality for acute exacerbations,^14,15 showing an age-related dysfunction of the immune system.¹⁶ Moreover, elderly COPD patients are more likely to suffer from comorbidities,¹⁷ that per se increase the risk of moderate and severe AECOPD.¹⁸

Conversely, ageing does not seem to affect health-related quality of life (HRQL) in COPD patients, since younger patients, aged 50–64 years, actually score higher in SGRQ than patients aged 65–80 years, probably due to a higher impact of dyspnea.¹⁹

Despite this evidence, to date, no specific studies have been carried out to assess the correlation between ageing and CID in COPD patients.

The meta-regression analysis is a statistical method which usually follows a traditional meta-analysis: on one hand, it is mainly used to identify possible confounding factors in a meta-analysis, on the other hand, meta-regression analysis can also be used as a tool to investigate if and how a specific variable, such as ageing, may modulate certain outcomes.^20–23

Therefore, the aim of this systematic review and meta-regression analysis was to investigate the impact of ageing on CID in COPD patients, according to the current PRISMA statement.²⁴

Materials and Methods

Search Strategy and Study Eligibility

This systematic review and meta-regression analysis was performed according to PRISMA-P guidelines.²⁴ The PRISMA 2020 flow diagram is shown in Figure 1. The PRISMA-P checklist is reported in Table S1.²⁴

Figure 1 PRISMA 2020 flow diagram for the identification of the studies included in the systematic review. Abbreviation: PRISMA, Preferred Reporting Items for Systematic Reviews and Meta-Analyses. Notes: PRISMA figure adapted from Page MJ, McKenzie JE et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ. 2021;372:n71. Creative Commons.25

A comprehensive literature search was carried out for Phase III RCTs or post-hoc analyses of RCTs written in English and assessing CID in COPD patients.

The patient problem, intervention, comparison, and outcome (PICO) framework were used for the literature search.²⁶ The “Patient problem” included COPD; the ‘Intervention’ regarded inhaled therapy; the ‘Comparison’ was performed vs placebo (PCB) or across different inhaled drugs; the ‘Outcome’ was the association between CID and ageing.

The search was performed in MEDLINE and Scopus to find relevant studies published up to February 9th, 2023.

The following search string was used in the database was: ((clinically important deterioration) OR CID) AND COPD. References of previous high-quality reviews were checked to identify further RCTs, if any.⁵

Eppi-Reviewer 4 (EPPI-Centre Software. London, UK) was used to manage data in literature and facilitate the collaboration across reviewers for the selection of the studies.

Study Selection

Phase III RCTs or post-hoc analyses of RCTs that analyzed CID in COPD patients were included in the systematic review and meta-regression analysis. Trials not including at least FEV₁ decline, SGRQ increase, and AECOPD in the CID definition were excluded from the analysis. Two reviewers (GMM and LC) independently checked for study selection and any difference in opinion was resolved by consensus.

Data Extraction

Data from the Phase III or RCTs post-hoc analyses included in this systematic review and meta-regression analysis were extracted from journal articles, and/or Supplementary Data Files, and/or the ClinicalTrials.gov database.

Data were checked for reference, clinical trial ID, duration and characteristics of the study, number of analyzed patients, treatments administered including the doses of medications, regimen of administration, main inclusion criteria, age, gender, smoking habit, post-bronchodilator FEV₁, AECOPD, SGRQ, COPD assessment test (CAT), modified medical research council dyspnea scale (mMRC), and Jadad Score.²⁷

Data were extracted according to the DECiMAL recommendations.²⁸ Cohen’s Kappa score was used to assess the inter- and intra-rater reliability for data abstraction (≥0.80: excellent agreement; ≥0.61 and <0.80: substantial agreement; ≥0.41 and <0.61 moderate agreement; <0.41 poor agreement).²⁹

Endpoint

The primary endpoint of systematic review and meta-regression analysis was to assess the risk of CID according to ageing.

Data Synthesis and Analysis

A pairwise meta-analysis was carried out to quantify the risk of CID in COPD patients enrolled in RCTs. Obtained results were reported as relative risk (RR) and 95% confidence interval (95% CI).

A common effect size cannot be assumed because data were extracted from a series of studies performed by investigators operating independently. Therefore, the binary DerSimonian-Laird random-effects model was used to balance the study weights and correctly assess the effect estimates and relative 95% CI. Subgroup analyses were carried out according to specific inhaled treatments.

A meta-regression analysis via random-effect method was also performed to investigate whether ageing, alone or in association with other potential effect modifiers, might modulate the risk of CID in COPD. The meta-regression analysis was carried out by plotting the outcome variables obtained from the pairwise meta-analysis with the explanatory variables reported in the included studies.^20–23 The resulting regression coefficient indicates how strongly the explanatory variables may modify the effect induced by a specific treatment.²² The meta-regression analysis reports a positive or negative correlation between the effect estimates and the potential effect modifiers for statistically significant regression coefficient. In this study, meta-regression analysis was used to investigate the impact of ageing on CID.

Study Quality, Bias, and Quality of Evidence

The risk of bias for the included studies was quantified via Jadad score, ranging from 1 to 5 (score ≤2: low quality; score = 3: medium quality; score ≥4: high quality).²⁷

Heterogeneity (I²) was assessed to analyze the between-study dissimilarity, as previously reported.²⁰

Funnel plot and Egger’s test were performed to investigate the origin and risk of publication bias related to significant and/or substantial (I²>50%) level of heterogeneity if more than 10 studies were included in the meta-analysis.^30–33

The quality of the evidence was assessed according to GRADE system (++++: high quality; +++: moderate quality; ++: low quality; +: very low quality).³⁴

Two reviewers (GMM and LC) independently assessed the study quality, bias, and quality of evidence and any difference in opinion was resolved by consensus.

Software and Statistical Significance

Open-MetaAnalyst Was Used to Perform

The pairwise meta-analysis and meta-regression analysis were analyzed via Open-MetaAnalyst software²⁰ and the quality of evidence assessed via the GRADEpro GDT software.³⁴ The statistical significance was identified for P value <0.05.

Results

Study Characteristics

Of the 332 potentially relevant records identified in the initial search, 17 studies were deemed eligible for qualitative and quantitative syntheses.^{2,6,7,35–48} Full-text papers were published between 2016 and 2021.

Data obtained from 55219 COPD patients were extracted from 15 post-hoc analyses of RCTs, a subgroup analysis of a RCT,⁴⁰ and a RCT prospectively assessing the CID composite endpoint.⁴² The duration of the studies ranged from 3 to 48 months and the age of the COPD populations enrolled in the studies ranged from 62.8 to 65.3 years.

A certain level of study population overlap was detected across the post-hoc analyses. Namely, the SHINE trial⁴⁹ was analyzed in both the studies by D’Urzo et al⁴⁵ and Anzueto et al,⁷ the ZEP117115 trial⁵⁰ was included in both the post-hoc analyses by Maleki-Yazdi et al⁴⁸ and Singh et al,² the study by Chen et al⁴⁰ was a subgroup analysis of the PINNACLE-4 trial,⁵¹ which was also included in the post-hoc analysis by Zheng et al.³⁹

Different treatments were analyzed as following: long-acting bronchodilators (LABD) vs PCB,^{2,37–40,45,47} long-acting muscarinic antagonist (LAMA) vs long-acting β₂-adrenoceptor agonist (LABA),^2,40 dual bronchodilation vs PCB,^2,39,40,47 dual bronchodilation vs LABD,^{2,7,36,39,40,42,47,48} dual bronchodilation vs inhaled corticosteroid (ICS)/LABA,^35,46 ICS/LABA vs LABD,⁴¹ triple combination vs LABD,⁴³ triple combination vs dual bronchodilation,^35,43 triple combination vs ICS/LABA.^6,35,43,44

The main characteristics of the studies included in the systematic review and meta-regression analysis are reported in Table 1, whereas Table 2 shows the different CID definitions for study.

Table 1 Main Characteristics of the Studies Included in the Systematic Review and Meta-Regression Analysis

Table 2 Different CID Definitions Used in the Studies Included in the Systematic Review and Meta-Regression Analysis

Pairwise Meta-Analysis

The overall pairwise meta-analysis indicated that the pharmacological treatment of COPD is effective in reducing the risk of CID in COPD (RR: 0.81, 95% CI 0.79–0.84; P < 0.001, GRADE +++) (Figure 2). The effect estimates were affected by high and significant heterogeneity (I² 92.28%, P < 0.001).

Figure 2 Forest plot of meta-analysis concerning the impact of pharmacological treatment on the risk of CID in COPD. Abbreviations: AB, aclidinium bromide; BUD, budesonide; CID, clinically important deterioration; COPD, chronic obstructive pulmonary disease; FF, fluticasone furoate; FOR, formoterol fumarate; FP, fluticasone propionate; GLY, glycopyrronium bromide or glycopyrrolate; ICS, inhaled corticosteroid; IND, indacaterol; LABA, long-acting β2-adrenoceptor agonist; LABD, long-acting bronchodilators; LAMA, long-acting muscarinic antagonist; O, olodaterol; PCB, placebo; SAL, salmeterol; T, tiotropium bromide; UMEC, umeclidinium bromide; VI, vilanterol. Notes: The subgroup and overall effect estimates resulting from the meta-analysis are reported in bold; the red-dashed line indicates the average relative risk resulting for the overall effect estimate.

More specifically, when compared to PCB, LABD and dual bronchodilation significantly (P < 0.001) reduced the risk of CID (RR: 0.76, 95% CI 0.70–0.84, GRADE +++ and RR: 0.67, 95% CI 0.60–0.76, GRADE +++, respectively). When compared to LABD, dual bronchodilation, ICS/LABA and triple combination significantly (P < 0.001) reduced the risk of CID (RR: 0.84, 95% CI 0.80–0.87, GRADE +++, RR: 0.89, 95% CI 0.86–0.93, GRADE ++++ and RR: 0.89, 95% CI 0.86–0.92, GRADE ++++, respectively). Triple combination significantly (P < 0.001) reduced the risk of CID (RR: 0.92, 95% CI 0.90–0.95, GRADE ++++) compared to dual bronchodilation. When compared to ICS/LABA, dual bronchodilation and triple combination significantly reduced the risk of CID (RR: 0.91, 95% CI 0.86–0.97, P < 0.01, GRADE +++ and RR: 0.76, 95% CI 0.66–0.87, P < 0.001, GRADE +++, respectively). Only the comparison between LAMA and LABA did not significantly modulate in the risk of CID (RR: 0.94, 95% CI 0.80–1.11, P > 0.05, GRADE +++).

The study arms comparing the same treatment groups were excluded from the pairwise meta-analysis. Namely, the study by Zheng et al, Maltais et al and Singh et al^39,42,47 included a subgroup comparing a long-acting muscarinic antagonist (LAMA) and a LABA, both belonging to the LABDs. The post-hoc analysis by Singh et al⁴³ included the TRINITY trial,⁵² comparing two different triple combinations, while the study by D’Urzo et al⁴⁷ compared different LAMAs.

Meta-Regression Analysis

An overall meta-regression analysis was performed for age, to assess whether it might modulate the risk of CID in COPD patients. Although age was not a significant (P > 0.05) potential effect modifier, it resulted a trend toward significance (P = 0.063) in the linear relationship with respect to the RR of CID (slope: 0.051, Figure 3).

Figure 3 Graphical representation of the meta-regression analysis for age with respect to the risk of CID. Abbreviation: CID, clinically important deterioration. Note: The size of the circles is proportional to the sample weights.

When considering specific treatment subgroups, age was a significant (P < 0.05) potential effect modifier for CID only for dual bronchodilation vs PCB and dual bronchodilation vs ICS/LABA subgroups (slope: 0.268 and 0.103, respectively).

The impact of age on CID was also analyzed in association with other characteristics of the study population, namely, gender, smoking habit, post-bronchodilator FEV₁, AECOPD, SGRQ, CAT, and mMRC. Only when associated with post-bronchodilator FEV₁, age significantly modulated the risk of CID (P = 0.027, slope = 0.061), indicating that the pharmacological treatment of COPD is less effective in older patients with reduced lung function. When associated with the other population characteristics, age did not result in a significant modulating factor (P > 0.05).

Among the study characteristics, the duration of the trials represented a significant (P < 0.001) albeit modest potential effect modifier for the risk of CID in COPD (slope: 0.008, Figure S1), suggesting a greater impact of pharmacological treatment in shorter-term studies.

Bias and Quality of Evidence

All studies (100.0%) were ranked as being of medium- to high-quality in agreement with Jadad score (Table 1). One study was of medium quality (Jadad score = 3),³⁷ and all the others were of high quality (Jadad score >3).

The assessment of the quality of evidence carried out via the GRADE system reported a general moderate (+++) to high-quality (++++) of evidence.

The visual inspection of the funnel plot confirmed the presence of overall heterogeneity, with data reported in an apparent symmetric way. Nevertheless, Egger’s test indicated that the overall effect estimates resulting from this meta-analysis were not affected by significant bias and confirmed symmetry for the reported data. Details on the funnel plots and Egger’s test analyses are reported in Figure 4.

Figure 4 Funnel plot (A) and graphical representations of Egger’s test (B). Abbreviation: SND, standard normal deviate. Notes: Funnel plot represents a visual approach to check for the existence of publication bias by assessing the symmetry of study distribution, whereas Egger’s test is a regression assay that permits to statistically quantify the extent of Funnel plot asymmetry; dotted lines in the Egger’s test represent 90% prediction bands.

Discussion

The findings of this quantitative synthesis indicate with moderate-to-high quality of evidence that the pharmacological treatment of COPD significantly reduces the risk of CID in the overall study population.

Moreover, three specific treatment subgroups, namely, LABD vs PCB, dual bronchodilation vs PCB, and triple combination vs ICS/LABA reached the minimal clinically important difference in the risk of CID, according to a detected RR value ≤0.75, as previously reported.^53,54 As expected, LAMA vs LABA did not modulate the risk of CID. Although data reported a certain level of heterogeneity, Egger’s test excluded that results were affected by significant bias.

Thus, the findings of this quantitative synthesis support the use of CID as an outcome for pharmacological trials in COPD, as already suggested in previous narrative reviews.^3,5

Of note, the meta-regression analysis indicated that age may represent a potential effect modifier for the impact of pharmacological treatment against the risk of CID in COPD, although just a trend towards significance was detected. As a matter of fact, despite the limited range of age between 62.5 and 65.5 years of the investigated population, it resulted that older patients may have a numerical greater risk of CID. Such a narrow age range is related to the intrinsic characteristics of the COPD populations enrolled in the primary RCTs, which evidently appear to differ from the broader demographic diversity detectable among real-life COPD patients. Therefore, it is expected that the impact of ageing could have a significant and detrimental impact on the risk of CID when considering a wider range of age in a real-life setting.

Interestingly, when the meta-regression analysis for age was carried out along with lung function, age significantly modulated the risk of CID when associated with post-bronchodilator FEV₁. On the other hand, the latter evidence suggests that older COPD patients with a larger impairment in lung function may experience a higher risk of CID.

In this regard, the Age, Dyspnea and airflow Obstruction (ADO) index was reported to be a prognostic factor in COPD.⁵⁵ Further evidence indicated that the ADO index may be a predictor of 3-year mortality in COPD.⁵⁶ It is important to underline that also the occurrence of a CID in the natural history of COPD patients correlates with all-cause mortality,^14,57 supporting the existence of a correlation between CID and age, especially when associated with lower FEV₁.

The evidence that older patients with impaired lung function are more prone to undergo a CID may have a substantial impact on daily clinical practice. In this respect, since the general population ageing is increasing as well as the age at death in COPD patients, and FEV₁ decline is greater in older patients,^9,58 it is expected that the prevalence of elderly COPD patients with poor clinical condition will increase in the future (Figure 5). This COPD population is per se characterized by a higher burden of comorbidities^17,59 that need to be managed properly, especially on the cardiovascular disease counterpart.^60,61 In this scenario, elderly COPD patients not only need to be screened and treated for comorbidities but might also benefit from a tight functional follow-up and an early pharmacological therapy in order to reduce the risk of CID.

Figure 5 CID risk increases in older patients with worse lung function. Abbreviation: CID, clinically important deterioration; COPD, chronic obstructive pulmonary disease; FEV1, forced expiratory volume in the first second. Notes: The meta-regression analysis suggested that age significantly modulates the risk of CID when associated with lower post-bronchodilator FEV1.

The main limitations to this study are intrinsic to the quantitative synthesis methods, which are based on large sample approximations, and to the intrinsic weakness of the included studies.^62,63 Indeed, most of the studies were post-hoc analyses of RCTs, and only one was an RCT prospectively assessing the CID composite endpoint. Moreover, the studies spanned a duration ranging from 3 to 48 months, introducing a certain level of temporal heterogeneity that could potentially act as an effect modifier, suggesting greater efficacy of pharmacological treatment in shorter-term studies. Finally, also the limited age range may represent a limitation of our analysis.

Conclusion

This systematic review and meta-regression analysis indicates that inhaled therapy is effective in reducing the risk of CID in COPD, although such a protective effect may be affected in older patients with reduced lung function. Certainly, further studies designed to directly assess the impact of ageing on CID are needed to confirm these findings in COPD.