We searched the following databases: MEDLINE, EMBASE, Cochrane Central Register of Controlled Trials, and the Cumulative Index to Nursing and Allied Health Literature, from their inception to December, 31st 2019, with no language restriction. We also reviewed the references of retrieved articles for additional studies. The search was limited only to RCTs using the terms: COPD, Fatigue, Fatigability, Muscle activation, Muscle endurance, Muscle Performance, Voluntary Activation, Motoneuron excitability, Force Development, Exercise, and Rehabilitation.

We included RCTs involving patients with COPD according to the Global Initiative for Chronic Obstructive Lung Disease recommendations in all stage of severity.18 We excluded studies involving patients with (1) an acute exacerbation within 4 weeks before starting the intervention; (2) major comorbidities such as chronic heart failure, asthma, and sleep-related disorders.

We included studies administering exercise training programs, with the following characteristics:

• -

  Involving in- or out-patients, and home- or community-based programs;

• -

  Programs of 2 week minimum duration;

• -

  Endurance and/or strength training of lower and/or upper limbs;

• -

  Comparing exercise training with usual care, (defined as conventional medical care without any prescription of exercise training or physical activity) and/or education.

We included any study assessing:

• -

  Perceived fatigue: by means of scales or questionnaires evaluating subjective perception of fatigue;

• -

  Performance fatigue: by means of objective measures evaluating changes in muscle performance after a fatiguing task such as muscle force or electromyography trace during a maximal voluntary contraction (MVC) or stimulated muscle resting Twitch.

Two investigators (LB, MP) independently conducted the search of the databases, screening all titles and/or abstracts for the inclusion criteria. Reviewers then retrieved abstracts and/or full-text papers of all potentially eligible studies and maintained records of all studies not meeting the inclusion criteria. They also provided the reasons for their exclusion.

The investigators inserted the data of potentially eligible articles in a Microsoft Excel (2013 version, Microsoft, Redmond, WA) dedicated database. Disagreement between investigators about eligibility was resolved by discussion and consensus. If consensus could not be reached, a third investigator (MV) adjudicated. All data were checked for accuracy. Missing data were requested by e-mail to the authors. An investigator (LB) retrieved the full-text of the included studies and inserted their data into a dedicated electronic database. Another investigator (MP) independently extracted data from the same studies.

The information collected included the background characteristics of the research reports: characteristics of participants in the study; number of participants who dropped out or withdrew from the study; a full description of the exercise training programs (setting, components, duration, and characteristics); assessments, and associated results. If a study reported multiple group comparisons (e.g., exercise training plus free walking vs exercise training alone vs conventional care), treatment groups performing exercise that could be relevant to outcomes were combined into one virtual intervention group, and this group was compared to the group receiving conventional care.

The investigators assessed papers for bias using the Cochrane Collaboration's tool for assessing risk of bias in RCTs.19 Risk of bias was assessed according to the following domains: sequence generation; allocation concealment; blinding of participants; blinding of personnel; blinding of outcome assessment; incomplete outcome data; selective outcome reporting, and other biases.

Statistical analysis was performed using STATA version 11.2 software (Stata, College Station, TX). All data were extrapolated from the corresponding full-text studies. For outcome, we recorded mean and standard deviation (SD) of variation from baseline to the end of the study. When SD was not available, we calculated them from standard errors, confidence intervals, or t-values or contacted the trial authors by email for clarification. We excluded studies with data other than mean and SD.20

The standardized mean difference (SMD) was calculated. Heterogeneity of studies was assessed by performing the Q-test considering values of P < 0.01 as significant. The first step of the evaluation was conducted by fixed-effect models using the Mantel-Haenzel method. If a significant heterogeneity among studies was found, a random effect evaluation by the Der-Simonian and Laird method approach was performed.21 Forest plots were used to detect publication bias for meta-analysis evaluation including more than 8 studies.22 For the meta-analysis, only the outcome measures included in at least two studies were analysed.

Seven studies24,26,28,32–34,36 were conducted mainly in an out-patient setting, six at home 25,27,30,31,37,38 and three studies23,29,35 in both settings.

The total duration of programs ranged from 6 to 24 weeks. There were two to seven sessions per week. The duration of each session ranged from 30 to 90 min in eleven studies23–28,30,32,33,35,36 and lasted two and a half hours in one study.34 Four studies29,31,37,38 did not report any duration.

Thirteen studies included exercise training of locomotor muscles: cycling in five studies,23,25,26,35,36 walking in three27,37,38 and a combination of these in five.25,28,31,32,34 In three studies the intervention consisted only of calisthenics exercises.29,30,33

All the studies included endurance training. The intensity ranging from 50% to 75% of either the maximum workload or oxygen consumption (VO2) reached during incremental tests in five studies.23–26,36 McNamara et al.32 used the Borg scale39 to set the intensity of the exercise (from 3 to 5 on modified Borg scale) and Hernandez et al.27 used the 70% of the maximum speed attained in the Incremental Shuttle Walking test (ISWT);40 nine others studies28–31,33–35,37,38 reported no training intensity.

In six studies23,24,26,34,35,36 patients performed continuous training on a cycloergometer, while only one25 used the interval training.

In four studies26,31,34,38 breathing exercises were administered in addition to the exercise program (e.g.: pursued lip breathing and diaphragmatic breathing). In eleven studies23–26,28,30,33–36,38 educational sessions were also administered. Inspiratory muscle training was added in two studies23,25

Twelve studies23–34 (463 patients: 271 study group and 192 controls) evaluated perceived fatigue by means of the dedicated domain of the chronic respiratory questionnaire (CRQ).41 Only in three studies35,37,38 (18.8%) was fatigue measurement the primary outcome. Fig. 2-Panel A shows the related forest plot: the intervention improved significantly more than the control group [SMD 0.708; 95% CI 0.510, 0.907; p < 0.001; I² = 34.3%; p = 0.116]. Figure 1SM describes funnel plot for studies on fatigue item in CRQ.

Figure 2.

Forest plot of effect of training on: fatigue item in CRQ scale (Panel A), Fatigue Index Scale (Panel B) and Fatigue Severity Scale (Panel C).

(0.19MB).

Two studies35,36 (68 patients: 41 study group and 27 controls) used the Fatigue Impact Scale (FIS), which examines patients' perceptions of their limitations caused by fatigue on the cognitive, physical, and psychosocial domains.42Fig. 2-Panel B describes the related forest plot: owing to the high heterogeneity in the “fixed analysis model” (P = 0.0001), a random effect model was performed. No differences between intervention and control group were found. [SMD-0.922; 95%CI −2.258, 0.413; p = 0.176; I² = 83.9%; p = 0.013].

Fig. 2-Panel C describes the forest plot for two studies36,38 (82 patients: 49 study group and 33 controls) assessing fatigue by the Fatigue Severity Scale (FSS), a 9-item scale assessing disabling fatigue. Each question evaluates patients’ perception in the form of 7-point modified Likert Scale.43 The intervention group improved significantly more than the control group [SMD -2.282; 95%CI −2.870, −1.699; p < 0.001; I² = 64.6%, p = 0.093]. Only one study37 used the COPD and Asthma Fatigue Scale.44

The assessment of the risk of bias by funnel plots on CRQ revealed high variability in the results of the included studies. Table 3 shows results of the risk of bias assessment. Half of the studies23,25,27,28,31,33,34,38 did not report any detail on randomization and allocation modalities, and there was a high detection bias rate. Almost all studies included showed a high risk of bias for the blinding of participants and personnel.