Risk factors for bronchiectasis in patients with chronic obstructive pulmonary disease: a systematic review and meta-analysis

The risk factors of bronchiectasis in patients with chronic obstructive pulmonary disease have not yet been established. This systematic review and meta-analysis aimed to investigate and identify potential risk factors for patients with chronic obstructive pulmonary disease accompanied by bronchiectasis. We reviewed eight electronic journal databases from their inception to November 2019 for observational studies with no language restrictions. The Newcastle-Ottawa Scale was applied to evaluate the quality of the literature. Binary variables were pooled using odds ratios and continuous variables using the standardized mean difference with 95% confidence intervals. The confidence of evidence was assessed according to the grading of the recommendations assessment, development, and evaluation method. Eight case-control studies met the inclusion criteria. Tuberculosis history, smoking history, hospitalization stays, admissions in the past year, and duration of symptoms were considered risk factors. In addition, the ratio between the forced expiratory volume in 1s and forced vital capacity, the percentage of forced expiratory volume in 1s, the forced expiratory volume in 1s as a percentage of the predicted value, purulent sputum, purulent mucus sputum, positive sputum culture, Pseudomonas aeruginosa infection, arterial oxygen pressure, daily dyspnea, C-reactive protein, leukocytes, and the percentage of neutrophils were found to be closely related to bronchiectasis. However, these were not considered risk factors. The evidence of all outcomes was judged as “low” or “very low.” Additional prospective studies are required to elucidate the underlying risk factors and identify effective preventive interventions.


' INTRODUCTION
Chronic obstructive pulmonary disease (COPD) is one of the main causes of global morbidity and mortality (1) and is characterized by partially reversible, persistent airflow limitation associated with chronic airway inflammation and emphysema (2). COPD is a complex heterogeneous disease (3). The clinical presentation and structural abnormalities of the lung can vary greatly between patients (3). With the increasing application of computed tomography (CT) in the evaluation of patients with COPD, previously unrecognized bronchiectasis is being identified (4). Ko et al. (33) defined the most accepted diagnostic criteria for bronchiectasis. Bronchiectasis is characterized by the irreversible widening of medium to small-sized airways, inflammation, chronic bacterial infection, and destruction of the bronchial walls (5).
Some studies have pointed out that bronchiectasis and COPD may co-exist as an overlap syndrome (6). Bronchiectasis was first defined as a comorbidity of COPD in the Global Chronic Obstructive Pulmonary Disease Initiative 2014 guidelines (7). This change was retained in the 2015 updated version and emphasized the impact of bronchiectasis on the natural history of COPD (6). Multiple studies have shown that bronchiectasis in patients with COPD is associated with increased bronchial inflammation, frequent colonization of potentially pathogenic microorganisms, and severe airflow obstruction (8). Bronchiectasis tends to adversely affect the clinical status of patients with COPD, lower their exercise capacity and quality of life, seriously influence the state of psychology, and cause a poor prognosis (9). Moreover, some cases may be obliged to adopt more efficient and sustained antibiotic therapy, and inhaled corticosteroids may not be suitable for patients with bacterial colonization or recurrent lower respiratory infections (10).
Therefore, identifying the potential risk factors for bronchiectasis in patients with COPD could lead to earlier detection and diagnosis, better guidance for management, more effective treatments, and improvement of health status. However, the risk factors for bronchiectasis in patients with COPD have not been fully confirmed. Several observational studies have investigated them but with small sample sizes.
In addition, some contradictory results were found in these studies. For example, Arram and Elrakhawy (11) found that age is a potential risk factor, but the studies by Martínez-García et al. (8) and Yu et al. (12) did not support this result. Thus, this systematic review and meta-analysis aimed to summarize the current evidence of observational studies and then investigate and identify potential risk factors for bronchiectasis in patients with COPD.

Research registration
This study was registered on the International Prospective Register of Systematic Reviews (PROSPERO no. CRD 42020171581) and was carried out according to the Metaanalysis Reporting Guide for Observational Research (13).

Search strategy
We conducted a comprehensive retrieval of eight electronic journal databases, including PubMed, Cochrane Library, Embase, Web of Science, China National Knowledge Infrastructure, Chinese Biomedical Literature Database, WanFang Database, and Chinese Scientific Journal Database. We reviewed these databases from their inception to November 2019 for observational studies with no restrictions placed on the language of publications. In addition, the bibliographies of identified articles and grey literature were also searched to avoid any omissions. The search strategy of the PubMed database is shown in Table 1, and we adjusted it according to the characteristics of others.

Eligibility criteria
The inclusion criteria were as follows: 1) eligible observational studies were identified if the risk factors for bronchiectasis in COPD were demonstrated; 2) diagnosis of COPD complies with any version of reliable and accepted guidelines with clear diagnostic criteria and bronchiectasis diagnosed by objective imaging methods such as highresolution CT (HRCT), CT scan, or chest X-ray (14); 3) studies with all study participants older than 18 years; 4) studies comparing patients with COPD and bronchiectasis in the research group to patients with COPD without bronchiectasis in the control group to identify risk factors; and 5) studies with complete experimental data and results.

Literature screening
All retrieved studies were imported into the Note Express 3.2.0.7350 software (Beijing Aegean Music Technology Co., Ltd.) to delete any duplicates. Two researchers (Zhang XX, Zhang HY) independently screened the titles and abstracts against the established inclusion and exclusion criteria and then downloaded the remaining studies for further screening by reading the full text. If any disagreements occurred, a consensus was reached through discussion or adjudication by a third senior researcher (Pang LJ).

Data extraction
The key characteristics of the included articles were extracted independently by two reviewers (Zhang XX, Zhang HY) using a predefined form. The following data items were collected from each study: the first author, publication year, primary locality of the study, sample size (research group/control group), outcomes, range of age (research group/control group), sex distribution (male/ female), diagnostic criteria, and funding. If any important information elements were missing, we attempted to contact the authors for the desired data. If any disagreements occurred during this process, the two reviewers reached a consensus through consultation or adjudication by a third senior investigator (Pang LJ).

Quality assessment
Two researchers (Zhang XX, Zhang HY) independently and separately applied the Newcastle-Ottawa Scale (NOS) (15) to evaluate the quality of the included literature, which contains three aspects: selection, comparability, and exposure/outcome. Those studies with a score of 5 or more were classified as high quality, while those with a score lower than 5 were classified as low quality (16). To ensure the reliability of the results, low quality literature were not be included in the meta-analysis. Any disagreement during this period was discussed with a third senior researcher (Lv XD). The AMSTAR 2 checklist was used to evaluate the methodological quality of this meta-analysis by two researchers independently (Zhang XX, Zhang HY). This checklist includes 16 criteria. The methodological quality score ranged from 0 to 16. Scores of 15-16, 12-14, 9-11, 6-8, and 3-5 items were evaluated as excellent, very good, good, acceptable, and deficient, respectively (17). Disagreements were resolved by consensus with a third investigator (Lv XD).

Statistical analysis
The Stata13.1 software (Stata-Corp LP, College Station TX77845) was used for the meta-analysis. The Q-test and I 2 values were applied to measure the inter-study heterogeneity. When the p-value of Q-test40.1 and I 2 o50%, a fixedeffects model was applied; otherwise, a random-effects model was used. Binary variables were expressed using the odds ratio with 95% confidence interval (CI) and continuous variables by the standardized mean difference with 95% CI. Forest plots were created using GraphPad Prism version 7.00 software. A subgroup analysis was used to explore the potential confounding factors for significant heterogeneity, such as age, country, literature quality, and publication year. A sensitivity analysis was carried out by removing individual studies to measure the robustness of the results. Egger and Peters tests (18) were performed to provide quantitative evidence of any publication bias (n410).
The grading of recommendations assessment, development, and evaluation (GRADE) algorithm (19) was used to assign quality levels to the meta-analysis evidence. The overall confidence could be judged as ''high,'' ''moderate,'' ''low,'' or ''very low.'' ' RESULTS

Literature selection
A total of 1034 studies were initially identified. Of these, 196 were excluded as they were duplicate studies, and 166 were excluded following a review of the title or abstract. A total of 672 studies remained for full text review. Of these, 664 were excluded as they did not meet the eligibility criteria. Finally, the eight remaining articles (7,8,11,12,(20)(21)(22)(23) were included in this meta-analysis, including four in Chinese and four in English. All of these were case-control studies. A flowchart of the literature screening and selection process is shown in Figure 1.

Characteristics of the studies and quality assessment
Two reviewers independently summarized the characteristics of the included studies according to the data extraction process. A total of 1669 patients were involved, which included 692 in the research group and 977 in the control group. The primary localities of the studies were distributed in three countries, six provinces, and municipalities. The median NOS score of the included studies was 6, with a range from 5 to 7, indicating that these studies were of high quality. The key characteristics of the included studies are presented in Table 2. As evaluated by the AMSTAR2 tool, this meta-analysis scored ''very good.'' Only questions 7 and 9 were evaluated as ''No,'' and the rest were evaluated as ''Yes.''

Data analysis
A meta-analysis was applied to the indicators of the eight included studies. The results show that the indicators were statistically significant between the research group and control group (po0.05), including tuberculosis history, smoking history, the ratio between forced expiratory volume in 1s and forced vital capacity (FEV 1 /FVC), the percentage of FEV 1 (FEV 1 %), the FEV 1 as a percentage of the predicted value (FEV 1 %pred), purulent sputum, purulent mucus sputum, positive sputum culture, Pseudomonas aeruginosa infection, arterial oxygen pressure (PaO 2 ), hospital stay, admission within the past year, duration of symptoms, daily dyspnea, C-reactive protein (CRP), leukocytes (WBC), and the percentage of neutrophils (N%). The results of the heterogeneity test, model, effect size, 95% CI, and p-values are shown in Table 3. The forest plots of the two types of variable indexes are described in Figures 2 and 3.
Reversed results of certain factors existed according to the sensitivity analysis. The lower heterogeneity and stable results emerged after excluding data on arterial carbon dioxide partial pressure (PaCO 2 ) and CRP. The specific results are listed in Table 3. The results of the remaining factors were unchanged after the sensitivity analysis, suggesting that the results should be more stable.
A subgroup analysis was used to explore the sources of heterogeneity for the indicators. For the factor of age, a subgroup analysis was conducted with two groups according to the country (Asian/non-Asian). There was no change in the Asian group but statistical significance in the non-Asian group. The factor of purulent sputum was analyzed in the subgroup analysis according to the country (Asian/ non-Asian). The results showed no change in the non-Asian group. In contrast, there was no statistical significance and lower heterogeneity in the Asian group. Therefore, the country where the study was conducted may be a confounding factor and source of heterogeneity, and more research will be needed in the future.

Sensitivity analysis and GRADE evaluation
The robustness of the results in the sensitivity analysis was good, except for smoking index, body mass index (BMI), mucous sputum, purulent sputum, PaCO 2 , PaO 2 , CRP, erythrocyte sedimentation rate (ESR), hemoglobin (Hb), plasma fibrinogen (FIB), WBC, and N%. The sensitivity analysis indicated heterogeneity in the strengths of the association due to the most common biases in observational studies. The GRADE evidence of all outcomes was judged as ''low'' or ''very low.'' The results are shown in Tables 4 and 5.

' DISCUSSION
The prevention of bronchiectasis is important in the treatment of patients with COPD. However, until now, the risk factors of bronchiectasis have not been confirmed. This study demonstrated a clear relationship between patients with COPD and bronchiectasis and certain risk factors, helping us to better understand the disease. Several casecontrol studies included in this article suggested some risk factors for bronchiectasis in patients with COPD (Table 3). The results showed that the risk factors for bronchiectasis in COPD might include tuberculosis history, smoking history, hospitalization stay, admission within the past year, and duration of symptoms. In addition, FEV 1 /FVC, FEV 1 %, FEV 1 %pred, purulent sputum, purulent mucus sputum, positive sputum culture, Pseudomonas aeruginosa infection, PaO 2 , daily dyspnea, CRP, WBC, and N% were clinical symptoms of bronchiectasis. They were closely related to bronchiectasis in COPD but were not regarded as risk factors. The lung lumens and parenchyma of patients with COPD with a history of tuberculosis were destroyed, which could lead to prolonged airway inflammation duration and acceleration of lung injury and severe airflow obstruction, thus increasing the incidence of bronchiectasis (22). Therefore, patients with a history of tuberculosis should also undergo regular follow-up, although the disease has been cured. Smoking tended to affect lung function. Therefore, it is necessary for patients with COPD to quit smoking. The lung function of patients with COPD was directly impaired due to irreversible airflow limitation. The lung function indicators progressively decreased, which negatively correlated with the number of damaged lung lobes (25). This tends to induce COPD deterioration, and the second most prevalent cause of bronchiectasis was COPD (26).
Consequently, patients with COPD need to monitor lung function indicators regularly to avoid further deterioration. COPD usually has recurrent attacks and is difficult to cure. If patients cough up purulent sputum, this can lead to a considerably greater magnitude of airway dysbiosis (27). However, purulent sputum is not regarded as a risk factor for bronchiectasis in COPD. Bacterial colonization of the airway was the main inducer of airway inflammation in bronchiectasis (24). Positive sputum culture in patients with COPD demonstrated an imbalance of autoimmune function, which increased the host's predisposition to diseases. The most common pathogenic microorganism,   N%, percentage of neutrophils; OR, odds ratio; PaCO 2 , arterial carbon dioxide partial pressure; PaO 2 , arterial oxygen pressure; SMD, standardized mean difference; WBC, leukocytes such as Pseudomonas aeruginosa (28), causes chronic inflammation and lung injury aggravations and increases the incidence of bronchiectasis. However, positive sputum culture and Pseudomonas aeruginosa infection were not considered risk factors for bronchiectasis in patients with COPD because these symptoms were present in bronchiectasis. The overall result of PaO 2 described in the literature was significant, but the results were reversed after removing the studies by Pan et al. (22) or Qin (20), indicating that the robustness of the results was poor. The inconsistent results of       than the optimal information size, p. The 95%CI of the pooled estimate included one or no effect; q. The total sample size was less than 400. The mean N% in the trial group was 0.91 standard deviations higher (0.09 lower to 1.72 higher) -1 9 3 2 ( 3 ) "JJJ the two studies above were likely because different blood collection times and instrument models were used in the blood gas analysis. In summary, more studies are needed to identify the relationship between PaO 2 and bronchiectasis in patients with COPD. The chance of contact with medical staff and patients in the same hospital increased after longer hospitalization, resulting in a greater risk of nosocomial infection. The hospitalized patients were more concentrated (29); therefore, the length of hospital stay directly affects the possibility of bronchiectasis in COPD. If patients with COPD were admitted to hospital within the past year, they might have had poorer disease control and acute exacerbation. The acute exacerbation of COPD resulted in repeated injuries to the lung tissue, leading to more severe airflow obstruction, which was susceptible to bronchiectasis (30). Thus, the disease should be strictly controlled according to the medical advice given to avoid admission for acute exacerbations to reduce the possibility of bronchiectasis. A longer duration of symptoms in COPD is a critical indicator of disease deterioration. Long-term clinical symptoms relieved the patient's resistance. The incidence of bronchiectasis was found to increase due to bronchial infection and the secretions blocking the airway (21). In conclusion, a longer duration of symptoms and hospital admissions within the past year were risk factors for bronchiectasis in COPD. The results of indicators such as purulent sputum, CRP, WBC, and N% were significant. Nonetheless, the results were reversed after removing some studies, indicating that the robustness of the results was weak. Some biases may exist in different clinical analysis instruments, and more rigorous studies are needed to identify these indexes. In addition to the above indicators, the results including smoking index, BMI, mucous sputum, ESR, Hb, and FIB were not significant. However, the results were all reversed in the sensitivity analysis, and more clinical studies are required for supplementary verification. The smoking index may have something in common with smoking history, which tends to aggravate airway inflammation in COPD and increase the incidence of bronchiectasis. A study (31) has shown that low BMI is accompanied by a decrease in muscle mass, which may lead to depression in the strength of the respiratory muscles. Hu X et al. (32) proposed that COPD and bronchiectasis should have a high commonality in clinical symptoms, pathophysiology, and other aspects. The social burden and psychological pressure of patients were increased with the severe airway limitations related to bronchiectasis in COPD. Therefore, we should be familiar with the risk factors for bronchiectasis in patients with COPD. This will help ensure the early prevention, detection, and treatment of bronchiectasis in patients with COPD. Thus, to reduce their risk of bronchiectasis, patients with COPD should quit smoking and drinking alcohol, maintain a balanced diet, and prevent infection. We should devote equal attention to each complementary risk factor. The articles were strictly selected according to the inclusion and exclusion criteria. This study set the sources from which the authors received the diagnostic criteria for COPD and bronchiectasis. However, the final result might be affected by The mean smoking index in the control group was 0 The mean smoking index in the trial group was 0.72 standard deviations higher (0.07 lower to 1.50 higher) -684 (5) "JJJ

BMI
The mean BMI in the control group was 0 The mean BMI in the trial group was 0.13 standard deviations lower (0.37 lower to 0.11 higher) -1022 (3) "JJJ

Age
The mean age in the control group was 0 The mean age in the trial group was 0.15 standard deviations higher (0.10 lower to 0.39 higher) The mean FIB in the trial group was 0.78 standard deviations higher (0.01 lower to 1.58 higher) -1155 (4) "JJJ

Albumin
The mean albumin in the control group was 0 The mean albumin in the trial group was 0.05 standard deviations lower (0.27 lower to 0.18 higher) -323 (2) "JJJ BMI, body mass index; CI, confidence interval; CRP, C-reactive protein; ESR, erythrocyte sedimentation rate; FEV 1 /FVC, the ratio between forced expiratory volume in 1s and forced vital capacity; FEV 1 %, the percentage of FEV 1 ;. FEV 1 %pred, FEV 1 as a percentage of the predicted value; FIB, plasma fibrinogen; GRADE, grading of recommendations assessment, development, and evaluation; Hb, hemoglobin; N%, percentage of neutrophils; PaCO 2 , arterial carbon dioxide partial pressure; PaO 2 , arterial oxygen pressure; RR, risk ratio; WBC, leukocytes.
the interference of some factors, and there are several limitations in this meta-analysis, such as uncertainty bias in the secondary data, a limited number of articles, small total sample size, and unpredictable differences between sample sizes.
In conclusion, the presented results can be valuable to the medical community. The strengths of this review and metaanalysis include the inclusion of articles that assessed the quality of evidence evaluation using the GRADE approach. However, more studies with larger sample sizes are required. Furthermore, a multi-center case-control study is required to identify the risk factors scientifically and comprehensively for bronchiectasis in COPD. This study can be beneficial in guiding clinicians to formulate targeted prevention and treatment measures. This paper can provide recommendations for improving survival and quality of life and reducing the psychological, family, social, and medical burdens of patients with COPD and clinical guidance for reducing the incidence of bronchiectasis in patients with COPD.