Risks of Pneumonia in COPD Patients with New-Onset Atrial Fibrillation

The association between Atrial Fibrillation (AF) and pneumonia remains unclear. This study aims to assess the impact of AF on high pneumonia risk group—chronic obstructive pulmonary disease (COPD)—In order to find the association between AF and the risk of pneumonia. The COPD cohort was extracted from National Health Research Institute of Taiwan. The AF cohort comprised all COPD patients with new-onset AF (International Classification of Diseases (ICD)-9 code 427.31) after COPD diagnosis. We further sampled non-AF cohort and performed 1:1 propensity score matched analysis to improve the balance of baseline characteristics between AF and non-AF cohort. The outcomes were pneumonia and pneumonia requiring mechanical ventilation (MV). From 2000–2011, a total of 6228 patients with COPD and AF, and matched 84,106 control subjects were enrolled. After propensity score matching, we identified 6219 patients, each with AF, and matched controls without AF. After propensity score matching, the AF cohorts had higher risk of mortality (adjusted hazard ratio (aHR), 1.24; 95% confidence interval (CI), 1.15–1.34), pneumonia (aHR, 1.17; 95% CI, 1.07–1.27), and pneumonia requiring MV (aHR, 1.33; 95% CI, 1.18–1.50) in comparison with the matched non-AF cohort. After adjusting for mortality from causes other than outcomes of interest as a competing risk, AF remains significantly associated with pneumonia and pneumonia requiring MV. The risks of pneumonia were higher in this population with AF than in those without AF, and the risk was still significant after the adjustment for the competing risk of all-cause mortality.


Introduction
Pneumonia is one of the most common types of infections, and it is associated with high morbidity and mortality. The risk factors of pneumonia include smoking, recent viral respiratory tract infection, elderly patients, difficulty swallowing due to neurologic disease, immunocompromised status, recent trauma or trauma, heart diseases, and chronic lung diseases, such as chronic obstructive pulmonary disease (COPD) [1][2][3]. To face this clinical entity, it is important to identify the population at high risk, such as vaccination for elderly and COPD patients.
Moreover, one recent study found that AF itself is an independent risk factor for hospital-acquired pneumonia [25]. However, the impact of AF on the development of subsequent pneumonia among patients at high risk is unclear. Patients with COPD are prone to respiratory infections; therefore, we hypotheses that AF may be associated with increasing risk of pneumonia in this vulnerable population. To clarify this issue, we conduct this study to assess the impact of new-onset AF on the patients with COPD, and to find out the association between AF and incidence of pneumonia.

Data Source
We used the database constructed by the NHRI (National Health Research Institute) of Taiwan. This database includes outpatient visits, hospital admissions, prescriptions, and disease and vital status data for 99% of the population (23 million people) in Taiwan. The NHRI used original reimbursement data from the National Health Institute (NHI) database to construct a longitudinal database of COPD patients from 1998 to 2010. This cohort included 2,200,000 patients representing 60.5% of all patients with heart or lung disease in the NHI database (n = 3,635,539). The patient records and information were anonymized and de-identified prior to analysis. Therefore, informed consent was not required and was specifically waived by the Institutional Review Board. Ethics approval was obtained from the Institutional Review Board of Cardinal Tien Hospital (IRB No.: CTH-106-3-5-058).

Study Cohort
The COPD cohort was extracted from National health insurance research database (NHIRD). All patients aged between 40 and 100 years, who had experienced a hospital admission or at least three outpatient visits with a COPD diagnostic code within one year from 1 January 2000 to 31 December 2010 were identified. COPD diagnoses were confirmed by the International Classification of Diseases, Ninth Revision (ICD-9) codes 491, 492, or 496. Patients were excluded for the following reasons: (1) incomplete demographic data, (2) had not undergone a lung function test within one year before or after the COPD diagnosis, and (3) had not received a COPD diagnosis after the lung function test. We also excluded those who were dead or diagnosed with AF prior to being indexed. Overall, 90,334 COPD patients were included in this study cohort.
The AF cohort comprised all COPD patients with new-onset AF (ICD-9 code 427.31) after COPD diagnosis. The index date was defined as the date of new-onset AF diagnosis. Index dates for subjects in the control group were randomly assigned dates of medical records. We further sampled non-AF cohort and performed 1:1 propensity score matched analysis to improve the balance of baseline characteristics between AF and non-AF cohort. In the non-AF cohort, patients who had previous history of AF were also excluded. A propensity score analysis was used to reduce potential confounding caused by unbalanced covariates. The propensity score, i.e., the probability of having AF was estimated using a logistic regression model conditional on the covariates of the time from COPD diagnosis to index date, age, gender, index year of AF, monthly income, hospital level, severe exacerbations of COPD in one year prior to index date (never, 1, or ≥2 times/year), medications for COPD, medications for hypertension, other medications and individual comorbidities. Finally, there are 6219 cases with AF and 6219 matched controls in this study ( Figure 1).  Figure 1).

Figure 1.
Study flow chart: a population-based cohort study.

Outcomes
The outcomes were pneumonia (ICD-9-CM codes 480-486, and 507), pneumonia requiring invasive and non-invasive mechanical ventilation (MV) (as presentation for severe pneumonia), and all-cause mortality. Because of the high mortality rate and older-aged in COPD patients, competing risk analysis using the Fine and Gray model was also performed [24]. All subjects were followed until the occurrence of events of interest, death or the end of the study (31 December 2011).

Statistical Analysis
Descriptive statistics were used to characterize the study population at baseline. Continuous variables were presented as mean ± SD; categorical variables were described as counts and percentages. Baseline characteristics were compared between groups using Chi-square tests for categorical variables and independent t-tests for continuous variables. p value < 0.05 was considered to indicate statistical significance. Cox regression models were used to calculate the crude and adjusted hazard ratios (HRs) of different outcomes in the two study cohorts. Adjusted HRs and 95% confidence intervals (CIs) were calculated using Cox regression models adjusted for propensity scores (continuous), cardioversion procedure, and amiodarone use. Amiodarone use was calculated as a time-varying covariate. The non-AF cohort was selected as the reference group. The crude incidence rate of different outcomes was calculated as the total number of events during the follow-up period, divided by person-years at risk. The competing risk analysis and subgroup analysis were performed to further assess the robustness of our study findings. We applied the Fine and Gray competing risk model [27] to derive sub-distribution hazard ratios and 95% CIs in relation to the primary outcomes. We used SAS software version 9.4 (SAS Institute Inc., Cary, NC, USA) for data analysis.

Characteristics of the Study Population
During the study period, a total of 6228 patients with COPD and AF, and matched 84,106 control subjects were enrolled ( Figure 1). Table 1 summarized the demographic characteristics of these groups. AF cohorts were older, longer duration between COPD diagnosis year and index date of AF and more male than control group (all p < 0.05). In addition, significant differences regarding distribution of monthly income, hospital level, the number of COPD with severe acute exacerbation (AE), anti-hypertension medication, most of the commonly used cardiovascular medication and almost all of COPD inhaled and oral drugs except LABA, were noted between AF cohort and control group (all p < 0.05). Additionally, AF cohorts had higher CCI, more myocardial infarction, congestive heart failure, peripheral vascular disease, cerebrovascular diseases, dementia, peptic ulcer disease, renal disease, liver diseases, cancer, and diabetes mellitus than the control group (all p < 0.05). The AF group had higher CHA2DS2-VASc score and CHADS2 score than the control group (both p < 0.05). After propensity score matching, we identified each of the 6219 patients with AF, and matched controls with similar characteristics including age, gender, duration from COPD diagnosis to index date, income, hospital level, all of the medications for COPD, hypertension, cardiovascular diseases and baseline comorbidities. COPD = chronic obstructive pulmonary disease; AF = atrial fibrillation; AE = exacerbation; LABA = long-acting beta agonist; SABA = short-acting beta-agonists; LAMA = long-acting muscarinic antagonist; ICS = inhaled corticolsteroid; ACEi = angiotensin-converting-enzyme inhibitor; ARB = angiotensin II receptor blocker; NSAID = nonsteroidal anti-inflammatory drug.

Discussion
This national population-based study, compromising two matched cohorts each comprising 6219 COPD patients with or without new onset of AF has several significant findings. There were no other differences in these two cohorts except AF or not. We found that the AF cohorts had a higher risk of mortality, pneumonia, and pneumonia requiring MV, in comparison with the matched cohort without AF. There was a consistent and negative effect of AF in pneumonia, even after adjusting for mortality from causes other than outcomes of interest as a competing risk. The negative impacts of AF were shown in subgroup analysis. None of the subgroups investigated appeared to modify the effect of AF on patients' outcomes.
The positive impact of AF on the outcomes in this study is consistent with previous studies [19,21]. In the analysis of Northern California Kaiser Permanente Medical Care Program [19], patients with AF had a higher risk of hospitalization than patients without AF (relative risk: 1.98, 95% CI: 1.73-2.25). Another study [21] had shown that AF could be an independent risk factor of death (OR: 2.66, 95%: 1. 39-5.09). In this study, we have used a nationwide population-based cross-sectional study that covers 99.0% of Taiwan's population and contains nearly complete follow-up information for the whole study population. In this study, we used propensity score matching to minimize the effects of possible confounding variables. Therefore, our findings should be representative and could be generalized. In summary, all of these should indicate that AF can directly affect the prognosis, and these results have clinical implications. In the clinical condition of increasing burden of high-risk groups in the whole word and AF remaining the common arrhythmia among them, we should devote more effort to identifying the patients with AF. After early diagnosis of AF in high-risk patients, we may give appropriate treatment and control for AF, and further ameliorate the negative effect of AF.
Inflammation may be the mechanism with accumulating evidence that progression of comorbidities is associated with AF. Theoretically, inflammation plays an important role in AF patients [28][29][30]. Currently, an increasing amount of evidence suggests that inflammation may participate in the onset and continuation of AF and AF-associated thrombosis through endothelial dysfunction, production of tissue factor, increase in the activation of platelet and increase fibrinogen expression [30,31]. Various inflammatory markers such as C-reactive protein (CRP), tumor necrosis factor (TNF)-α, interleukin (IL)-2, IL-6, IL-8, and monocyte chemoattractant protein (MCP)-1 have been demonstrated to be associated with AF [31,32]. Besides this, for high-risk patients (such as COPD in this study), persistent and systemic inflammation is considered to play a significant role in its pathogenesis [29][30][31][32][33]. Elevated levels of CRP, IL-6, IL-8, and TNF-α have been reported in patients with COPD [34][35][36]. Currently, there is some discussion about the association between inflammation and the respiratory tract microbiome [37,38]. Some byproducts of inflammation may serve as growth factors for bacteria. This may contribute to the incidence of pneumonia [38].
Besides, several other factors may help to explain the effect of AF on COPD patients. Clinically, agents used to treat COPD, including beta-adrenergic agonists and theophylline can result in tachyarrhythmia [39]. In contrast, medication used for controlling AF, such as sotalol, propafenone and non-selective-β-blockers may cause bronchospasm [39]. In addition, the symptoms of COPD patients could be worse due to AF associated with irregular heart beat and reduced diastolic filling of the ventricle [20]. Thus, it is difficult to control COPD when the patients have COPD and AF at the same time. Therefore, it is possible that AF and COPD share the common pathway of inflammation, and interact between themselves through a similar mechanism.
CHA2DS2-VASc [40], CHADS2 [41] and CCI scores [42] were applied in this study. In all three scoring systems, patients with higher scores are at higher risk of pneumonia and pneumonia with MV. CHA2DS2-VASc and CHADS2 are used to calculate the risk of stroke in AF patients. CCI score is a score system to evaluate disease severity. In this study, all three systems work well to find high-risk groups. However, CHA2DS-VASc has better ability than the other two scoring systems. Maybe CHA2DS2-VASc score (Congestive heart failure (1 point), Hypertension (1 point), Age ≥75 (2 points), Diabetes (1 point), Stroke or TIA (2 points), Vascular disease (1 point), Age 65-74 (1 point), Female sex (1 point)) cover more risk factors and age ranges.
Nevertheless, several inherent limitations must be considered. First, as with all claims databases, the data describing lifestyle factors such as body mass index and smoking are not available. Second, we defined AF by ICD-9 codes from administrative data reported by physicians. Although the diagnostic accuracy of AF has been validated in previous studies [43,44], this issue remains a concern. Third, the data of pulmonary function test and the clinical symptoms and signs were not available from NHIRD database. However, here we offer two matched groups without differences except AF or not. We also have tried to adjust the commonly used medications for COPD including LABA, LAMA, SABA, and ICS, and the frequency of AE between COPD patients with and without AF. Therefore, we almost match the severity of COPD in these two groups.

Conclusions
The risk of pneumonia was higher in patients with AF than in those without AF, and the risk was still significant after the adjustment for the competing risk of all-cause death. Thus, the net clinical benefit of pneumonia prevention for high-risk patients with AF needs to be emphasized, especially considering the impact of the high mortality burden. It is noteworthy that our results stressed the need for paying attention to new-onset AF in high-risk patients.