Impact of self-reported Gastroesophageal reflux disease in subjects from COPDGene cohort

The COPDGene Study (http://​www.​COPDGene.​org/​) is an ongoing NHLBI-funded multicenter study of the genetic epidemiology of smoking-related lung disease (Clinical Trials Registration # NCT00608764). A complete description of the protocol has already been published [8]. Briefly, inclusion criteria other than ability to give informed consent are: age 45-80 years; at least 10 pack-years cigarette smoking history; self-defined non-Hispanic white or African-American ancestry; and willingness to undergo study-related tests, including spirometry, CT scan of the chest and blood collection for biomarker and genetic analysis. For the current analysis, we selected from the full cohort of 10,300 enrolled subjects participants with COPD, both former or current smokers, who met criteria for GOLD stage 1 or greater (fixed airflow obstruction with a post-bronchodilator FEV₁/FVC [forced vital capacity] ratio ≤0.7),. Additionally, all subjects had CT measurements of emphysema and airway abnormalities completed at the time of data analysis. Parenchymal analysis was performed using Slicer (http://​www.​Slicer.​org); airway analysis was performed using VIDA Pulmonary Workstation 2 (http://​www.​vidadiagnostics.​com). Lung areas with attenuation value of less than-950 Hounsfield Units (HU) on the inspiratory scans were considered emphysematous. CT-measurements of airway disease included mean bronchial wall thickness calculated as an average of six segmental values for each subject, wall area percent (100* wall area/total bronchial area) and the square root of the wall area of a theoretical airway of 10 mm luminal perimeter (Pi10) [8]. The research protocol was approved by the institutional review board at each participating institution (University of Michigan Health System Research Committee IRB approval HUM000014973, 07/16/2010), and all participants provided written informed consent.

Demographic data, smoking, and medical history were collected using self-administered questionnaires. Symptoms and self-reported acute exacerbation frequency were quantified using a modified version of the ATS Chronic Respiratory Disease Questionnaire (ATS-DLD-78) [9] with the question: “Have you had a flare-up of your chest trouble in the last 12 months?” If the answer was “No”, zero exacerbations were recorded, and when the answer was “Yes”, additional questions on the presence, severity, management and number of exacerbations followed. Dyspnea was assessed using the Modified Medical Research Council scale [10], health-related QOL with the St. George’s Respiratory Questionnaire (SGRQ) [11], Medical Outcomes Study Short Form (SF) 36 version 2, and the BODE index score was calculated for all patients [12]. Comorbidities were assessed based on subject self-report. At enrollment all participants agreed to be contacted on a regular basis and to provide information about development of new exacerbations in the interval since the initial visit, using the same initial questions. Participants were followed longitudinally via either an automated telephone or web-based system on an every 6-month basis. Subjects not reached through the automated system were contacted by a research coordinator for a phone-based interview. Both baseline history of exacerbations as well as exacerbations during longitudinal follow-up were dichotomized on “frequent or infrequent”, based on the currently accepted definition of ≥2 exacerbation per year [13].

Participants underwent spirometry pre- and post- administration of short-acting bronchodilating medication (albuterol), using the EasyOne™ spirometry system (Zurich, Switzerland). Predictive values were obtained using NHANES III data [14]. Quality control was performed for all spirometry tests using both an automated system and manual review. Six minute walk distance was measured in standard fashion [15].

Ascertainment of physician-diagnosed GERD was based on self-report. The patient was presented with the question “Have you ever been told by a physician that you have…” and a list of different diseases, including GERD. Information on medications, including GERD-related prescriptions PPIs, antihistamines, antacids), was also collected.

Clinical characteristics between patients with and without GERD were compared with t-tests or chi-square tests where appropriate. Multivariate linear regression was used to model SGRQ and MMRC scores. Given the skewed distribution of exacerbation frequency, we used a zero-inflated Poisson model. All models were adjusted for age, gender, current smoking history, BMI, FEV₁% predicted and clinical center. In the final models, we also evaluated the influence of PPIs and of the GERD-PPIs interaction. Because GERD could be a marker for the presence of other serious diseases and conditions independently associated with worse COPD outcomes, we also conducted additional analyses to determine whether differences in the GERD population as compared to the entire population contributed to differences in outcomes. To better isolate the impact of GERD, we generated a covariate representing the propensity to be diagnosed with GERD, based on a logistic regression model. In this model, the response variable of GERD diagnosis was generated, using as predictors either the variables with association <0.10 in univariate analysis or those accepted in the literature as associated with GERD (age, gender, BMI, smoking history, presence of cardiovascular disease, osteoporosis and gastric ulcers) [16, 17]. These estimated propensity scores were included as a covariate in fully-adjusted regression models for outcomes of interest. All statistical analyses were conducted using STATA v.12 statistical software (College Station, TX.). A p-value <0.05 was considered significant.

Data were available for 4,483 participants who met GOLD criteria for COPD, of any GOLD stage, and who had answered the question of whether they had been previously diagnosed by a physician as having GERD. Table 1 outlines patient demographics, clinical characteristics, spirometry, and disease impact (SGRQ, BODE, MMRC dyspnea score and exacerbation frequency), stratified by history of GERD. Physician-diagnosed GERD was reported by 29.1% of the subjects, more frequently among women (50.9% vs. 41.3%, p < 0.0001) and older individuals (63.9 vs. 62.7 years of age, p < 0.0001). Statistical differences of no clinical significance were found in severity of airflow obstruction among GERD subjects (56.3% vs. 57.9%, p = 0.02) as well as those with higher BMI (28.8 kg/m² vs. 27.5 kg/m², p <0.0001); the distribution by GOLD stages was also similar. Although those with GERD reported greater tobacco consumption (53.4 vs. 50.0 pack-years, p = 0.003), they were less frequently current smokers (34.3% vs. 47.1%, p < 0.0001). Compared to patients with COPD without GERD, a greater frequency of subjects reporting coexistent GERD also fulfilled the diagnostic criteria for chronic bronchitis (29.0% vs. 24.7%, p = 0.002). Subjects with coexistent GERD were more frequently prescribed all classes of COPD-related medications, including short-acting medications (61.0% vs. 48.5%, p < 0.0001) and controller medications (e.g. combined steroid and long-acting beta-agonist 41.7% vs. 33.6%, p < 0.0001; long-acting anti-muscarinics 40.2% vs. 30.8%, p < 0.0001).

Significant differences were also noted in the impact of GERD on QOL measures and exacerbations (Table 2). Those with GERD had a shorter six minute walk distance (1,192 vs. 1,250 feet, p < 0.0001); higher BODE index (2.7 vs. 2.4, p = 0.001), more severe MMRC dyspnea scores (2.2 vs. 1.8, p < 0.0001); higher total SGRQ score (41.8 points vs. 34.9 points, p < 0.0001); lower SF36 PCS score (38.2 vs. 41.4 points, p < 0.0001), and more frequent exacerbations in the year prior to enrollment (0.9 vs. 0.6, p < 0.0001). Importantly, those with GERD more frequently reported ≥2 exacerbations in the prior year (22.5% vs. 12.9% of those without GERD, p < 0.001).

COPD subjects with GERD also had significantly increased frequency of comorbidities (Table 2). We found significantly increased frequency of myocardial infarction (10.9% vs. 6.4%, p < 0.0001), coronary artery disease (11.9% vs. 7.6%, p < 0.0001), angina (8.6% vs. 4.0%, p < 0.0001), as well as modifiable cardiovascular risk factors (two or more of the following: hypertension, hyperlipidemia, obesity and smoking, 60.5% vs. 50.5%, p < 0.0001), and sleep apnea (22.2% vs. 13.5%, p < 0.0001). By contrast, we found no differences in the reported frequency of diabetes (13.5% vs. 11.7%, p = 0.22).

After identifying the imbalance between GERD and no-GERD patients in the distribution of demographics and comorbid conditions, we generated a propensity score for GERD diagnosis, as described in the Methods section. This propensity score was well-balanced with respect to the distribution of important covariates. Next, in separate multivariate models, each controlling for age, gender, BMI, current smoking and severity of airflow obstruction, we examined the effect of GERD on health-related QOL and exacerbation frequency in COPD. The three models tested the independent effect of GERD (model 1); of GERD plus propensity to be diagnosed with GERD (model 2); and with the addition of use of PPIs plus GERD-PPI use interaction (model 3) (Table 3).

These analyses demonstrate that the presence of GERD in COPD is associated with worse QOL, with an associated 3.4 point (95% CI 1.8, 4.9) higher SGRQ total score, which is within the accepted minimal important difference for this measure of health status [18].

Using data on enrollment, GERD in COPD is also associated with higher frequency of exacerbations in the previous year (1.17 fold increase in frequency; 95% CI 1.05, 1.31; p = 0.005) and more risk of being a “frequent exacerbator” (OR 1.40; 95% CI 1.10, 1.77; p = 0.006). Additional analysis after a mean follow-up of two years showed a sustained association between GERD and exacerbation frequency fulfilling the definition of frequent exacerbation (OR 1.40; 95% CI 1.10, 1.79; p = 0.006).

All models (Table 3, third column) include an analysis of the effect of PPIs. They uniformly demonstrate that PPI use does not change the associations between GERD and exacerbations and are actually associated with poor outcomes. Further analysis of the interaction of GERD by PPI use demonstrated significance only when modeling QOL, but not for exacerbations. In additional sensitivity analyses we included in the propensity score and all multivariate models the imbalance in the use of respiratory medications, and there was not change in the strength or the direction of the associations.

To explore hypothetical mechanistic explanations for GERD-COPD outcome relationships, we also used our data to examine two additional associations. First, we looked for differences in indirect markers of recurrent aspiration or microaspiration, and found all to be more frequent among those with GERD: reported pneumonia (58.1% vs. 43.9%, p < 0.0001); recurrent wheezing (69.8% vs. 55.5%, p < 0.0001); and physician-diagnosed asthma (28.8% vs. 20.0%, p < 0.0001). Second, we searched for radiologic changes that might accompany GERD. This analysis disclosed no differences between COPD patients with versus without GERD in severity of emphysema, or in any of the measures of airway thickness (Table 4).