Pulmonary symptoms and diagnoses are associated with HIV in the MACS and WIHS cohorts

The MACS and WIHS cohorts have been described previously [20, 21]. Briefly, the MACS is a prospectively followed cohort of 6,972 men with or at risk for HIV started in 1984, and the WIHS is a prospectively followed cohort of 3,766 women with or at risk for HIV started in 1994. The MACS is primarily a cohort of men who have sex with men, and WIHS is a cohort of women who engage in high risk behavior for HIV infection, primarily intravenous drug use. Participants in this study were HIV-infected and HIV-uninfected individuals who completed a pulmonary questionnaire at an ongoing MACS or WIHS study visit between January 2008 and December 2010. All participants were eligible. If participants did not complete a baseline pulmonary questionnaire during the 2008 visit cycle, they were excluded from further study. All participants gave written informed consent to study protocols approved by the institutional review boards of participating sites (University of Pittsburgh; Women’s Interagency HIV Study Data Management and Analysis Center; Johns Hopkins University; University of California, Los Angeles; University of California, San Francisco; State University of New York Downstate Medical Center; Northwestern University; and the Center for Analysis and Management of Multicenter AIDS Cohort Study).

At yearly MACS or WIHS visits from 2008 to 2010, participants responded to a modified American Thoracic Society-Division of Lung Disease-78 Respiratory Disease Questionnaire [22] that collected information regarding frequency of respiratory symptoms (cough, phlegm production, wheezing, and dyspnea) and prevalent (2008) and then incident (2009, 2010) therapies (medication, oxygen therapy, nocturnal non-invasive ventilation), diagnostic testing (self-report of having had sputum sampling, bronchoscopy, chest computed tomography [CT], pulmonary function testing, echocardiography, or polysomnography), and chronic diagnoses (self-report of having been told they had asthma, COPD, chronic bronchitis, lung fibrosis, sarcoidosis, pulmonary hypertension, pulmonary emboli, and sleep apnea). Baseline pulmonary questionnaire asked if the participant had ever had diagnostic testing, and follow-up pulmonary questionnaires if they had diagnostic testing in the past year. Participants were considered to have significant respiratory symptoms if they reported that symptoms occurred more frequently than just with colds or infections and were experienced at least monthly. The presence of any COPD was defined as having been told they had a diagnosis of COPD, chronic bronchitis, or emphysema.

Time-dependent data were determined from the visit at which the prevalence pulmonary questionnaire was administered (2008) and included age, HIV status, duration of HIV infection, body mass index (BMI), cigarette smoking status, cumulative pack-years of cigarette smoking, alcohol consumption, use of intravenous drugs in the past 6 months, use of cocaine in the past 6 months, diagnosis of cardiac disease (by self-report of myocardial infarction, hospitalization for congestive heart failure hospitalization for chest pain/angina, surgery on heart vessels, or other heart disease) use of HAART, current CD4 count, and current plasma HIV RNA level. Data from all visits prior to and including baseline visit were used to determine participants’ ever-use of intravenous drugs or cocaine, history of tuberculosis, history of Pneumocystis or bacterial pneumonia, duration of HAART use, and mean and nadir CD4 counts. HIV status was determined by HIV serological testing at each visit in previously seronegative participants, and participants who were HIV positive at baseline visit were assumed to have duration of HIV infection from time of MACS or WIHS enrollment to the date corresponding to the prevalence pulmonary questionnaire visit. Race, ethnicity, cigarette smoking history, alcohol consumption, and illicit drug use were determined by participant self-report. Alcohol consumption was categorized as none, light drinking (<3 drinks/week), moderate drinking (3-13 drinks/week), or heavy drinking (≥14 drinks/week). HAART was defined previously [23] and categorized as never, past HAART use, or current HAART use.

The MACS and WIHS cohorts were analyzed separately because of variations in data collection procedures and differences in the populations sampled, including gender, proportions of minorities, socioeconomic status, access to health care, and substance use. To assess selection bias, characteristics (age, race/ethnicity, smoking history, alcohol use, drug use, HIV status, and CD4 counts and viral loads) were compared between participants in each cohort who completed the prevalence pulmonary questionnaire to those who were enrolled in the cohort and did not complete the questionnaire. For participants who had not completed the prevalence pulmonary questionnaire because they failed to show for the scheduled visit, data were carried forward from the last visit they had attended.

Personal characteristics, clinical data, presence of respiratory symptoms, medical therapies, diagnostic testing, and diagnoses were compared between HIV-infected and HIV-uninfected groups in each cohort using parametric (t-tests) and non-parametric (chi-square, Wilcoxon) tests as appropriate. Respiratory symptoms, medical therapies, diagnostic testing, and diagnoses were also compared between HIV-infected and HIV-uninfected groups in each cohort stratified by smoking status because cigarette smoking is such a dominant risk factor for lung disease. To determine incidence of pulmonary diagnoses from follow-up questionnaires (2009-2010), incidence rates for specific pulmonary diseases were computed in those who did not report that disease at the prevalence pulmonary questionnaire (2008) as the number of observed incident events divided by the number of person-years of follow-up, where the follow-up time available for each person was the number of years from the initial pulmonary evaluation visit until the earliest of the following: disease diagnosis date, death, loss to follow-up, or the date of the last visit with a pulmonary questionnaire.

Multivariable modeling was then performed for the most prevalent conditions: symptoms, diagnostic testing, asthma, COPD, and sleep apnea. Poisson regression methods [24] were used to examine the association of baseline measures and HIV status with baseline prevalence of respiratory symptoms, diagnostic testing, and diagnoses, and to identify potential confounders. Baseline characteristics examined included age, African-American race, Hispanic ethnicity, smoking status, pack-years smoked, alcohol status, intravenous drug use ever, cocaine use ever, and hepatitis C status. Adjusted prevalence ratio models were fitted with robust variance to quantify the association of HIV status with the outcomes independent of the effect of race, smoking status, intravenous drug use ever, age, and any additional significant factors (p < 0.05) added from the above multivariable model. Similar models were repeated restricting to HIV-infected participants to assess HAART use, CD4 cell counts, and plasma HIV RNA levels, and an additional set of models restricted to HIV-infected participants were performed to include history of cardiovascular disease and history of Pneumocystis or bacterial pneumonia to assess for contributions of these variables. We also assessed if protease inhibitors were related to symptoms or diagnoses by including protease inhibitor use in the model instead of HAART use, but there were no associations with this specific antiretroviral category (data not shown). To evaluate the associations between incident diagnoses during the two-year follow-up and HIV status, we fit separate multivariable Poisson regression models to incidence data [25, 26]. For covariates with similar origin (for example current or ever using intravenous drugs or the current or mean CD4 count), the variable with the lowest bivariate p-value was included in the model. Prevalence ratios or rate ratios and 95% confidence intervals were reported for all outcomes. An interaction between HIV infection and smoking status was also test by adding an interaction term to the final multivariable models. A p-value <0.05 determined statistical significance. We analyzed data using SAS Version 9.2 (SAS, Inc., Cary, NC).

The prevalence pulmonary questionnaire was completed at a study visit in 2008 by 907 HIV-infected and 989 HIV-uninfected participants in the MACS cohort and 1405 HIV-infected and 571 HIV-uninfected participants in the WIHS cohort. There were 223 MACS participants and 488 WIHS participants who did not complete the prevalence pulmonary questionnaire (Additional file 1: Table S1). In both cohorts, equal proportions of HIV-infected participants completed the questionnaire. Participants who completed the questionnaire were more likely to be older, Caucasian, use recreational drugs, and have lower plasma HIV RNA levels. In WIHS, there were fewer never-smokers and more never drinkers of alcohol in participants who completed the questionnaire, and Hispanics were more likely to complete the questionnaire.

There were significant differences in demographic and clinical factors between the HIV-infected and HIV-uninfected participants in each cohort at the time of the prevalent questionnaire (Table 1). In the MACS cohort, HIV-infected individuals were younger, more likely to be African-American, smoked more, and used cocaine more than HIV-uninfected individuals; while in the WIHS cohort, HIV-infected individuals were older, less likely to be African-American, and less likely to smoke than HIV-uninfected participants. HIV-infected participants in both cohorts had lower BMI and were less likely to drink, but were more likely to use intravenous recreational drugs, be hepatitis C seropositive, and have had pneumonia compared to HIV-uninfected individuals.

The majority of HIV-infected participants in both cohorts were on HAART (78.4% and 76.2% in MACS and WIHS, respectively) at the time of the prevalence pulmonary questionnaire with a mean (SD) duration of HAART use of 14.1 (7.9) years in MACS and 15.5 (8.3) years in WIHS. HIV-infected participants had a mean (SD) CD4+ T-cell count at the time of the prevalence pulmonary questionnaire of 571.8 (277.4) cells/uL in MACS and 502 (291.2) cells/uL in WIHS, and the majority of participants had undetectable viral loads (MACS: plasma HIV RNA level < = 40copies/mL: 590/843 = 70%; WIHS: plasma HIV RNA level < =80 copies/mL: 764/1365 = 56%).

Respiratory symptoms and inhaler use were common in both cohorts at the time of the prevalence pulmonary questionnaire (Additional file 1: Table S2), and HIV was an independent risk factor for certain symptoms. In the MACS cohort, phlegm production, dyspnea, any respiratory symptoms, and inhaler use were reported more commonly by HIV-infected participants (Figure 1a), and in the WIHS cohort, only oxygen use was reported more commonly in HIV-infected participants, although oxygen use was uncommon overall (Figure 1b). In regression models, HIV infection was independently associated with greater prevalence of cough (prevalence ratio [PR], 1.11; p = 0.03), dyspnea (PR, 1.25; p = 0.03), wheezing (PR 1.20; p = 0.04), and having any respiratory symptom (PR, 1.11; p = 0.003) in the MACS cohort and associated with wheezing (PR, 1.19; p = 0.02), inhaler use (PR 1.14; p = 0.04), and oxygen use (PR 3.43; p = 0.01) in the WIHS cohort (Table 2).

Patterns of the use of respiratory diagnostic testing varied significantly according to HIV status. Diagnostic tests commonly used to evaluate acute respiratory infections such as sputum testing, bronchoscopy, and chest CT scans were more likely in HIV-infected participants (Figure 1c-d). Despite a greater prevalence of chronic respiratory symptoms in HIV-infected participants, HIV-infected individuals in either cohort were no more likely to have mainstay testing for chronic respiratory disease such as pulmonary function testing (Table 2). HIV-infected participants in MACS were more likely to have an echocardiogram and polysomnography.

Obstructive lung diseases were the most commonly reported prevalent diagnosis (Figure 1e,f). In both cohorts, asthma was the most common diagnosis. Asthma prevalence was strikingly high in the WIHS cohort in both HIV-infected (22.9%) and HIV-uninfected (25.7%) participants, but not significantly different by HIV status. COPD diagnoses were also common and mostly attributable to chronic bronchitis, but were no more likely in HIV-infected compared to HIV-uninfected in either cohort.

Sleep apnea appeared to occur at similar prevalence rates in HIV-infected and HIV-uninfected men as well as women, however, when adjusting for confounders such as age and BMI, HIV-infected participants were more likely to have been diagnosed with sleep apnea in both MACS, (PR, 1.42; p = 0.01) and WIHS (PR, 2.10; p = 0.002) (Table 2).

The prevalences were low and did not differ between HIV-infected and HIV-uninfected participants for the diagnoses of pulmonary hypertension (MACS: HIV-infected, 0.6% vs. HIV-uninfected, 0.4%; WIHS: 0.7% vs. 1.0%), interstitial pulmonary fibrosis (MACS: 0% vs. 0.2%; WIHS: 0% vs. 0.2%), sarcoidosis (MACS: 0.4% vs. 0.2%; WIHS: 0.2% vs. 0.9%), and pulmonary embolism (MACS: 0.5% vs. 1.0%; WIHS: 0.4% vs. 0.7%).

When analysis was performed comparing prevalence of symptoms, diagnostic testing, and diagnoses between HIV-infected and HIV-uninfected subjects stratified by smoking, there appeared to be increased prevalence of outcomes with former and current smoking compared to never smokers (Additional file 1: Table S4a and b). Compared to HIV-uninfected never smokers, HIV-infected men who were never smokers were more likely to have dyspnea and have had bronchoscopy and sputum sampling performed. Compared to HIV-uninfected former smokers, HIV-infected men who were former smokers were more likely to have phlegm production, use inhalers, and have had chest CT, bronchoscopy, and sputum sampling performed, and compared to HIV-uninfected current smokers, HIV-infected men who were current smokers were more likely to have had chest CT, echocardiogram, or sputum sampling performed. HIV-infected women were more likely to have sputum sampling done independent of smoking status, but compared to HIV-uninfected former smokers who were women, female HIV-infected former smokers were more likely to have had chest CT. Compared to HIV-uninfected current smoker women, HIV-infected current smokers were more likely to have wheezing, use an inhaler, and be diagnosed with COPD and chronic bronchitis. An interaction between smoking status and HIV infection was not found to be significant for any respiratory symptoms or diagnoses.

Incidence of diagnoses was determined during 2 years following the prevalence questionnaire. Men had a higher incidence rate of sleep apnea compared to women, while women had higher incidence rates for asthma and COPD compared to men (Table 3). HIV-infected individuals were more likely to report a new diagnosis of COPD in the MACS cohort (rate ratio [RR], 2.21; p = 0.03), and there was a borderline association with HIV and increased incident sleep apnea (RR 2.33; p = 0.08) in the WIHS cohort.

HIV-related variables such as nadir CD4 cell count, HIV viral load, and HAART use were also associated with respiratory symptoms, diagnostic testing, and diagnoses (Additional file 1: Table S3a). In HIV-infected participants in both the MACS and WIHS cohorts, diagnostic testing such as sputum and bronchoscopy were more likely in those with lower nadir CD4 cell counts (Additional file 1: Table S3a). In WIHS HIV-infected participants, COPD was more likely with higher plasma HIV RNA levels.

We also included cardiovascular disease and history of bacterial or Pneumocystis pneumonia in models (Additional file 1: Table S3b). In MACS, there were associations of cardiovascular disease and history of pneumonia with various respiratory symptoms (cardiovascular disease, bacterial pneumonia), increased likelihood of bronchoscopy (cardiovascular disease, bacterial pneumonia) or sputum induction (bacterial pneumonia, Pneumocystis pneumonia), and with diagnoses of asthma (bacterial pneumonia) and COPD (bacterial pneumonia). Lower nadir CD4 counts were no longer associated with increased chance of sputum testing and bronchoscopy. In WIHS, dyspnea was significantly increased in those with cardiovascular disease or a history of bacterial pneumonia with a trend to being more common in those on HAART (Additional file 1: Table S3b). Wheezing was also more common in those with a history of bacterial pneumonia or in those using HAART. Individuals with a history of cardiovascular disease were more likely to have had an echocardiogram and those with bacterial pneumonia or Pneumocystis pneumonia were more likely to have had a bronchoscopy of sputum induction. A history of pneumonia was more common in individuals with asthma (bacterial pneumonia) or COPD (bacterial or Pneumocystis pneumonia). Higher HIV RNA level was no longer significantly associated with a COPD diagnosis.