Abstract
Objective: We aim to make use of clinical spirometry data in order to identify individual COPD-patients with divergent trajectories of lung function over time. Study Design and Setting: Hospital-based COPD cohort (N = 607) was followed on average 4.6 years. Each patient had a mean of 8.4 spirometries available. We used a Hierarchical Bayesian Model (HBM) to identify the individuals presenting constant trends in lung function. Results: At a probability level of 95%, one third of the patients (180/607) presented rapidly declining FEV1 (mean -78 ml/year, 95% CI -73 to -83 ml) compared to that in the rest of the patients (mean -26 ml/year, 95% CI -23 to -29 ml, p ≤ 2.2 × 10-16). Constant improvement of FEV1 was very rare. The rapid decliners more frequently suffered from exacerbations measured by various outcome markers. Conclusion: Clinical data of unique patients can be utilized to identify diverging trajectories of FEV1 with a high probability. Frequent exacerbations were more prevalent in FEV1-decliners than in the rest of the patients. The result confirmed previously reported association between FEV1 decline and exacerbation rate and further suggested that in clinical practice HBM could improve the identification of high-risk individuals at early stages of the disease.
Introduction
Lately several clinical subphenotypes of Chronic Obstructive Pulmonary Disease (COPD) have been recognized, providing differences in the natural course of the disease as well as response to treatment have been recognized (Citation1). Even though accelerated decline of lung function has been a recognised feature (Citation2) in COPD for decades, the patient-to-patient variation in decline rates has remained poorly understood until recently. The proportion of patients who rapidly lose lung function and develop progressive airway obstruction has been estimated to be between 20–25% of all COPD patients. Rapid annual decline of FEV1 (> 40 ml) has been associated with poor prognosis and thus, these patients should be identified at early stages of the disease (Citation3–Citation6). Another clinically important phenotype is a frequent exacerbator. A stepwise decline in the FEV1 trend is often associated with exacerbations of the disease suggesting that these phenotypes might be at least partially overlapping (Citation4,Citation7).
Longitudinal information on patients’ spirometry values is needed when differences in individual decline rates are studied. Sequential spirometry measurements however include intra-patient variation due to both natural oscillation of disease process due to e.g., pharmacotherapy or disease exacerbations and measurement error related to spirometry performance. This variation needs to be analyzed before a patient can be labeled as a “decliner.” FEV1 decline rate can be considered as a marker of disease activity, which can be either a continuous process or manifest in stages. It can also vary according to the severity stage of the disease. Understanding of the disease activity profile may be of importance for our success in medical interventions.
To our knowledge, there's only one longitudinal study, which has addressed the individual decline of lung function (Citation5), as previous studies remain to infer based on development of group of patients as in the seminal study by Fletcher and Peto (Citation2). Tobacco smoke being the primary cause of COPD, the findings that pack-years are only loosely associated with the decline rate suggest that some of the COPD patients are extremely sensitive to cigarette smoke due to genetic or environmental triggers (Citation3,Citation5). Given the progressive nature of COPD, another unexpected result has been the abundant number of patients whose FEV1 remained stable or even improved, while 15% of the ECLIPSE patients (Citation4) showed an improving FEV1 trend.
The recognition of individual trends with unfavorable lung function development has proven challenging when traditional regression methods are used. For diagnostic purposes we need to develop approaches enabling us to estimate the uncertainties in individual slopes and identify the rapid decliners with high probability. In the present study we aimed to make use of the Hierarchical Bayesian model in recognition of COPD patients with rapidly declining trajectories in FEV1. In addition we studied whether the disease exacerbations were associated with rapid FEV1 decline and whether smoking cessation had a positive effect on the decline rate.
Material and Methods
Study subjects
The patients with smoking-related COPD or chronic bronchitis were recruited from the Turku and Helsinki University Hospitals in Finland in the years 2005–2007 (Citation8,Citation9) by mail. A total of 844 patients contacted the research staff again by mail, volunteered to the study, participated to an interview and donated blood for DNA extraction. All the hospitals and outpatient clinics that had been treating the study subjects were contacted in order to gather each participant's unbroken medical history. The patients have been followed through medical records last ten years prior the enrolment or from the onset of the COPD diagnosis onwards. The data gathering will continue in total for 10 years or up to the death of the patient.
The Coordinating Ethics Committee of the Helsinki and Uusimaa Hospital District (Coordinating Ethics Committee decision 125/E0/04) approved the study approach and the permission to conduct this research was granted by the Helsinki and Turku University Central Hospital.
Assessment of lung function measurements and co-morbidities
The medical history of 739 patients was well documented and the disease history could be confirmed. Baseline clinical characteristics and physiological examinations are all based on retrospective medical records covering at least 5 years prior to the enrolment. Diagnosis of COPD including Asthma COPD Overlap Syndrome (ACOS) was confirmed by a pulmonologist from the electronic medical records covering more than over 5 years of diagnostic/monitoring of the disease. The diagnosis was based on longitudinal symptoms, signs treatment response and lung function testing results. Patients presenting with pure asthma, heart failure or lung parenchymal diseases as a cause of dyspnea were excluded.
Spirometry data were available from year 2000 up to year 2012 (a total of 5291 measurements). The Finnish reference values were used (Citation10). All patients with at least 3 pre-bronchodilatory measurements available were included into the analysis (607 patients with 5088 measurements, yielding 2816 patient years). Forced expiratory volume in 1 s (FEV1) was analysed as markers of airway obstruction.
We collected the patients’ weight, height and smoking status from the medical records. All the co-morbidities stated in the medical records were carefully evaluated, especially time of the onset and the diagnostic criteria of the disease, often made by a specialist of the field. Diffusion capacity was available for about 71% of the patients, but computerized tomography (CT) imaging only for sporadic patients and thus CT imaging was omitted from the analysis.
Assessment of COPD exacerbations
During the years 2006–2011 all the visits at the department of emergency and all hospital admissions due to COPD exacerbations and pneumonias were gathered from the medical records by a pulmonologist. Each hospital admission was defined as a separate event and recorded as a severe exacerbation. Pneumonia was defined by clinical symptoms and signs such as fever, crackles by auscultation, elevated C reactive protein and leukocytosis suitable for respiratory tract infection and a matching infiltrate in the chest x-ray.
Drug claims were collected from the National Drug Imbursement Registry covering the years 2007–2011. Drug claims were available for 514 COPD patients who had taken part in a questionnaire-based study on Physical activity (COPEX-study) (Citation11). Anatomical Therapeutic Chemical (ATC) classification was used to identify the product and Nordic article number (VNR) to recognize the format of the product and the medicine package size. All purchases of antibiotics (amoxicillin, doxycycline, macrolides, and fluoroquinolones) commonly used in respiratory infections and oral steroids (prednisolone, methylprednisolone) were considered as a single event. Courses of oral steroids and antibiotics require prescription and were considered as markers of mild or moderate, home-treated COPD exacerbations. Physical inactivity defined in COPEX study was also tested for association of decline of lung function.
Statistical analysis
The Hierarchical Bayesian model (HBM) is based on the sequential spirometry data on measurements of each individual patient over time. The length of follow-up and the number of measurements varied between patients. Hierarchical Bayesian model is well suited for unbalanced data as in the present study. Uninformative priors for lung function development were used to model the linear change of lung function over time. The trajectory was let to correlate to the spirometry measurement at study baseline. A Markov chain Monte Carlo algorithm was run to estimate the posterior distributions of lung function development for each patient. Linear model was found to fit data well and is robust against over-fitting. Extrapolation or imputation of missing data was not done. Correlations between continuous variables were examined using Pearson correlation coefficient. Differences in mean values between two groups were determined using Student's t-test or Mann-Whitney test when appropriate.
Results
Development of airway obstruction over time
The cohort represents smoking related COPD patients (a total of 607 patients) of which 65% were men, mean age 58 (SD 6.9) years with varying levels of airway obstruction, varying profiles of co-morbidities and exacerbations (Table ). Patients lost to follow-up (N = 132) due to low number of spirometry measurements had significantly better FEV1 (63.9% vs. 55.9% of predicted, p ≤ 6.9 × 10-5) suggesting that patients with more severe disease were not lost to follow-up. The mean number of pre-bronchodilatory spirometry measurements available for each patient was 8.4 (SD 5.5 measurements) during the mean follow-up time of 4.6 years (SD 2.9 years). Using the Hierarchical Bayesian model (HBM), the mean annual decline across the cohort in FEV1 was 41 ml (SD 39 ml). Of the total 607 patients, 14 were above the lower limit of normal (LLN) of FEV1/FVC at the start of the follow-up, while 8 of them reached LLN during the follow-up period.
Table 1. Clinical characteristics of the patient population.
Identification of the rapid decliners
The great majority of patients (N = 548, 90.2%) showed a declining slope for FEV1 development (). The accuracy of the slope, however, is dependent on the within-patient variation between the sequential spirometry measurements and the number of measurements available. We identified 180 out of 607 (30%) patients who showed constant and rapid decline in their FEV1 (Table , ). Even though 59 (10%) patients had a positive slope in , only 11 (1.8%) of them were shown to have significantly improving FEV1. The mean annual volume lost in FEV1 among rapid decliners was significantly larger than among the rest of patients: 78 ml (SD 33 ml) per year and 26 ml/year (SD 29 ml), p ≤ 2.2 × 10-16 (Table , ).
Figure 1. Distribution of estimated annual change in FEV1 among the COPD patients.
Table 2. Comparison of lung function at the beginning of the follow-up, the change during the follow-up and certain demographic characteristics that potentially affect the results among the decliners and non-decliners based on FEV1.
Figure 2. Distribution of estimated annual decline in FEV1 in the rapid decliners.
Trajectories presented in distributions () do not demonstrate the uncertainty related to each trend. The trajectories consist of multiple measurements over time with individual patterns, and the probability of observing a significant individual trajectory needs to be assessed, as done by way of illustration in .
Both baseline FEV1 % predicted and absolute FEV1 were significantly but weakly correlated with FEV1 decline rate (RFEV1% = -0.23, p ≤ 2.2 × 10-16, RFEV1 = -0.23, p ≤ 2.2 × 10-16). The loss of FEV1 was most notable in mild airway obstruction and better lung function was a risk factor for rapid decline (OR = 1.35 for each additional liter in FEV1 volume, p = 0.0171).
Since the follow-up time and number of the measurements available varied between patients, we assessed the correlation between the decline rates and the follow-up time between the first and last spirometry and the number of measurements. No correlation was found between the FEV1 decline and the length of the follow-up (r = -0.07, p = 0.07) or the number of the measurements (r = -0.04, p = 0.37).
Risk factors associated with rapid decline
To assess the capability of the method and hospital based data used in this study we assessed whether the rapid decliners identified were overlapping with the patients suffering from frequent exacerbations. Several markers of exacerbations were tested: use of antibiotics and oral steroids as markers of mild home treated exacerbations and emergency visits, hospital admissions and X-ray verified pneumonias as markers of severe exacerbations (Table ). All these events were significantly more frequent in rapid FEV1-decliners than among the rest of the patients. More than five emergency visits (crude OR = 2.47, p = 0.0003), more than one hospital admission (OR = 2.43, p < 0.0001), and at least one pneumonia (OR = 1.82, p = 0.0071) during the 5-year follow-up time increased the risk of being rapid decliner.
More than one purchase of oral corticosteroids was associated with the increased risk of being a rapid decliner (OR = 2.49, p < 0.0001) as well as more than five purchases of antibiotics (OR = 2.22, p < 0.0001). Gender, age, response to bronchodilators (change in FEV1 % predicted) or common co-morbidities were not significantly associated with the rapid decline, while the effect of decreased diffusion capacity was found suggestive. Physical inactivity was found to be associated with increased risk of rapid decline of FEV1 (OR = 2.16, p < 0.0001).
Figure 3. A panel chart of FEV1 slopes for a random sample of study subjects with varying number of measurements available. At 95% probability level HBM identified patients as rapid decliners (black), improvers (white) or patients whose lung function development was considered to present random variation (grey).
Table 3. Clinical characteristics and events associated with the risk of being a rapid FEV1 decliner.
Smoking habits were processed as continuous (pack-years, quitting-age) and dichotomous (continuous or quitters, quitting age < 61) risk factors of rapid decline. Continuous smoking had a borderline significant effect (OR = 1.44, p = 0.06) as did late cessation-age (60+ years) (OR = 1.42, p = 0.07), while continuous smoking related factors remained non-significant.
Discussion
Recently a substantial amount of variation in the development of lung function in COPD has been described (Citation3,Citation4,Citation12,Citation13). In the present study we found that HBM was able to identify the rapid FEV1 decliners with a high probability. These patients represented 30% of all study subjects and were more frequently found among mild/moderate than among the patients with severe airway obstruction. Among the rapid decliners the mean annual loss was 78 ml/year (range -274 to -23 ml), while among the remainder of the patients the decline was only 26 ml/year (-78 to 170 ml). These estimates are close to what has been reported earlier among rapid FEV1 decliners. A prospective study of 2163 patients with clinical features close to those of the present cohort (age 63 years, proportion of women 35%, 50 pack-years, FEV1 50% of predicted) was followed for 3 years. Based on 3–8 spirometry measures, 38% of the patients were estimated to lose FEV1 more than 40 ml/year (Citation4).
Only in one previous study, the uncertainty in individual decline rates of the patients has been assessed (Citation5). In this prospective cohort of 751 patients (age 66 years, proportion of women 8%, 66 pack-years, FEV1 50% of predicted), 18% of patients were determined as rapid decliners with the FEV1 decline rate of 86 ml/year (95% CI 32–278 ml/year)5. Airway obstruction at baseline or the decline rate was not associated with pack-years. The rapid decliners also had a higher mortality rate. In a 5-year Japanese observational study (N = 279) (age 70 years, proportion of women 5%, 64 pack-years, FEV1 58% of predicted) the patients were divided into three groups: rapid (mean loss of FEV1 63 ml/year) and slow decliners (mean loss of FEV1 31 ml/year) and sustainers (mean loss of FEV1 2 ml/year). Rapid decline was associated with CT verified emphysema, but not baseline FEV1 level (Citation3).
In regard of the progressive nature of COPD, previous studies have reported an unexpectedly large number of patients who show an improving trend of FEV1 (Citation3–Citation5). A similar distribution in mean trends was found also in our cohort (), when compared to prospective studies. However, when the probability of the finding was estimated, only 11 out of 607 patients (1.8%) were shown to be improvers. In the rest of patients the apparent improvement could be explained by random variation only. These observations underline the importance of the estimation of significance of the individual trends before the final inferences are made.
As in normal clinical practice the number of post-bronchodilatory values was sparser than pre-bronchodilatory spirometries and thus we chose to use pre-bronchodilation values in the analysis, which could slightly increase the intra-patient variation in the course of FEV1. A large epidemiological study (N = 4484) (Citation14) among mild and moderate COPD patients however showed that pre-bronchodilatory values are applicable for longitudinal lung function analysis. The slopes as well as the standard errors for pre-bronchodilation values were only minimally higher compared to those for post-bronchodilatory values. In another study (Citation15) with data from the UPLIFT trial (Citation16) on tiotropium, the post-bronchodilatory estimates of lung function were found to steepen the slope of FEV1 and FVC whereas the variation of lung function development was found to remain the same. The effect of bronchodilation in reducing the intra-individual variation of longitudinal lung function remains to be determined (Citation17).
The present study is based on real life clinical spirometry data and is thus exposed to additional amount of variation when compared to that in prospective studies. On the other hand, our study design, which reflected the real world clinical practice in hospital-based care, suggested that intrapatient variation can be controlled and individual level decisions are possible with a high probability. Also, the patients lost to follow-up were largely those with less advanced COPD, in contrast to many randomized controlled trials and can be considered as strength in our study.
Clinically the most relevant finding of our study was that the 30% of patients identified as rapid decliners also more often suffered from COPD exacerbations. This highlights the ability of this method to identify the patients in risk of a disabling disease in a clinical setting. Even though the rapid decliners had less advanced airway obstruction, they were at greater risk of having emergency visits (OR = 2.47), recurrent hospital admissions (OR = 2.43) and pneumonia(s) (OR = 1.82) than the non-decliners. Based on drug claims in the sub-cohort (N = 514), the rapid decliners had also higher risk of oral corticosteroid (OR = 2.49) and antibiotic (OR = 2.22) treatments as markers of milder, home treated exacerbations. BMI, comorbidities, gender or age were not associated with lung function development. Smoking cessation showed only borderline significance in the decline of lung function in the present study and its role has been debated also in the earlier publications (Citation3–Citation5,Citation17). Very often COPD patients are all heavy smokers, and the smoking variables used in the analyses are rough estimates of smoking behavior that might dilute the results. On the other hand, the sensitivity to tobacco-smoke varies between patients as some patients may continue to lose lung function even after smoking cessation while others sustain their level after decades of smoking (Citation18).
Clinical, radiological, and molecular phenotyping enables us to divide COPD into smaller and smaller subphenotypes (Citation1). At the same time treatment options are rapidly growing which has created the demand of statistical methods that can identify divergent individuals in an epidemiological setting. Traditionally the inferences from clinical trials and observational studies are made based on mean rates of change across patients and the treatments have been designed for these average patients. This approach is often taken to improve the statistical power or analysis but at the same time the aggregation of data leads to loss of information, bias known as the ecological fallacy.
This might lead to biased inferences when the uncertainty related to each individual's trend is discarded and the population means of development are used to estimate the trajectories of individuals. That has also concerned COPD in which treatment guidelines have until recently been based solely on cross-sectionally measured lung function (Citation19). Useful tools such as the SPIROLA-software (Citation20) exist to assess the decline of lung function and whether it could be considered as excessive when compared to reference populations. Our method, however, relies on past measurements for each patient and is not subjective to reference populations as the aim is to detect significant trajectories in individual FEV1.
Selection of rapid decliners for genetic analysis might unveil new insights into the underlying mechanisms of disease progression in COPD, while only two genome-wide association studies on lung function development have been published so far (Citation21,Citation22). In the future clinical medicine needs to be more predictive, preventive and personalized to recognize the high-risk individuals and to meet the needs of diagnostics for the tailor-made treatments in COPD. This study presents a novel approach to understand the concepts of individual trajectory of FEV1 and probability of significant change over time.
Acknowledgments
The authors would like to thank clinical research nurses Kerstin Ahlskog, Kirsi Elorinne, and Päivi Laakso for their skillful patient recruitment, Tuula Lahtinen for the monitoring of the project and research assistant Sari Nummijoki for data management. The authors would also like to thank Prof. Dr. D.S. Postma for assistance in the writing process.
Declaration of interest statement
All authors have completed the ICMJE uniform disclosure form and declare that we have no competing interest. The authors alone are responsible for the content and writing of the paper.
Funding
This study was supported by the funding of Tekes (Intelligen Monitoring), Finnish Anti-Tuberculosis Association Foundation, Research Foundation of the Pulmonary Diseases, Ida Montin Foundation, Väinö and Laina Kivi Foundation and Kela (Social Insurance Institution of Finland).
References
- Friedlander AL, Lynch D, Dyar LA, Bowler RP. Phenotypes of chronic obstructive pulmonary disease. COPD 2007; 4(4):355–384. doi: 785381889 [pii].
- Fletcher C, Peto R. The natural history of chronic airflow obstruction. Br Med J 1977; 1(6077):1645–1648.
- Nishimura M, Makita H, Nagai K, Annual change in pulmonary function and clinical phenotype in chronic obstructive pulmonary disease. Am J Respir Crit Care Med 2012; 185(1):44–52. doi: 10.1164/rccm.201106-0992OC; 10.1164/rccm.201106-0992OC.
- Vestbo J, Edwards LD, Scanlon PD, Changes in forced expiratory volume in 1 second over time in COPD. N Engl J Med 2011; 365(13):1184–1192. doi: 10.1056/NEJMoa1105482; 10.1056/NEJMoa1105482.
- Casanova C, de Torres JP, Aguirre-Jaime A, The progression of chronic obstructive pulmonary disease is heterogeneous: The experience of the BODE cohort. Am J Respir Crit Care Med 2011; 184(9):1015–1021. doi: 10.1164/rccm.201105-0831OC; 10.1164/rccm.201105-0831OC.
- Tantucci C, Modina D. Lung function decline in COPD. Int J Chron Obstruct Pulmon Dis 2012; 7:95–99. doi: 10.2147/COPD.S27480; 10.2147/COPD.S27480.
- Chronic obstructive pulmonary disease exacerbations: latest evidence and clinical implications. Qureshi H, Sharafkhaneh A, Hanania NA. Ther Adv Chronic Dis 2014; 5(5):212–227. doi: 10.1177/2040622314532862. Review.
- Kauppi P, Kupiainen H, Lindqvist A, Overlap syndrome of asthma and COPD predicts low quality of life. J Asthma 2011; 48(3):279–285. doi: 10.3109/02770903.2011.555576; 10.3109/02770903.2011.555576.
- Laitinen T, Hodgson U, Kupiainen H, Real-world clinical data identifies gender-related profiles in chronic obstructive pulmonary disease. COPD 2009; 6(4):256–262.
- Viljanen AA, Halttunen PK, Kreus KE, Viljanen BC. Spirometric studies in non-smoking, healthy adults. Scand J Clin Lab Invest Suppl 1982; 159:5–20.
- Katajisto M, Kupiainen H, Rantanen P, Physical inactivity in COPD and increased patient perception of dyspnea. Int J Chron Obstruct Pulmon Dis 2012; 7:743–755. doi: 10.2147/COPD.S35497 [doi].
- Jenkins CR, Jones PW, Calverley PM, Efficacy of salmeterol/fluticasone propionate by GOLD stage of chronic obstructive pulmonary disease: Analysis from the randomised, placebo-controlled TORCH study. Respir Res 2009; 10:59. doi: 10.1186/1465-9921-10-59; 10.1186/1465-9921-10-59.
- Tashkin DP, Celli B, Senn S, A 4-year trial of tiotropium in chronic obstructive pulmonary disease. N Engl J Med 2008; 359(15):1543–1554. doi: 10.1056/NEJMoa0805800; 10.1056/NEJMoa0805800.
- Tashkin DP, Wang HJ, Halpin D, Comparison of the variability of the annual rates of change in FEV1 determined from serial measurements of the pre- versus post-bronchodilator FEV1 over 5 years in mild to moderate COPD: Results of the lung health study. Respir Res 2012; 13:70 doi: 10.1186/1465-9921-13-70; 10.1186/1465-9921-13-70.
- Tashkin DP, Li N, Halpin D, Annual rates of change in pre- vs. post-bronchodilator FEV1 and FVC over 4 years in moderate to very severe COPD. Respir Med 2013; 107(12):1904–1911. doi: 10.1016/j.rmed.2013.08.001 [doi].
- Tashkin DP, Celli B, Senn S, A 4-year trial of tiotropium in chronic obstructive pulmonary disease. N Engl J Med 2008; 359(15):1543–1554. doi: 10.1056/NEJMoa0805800 [doi].
- Tashkin DP. Variations in FEV1 decline over time in chronic obstructive pulmonary disease and its implications. Curr Opin Pulm Med 2013; 19(2):116–124. doi: 10.1097/MCP.0b013e32835d8ea4; 10.1097/MCP.0b013e32835d8ea4.
- Cosio MG, Saetta M, Agusti A. Immunologic aspects of chronic obstructive pulmonary disease. N Engl J Med 2009; 360(23):2445–2454. doi: 10.1056/NEJMra0804752 [doi].
- Global initiative for chronic obstructive lung disease (GOLD). Global strategy for the diagnosis, management, and prevention of COPD. Available from: www.goldcopd.org. Updated 2014. (accessed 2 May, 2014).
- SPIROLA Software. Available from: http://www.cdc.gov/niosh/topics/spirometry/spirola-software.html (accessed 1 February, 2015).
- Tang W, Kowgier M, Loth DW, Soler Artigas M, Joubert BR, Hodge E, Gharib SA, Smith AV, Ruczinski I, Gudnason V, Mathias RA, Harris TB, Hansel NN, Launer LJ, Barnes KC, Hansen JG, Albrecht E, Aldrich MC, Allerhand M, Barr RG, Brusselle GG, Couper DJ, Curjuric I, Davies G, Deary IJ, Dupuis J, Fall T, Foy M, Franceschini N, Gao W, Gläser S, Gu X, Hancock DB, Heinrich J, Hofman A, Imboden M, Ingelsson E, James A, Karrasch S, Koch B, Kritchevsky SB, Kumar A, Lahousse L, Li G, Lind L, Lindgren C, Liu Y, Lohman K, Lumley T, McArdle WL, Meibohm B, Morris AP, Morrison AC, Musk B, North KE, Palmer LJ, Probst-Hensch NM, Psaty BM, Rivadeneira F, Rotter JI, Schulz H, Smith LJ, Sood A, Starr JM, Strachan DP, Teumer A, Uitterlinden AG, Völzke H, Voorman A, Wain LV, Wells MT, Wilk JB, Williams OD, Heckbert SR, Stricker BH, London SJ, Fornage M, Tobin MD, O'Connor GT, Hall IP, Cassano PA. Large-scale genome-wide association studies and meta-analyses of longitudinal change in adult lung function. PLoS One 2014; 9(7):e100776. doi: 10.1371/journal.pone.0100776.
- Imboden M, Bouzigon E, Curjuric I, Ramasamy A, Kumar A, Hancock DB, Wilk JB, Vonk JM, Thun GA, Siroux V, Nadif R, Monier F, Gonzalez JR, Wjst M, Heinrich J, Loehr LR, Franceschini N, North KE, Altmüller J, Koppelman GH, Guerra S, Kronenberg F, Lathrop M, Moffatt MF, O'Connor GT, Strachan DP, Postma DS, London SJ, Schindler C, Kogevinas M, Kauffmann F, Jarvis DL, Demenais F, Probst-Hensch NM. Genome-wide association study of lung function decline in adults with and without asthma. J Allergy Clin Immunol 2012; 129(5):1218–1228. doi:10.1016/j.jaci.2012.01.074.