Evaluation of Small Airways Disease in asthma medical research: Application of Structural Equation Modeling (SEM) and Latent Transition Analysis (LTA)

Transcription

1 Evaluation of Small Airways Disease in asthma medical research: Application of Structural Equation Modeling (SEM) and Latent Transition Analysis (LTA) Thomas Zwingers Marco Pannacci

2 Agenda Study Design Small Airways Disease (SAD) in asthma Structural Equation Model Cross-sectional Model Longitudinal Model Summary 7

3 Study Design and Objectives Study Design: Multinational, multicenter, nonpharmacological intervention study with a cross-sectional and a longitudinal phase. Follow-up time : 1 year Major objectives: To evaluate which (combination of) clinical methods best assesses the abnormalities of small airways disease in asthma and their role in the clinical manifestations of asthma, both cross-sectionally and longitudinally. 7

4 Study Population Planned Sample Size: 900 patients (800 asthmatic patients healthy volunteers). No. of Asthmatic Patients No. of Healthy Volunteers Screened Set Evaluable Set

5 ASTHMA Asthma is a common long-term inflammatory disease of the airways of the lungs. It is characterized by variable and recurring symptoms, reversible airflow obstruction, and bronchospasm. There is no cure for asthma. Poorly controlled inflammation in small airways may contribute to asthma exacerbations, air trapping, lung function decline, and irreversible structural changes Source: 7

6 ASTHMA Asthma Prevalence An estimated 300 million people worldwide suffer from asthma, with 250,000 annual deaths attributed to the disease. It is estimated that the number of people with asthma will grow by more than 100 million by Source: 7

7 ASTHMA Categories of asthma severity Mild asthma: Well controlled with SABA or low dose ICS GINA Steps 1 or 2 Moderate asthma: Well controlled with low dose ICS/LABA GINA Step 3 Severe asthma: requires moderate or high dose ICS/LABA ± add on, or remains uncontrolled despite this treat GINA Steps 4/5 Source: 7

8 Structural Equation Modeling (SEM) SEM is a very general statistical modeling method. It can be viewed as a combination of exploratory factor analysis, confirmatory factor analysis, path analysis or regression and others. Typical example: measurement of intelligence. We can t measure the concept of interest (intelligence, socioeconomic status..) directly latent variables. Hovewer we can refer to observable indicators that are influenced by the latent variable, e.g scores obtained by a series of tests about linguistic or arithmetic abilities observed variables. 8

9 Structural Equation Modeling (SEM) Structural Equation Models consist of a structural model representing the relationship between the latent variables of interest, and measurement models representing the relationship between the latent variables and their manifest or observable indicators Measurement model A I6 I7 Structural model A Measurement model B I5 I4 I3 B I2 C I1 Measurement model C

10 Specification of SEM model Cross-sectional phase The Path Diagram describes the theoretical framework that forms the basis for specification of our initial model. The rationale about hypotheses and directionality specifications in the diagram was built by the statisticians consulting the experts of the clinical team during two TC meetings. Observed variables in the Path Diagram are represented with squares or rectangles; latent variables with circles; hypothesized directional effects (causal relationships) of one variable on another, with a line with a single arrowhead; correlations between variables with a curved line with two arrowheads. 7

11 Specification of SEM model BHQ Total Score Sputum induction FeNO Small/large airways Inflammation Haematology??? Systemic Inflammation Healthcare resource consumption? Phadiatop test ACT Total Score EuroQol VAS Subjective status? Small/large airways disease Asthma history Age at asthma diagnosis Duration of disease ACQ-6 Total Score Lung physiology Number of Exacerbations (last year) Mni Total Score Age Smoke history Body plethysmography Markers Methacholine challenge test Spirometry Markers Impulse Oscillometry System (IOS) Markers Multiple-Breath Nitrogen Washout Test (MBNWT)

12 % of patients Outcome of SEM model: Asthma clusters Overall the clusters represent GINA severity stages predominantly, suggesting that SAD increases with asthma severity GINA Steps vs. Asthma Groups Asthma Group 1 Asthma Group 2 Asthma Group 3 Asthma Group 4 Step 1 Step 2 Step 3 Step This analysis identifies specific asthma, not SAD, clusters.

13 Statistical Analysis Plan - New approach only the variables reflecting small airways disease and both small and large airways disease were considered for the SAD measurement model Small Large Both Systemic effects Patient-reported outcome % fall in FVC at PC20 R20 (kpa L-1 s-1) PC20 (mg/ml) WBC Asthma Control Test (ACT) R5-20 (kpa L-1 s-1) FENO single flow (50ml/s) PD20 (mg) Neutrophils Mini asthma quality of life questionnaire X5 (kpa L-1 s-1) Sputum PD20 & PC20 categories Monocytes Asthma control questionnaire (ACQ-6) AX (Hz kpa L-1 s-1) FEV1 %pred Lymphocytes Bronchial Hyperresponsiveness Questionnaire (BHQ) Scond*VT in liter FEV1/FVC (L/sec) Eosinophils EuroQol-5D-5L Sacin*VT in liter IVC %pred Basophils EuroQol VAS (respondent s self-rated health) RV %pred RV / TLC %pred FRC %pred FEF50%/FVC FEF25-75%/FVC SAD FENO multiple flows (100, 150, 350, other) Phadiatop test Total number of hospital admission days in the past 12 months before V1 Number of asthma-specific emergency room or urgent care visits Number of unscheduled consultations visits due to symptoms worsening Total number of Unscheduled tests for asthma Number of exacerbation in the last year

14 Study Population No. of Asthmatic Patients No. of Healthy volunteers Evaluable Set SAD SEM Analysis* 761 Not used *We lost only 6 patients with no data for small airways disease variables and we excluded 5 patients that were outliers (based on the model indicators for outliers detection).

15 Clinical SAD SEM Cross Sectional Model (.000) SAD.617 (.047).518 (.046).981 (.004).619 (.058).731 (.047) LUNG1 LUNG2.285 (.053) (.005).904 (.009).996 (.000).291 (.056).964 (.021).925 (.021) R5-20 X5 AX Sacin FEF50/FVC FEF25-75/FVC.108 (.009).183 (.016).009 (.001).782 (.042).070 (.041).145 (.039).038 (.007) LUNG3.512 (.043).645 (.058) RV/TLC % pred Scond.738 (.044).585 (.074) Model Fit: Root Mean Squared Error of Approximation(RMSEA): Comparative Fit Index (CFI): 0.990

16 SAD SEM model based clustering Count Clinical SAD SEM Score Clinical SAD SEM Groups 1 2

17 Relationship between Clinical SAD (SEM) groups and GINA classes Clinical SAD SCORE Mean (n) Gina 1 Gina 2 Gina 3 Gina 4 Gina (134) Anova p-value < (83) Mean Clinical SAD SCORE (205) (296) (46) -0.2 Step 1 Step 2 Step 3 Step 4 Step 5

18 Clinical SAD SEM Longitudinal Model Latent Transition Analysis VISIT 1 VISIT 3 I5 I4 I3 I5 I4 I3 I6 I7 A B I6 I7 A B C I2 I1 C I2 I1

19 Clinical SAD SEM Latent Transition Analysis VISIT 1 VISIT 2 VISIT 3 SAD Groups G1 = 59% G2 = 41% SAD Groups G1 = 61% G2 = 39% 6 months 6 months SAD Groups G1 = 60% G2 = 40% 14 patients move from G1 to G2 28 patients move from G2 to G1 14 patients move from G1 to G2 9 patients move from G2 to G1 VISIT 1 27 patients move from G1 to G2 VISIT 3 36 patients move from G2 to G1

20 Summary We were able to define a score that reflects the amount of physiological small airways impairment The score associated significantly with measures of asthma control, disease instability and asthma severity as defined by GINA stages SAD appears to be clinically meaningful as suggested by the association of physiologic parameters of SAD with asthma severity, asthma control and health care utilization We can detect asthma subtypes based on small airways dysfunction

21 Literature 1. Bollen, K. A. (1989). Structural equations with latent variables. New York, NY: John Wiley & Sons, Inc. 2. Bentler P M 1990 Comparative fit indices in structural models. Psychological Bulletin 107: Chung, H., Park, Y., & Lanza, S.T. (2005). Latent transition analysis with covariates: pubertal timing and substance use behaviors in adolescent females. Statistics in Medicine, 24, Muthen B 1984 A general structural equation model with dichotomous, ordered categorical, and continuous latent variable indicators. Psychometrika 49: Stephanie T. Lanza, Linda M. CollinsA : New SAS Procedure for Latent Transition Analysis: Transitions in Dating and Sexual Risk Behavior. Dev Psychol March ; 44(2): William McNulty* and Omar S. Usmani: Techniques of assessing small airways dysfunction. European Clinical Respiratory Journal 2014, 1: 25898