Assessing causal relationships between diabetes mellitus and idiopathic pulmonary fibrosis: a Mendelian randomisation study

Abstract

Background: Idiopathic pulmonary fibrosis (IPF) is a disease of progressive lung scarring. There is a known association between diabetes mellitus (DM) and IPF, but it is unclear whether a causal relationship exists between these traits.

Objectives: The objectives of this study are to examine causal relationships among DM, diabetes-associated traits and IPF using a Mendelian randomisation approach.

Methods: Two-sample MR approaches, including bidirectional inverse-variance weighted random effects and routine sensitivity models, used genetic variants identified from genome-wide association studies for type 1 diabetes (T1D), type 2 diabetes (T2D), glycated haemoglobin level (HbA1c), fasting insulin level and body mass index (BMI) to assess for causal effects of these traits on IPF. Further analyses using pleiotropy-robust and multivariable MR (MVMR) methods were additionally performed to account for trait complexity.

Results: Results did not suggest that either T1D (OR=1.00, 95% CI 0.93 to 1.07, p=0.90) or T2D (1.02, 0.93 to 1.11, p=0.69) are in the causal pathway of IPF. No effects were suggested of HbA1c (1.19, 0.63 to 2.22, p=0.59) or fasting insulin level (0.60, 0.31 to 1.15, p=0.12) on IPF, but potential effects of BMI on IPF were indicated (1.44, 1.12 to 1.85, p=4.00×10^-3). Results were consistent in MVMR, although no independent effects of T2D (0.91, 0.68 to 1.21, p=0.51) or BMI (1.01, 0.94 to 1.09, p=0.82) on IPF were observed when modelled together.

Conclusions: This study suggests that DM and IPF are unlikely to be causally linked. This comorbid relationship may instead be driven by shared risk factors or treatment effects.

Keywords: Clinical Epidemiology; Idiopathic pulmonary fibrosis.

Figure 1. An overview of study design and analytical approaches. (A) Example directed acyclic graphs (DAGs) presenting a bidirectional Mendelian randomisation (MR) study design testing for causality between type 2 diabetes (T2D) and idiopathic pulmonary fibrosis (IPF). In each DAG, arrows indicate the directionality of the potential causal relationship that is being tested. Genetic variants that are strongly correlated with an exposure variable (G_T2D for a T2D exposure, G_IPF for an IPF exposure) are used as instrumental variables to test for causal effects on particular outcome (IPF and T2D, respectively). (B) A DAG presenting the multivariable MR design applied in this study. Genetic variants associated with multiple correlated exposures (G_Exp) (in this example: T2D, glycated haemoglobin level (HbA1c), fasting insulin and body mass index (BMI)) are combined and integrated into the MR analysis framework to test for direct causal effects of each exposure on a single outcome (IPF). (C) A diagrammatic overview of this study, including potential causal relationships tested in main, multivariable and secondary analyses and MR methods used at each stage. IVW-RE, random-effects inverse-variance weighted; T1D, type 1 diabetes.

Figure 2. A forest plot and summary table of causal estimates from bidirectional Mendelian randomisation (MR) analyses testing for potential causal relationships between diabetes mellitus (DM) (type 1 diabetes (T1D) and type 2 diabetes (T2D)) and idiopathic pulmonary fibrosis (IPF). Error bars show 95% CI for overall estimates (OR) from each MR method (left). IVW-RE, random-effects inverse-variance weighted.

Figure 3. A forest plot and summary table of causal estimates from Mendelian randomisation (MR) analyses testing for potential causal effects of continuous diabetes mellitus (DM)-associated variables (glycated haemoglobin level (HbA1c), fasting insulin (FI) and body mass index (BMI)) on idiopathic pulmonary fibrosis (IPF). Error bars show 95% CI for overall estimates (OR) from each MR method (left). IVW-RE, random-effects inverse-variance weighted.

Figure 4. A forest plot and summary table of causal estimates from multivariable Mendelian randomisation (MVMR) analyses testing for direct causal effects of type 2 diabetes (T2D) and continuous diabetes mellitus (DM)-associated variables (glycated haemoglobin level (HbA1c), fasting insulin (FI), and body mass index (BMI)) on idiopathic pulmonary fibrosis (IPF). Error bars show 95% CI for overall estimates (OR) from each Mendelian randomisation method (left).