Statins are widely used as first-choice treatment for preventing conditions affecting the heart or blood vessels. These conditions are known as cardiovascular diseases (CVD). Multiple clinical trials support the use of statins to reduce the risk of CVD in people with pre-existing CVD and there is growing evidence on their benefits in primary prevention among people with low cardiovascular risk.
Previous observational studies indicate that patients taking statins may have a higher risk of developing type-two diabetes compared to patients that do not take statins. These studies used methods that may not adjust for all differences between treatment groups. I will use methods that can, in theory, adjust for such differences to estimate the effect of statins on the development of type-two diabetes. The aim is to provide guidance on the true risks of taking statins although the latter is unknown.
We will test methods on made-up data in which we know the right answer. Using these data, we will assess which methods perform best in different situations. The simulated data will be created based on the analysis above so that they look like real patient data.
There are concerns that the use of statins can increase patients risk of developing type-2 diabetes. The health problems and complications associated with type 2 diabetes may affect the benefit-risk ratio of prescribing statins to patients with low risk of cardiovascular disease.
A previous study found the risk of developing type 2 diabetes to be 57% (95% CI [54%, 59%]) higher in patients on statins compared to those not on statins. The authors used a propensity score method to adjust for measured confounding. However, their results may still be affected by unmeasured confounding factors.
Instrumental variables methods can be used to obtain causal effect estimates even in the presence of unmeasured confounding provided the assumptions have been satisfied. To investigate potential bias, the previous analysis by Macedo et al. will be extended using instrumental variables, structural equation models and regression discontinuity analysis to obtain unbiased estimates of the effect of statins on the development of type-2 diabetes.
The Cox-proportional hazards model with propensity score adjustment used in the previous study will be replicated on an updated data cut. These will be compared with estimates obtained from an Aalen additive hazards instrumental variables approach. Structural equation model and regression discontinuity approaches will also be applied.
This analysis will form the basis of a simulation study to compare the different methods empirically. The data will be simulated based on features of the study data to ensure that the simulated data are as realistic as possible.
Incident type 2 diabetes
Secondary outcomes:
Diabetes risk in patients with major type 2 diabetes risk factors: metabolic syndrome, fasting plasma glucose 100-125 mg/dl, BMI >30 and HbA1c > 6%.
Cholesterol
Cardiovascular Disease Events
Nuala Sheehan - Chief Investigator - University of Leicester
Ellie John - Corresponding Applicant - University of Leicester
Kate Tilling - Collaborator - University of Bristol
Keith Abrams - Collaborator - University of York
Liam Smeeth - Collaborator - London School of Hygiene & Tropical Medicine ( LSHTM )
Michael Crowther - Collaborator - Karolinska Institute Sweden
Paul Lambert - Collaborator - University of Leicester
Umesh Kadam - Collaborator - University of Leicester
Michael Crowther - Collaborator - University of Leicester
HES Admitted Patient Care;ONS Death Registration Data;Patient Level Index of Multiple Deprivation;Pregnancy Register