April Forum- ASA for prevention of MI in RA and propensity scoring

Welcome to the April edition of #chatJRheum

The article for discussion this month is on the emerging interest in cardiovascular (CV) disease in RA, and the potential role of low dose aspirin (ASA) for primary prevention.

The efficacy of ASA for primary prevention is not clear. While benefits may outweigh risks in certain high-risk groups, this has not been evaluated in patients with RA who are known to be at increased risk of CV events.

This study was carried out as a prospective cohort study of 1836 patients with RA and no previous CV event, of which 918 patients initiated on ASA were matched to 918 RA patients not on ASA. To minimize confounders, propensity score matching was utilized.

Admittedly this methodology is not familiar to me and can be confusing for those of us not proficient in complex biostatistics. If you felt the same way, continue reading.

• I asked the authors, Drs. Josefina Durán and David Felson, to explain why this methodology was chosen and how it adds value to interpretation of the clinical study. Here is their explanation of propensity score matching 101:

This was an observational study and as such the allocation of the intervention (aspirin) was not random. It is likely that persons were prescribed aspirin because they were perceived to be at high risk of myocardial infarction (MI) so that aspirin use is ‘confounded by indication’----the indication being a high risk of MI.

To make a valid estimate of whether the risk of MI in persons with RA is affected by ASA use, ASA users and nonusers need to be at equal risk of MI EXCEPT for ASA use. Because of confounding by indication they are not otherwise at equal risk.  Propensity Scores (PS) allow us to adjust for the effect of confounding. PS’s summarize the confounding variables into one score per subject. Two subjects with same PS, one on ASA and the other not on ASA, should have a similar risk of MI.  When matching the ASA and non-ASA users on PS, we hope to simulate a randomized trial by creating two groups whose risk of MI is equivalent except for one group’s use of ASA. 

A limitation of using a PS is that you have to carefully choose all variables included… only real confounders must be included and none missed. If we leave some confounders out of the PS (e.g. current hypertension or smoking), the PS will fail to fully control for confounding.  In this study not all variables were measured, such as family history of CV disease, which is a relevant risk factor. Also obesity, alcohol consumption and smoking were not recorded in all patients (and we left those subjects with missing values out of the analysis).

Finally, ideally, we would have had to quantify some variables: well-controlled hypertension does not confer the same risk as subjects who have severe uncontrolled hypertension. Similarly, the risk of smoking a cigarette daily is very different from the risk of smoking more than a pack of cigarettes daily.

To my surprise, the results from the study did not detect a statistically significant difference. This outcome has the potential to influence our day-to-day practice by discouraging the use of ASA for primary prophylaxis in this high-risk group.

• Before we abandon cause, are there reasons this study could have had a negative outcome? The authors provide a plausible explanation that the prothombotic risk of chronic inflammation in RA is not controlled by ASA, and that additional factors are at play. I’d be interested to hear from you, the JRheum community, on this issue.
• Furthermore, what, if any additional studies might be needed to better answer the pervasive question of how to reduce risk of CV events among these patients? Perhaps this same study design might be conducted with combined ASA and statin therapy, which may have added CV protection and pleiotropic effects. Or, dare I say, do we need a randomized control trial? 

Readers, I hope you will find this to be an interesting and timely discussion, and that you will weigh in with your questions, thoughts and comments. Let’s chat!

Dr. Sarah Troster, Forum Editor

(Thank you to authors Josefina Durán and David Felson of the Rheumatology Department, Pontificia Universidad Católica de Chile and Clinical Epidemiology Unit, Boston University)