Background/Purpose: Observational studies of both chronic and short term glucocorticoid (GC) use have suggested an elevated risk of acute myocardial infarction (MI). However this could be a result of confounding by indication; i.e. some condition for which a GC is prescribed is also a risk factor for MI. This study’s aim was to examine whether ‘burst’ GC use serves as a trigger for acute MI or whether any effect is likely to be the result of confounding by indication.

Methods: We used national Veterans Administration (VA) data from fiscal year 1998 through 2008 to compare the risk for MI during and around periods of GC use compared to periods of non-use using the self-controlled case series study design, a case only design which compares risks within persons, thus limiting confounding. Based on prescription records we first developed a cohort of subjects who had only used ‘burst’ GCs, defined as dispensed oral prescriptions of ² 30 days with at least 42 days between consecutive prescriptions; we excluded persons who received GCs during their first 60 days of follow-up or who could have received GCs during a prior hospitalization.  From these, we selected those persons who had a first MI requiring hospitalization (cases) using validated ICD-9 codes; we excluded cases who had an MI within the first year of follow-up (to limit the number with possible recurrence). We focused on the period in which each subject was using GCs, as well as 42 days before and after to account for confounding by indication and any residual GC effects. The risk of MI in each of these periods was compared to that in the remaining follow up time period.  We controlled for age in 5 year bands (18-24, 25-29, 30-34, … , > 80).

Results: There were 1632 cases of MI among burst GC users. 97.5% were men. Mean age at MI was 66.8 (sd 11.4) years; 66.9 (11.4) in men, 59.5 (12.9) in women.  75.4% of subjects were white, 10.1% African American, 5.0% Hispanic, 10.0% other. All cases had only 1 GC prescription for a median duration of 6 (Q1, Q3: 6, 10) days and median average daily prednisone equivalent dose of 17.5 (17.5, 22.0) mg.  The incidence rate ratio (IRR) of MI rose over the 42 days prior to the GC prescription from baseline to 7.5, and then dropped while the GC was being used.  The risk essentially returned to baseline after GC was discontinued.  See table.

Conclusion: Our results suggest that any elevated risk of MI from GC use could be due to confounding by indication: in this study the risk rose before the GC prescription was issued, suggesting that the MI was associated with some other factor for which the GC might also have been prescribed at a later time.