Effects of Changing Guidelines on Prescribing Aspirin for Primary Prevention of Cardiovascular Events

Jennifer Hissett, MD, Brittany Folks, MD, Letoynia Coombs, MS, William LeBlanc, PhD, and Wilson D. Pace, MD


J Am Board Fam Med. 2014;27(1):78-86. 

In This Article


Despite evidence disputing the benefit of aspirin for primary prevention of CVD events, as well as changing guidelines, patients in the observed practices demonstrated increased aspirin use for primary prevention across all groups with CVD risk equivalents. The data do not indicate any effect on the use of aspirin for primary prevention as a result of new negative primary prevention trials and minor changes in ADA/AHA guidelines that were published between 2007 and 2010. After the time period evaluated in this project, Sheshasai et al[23] published a meta-analysis that includes all 9 randomized controlled trials of aspirin used for primary prevention of CVD, extending previous work through the addition of the latest 3 trials. The analysis included data on >100,000 people and >700,000 person-years. An approximate 10% decrease in nonfatal myocardial infarctions was seen, but there was no change in all-cause cardiovascular or cancer death rates. The nontrivial bleed rates exceeded the decrease in nonfatal myocardial infarctions. The positive results related to nonfatal myocardial infarctions were heavily influenced by older studies, which were conducted before more aggressive treatment of hypertension and hyperlipidemia. The authors conclude that "routine use of aspirin for primary prevention is not warranted."

The data in this study indicate that there has been no decrease in the clearly appropriate use of aspirin in individuals with known CVD. It is reassuring that negative information related to aspirin in primary prevention does not seem to have adversely affected the prescription of aspirin for secondary prevention in these practices.

All practices involved in this study use clinical decision support software that reminds clinicians to consider aspirin in individuals with CVD. Until 2009 this system also prompted clinicians to consider aspirin in individuals with diabetes, chronic kidney disease (based on International Classification of Diseases, Ninth Revision, diagnoses or calculated creatinine clearance), or a calculated Framingham risk score >10%. The use of this software may have increased the recording of aspirin usage from 2007 to 2009 without any actual change in true underlying usage. From 2009 on these recommendations were turned off in some but not all 33 practices.[24]

There are of course limitations to this analysis. Because aspirin is an over-the-counter medication, the use of EHR data to capture usage likely underestimates the percentage of the population actually using the medication. There is no reason to believe that the underestimated usage dramatically changed between study periods, in particular as it relates to newly diagnosed individuals. This dataset comes from 11 distinct clinical organizations across 6 states that are not related except through their participation in eNQUIRENet; thus it is unlikely that widespread, independent medication reconciliation activities fully account for the results. The results for individuals who newly entered the cohorts across time periods also reduce the likelihood that medication reconciliation efforts over time could account for the findings. Thus the observed changes are not likely to be entirely due to recording artifact.

It is clear that many of the newly diagnosed patients placed on aspirin met criteria for primary prevention as outlined by the US Preventive Services Task Force[2] or ADA/AHA guidelines[16–18] at the time. Thus the clinicians in these practices cannot be criticized for following published guidelines, and some were subject to pay-for-performance evaluations based on increasing the use of aspirin for primary prevention. The inclusion of aspirin as a quality metric is likely to have blunted any impact of the new data from controlled trials or the subtle changes in the ADA/AHA guidelines. In addition, during this time period data were emerging that indicated low-dose aspirin may be beneficial in reducing risk of colon cancer. This could have blunted the impact of repeated negative studies for primary prevention of cardiovascular events, although meta-analysis supporting the use of low-dose aspirin for colon cancer prevention appeared long after the time period in question.[25,26] Furthermore, the meta-analysis by Shesasai et al[23] evaluated cancer mortality as well as CVD mortality and found no overall effect of aspirin use on cancer deaths.

With regard to aspirin prescriptions in the EHR, it is impossible to know the validity of the data at the population level. In some cases, the timing of a patient's diagnosis and initiation of aspirin therapy may not be captured accurately. It is possible to imagine patients who are well known to a practice, having been previously diagnosed with diabetes and prescribed aspirin, but the aspirin data are entered later, only after a prompt from the EHR. Likewise, a patient who was placed on aspirin in 2007 may have been discouraged from aspirin use at a later visit but refused to stop taking it. In any case, it is not likely that the reliability of the data changed enough over time to actually mask a decrease in the use of aspirin for primary prevention. The analysis of new diagnoses per time period would argue against this scenario; it is unlikely that new usage would go up over time if clinicians were actively discouraging aspirin use for primary prevention, and medication reconciliation activities would not be a major factor in individuals just entering the high risk cohorts. As aspirin can be obtained without a prescription and is generally considered to be harmless, it is more likely that patients did not report use and that true aspirin consumption is actually higher. While this is good news for the population with known CVD, it is potentially concerning for the primary prevention populations.

A strength of this analysis is that it includes multiple primary care organizations across multiple states using varying EHRs; thus it is unlikely that any organization-specific activity, such as medication reconciliation improvements, could explain the findings. Furthermore, the data are drawn from all individuals >17 years old in these practices and therefore do not represent sampling activities and as such do not require further statistical analysis to represent true changes in this population.