Journal of Invasive Cardiology

Study limitations. There are several limitations to our study. First, our analysis is a non-randomized observational cohort study: FFR of intermediate coronary lesions was performed according to operator discretion — not per protocol —as would be the case in a randomized controlled trial. As such, unrecognized confounders may have entered into the clinical decision to initially perform an FFR measurement, or in the ultimate decision regarding revascularization. Our inclusion of only a small number of lesions over a decade-long period suggests this is the case, despite the fact that the same operators performed the procedures over the entire course of the study period. Since we did not examine data on patients in whom FFR was not performed, it is difficult to discern the rationale behind FFR use in each case compared to those in which it was not performed.

One possibility is that these unmeasured confounders may have selected for a lower-risk patient group, as higher-risk patients may have been felt to have a greater need for aggressive revascularization without further ancillary coronary testing. Given the comparatively high burden of co-morbid conditions in our patient group, and the mortality rate of approximately 28% over a mean of 6 years, we feel this is unlikely to be true.

Certainly, operator comfort and expertise in performance and interpretation of FFR, as well as trends in the overall use of FFR, changed over the study period. As the accumulated data  pool supporting the use of FFR has grown, so has the use of FFR in our laboratory, raising the possibility that there may also be unrecognized differences in FFR use among patients included in earlier and later parts of the study. Additionally, more recent trends in technique of FFR measurement, including utilization of increased adenosine doses for increased sensitivity, are not captured with our data.¹⁸

The ability to capture follow-up events that did not occur within our medical system was also limited. Although many of our patients exclusively receive medical care within the VA healthcare system, and significant hospitalizations and cardiac events were likely to be captured in the electronic medical record even when occurring outside of our system, it is possible that significant clinical events were missed. Likewise, while patient deaths occurring outside of the VA system are recorded in the medical record, data useful in determining a cause of death were limited. Follow-up was also incomplete in 3 patients. An additional analysis in which a “worst case scenario” of repeat revascularization, MI, and cardiac death was assigned to the 3 missing patients was performed and no significant alterations in the overall findings of this cohort study were found.

Our study began in the era prior to drug-eluting stents and concluded in 2005 when a large majority of procedures were done with these devices. It is possible that our results would have been different if all the procedures were done in the current era, but this limitation is inherent in any study evaluating very long-term outcomes since therapies change over time.