April 12th, 2012
Ascertaining ASCERT: How Well Do Registry Data Measure Up to the ‘Bedside Test’?
In the ASCERT observational study, stable patients (age 65 or older) with double- or triple-vessel CAD, but not left-main disease, were found to have better long-term survival after CABG than after PCI. ASCERT was a laudable achievement in terms of its scope and the level of collaboration it represents.
The ASCERT investigators used inverse probability weighting (propensity scores) in an attempt to mitigate the impact of selection bias. However, this technique depends on measured variables. Experts are currently debating whether the registries analyzed in ASCERT — the STS Adult Cardiac Surgery Database and the ACCF National Cardiovascular Disease Registry (NCDR) — adequately captured clinical characteristics that both influenced treatment selection and were associated with outcomes.
In a study published last year, my colleagues and I leveraged the granularity of the Kaiser medical record to assess, in 101 patients undergoing nonemergent, unprotected left-main PCI, the prevalence of typically uncaptured clinical conditions and their impact on post-PCI outcomes. The majority of patients in our series were considered by their treating clinicians to be ineligible for CABG.
We extracted the reasons that the treating clinicians cited for CABG ineligibility. The most common reasons were advanced age, severe lung disease, and poor targets or inadequate conduits, but frailty, malignancy, and severe aortic calcification were also frequently cited. Notably, the majority of the reasons for CABG ineligibility were not captured by the NCDR. Not surprisingly, both CABG ineligibility and the presence of a “non-NCDR captured risk factor” were associated with markedly worse long-term post-PCI outcomes, even after accounting for ascertainment bias and adjustment with standard risk scores.
Ours was a single-center study restricted to patients with left-main CAD. However, selection bias (related to factors typically not captured in registries) is likely to also be pertinent to patients with multivessel CAD. That bias could confound results from observational studies that rely on registry data. It remains to be seen whether such registries can ever capture sufficient detail to account for all the factors that influence bedside clinical judgment and, therefore, to allow for truly effective comparisons.
Are data from the NCDR suitable for comparing outcomes after PCI versus CABG if important clinical factors, like those identified in the Kaiser study, are missing?