SAFE-PCI for Women Offers Clues on How Registry-Based Feedback Might Inform Future Trials

A comparison of SAFE-PCI participants versus NCDR CathPCI registry patients points to selection bias, not physician behavior.

SAFE-PCI for Women Offers Clues on How Registry-Based Feedback Might Inform Future Trials

Randomized controlled trials that harness existing registries for patient recruitment can provide an uncommon perspective on sources of bias, suggests a new analysis using data from the SAFE-PCI for Women trial.

“There’s increasing interest in how we make clinical trial populations more generalizable, to make them look like what our real-world patients look like,” lead author Jennifer A. Rymer, MD, MBA (Duke Clinical Research Institute, Durham, NC), told TCTMD.

SAFE-PCI, which compared radial versus femoral access in women undergoing cardiac catheterization or PCI, had lower-than-expected event rates and thus was terminated early. However, Rymer and her co-authors point out in their paper published recently in Circulation: Cardiovascular Interventions, whether these better-than-anticipated outcomes were due to selection of lower-risk patients or to differences in physician behavior has not been clear.

As the investigators note, their findings—namely, that SAFE-PCI trial participants were at lower risk, had fewer bleeding events, and were younger than their peers in the same registry—point to a way that trialists might receive feedback throughout enrollment that could in turn be used to curb bias. “This study highlights the opportunity for registry-based randomized control trials to allow for point-of-care data entry and embedding of trial inclusion and exclusion criteria into the electronic health record to allow for close to real-time assessment of trial eligibility,” they explain.

NCDR CathPCI Registry Comparison

SAFE-PCI for Women was conducted within the American College of Cardiology (ACC)’s National Cardiovascular Data Registry (NCDR) CathPCI registry. This allowed Rymer and colleagues to compare 496 women who underwent PCI as part of the trial with registry patients in three other groups:

  • Group 1: 15,904 trial-eligible registry patients treated during the SAFE-PCI trial
  • Group 2: 13,315 treated in the year before SAFE-PCI
  • Group 3: 11,643 treated in the year after SAFE-PCI

On the whole, trial participants were younger, had lower predicted risks of bleeding and mortality, and experienced less post-PCI bleeding within 72 hours versus the registry patients. Vascular complications trended lower for the SAFE-PCI subjects, but the difference was not significant.



(n = 496)

Group 1

(n = 15,904)

Group 2

(n = 13,315)

Group 3

(n = 11,643)

P Value

Median Age, years





< 0.01

Predicted Risk of Bleeding





< 0.01

Predicted Risk of Mortality





< 0.001

Bleeding ≤ 72 Hours





< 0.01

And an analysis that employed Centers for Medicare & Medicaid Services data on 12,212 women, the combined rates of 30-day death and revascularization were similar across all four groups (3.7%, 1.6%, 1.7%, 1.6%, respectively; P = 0.66). Rymer observed to TCTMD that it’s impossible to draw any conclusions about the numerically higher event rate among SAFE-PCI participants, given that the sample size is so small.

The investigators conclude: “Our analysis suggests that selection bias played a role in the SAFE-PCI trial and its lower-than-expected rate of bleeding or vascular complications.”

The NCDR infrastructure was a key factor that made this comparison among groups possible, Rymer stressed. “A lot of credit goes to the ACC for working with various institutions to try to start integrating trials into registries, to help with quicker recruitment and to make sure that in real time you’re getting patients that look like real-world patients.”

Ideally, a registry-based process would be seamless thanks to centrally housed data, she said. “You wouldn’t need a lot of work by study coordinators in order to go into these complex data collection forms and fill in all this information on the patients because that would already be being done. That’s the vision for the future.”

Being able to examine the population on a week-by-week basis might identify certain sites that are enrolling lower-risk patients and hint at why that’s happening, she said. “Nobody does this purposefully, but when you’re thinking about clinical trials, they have very stringent inclusion/exclusion criteria. . . . Even when you’re approaching the patients, there’s a lot of selection bias in just who will sign up and who will agree to be in a trial. Some of that’s unavoidable, but what’s not unavoidable is only approaching patients who are [ideal].”

SAFE-PCI also showed the potential for reducing the cost of conducting a trial, Rymer added, because “a lot of the expense of big clinical trials comes from the effort needed at the sites to input that data.”

Another “pragmatic trial” to keep an eye on, Rymer noted, is ADAPTABLE. That study is using electronic health records to streamline enrollment, as well as an online portal to gather patient-reported outcomes.

  • This study was funded by the NCDR.
  • Rymer reports receiving a research grant from the American College of Cardiology.

We Recommend