News Daily News

Model Tries to Simplify Choice Between Drug-Eluting, Bare-Metal Stents

By L.A. McKeown

Download this article's Factoid in PDF (& PPT for Gold Subscribers)


A new predictive model may help clinicians decide between drug-eluting stents (DES) or bare-metal stents (BMS) and allow for more individualized assessments of revascularization needs, researchers conclude in a study published online September 6, 2011, ahead of print in Circulation.

Robert W. Yeh, MD, MSc, of Harvard Medical School (Boston, MA), and colleagues developed and validated the model based on the likelihood of TVR occurring within 12 months of PCI. Preprocedural and clinical data came from National Cardiovascular Data Registry (NCDR) CathPCI records of 27,107 PCI hospitalizations that occurred in Massachusetts between October 1, 2004, and September 30, 2007.

The model incorporated a number of clinical variables including:

  • Sociodemographic information
  • Medical history
  • Cardiovascular history
  • Clinical status variables at admission
  • PCI status (elective, urgent, emergent/salvage)

In addition, the researchers identified relevant angiographic variables including:

  • Minimum stent diameter (< 3 vs. ≥ 3 mm)
  • Total stent length (< 30 vs. ≥ 30 mm)
  • Bifurcation lesion
  • Previously treated lesion
  • Lesion within saphenous vein graft
  • Treated lesion within LAD
  • Multivessel disease
  • Number of treated lesions

Commonly Collected Variables Aid in Prediction

At 12 months, TVR had occurred in 7.6% of patients overall (6.7% of DES patients and 11.0% of BMS patients). Patients who developed TVR were younger, more often diabetic, and had higher rates of prior MI, peripheral vascular disease, multivessel disease, prior PCI, and prior CABG. They also tended to receive stents with smaller diameters and longer lengths at the index procedure compared with patients who did not develop TVR. The mortality rates at 1 year were similar for patients with and without TVR (5.3% vs. 5.2%; P = 0.78).

In a model that considered only clinical factors, DES use was associated with a 43% decrease in 1-year TVR risk compared with BMS (OR 0.57; 95% CI 0.52-0.64). Adding angiographic variables to the mix further improved the model discrimination (c statistic 0.62 for clinical model vs. 0.66 for full model; P < 0.001) such that treatment with DES was associated with an even greater reduction in 1-year TVR (OR 0.53; 95% CI 0.47-0.59).

The predicted number-needed-to treat with DES vs. BMS to prevent 1 occurrence of TVR varied broadly, depending on clinical and angiographic factors and ranged from 6 to 80.

Results Supported by Real-World Data

Results were generated from a population-based sample undergoing PCI in routine clinical practice unlike clinical trials or volunteer registries that are subject to selection biases, the researchers point out. In addition, the analysis was performed during a time period that included a large number of both BMS and DES patients, allowing for precise estimation of predicted TVR reduction with DES across the entire spectrum of clinical profiles.

The decrease in TVR with DES use was expected, the study authors say, adding that “[n]one of the interaction terms we explored to examine whether the benefit of DES was modified by clinical factors were found to be significant, so the relative risk reduction in TVR (~45%) was constant for all patients in the study.”

From a practical standpoint, the findings create an opportunity to prospectively identify and use DES in patients who stand to derive the greatest benefit from the devices. BMS can be considered in patients with low anticipated benefit, given the added bleeding risk and costs associated with DES and their requirement for prolonged dual antiplatelet therapy, Dr. Yeh and colleagues add.

A Modest Model

In a telephone interview with TCTMD, Sorin J. Brener, MD, of Weill Cornell Medical College (New York, NY), said the main limitation of the model is that there is no way of knowing why some patients received DES rather than BMS.

“These patients were not propensity matched,” he said. “The models are parsimonious, which means that they selected a number of variables but not all the variables. Granted, they are good variables, but without talking to the patient, it is impossible to know what the biases may have been. Maybe a patient was judged to be not very good at taking their medications and therefore they got BMS. These are things you just don’t know that could sway physicians to choose one stent over another.”

What is still needed, Dr. Brener added, are good models where as much information is taken from the patient as possible and put into the model to look for interaction trends. He acknowledged, however, that the predicted and observed rates of TVR in the new model were closely aligned, suggesting that the study authors likely captured the most important variables.

And yet, the c statistic was only 0.66, meaning that only 66% of the variability in TVR can be explained by the variables they chose and another 34% is unexplained, Dr. Brener concluded.

“That’s why it’s important to understand that while the model is well calibrated, the overall discrimination is modest,” he said.

 

Source:
Yeh RW, Normand S-LT, Wolf RE, et al. Predicting the restenosis benefit of drug-eluting versus bare metal stents in percutaneous coronary intervention. Circulation. 2011;Epub ahead of print.

 

 

Related Stories:

Disclosures
  • Dr. Yeh reports being an investigator for the Harvard Clinical Research Institute and a consultant for the Kaiser Permanente Division of Research.
  • Dr. Brener reports no relevant conflicts of interest.

Comments

DES/BRS/DCB