Randomized clinical trials have been providing high-quality evidence for decades, but there are limitations to the traditional design. At the 2018 ASCO Quality Care Symposium, George J. Chang, MD, MS, FACS, FASCRS, discussed the need to modernize clinical trials, so they continue to provide clinicians with the highest value intervention for quality improvement.1
“The biggest challenge in terms of quality improvement in cancer care is taking evidence that’s generated in the tightly controlled settings of randomized trials and translating that into clinical practice,” said Dr. Chang, of the Department of Surgical Oncology at The University of Texas MD Anderson Cancer Center. “Modern clinical trials engage stakeholders at the front lines of care. They leverage an existing infrastructure and data systems, so they’re quite practical.”
The biggest challenge in terms of quality improvement in cancer care is taking evidence that’s generated in sometimes rarefied situations of randomized trials and translating that into clinical practice.— George J. Chang, MD, MS, FACS, FASCRS
Tweet this quote
As Dr. Chang explained, many traditional randomized trials tested patients under idealized conditions and had strict inclusion criteria, which might not match the patients seen by clinicians on a regular basis. In other words, it was research focused—conducted by investigators in a research environment with a lot of research infrastructure. The goal of the modern clinical trial design, on the other hand, is pragmatism.
“So-called pragmatic trials are studies of effectiveness under real-world conditions, where we now have an opportunity to address diverse patient populations with broad inclusion criteria,” said Dr. Chang. “The modern trial leverages an existing infrastructure, particularly a data infrastructure, so we can monitor the outcomes and collect data that don’t require such a complex research infrastructure. It’s not only research-focused but clinician-focused.”
According to Dr. Chang, these studies offer greater flexibility at a lower cost and can provide a better patient experience because they are performed at multiple levels within an organization. And because these studies are being conducted at the point of care, they can more proximally deliver evidence to practice. The two most common pragmatic designs, said Dr. Chang, are the cluster randomized trial and the step-wedge design.
Cluster Randomized Trials and Stepped-Wedge Design
“Cluster randomized trials are particularly useful for quality-improvement interventions because the intervention can be targeted not at patients but at providers, clinics, and hospitals,” he explained. “These trials will randomize social units or clusters of individuals—hospitals, providers, practices, even entire communities—and in so doing can avoid contamination of the intervention from the control to the intervention group.”
One such unit of randomization could be a medical practice to evaluate the efficacy of practice-level improvement programs. Another unit of randomization could be hospitals to evaluate system-level interventions that are targeting multiple levels of the medical staff and administrators.
“With cluster randomized trials, quality improvement is not isolated at the level of a single point in the care continuum,” said Dr. Chang. “It’s not simply focused on patients or on providers but often requires the engagement of multiple levels.”
Cluster randomized trials are particularly useful for quality-improvement interventions because the intervention can be targeted not at patients but at providers, clinics, and hospitals.— George J. Chang, MD, MS, FACS, FASCRS
Tweet this quote
Furthermore, the unit of analysis might not be the same as the unit of randomization. The data generated could be analyzed at the patient or hospital level or at a different level and may differ from the unit of inference as well. There’s a lot of flexibility in the design, Dr. Chang explained.
One example of a cluster randomized trial shared by Dr. Chang studied management of pain among inpatients at cancer centers. After data were collected, the centers were randomized to a pain management intervention.
“This was a facility-level randomization, with the primary outcome being a change in the percent of patients with significant improvement in pain,” said Dr. Chang. “This is a pragmatic way of studying the difficult and challenging issue of pain control in the inpatient setting.”
In a separate cluster randomized trial, investigators looked at ways to decrease the risk of falling among elderly patients in nursing homes based on preliminary evidence that improving communication among nursing home providers can improve patient outcomes. When nursing home employees were randomized to receive detailing for improving communication or not, however, no improvement was observed in either fall risk or counseling about fall risk.
“Clearly, this study highlights the fact that we need to be more certain about what interventions truly affect outcomes of interest before we embark on a whole-scale effort to change the way we practice,” stated Dr. Chang.
According to Dr. Chang, the second most common pragmatic design, the stepped-wedge trial, reconciles the need for a rigorous study within logistic constraints. The stepped-wedge design involves the random, sequential crossover of patient clusters from a control group to an intervention until all patients have received the intervention. These trials, which are being utilized with greater frequency, are ideal for situations in which all organizations should have an opportunity to use a new intervention.
“For complex interventions where large-scale, immediate rollout is not feasible, the stepped-wedge design is -effective because all clusters will eventually be exposed to the intervention, but they also serve as their own controls,” said Dr. Chang.
Other Approaches to Quality Improvement
Another approach to quality improvement involves incident learning systems. “As a surgeon, we do this every week at our morbidity and mortality conference; we identify a bad outcome that’s happened and then try to figure out where things went wrong,” revealed Dr. Chang, who noted, however, that the same complication may occur again the following week “unless we are able to implement effective processes to prevent that outcome from happening.”
In addition, because incident learning relies on voluntary reporting, Dr. Chang acknowledged a significant risk for bias. “In many of these systems, there is poor participation by physicians, and because it supplies only the numerator, it is difficult for us to have insight into the true impact on the organization,” he observed.
For more from the ASCO Quality Care Symposium, see an interview with Monika K. Krzyzanowska, MD, MPH, on avoiding pitfalls in quality improvement on The ASCO Post Newsreels at
Finally, quality measurements offer another approach to quality improvement, said Dr. Chang, but they also present a number of challenges, including extensive resource utilization. The Merit-based Incentive Payment System (MIPS), for example, one of two tracks under Medicare’s Quality Payment Program, is estimated to cost $1.3 billion. There are also ways to “game the system.” In addition, what we measure does not always achieve our intended quality-improvement objective. One study looking at compliance with Surgical Care Improvement Project measures and outcomes showed improvement in measured compliance over time without the desired effect on outcomes.
“Our goal is to be pragmatic,” Dr. Chang concluded. “Ultimately, clinical trials are about data-guided quality improvement. Modern trials should be conducted at the point of care, where the evidence is to be applied. It still remains the gold standard.” ■
DISCLOSURE: Dr. Chang has served as a consultant or advisor to Johnson & Johnson and More Health.
1. Chang GJ: Clinical trials: Struggling to stay relevant? 2018 ASCO Quality Care Symposium. Welcome and General Session 1. Presented September 28, 2018.