November 2020 Edition Vol.12, Issue 11

Improving Cancer Care Delivery Using Real-World Designs

By Lynne Lederman, PhD

The final session of the virtual 2020 ASCO Quality Care Symposium, Improving Cancer Care Delivery Using Real-World Designs, was chaired and moderated by Hanna Kelly Sanoff, MD, MPH, University of North Carolina Lineberger Comprehensive Cancer Center.

How Pragmatic Trials Can Inform Practice Change

Deborah Schrag, MD, MPH, Dana-Farber Cancer Institute, provided an overview of pragmatic clinical trials (PCT). She then presented a specific example, the CANVAS trial, a PCT designed to determine the best strategy for anticoagulation to prevent venous thromboembolic events (VTE) in patients with cancer who experienced a VTE within the past 30 days, and for whom anticoagulation for at least 3 months was indicated.

PCTs are designed for the primary purpose of informing decision makers about the comparative balance of benefits, burdens, and risks of a biomedical or behavioral health intervention at the individual or population level. Dr. Schrag said this is a hallmark of much of the research funded by the Patient-Centered Outcomes Research Institute (PCORI).

The four key components of PCT are 1) intent to inform decision makers as opposed to elucidating a fundamental, biological, social, or behavioral mechanism; 2) intent to enroll a population relevant to a real decision encountered in practice, and a decision that represents the needs of patients for whom particular clinical decisions are relevant; 3) intent to streamline procedures and data collection, so trials can focus on gaining adequate power, speed, and representation to inform clinical and policy decisions; and 4) focus on a range of outcomes that are easily observable and salient to decision makers.

Whereas randomized clinical trials (RCTs) ask whether interventions work under ideal circumstances, PCTs ask if interventions actually work under typical circumstances. These contrast with quality improvement initiatives that focus on how to make the intervention work best for patients in routine care.

In CANVAS, patients were randomly assigned to direct oral anticoagulants (DOAC) (intervention) or to low molecular weight heparin (LMWH) with or without transition to warfarin (usual care). Outcomes included cumulative incidence of recurrent VTE, major bleeding, and survival at six months; and anticoagulation burden and health-related quality of life at three and six months. Funded by PCORI, the study has accrued 811 patients, with results expected in January.

Pragmatic features of CANVAS include streamlined data collection and reporting, no specified follow-up visits and intervals, no required follow-up evaluations such as scans, no required lab tests, no biospecimens, and flexibility regarding the intervention itself, i.e., physician choice of agent within treatment class. Patients assigned to a DOAC can switch to another DOAC if necessary, e.g., if prescription drug plans change. In the other arm, switching between LMWH and warfarin was allowed. Anticoagulation breaks and gaps, e.g., to undergo dental work, were not considered to be protocol violations, which they would in a traditional RCT.

Dr. Schrag said PCTs are RCTs “with features that facilitate conduct and implementation in diverse settings that do not require special resources. These studies can integrate with routine workflow relatively easily with relatively easy observable endpoints, and often no special visits, evaluations, or need for expert research personnel.”

Dr. Schrag concluded that PCTs “increase the generalizability of the research that we do in oncology. They engage more participants and they reflect the full spectrum of oncology care.” PCTs can be executed quickly and efficiently, so that research is nimbler and has broader scope, and that benefits patients.

Real-World Evidence and Practice Change

George J. Chang, MD, University of Texas MD Anderson Cancer Center, Houston, Texas, spoke about the role real-world evidence (RWE) could have in informing clinical practice. RWE can be generated from clinical medical records, claims, government data sets, disease, device, and patient registries, and patient-sourced biometric survey, social media, and populational data.

Obtaining evidence using RCT is slow and costly, and may have limited generalizability due to strict eligibility criteria, although there is a high degree of internal validity. There are also many clinical situations or conditions for which it may not be possible to generate clinical trial evidence.

RWE studies are efficient, applicable to a broader array of situations, generalizable, and benefit from information diversity. However, there is great risk for internal validity, as the data have not been collected for research. RWE is not intended to replace evidence from RCTs. However, RWE can inform clinical practice through hypothesis generation, and confirm the generalizability of RCT findings. RWE can be used for post-marketing surveillance and safety studies, assessment of disease burden in routine practice, or for studies of patterns of care, outcomes research, and systems and quality improvement studies.

RWE from population-based cancer registries has been used to show that Medicaid expansion has reduced disparities in cancer care associated with low income or rural location. RWE has confirmed findings from RCTs, such as the treatment benefit associated with adjuvant chemotherapy in lung or colorectal cancer and has failed to confirm findings from an RCT of sorafenib in hepatocellular carcinoma. RWE can be used to determine if the criteria in RCTs were so rigorous that the findings are not generalizable to routine practice, as well as augment findings from RCT that are underpowered in certain settings, e.g., in rare diseases. RWE can also provide evidence when RCTs are not available or cannot be performed.

The annual number of citations of RWE in PubMed has increased exponentially over the last five decades. Dr. Chang said that in reviewing these studies, it is important to consider the methods used, because they are subject to biases threatening internal validity, including the accuracy of data collection, confounding patient characteristics and comorbidities, non-adherence to treatment, and selection bias.

He said, “In assessing real-world evidence, we have to first start with the sniff test. Do the results make sense? There are several studies with conflicting results or implausibly large effect sizes of treatment that are clearly due to bias or confounding.”

Analytic methods need to be rigorous with multivariate adjustment, although this applies only to measurable variables. Other ways to reduce bias include the use of propensity scores to estimate the probability that a patient would have been assigned to treatment, and instrumental variable analysis, where the instrument is related to the outcome only through its relationship with the treatment of interest.

Among the examples Dr. Chang presented was a study that examined the observed survival benefit associated with aggressive local therapy for prostate cancer. Using propensity-score-adjusted methods, the investigators identified that aggressive local therapy was really associated with an improvement in non-cancer mortality as well as cancer-associated mortality, indicating that it was the selection of patients for that aggressive local therapy that was likely attributable to the large survival benefits seen in the overall cohorts.

Many studies had shown that primary tumor resection was associated with a dramatic improvement in survival for patients with metastatic colorectal cancer (CRC). Dr. Chang’s group showed in one RWE study that patients in the non-surgery group experienced a much higher rate of early mortality, likely due to underlying factors such as tumor burden or comorbidity. This early mortality accounted for the entire apparent benefit of primary tumor resection which in reality provides no survival benefit.

Although some RWE studies are unlikely to affect practice guidelines, results of some RCTs have confirmed findings previously generated from RWE, e.g., the JCOG1007 RCT confirmed that adding primary tumor resection to chemotherapy for stage IV CRC conferred no survival benefit. Likewise, minimally invasive surgery for cervical cancer was associated with worse outcomes than open surgery using RWE, which was also then confirmed by a RCT.

Dr. Chang asked if these RCTs were necessary in order to change practice or inform practice guidelines. “It seems that the bar is exceptionally high for real-world evidence to inform practice change,” he said. “One can ask, when is real-world evidence most believable? Is it when it confirms or doesn’t affirm a treatment benefit? Or is it when it fails to demonstrate treatment benefit? Or is it when it demonstrates treatment harm or it shows non-inferiority?”

Dr. Chang concluded that RWE could be used to inform clinical practice and treatment guidelines provided threats to internal validity are addressed. “Real-world evidence should be held to a high standard for these rigorous methodologies, but when these methodologies are met, they should support practice change.”

Linkage is Key

During the discussion, a participant asked about opportunities to leverage multiple data sources, such as the EHR, claims, patient-reported outcomes, and trial data, to overcome limitations. Dr. Schrag said, “For us, the word is linkage. There is incredible power through data linkage.” This doesn’t necessarily address problems with missing data, but it can expand the scope and kinds of questions that can be asked and add nuance.

Dr. Schrag said the help of the National Cancer Institute was needed to link data. “We need more data linkages. It drives us all bananas that we don’t have linkages for every single cancer registry. We all use SEER-Medicare, but that’s just one measly data set. We need much larger, more powerful data sets, and we could all answer questions a lot faster.” This, she said, requires some government coordination and help to help make those data sets available.

Post a Comment

OBR Archives

To view previous issues of OBR green you can visit our archives. The entire library of OBR green articles is searchable.