This article is more than 5 years old. Information may no longer be current.
Mortality data vary widely between randomized trials, observational cohorts
Click Here to Manage Email Alerts
Data regarding the effect of radiotherapy on mortality in patients with breast cancer diverged widely between randomized clinical trials and observational data, even after adjustment for potential confounding factors, according to the results of a meta-analysis.
Nonrandomized comparisons appeared to provide misleading estimates of treatment effects, results showed.
Sarah C. Darby
“The increasing complexity and cost of conducting randomized clinical trials, together with financial pressures in both North America and Europe, have stimulated interest in using observational data sets to determine the impact of cancer treatments,” Sarah C. Darby, PhD, professor of medical statistics at University of Oxford, and colleagues, wrote. “There is little doubt that observational data may be valuable in characterizing treatment outcomes in complete populations of cancer patients, rather than just in patients selected to take part in randomized trials. However, determining the causal effect of treatments from observational studies has long been recognized as challenging.”
Thus, Darby and colleagues sought to compare estimates on the effect of breast radiotherapy from observational data with findings from randomized trials.
To do so, they collected data from 13,932 women randomly assigned to radiotherapy or not in trials recently used in meta-analyses by the Early Breast Cancer Trialists’ Collaborative Group (EBCTCG). The analysis also included data from 393,840 women registered with breast cancer in the U.S. SEER registries between 1973 and 2008.
Among patients enrolled in randomized trials, radiotherapy after breast-conserving surgery appeared associated with reduced mortality from breast cancer (rate ratio [RR] = 0.82; 95% CI, 0.75-0.9) and all causes (RR = 0.92; 95% CI, 0.86-0.99).
Similar reductions in mortality occurred among women assigned to radiotherapy following mastectomy for node-positive breast cancer (breast cancer mortality, RR = 0.84; 95% CI, 0.76-0.94; all-cause mortality, RR = 0.89; 95% CI, 0.81-0.97).
In contrast, the researchers found that the effect of radiotherapy on mortality varied widely in the observational datasets.
These data showed an association between radiotherapy after breast-conserving surgery and a large reduction — greater than that seen in the randomized data — in mortality (breast cancer mortality, RR = 0.64; 95% CI, 0.62-0.66; all-cause mortality, RR = 0.63; 95% CI, 0.62-0.65).
However, radiotherapy after mastectomy appeared to increase mortality (breast cancer mortality, RR = 1.34; 95% CI, 1.31-1.37; all-cause mortality, RR = 1.23; 95% CI, 1.22-1.25). Researchers noted this effect strongly differed from the reduction in breast cancer mortality observed in the SEER data among women who underwent breast-conserving surgery, as well as the effect observed among women from randomized trials who underwent mastectomy and axillary dissection.
The researchers reported that adjustments for potential confounding factors did not reduce the deviation compared with the randomized data.
“In light of the findings presented here, the authors advocate increased resources to conduct randomized trials, particularly in diseases where evidence is scant, and to allow extended follow-up in randomized trials,” Darby and colleagues wrote. “Although observational analyses can complement our understanding of the comparative effectiveness of treatments in real-world settings, it remains imperative to increase the body of evidence regarding treatment efficacy through randomized trials.”
However, observational studies can still provide vital data for practice management, despite inherent limitations, Mariana Chavez-MacGregor, MD, and Sharon H. Giordano, MD, MPH, both of the departments of breast medical oncology and health services research at The University of Texas MD Anderson Cancer Center, wrote in an accompanying editorial.
“Although we must recognize that observational studies are also prone to limitations associated with misclassification, underreporting, or ascertainment, criteria have been established to design and to critically appraise the quality of observational data,” Chavez-MacGregor and Giordano wrote. “Observational studies, despite their limitations, can be a powerful resource for comparative effectiveness research and can help answer relevant questions. ... While we will continue to obtain high-quality data from randomized controlled trials that will determine and guide our treatments, we believe there is — and will continue to be — a place for population-based studies. In the era of comparative effectiveness, there should be no battle between randomized controlled trials and observational studies; both designs can provide valid and important knowledge.” – by Cameron Kelsall
Disclosure: Darby reports no relevant financial disclosures. One study researcher reports institutional research funding from AbbVie, honoraria from International Journal of Radiation Oncology Biology Physics and a consultant role with Eviti. Chavez-MacGregor reports travel expenses from and consultant roles with Genentech/Roche, Genomic Health, InVitae, Pfizer and Novartis, as well as institutional research funding from Novartis. Giordano reports no relevant financial disclosures.
Click Here to Manage Email Alerts