Chicago, IL—Oncology has been identified as a high-priority area for comparative effectiveness research (CER), given the accelerating pace of diagnostic and therapeutic interventions and the often high costs attached to these interventions. CER in oncology was the subject of a well-attended symposium at the 2010 ASCO meeting.
The goal of CER is to reflect care in the real-world setting, said Gary H. Lyman, MD, professor of medicine and director of health services, effectiveness, and outcomes research in oncology, Duke Comprehensive Cancer Center, Durham, NC. This means going beyond the confines of study populations to the broader context of real-life clinical practice.
The toolbox for comparative effectiveness contains randomized controlled trials (RCTs), which are considered the gold standard of CER, as well as systematic reviews and meta-analyses of RCTs. However, because most clinical questions in oncology have not been the subject of RCTs, the toolbox also includes cohort studies; population studies; prognostic and predictive studies; and modeling approaches.
Even RCTs have limitations for use in comparative effectiveness, Dr Lyman said, because they are often short term and use surrogate efficacy outcomes; they have a limited sample size; a duration of follow-up too short to address important safety and toxicity questions; and they frequently exclude patients with major but common comorbidities.
Clinical decision models (eg, Markov models) are increasingly being used as reasonable estimates of comparative effectiveness, he said. Such modeling studies have the advantage of comparing efficacy over a range of assumptions and can provide measures of incremental effectiveness that reflect quality of life and costs.
Imaging in cancer is in the top quartile of the 100 Institute of Medicine priority topics for CER. The use of diagnostic imaging in cancer care has increased by 40% to 50% per year since 1999, said Neal J. Meropol, MD, chief, division of hematology/oncology, University Hospitals Case Medical Center and Case Western Reserve University, Cleveland, OH.
Other priority needs for CER in cancer screening and prevention are molecular risk stratification, prescreening molecular tests, risk modifiers, and the optimal frequency of diagnostic testing. Gene profiling also has taken center stage of late and represents another area that is ripe for CER.
“What we need are models that integrate clinical and molecular risk assessment,” said Dr Meropol.
Treatment must include more research on adjuvant therapy decisions, selection of treatment intensity, and better knowledge on the best time to change treatments. Obtaining support for many of the treatment studies “is difficult when competing treatments are marketed by different entities,” he said, calling for more publicly funded phase 3 trials in cancer research.
Katrina Armstrong, MD, associate director, outcomes and delivery research, Abramson Cancer Center, University of Pennsylvania, Phila delphia, elaborated on alternatives to RCTs for conducting CER, focusing on observational studies with the use of propensity scoring or instrumental variable analysis, and evidence synthesis using decision analysis.
Propensity scoring is a method to adjust for differences between treatment groups, and as such “can address 90% of the bias in observational data,” she said. One example of using propensity scores to adjust for potential confounders in observational studies is in the treatment of localized prostate cancer, in which Wong and colleagues determined that men aged 65 to 80 years experienced a survival advantage with treatment versus observation.1
Instrumental variable analysis adjusts for differences in groups when relevant differences between groups are not available, said Dr Armstrong. One such example is an analysis that simulated the conditions of a randomized trial of chemotherapy for advanced lung cancer in patients older than 65 years.2 The analysis showed that chemotherapy in this population led to a median survival increase of 33 days and a 1-year survival rate of 9%.
Decision analysis is a quantitative approach to synthesizing evidence to compare expected outcomes of different interventions. She cited her own study, a Markov decision-analytic model, to determine that women with BRCA1/2 mutations should make decisions about the use of hormone replacement therapy after prophylactic oophorectomy based on quality-of-life issues rather than life expectancy.3
“Cost is a growing focus of comparative effectiveness research,” Dr Armstrong said, adding that it is increasingly being incorporated into Markov models. A recent example is a Markov state- transition model aimed at determining how fracture risk should be managed in men receiving androgen-deprivation therapy for the treatment of localized prostate cancer.4 The authors concluded that in men at higher risk for hip fracture, routine use of alendronate without a bone mineral density test is a cost-effective strategy.
As more targeted therapies for cancer are developed, data collection will have to change to include the presence of biological markers to conduct CER of these therapies, Dr Armstrong said.
