Improving cancer outcomes: What’s your policy?
What do clinical investigators, statisticians, health care economists, and health services researchers have in common? They collaborate on outcomes research, says John Ayanian, MD, MPP (HMS/BWH), leader of the Outcomes Research Program at DF/HCC. As cancer is detected earlier and patients live longer, outcome studies – which require the broad expertise of a multidisciplinary team – play an increasingly crucial role in understanding the long-term impact of interventions to prevent and treat cancer.
Expanding traditional study endpoints, outcomes research examines the quality of care, quality of life, effectiveness, and cost-effectiveness of interventions in routine care settings, while aiming to translate findings into better health care policy. The observational studies, randomized trials, and data modeling featured in this story are just three of the rigorous approaches DF/HCC investigators are taking to extend the scope of research across time and populations to improve cancer outcomes.
Does cost-sharing reduce mammography?
Good question. To find an answer, Ayanian and colleagues conducted an observational study, published in the New England Journal of Medicine, which analyzed data on 366,475 women, aged 65 to 69, who were enrolled in Medicare plans from 2001 through 2004. Investigators examined whether women whose plans required copayments were as likely to participate in screening mammography as were women whose plans did not. Although copayments were only $20 on average, says Ayanian, “We found that screening rates were about 8 percent lower among women required to make a copayment, and that this negative effect was greater for women with lower incomes or education levels.” Researchers also evaluated a subset of plans that did not require copayments at the beginning of the study but did so later on – a more powerful analysis, says Ayanian, professor of medicine and health care policy at HMS and BWH. “When we compared those plans in the same regions of the country, we saw that screening rates went down by 5.5 percent in plans that added copayments, whereas in matched plans in which mammography remained fully covered, rates went up by 3.4 percent – a relative change of 8.9 percent.” The results suggest that exempting elderly women from copayments for mammography will increase screening rates in this group, concludes Ayanian.
In search of best practices
In a randomized trial, Ayanian collaborated with Thomas Sequist, assistant professor of medicine and health care policy at HMS and BWH, to evaluate whether direct-to-patient reminders or electronic alerts to physicians increased screening rates for colorectal cancer (CRC). “We asked, which is the best method that a large group practice can implement to improve CRC screening,” says Sequist, who is also a primary care physician at Harvard Vanguard Medical Associates (HVMA). Altogether, 21,680 patients aged 50 to 80 who were overdue for CRC screening and 110 HVMA physicians responsible for their care participated in the study (Sequist did not). Patients were randomly assigned to receive mailings with reminders to return the fecal-occult blood test or to schedule a sigmoidoscopy or colonoscopy; physicians were randomly assigned to receive an electronic alert when a visiting patient was overdue for screening. “We’re not sure whether physicians ignored the alerts,” says Sequist, or simply had more urgent issues to discuss with the patient. Whatever the reason, physician reminders had no impact on screening rates whereas patient reminders did, especially among the 70- to 80-year age group, says Sequist, “where we saw a 10 percent absolute increase in screening.”
Since patient mailings were by far the stronger intervention, says Sequist, Harvard Vanguard has now adopted mailings as official practice for CRC screening − and turned off its physician alerts.
Looking at long-term complications
Outcomes research helps inform clinical practice in other important ways. In his work as a prostate cancer physician-scientist, Matthew Smith, MD, PhD (MGH) had observed that men treated with gonadotropin-releasing hormone (GnRH) agonists developed insulin resistance and increased fat mass, among other effects; he wondered whether these complications could later lead to diabetes and cardiovascular disease. (Smith’s work on androgen deprivation therapy and osteoporosis was featured in the Fall 2008 issue of eNews.) To answer that question, physician and health services researcher Nancy Keating, MD, MPH, associate professor of medicine and health care policy at HMS, collaborated with Smith and statistician James O’Malley (HMS) on an observational study. Investigators analyzed a cohort of 73,196 Medicare recipients, aged 66 or older, who had been diagnosed with non-metastatic prostate cancer; these data derived from the Surveillance, Epidemiology, and End Results (SEER) program linked with Medicare data about utilization of services.
“We found that men who were treated with GnRH agonists had a significantly increased risk of developing incident diabetes, coronary heart disease, heart attack, or sudden death,” says Keating, who cautions that these associations are not necessarily causal, but raise questions about the use of GnRH agonists. “These drugs are often given in settings where they have not been shown to decrease death from prostate cancer,” she explains. “The fact that we could be doing potential harm using these drugs is of concern.”
Keating is now trying to replicate these findings in a different population: younger men in the Veterans Affairs healthcare system. The new study of 38,000 men also aims to identify those at higher risk of developing diabetes or cardiovascular disease. “Clinicians can then use such information in making decisions about treatment,” suggests Keating, “perhaps prescribing these drugs less often in patients at higher risk for adverse outcomes, or more closely monitoring for these outcomes in men who require GnRH agonist therapy.”
Not a parlor game
Though the Lung Cancer Policy Model can predict outcomes, it’s not a parlor game, observes Jane Weeks, MD, co-leader of the Outcomes Research Program and chief of the Division of Population Science at DFCI. (See companion story about Weeks’ work in CanCORS.) Indeed, as Scott Gazelle, MD, MPH, PhD (MGH) explains, the model simulates individual lung cancer patients using real-world data about the likelihood of certain events occurring over time − the tumor will be malignant or benign, diagnosed early or not, treated or untreated, and so forth. “We use data from countless sources to estimate probabilities based on what we know of lung cancer to project outcomes over a lifetime,” says Gazelle, a radiologist and director of the Institute for Technology Assessment at MGH. Moreover, the Monte Carlo-based model can simulate 10,000 patients − or 10 million.
Recently, Gazelle and colleagues used the model to evaluate the effectiveness of lung cancer screening, based on data from the Mayo Clinic’s single-arm study of helical CT screening in 1,500 patients. “First we simulated the Mayo study and tweaked our model until we replicated their results,” explains Gazelle. Investigators then simulated a control arm and compared it to the screening arm: At six-year follow up, the screening arm showed a 37 percent relative increase in lung cancer detection – results that the Mayo study had not been designed or powered to answer. “Clinical trials are absolutely necessary,” says Gazelle, “but no one trial can answer every question.” That’s where the microsimulation model can help, by putting clinical trials into perspective and exploring what-if questions about health care outcomes and costs.
What if, for example, lung cancer screening had been clinical practice from 1990 through 2005? Would it have saved lives? That’s the question Gazelle posed in another study, tailor-made for microsimulation since a control arm already existed (patients were not screened for lung cancer during that time period). The results of the study showed a modest decrease in lung cancer mortality, says Gazelle, who is now investigating the cost-effectiveness of lung cancer screening.
Does he expect to make policy recommendations based on the results? “Absolutely,” he replies. “We always conduct our research to inform health care policy.”