Government agencies have responded with tighter regulations, but investigators say the only real solution is to identify the most suitable candidates for opioid treatment: those patients most likely to experience effective analgesia with minimal adverse consequences.
In a recent paper, a panel of prominent pain researchers and clinicians outlines a research agenda for achieving personalized opioid prescribing. Central to that plan is a call for large, long-term observational studies aimed at finding associations between patient characteristics and treatment outcomes. In particular, the panel zeroes in on "practice-based evidence" (PBE), an approach that depends on registry systems that log comprehensive information about large numbers of patients over the course of routine care. After collecting an abundance of data, researchers can comb through the information for factors that associate with outcomes.
"There's a raging question: Who are the people who benefit from long-term opioid therapy?" said Charles Inturrisi, Weill Cornell Medical College, New York, US, who has launched a registry study that is seen as a pilot project for practice-based studies in pain. "Some people would argue, no one. Some would argue, lots of people. Does it make any difference whether you have a cancer pain diagnosis, or a rheumatoid diagnosis, or a spinal stenosis diagnosis?" These are all questions that researchers can begin to address with practice-based data, Inturrisi said.
The paper, written by Inturrisi and other US pain experts, stemmed from a September 2011 US National Institutes of Health (NIH) workshop, "Pathways Toward Evidence-Based, Personalized Analgesic Medication," organized by the National Institute on Drug Abuse (NIDA). The paper appeared in the February issue of the Journal of Pain.
In search of solutions
Evidence for harm from opioid use is accumulating. For example, researchers from the US Centers for Disease Control and Prevention reported recently that US drug overdose deaths increased for the eleventh consecutive year in 2010; roughly half involved prescription opioids (Jones et al., 2013). Furthermore, epidemiological studies of long-term opioid use have yielded "clear, overwhelming evidence of lack of safety: high death rates, high fracture rates, high rates of cognitive dysfunction, and high rates of abuse, particularly in patients who get to high doses," said Jane Ballantyne, University of Washington, Seattle, US, an author on the new paper. In addition, whether or not long-term opioids actually relieve chronic non-cancer pain is controversial, with numerous expert panels concluding that the existing evidence base is too weak to provide an answer (see Chou et al., 2009; Chapman et al., 2010;Reid et al., 2011).
Ballantyne said the existing data, and her own clinical experience, make it look unlikely that long-term opioid therapy is effective and safe. But she accepts that she may not have the whole picture—and she wants more information. "When we try to put controls on opioid use … physicians will stand up and say, 'I've got patients who are doing really well'" on long-term opioid therapy, she said. "But nobody knows who [those patients] are and how to identify them so that you don't start people on opioids who become the patients who do so badly in the long term."
And the need for personalized information is not limited to opioids, Inturrisi said. For many pain treatments—anticonvulsants, tricyclic antidepressants, and others—"they on average only work in about a third of patients. But we don't know which third."
Ballantyne and others say the best hope for obtaining that kind of information is through large observational studies of patients during routine practice—the PBE model.
The approach offers certain advantages over randomized controlled trials (RCTs). Two are time and size: RCTs are so expensive that they are generally small and short, enrolling a limited number of patients and ending in a matter of weeks. Registry studies can follow large numbers of patients for months or years, allowing a fuller look at long-term treatment in multiple patient subgroups.
Another advantage is that the PBE studies look at real-world patients, in all of their complexity. In an RCT of an analgesic, "you exclude everybody who has a history of drug abuse, everyone that has any kind of comorbidity, like [poor] renal function, a history of depression … I could go down the list," said Inturrisi.
A pilot project
In pursuit of personalized pain medicine, Inturrisi struck up a collaboration with Susan Horn, a statistician at the Institute for Clinical Outcomes Research in Salt Lake City, Utah, US, who had already used the PBE approach in other settings, including stroke and spinal cord injury rehabilitation (see Horn et al., 2012; Horn and Gassaway, 2010). But the model was new to pain.
In July 2011, with support from NIDA, Horn and Inturrisi launched the New York City Tri-Institutional Chronic Pain Registry. The registry enrolls all patients from two pain clinics at Memorial Sloan-Kettering Cancer Center; a clinic at New York-Presbyterian/Weill Cornell that sees patients with neuropathic and other forms of non-cancer pain; and the Hospital for Special Surgery (a hospital for orthopedic and rheumatoid conditions). "We have the whole spectrum of pain patients," Inturrisi said, "who have all the comorbidities that come along with being real-world patients."
When a patient comes to one of the clinics, the registry system "talks" to the patient's electronic medical record to collect demographic data like age and sex, diagnoses and comorbidities, and medical and social history. Then, the system tracks all of the treatments the patient receives—medications as well as surgical procedures, physical therapy, cognitive-behavioral therapy, or complementary and alternative therapies. And before every appointment, the patient uses an iPad to answer questions about outcomes including pain, function, adverse events, and red flags for opioid abuse.
The registry-based approach is "integrated into the actual practice of care," Horn said, since information is collected as doctors see patients during clinical care. That means she and her colleagues had to design systems in which physicians and patients can record information in ways that are practical for them and that also are useful for research. "The problem with most clinical information systems today is that, other than lab values and vital signs, almost everything else is text, which means that you have to read it manually," she said. With standardized formats for recording clinical information, researchers can easily capture and compare data among patients. "If you standardize the documentation," then these practice-based studies "can just flow right out," Horn said.
Thus far, the New York registry has enrolled almost 2,000 patients. Currently, all of them come from pain clinics at research hospitals. But Inturrisi wants the registry to capture a fuller range of patients by going into community-based clinics around the city, and eventually into primary care. "That's where 90 percent of pain is being managed," he said.
Moving forward
Inturrisi said that the PBE approach, which is limited in its ability to infer causality, is not a replacement for RCTs. "Ultimately, it will be important to validate [results from PBE studies] with the gold standard—the randomized controlled trial," he said.
But the PBE strategy could help "narrow down the universe of predictors" of opioid response, said Stephen Bruehl, Vanderbilt University School of Medicine, Nashville, Tennessee, US, lead author on the paper.
For instance, Inturrisi and his colleagues are seeking funding to perform genomic analyses on some of the patients in their registry to look for gene variants that associate with drug effects. Inturrisi also said metabolomics analyses might reveal differences in the way patients metabolize drugs and factors that predict drug response. And brain imaging, although too expensive to perform routinely, might help to characterize certain patient subsets and find patterns that predict opioid response, or that could serve as objective measures of analgesia.
"The cost of capturing these biomarkers will continue to drop, so we will be able to capture larger and richer datasets on our patients," said Sean Mackey, Stanford University School of Medicine, California, US, an author on the paper. Mackey is developing another registry platform for patient assessment, based largely on the NIH Patient Reported Outcomes Measurement Information System (PROMIS®), a standardized set of clinical questionnaires that assess patient health and well being. Mackey and his colleagues at Stanford rolled out their system to new patients in August 2012. Currently, the project's aim is to improve clinical care—but in the future, the system should also be useful for research, he said, and he intends to make the platform open source so that other investigators can implement it at their own sites.
For any pain registry study, developing suitable informatics tools is a major challenge, said Jörn Lötsch, Goethe Universität, Frankfurt am Main, Germany, who has studied genetic factors associated with opioid response but was not involved in the NIH workshop or paper. Lötsch's group and others are trying to develop the sophisticated tools needed to crunch the data. "What we get in a complex trait such as pain is high-dimensional data"—information on pain mechanisms, patient characteristics, treatment combinations, genetics, biomarkers, and more. "And we still have not developed the means to deal with high-dimensional data."
Will registry-based studies actually reshape opioid prescribing?
The goal of the PBE approach when it comes to opioids is to pinpoint the patients most likely to experience good pain relief with few side effects and a low risk of abuse. Will Inturrisi and others achieve that aim?
Preliminary data from the New York registry, Inturrisi said, suggest that some patients do benefit from long-term opioids. "The possibility that no one benefits from opioids—I would say that's a bit of a longshot," he said. For those deemed unlikely to do well, physicians could then consider other treatments.
Still, the predictors of individual success with opioid treatment are unknown—and likely numerous. "Is [the predictor] diagnostic code? Is it demographics? Is it concurrent treatments? My guess is it's going to be multifactorial, because I don't think there's a single factor that will predict a particular outcome for our patients," Inturrisi said.
REFERENCES:
Personalized medicine and opioid analgesic prescribing for chronic pain: opportunities and challenges.
Bruehl S, Apkarian VA, Ballantyne JC, Berger A, Borsook D, Chen WG, Farrar JT, Haythornthwaite JA, Horn SD, Iadarola MJ, Inturrisi CE, Lao L, Mackey S, Mao J, Sawczuk A, Uhl GR, Witter J, Woolf CJ, Zubieta J-K, Lin Y.
J Pain. 2013 Feb; 14(2):103-13.
http://www.painresearchforum.org/news/26207-path-personalized-pain-treatment
