In the April 11 New England Journal of Medicine, researchers report an advance toward the goal of an objective measure of pain. A team led by Tor Wager, University of Colorado, Boulder, US, describes a new functional MRI-based test that is more accurate and specific for physical pain than any previously described method.
The measure, which the investigators call a “neurologic signature of pain,” relies on a computer algorithm they developed to recognize a complex pattern of neuronal activation that occurred in the brains of study subjects exposed to painful heat.
The study represents “a further step towards a biomarker of pain,” said Markus Ploner of the Technische Universität München, Germany, a neurologist who studies brain mechanisms of pain but was not involved in the new work. “It’s not the first step, and for sure not the last step, but a further step.”
The work is a significant advance, but the field is still “years to decades” away from the use of imaging biomarkers for pain in the clinic, Ploner said. Most importantly, the new study looked at a short-lived, defined heat pain stimulus in healthy volunteers, which is very different from pain in the context of disease, including ongoing chronic pain.
Locating pain in the brain
The gold standard for measuring pain in the clinic is self-report—clinicians ask and patients tell. The search for objective measures of pain—telltale physiological responses that signal a person’s pain—has gained steam with the use of functional brain imaging, which uses blood oxygen level-dependent (BOLD) signal changes to pinpoint areas of the brain that become activated when people experience pain.
Such studies have revealed that painful stimuli evoke activity in many brain regions, and it has become clear that no one area of the brain is responsible for pain. In the last several years, multiple groups have used sophisticated machine-learning techniques to analyze whole-brain or regional fMRI data to identify complex patterns of BOLD signal changes that seem to underlie pain in experimental settings (Marquand et al., 2010; Brown et al., 2011; Brodersen et al., 2012; Cecchi et al., 2012). But accuracy and specificity have been a problem.
To improve both, Wager and colleagues focused on the brain regions most implicated in pain, based on a meta-analysis of 6,000 published MRI studies on pain and other conditions (the NeuroSynth database; see PRF related news story). First, they measured BOLD signal changes in 20 healthy young adults during the application of painful or non-painful heat to the forearm. Using the readings from the brain volumes (voxels) in pain-related areas, they employed a machine-learning algorithm to define a pattern of brain activity that correlated with the subjects’ pain scores.
Next, the researchers asked whether the signature could accurately predict pain in a new group of 33 people tested in a different scanner. In the new subjects, the signature distinguished painful heat from innocuous warmth with 93 percent accuracy. In addition, the response was graded: The magnitude of the response correlated with temperature and with self-reported pain intensity. The signature was specific for actual pain; it was not elicited during pain anticipation or pain recall in the subjects.
The team also tested the specificity to physical pain versus the feeling of social rejection (“social pain”). Wager’s group had previously shown that asking subjects to look at pictures of an ex-partner after an unwanted romantic breakup activates many of the same brain regions as physical pain (Kross et al., 2011). In a new experiment with 40 such subjects, the signature response correctly distinguished 95 percent of the time whether subjects were looking at a picture of their ex-partner or feeling painful heat.
In a test of clinical relevance, the investigators showed that the signature response decreased reliably after subjects were treated with the potent opioid analgesic remifentanil.
Interestingly, the signature response decreased even when the subjects did not know they were receiving the drug. Because hidden dosing usually diminishes the efficacy of the opioid (e.g., see PRF related news story), this result indicates that the signature response may not always track with self-reported pain.
In fact, that dissociation may lead to new insights into mechanisms of pain and analgesia. Wager said that his group is following up on the remifentanil work by testing the effects of other treatments, including psychological manipulations such as distraction, mindful acceptance, and placebo on the signature response. “If some kinds of treatments are influencing this pain signature, and others are influencing other systems, that’s very exciting.” And being able to compare responses in different people “will help us to take steps towards being able to personalize treatments,” he said.
Just a start
While the study is, in principle, a step forward for clinical use of imaging in pain, Ploner said, the practical applications are far from certain. “What this study and all the studies before have done is to measure brief, experimental painful stimuli,” he explained. But that is a simplistic model compared to clinical pain. “The experience [of medical pain] is very different, and the brain mechanisms and maybe the brain signature of ongoing chronic pain may be very different.”
Wager agreed, and said his group is now starting to look for brain activity patterns that characterize different kinds of clinical pain.
Karen Davis, University of Toronto, Ontario, Canada, who studies pain using brain imaging but was not involved in the new study, is skeptical about whether brain imaging will ever be able to report accurately on chronic pain in an individual patient. Brain imaging results are “extremely unreliable at a single-subject level, and are confounded by sex and age effects, important factors that co-vary with pain,” she said. “To be able to put people in the magnet, and give them some stimuli and try to predict whether they’re feeling pain might work. But that’s a different ballgame from saying somebody has chronic pain.”
The new study, published in a high-profile medical journal, is sure to feed a growing debate over the potential clinical uses, and possible misuses, of imaging biomarkers for pain. Besides raising technical issues, Davis questioned the rationale behind the hunt for a brain biomarker for the experience of pain. She said biomarkers that give doctors insight into neurological pathology, or suggest which treatments are likely to work in individual patents, or predict side effects, will be useful. But, she said, “Pain by definition is a subjective experience, and so the self-report of the pain experience is what tells us if someone is in pain, and should not and need not be replaced by an imaging measure."
Wager emphasized that brain imaging should never be used to disprove that a person is in pain. “It’s not to be used as a pain lie detector, or some way of disbelieving people,” he said. If a group of patients in pain had brain activity that did not match any known pain-related pattern, “would we then say they must not be feeling real pain? No. We’d say they’re experiencing real pain … driven by another system.” Instead, he said, brain imaging should only be used to help confirm a patient’s pain and to understand it better. “I think we need to find a way of respecting pain reports, while also bringing in these other measures that are coming down the line and could potentially really help people.”
The latest in the debate over imaging biomarkers for pain is laid out by Davis and others in six commentaries and responses in the April Journal of Pain (Sullivan et al., 2013; Robinson et al., 2013; Mackey, 2013; Davis, 2013; Robinson et al., 2013; Sullivan et al., 2013).
For more perspective on the Wager study, see the New England Journal of Medicine editorial that appeared along with the paper (Jaillard and Ropper, 2013).
Top image credit: Tor Wager.