Novel Biomarkers Predict Pain Sensitivity

A novel biomarker signature that assessed cortical activity predicted individual pain sensitivity, the PREDICT validation study showed.

The signature consisted of two measures: sensorimotor peak alpha frequency (PAF) and corticomotor excitability (CME). In the training set, it correctly classified study participants with high or low pain sensitivity with an area under the curve (AUC) of 1.00.

In the test set, the signature had an AUC of 0.88 (95% CI 0.78-0.99), reported David Seminowicz, PhD, of the University of Western Ontario in London, Canada, and co-authors in JAMA Neurology.

Results were reproduced across a range of parameters. The PAF and CME biomarkers showed good to excellent test-retest reliability.

“The combination of biomarker accuracy, reproducibility, reliability, and pain model validity suggests high potential for clinical translation, particularly in predicting the transition from acute to chronic pain,” Seminowicz and colleagues wrote.

PAF is the dominant sensorimotor cortical oscillation in the 8-12 Hz (alpha) range. CME is the efficacy of relaying signals from the primary motor cortex to peripheral muscles. Previous work showed that slower PAF before pain onset and reduced CME during prolonged pain were associated with more pain, while faster PAF and increased CME were associated with less pain.

“Given that individuals who experience higher pain in the early stages of a prolonged pain episode (e.g., postsurgery) are more likely to develop chronic pain in the future, slow PAF before an anticipated prolonged pain episode and/or CME depression during the acute stages of pain could be predictors for the transition to chronic pain,” the researchers noted.

Identifying objective biomarkers to track pain severity has been dubbed “the holy grail” of pain neuroscience, observed Prasad Shirvalkar, MD, PhD, of the University of California San Francisco, and Christopher Rozell, PhD, of the Georgia Institute of Technology in Atlanta, in an accompanying editorial.

“While pain is among the most fundamental, ubiquitous, and adaptive experiences that can befall an organism, there is still a murky understanding of how pain is generated in the nervous system,” they noted. The consensus on mechanisms underlying chronic pain — pain that persists for more than 3 months, which affects 21% of U.S. adults — is even less clear.

The PAF and CME signature “will likely have broad applicability across many medical fields,” Shirvalkar and Rozell said. “If successfully translated into clinical practice, biomarkers that predict a transition to chronic pain would have a tremendous impact for the treatment of millions of individuals.”

Advances in pain biomarkers also need to incorporate advances in global neuroethics guidance and address ethical concerns about pain treatment, the editorialists pointed out. “We must take care to ensure that quantitative measures do not supplant lived experience reports, introduce distrust in the physician-patient relationship, set unrealistic patient expectations, or exacerbate existing inequalities in pain treatment across this vulnerable population,” they wrote.

The PREDICT validation study included 150 people (100 in the training set, and 50 in the test set) who were given an injection of nerve growth factor into the right masseter muscle on day 0 and day 2 to induce prolonged jaw pain that lasted up to 4 weeks.

Participants were healthy adults recruited in Australia with a mean age of 25. They had no history of chronic pain or a neurological or psychiatric condition, and 84 participants (56%) were men.

The research aimed to determine whether individuals could be accurately classified as having high or low pain sensitivity based on baseline PAF and CME readings. The researchers used electroencephalography to assess PAF and transcranial magnetic stimulation with resulting evoked potentials to assess CME on day 0, day 2, and day 5.

The primary outcomes were jaw pain on chewing and yawning. Pain sensitivity was assessed twice daily from day 1 through day 30 through self-reported pain scores.

Seminowicz and colleagues used five machine learning models on the training set. Of these, the winning classifier was logistic regression. Including sex and pain catastrophizing as covariates did not improve model performance.

The study assessed healthy participants using an experimental pain model; results may not apply to other people or other circumstances, the researchers acknowledged.

Please enable JavaScript to view the comments powered by Disqus.