Indulge me for a moment. Let’s say you just arrived at your physician’s office with a troubling symptom. She says “Hold on, I need to put you in the MRI to see if this symptom is pain, or if you are a pain patient.”
There have been a number of scientific papers and popular press releases that suggest we need to replace self-report of pain because it is flawed, and we need to employ brain imaging as a “biomarker” or “objective” measure of pain.
The basic assumption is that self-report is “unreliable” or otherwise flawed. That’s a claim often made in these papers despite a large body of evidence that self-report of pain is 1) quite reliable, 2) sensitive to treatment change, and 3) predictable. If it were unreliable (full of error), it could not be predicted, and it would not be sensitive to any sort of treatment effect. As we know from our own experience, pain varies. Variability is not necessarily error. And finally, brain imaging of pain was originally, and continues to be, validated by correlations with…. wait…. SELF–REPORT!!
Whenever I bring this up, someone is quick to point out the exception in one of their patients, or relatives, or friend’s pain report that didn’t make sense, or they had trouble using the scale. Of course, there are exceptions, but the empirical data are quite clear. Self-report works. There is not a compelling reason to replace it.
Back to the doc’s office: Didn’t you go to her office because you were already pretty sure something hurt?
Why do we have this quest for a “biomarker” for pain? Here is my speculation. Our treatments, particularly pharmacological, don’t work very well. The people making them are heavily invested in producing successful outcomes in order to continue making and selling them, and believe in them. Rather than the depressing admission that the drugs themselves lack efficacy, they conclude there must be something wrong with the patients. “That lousy self-report is too variable to show our drug works. A biomarker would demonstrate that our treatment works.”
Back to the doc’s office: “Mr O’Mahoney, despite your self-rated 8/10 pain level, the MRI clearly shows that this treatment is working.”
Let’s indulge those folks who want to find the brain biomarker. Shouldn’t the brain biomarker be subjected to the same standards that they claim (erroneously) is lacking in self-report? At the time we wrote the original paper , I was surprised to find virtually no information available on the reliability of brain imaging of pain used to derive “biomarkers”.
So, we took data from our lab and put the brain biomarker business to the test. In the first study, we looked at reliability of functional magnetic resonance imaging (fMRI) and self-report, obtained at the same time. Self-report was highly reliable. fMRI was statistically reliable, but only at a modest level. The reliability of the fMRI suggests it’s good enough for investigating how the brain works, but certainly not high enough for any diagnostic test currently used in hospitals.
The paper that prompted this blog entry was the 3rd step in this line of investigation. We simply took data from another one of our projects and applied the impressively sophisticated mathematical algorithms that others have employed to classify people with chronic pain to both the neuroimaging data and the self-report data. The message from this paper is quite simple. Self-report was nearly perfect in classifying patients vs non-patients on the basis of single item pain or mood scores. We employed over 50 brain regions to replicate what other authors report as their significant classification rates. The self-report algorithms significantly outperformed the “brain biomarker” algorithms. The self-report classification rates were excellent (90-100%). The best “brain biomarker” classification was in the 70-percent range. There is a certain irony in these reports: “Our brain biomarker is 70-percent as good as that self-report that is so lousy we want to replace it?” What?!!
Back to your doc’s office: Which do you want to use, your self-report or the MRI scan?
There are philosophical and ethical considerations to be grappled with. I have employed brain imaging in my own work on pain perception for many years. I will continue to do so. There is a very important place for brain imaging in the investigation of pain. But, which should take precedent: your report of your pain, or the scan of your brain? What if they differ? Which should be believed? Which one is “correct” and which is in “error”? Who should use these data, your doctor, your insurance company, or your lawyer? Is it worth putting an MRI scanner in every office?
The measurement will improve, and perhaps the classification rates will begin to approach that of self-report. I’m certain of that because the technology improves rapidly, and people will do the appropriate studies to adequately test reliability and validity. The philosophical and ethical considerations will not be adequately answered at the same pace.
About Michael Robinson
Dr Robinson is a Professor of Clinical and Health Psychology, Anesthesiology, and Physical Therapy at the University of Florida. He is the Director of the Center for Pain Research and Behavioral Health at the University of Florida.
- Robinson, M.E., R. Staud, and D.D. Price, Pain measurement and brain activity: will neuroimages replace pain ratings? The Journal of Pain, 2013. 14(4): p. 323-327.
- Letzen, J.E., et al., Test-Retest Reliability of Pain-Related Brain Activity in Healthy Controls Undergoing Experimental Thermal Pain. The Journal of Pain, 2014. 15(10): p. 1008-1014.
- Robinson, M.E., et al., Comparison of Machine Classification Algorithms for Fibromyalgia: Neuroimages Versus Self-Report. The Journal of Pain, 2015. 16(5): p. 472-477.