Faces of digital health

View Original

F131 How Can We Better Measure Pain? (Sara Berger, IBM)

Pain can be acute or chronic. They differ not only by how they feel but also how they appear in the brain, spinal cord, and elsewhere. Historically, chronic pain was treated in the same way as acute pain.

The academic definition of pain is more or less agreed upon, however, aspects of studying pain can differ substantially.  Science is also in the beginnings of understanding the impact of fear and emotions on pain. 

Understanding how multifaceted chronic pain is, is shifting care to a more holistic approach, says Sara E. Berger, a researcher at IBM with over a decade of experience in the pain field. 

“Particularly in a clinical setting, when someone asks you what's your pain level right now on a scale of one to ten, you are automatically bringing up memories of previous pain for comparisons. Are you scared in that moment? Are you calm in that moment? All of these things color your perception of pain in that moment. That makes it difficult to study pain. Additionally, everybody's coming at their pain experience with a variety of other life experiences. This makes it also really tough to say two chronic pain patients are the same. Maybe one is a female, one is a male. Or one is white or one is black. With technology we can now start using all of these differences to cluster them, and make sense, find patterns in individual differences,” explains Sara E. Berger. 

She among other things researched the mechanistic standpoint of a placebo response. 

How far is research in making most of the placebo effect?

A placebo effect happens when a patient gets a fake treatment and still benefits from it because the brain managed to convince the body that it received real medication. The opposite of placebo is nocebo - when the patient suffers negative consequences without getting a harmful substance. 

Placebo has traditionally been associated with deception. But research shows that you can get the same benefits from open-label placebo when people actually know that they are taking some sort of sugar pill, or sham treatment, says Sara E. Berger. 

Sara E. Berger.

“Placebo is all about expectation. Statistically speaking, there's always going to be a number of people that are randomly assigned to an active treatment group who would be placebo responders. They're likely to respond no matter what was given to them. The challenge is that if you have a group of people like that already in your active treatment group, then you're already, in some ways biasing the treatment effects of that group. If we could recognize placebo responders we could identify them ahead of time and make sure that groups in clinical trials are equally stratified. That then makes clinical trials more likely to be accurate.”

There’s still a lot of unknown in the placebo research. For the last 10 years, placebo has been more or less studied in otherwise healthy people without chronic problems. It’s still not clear if a placebo response is a trait or a state, Sara E. Berger mentions. In the past researchers often gave people a placebo and said it will help them, which set additional pre-defined expectations in study participants.  

Detecting and assessing pain based on language

In one of the research projects Sara E. Berger worked on, research showed that certain brain networks and associated personality characteristics are predictive of a placebo response before patients even get a pill.

“We found that there are certain ways that people talked about their pain experiences overall, that was also predictive. We see language as a window into the mind, right? It's both a consequence of thought, but it's also a cause of thought. And so the idea is if the pain is in the mind then how you think about pain should affect how you talk about pain. We essentially asked patients a variety of things that weren't really even about their time in this study. We asked them to talk about themselves, what they like doing, what dreams they had lately. Then we also asked them, how has your pain experience been so far? When did it start? What are some previous experiences you've had in the medical system? What we discovered was that there were very clear distinctions between the people that responded to the placebo and the people that didn't. We - researchers - were blinded at the time, so we didn't know who did or didn't get a placebo until later. So the perception of pain was heavily related to identity and self and leisure activities,” Sara E. Berger says. 

In this research, patients were told they may or may not get active treatment. They were told that what they received may or may not help them. Researchers did not encourage predefined expectations.

“We wanted to see where patients would naturally go. The study was double-blinded so I had no idea as I was interacting with patients, what they were assigned to. They didn't have any interaction or knowledge of what they were assigned to throughout the entirety of the study. We even triple blinded the data analysis to see if our model results were by chance or if they are actually really robust.”

The multidimensionality of pain research

Apart from the content, scientists are also looking at emotions that can be recognized in how we say things and how this can be used for diagnosis. “We did not look at that aspect in our particular study, but commenting on it generally,  voice has acoustic properties. With acute pain you might imagine that your pitch might get higher if you're in more pain. Those give us a hint at emotionality but also can give us a hint if other cognitive differences are happening. For example, speech rate can be indicative.  If we combine the acoustics of what you say and the content of what you say, that's where you probably get the best signals.”  

Tune in for the full discussion.

See this content in the original post

Some questions addressed: 

  • For starters: how does science define pain since pain is a very personal and subjective matter? How diverse are definitions of pain in the research space? 

  • The challenge of pain is that it’s very difficult to quantify because it depends on the emotional state of an individual. The same physical presentation of a condition can cause different levels of pain in different people because pain is a matter of perception and tolerance. Can you talk a bit about what has science tried so far when designing assessment criteria? 

  • Now let’s talk about the role of placebo. We mentioned that psychology has a huge impact on pain levels. How does that translate into medicine?

  • In the era of data and new technologies, new solutions are invented daily to curb pain. Novel therapies include VR, mental health aids such as meditation and mindfulness. What’s your view on these approaches? (f there’d a comment based on publications and research in the field you’re following closely)

  • You recently posted research about how Speech-to-text AI could help doctors prescribe a placebo to ease chronic pain. Sounds intriguing. Can you share the research basics: what were the hypothesis, the sample size, and other details? 

  • This research refers to voice analysis for health prediction. Researchers are looking at what can be discovered in how we speak but also what we say. Can you tell us a bit more about some of the findings in the field so far? 

  • Placebo and nocebo are a known yet puzzling phenomena in medicine. How much do we know about them in the quantified/controllable sense? 

  • If researchers figure out what kind of personality types respond to placebo, what consequences can that have on clinical trials? Could people that respond to the placebo be eliminated from the trials because it would be evident in advance that they might experience improvement because of the placebo? 

  • There are clear ethical aspects to consider when thinking about prescribing someone a placebo medicine. In your experience/research, what do clinicians think about this? Liability issues, etc. 

  • “In the future, for example, a standardized list of pain-related questions given to patients during a visit could potentially be automatically transcribed with AI speech-to-text tools and analyzed in real-time to provide a physician the likelihood of that patient’s response to a prescribed treatment.” How far do you think we still are from there? How many trials and research still need to be done?