Brave New Precision Medicine
New techniques can predict relapse: Is this a good thing?
If the market for baseball players was inefficient, what market couldn’t be? If a fresh analytical approach had led to the discovery of new knowledge in baseball, was there any sphere of human activity in which it might not do the same?
– Moneyball author Michael Lewis, about his best-selling book and the film it spawned
Moneyball is a story about discarding old-fashioned ways to assess athletes’ skills, in favor of modern statistics. In the plot, new math spawns serious athletic wins, along with the question: Is there any place it can’t do the same?
Can new statistics be used to treat addiction? Novel ideas are spilling from research in medicine, psychology — and statistics — at a nimble pace. Scientists aiming to help addicts find relief from suffering are constantly seeking robust, novel approaches. For example, in efforts to unravel how relapse happens, various attempts to take pictures of the inside of an addict’s brain have included Magnetic Resonance Imaging, or MRI scans. X-rays, EEGs (Electroencephalograms), PET (Positron Emission Tomography) scans, and DTI (Diffuser Tensor Imaging) have also been tried.
Another innovation is SPECT — Single Photon Emission Computed Tomography. In one 2016 study, SPECT neuro-images showed abnormally low blood flow in marijuana users’ brains, and normal blood flow in non-users’ brains. Statistical analyses of SPECT images told doctors which brains were users’ and which brains were non-users’. And a 2013 article titled “Researchers say MRI could help predict success of drug addiction treatment” describes how during an MRI of an addict’s brain, “reward” neurons lighted up, following even brief cues — like a photo of crack rocks flashing on-screen for just milliseconds — while the addict was looking at a neutral image, like a stapler.
In addition to highlighting “reward” areas in the brain, MRIs show which brain parts seem linked to addictive compulsiveness. Newly developed neuro-images like MRIs help to light up brain pathways that are active in addiction, something never before seen in medicine.
MRIs also show why self-reported craving is a known contributor to cocaine relapse. Recovering addicts experiencing less craving tend to stay drug-free. In a study of craving, cocaine addicts’ self-reports matched their MRIs — which showed activity in the craving area of their brains. MRIs showed craving would occur when addicts reported feeling it. Treatment strategies aimed at craving are thus likely to be beneficial for cocaine addiction.
A 2016 article, “The Neurobiology of Addiction: The perspective from MRI, present and future,” describes mapping what appear to be relapse pathways in cocaine-, alcohol- and opiate-dependent patients.
When MRIs came into frequent use, their pictures of the inside of a living body resulted in clearer, sharper images than old-fashioned X-rays. The difference between abnormal tissue and normal tissue was clear to see. The flow of fluids and electrical impulses between body parts was visible.
MRI is like a camera, not a surgery. It is non-invasive yet can show how atoms in the brain emit energy that is different for healthy tissue than for damaged tissue. A doctor can review an MRI and see not only what brain parts may be diseased due to addiction, but also whether nerves that connect various parts are working like they should be. MRIs can show how well neurons are connecting the brain’s various parts. A doctor can see if brain parts have decreased in size, or whether messages are not getting through because the connecting nerve bridges are diseased. If a brain part has shrunk due to substance abuse, or has changed shape, it will not be working, and that will show up. MRIs can show whether the oxygen levels and electricity flows inside a brain are normal or abnormal.
Reward systems inside a healthy brain cause people to use what they’ve learned about rewards: “If I’m high all day, I will not be rewarded with a paycheck. If I get high, my loved one will not reward me with sex. If I get high, I might end up arrested, like the other times. I might hurt a child or crash my car.” The greater the amount of brain disease seen on an MRI, the greater the chance the brain’s reward systems are not working right.
Conversely, in healthy brains with normal reward systems, the need for a paycheck or for love can overcome craving. In brains damaged by drug abuse, craving can overwhelm the need for a paycheck — or for sleeping or eating. Craving can overcome fear of going to jail or being injured. And the parts of the brain involved with craving, and with reward, show up on MRI scans.
Amount of diseased tissue has been correlated with relapse in a number of studies. The 2016 study “Neuroimaging Findings in Methamphetamine Abusers,” describes abnormalities in the brains of users of about five years, and in size changes in portions of brains of those using for about nine years. This kind of neuro-degeneration leads to the involuntary movements sometimes seen in meth users. The study also describes how MRIs can light up dynamics such as reaction time or memory accuracy.
For creating new treatments and for engineering new MRI machines, this research is great news. For recovering addicts who’d rather not know their personal chances of relapsing, it’s not. And for addicts who hope there’s no chance someone can know their personal probability – a short list would include courts of law, social media trolls, family, insurance companies or employers – these advances are unwelcome. Just as it has in business and government, once technology enters the equation, the pace of discovery speeds up, and comes astride a sci-fi scent.
Every scientific innovation has pros and cons. The pro here is that greater MRI precision leads to more successful treatment planning. The con is that greater precision fuels onerous levels of stigma.
Estimates of the number of brain cells inside a human head range from 86 billion to 100 billion. Joined together by synapses and impulses that travel from neuron to neuron, brain cells can be connected to millions of other brain cells. MRIs have gone a long way toward mapping all this out. Still, neurons in the brain are so dense and intertwined their number is difficult to define. Given that the estimates are so huge, neural network functioning is big data – really big data.
As techniques expanded, so did the building of more precise MRI machines. Next came newly invented statistics to harness the results of new-and-improved MRI machines. Years ago, personality tests alone were used to plan treatment. Then later, personality tests were combined with MRI results. Doctors not only gave personality tests, they also looked at whether addicts’ personalities matched their MRIs — a bit like a lie detector tracking whether a person’s spoken words match their physical condition.
Now there’s a triple combination being used: personality test, plus MRI, plus new statistics.
Math advances are marching — or some would argue, slithering — into the field of MRI interpretation. In one area of biomedicine known as Data Science for Personalized Medicine, or DS4PM, the goal is to “apply cutting-edge mathematics to improve decision quality.” Statisticians are “creating powerful decision-support algorithms that assist clinicians” in predicting outcomes. This augurs “more efficient methods that push the boundaries of modern statistics.”
One newcomer is the “random-forest” statistic. A 2015 study suggests that MRI results can be fed into the random-forest statistic to generate predictions of relapse for meth addicts, with greater precision than previously possible. A “forest” of data is analyzed, and re-ordered from different angles, not just one way, as older statistics might do. Mixing the personality test, the MRI scan and the random-forest statistic was a good predictor of relapse.
Random-forest is not the only new kid on the block. Others are on the way. This alloy of MRIs and personality tests, all funneled into new analyses like random-forest, fuels the Moneyball question: Can new kinds of stats work for anything?
The accelerated pace of innovation can leave some addicts feeling unmoored. This is where the sci-fi scent wafts in. Authors of the book Addiction Neuroethics: The Promises and Perils of Neuroscience Research on Addiction, note that in Australia, researchers cannot guarantee human subjects their data will not be subpoenaed by police. Many subjects thus remain anonymous, making follow-up studies nearly impossible.
Still, progress dazzles. A 2008 study by Martina Reske and Martin Paulus, “Predicting Treatment Outcome in Stimulant Dependence,” advocates for more:
“We need to push the limits of this technology to clearly show its ability to define clinically-relevant information. . .”
It’s unclear who will be defining “push.” Or “limits.” Their push is somewhat tempered by Reske and Paulus’ recognition that stigma is hovering overhead:
“Researchers will need to better anticipate and evaluate implied ethical concerns. The medical and social consequences of predicting drug dependency or relapse, based on functional brain images, have to be acknowledged. Effects on health insurance and stigmatization are obvious issues. Measures to guarantee confidentiality will have to be identified.”
Do The Math
The aforementioned article, “The Neurobiology of Addiction: The perspective from MRI present and future,” covers how new techniques track brain circuitry breakdowns within addicts’ MRI images. The neuro-circuitry of addiction is better understood every day, according to the study authors: “There have been large improvements in MRI technology as a result of the evolution of precision engineering and electronics. . . . the greatest advances may not come from new (MRI) measurements, but from applying innovations in data processing and computer modelling.”
However, as the drive toward new math rolls forward, caution grows. Fake-news scandals are narrowing the veneration of statistical algorithms that determine what “news” is spread on the Internet. In her 2016 article “Alien Algorithms,” data scientist Cathy O’Neil writes about analysts who “worship at the altar of algorithms.” She describes algorithms that have become widespread, to evaluate success and failure in various categories and “. . . to illustrate how extreme our blind trust in algorithms has become, even when they are statistically unstable. Our faith in and fear of mathematics has allowed” this to happen, according to O’Neil:
“We’ve developed all sorts of scoring systems, ranging from scores that determine who among us is likely to end up in jail …to scores that predict when we’ll get sick, whether we’ll quit our job, whether we’re vulnerable to the predatory loan industry. Moreover, they’ve all got these three things in common: they’re widespread, secret, and destructive. I have a name for algorithms like that. I call them Weapons of Math Destruction, or WMDs.”
And despite their “pushing of the limits,” Reske and Paulus do recognize what addiction professionals know all too well:
“Individuals at high risk for relapse may be liable to increased stress around such testing and predictions, which themselves may increase relapse risks.”
The irony is that if addicts learn they have a solid chance of relapse, the emotions that follow – hopelessness, anger, depression – might themselves trigger relapse. A prediction of relapse could itself breed relapse. Finding out that one will relapse does not exactly light up the reward neurons, but it might light up the need to escape reality.
Published on TheFix.com, February 2017
© 2016 Kathy Jean Schultz