Who is most likely to respond to TMS?

This is unfortunately a question we don't have definitive answers to. Every case is different. It's certainly a question we're trying to answer, and Dr Jonathan Downar's lab is at the forefront of this effort. Here I breakdown a recent study of his that tries to pin down the answer to this question.

First, some caveats:

• This study only looks at the first 4 weeks of treatment. Many people respond only towards the end of treatment, or even up to 5 weeks after completion. So the "poor responder" group could very well have had a full response after study end.

• This study only uses standard TMS (targeting the left DLPFC brain area). It focuses a lot on anxiety as a predictor. When I treat someone with anxiety, I frequently add on right DLPFC stimulation. So poor response in this study doesn't necessarily translate to poor response in my clinic.

• No modifications are made to treatment along its course in this study. That's great for controlling variables, but not how I operate my practice. I make frequent adjustments if we're not seeing an early response.

• Causation does not equal correlation! Meaning, for example, just because you'll see that there are more benzo users in the non-responder group, we cannot conclude that benzo use causes a non-response in and of itself.

Caveats aside, the more data we have the better! Just be careful not to let any of it discourage you from pursuing this treatment, and let's discuss it in more detail if you're considering TMS.

Okay enough disclaimers, let's look at the actual study...

Drs. Kaster, Chen, Downar, and their colleagues looked back at all of the data collected from a prior randomized clinical trial (conducted from 2013 to 2016 at 3 Canadian academic hospitals). After cleaning up the data and excluding people that didn't qualify for the analysis, they ended up with 388 subjects.

Their goal was to help us move away from viewing depression one-dimensionally. Depression is such a heterogeneous condition, that it would be very helpful to break down TMS response based on certain symptom clusters. That way when we treat, instead of using a one-size-fits all model, we can eventually choose exactly the right type of TMS for the particular person.

They identified four trajectories of symptoms that they wanted to track along the treatment course: anxiety, mood, insomnia, and somatic symptoms. Then they pursued the answer to this question: What baseline characteristics predict which of those trajectories a person would end up following?

Here are their results, followed by an explanation of how to interpret them:

First, let's be cynical and analyze the minimal response group (magenta). These subjects didn't get any improvement from their TMS course overall. We can see from the graphs that they also had the worst starting mood symptoms, worst insomnia, and worst somatic symptoms (but not the worst anxiety symptoms). I believe the most important take-away is the shape of their graphs: all pretty flat-line, or even curving upwards. For me, this confirms my current practice: if we are not seeing any glimmer of anything improving by session 10-15, we're going to make some type of change to your protocol.

You may also notice that the magenta group started out worst on almost every measure. I don't think this means that if you're in really bad shape there's no hope. But it may point to requiring a greater intensity of treatment, and a collaborative approach with other modalities.

Okay let's move on to the positive. The yellow group are the best responders. Like the other two responder groups, their trajectory shows a steady improvement in all symptoms over time, which is what we expect. Now, given that the non-responders had the worst symptoms on almost all measures, you might wonder if the opposite will be true for the responders -- did they do better because they started out better on all measures? Nope! This is good news. It means TMS doesn't just work on people who are already feeling better.

Now we can look at the partial responders, who are separated out by high anxiety (light blue) and low anxiety (light green). I find it interesting and reassuring that their trajectories move mostly in parallel. Having high anxiety doesn't block your response to TMS, otherwise we would have seen these two lines diverge. They all showed the same symptom improvements, just from different starting points.

Alright, now let's look closer at the groups themselves, so we can see if any element of the baseline characteristics can predict which of these 4 groups a person might end up falling into, and whether we can increase the odds of some falling into the optimal response group in the future.

Among groups we see that sex, average age at depression onset, education level, and episode are not significantly different (that's what the statistics in the right-most columns tell us). Their age is the only significant difference (a p value of less than .05 tells us that a finding is significant and not just due to random chance). Participants in the optimal response group tend to be older (average 44.08 years) compared to those in the minimal response group (average 39.42 years).

Here are the other characteristics that do not show significant differences among groups: antidepressant treatment, antidepressant combos, antidepressant augmentation, and number of previous antidepressant trials. This is good news too - it doesn't really matter what else you are taking during TMS as far as antidepressants go.

Unfortunately it may matter for another class of medications: benzodiazepines, which are commonly prescribed for anxiety.

These results tell us that there is a significant difference among groups in their use of benzodiazepines. At 48%, the minimal response group contained a significantly greater number of benzo users than the optimal response group at 26.05%. Partial responders fell somewhere in between. (P is less than .02 so the difference is significant, not just random variation.)

Is this merely a marker of more severe illness, or does it tell us that benzo use itself makes it less likely someone will respond to TMS?

This last table helps to make sense of whether these two factors (age and benzo use), being significantly different among the groups, also increase the odds of falling into one group versus another. This is another statistical measure known as an Odds Ratio (OR). I'm going to get a bit uber-statistical here so if your head is already spinning, skip ahead.

Remember p value? Let's start there: only p values less than .05 are significant, the rest are due to random chance. So let's eliminate any findings except those significant ones. I'm also going to eliminate the "Symptoms" bottom half of the chart -- we know these differences are significant because that's how the groups were formed.

Ahh, isn't that nicer to read?

There were only two baseline characteristics that significantly affected which group a person fell into, and that applied only for falling into the minimal response group.

Here is the summary of those findings in English:

• Younger age makes you 0.95 times less likely to respond to TMS.

• Benzo use makes you 2.73 times less likely to respond to TMS.

The age thing is a bit of a "whatever" result, since there's nothing we can do about it and if you recall from up above, we're only talking about 3-4 year variations in average age.

The benzo result is probably the most significant take-away of the whole study. For standard TMS to the left DLPFC, benzos may reduce your odds of response.

Should everyone stop benzos before starting TMS? NO. We've suspected for a while that this might be an issue with benzos based on mechanisms of action of both. I have treated plenty of patients with TMS who were on benzos and responded to treatment well. This would be my recommendation if you are on a benzo and considering TMS:

• If your anxiety or insomnia is such that you cannot function without a daily benzo, then you're probably not going to be able to regularly get to TMS treatment if you try to stop it first. And TMS definitely does not work if you can't get to treatment! So in this case you should continue, and perhaps consider a taper later in treatment once you're feeling better. (Remember, this study did not make use of specific anxiety-targeting protocols, which may facilitate stopping benzos.)

• If you're not in the above category, and you've been considering tapering your benzo, then doing so before TMS treatment may be a good idea in order to boost your odds of response.

• If you only use a benzo sporadically, i.e. for a panic attack, you probably do not need to worry about any of this.

This leaves me with a lot of questions - were these benzos standing or only as needed? Would the results be different if anxiety-targeting protocols had been used? Luckily I have my own data set, from 7 years of treatment. I'm going to crunch those numbers.... stay tuned for part 2! And please post any questions or comments below.

~ Dr Mednick

AI disclaimer: AI was used in the creation of this post, but only to help interpret the statistical values in the study. Sorry folks, not my strong suit.