The Treatment-as-Prevention Hypothesis

The treatment as prevention hypothesis is the untested idea that therapeutic treatment early in life may prevent or delay the development of serious psychiatric issues later in life. Despite the lack of evidence for its safety or efficacy, proponents argue that this approach is reasonable, since we do know the harm of doing nothing.

The principle is not unlike the idea of taking a baby aspirin to prevent a heart attack, and in principle, I think many of us consider it a no-brainer.

I’ll take it a step further and say that the idea I’ll call “therapy early, therapy often” seems reasonable enough that nobody really objected to it until recently.

Why recently?

Well, yesterday’s hypotheticals are today’s realities.

Science™ takes time, and obviously we needed to wait for The Children to come of age before we can begin to honestly evaluate The Evidence.

And so far, it’s not looking good.

The Evidence

When we speak of The Evidence, we’re referring to a combination of Data, Information, Knowledge, and Wisdom. The Data, raw, unprocessed, and context-free, must be wrought into The Evidence, and this takes time.

Once The Data is collected, the researcher organizes and processes it, creating Information that tells us something about the world. Information still has its limits - we frequently lament our inability to extract Knowledge from the ever-increasing deluge of Data and Information, for example.

When multiple pieces of Information are pieced together, we arrive at Knowledge: an understanding of how something works or happens. Sometimes, we must consider ourselves fortunate to get to even this state.

But the pinnacle of understanding is Wisdom, a heady blend of Knowledge, experience, intuition, perspective, personal and institutional values, ethical judgement, and at least half a dozen other secret ingredients.

Put another way, data tells us little on its own; Information answers who, what, where, and when; Knowledge tells us how, and Wisdom tells us why. Or, my personal favorite:

“Knowledge is knowing that tomato is a fruit. Wisdom is not putting tomatoes in a fruit salad.”

Consider two farmers, each with their own ideas about how to best grow a crop. They challenge each other to a bet, based on some measurable outcome: the biggest yield, the heaviest pumpkin, the most beautiful flower.

Certainly, the townsfolk have their own thoughts, and some perhaps even posess wisdom that seemingly allows them to fortell that Farmer John’s pumpkin will be the biggest, most beautiful pumpkin the world has ever seen. Of course it could also be luck.

But, a long time ago, we put Science in charge because it gave us better results than pure chance. For all of its shortcomings, on the whole, it’s much better than placebo.

To Science™ the problem, we’ll need to wait til the end of the season to know for sure.

So with the “mental health generation” beginning their transition into adulthood and the GSMS participants well into their 30’s, it’s an exciting time for people who care about longitudinal studies on the adult outcomes of early childhood, childhood, and adolescent interventions.

Better Than Placebo

Better than placebo is the minimum standard by which we measure the effectiveness of an intervention (classically, a medication).

Western medicine has only gained its impeccable reputation (“surgical precision”, “clinical efficiency”, “antibiotic effect”, etc.) in the latter half of the 20th century, with the introduction of antibiotics, medical imaging, etc.

In the era of phantomplasia hyperghoulemia, the public was far more likely to consider physicians quacks, and it was the state of medicine in this era (and not the bumbling, superstitious nature of the patients) that allowed homeopathy to flourish.

Homeopathic medications are diluted many times, resulting in the patient receiving what’s basically just water. Treating every ailment with a small amount of water is obviously no better than chance or placebo (because it’s literally a placebo), but it also doesn’t cause harm.

Compared to mercury, arsenic, heroin, or cocaine, plain old water must have seemed like a panacea.

Risks & Benefits

Inert substances - placebos or homeopathy, take your pick - are one thing, but it’s a reasonable generalization to say that any substance potent enough to heal is potent enough to harm. That is, there’s more to the equation than better than placebo.

The general balance is one of Risk vs. Benefit, and there are lots of ways to calculate it, which can quickly get messy - even for The Experts.

The NNT (Number Needed to Treat) was developed as a practical metric for providers more focused on medicine than statistics:

The NNT offers a measurement of the impact of a medicine or therapy by estimating the number of patients that need to be treated in order to have an impact on one person. The concept is statistical, but intuitive, for we know that not everyone is helped by a medicine or intervention — some benefit, some are harmed, and some are unaffected. The NNT tells us how many of each.

An accompanying measure, the Number Needed to Harm (NNH), is of course the measure of harm. When taken together, the NNT and NNH paint a picture of how beneficial (or harmful) a given intervention is.

A best-case intervention has a low NNT and a high NNH. An intervention with a high NNT means that we have to administer it to more people in order to see anyone benefit. A high NNT is sometimes acceptable as long as the NNH is not too small. A smaller NNH means patients will be harmed more frequently.

Remember aspirin to prevent heart attacks as a no-brainer example of the treatment as prevention hypothesis? That was a setup.

Let’s take a look at The Evidence.

Fortunately, the NNT gives us several aspirin evaluations:

1. Aspirin Given Immediately for a Major Heart Attack (STEMI) (NNT: 42, NNH: 167)
2. Aspirin For Preventing A First Heart Attack or Stroke (NNT: 233, NNH: 250)
3. Aspirin to Prevent a First Heart Attack or Stroke (NNT: 1667, NNH: 3333)
4. Aspirin to Prevent Cardiovascular Disease in Patients With Known Heart Disease or Stroke (NNT: 50, NNH: 400)

The first one doesn’t quite apply to our specific hypothesis of prevention (though it may help inform our general impression about aspirin’s overall utility). The second and third results apply specifically to preventing a first heart attack or stroke, while the last one looks at prevention in patients with existing heart disease.

Since #4 is a well-known practice, it’s probably the most fair interpretation of our vauge hypothesis. Also, it has convenient numbers, so let’s start there.

In this case, aspirin has an NNH of 400, meaning that we can expect a single adverse event per 400 patients. Based on the NNTs, in this ideal population of 400, we’d also expect to have prevented:

• 8 cardiovascular prolems (NNT=50)
• 5 non-fatal heart attacks (NNT=77)
• 2 non-fatal strokes (NNT=200)
• 1 death (NNT=333)

So, out of 400 patients, one person was injured and had to be kept in the hospital, but we saved a life and prevented over a dozen instances of harm. Great success!

But what if we expand access to aspirin? What if we recommend that everyone take aspirin, just in case?

Let’s look at The Science.

Well, shit.

It’s all red.

Red isn’t good (in the US, anyway).

With an NNH of 250 and an NNT of 333, we’ll be hurting people more than we’re helping. In a group of 300 patients, we’d expect no benefit (NNT=333) and one adverse outcome (NNT=250).

The LCM of the two numbers is over 83k, so let’s pretend aspirin is a little more beneficial than it really is and round its NNT down to 300. In some hypothetical, ideal population of 1500 patients, we’d expect to see:

• 5 patients helped (NNT=300)
• 6 patients harmed (NNT=250)

So that’s a no-go, at least if you’re worried about patients (and not aspirin sales).

Maybe #3 will show us how right we are.

Nope.

Still red.

And the NNTs are even higher!

Who would have thought Science is so anti-aspirin?

Considering these values in an imaginary, ideal population of three thousand some odd people, we’re pretty much breaking even:

• We prevented maybe one or two cardiovascular problems (NNT: 1667)
• We prevented maybe one non-fatal heart attack (NNT: 2000)
• We prevented perhaps one non-fatal stroke (NNT: 3000)
• One patient or so was injured (NNH: 3333)

So, out of 3,000+ people, we maybe helped 3-5 of them and perhaps injured someone. 99.8% of patients didn’t notice one way or another.

But we sold 3,000+ aspirin!

The reason for this is a bit obvious: aspirin is a relatively mild drug that probably won’t affect most people all that much, all things considered. That’s probably why it’s available in vending machines. But literally any (competent, medication-administering) healthcare provider will be quick to point out that aspirin is not without risk.

Furthermore, in low-risk populations, there are simply no heart attacks or strokes to prevent. This means that there cannot be any possible benefit, leaving only risk. An intervention with an NNT approaching infinity is not a good intervention - especially with non-zero risk.

The authors explain this in more detail (emphasis added):

The apparent failure of aspirin to be helpful in this population highlights an important fact about medical treatment and the results of research on medical treatments: the more likely that patients in a study will have an event (a heart attack or a stroke, etc.) the more likely it is that they can potentially benefit from an effective intervention. Conversely in a group of healthy patients who are unlikely to have a heart attack or stroke, it is very difficult for a drug to successfully reduce heart attacks or strokes. This is intuitive: if there are no heart attacks in a group, it is not mathematically possible to reduce heart attacks. If there are very few heart attacks, it is mathematically difficult to reduce heart attacks. The more heart attacks in a group, the more room for improvement. Thus drugs that are effective in preventing events like heart attacks are always increasingly effective as the risk of heart attacks in the group increases.

Yes, heart attacks kill lots of people each year.

Yes, strokes are Bad.

Yes, most people will probably be fine if they take a baby aspirin for no reason whatsoever.

Nevertheless, we don’t make blanket recommendations that encourage everyone to take aspirin “just in case”, because there’s so much more to it than being better than nothing (or being “pro-Aspirin” or “anti-Aspirin”).

The Kids Aren’t Alright

At the center of this emerging debate is the treatment-prevalence paradox (TPP), which points out a puzzling situation: even though we’re putting more time, effort, and money into mental health awareness and interventions than ever, the overall state of mental health seems to be on the decline.

There are at least seven hypotheses for this:

1. People are more open about talking about their mental health and doctors are more likely to diagnose someone with anxiety or depression. Known as diagnostic inflation, it inflates the denominator, making it appear that patients are worse-off.
2. There is a genuine increase in the number of patients (e.g., suffering from anxiety and depression)
3. Studies performed in a controlled setting may make interventions appear more beneficial than they are (e.g., studying aspirin on the cardiac ward)
4. Study results may not pan out when applied to real-life situations
5. While patients with occasional or minor complaints might see a great improvement, they don’t do much for the chronic, serious cases we’re worried about.
6. Some treatments are doing more harm than good.

The author does not state how he chose the order for that list, but it appears to me that the most frequently-claimed “excuses” from the “pro-Therapy” crowd appear at the top, while the complaints of the “anti-Therapy” crowd tend more towards the bottom.

It’s probably not that deep, but I mention it to illustrate the divide I see here: anecdotally, proponents of “therapy early, therapy often” tend to dismiss the paradox as artifact, while the “get away from my kids, creep” crowd focuses on the futility or harm. However, like most things, the truth is probably multi-factoral - a little from column A, a little from column B.

The Middle Path

More research is absolutely needed - and it’s coming.

Some cursory searches reveal a number of studies published in the last couple of years, and I only expect more as these kids come of age and the various school-based mental health surveillance/intervention programs come under increasing scrutiny.

As the funding and number of interventions continues to rise, we need to do our best to keep that gauge between the NNT and the NNH. With mental health at an all-time low and a finite number of resources, we need to ensure our efforts are focused on the serious, chronic cases that affect society and not squander them on homeopathy for entitled, privileged heirs.

However, as long as individuals with serious, debilitating mental illness remain without homes or jobs while the therapists’ paychecks and Yelp reviews are written by affluent families, facing this challenge will continue to be an uphill battle.

Unfortunately, “it’s complicated” doesn’t make a great political slogan, so we’re stuck with the stupid anti-Therapy vs. pro-Therapy bullshit until we decide to tell those simpletons to sit down, shut up, and quit politicizing our childrens’ fucking safety.