The Story of Dr. Matthew Walker and The Fear of Sleeping Pills

By David E. McCarty, MD, FAASM (but you can call me Dave)

~ ~ ~ ~ ~

Is it possible for a book to be a frenemy?

That’s how I’ve come to think of Dr. Matthew Walker’s runaway train of a bestseller: Why We Sleep.

It’s a frenemy. I love it, but it bugs me.

Let’s start with the stuff I like about it. That feels right.

I can’t overlook the fact that the book is essentially a biography of a Science that I love deeply and dearly.  

In the introduction, Dr. Walker recalls distinctly the time when sleep became his “obsession.” My journey feels the same. I am fascinated by this stuff, crazy about it, drunk with it. I get a silly thumpy-hearted bibbliness when I think about Kleitman’s chronobiology experiments, deep within the earth at Mammoth Cave. I still think it’s so cool that his graduate students Gene Aserinsky  and Bill Dement discovered REM sleep kind of by accident. The discovery of the machinery beneath homeostatic sleep pressure—the glymphatic system—nearly blew the back of my head clean off.

These stories are legendary, the science epic and marvelous.

So I love the fact that Dr. Walker captured all of that, for the record. His book is a beautifully written love letter to a field that has been my obsession, as well.

So for that, I’m very grateful. Thank you, Dr. Walker. I mean that.

Here’s the thing, though, and when I say this, I’m specifically talking about the narrative about sleeping pills.

This book scares people and it does so on the basis of weak science.  

Scaring people has consequences.

I have something to say about this punchy little combination.

Legitimate harms, including death, can result as a direct consequence of taking sleeping pills

[as Jeff Spicoli famously said: People on ‘ludes should not drive!].

Let’s talk about what I mean.

I’ll start with this: Sleeping pills are a problematic drug class that can harm you in interesting and inventive ways. The sleep that one obtains whilst using sleeping pills is not exactly the same as sleep without pills. Legitimate harms, including death, can result as a direct consequence of taking sleeping pills [as Jeff Spicoli famously said: People on ‘ludes should not drive!].  Therefore, sleeping pills must be considered very carefully.

All of these statements are very, very true.

However.

Sleeping pills causing cancer?

Let’s discuss this. 

Before we get into it, let’s review the concept of proving scientific causality. It’s actually not trivial to prove that something causes something else to happen. 

Let’s explore that.

You’re walking in the supermarket floral area. You look up to notice a giant mylar unicorn balloon. The moment you look at the balloon, it deflates and sags to the ground.

I think we all would agree that your GAZE didn’t sink the balloon. It didn’t pop because you looked at it. It popped whilst you were looking at it. 

There’s a difference. 

Statements about RISK can be kinda tricky for a feller to interpret, sometimes.

In medical and health science, it’s hard to prove causality. Really hard. It requires special tools.

One of these tools is called a Prospective, Randomized, Placebo-Controlled Trial (PRPCT). That’s a study design where you define a population you want to study, and you randomly assign part of the group to receive the exposure and part of the group receives an exposure to something that is known to be innocuous (the famous placebo “sugar pill”).

Researchers use this type of study design to prove the efficacy of a drug. The same model could also be used to prove that a drug caused harm.

Another powerful tool is called a Prospective Population-Based Observational Cohort. This type of study design is used for population health questions. For this type of study, a population is defined, and data are collected at the time the observation period begins, and then on an ongoing basis going forward. Observational cohorts can’t deliver the message of causality as clearly as a randomized trial, because the behaviors of those under study are self-selected. However, it can get pretty close. If you’re really careful and patient, you can collect enough data along the way to prove your point of causality pretty clearly.

Let’s explore the concept of self-selection, to see why that’s important. Let’s say we were studying a population of islanders. We discover that half of the people on the island eat the berries from the FruFru tree every day, and half the people don’t touch them. Self-selected groups, see? We decide to compare the two groups, to see if there are any health benefits or risks associated with eating FruFru berries.

After we collect our data, we come to a startling conclusion: everyone who eats the berries has brown eyes. Among the non-berry eaters, there are eyes of all different colors.

We start to feel a heady sense of discovery. We know that FruFru berries have certain chemical characteristics. Perhaps FruFru berries are changing the genetic code! Maybe eating FruFru berries changes eye color!

We will publish our research! We will go on Oprah! We are sure we are going to win the Nobel Prize!

And then one of the islanders points out that it’s common knowledge that being brown-eyed makes FruFru berries taste good. If you don’t have brown eyes, they taste like soap, like cilantro does for some folks.

And just like THAT! Our Nobel-Dreams are dashed.

The point here is that self-selection matters, and it makes drawing conclusions about causality really really really really tricky.

So let’s get back to the fear.

Me and my frenemy.

Showing off a new modification strategy which helps us get along a bit better.

I want to discuss a single sentence in Dr. Walker’s book, the sentence that launched a thousand panicked phone calls. It’s actually just part of a sentence. Anyway, it’s this one:

“…it is equally possible that sleeping pills do cause death and cancer.”

Dr. Walker based this statement on a research study that was not a Prospective Randomized Placebo-Controlled Trial. Nor was it a Prospective Population-Based Observational Cohort.

It was a retrospective chart review (1).

Here’s what that means, and why this point is so darned important.

Instead of a fixed population studied going forward, Dr. Kripke and colleagues used records they could easily access—primary care electronic health records of 250,000 unique persons, living in a rural area in Pennsylvania. They peered into the charts of these folks between the years 2002 and 2006, and they wrote down some of the stuff they found.

Amongst these records, they found that 12,465 persons had been provided with at least one prescription for a sleeping pill. Which means that in each case, the patient had to physically trudge themself into the doctor’s office, and complain about their sleep enough, so that the provider gave them a sleeping pill.

Are you getting a visual of what these people might be experiencing?

Mostly, when they got sleeping pills, they got zolpidem. A few received temazepam (different drug class with different pharmacologic characteristics).

For the kids dozing at the back of the class, I’d like to point out that now we are looking at two different drugs, with different effects on the brain and different effects on sleep.

Stay with me, it gets even better.

The next step in the process is to match these people to other folks in the database who didn’t receive a prescription for a hypnotic. We have to compare those who ate the FruFru berries with those who didn’t, right?

Ideally, we’d like the groups to be as similar to each other as possible, in all other measurable ways. After all, we’re still trying to live up to the angel-choired promise of the Randomized Controlled Trial.  

Ah, but dear reader, you already know the punch line, don’t you?—the two groups will never be the same. They started with a SERIOUS FUNDAMENTAL DIFFERENCE: some had complaints of a systemic and nonspecific nature (insomnia bad enough to request drugs), and some didn’t.  

But, try to correct for confounders, they must, as Yoda would say. So here’s what they did:

Kripke and colleagues made a database that consisted of stuff they could find in their retrospective chart review: age, gender, BMI, marital status, ethnicity, smoking status (current, quit, or never being the categories, not a pack-year-history burden of exposure), and alcohol use (which was also a binary “yes” or “no”; the burden of exposure was not quantified). Conceptually, these are not mere details. The evidence is pretty conclusive that cigarettes and alcohol are carcinogenic, in a dose-responsive pattern. The more you use, the greater the risk.*

*For a study that purports to shed light on causality of cancer, this omission should be considered quite warty.

And wouldn’t you know it? After comparing the FruFru eaters with the non-eaters on all of these elements, the authors discovered that the two groups weren’t similar at all.

The group who had self-selected hypnotic therapy was sicker. Much sicker.

People who were in the self-selected hypnotic group were significantly more likely to carry diagnoses of asthma, cerebrovascular disease, cardiovascular disease, diabetes, heart failure, hypertension, obesity, and reflux. The prevalence of most of these problems was essentially doubled in the hypnotic group, compared to controls.

 To this point, I’ll only make the observation that all of the above-mentioned problems are known to co-migrate with Sleep Apnea, the existence of which was never assessed in either group. This is also a good time to point out that one of the most common symptoms of Sleep Apnea is…you guessed it!...insomnia.

In middle aged women, it happens to be the most common symptom of all.

Large study groups are great, but even a huge retrospective database like Kripke’s doesn’t overcome the logical limitation that the groups were different from the start. The group exposed to sleeping pills was self-selected, and objectively sicker than the group who didn’t seek and obtain hypnotics.

Given their comorbidities, odds are, more undiagnosed Sleep Apnea was present in the hypnotic group, a condition which has been shown in two large Population-Based Prospective Observational Cohort studies (Sleep Heart Health Study, Wisconsin Sleep Cohort) to have a likely causal association with cancer. In light of the above, I believe the following conclusion is more in keeping with what we know and what we don’t know.

Say it loud, and say it proud:

Amongst an underserved rural population who have limited access to specialty services and who historically don’t come in for preventative care (only for complaints), those who complain about their sleep satisfaction enough to receive sleeping pills from their providers are more likely to be medically sick, and are more likely to die of cancer in the next four years. The ones who needed pills more frequently had the worst overall prognosis.

There. That felt good.

Sleeping pills can be dangerous, folks. Don’t misinterpret my comments to be a ringing universal endorsement to take medication.

But, for Pete’s sake!

Please don’t tell people they cause cancer.

The science is not there yet.

I’ll close with this: it’s OK to love this book. Just put a sticker on the front to remind you to be empowered.

Recommended Reading

(1) Kripke DF, Langer RD, Kline LE. Hypnotics Association with Mortality or Cancer: A Matched Cohort Study. BMJ Open 2, no. 1 (2012): e00850. The fateful paper itself.

(2) Brandt J, Leong C. Benzodiazepines and Z drugs: An Updated Review of Major Adverse Outcomes Reported in Epidemiologic Research. Drugs R D (2017) 17: 493-507. This is a sane and well-written paper that explains everything I just talked about, just in more academic prose.

___________________________________

ADDENDUM 16 JUNE 2022

I’ve never written an addendum to an essay in my life. Essays can be revised, but I only think about addendums for office notes.

The reason for my habit change is this: the author of my frenemy wrote me a letter!

A really beautiful and thoughtful letter. An unusually kind letter.

That kind of thing justifies an Addendum.

In his letter, Dr. Walker explained that he is no longer amplifying the message in Kripke’s paper, for precisely the reasons described above, and because he had picked up on the signal of how triggering that message can be.

I mean. You had me at hello.

I previously mentioned that there was one sentence that turned a magnificent book into my frenemy. You recall how I mentioned it was only part of a sentence?

Well, it’s time for the big reveal. The first part of the sentence was this: “But,”

The sentence that started the paragraph before that one? “Do these findings prove that sleeping pills cause cancer? No. Absolutely not”,

I left that part out of my essay before, not to be misleading, but because it seemed like nobody ever remembered that it was in there. The only part anybody remembered was the cancer and the death.

So, I’m glad Dr. Walker reminded me of the amount of thought he put into writing that sentence, hoping that it would allow a balanced read. It brings me joy that I can put the book down, and remember that sentence more clearly.

As for my frenemy?

Perhaps this is the beginning of a beautiful friendship, after all.

Recommended Reading

Walker, Matthew. Why We Sleep. Scribner. New York, NY. 2017