But now suppose you’ve never thought of it that way, and neither has anyone else. And it’s not your job to run experiments to figure things out, and in fact that’s not the job of anyone you’ve ever heard of. Maybe you’re a doctor. Your patients come to you with various maladies, and you give them the best treatments you know of. Sometimes they recover, sometimes they don’t.

Would you invent the randomized controlled trial?

Uncontrolled

To improve your practice as a hypothetical historical doctor, you might seek the advice of more experienced doctors, study existing medical texts, or contemplate theories of anatomy and pathology. Tradition, authority, and pure reason can get you a long way. If you’re unsatisfied with these methods, you might decide that empirical observation and experiment is the best way to figure something out. But even then, you won’t necessarily arrive at the idea of a controlled trial.

Galen, the Greek physician who lived around 200 AD, is considered a key figure in experimental medicine. He studied anatomy by dissecting and vivisecting animals, and his experiments took the form of applying various surgical techniques to animals and seeing what happened. For example, consider his research on arteries. When you dissect a dead animal, you find that the arteries are mostly empty (because the blood has drained out), so people believed that the arteries were a system of tubes for conveying inhaled air to the rest of the body. They realized that when you cut the artery of a living animal, it bleeds, but they theorized that this was because the cut lets the air out of the artery, and then blood rushes in. Galen’s experiment was to expose a length of artery in a living animal, tie ligatures in two places, and then open the artery in between. There was, of course, already blood in this isolated section of artery.

This kind of experiment wouldn’t really benefit from a control. The question under investigation wasn’t fundamentally about the effect of the surgery, so a second animal standing by, not being operated upon, wouldn’t have been helpful. Galen’s surgical intervention was just his method of peering into the mysterious black box of the living body, and cleverly distinguishing the predictions of two different theories.

About fourteen centuries later came what we now call the Scientific Revolution in Europe. There was renewed interest in empirical observation and experimentation, there were innovations in precise measurement and mathematical modeling, and significant advances in our understanding of astronomy, optics, physics, chemistry, etc. These developments are foundational for modern science, but they don’t really prefigure the modern controlled trial. Early scientists aimed to observe nature, uncover its hidden regularities, and develop mathematical laws to describe and predict its mechanisms. An experiment was just a little corner of nature set up for close observation. A bronze ball on a wooden ramp, for example, allows you to take a hundred repeated measurements and conclude that “the spaces traversed were to each other as the squares of the times.”

Even when studying the efficacy of a drug, which we now consider the perfect occasion for an RCT, it’s possible to think very hard about what experiments to conduct and never come up with the idea of a controlled trial. In The Canon of Medicine (1025 AD), Persian physician and philosopher Avicenna laid out “the rules that must be observed in finding out the potency of medicines through experimentation.” The rules are all described in terms of Avicenna’s theory of the four temperaments: hot, cold, moist, and dry. For example, we are instructed to test every drug against two “contrary” conditions, such as a cold disease and a hot disease; if the drug works against the cold disease but not the hot disease, this shows that the drug has a hot effect. Other rules include: test drugs in patients with no comorbidities (because if the patient is suffering from two contrary conditions, we won’t know which one the drug is acting on), and test drugs in humans (because “the medicine might be hot compared to the human body and be cold compared to the lion’s body”). Although the justifications seem bizarre to us now, it’s a thoughtful list, and yet the concept of controls makes no appearance.

Gaining Control

The first proper controlled trial that I know of was conducted by the Persian physician al-Razi, and described in his Comprehensive Book of Medicine around 900 AD. He wrote: “So when you see these symptoms, then proceed with bloodletting. For I once saved one group by it, while I intentionally neglected another group. By doing that, I wished to reach a conclusion.”

You couldn’t ask for a clearer description of a controlled trial: treating one group and intentionally neglecting another in order to reach a conclusion. It is a bit awkward that the first controlled trial proved the therapeutic efficacy of, well, bloodletting. Perhaps that’s one reason that the story of James Lind’s test of cures for scurvy is a more popular touchstone in the history of controlled trials. According to Lind, in 1747 he took a group of twelve sailors suffering from scurvy, whose conditions were “as similar as I could have them,”1 and assigned them in pairs to six different treatments, including cider, sea-water, and citrus. The two sailors eating citrus fruits recovered best, which fits perfectly into our modern understanding that scurvy is a deficiency of vitamin C.

Another interesting early reference to a controlled test appears in Bencao Tujing, a pharmacopeia compiled and edited by the Chinese naturalist and engineer Su Song around 1060 AD. “It was said that in order to evaluate the effect of genuine Shangdang ginseng, two persons were asked to run together. One was given the ginseng while the other ran without. After running for approximately three to five li, the one without the ginseng developed severe shortness of breath, while the one who took the ginseng breathed evenly and smoothly.”

I love this example, because I think it beautifully demonstrates both the intuitiveness and the unintuitiveness of controlled trials. Clearly, the reader is supposed to consider this anecdote compelling evidence for the effectiveness of genuine Shangdang ginseng. There’s no explanation of why this constitutes good evidence; it’s just obvious. At the same time, the story is presented purely as a secondhand, one-off event. There’s no suggestion that anything else ever was or should be tested this way. It’s just a fact about ginseng.

One more early example of controlled trials: the work of Ambroise Paré, a French surgeon in the 1500s. He recalls in his memoirs that as an inexperienced battlefield medic, he cauterized gunshot wounds with boiling-hot “oil of elder,” as instructed by a medical text and by his fellow surgeons. But then he ran out of oil of elder: “At last my oil lacked and I was constrained to apply in its place a digestive made of the yolks of eggs, oil of roses and turpentine. That night I could not sleep at my ease, fearing by lack of cauterization that I should find the wounded on whom I had failed to put the said oil dead or empoisoned, which made me rise very early to visit them, where beyond my hope, I found those upon whom I had put the digestive medicament feeling little pain, and their wounds without inflammation or swelling having rested fairly well throughout the night; the others to whom I had applied the said boiling oil, I found feverish, with great pain and swelling about their wounds. Then I resolved with myself never more to burn thus cruelly poor men wounded with gunshot. [...] See how I learned to treat wounds made by gunshot, not from books.”

Paré didn’t set out to do a controlled trial, but apparently the results of running out of oil were so clear, dramatic, and unexpected that he took the findings to heart. He may also have picked up the idea of a controlled trial more generally, because elsewhere he describes a deliberate within-subject trial of treating burns with onion paste: “I applied onions to one half of his face and the usual remedies to the other. At the second dressing I found the side where I had applied the onions to have no blisters nor scarring and the other side to be all blistered; and so I planned to write about the effects of these onions.”

A common thread in the work of al-Razi, Paré, and Lind is that they found themselves in situations where many people were simultaneously suffering from the same ailment: medieval Islamic hospitals, a battlefield of soldiers with gunshot wounds, a shipful of sailors with scurvy. Other contexts for practicing medicine, like working in a small town, or serving as a personal physician in a powerful household, wouldn’t provide this opportunity. Of course this doesn’t entirely explain the historical paucity of controlled trials, because the world is full of other kinds of opportunities to conduct them. Paré’s onion paste trial and the ginseng running race are examples that could have been much more common, and you can easily imagine controlled tests of anything from the best bait for catching fish to the best well for throwing curse tablets into.

Then again, maybe some such tests were carried out, and lost to history. Another common thread in the work of al-Razi, Paré, and Lind is simply that they wrote down their experiences and their texts have survived to the present day.

Stay tuned for more installments from the history of the RCT, covering Peruvian bark, imperial decrees, cadaverous particles, and, eventually, randomization.

The one-sentence answer to the first question is: in rare cases, honey can cause infant botulism. Arguably that’s all you need to know – it sounds bad, it is bad, your baby doesn’t need honey and probably doesn’t even know what honey is; don’t give them any till their first birthday. But, obviously, I wanted to know more!

The toxin and the spore

Botulism is the disease caused by a toxin called botulinum. The toxin interferes with nerve and muscle function, causing a kind of floppy paralysis. In severe cases, the paralysis can progress to respiratory failure and death.

You might know botulinum toxin under the name Botox – injections of botulinum toxin for cosmetic or medical reasons. The idea of Botox is that a controlled dose of the toxin relaxes and immobilizes muscles, which can smooth out wrinkles, or help with painful muscle spasms, migraines, lazy eye, and a few other conditions.

You might also know something about botulinum toxin if you’re into home canning. The toxin is produced by bacterial spores called Clostridium botulinum. The spores themselves are not harmful, until they end up in the right environment to multiply and produce botulinum toxin. The moist, low-oxygen, low-acid, moderate-temperature conditions of some kinds of canned food can be the right environment.1

And, unfortunately, it turns out that a baby’s intestinal tract can also be the right environment. If a baby ingests C. botulinum spores, they sometimes colonize the baby’s intestines and produce botulinum toxin from the inside. This almost never happens to older children or adults. So that’s the special risk for babies: ingesting the toxin is dangerous for everyone, but babies are also endangered just by ingesting the spore.

The numbers

We have good data about infant botulism diagnoses in the US, because botulism in all its forms is a ‘notifiable disease,’ meaning that doctors are required by law to report all cases to public health officials. In 2015-2019 (the most recent five years for which I found CDC summaries), about 150 babies per year were diagnosed with infant botulism. This is 3.8 cases per 100,000 births, or 0.0038% of all babies.2

Babies usually aren’t tested for infant botulism unless they are already seriously ill, so these 150 cases per year are quite severe: more than 99% are hospitalized, and about 20% get intubated.3 But, fortunately, they usually recover well. Only one baby died from infant botulism in the US during these five years.

Outside the US, infant botulism diagnoses are even rarer. This is probably due to less testing and less awareness among doctors, but it’s also possible that the US genuinely has a higher rate of infant botulism than most other countries.

Let’s get back to the topic of honey. In the US, about 4% of infant botulism patients were fed honey before their illness. Now, this clearly shows that honey is not currently a major cause of infant botulism in the US. However, that’s in part because parents dutifully avoid honey! In the first few years of infant botulism diagnoses, before the no-honey recommendation was widespread, about 33% of patients had been fed honey. And in Europe, where the recommendation still isn’t as well-known as it is in the US, about 28% of patients have been fed honey.4

Based on this, we can roughly estimate that without parents avoiding honey, we would see an additional 55-60 cases of infant botulism in the US every year.

These are not enormous numbers. As I see it, I might as well avoid feeding my baby honey, since I find it easy to do so. But if, for example, your baby eats something that you later find out was sweetened with honey, that’s really nothing to worry about.

The history

Infant botulism is a pretty recent discovery! It was first recognized in 1976, in California, when a two-month-old boy became weak and unresponsive. He was referred to a UCSF hospital, where, his case report reads, “He lay without movement.” His doctors considered everything from heavy metal poisoning to snakebite, including botulism, which didn’t make sense epidemiologically (he was a baby, and hadn’t had solid foods, and none of his family members were sick!) but did match his symptoms. They sent blood and stool samples for testing, and the stool sample was positive for botulinum toxin and spores.

The same year, a three-month-old boy with similar symptoms was also referred to UCSF. His doctors also had serum and stool samples tested for botulism, and again the stool sample contained botulinum toxin.

Both of these babies recovered fully with supportive hospital care, without any botulism-specific treatments. But meanwhile, at the microbial diseases testing laboratory run by the California Department of Health, interest was piqued. They had been sent botulinum-positive samples from two separate young babies, an age group in which botulism was unknown. Now alert to the possibility, doctors affiliated with the Department of Health identified three more cases within the year.5 They also started looking for the sources of the infection. They tested dozens of baby foods and supplements from the households of the five patients, and they didn’t find any evidence of botulinum toxin, the normal cause of botulism. However, they did find C. botulinum spores, in honey that had been fed to one of the babies.

Over the next few years, more cases of infant botulism were identified, and for a handful of patients, sources of C. botulinum exposure were identified too. A CDC report from January 1978 explains that out of the 58 cases known so far, one was linked to household dust, one to the soil in the household’s yard, and four to honey. And so, despite being linked to only a small minority of cases, honey emerged as the most commonly identified source of exposure, and also by far the most avoidable (I mean, good luck never exposing your baby to “household dust”).

Warnings from public health officials about babies and honey were imminent, but when I went to read the actual text of the first such warning, I was surprised to find that they had been pre-empted by the honey industry itself. In July 1978, the California Department of Health stated: “Since honey is not an essential food for infants, the California Department of Health concurs with the recent recommendation of the Sioux Honey Association that honey not be fed to infants under 1 year of age.” And indeed, in the New York Times a week earlier: “Honey and other raw agricultural products fed to infants less than a year old may produce botulism poisoning, suspected cause of some unexplained infant crib deaths, the world's largest honey‐producing cooperative [the Sioux Honey Association] said in statement issued yesterday.”

The connection with “crib death” (aka Sudden Infant Death Syndrome, or SIDS) is, I suspect, an important part of the story. In one of the first five identified cases, a baby suddenly stopped breathing in his mother’s arms. He was successfully resuscitated, but that experience led doctors to suspect that botulism could be a cause of SIDS, maybe the cause. Respiratory arrest as the mechanism of death was a match; the age distribution of infant botulism cases also closely matched the age distribution of SIDS deaths. Referring to the possibility that botulism was the cause of many previously-mysterious infant deaths, a leading infant botulism researcher told Science News: "We think we have a tiger by the tail.”

Over time, this emphasis has faded. A 1978 study from the California Department of Health found that 4.3% of SIDS cases (9 out of 211) tested positive for C. botulinum. Now, to me this seems quite significant. In 2022 in the US there were 2,660 sudden infant deaths with no known cause; 4.3% of that figure would be 114 deaths in one year, from a disease that’s recorded as causing only one death across five years! But I get the impression that perhaps researchers had hoped-slash-feared that botulism was causing a much higher proportion of SIDS deaths. Further research was inconsistent,6 and as far as I can tell, botulism is not currently considered a significant cause of sudden unexpected infant death.

Considering this history, I have a few ideas about the factors that contributed to the honey rule’s ubiquity:

• It aims to prevent a specific illness with a simple causal chain. For comparison, the American Academy of Pediatrics also doesn’t want you to give babies under 1 year old any juice. But that’s based on more general concerns about sugar, obesity, and cavities, not any immediate risk of disease. Immediate risk of disease is especially motivating!

• The link to honey was discovered right alongside the discovery of the illness itself. Pretty much as soon as people started considering this new disease, and how it could be prevented and treated, honey was a known cause of at least one case.

• The broader illness category (botulism) had a preexisting public health apparatus! There was an active surveillance system, and an attitude that every case of botulism was due to an identifiable food-handling mistake, which could and should be prevented in future. When infant botulism arrived on the scene, it inherited some of that aggressive preventative attitude.

• It was thought that infant botulism might be the mystery cause of some large proportion of SIDS deaths. It would have been very exciting to suddenly understand these deaths, and hopefully be able to prevent them.

• The honey industry was impressively proactive. The Sioux Honey Association, which produced a third of US honey at the time, warned the public not to give honey to infants before official public health organizations did; they also sponsored and published early research on the prevalence of C. botulinum spores in their own honey.

• Honey was, and still is, the clear frontrunner among known causes of infant botulism – it definitely isn’t the cause in a majority of cases, but it’s by far the most commonly identified avoidable source of exposure. “One food” is a a very manageable number of foods to avoid.7

• It was clear from the beginning that this was only a problem for babies. If honey was considered unsafe for everyone, we might have gone in the direction of changing its production or regulating its sale, rather than leaving the product as-is while trying to spread warnings about it. Plus, the honey industry would probably be less enthusiastic about the whole idea.

I wouldn’t claim that every one of these factors was necessary to produce the honey avoidance we see today, but I do think that multiple exceptional circumstances needed to combine. I can easily imagine a world in which the honey connection was not discovered as early or pursued as aggressively. In that world, we don’t have warnings on millions of jars of honey sold every year; maybe we have recommendations directly from pediatricians and a few government websites (more like the recommendation to use liquid formula). I hope you’ve enjoyed the story of how we ended up in our world instead.