Notes on Foxglove
Measurement, Dose, Antidote, and Value—a 40-year Anniversary
In 1785, a Birmingham physician named William Withering published a slim book with a long title: An Account of the Foxglove, and Some of Its Medical Uses. Tucked into its introduction is a short story about a travelling Yorkshire tradesman. Withering found him vomiting without pause, his vision blurred, his pulse down to forty beats a minute. The man's wife, it turned out, had stewed a large handful of green foxglove leaves in half a pint of water and given him the liquor in a single draught to cure his asthma. He narrowly survived. Withering's verdict is one of the better sentences in the history of pharmacology: “This good woman knew the medicine of her country, but not the dose of it.”
The foxglove, Digitalis purpurea, has always been two plants at once. Its folk names, dead man's bells among them, warned of a poison. Its place in cottage medicine promised a cure. The difference between the two was never in the plant, though. It was in the quantity, and in whether anyone could tell how much was too much before it was too late.
Measurement
A medicine with a narrow margin between help and harm (its so-called therapeutic index) is not mainly a problem of discovery. It is a problem of engineering. It asks three things of us: that we find the right dose, that we measure what the body actually contains, and that we have some way back when the first two fail. This is the story of how one of the oldest drugs in the pharmacopeia acquired all three over roughly two hundred years of practice. It also happens to be the story behind the name of our group.
Dr. Withering came to the plant in 1775 when he was asked his opinion about a family recipe for dropsy, the old name for the swelling we now associate with heart failure. He was told the recipe had long been kept secret by an old woman in Shropshire, and that it contained twenty or more herbs. Withering was a serious botanist and he concluded that the active ingredient in the mix must be the foxglove. The woman supposedly behind the recipe has since acquired a name, “Mother Hutton,” along with a romantic legend of an herbalist robbed of appropriate credit. Both the particular credit and the robbery appear, however, to be inventions of a 1928 Parke-Davis advertising campaign. As the cardiologist Dennis Krikler showed in 1985, Withering's own account is the only actual source behind how he came to learn of digitalis, and it fails to include any mention of the mythic Mrs. Hutton. Readers of last month's essay on undiscovered public knowledge will recognize a pattern here, though: valuable knowledge (i.e., the family recipe) sitting in plain sight, waiting for someone able to read and understand it.
What Withering did next is why we still remember him. He saw the poor of Birmingham without charge for an hour each day, between two and three thousand people a year, and he used that clinic to learn the attributes of foxglove as a drug. At first, by his own admission, he gave doses “very much too large,” on the theory that a diuretic worked best when it made the patient sick. He came to see nausea as a warning rather than a mechanism. He abandoned boiled decoctions, whose strength depended on whoever did the boiling, in favor of dried and powdered leaf, gathered only when the plant was in flower because potency varied with the season. “The more I saw of the great powers of this plant,” he wrote, “the more it seemed necessary to bring the doses of it to the greatest possible accuracy.” His working rule was to give the drug until the body showed the first sign of its action, and then stop.
Withering could easily have published only his successes. Instead, he reported every case in which he had prescribed the foxglove, “proper or improper, successful or otherwise,” because “Truth and Science would condemn” the practice of choosing only the flattering examples. He even recorded a patient who died suddenly before taking a single dose, and asked whether, had she taken it, her death would not have been blamed on the medicine. Two centuries before anyone spoke of publication bias or confounding, a provincial doctor was guarding against both. Historians have observed that a clinician could learn to use digitalis safely with nothing but Withering's book as a guide.
Withering's book did not settle the matter. It was received with acclaim, but the use of digitalis remained sporadic through much of the nineteenth century, swinging between enthusiasm and fear. Withering himself predicted as much. He worried that the Edinburgh physicians who had adopted the drug would soon reject it again if they kept prescribing it in the “enormous doses” then fashionable in London.
The reasons were structural. Dried leaf is a variable product. As late as 1907, physicians were still debating how to standardize digitalis tinctures, and the best available answer was to inject them into frogs and see how much it took to stop a heart. More fundamentally, the doses that help and the doses that harm sit close together, and the margin moves with a patient's age, kidney function, and electrolytes. Withering's patients had taught him what toxicity looked like: sickness, a slowing pulse, disordered vision. One man who kept taking his infusion against instructions reported that every object appeared green to his eyes. But a bedside sign is a lagging indicator. By the time it appears, the harm has already begun.
Chemistry narrowed the problem without solving it. In 1930, Sydney Smith, a chemist at Burroughs Wellcome in London, isolated digoxin from the leaves of the woolly foxglove, Digitalis lanata. (The purple foxglove that Withering used yields mainly digitoxin, a close chemical cousin.) A pure molecule could be weighed, which meant a dose could be written in micrograms rather than leaves. Yet a weighed dose is not the same thing as a delivered dose. Two patients swallowing the same tablet can end up with very different amounts in their blood. Well into the 1960s, physicians still titrated digoxin largely by watching the heart rhythm and waiting for symptoms, much as Withering had done.
The missing tool came from an unexpected direction. In the 1950s, at the Bronx Veterans Administration Hospital, the physicist Rosalyn Yalow and the physician Solomon Berson developed a new technology, the radioimmunoassay. The idea is elegant: let the substance you want to measure compete with a radioactively labeled version of itself for a limited supply of antibodies, and read the result off the radioactivity. They first used it to measure insulin and published a detailed account of the method in 1960. It was astonishingly general. Anything an antibody could be raised against could, in principle, be counted at concentrations no chemical method could reach. Yalow and Berson also declined to patent it, which helps explain how quickly it spread through medicine. Yalow received the Nobel Prize in 1977; Berson had died in 1972 and thus was unable to share in the honor.
In 1969, Thomas W. Smith, a young cardiologist at Massachusetts General Hospital, working with Vincent Butler and Edgar Haber, applied the approach to digoxin. Their paper in the New England Journal of Medicine, “Determination of therapeutic and toxic serum digoxin concentrations by radioimmunoassay,” gave clinicians, for the first time, a practical way to see how much of the drug was actually circulating in a patient's blood.
Dose
A follow-up study by Smith and Haber the next year showed what the number meant. Among patients with rhythm disturbances caused by digoxin, the average serum level was 3.7 nanograms per milliliter. Among patients doing well on maintenance doses, it was 1.4, even though both groups were taking comparable daily doses. Ninety percent of patients without toxicity had levels of 2.0 or below, and 87 percent of the toxic patients had levels above it. The overlap, roughly 1.6 to 3.0, was as instructive as the separation. The assay did not replace clinical judgment, and its authors were careful to say that the level was useful when weighed in the clinical context. Readers of our note on the error bar will recognize the shape: a measurement that sharpens a decision precisely because it is honest about the zone in which it cannot decide.
Measurement changed the character of the drug. Dose no longer had to be inferred from symptoms after the fact. It could be checked before harm was done, adjusted for the particular patient, and studied across populations. The plant that Withering had tamed through patient observation was now governed by measurements and numbers.
But while measurement tells you when you have gone too far, it does not bring you back. Digoxin binds tightly in heart tissue and leaves the body slowly, so a severe overdose, whether by accident or by intent, was often fatal. Treatment meant pacing the heart and giving antiarrhythmic drugs while waiting for the poisonous effects to clear.
Antidote
The antibodies that made the assay possible suggested a way out. If an antibody could capture digoxin in a test tube, perhaps it could capture it in a body. Early animal work pointed in that direction: rabbits immunized against digoxin survived doses that killed every unprotected animal. In 1971, Donald Schmidt and Vincent Butler reported that digoxin-specific antibodies reversed established toxic arrhythmias in dogs. Six of eight treated animals returned to a normal rhythm within ninety minutes, while nearly all of the untreated dogs died within a day. That same year, John Curd, Thomas W, Smith, Jean-Claude Jaton, and Edgar Haber described purifying the antibodies, raised in sheep, and showed that fragments produced by digesting them with the enzyme papain could neutralize digoxin's effects on heart cells and on the intact heart. Fragments mattered. Smaller than whole antibodies, they acted faster and were less likely to provoke an immune reaction of their own.
The first successful clinical use of digoxin-specific Fab fragments was reported in 1976. By 1982, Smith and colleagues could describe 26 patients with advanced, life-threatening toxicity that had not responded to conventional treatment. Every one responded at first. Twenty-one recovered fully. Four died after prolonged periods of low blood pressure that had begun before treatment. In the fifth death, the supply of fragments on hand was simply too small to neutralize a massive deliberate overdose.
That last case points to the quiet elegance of the whole design. The dose of antidote was calculated to match, molecule for molecule, the amount of drug in the patient's body, an amount estimated from the history or from the serum level. Reversal depended on measurement, which in turn had depended on a precise understanding of dose. The three problems Withering had wrestled with had become a single system. A multicenter study eventually extended the experience to 150 patients, from a newborn treated within hours of birth to a 94-year-old, with no allergic reactions. The FDA licensed the product as DigiBind in April 1986, forty years ago this year, and approved a second version, DigiFab, in 2001.
Value
There is a final chapter, and it is a humbling one. Making a drug safe to use does not prove that it is worth using. Two centuries after Withering, digoxin was still prescribed to millions of people with heart failure on the strength of clinical experience rather than a definitive clinical trial. In the early 1990s, the National Heart, Lung, and Blood Institute and the Department of Veterans Affairs sponsored a test of exactly that question. The Digitalis Investigation Group (DIG) randomized 6,800 patients to digoxin or placebo and followed them for about three years.
The results, published in February 1997, were clear and a little deflating. Digoxin did not reduce overall mortality: roughly 35 percent of patients died in each group, 34.8 percent on digoxin and 35.1 percent on placebo. It did reduce hospitalizations, both overall and for worsening heart failure. Later analyses of the same data, including by sex, kept refining the picture, often within wide confidence intervals. Thomas Smith was one of the four investigators who took responsibility for the paper on behalf of the trial group. The scientist who had helped make digoxin measurable and reversible also helped establish the limits of what it could do.
Read closely, the verdict is not far from what Withering himself had claimed. He never promised a cure for heart disease. He promised relief of dropsy in many patients, and he warned that no general conclusions should be drawn from cases that were, by their nature, the most hopeless. He ended the preface of his book with a sentence that any trialist could adopt as a motto: “Time will fix the real value upon this discovery.” It took 212 years, a new class of measurement, and one of the largest randomized trials of its era, but time did eventually affix value.
This is also what measurement buys. It is not only a safety device. It is the precondition for the kind of honest accounting that tells us when a familiar remedy is doing less than we believed, which is its own form of the intellectual debt we wrote about in one of our earliest Notes.
Step back from the foxglove and the pattern is general. Powerful interventions become governable when three capabilities line up: a disciplined understanding of dose, a way to measure what is actually happening inside the system, and a way to undo the intervention when it overshoots. Each without the others is fragile. Dose without measurement is guesswork. Measurement without reversal is a well-instrumented emergency. Reversal without measurement cannot be dosed.
The pattern recurs wherever medicine handles sharp tools. When the anticoagulant dabigatran reached the market in 2010, it arrived without a specific antidote. Five years later, the FDA approved idarucizumab, an antibody fragment built on the same logic as DigiBind, as the first specific reversal agent for a drug of its class. The manufacturer expected it to be used rarely, and said so. Its real value was the assurance it gave physicians and patients in choosing the anticoagulant in the first place. An antidote changes the worth of the drug it reverses. Naloxone has played a similar role for opioids, and the logic is now being tested on interventions far more permanent than a pill. Gene editing, engineered cells, and long-acting implants raise a hard version of the old question: what is the equivalent of an antidote for something designed to last a lifetime?
There is an economic lesson here as well. Antidotes are almost by definition small markets. The first clinical series of digoxin Fab fragments included a death because there was not enough product on hand. Reversal agents protect far more people than they ever treat, which makes them a textbook example of the goods we described in Funding the Unfundable: valuable to society, chronically under-supplied by markets. Measurement tools have a public character too. It is worth remembering that radioimmunoassay spread as fast as it did in part because its inventors chose not to own it.
The same frame applies to the technologies dominating today's headlines. Much of the anxiety about artificial intelligence, and much of the work on making it safe, is an argument about dose, measurement, and reversal under another name. How much autonomy should a system be given? How would we know what it is actually doing? And what is our way back if it goes wrong? Those are Withering's questions, asked of a new and far less familiar plant.
When we founded Digitalis, the name was in part a nod to my father, whose work on the digoxin assay and on the antibody that became DigiBind is described above. He died in March 1997, a few weeks after the DIG results appeared. But the name was never only a tribute. It is also a compressed description of a method.
The foxglove's story runs from a secret kitchen recipe to a molecule that can be weighed, counted in the blood, and neutralized on demand. At each stage, progress came less from finding something new than from making something known governable: Withering's honest ledger of cases, Yalow and Berson's unpatented assay, a sheep antibody cut down to size, a trial large enough to tell the truth. That is what we mean by Engineering Health. The goal is not only to discover powerful things, but to make them safe to use, measurable in practice, and correctable when we are wrong.
Foxgloves still bloom each summer in dry, gravelly soil along English roadsides, much as Withering described them. It is worth remembering what the Yorkshire tradesman's wife was missing. She had the plant. She had the knowledge of her country. What she did not have was the dose, and the means to know when she had exceeded it. Most of the work of medicine, then and now, lives in that gap.♦
