Dear reader, there is a question in pharmacology that sounds simple but is not: why does the same tablet, at the same dose, in two patients of the same age and weight, produce a cure in one and no effect in the other?
For fifty years, the answer has been sought in the human genome, and pharmacogenomics has delivered real results. But much of the variability was never in our genes. It was in the genes of the passengers, and the drug met them first.
The first clear case appeared in the early 1980s with digoxin, the old heart medicine derived from foxglove. In roughly 1 patient in 10, much of the dose was converted to cardio-inhibitorymetabolites, and the reason was a gut organism now called Eggerthella lenta, which reduces digoxin to an inactive product before it can be absorbed. Patients carrying a particular strain were, in effect, treated with nothing. Decades later, researchers identified the bacterial genes responsible and showed, in gnotobiotic mice, that whether the bug inactivates the drug depends on diet, because arginine, abundant in dietary protein, represses those genes. A cardiac drug's efficacy, mediated by a bacterium, is modulated by dinner.
For a long time, this looked like a curiosity. Then, in 2019, a Yale group led by Andrew Goodman decided to find out how common the phenomenon was. They exposed 76 strains of human gut bacteria, drawn from the major groups that colonize the gut, to 271 orally administered drugs and measured what happened. In culture, roughly two-thirds of those drugs were chemically modified by at least one strain, and each strain handled between eleven and ninety-five of them. This was not a curiosity. It was a second liver, unmapped, sitting upstream of the first, with a different set of enzymes in every patient.
The direction of the effect varies, which makes this clinically dangerous rather than merely interesting. Sometimes bacteria can render a drug ineffective, as with digoxin. Sometimes they are required to activate it: sulfasalazine, used in inflammatory bowel disease, is a prodrug that only becomes therapeutic when gut bacteria cleave it to release its active half, so a patient whose flora has been flattened by antibiotics may take the tablet faithfully and receive nothing. Sometimes they reactivate a drug the liver has already disarmed, the mechanism behind the severe diarrhea caused by the chemotherapy agent irinotecan: bacterial enzymes strip off the liver's protective modification and release the toxic metabolite onto the intestinal lining.
The most consequential example may be levodopa, the mainstay of Parkinson's treatment. Gut bacteria, including strains of Enterococcus faecalis, decarboxylate levodopa in the intestine before it can reach the brain, and the carbidopa co-administered precisely to prevent peripheral decarboxylation does not inhibit the bacterial enzyme. The abundance of those organisms varies enormously between patients. Some of what neurologists have long treated as unexplained variation in dose requirement is a microbial census.
Which brings me to a drug interaction almost nobody writes on a chart. If antibiotics reshape the microbial community, and that community determines how much of a drug survives to be absorbed, then every course of antibiotics is a pharmacokinetic event for every other medicine the patient takes. This is not speculation. In that same 1981 study, five days of erythromycin or tetracycline abolished digoxin inactivation in the patients who had been inactivating it, and their serum digoxin concentrations rose as much as twofold. The experiment has been sitting in the New England Journal of Medicine for forty-five years. We do not think this way, and we should.
There is a regional dimension I find hard to overstate. The gut microbiome is shaped by diet, climate, antibiotic exposure, and mode of birth, and differs substantially between populations. Nearly every pivotal trial that set the doses of the medicines we prescribe here was conducted in North America, Europe, or East Asia, in people whose microbial communities were shaped by different diets and medical practices. Adopting those doses unchanged assumes the second liver is the same everywhere. What evidence exists points the other way. When Lindenbaum's group compared populations, roughly a third of Americans tested in New York inactivated digoxin, against one in seven Bangladeshis tested in Dhaka, and the difference held regardless of where the subjects were living at the time. In a region with high antibiotic consumption and a diet that has changed within a single generation, the assumption deserves examination rather than inheritance.
The field now has a name, pharmacomicrobiomics, and its promise differs in kind from that of genomics. Your genome is fixed and can only be read. Your microbiome can be read and, in principle, changed. That makes it the more actionable of the two, and it receives a small fraction of the attention.
Dear reader, we have built medicine on the comforting assumption that a milligram is a milligram. It is not. Between the tablet on your tongue and the molecule in your bloodstream stands a community of organisms that is unique to you, that changes with what you eat and what you were last prescribed, and that is entirely capable of deciding that your medicine is not a medicine at all.