A Nature Study Says Sugar Makes Antibiotics Hit Your Gut Harder. Our Database Already Has a Sugar Column, and It's Always Zero.

Researchers at NYU Langone, working with City of Hope and Memorial Sloan Kettering, published a study on September 30th in Nature that pins a real mechanism on something that used to be a vague “eat healthy while you’re sick” suggestion. The team tracked every single food item in 9,419 meals eaten by 173 hospitalized cancer patients, a population that gets antibiotics constantly, then measured what happened to their gut microbiomes afterward. The finding: for every additional 100 grams of sugar a patient ate in the 48 hours before starting antibiotics, their gut’s bacterial diversity dropped an extra 21 percent on top of whatever the antibiotic itself did. Mouse experiments backed up a mechanism for why. Mice given both the antibiotic biapenem and dietary sucrose saw their Enterococcus population, a genus behind some genuinely hard-to-treat infections, balloon 16.3-fold by day three and 33.4-fold by day six. Sugar wasn’t just sitting there. It was feeding the exact bacteria that survive the antibiotic and take over the space the antibiotic clears out.

What makes this different from the usual “sugar is bad for your gut” coverage is how narrow and specific the actionable window is. This isn’t “eat less sugar forever.” It’s “the 48 hours before you start a course of antibiotics is a moment when your sugar intake has an outsized, measurable effect on an outcome that actually matters.” That’s a genuinely useful thing for a coach to know, and it’s the kind of narrow, well-timed nudge an app that already logs what you eat should, in theory, be able to deliver. Acting on it only requires an app to know two things: that a user is about to take or is taking antibiotics, and how much sugar they’re actually eating. So I checked whether Bite can do either one.

The medication side is the simpler problem, and I already half-knew the answer from an issue I wrote back in August about GLP-1 drugs. Bite has exactly one place to disclose anything medication-related: a single onboarding checkbox, “medical condition or medication affects diet/weight,” that feeds a HealthSafetyReason enum with no follow-up question asking what the medication actually is. I went looking this week for anything more specific, a GLP-1 flag, a drug class, anything, on the theory that if we’d built one for weight-loss drugs we might have a pattern to extend to antibiotics. There isn’t one. The only place the string “GLP” appears anywhere in the codebase is a line in an internal ad-strategy document telling the marketing team to imply GLP-1 relevance in Instagram creative without naming the drug outright. There was never a real field. Checking that one box, for any reason at all, just flips a user into clinician_required mode, which stops Bite from calculating or saving any macro target for them, full stop. There’s no narrower setting for “this person has a time-limited reason to watch one number closely,” only the one blunt switch that turns off coaching entirely. A user starting a five-day course of amoxicillin has no way to tell us that, and if they tried to via the only box available, we’d stop giving them numbers altogether rather than the one number that would matter this week.

The sugar side turned out worse than I expected, and in a more specific way than “it’s just not summed.” Sugar does get fetched: Open Food Facts returns it on a barcode scan, Spoonacular returns it on a recipe, and our NutritionLabel component dutifully renders a “Total Sugars” row right next to Dietary Fiber, the nutrient I found in the same unsummed state two issues ago. But sugar’s problem starts a step earlier than fiber’s did. The data type we use for an actual logged meal entry, the object that represents a food once it’s been added to today’s log, has four fields: calories, protein, carbs, fats. No sugar key exists on it at all. It isn’t that sugar gets fetched and then dropped before reaching a total; it never makes it into the log in the first place, so there’s nothing to total even if we wanted to.

Then I found something stranger. The Postgres table that stores daily nutrition totals has real fiber and sugar columns, decimal fields sitting right next to calories and protein, which means someone, at some point, explicitly planned for this. But the one function that writes to that table, the nightly job that materializes a day’s totals, builds its insert from exactly four fields: calories, protein, carbs, fat. Fiber and sugar are left out of the object entirely, so those two columns sit at their database default of zero, permanently, for every user, every day. That’s not a display gap. That’s a column someone built space for and a pipeline that was never connected to it.

I don’t have a tidy fix to report, and the pattern across these last few issues, sodium, fiber, now sugar, tells me the honest fix isn’t patching one more field. It’s that calories, protein, carbs and fat became the four numbers this entire product category agreed to compete on, and everything else gets bolted onto a display screen without ever being plumbed into the pipe that actually sums a day. A study with a 48-hour window and a specific number attached to it is exactly the kind of thing that pipe would need to exist for. Right now it doesn’t, on either end: we can’t ask what medication someone’s on without shutting off their coaching, and we can’t tell them their sugar total even if they asked.

Three Quick Bites

1. CVS Caremark added Zepbound back to its standard commercial drug list on October 1st, ending a roughly 15-month exclusion that had sparked patient protests and lawsuits, and added coverage for the new oral GLP-1 pill Foundayo (orforglipron) as well. The company says eligible patients will be able to see transparent, same-day cash-pay pricing in the CVS Health app by the end of the year. Coverage still depends on what each employer’s plan actually elects to follow. CVS Health

2. The FDA classified a recall of nearly 1.7 million pounds of United, Good & Gather and Crystal brand light brown and powdered sugar as Class II on September 21st, after the Minneapolis-based producer found possible undeclared wheat contamination. Wheat is one of the FDA’s nine major food allergens, and the affected sugar shipped to eight states. It’s a strange thing to be the second sugar-related recall or finding in this issue, but it’s a reminder that the label on a bag of sugar can fail you in ways that have nothing to do with the sugar itself. Newsweek

3. A new Connecticut law regulating AI companion chatbots took effect October 1st, requiring operators to run evidence-based self-harm and suicide detection, refer at-risk users to resources like the 988 Lifeline, and disclose plainly that users are talking to AI. It’s the same category of law as California’s Adam’s Law, which Collins wrote about last month after finding Bite’s own chat safety filter has no self-harm category at all. That’s still true as of this week. CT Mirror

Tool of the Week: SugarScience’s Hidden in Plain Sight

Since this issue is about a nutrient that’s genuinely hard to see coming, this week’s pick is UC San Francisco’s free guide to where added sugar actually hides. SugarScience.ucsf.edu’s “Hidden in Plain Sight” breaks down how added sugar shows up in roughly 74 percent of packaged foods, including savory ones like bread, pasta sauce and salad dressing, and lists the roughly five dozen different names it can go by on an ingredient list, from the obvious (sucrose, high-fructose corn syrup) to the easy to miss (barley malt, dextrose, rice syrup). It won’t total your week for you, which is exactly the gap this issue is about, but it’s a fast way to stop a label from hiding something from you. SugarScience: Hidden in Plain Sight

Thanks for reading. If you want a coach that’s honest about the number it isn’t tracking yet, Mr BITE is free on iOS and Android.

Michael, building Mr BITE