A New Meta-Analysis Says the Vitamin D Pill You Pick Can Work Against You. Mr Bite Doesn't Track Vitamin D At All.

A meta-analysis out of the University of Surrey, with the Quadram Institute and the John Innes Centre, published October 7th in Nutrition Reviews settled a question that sounds pedantic until you see the number. Vitamin D supplements come in two forms, D2 and D3, and they’re usually sold as interchangeable. The researchers pooled 20 randomized controlled trials, 11 of them with usable data, comparing people who took D2 against people who took a placebo or nothing. The D2 group’s blood levels of D3, the form your body actually uses and the one most lab tests report on, came out an average of 17.99 nmol/L lower than the control group’s (95% CI, -25.86 to -10.12; P < .00001). That’s not D2 simply doing less. It’s D2 measurably suppressing the other form, in the opposite direction of what anyone buying a vitamin D pill thinks they’re doing. The researchers’ own read is blunt: for most people, D3 is probably the better buy, and the two aren’t the swap-in-swap-out options the supplement aisle treats them as.

Bite doesn’t manage supplements, but it does carry itself as the place that knows what’s in your food, so I wanted to know what “Vitamin D” means inside our own pipelines before I wrote a word about someone else’s pill bottle. I expected to find the usual shape of gap from this newsletter: a field that gets fetched and dropped before it reaches a total. What I found was flatter than that. Vitamin D isn’t dropped anywhere in Bite. It’s never asked for in the first place, on any of the three paths that could carry it.

Start with recipes, where Bite actually does show a vitamin today. spoonacularToRecipe() in src/utils/spoonacularMapper.ts pulls a recipe’s nutrition from Spoonacular’s API and builds it with a helper, getNutrientAmount(), that takes a list of acceptable names and matches them against whatever Spoonacular returns. The calls are explicit: ['Vitamin C'], ['Iron'], ['Calcium'], alongside the usual calories, protein, carbs, fat, fiber, sugar, sodium and saturated fat. Those three micronutrients make it all the way to the screen, too. RecipePreviewScreen.tsx builds a row for each one with its own icon, Citrus for Vitamin C, Magnet for iron, Bone for calcium, and renders them under the recipe’s extended nutrition section. Vitamin D never appears in either list. Nobody deleted it; getNutrientAmount() was simply never called with that name, so there was never a value to carry or a row to build.

Barcode scans are the blunter version of the same gap. openFoodFactsService.ts is what turns a scanned product into a logged food, and I read through its nutrient mapping end to end looking for anything, a vitamin, a mineral, a single micronutrient of any kind, that survives a scan. There isn’t one. The function pulls calories, protein, carbs, fat, fiber and sugar from Open Food Facts’ nutriments object and stops there. It’s not that Vitamin D specifically got left off a longer list. The barcode pipeline doesn’t have a concept of a vitamin at all, for any product, scanned from any box.

Then there’s the component that actually knows how to draw a “Vitamin” row with the word Vitamin in it. NutritionLabel.tsx renders Vitamin A, Vitamin C, calcium and iron as percentage-of-daily-value rows, laid out like the label on the side of a cereal box. I went looking for where it’s used, expecting a packaged-food detail screen. It’s exported from the app’s component barrel and nothing else. No screen imports it. It isn’t wired to any food in Bite, scanned, searched or logged. The one piece of UI in the entire codebase purpose-built to show a vitamin, Vitamin D included in spirit if not by name, isn’t connected to anything a user can open.

So the honest inventory is: one pipeline that asks for three specific micronutrients by name and could ask for a fourth as easily as adding ['Vitamin D'] to one line, one pipeline that asks for none at all, and one finished display component sitting unused. None of that changes what a user should do with this week’s study, ask a pharmacist about D3 over D2 if they’re buying a supplement at all. It just means if they logged that pill in Bite afterward, we wouldn’t have told them anything about it either way.

BITE of the Week

Photo-scan logging got a real accuracy fix this week. The scanner’s prompt used to ask for “plausible hidden oil” on every plate it saw, and it showed: 136 of 227 photo logs since September came back with an oil or butter row added, including on packaged snacks and steamed vegetables that had none, and people were telling us their logs ran high. Fat is now counted only where the photo actually shows it. Alongside that, every ingredient the scanner finds now gets its own grams and macros instead of one lumped total, the log card shows each row’s calories, and editing one ingredient rescales just that row rather than the whole entry. The scanner also reads a person’s past corrections back as context for their next photo, so a rename like “fries” to “air fried fries” applies immediately and a pattern of removals or portion cuts kicks in after a couple of repeats. Separately, a new OraclePT integration lets a connected coach view a client’s week, goals, plans and recipes from their own dashboard, with the app checking for updates and reloading automatically when something changes on either side; the coach’s ability to edit targets is still behind a flag we haven’t turned on broadly. All of this landed on main this past week, after our last store release (1.16.0, September 28th), so it’s shipping in the next build rather than live in your app today.

Three Quick Bites

1. California Governor Gavin Newsom signed AB 2244 in late September, creating the country’s first state-run “Non-Ultraprocessed Certified” seal, a voluntary label that manufacturers can apply for through third-party certifiers accredited by the state health department. Large grocery chains, those with over $10 million in annual sales carrying at least 25 certified products, will have to prominently display some of them once the program is live, though the department doesn’t have to start accrediting certifiers until June 2029. Good Morning America

2. A University of Reading-led review of 22 studies, published in Nutrition Research and covered this week, found that trans fats occurring naturally in dairy don’t raise the risk of heart disease, stroke or type 2 diabetes the way industrial trans fats do. Ten controlled trials covering dairy trans fat intakes from 1.3 to 13.2 grams a day showed no meaningful shift in blood lipids, and twelve cohort studies found no added long-term risk. It’s a reminder that “trans fat” on a label and “trans fat” in a research headline aren’t always the same molecule. University of Reading

3. An MIT study published in Nature and covered October 5th found a ketogenic diet had opposite effects in different parts of the gut in mice genetically prone to intestinal cancer: more tumors in the small intestine, fewer in the colon, with the risk traced to how fat drives intestinal stem cells to multiply rather than to ketone bodies themselves. The lead researcher was explicit that the mouse model is a stand-in for a rare inherited condition, not a reason for a healthy person to change course. MIT News

Tool of the Week: NIH’s Vitamin D Fact Sheet

Since this issue is about a nutrient we don’t track at all, the pick this week is the free, plain-language reference for the thing the study is actually about. The NIH Office of Dietary Supplements’ consumer fact sheet on vitamin D lays out the daily recommended amounts by age, which foods and how much sun exposure actually provide it, and the difference between D2 and D3 in terms a non-scientist can use before walking into a pharmacy. It won’t log anything for you, but it’s a better five minutes than guessing which bottle on the shelf is the right one. NIH Office of Dietary Supplements: Vitamin D

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

Michael, building Mr BITE