Seven Gut Microbiome Tests, One Sample of Poop

One sample. One donor. One gut microbiome testing company. Three identical kits, mailed off and then waited on for two to eight weeks. Two reports came back clean. The third declared the gut unhealthy, flagged seven of ten functional categories as below average, and gently nudged the customer to do something about it. And no one at the company noticed: the dodgy report sailed through their own quality checks and posted to the account like any other.

That sample never came from a real customer. It came from a freezer at the National Institute of Standards and Technology, where a team of microbiologists had been buying up at-home gut microbiome kits to see what these tests actually deliver.

The premise is by now familiar from the advertising. Spit in a tube and learn about your ancestry; deposit something rather less pleasant on a swab and learn, supposedly, about the bustling ecosystem in your gut and what to eat to improve it. The microbes down there, the bacteria and fungi and viruses lining your gastrointestinal tract, really do seem tangled up with everything from mood to obesity to how well some cancer treatments work. So the appetite for a peek inside is understandable. The question NIST set out to answer was narrower and more awkward: if you send the very same gut to seven different companies, do you get the same gut back?

You do not. Not even close.

Putting the Tests to the Test

To run a fair trial you need samples that are identical, and this is where NIST had an unusual advantage. The institute produces and sells standard reference fecal material, painstakingly homogenised and characterised, the kind of thing used to anchor research into fecal transplants (which are exactly what they sound like, and which can help treat conditions like colitis and may even nudge cancer patients toward better responses to treatment). Leftover donor material from that effort was perfect for the job. The researchers ordered three kits each from seven companies, twenty-one in all, created accounts under personal email addresses so nobody knew they were being watched, and inoculated every kit from the same homogenised pool.

“A team of microbiologists at NIST recently put these tests … well … to the test using precisely measured stool samples we already had in our labs,” writes Stephanie Servetas, the microbiologist who led the work.

The reports came back weeks later, and they disagreed with each other as much as if they’d come from eight different people. The team split one identical sample among several companies, then ran eight separate donors through a single workflow. The noise from the testing process was as large as the real biological differences. For 17 of 18 common microbial genera, the variation from the testing matched or exceeded the variation between actual people. Only one genus, Haemophilus, held steady. In plain terms: the lab is shouting louder than the gut.

Why so much scatter? Because turning a smear of stool into a tidy pie chart is a long chain of fiddly steps, and bias can creep in at every link. Some companies asked customers to sample a whole bowel movement, others to swipe used toilet paper. Some had you drop the swab in a buffer, some wanted it shipped back neat. None packed any ice, so everything travelled at whatever temperature the postal service happened to provide. Then come the choices invisible to the customer: how the DNA is pulled out, which stretch of bacterial gene is read, how deeply, and which software decides what’s what. More sequencing did not even guarantee more answers. One firm reported over 600 species per sample on a shallow read; another, reading roughly five hundred times deeper, found fewer than a hundred.

Even the headline question of who is living in your gut went unsettled. Across all nine analyses, only three genera showed up in every single sample. And the test for Clostridioides difficile, a nasty pathogen you’d want flagged accurately, split the room: three of the seven companies said it was present, four said it wasn’t. Same sample.

What This Means If You Spit, Swab and Hope

None of which, Servetas is careful to say, amounts to a verdict of fraud or quackery. “This variability in results doesn’t mean I would advise people not to take these tests. In fact, I may try one myself one day,” she writes, adding that she’d quite like to test her whole family someday and compare notes (that’s a microbiologist’s idea of fun, apparently). The trouble is mostly that the science is young, barely twenty years old, and there’s still no agreed picture of what a “normal” microbiome even looks like. Many of the companies recommend dietary tweaks, which seems harmless enough. Some go further and sell their own probiotic supplements, the clinical evidence for which is thin.

The people most likely to buy these kits, it’s worth remembering, are often those wrestling with chronic gut problems and running out of options. For them a wobbly result isn’t a curiosity; it’s a possible wrong turn, or a delay before they see someone who can actually help. “I generally know how microbes interact with the body but not necessarily how the body responds to microbes,” Servetas writes, deferring, as she suggests the rest of us should, to a doctor.

So the advice is unglamorous but sound: read your report with a healthy dose of skepticism, and don’t overhaul your life on the strength of it without talking to someone who knows your medical history. The longer game, though, is more hopeful. Having finished the study, NIST went back to the companies, most of whom were game to keep working together, and the institute is now collaborating with the field to tighten up the measurements. The dream is to drag these tests closer to the boring reliability of a blood panel, where a result means roughly the same thing whoever runs it.

“Our ultimate goal is to help these tests become more like a blood test, so they can potentially be part of your regular health screenings,” Servetas writes. Until that day arrives, your gut, it seems, will keep telling slightly different stories depending on who you ask.

Frequently Asked Questions