6.4 Do Probiotics Work?

Probiotic supplements are a $34 billion business, of which at least $6 billion is supplements, according to a May 2016 report by BCC Research. With unregulated health claims that range from the benign (“helps digestion”) to the fantastic (“A miracle cure!”), do they make a significant difference in my own gut microbiome? I tested myself to find out.

Among unhealthy people, there is evidence that, under a doctor’s care, probiotics can help with antibiotic-associated diarrhea and similar conditions in children or among people recovering from C. difficile infections.

On the other hand, a recent scientific review of all well-done studies of probiotics among healthy people couldn’t find evidence that probiotics made much difference compared to a placebo in randomized controlled trials. When the data-heavy web site FiveThirtyEight did a week-long series on Gut Science, including a detailed survey of what’s known about probiotics, they concluded: “There’s no evidence in humans, however, to support taking probiotics just to generally improve your gut health.”

A literature review by the US Agency for Healthcare Research and Quality found no safety issues in healthy adults, but there is suprisingly little research to show that the pills actually do anything. The independent lab Labdoor tests most common brands to see which actually contain the organisms claimed on the label, but I couldn’t find anyone who tests whether the body can absorb them or not. There have been a few peer-reviewed studies showing that some microbes in supplements can make it to the gut83, but these studies almost feel like special cases, where they try lots of microbes and only a few make it. It’s not clear that organisms in a typical off-the-shelf bottle of probiotics have ever been tested that way.

I’m especially interested in learning whether the probiotics in the supplements actually “stick” in my gut. Taking so many billion organisms in pill form all at once may very well show up in a single gut test result, but how do I know they’re not simply being flushed out of my system? Or worse, how do I know I’m not just crowding out something more important?

To find out, I tracked my microbiome daily while taking a high quality probiotic supplement, one that I received directly from the manufacturer. To be a fair test, one worth publicizing the brand name for better or worse, I’d want to try it out on multiple people, at multiple times. Because I didn’t do that this time, I won’t name the product other than to say that it’s from a “good” brand and well-recommended.

I took the supplement once per day for nine days. I would have continued for an even ten, but I was starting to feel uncomfortably bloated those last few days. While that’s an encouraging sign that the pill is working, I didn’t want to do anything to seriously mess up my gut. I’m doing this experiment for fun, and it won’t be fun if I get sick as a result.

Let’s look for at the overall abundances for the two genera that were in the supplements: Bifidobacterium and Lactobacillus. (Figure 6.14)

Percent abundance of key microbes (genus-level) found in the gut while taking a probiotic supplement.

Figure 6.14: Percent abundance of key microbes (genus-level) found in the gut while taking a probiotic supplement.

The red dots represents days when I took a gut sample after consuming the probiotic. Unfortunately, despite taking and submitting samples daily, several of my results just didn’t have enough reads to be useful. This chart shows only the days when I have a sample of at least 10,000 reads.

Even with that caveat, it’s hard to see clear-cut evidence that the pill had a significant effect. Yes, I have slightly more of those two taxa by the end of the experiment, but seriously, not that much more.

Let’s look at a longer time horizon (Figure 6.15).

Percent abundance of key gut microbes over a three month period after taking a probiotic supplement.

Figure 6.15: Percent abundance of key gut microbes over a three month period after taking a probiotic supplement.

Hmmmm, it seems the levels of those particular genera did increase a tiny bit at the end of the experiment, but there are plenty of other times on the chart where I see spikes too. In fact, the biggest increase happened in September when I was living it up in New Orleans, eating red beans and rice – and no probiotic pills.

Maybe my view of the microbe ecology, hoping to see results in only one or two genera, is too simplistic. We know that the gut is an ecosystem. If you add lots of one type of organism, maybe that affects the abundances and ratios of other microbes, all of whom are in constant competition with one another. Is there a way to tell overall how the microbes are changing?

Let’s apply an ordination analysis. Essentially this means we look at all the samples together and work out how different the samples are from one another, based on some “distance metric” that compares the abundances of specific microbes. If the abundances of two samples are roughly the same, or if they tend to rise and fall together, then we plot them next to each other, and vice versa if they are not well-correlated. There is a mathematical way to do this where we combine all these different correlations over and over and pick just the two that seem to matter the absolute most, which we’ll plot on a two-dimensional graph (Figure 6.16)

NMDS ordination (Bray-Curtis) of gut samples for ten months before and after taking probiotic supplements.

Figure 6.16: NMDS ordination (Bray-Curtis) of gut samples for ten months before and after taking probiotic supplements.

Hmm… that looks pretty random to me.

6.4.0.1 Other people

Since doing this experiment on myself, I’ve spoken with numerous others who’ve tried something similar: take a gut test, then start some type of probiotic supplement, and finally take another followup test a few days or weeks later.

Here’s an example, “Jeremy”, a healthy man in his 50s took this probiotic supplement: $42 for one month of pills:

Super Bifido Plus Probiotics contains high amounts of live Bifidobacteria and Lactobacillus.

Figure 6.17: Super Bifido Plus Probiotics contains high amounts of live Bifidobacteria and Lactobacillus.

and here’s the high level result:

Next let’s look just at the microbes reported to be in the probiotic pills. Jeremy has three samples of interest: (1) taken in mid-summer, a month before starting the probiotics, (2) right before the month of pills, and (3) after completing 30 days of faithful pill taking.

So although we do see a slight increase in both taxa, it’s hard to pin it solely on the probiotics. After all, he was at even higher levels a month before starting the pills.

Also, looking more closely at the read counts, I see that the final sample had the lowest, about 36,000 reads versus the 80,000+ reads of the other samples. When dealing with low-abundance bacteria, this can matter, but it’s impossible to tell precisely how much. The bottom line is that it’s possible that the probiotics had no effect whatsoever, and even if there was an effect, it was probably quite slight.

In fact, probiotics appear to have less of an effect even than travel. Here’s “Kevin”, a European man who moved to the United States.

Notice how Kevin’s microbiome shifted dramatically a month after arriving in the US. Soon after that, he began taking a probiotic supplement, but his gut – while different – hasn’t shifted as much as it did from the international move.

6.4.1 VSL

The most tested probiotic is VSL#3, and recently a woman sent me her microbiome test results after taking Optibac for 4 days prior to her second test. (Figure 6.18).

In this case the abundances of these microbes went up significantly. Is that a coincidence? Hard to tell from a single sample, but perhaps this probiotic is one that makes it through and shows up in the results.

Change in key microbe levels after a course of VSL#3

Figure 6.18: Change in key microbe levels after a course of VSL#3

6.4.1.1 (Tentative) Conclusions and next steps

It is very difficult to say with this analysis that the probiotics had any effect that is detectable by the uBiome Explorer test.

Further analysis required:

  • Consider other statistical analysis. Although the two strains contained in the probiotic pill don’t appear to cause a change in the gut microbiome results, are there other changes that can be detected statistically. Perhaps there are other taxa that show a significant change.

  • Other time horizons. Maybe the changes don’t happen immediately. Although at a high level, there doesn’t appear to a noticable lag in the levels of the probiotic strains, perhaps a more sophisticated data transformation would uncover something.

##Experiment: Gut Cleanse* {#experimentCleanse}

Microbiome experiments are complicated by the difficulty of holding everything constant. Even if you are careful with precise amounts of the same food and exercise, you are still dealing with your existing microbiome with all its uncertainties, making it difficult to tell precisely what caused a particular change. What if you could wipe the slate clean; start over with a completely new biome and just track that, along with precisely what you eat afterwards? What could you learn?

In this experiment, I tried exactly that, using a colon cleanse – the kind you do before a colonoscopy screening. By flushing all the bacteria from my system and carefully watching them grow back with day-to-day testing, I was able to get a better picture of the resilience of my microbiome.

The bottom line:

My gut microbiome recovers pretty quickly. Unlike antibiotics, which are known to cause long-term (and possibly permanent) changes, losing bacteria this way seems to matter only for a day or two. The missing microbes sprout right back just like a haircut. In two weeks it was as if nothing had happened.

Figure 6.19 is a broad, phylum-level look at how the various microbes shifted in abundance. As you can see, all of these high-level colonies were back to the same proportions that had been before the cleanse.

Overall phylum-level summary, from baseline (2 weeks before the cleanse) to CC (colon cleanse) to one month after CC.

Figure 6.19: Overall phylum-level summary, from baseline (2 weeks before the cleanse) to CC (colon cleanse) to one month after CC.

Even at the more detailed, genus-level, whatever shuffling occured didn’t look much different than the normal random variation I see in any month-long survey.

Overall Genus-level summary, from baseline (2 weeks before the cleanse) to CC (colon cleanse) to one month after CC.

Figure 6.20: Overall Genus-level summary, from baseline (2 weeks before the cleanse) to CC (colon cleanse) to one month after CC.

Amounts and ratios changed, but not the specific organisms. Of course I lost a bunch of bacteria – that was the point – but surprisingly I didn’t seem to gain anything really new, even after an aggressive attempt at re-seeding. I didn’t gain or lose a single phyla. Other than amounts and ratios, I had to dig down to the Class level of the biological hierarchy to find anything that was permanently lost, and even at the very fine-grained Genus level, only two taxa that had been regularly present beforehand were now extinct. (Holdemania and Methanomassiliicoccus).

My overall gut diversity spiked the day of the cleanse and then plunged the following day, but soon it was right back to normal (Figure 6.21)

Shannon level diversity measures before and after the cleanse period.

Figure 6.21: Shannon level diversity measures before and after the cleanse period.

A couple of weird microbers, at small amounts, made a brief appearance. I was intrigued by five new taxa that showed up just once, the day after the cleanse, and then disappeared. Maybe I found some that ordinarily get lost in the noise of the microbiome and only show up when the rest of the forest has been cleared. These are some hardy guys and I’m glad I know their names and can watch for them again: Abiotrophia, Bacillus, Catonella, Christensenella, Parvimonas.

It’s pretty hard to make a significant change. These days a little googling will find plenty of web sites, books, diets, and supplements that claim to “fix” or “change” your microbiome. I’m a healthy, reasonably fit adult, so I’m not as motivated as somebody with a specific health problem, but I thought simply popping probiotics and eating a variety of new and fermented foods would have a big effect. Nope. There are exceptions – new microbes will sprout when I drink homemade Kefir, or travel to China but it’s much harder than you’d think.

Of course, I’m not the first to study microbiome changes after a colon cleanse. A 2015 European study found increases in Dorea, which interestingly I found as well (Figure 6.22) .84

A rise in Dorea after a colon cleanse.

Figure 6.22: A rise in Dorea after a colon cleanse.

A more recent Japanese experiment Nagata et al. (2019)85 also found, like me, almost no difference after two weeks. They also used a mass spectrometer to study the specific metabolites present in each sample, but again, after two weeks it’s as though nothing had happened.


When I began this experiment I thought for sure I’d find something unusual and perhaps uncover a new way to modify the microbiome. Ultimately the main thing I learned is that the microbiome is incredibly robust. Even a complete reset won’t change much.

References

Nagata, Naoyoshi, Mari Tohya, Shinji Fukuda, Wataru Suda, Suguru Nishijima, Fumihiko Takeuchi, Mitsuru Ohsugi, et al. 2019. “Effects of Bowel Preparation on the Human Gut Microbiome and Metabolome.” Scientific Reports 9 (1): 4042. https://doi.org/10.1038/s41598-019-40182-9.


  1. L. reuteri DSM 17938 and L. rhamnosus GG in Dommels et al. (2009)

  2. Read a detailed description of what microbes are noted in a bowel cleansing: Jalanka et al. (2015)

  3. Full text available here: https://www.nature.com/articles/s41598-019-40182-9.pdf