Full disclosure: I know two senior people at Seres Therapeutics from previous positions; however, I have no ties with the company whatsoever, nor do I have any ‘insider’ information.

Last week, the prospect of altering the microbiome–those microorganisms that live on and in us–to fight disease suffered a setback with the apparent failure of Seres Therapeutics’ SER-209 treatment (aka ‘the poop pill‘). What looked like a very promising avenue pretty much vanished.

To recap, Seres found incredibly promising results during their Phase I trials. Phase I trials are only designed to determine the basic safety of the treatment (i.e., don’t kill the patients!), and are not designed to determine if the treatment actually works compared to the current standard of care. Nonetheless, when examining the effects of SER-209 on thirty people, here’s what Seres found:

Results demonstrated that 87 percent of patients (26 of 30) met the predefined endpoint of preventing recurrent CDI within eight weeks following administration of SER-109. Most importantly, 97 percent of patients (29 of 30) achieved clinical cure during the eight-week period after SER-109 dosing, as defined by the absence of CDI requiring antibiotic treatment. By contrast, the expected cure rates in people treated with the standard of care, i.e. antibiotics, for recurrent CDI range from 23-31 percent, according to a recent well-controlled study.

It’s hard to explain how phenomenal this result is, but this is pretty incredible. Even if one were to claim that the true expected cure rate was fifty percent–which is higher than seen in the literature–we’re talking about millions to one odds that this is a chance fluke.

And then the Phase II results came in (boldface mine):

The predefined study primary efficacy endpoint is the relative risk of CDI recurrence up to 8 weeks after treatment comparing subjects in the placebo arm with the SER-109 arm. CDI recurrence is defined as diarrhea for 2 or more consecutive days, a positive CDI test, and the requirement for antibiotic treatment. Based on 8-week data, CDI recurrence occurred in 44% of subjects (26 of 59) who received SER-109, compared to 53% of subjects (16 of 30) who received placebo. The relative risk of CDI recurrence for the placebo population compared to the SER-109 population was not statistically significant. As part of the prespecified design, subjects were stratified into two groups: <65 years old and ≥65 years old. In subjects <65 years old, CDI recurrence occurred in 43% of subjects who received SER-109 (12 of 28) and in 27% of subjects who received placebo (4 of 15). In subjects ≥65 years old, CDI recurrence occurred in 45% of subjects who received SER-109 (14 of 31), and in 80% of those who received placebo (12 of 15)….

Based on the eight-week data, we did not observe any difference in the adverse event frequency or type in the subjects receiving SER-109 compared to those receiving placebo. The most commonly reported adverse events in both the SER-109 and placebo arms were in the gastrointestinal category. The most common adverse events reported in the SER-109 arm were diarrhea, abdominal pain and flatulence. No drug-related serious adverse events were observed.

So what the hell happened? This isn’t the Decline Effect or regression to the mean–the initial effect was so overwhelming, I think we can rule out a sample size problem Paul Orwin suggests that this might have something to do with sporulation–the pill only contained bacterial spores, and spore germination can depend on a complex interaction of environmental variables. Something else that jumps out at me is how low the recurrence frequency is in the non-SER-109 control group (57% versus 70 – 80% in other studies).

A major lesson here is that biology (including ‘microbiome-ology’ or ‘microbiomics’ which in days of yore were simply known as microbiology) is twitchy–or as a dissertation committee member once told me, “It all comes down to those stupid fucking natural history facts.” Here’s a non-microbiome example:

In one study, a[n antibiotic] resistance plasmid found in cattle was advantageous even in the absence of antibiotics because the plasmid enabled an E. coli strain to use vitamin D. In essence, what we call a ‘resistance’ plasmid (because that’s the trait we look for) could be an ‘eat your vitamins’ plasmid, or have some other function, such as protection against bacterial viruses (‘phage’).

Because we’re not solely trying to understand things in idealized conditions–though that’s often a first, crucial step:

But biologists aren’t only trying to derive general principles, we’re also trying to figure out how organism- or system-specific processes work.

To use a very macabre example, we are interested in a how a gun fires a bullet–and in a controlled environment, we can estimate very precisely how that bullet will travel. But, a biologist is also faced with the task of trying to figure out what happened on the grassy knoll in Dallas. The study of ballistics is necessary, but not sufficient. At the risk of completely tasteless overkill, does anyone view September 11th primarily (or even entirely) as a structural engineering problem? Those stupid natural history facts matter too. Understanding and predicting particular events also requires a knowledge of phenomena that can not be generalized and reduced to simple general theory.

Biology is hard. Hopefully, Seres learns from this, and makes the needed changes, if possible, because we need better treatments for C. difficile.