I. Microbiota Transplants Lack High-Quality Evidence

Despite professional and public enthusiasm for FMT as a potential treatment for many ailments, the scientific and medical communities currently lack high-quality evidence of efficacy for the use of FMT to treat any disease.^{Reference Panchal, Budree and Scheeler24} This article defines “high-quality evidence” as data generated from clinical trials that justifies valid causal inference about the effects of a medical intervention.^{Reference Friedman, Furberg and DeMets25} This section focuses primarily on evidence for the efficacy of stool for treating medical conditions, particularly rCDI.

Statisticians have long argued that high-quality evidence is best generated through prospective, randomized, controlled, double-blinded, clinical trials.^{Reference Fisher, Bothwell, Greene, Podolsky and Jones26} Ideally, project personnel conduct such trials according to detailed protocols; they carefully monitor both participants and researchers for adherence to trial protocols; they collect data using detailed case report forms separate from ordinary health records; they use laboratory and clinical tests that elicit information most useful for answering the research question; they rigorously monitor participant outcomes; and they design the study and analyze its results using appropriate statistical expertise, such that the results will most likely lead to valid conclusions.^{Reference Kendall, Friedman, Furberg and DeMets27} When properly implemented, these features of randomized controlled trials (RCTs) help to produce precise, complete data from which valid conclusions about the safety and efficacy of an investigational medical intervention can be drawn.

Techniques such as randomization^{Reference Dettori, Bothwell, Greene, Podolsky, Jones and Kendall28} and double-blinding^{Reference Podolsky, Jones, Kaptchuk, Kesselheim, Robertson, Hróbjartsson, Kesselheim, Robertson, Begley and Ioannidis29} help minimize bias and confounding.³⁰ Appropriate controls also help ensure lack of confounding by known prognostic and risk covariates.³¹ Together, these techniques help researchers make valid causal inferences. They help researchers determine whether a trial's outcomes are caused by the test intervention or by other factors, such as the researchers' or participants' expectations, the natural course of the disease, or some exposure common among people in the treatment arm and not among people in the control arm.^{Reference Bothwell, Greene, Podolsky and Jones32} Lack of randomization, blinding, and proper controls are reasons why researchers generally cannot use observational studies to determine causality.^{Reference Sherman, Anderson, Dal Pan, Ejima, Li and Smith33} When the medical community and patients rely primarily on observational and anecdotal evidence, they can engage in medical interventions that do more overall harm than good, or that impose burdens on patients and costs on the medical system without commensurate benefit to patients.^{Reference Henderson, Paganini-Hill, Ross, Stampfer, Colditz, Grady, Rubin, Petitti, Hulley, Grady and Bush34}

When a clinical trial is conducted by drug developers who intend to submit data to the FDA (or a comparable agency) in support of a marketing application, the agency provides oversight of the trial's design, data acquisition, data management and integrity, and data analysis.^{Reference Lietzan35} Trial sponsors often audit trial sites, and the FDA inspects them. During the review of a sponsor's marketing application, the FDA often conducts an independent analysis of the sponsor's data.³⁶ The FDA expends substantial amounts of money to oversee companies' design and conduct of clinical trials, and then to validate the evidence these trials produce.³⁷ Agency oversight helps clinical trials produce high-quality evidence.³⁸

Enthusiasm for FMT developed in advance of evidence from any double-blinded or well-controlled clinical trial.^{Reference Kelly, Kim, Laine and Wu39} The strongest evidence for FMT's efficacy is for the treatment of rCDI.^{Reference Allegretti, Kassam, Osman, Kelly, Khoruts and Staley40} Clostridioides difficile, often a nosocomial infection, causes uncontrollable diarrhea. In 2015, the U.S. Centers for Disease Control and Prevention announced that almost half a million patients were infected with Clostridioides difficile in a single year.⁴¹ Of those patients, 29,000 died within 30 days of their initial CDI diagnosis; 15,000 of those deaths were directly attributable to CDI. The number of CDIs is increasing in the U.S. and across the developed world.^{Reference Lessa, Mu, Bamberg, Edelstein, Daw and Kassam42} Clearly, CDI is an important public health problem in search of an effective treatment.

Arguments for using FMT as a treatment for rCDI were initially based on observational studies, such as case series, and data from small, unblinded trials.^{Reference Scheeler43} These studies were not designed to produce evidence of efficacy for FDA approval of stool as a treatment for rCDI and were not regulated by the FDA. Results from these initial studies indicated that FMT led to resolution of rCDI symptoms in 89% to 94% of patients.^{Reference Quraishi, Widlak, Bhala, Gough, Shaikh, Manges, Drekonja, Reich, Gezahegn, Kassam, Lee and Hunt44} Some articles reported a 100% cure rate.^{Reference Kelly, Kahn and Kashyap45} The first published clinical trial to evaluate the efficacy of FMT for rCDI was a small, unblinded, randomized, active-controlled trail.^{Reference van Nood, Vrieze and Nieuwdorp46} It was terminated early because FMT appeared quite effective. Taken together, early studies suggested that FMT treated rCDI much more effectively than standard-of-care treatment with antibiotics, and the effect size was so large that many medical professionals may have brushed aside concerns about evidence quality.

Results from the first double-blinded RCT for treatment of rCDI were published in 2016, by which time many patients were clamoring for the intervention and a community of enthusiasts for DIY FMT had already emerged. In the double-blinded study, investigators compared cure rates between participants in the experimental arm, who received donor FMT through colonoscopy, and participants in a control arm, who received autologous FMT.⁴⁷ “Autologous FMT” means that patient-participants in the control group were administered their own C. difficile-infected feces, prepared the same way as the donor feces were prepared for people in the experimental arm of the study.

The Kelly et al. trial ran at two sites and, unfortunately, results varied substantially between them. Overall, 15 of 24 participants (63%) who received autologous stool were cured (a very high cure rate when compared to antibiotic treatment), whereas 20 out of 22 participants (91%) who received donor stool were cured.⁴⁸ At one of the two sites, 90% of patients who received autologous stool were cured, as compared to 92% who received donor stool — there was no statistically significant difference between the experimental and control interventions. At the other site, 43% of individuals who received autologous stool were cured, compared to 90% of participants who received donor stool. Across both sites, participants infused with donor stool subsequently showed increased diversity of their gut microbiota. Such diversity is commonly associated with gut health. On the other hand, participants infused with autologous stool continued to display dysbiotic (abnormal) gut micro-biota, with low diversity, even when the participants were cured.

This trial's high between-site variation in cure rates in the control group, and higher-than-expected overall cure rates in the control group, make it difficult to use the data for drawing causal inferences about the effects of donor stool for FMT. The high cure rate for the control group could have occurred by chance. Alternatively, the high cure rate among participants who received autologous stool, and despite their ongoing gut dysbiosis, could indicate that something other than a healthy donor's microbial community cures rCDI when FMT is performed. Based on earlier data, experts have inferred that microbiota in healthy donor stool caused the resolution of symptoms;^{Reference Seekatz, Rao, Santhosh and Young49} however, that conclusion cannot be drawn from existing data.

The NIH recently funded a placebo-controlled, double-blinded RCT to assess the safety and efficacy of treating rCDI using FMT delivered by enema.⁵⁰ This trial started in 2018 and is now recruiting participants.⁵¹ Whether this public investment will be sufficient to answer important questions concerning the causal relationships between microbes in stool and resolution of rCDI symptoms remains to be seen.

While there is little high-quality evidence that microbes in stool cure rCDI, the anecdotal evidence for FMT as a cure is notable. There are cases in which people on the verge of death from rCDI received FMT and made a full recovery.⁵² This article's questioning of the evidence base is not meant to denigrate the profound experiences of people who have undergone FMT and been cured, but rather to say there is insufficient evidence that fecal microbiota cure rCDI. Perhaps, a colonoscopy that delivered saline would also have cured some or most of those people. Even if some people are cured by fecal microbiota, without good evidence nobody knows who is likely to benefit from FMT or by which microbes they would be helped.

In 2018 the Infectious Disease Society of America and the Society for Healthcare Epidemiology of America updated their clinical practice guidelines to include a strong recommendation for FMT to treat adults with rCDI.^{Reference McDonald, Gerding and Johnson53} Some private health insurers now cover this treatment.⁵⁴ Medicare and its contractors also pay for it.^{Reference Lessa, Mu and Bamberg55} Due to the absence of conventional medical alternatives, FMT as a treatment for rCDI has become widely available based on low-quality evidence showing strong effects of FMT.

The apparent success of FMT as a treatment for rCDI, and other evidence that dysbiosis of the gut microbiota is associated with many illnesses, have led to widespread enthusiasm for FMT as a possible treatment or cure for a wide variety of ailments other than rCDI.^{Reference Vrieze, de Groot and Kootte56} In June of 2019 at least 310 clinical studies were investigating FMT for some indication, although only a few of those studies were double-blinded RCTs.⁵⁷

Despite the value of well-designed RCTs, various trends in science and medicine have decreased public, medical professional, and Congressional enthusiasm for them. These trends include: their high costs; what some stakeholders perceive as their excessively slow pace; the difficulty of generalizing RCT results to broader, real-world patient populations; the difficulty of conducting RCTs for rare or “molecularly subsetted” diseases;⁵⁸ the difficulty or impossibility of detecting rare but serious side-effects in an RCT; the rise of patient-driven research; and the increasing availability of health care data for use in big data analyses.^{Reference Sherman, Anderson and Dal Pan59} In response to the preceding concerns and opportunities, many stakeholders hope to use real world evidence (RWE) to supplement or supplant evidence from RCTs. Data for producing RWE include (but are not limited to): patient-reported outcomes, electronic health record information, claims and billing information, data collected from product and disease registries, data from healthcare operations, social media posts or queries, and information gleaned from wearables and apps.⁶⁰

RWE could be generated for FMT even though stool for FMT has not been approved for any indication, because FMT is occurring in clinical practice and DIY settings. FMT is also being conducted outside of the U.S. for a variety of indications.^{Reference Edelstein, Kassam and Daw61} A coalition of U.S. medical professional societies and medical philanthropies has created an FMT national registry that includes some data from which RWE could be generated.⁶² Such evidence could be useful in understanding long-term or low frequency safety issues associated with FMT. However, given the complexity of the gut microbiota and their relationship to health and disease, it seems unlikely that decision-makers, including researchers, clinicians and regulators, could or should use RWE as a basis for valid judgements about FMT's efficacy for treating any indication.

To the extent that RWE derives from observational data, it suffers from the same problems as other observational evidence discussed above — it is not generally suitable for making causal inferences. Even when RWE is generated in a prospective experiment, the data sources for RWE are prone to error, missingness, and numerous biases.^{Reference Sherman, Anderson, Dal Pan, Pearl, McMurray, Freemantle, Marston and Walters63} Data for RWE typically lack the quality controls imposed in a well-designed and properly-conducted RCT. Analysts may find it impossible to sort out fundamental issues of confounding and bias in such data. Data scientists are working to develop methods for producing high-quality RWE,^{Reference Berger, Sox and Wilke64} and even for drawing causal inferences from observational data,^{Reference Kuang, Bao, Thomson and Vanhaelen65} but to date most experts caution that RWE should supplement, rather than supplant, evidence from well-designed RCTs.^{Reference Gottlieb66} Governance of microbiota for transplantation cannot rely on RWE to fill in evidence gaps left by a dearth of RCTs.

Thus far, this article has focused on evidence for FMT's efficacy; however, a brief discussion of safety is also in order.^{Reference Ossorio and Zhou67} Experts worry about the risks of transmitting dangerous microorganisms and unhealthy phenotypes (such as obesity or depression), physically injuring recipients when administrating FMT (e.g., by colonoscopy or nasogastric tube), and exacerbating a recipients' disease.^{Reference Baxter and Colville68} Unfortunately, much existing safety data comes from metanalyses of case studies and case series in which safety information was not systematically collected.⁶⁹ The data are generally low-quality and relate primarily to short-term risks. Studies in which safety data were systematically collected suggest that FMT administered by a healthcare provider has some risks, but their probability is low and they are generally mild and tolerable.^{Reference Qazi70}

Transmission of pathogenic organisms is a primary concern for experts, patients, and research participants.^{Reference Paramsothy, Borody, Lin, Park, Mone and Price71} When FMT is conducted by health care providers the donor and stool are screened to avoid transmission of infectious diseases and unhealthy phenotypes.^{Reference Dubois, Ling and Osman72} There are no widely accepted professional standards for how donors and stool should be screened,^{Reference Wood-worth, Neish, Miller, Terveer, van Beurden and Goorhuis73} and there is little guidance from the FDA on this issue;⁷⁴ however, responsible stool providers conduct extensive screening.^{Reference Dubois, Ling and Osman75} Despite donor and stool screening, the possibility of transmitting pathogens via FMT remains, either because a screening test missed a known pathogen or because a new pathogen has entered the human microbiota and there are no tests to detect it.⁷⁶

While this article was in production, the FDA announced that two immunocompromised patient-participants in an FMT study had contracted a multi-drug resistant, pathogenic strain of E. coli from FMT. One participant died from the infection.⁷⁷ The FDA issued instructions for screening donors and stool for multi-drug resistant microorganisms, and required that all researchers licensed to conduct FMT studies (all researchers using FMT under an IND⁷⁸) implement such screening by July 15, 2019.⁷⁹

Finally, note that scientists and regulators cannot evaluate the safety of FMT vel non. FMT could have different safety profiles in patients with different diseases. The way in which transplanted microbes engraft in a recipient, and whether or not they cause harm, could depend on how a recipient's gut has been altered or damaged by her underlying disease.

This section has presented the case that the evidence base for FMT as a safe and effective treatment is weak. Past governance approaches have not incentivized the production of high-quality evidence. The next section will compare FMT to stool-derived microbial products, and argue that stool-derived microbial products will likely be as good or better than stool for treating disease.

II. Stool Versus Stool-derived Microbial Products

Stool is a complex mixture of bacteria, viruses, fungi, human cells, mucous, extracellular proteins, and metabolites.^{Reference Hoffmann, Palumbo and Ravel80} Stool for FMT can be delivered via colonoscopy, nasogastric tube, enema, or capsules.⁸¹ Stool can be prepared and screened “in-house” by clinicians and pharmacists at a health clinic or hospital, or it can be produced and shipped by a stool bank.⁸² For purposes of this article, stool banks are entities that obtain, prepare, and store stool for later distribution to other institutions for use in patient therapy or clinical research.⁸³

One problem in developing an evidence base for the efficacy and safety of FMT is that stool has not been well-characterized. Over half a decade ago, work from the HMP suggested that human health and disease might not be determined by the taxa of microbes in people's microbial communities, but rather by the collection of molecular functions performed by the specific strains in those communities.⁸⁴ Yet, a recent commentary notes that researchers have done little to determine which changes in microbiota composition or function represent causes of disease, rather than effects or associations.^{Reference Lloréns-Rico and Raes85} If microbes in stool produce beneficial effects through FMT, researchers still do not know which microbes do so, or why.

Complexity does not make stool unusual among items regulated as biological products, which the FDA describes as “complex mixtures that are not easily identified or characterized.”⁸⁶ Stool is, however, orders of magnitude more complex than many other biological products. Because most biological products cannot be fully characterized, regulators define them in part by their manufacturing processes, which are tightly controlled.⁸⁷ The FDA's oversight of stool manufacturing focuses on the processes for donor and stool screening, methods for stool storage and preparation, effects of adding saline or stabilizers, techniques for characterizing stool, and other relevant considerations.^{Reference Slater88}

However, the FDA has no jurisdiction over the processes by which stool is initially produced in and excreted from the human body. Stool is difficult to characterize not only because it is complex, but because it is dynamic and responsive to the environment.⁸⁹ The composition of a donor's stool will vary with the person's diet, general health, behaviors (such as the amount of sleep the person had before donating), perhaps time of day, and other unknown factors. Characterizing stool from every donor at every donation, for instance by developing a genomic profile of all microorganisms the stool contains, would be impossible or prohibitively expensive using current technology.⁹⁰

Because stool for FMT cannot be well-characterized, it is all the more important for researchers, regulators, and clinicians to understand which, if any, of its components have therapeutic effects. If scientists gain a better understanding of which microbes, and which microbial functions or molecules, cause changes in a person's health, then clinicians and researchers should be able to adequately characterize stool: they should be able to measure the components that matter when FMT is used as a treatment. If researchers identify which microorganisms and microbial functions cause health improvements, they can develop metrics for stool's potency and stability in relation to different methods of storage, preparation, and delivery. Currently, these basic characteristics of a biological drug are not determinable for stool.

Users cannot determine what dose of a therapeutic agent is being administered in a specified quantity of stool. With microbiota the starting dose might not be as important as it is for a small molecule drug; after all, microbes can multiply. However, without knowledge of which microorganisms or molecules are therapeutic for a given indication, researchers, regulators, and clinicians cannot know whether any therapeutic agent has been delivered to a recipient or whether one donor's stool will be more therapeutic than another donor's stool for a particular indication. They cannot know whether the therapeutic agent has grafted at its target body site, whether it has multiplied, or whether it is producing therapeutic molecules or functions in the new human host. In addition, the scientific community has little understanding of which genetic, metabolic, or other “ecological elements” of an FMT recipient influence the activity or treatment effects of donor microbiota.⁹¹

Bacteria are the easiest microorganisms to identify and characterize in the gut microbiota, but they are not the only gut microorganisms influencing human health. Fungi and viruses are also present and may influence the course of many diseases.^{Reference Sokol, Leducq, Aschard, Chehoud, Dryga, Hwang, Lloyd-Price, Arze and Ananthakrishnan92} Adequately characterizing the gut microbiota, and their transformations, will involve a better understanding of viruses and fungi.

Private firms face disincentives to developing stool as an approved drug. As a product of nature, minimally-manipulated stool will be difficult to patent.⁹³ Patent protection will be available for new and non-obvious methods of storing, freezing, or administering stool, and for products containing stool in specialized capsules or other delivery vehicles.^{Reference Stevens, Sandowsky, Khoruts, Weingarden and Hamilton94} Thus far, such patents have not incentivized substantial private investment in developing stool as a drug to treat rCDI or any other indication, perhaps because stool is widely available outside of firms' control, the types of patents available are easy for people to work around, or patent holders might have difficulty enforcing their patents against large numbers of small-volume infringers. A combination of public funding, philanthropic, and non-profit investments could lead to development of stool as a drug, but thus far has led to numerous small studies and a few RCTs that have not produced the quality of data that generally would lead to marketing approval.

Given clinicians' and patients' enthusiasm for microbiota transplants as a treatment, and also the difficulties and risks of using stool as a drug, it should be no surprise that biotechnology companies are attempting to develop stool-derived microbial products for treating rCDI and other medical conditions.^{Reference Gertner95} For purposes of this article, “stool-derived microbial products” consist of defined communities of microbes that were originally isolated from stool samples. Such communities could range from complex consortia to relatively simple mixtures.⁹⁶ They will be composed of known strains, and to create such a product scientists must identify which strains have therapeutic properties. Compared to stool, a stool-derived microbial product will contain few or no human cells and proteins, and far fewer taxa of bacteria, viruses, and fungi.

Stool-derived microbial products will raise some of the same regulatory challenges as stool and are likely to compete in the same markets as stool.⁹⁷ These products are regulated as biological drugs and do not receive any enforcement discretion for their use. No such product is currently on the market, although several are in development.

As drugs, stool-derived microbial products will likely be as good as or better than stool.⁹⁸ Once such defined mixtures or consortia of microbes are created, specifying and characterizing their important functionalities, and standardizing the microbial product, should be more straightforward than doing so for stool. Knowledge of which microbes and molecules provide therapeutic benefit will facilitate assessment of the product's purity, potency, dosing, stability and other important characteristics related to delivering safe and effective drugs to patients. Stool-derived microbial products might also have fewer off-target effects than stool, because stool-derived products will contain fewer microbes and other biologically active materials.

As a practical matter, stool-derived microbial products likely will be grown in laboratories rather than repeatedly isolated from stool. Laboratory-grown products would be less likely than stool to acquire human gut pathogens. And because even a complex microbial consortium would be less complex and far better characterized than stool, a stool-derived microbial product would be easier to screen for contamination by microorganisms not intended as part of the product.

For the above reasons, stool-derived microbial products likely will be safer and as or more effective drugs than stool for many indications for which microbes cause resolution of symptoms. FMT using stool from human donors might best be viewed as a temporary solution to treating medical problems, and used only until a stool-derived microbial product or other safe and effective drug becomes available.⁹⁹

In light of the scientific community's current lack of knowledge regarding the therapeutic functionalities of microbes in stool, biotechnology companies are undertaking extensive preclinical and clinical research to develop stool-derived microbial products. Such research takes time and costs significant sums of money, and when stool-derived microbial products come on the market they will likely be more expensive than well-screened stool.¹⁰⁰ Whether clinicians and patients find the additional costs justifiable will depend on the price differentiation, the relative effectiveness of the stool-derived microbial products, their ease of administration, and their marketing.

Whether firms continue to invest in development of stool-derived products will depend, in part, on how stool for FMT is regulated. Expected future returns shape firms' incentives to make present investments in research and development of new drugs (including biologics), and regulation can influence a firm's expected returns.^{Reference DiMasi, Grabowski, Danzon and Nicholson101} If the FDA continues its current enforcement discretion policy, and particularly if the FDA signals that it might leave unapproved stool on the market after a stool-derived microbial product is approved, biotechnology companies will have a dis-incentive to invest in stool-derived products. If unap-proved stool, sold by stool banks that did not need to recoup the costs of RCTs and FDA approvals, could compete in the same market as an approved stool-derived microbial product, then rational firms would be less likely to invest in producing a stool-derived product.¹⁰²

III. Enforcing Drug Regulations for Stool and Stool-Derived Products: The FDA's Current Approach

Since 2013, the FDA has regulated microbiota for medical transplantation as a biological drug, whether the microbiota consist of stool or a stool-derived microbial product.¹⁰³ A new drug, including a new biological drug, must undergo premarket review in which the sponsor presents the FDA with substantial evidence for the drug's safety and efficacy.¹⁰⁴ Substantial evidence has traditionally meant evidence from two prospective, double-blinded RCTs. However, that quantity and quality of evidence is not required by law, and the FDA has flexibility regarding the amount and type of evidence it accepts.¹⁰⁵

For a biological drug such as stool, a Biologics License Application (BLA) must be approved prior to the drug entering the market.¹⁰⁶ Approval of a BLA, and the twelve years of data exclusivity that accompanies such approval, are mechanisms by which FDA oversight incentivizes firms to invest the hundreds of millions, perhaps billions, of dollars required to develop high-quality evidence of a biological drug's safety and efficacy.¹⁰⁷ Firms sometimes game the evidence-production system, for instance, by designing clinical trials to avoid learning important information,¹⁰⁸ but no other set of institutional arrangements and incentives has produced as much or as good medical evidence as the FDA's authorities to regulate new drugs.

Prior to approval of a BLA, researchers must obtain an investigational new drug application (IND) before using the drug in clinical research.¹⁰⁹ Outside of research, manufacturers can only provide an unap-proved drug to clinicians and patients under a narrow set of circumstances; although, the U.S. market for pharmaceuticals includes numerous unapproved drugs that are sold openly but illegally.¹¹⁰ Clinicians and patients may attempt to legally obtain an unap-proved drug using the FDA's expanded access program (often referred to as “compassionate use”),¹¹¹ or using a pathway created by the new federal “Right to Try” law.^{Reference Wendler, Mongiello, McLinn and Bellina112} Both of these pathways restrict the kinds of patients who are eligible to receive a drug and the types of drugs that can be provided, and neither pathway requires that a manufacturer provide the unap-proved drug.

As mentioned above, prior to 2013 stool for FMT was not regulated and FMT was incorporated into medical practice as a treatment for rCDI. Now, facing the access-before-evidence dilemma, the FDA is struggling to regulate stool. In May of 2013, “everyone kind of flipped out”^{Reference Fischer, Kassam, Kelly and Sobcinski113} when the agency announced that it would regulate FMT as a biological drug. Later that year, the agency published guidance announcing that it would exercise enforcement discretion for microbiota used to treat “C. difficile infection not responding to standard therapies.”^{Reference Price and Price114} This policy permits manufactures to produce and distribute stool without an IND when the stool will be used to treat rCDI. Distribution of stool for all other indications still requires an IND.

Under the 2013 guidance, stool banks have shipped over 40,000 stool preparations to clinicians for use in FMT,¹¹⁵ and clinicians have conducted an unknown number of FMTs using in-house-produced stool. While the guidance only extends enforcement discretion to stool for treating rCDI, the FDA will not always know whether clinicians are using stool to treat other indications, particularly if that stool was produced in-house. One biotechnology firm stated that some entities “are investigating and marketing FMT products for other indications [not rCDI] without complying with the applicable regulatory requirements.”¹¹⁶

Numerous commentators have argued that the FDA should not regulate stool for FMT as a drug, or that the agency should not to regulate stool at all.^{Reference Hoffmann, Palumbo, Ravel, Smith, Kelly, Alm, Sachs, Edelstein, Riley and Olle117} Such arguments entail a claim that stool should not be required to undergo the type of clinical trials and regulatory scrutiny that could produce high-quality evidence of its safety and efficacy for treating particular medical conditions. Such a claim might be credible if high-quality evidence for medical use of FMT, even to treat rCDI, had been produced during the pre-2013 period in which stool was unregulated. However, in the absence of FDA drug regulation there were and are no incentives strong enough to promote investments of the magnitude needed to produce high-quality evidence for FMT.¹¹⁸

In 2014 and 2016 the FDA attempted to update its 2013 policy regarding enforcement discretion for the use of stool to treat rCDI.¹¹⁹ The agency published draft guidance announcing that it would not extend enforcement discretion to stool banks. It proposed that a stool bank should either hold an IND or ship product only to an entity with an active IND. The agency raised concerns that centralized manufacturing and distribution of stool could create significant public health risks.¹²⁰ If a stool bank failed to adequately screen donors or stool, and distributed a dangerous product to numerous different hospitals for use in numerous patients, a great deal of harm could occur. Under enforcement discretion, the FDA would not know which stool banks were distributing products to which medical providers. It would not know how the donors and stool were screened. The agency would not receive systematically collected adverse event reports. Under such circumstances, the FDA and the medical community would be slow to detect safety problems. The 2014 and 2016 draft guidances likely were influenced by the specter of a public health fiasco that occurred in 2012, in which a single compounding pharmacy distributed contaminated drugs across the U.S., causing fungal meningitis in 753 people and killing sixty-four.^{Reference Outterson121}

Many commentators voiced strong opposition to both the 2014 and the 2016 proposals. Some medical experts argued that the proposed policies could move the preparation and screening of stool from stool banks to small healthcare organizations that lack the scientific expertise, logistical capacity, and financial wherewithal to conduct the extensive screening that stool banks currently perform.¹²² Such an outcome would make FMT more dangerous, undermining the FDA's public health objectives. Other commentators were concerned that requiring an IND to use materials from stool banks would make FMT for rCDI less available to patients, because clinicians with the expertise to administer FMT might lack the time or expertise to file IND paperwork. After considering such criticism, the FDA did not finalize either of the draft guidance documents.

Operating under the 2013 guidance, stool banks still distribute material for FMT to physicians, without an IND, for use in treating rCDI.^{Reference Edelstein123} The first and best-known stool bank in the U.S., the non-profit OpenBiome, also helps clinicians to obtain INDs so the bank can distribute stool to them for uses other than treating rCDI.¹²⁴ OpenBiome has a “biologics master file” with the FDA,¹²⁵ and clinician-investigators not employed by the stool bank can reference this file when applying for their own INDs.

Some microbiome scientists have argued that the FDA's enforcement discretion policy has slowed evidence generation for stool and for stool-derived microbial products.¹²⁶ One expert lamented that “only a tiny fraction of data on the >10,000 patients treated by OpenBiome products has been systematically collected and published.”^{Reference Khoruts127} In addition, several commentators believe that widespread access to FMT has slowed enrollment in RCTs testing both stool and stool-derived products as treatments for rCDI. Potential participants may be less likely to enter a trial in which they might be randomized to a placebo or other control arm;^{Reference Kelly, Fischer, Grinspan and Allegretti128} they might not receive FMT immediately; or they might receive a stool-derived microbial product instead of donor stool.¹²⁹ When exercising enforcement discretion, the FDA has no opportunity to assess whether access to the unapproved product is hindering the evidence generation process.

Another reason some patients might not be enrolling in RCTs testing potential treatments for rCDI is because their clinicians believe that the medical community is no longer in equipoise about the value of FMT. Based on recently promulgated practice guidelines recommending FMT for treatment of rCDI,¹³⁰ and insurers' increasing willingness to pay for the intervention, clinicians might believe it is unethical to recommend that their patients participate in a trial if they could be randomized to an arm in which they do not receive FMT. Note that both physician and patient concerns about participation in RCTs are likely premised on the belief that microbiota transplants are safe and effective treatments for rCDI or other indications. This article argues that belief is not yet well-justified, even for rCDI.

IV. A Proposal for Governing FMT and Stool-derived Microbial Products

Enforcement discretion should not be a permanent solution to the access-before-evidence dilemma with regard to FMT for rCDI (or for other potential drugs). Access-before-evidence leaves society with the continuous accumulation of low-quality and potentially misleading evidence, while inhibiting the production of high-quality evidence for the use of stool or stool-derived microbial products. Enforcement discretion may decrease incentives for private investment in the development of stool-derived products, even though such products would likely be safer and as or more effective than stool. The FDA would better fulfill its public health mission by curtailing its use of enforcement discretion for stool. Both the 2014 and 2016 draft guidance documents indicate the agency's intention to end enforcement discretion in regard to stool distributed by banks.

The policy articulated in the 2016 draft guidance is practicable, and is superior to the current enforcement discretion policy; however, this article proposes a slightly different approach:

• Enforcement discretion should not extend to any use of stool by institutions where any investigator is currently conducting research under an IND, or where any investigator has done so in the recent past (probably, the last three years).

• Stool to treat rCDI meets the criteria for expanded access for an intermediate-size patient population.¹³¹ When appropriate, the FDA should grant expanded access to stool for FMT to treat intermediate-size patient populations with rCDI.¹³² The agency should prioritize expanded access for organizations that (1) sponsor an IND for treatment of rCDI with stool or a stool-derived microbial product; (2) employ at least one investigator¹³³ who, under an IND, is conducting studies of stool or a stool-derived microbial product to treat rCDI; or (3) can demonstrate that they have successfully referred at least one patient to a relevant RCT.

  • – Consistent with the regulations, the FDA should only grant expanded access if “providing the investigational drug for the requested use will not interfere with the initiation, conduct, or completion of clinical investigations that could support marketing approval of the expanded access use or otherwise compromise the potential development of the expanded access use.”¹³⁴

• The FDA should undertake extensive and varied stakeholder engagement activities, with the aim of developing RCT designs for FMT and stool-derived microbial products that are (1) attractive to patients and their clinicians; (2) capable of efficiently producing data the FDA will accept in support of a BLA; and (3) feasible and affordable.

This proposal for changing regulatory enforcement aims to promote the generation of high-quality evidence for the use of stool in FMT, and for stool-derived microbial products. It would require that clinicians at organizations with significant research and regulatory sophistication obtain an IND before undertaking FMT for any reason, including treatment of rCDI. They would need an IND for using any stool, whether from a bank or produced in-house. Although obtaining an IND is burdensome, the proposal would impose these burdens on investigators at “research-capable” institutions, which already possess the infrastructure to participate in FDA-regulated research and are capable of producing high-quality evidence.

Obtaining an IND is feasible for research-capable institutions and their scientists. Sponsor-investigators¹³⁵ at some research-capable healthcare institutions already hold research-INDs for use of stool to treat people with rCDI or other conditions,^{Reference Kelly, Kunde and Kho-ruts136} and many hold INDs for studying other drugs. The medical literature instructs clinician-scientists in how to obtain a research-IND;^{Reference Holbien137} the FDA provides instructions;¹³⁸ and some academic centers have offices that help clinician-scientists apply for INDs. Just as Open-Biome currently does for non-rCDI indications, stool banks should have a biologics master file and permit sponsors to reference that file when submitting their INDs.

When the product under study is a stool-derived microbial product, the IND likely will be held by a commercial sponsor and researchers at the healthcare institution will be investigators¹³⁹ (not sponsor-investigators). Healthcare institutions whose investigators conduct research under a commercial sponsor's IND are selected by the sponsor because of the institutions' infrastructure for conducting regulated research and its researchers' superior reputations. Such institutions should be producing high-quality evidence for the use of FMT and stool-derived products.

Under this article's proposal, healthcare providers and institutions with less research sophistication — often smaller, non-academic, less-well-resourced, or in less populous areas — would not need an IND for obtaining materials from a stool bank or for preparing stool in-house. Most of these institutions probably see few patients with rCDI. If they provide FMT, they may lack the resources to safely prepare and store stool in-house, so policies should permit and even encourage them to obtain FMT materials from a stool bank. Continued enforcement discretion for them likely will not undermine the generation of high-quality evidence, but it will promote patient safety.

Requiring that certain providers of FMT have an IND is not the same as requiring that they all conduct RCTs. The IND regulations only specify that the drug be used in a “clinical investigation,” which is defined as “any experiment that involves a test article and one or more human subjects…”¹⁴⁰ Clearly, some patients who seek FMT for treatment of rCDI would be ineligible or unable to participate in an RCT. Some rCDI patients might not have an RCT geographically accessible. Some rCDI patients would not meet eligibility criteria for an RCT. For instance, a recent study found that 75% of patients with rCDI were not eligible to participate in an RCT testing a stool-derived product.¹⁴¹ The problem of restrictive eligibility criteria that slow accrual to RCTs is not unique to studies of FMT;^{Reference Kim, Atlas, Ison and Ersek142} the FDA and other stake-holders are working to address this issue. In addition to requiring that sponsors justify eligibility criteria for trials of FMT or stool-derived products to treat rCDI, the FDA can work with sponsors to design some non-RCT studies for which some people who cannot participate in RCTs would be eligible. This approach will not always produce the highest quality data, but it ought to produce some data and systematic adverse event reporting, which is more than society receives under the current enforcement discretion policy.

Although the regulations permit INDs for studies that are not RCTs, this article's proposal aims to encourage faster accrual for RCTs. It does so by using the FDA's expanded access authorities rather than enforcement discretion. When considering whether expanded access would “interfere with the initiation, conduct, or completion of clinical investigations,”¹⁴³ the FDA can and should consider whether it is receiving enough INDs for RCTs for stool and stool-derived products, and whether existing RCTs are accruing participants. Further, the regulation for expanded access for intermediate-size patient populations states that if a relevant clinical trial is underway, sponsors of the expanded access IND/protocol must “explain why the patients to be treated cannot be enrolled in the clinical trial and under what circumstances the sponsor would conduct a clinical trial in these patients.”¹⁴⁴ The FDA probably has authority to require that patients be referred to a relevant RCT (one that is currently accruing participants) before they are considered for treatment under expanded access. This article's proposal would also incentivize clinicians at research-capable institutions to refer patients to RCTs by prioritizing the grant of expanded access INDs to entities that successfully refer patients to RCTs.

The proposal would decrease patients' access to FMT outside of studies, at least for patients who receive care at research-capable institutions. Given the goal of producing better evidence, incentivizing patients to receive FMT or a stool-derived microbial product through a study is a feature of the proposal. Patients who are ineligible or unable to participate in a study often would be eligible for expanded access. If a patient did not qualify for expanded access, there might be a good reason why she or he should not be receiving FMT. Overall, the aim of this proposal is not to substantially change the number of people who receive unapproved stool for FMT, but to change the context in which they receive it to one in which more and better evidence can be generated.

Expanded access requires submissions to the FDA, and therefore places burdens on clinicians and their institutions; it also provides some social benefits that enforcement discretion does not. Under expanded access the FDA would know who is doing FMT, for what purpose, and by what method. It could ensure that clinical protocols for performing FMT reflect up-to-date knowledge. The agency would also receive some adverse event reports related to the expanded access use.¹⁴⁵ These features would make FMT safer for patients. Furthermore, expanded access for an intermediate-size patient population aims to minimize administrative burdens by permitting many patients to be treated under a single expanded-access IND or protocol.

This article's proposal places stool for FMT on a similar regulatory footing to stool-derived microbial products by removing much of the enforcement discretion now granted to stool. It also provides similar incentives for institutions to participate in or refer patients to studies of stool for FMT or studies of stool-derived microbial products. This more even regulatory approach should help secure investment in research and development for stool-derived microbial products.

Another salutary feature of this article's proposal is that it would retain much of the market for stool banks. Stool banks would not be regulatorily disadvantaged in comparison to in-house stool production, and could still distribute their products to some institutions under enforcement discretion (although this sector of the market would shrink considerably). Existing stool banks have played an important role in educating patients and advocating for them, and have played an important role in research on FMT.¹⁴⁶ For these reasons, and because a product from stool banks likely is safer for some patients than in-house-produced stool, a regulatory policy should seek to retain stool banks. However, because the risks posed by centralized manufacturing and distribution of stool products are real, either the banks or the frequent recipients (heavy users) of their products should have INDs. The FDA should know who is producing stool for wide-spread distribution, how they are screening and tracking it, and to whom they are distributing it.

Governance involves more than regulation. Modern scholarship views governance as a polycentric activity in which the expertise to ensure adequate oversight is distributed among many actors.^{Reference Freeman, Dorf, Sabel, Lobel, Burris, Drahos, Shearing and Solomon147} For this and other reasons, scholars view agencies' engagement with stakeholders as a crucial tool of governance.^{Reference Lobel, Levi-Four and Trubek148} This article proposes that the FDA should undertake extensive and varied stakeholder engagement to generate ideas regarding RCT designs that are: (1) acceptable and attractive to patients and their clinicians; (2) capable of efficiently producing data the FDA will accept in support of a BLA for a stool-derived microbial product or for stool used in FMT; and (3) feasible and affordable.

To generate high-quality evidence, researchers and the FDA must find RCT designs that attract participants and clinicians. A new approach to regulatory enforcement will not promote evidence generation for FMT or stool-derived microbial products if patients turn to DIY FMT instead of clinical trials. Given patients' current enthusiasm for FMT, and practice guidelines recommending it as a treatment for rCDI, it seems likely that RCTs for both stool and stool-derived microbial products to treat rCDI will need to offer FMT to participants whose symptoms do not resolve during the trial. The obvious approach is to offer FMT as an open-label extension following the RCT portion of a trial.¹⁴⁹ Open-label extensions could provide some information on the longer-term safety of FMT; however, their main purpose would be to encourage enrollment in RCTs. Also, such a trial design could slightly expand the market for products from stool banks. Open-label extension with FMT has been used in some current or recent RCTs.^{Reference Johnsen150}

Stakeholder engagement might elicit other study design features that could attract participants to RCTs. For instance, a survey indicated that patients' most significant concern about FMT is disease transmission.¹⁵¹ RCTs that emphasize product screening and patient safety monitoring might, therefore, excel at accruing participants.

Stakeholder engagement should also help the FDA and research sponsors agree on trial designs that are efficient and likely to produce evidence that can support a BLA. They could address questions such as whether there are surrogate endpoints that could be measured in RCTs for FMT or stool-derived products. Engagements could help identify biomarkers or other tools that researchers in industry and academia should be trying to qualify.¹⁵² The use of biomarkers and valid surrogate endpoints could lead to more efficient production of high-quality evidence.

In addition to its merits, this article's proposal has some drawbacks. Most significantly, the FDA could generate powerful opposition by regulating in-house production of stool for FMT. Some organizations will argue that the FDA has no jurisdiction over in-house manufacturing and use. However, if stool produced by banks is regulated while stool produced in-house is not, then some (perhaps many) research-capable healthcare institutions will simply switch from using stool banks to producing stool in-house. These research-capable institutions would continue providing FMT as an unapproved treatment without generating evidence regarding its safety and efficacy.

Under the FDA's interpretation of its authorities, it has jurisdiction if a product meets the definition of a drug and is manufactured using materials obtained through interstate commerce. One federal appellate court and one federal district court have affirmed this legal interpretation with respect to production and use of stem cells.^{Reference Grady and Can153} Nonetheless, some health care organizations would view regulation of in-house stool production as agency overreach.

The FDA's approach to implementation would influence the degree of opposition to this article's proposed policy. For instance, before ceasing enforcement discretion the agency should provide enough time for investigators at research-capable healthcare organizations to find collaborators and initiate studies, begin referring patients to RCTs, and otherwise adjust to the coming policy change. Stakeholder engagements could help research-capable institutions better understand the proposed policy and help the FDA identify effective processes for implementation.