very little association of use in a normal population with any adverse effects of real concern. There are some notable exceptions, she said. For example, Besselink et al. (2008) reported increased mortality after administration of probiotics to patients with severe pancreatitis. Sanders cautioned that not only do different microorganisms have different safety patterns, but that the health status of the host also influences safety.
Dan Levy emphasized the need to examine strain-specific safety effects. He said, “The tendency has been to say, well, we’re exposed to all of these lactic acid bacteria with no adverse effects.” But not all lactic acid bacteria are the same. While some strains may cause no adverse effects, that is not necessarily the case for all strains. This is especially worrisome when a strain is chosen for a unique property, that very uniqueness suggesting that the strain does not behave like others, particularly when selecting for strains intended to have specific and perhaps novel physiological effects on the consumer. “You have to stop lumping all lactic acid bacteria together,” Levy said. “We’re looking at specific physiological mechanisms associated with specific strains, and we have to study the good, the bad, and the ugly.”
Experimental Design and Study Size
The discussion of microbiome markers of health and disease prompted a couple of comments on experimental design. Ellen Silbergeld expressed concern that too many studies on the microbiome are “almost entirely underpowered,” raising serious questions about the value of the information being provided by those studies. If a study is too underpowered to provide any evidence of an effect, then what is the value of that study? “If you really want to move this field forward,” she said, “you really have to start considering your study design.” In response, Johan van Hylckama Vlieg remarked that many small studies, such as the twin study that he mentioned, are intended to be exploratory and hypothesis-generating. They are not intended to yield the same type of results that large-scale clinical trials provide. Silbergeld then wondered whether any of the larger studies actually meet the criteria of a clinical trial. It is not clear that any do. Versalovic questioned whether large-scale clinical trials are even appropriate for microbiome studies. Given such extreme individual-level variation in microbiome composition and function, he said, “I don’t know that we can do trials the same way anymore,” indicating the need for further discussion on issues of study design.