Daniel Tancredi PhD, Professor of Pediatrics, UC Davis School of Medicine and Center for Healthcare Policy and Research, Sacramento, CA.
Decision makers frequently rely on p-values to decide whether and how to use a study to inform their decisions. Many misinterpret what a low p-value actually means, however. I will attempt to correct this common misinterpretation and explain how to use small p-values to evaluate whether a null hypothesis is plausible. I will show that a low p-value should be used in the same way that a clinician should use a valuable but imperfect clinical diagnostic test result; as one factor, but not the only one, on which to base a decision.
Typically, readers assume that if a p value is low, such as less than 0.05, and thus the test statistic is statistically significant at the conventional level, that there is a good chance that the study results are not “due to chance”. But it is not as simple as that. Let’s suppose that one has a p-value that was generated in a well-designed placebo-controlled randomized clinical trial. We will assume that the trial had a sample size that would provide 80% power to detect what the investigators considered to be the minimum clinical significant difference. The null hypothesis in a trial like this would predict that there is no difference between the placebo and intervention arms for the primary outcome. Once the data were analyzed and reported properly, the p-value was estimated to be just under 0.05. Does this p value less than 0.05 mean that the null hypothesis has no more than a 5% probability of being true? Does it even ensure that the null hypothesis is unlikely? If the calculated p-value was greater than 0.05 (let’s say p=0.10) would that be enough to ensure that the probability of the null hypothesis being true is definitely greater than 5%?
The answer to each of these questions is “no”! This is where, unfortunately, it begins to get (a little) complicated for many users of statistics. The p-value is calculated under the assumption that the null hypothesis is true, and so it does not and cannot measure the probability of that assumption being correct.1 Even though it is common to interpret a p-value as though it is an objective and sufficient statement about the probability that the null hypothesis is true, that is not the case. Statisticians have been trying to communicate this nuance for decades, including the issuance of a statement in 2016 on p-values by the American Statistical Association, a rare statement on statistical practice in that prominent organization’s long history1.
Using a p-value to calculate the probability that the null hypothesis is true
If one wants to use a p-value as one factor in a procedure that can produce a statement about the probability that the null hypothesis is true, one needs to supply an additional input that can be very difficult to obtain. This is a prior (or pre-study) probability, a quantitative estimate of the probability that the null hypothesis is true. This is based on a considered judgment of the state of existing knowledge, what is already known (outside the study results) about how and by how much the intervention may affect the outcomes being assessed.2 Of course, such a judgment can vary a great deal from one individual to the next, according to his/her ability to gather and appraise that knowledge. These judgments can also be influenced by other interests, including financial and ideological, how much of a stake one has in each of the various competing scientific explanations. Depending on the prior (or pre-experiment) probability for the null hypothesis, a p-value of 0.01 may not be enough to ensure that the null hypothesis has less than a 50% posterior (or after-experiment) probability of being true, whereas a p-value of 0.10 may be enough to make the posterior probability of the null hypothesis be comfortably under 5% (for those interested in taking this discussion further, the final section of this post illustrates this in more detail and with examples).
Determining whether an intervention works
Fundamentally, p-values cannot be used by themselves as though they are objective and reliable ways to make prudent decisions about whether an intervention works. Statisticians emphasize the necessity for the results of individual studies to be interpreted in a broader context, one that involves both statistical judgment and judgment on the underlying scientific plausibility of the hypothesized effects. It is well known that when sample sizes are very large, such as in many observational studies involving tens or even hundreds of thousands of observations, p-values can be very low, even for effect sizes whose confidence intervals are relatively narrow yet do not include any effects that would be of practical importance. In evidence-based medicine we typically face the opposite challenge, where small sample sizes and/or relatively infrequent outcome events result in p-values greater than 0.05 and 95% confidence intervals that are ambiguous because they include the null value (as is implied by p>0.05), but with outcomes that would be very important clinically. Thankfully, in my own career, it does seem to me to have become better appreciated that simply describing studies as positive or negative depending on which side of 0.05 the p-value falls is an unreliable method for evaluating evidence.
p-values and meta-analysis
Another thing to keep in mind is that even when a majority of individual studies that address the same research question may have p-values above 0.05, the meta-analysis of those study results can still indicate a statistically and clinically significant effect. As an example I will use a 2017 Cochrane review of the use of probiotics for the prevention of Clostridioides difficile‐associated diarrhea (CDAD) in adults and children.3 The overwhelming majority of studies, 17 of 21, were supposedly “negative” in that they have confidence intervals that include the null value, but the overall pooled estimate reports a statistically significant and clinically important range of effects. Also note that the overwhelming majority of the studies report confidence intervals that are consistent with the confidence interval for the overall pooled estimate, when one considers the degree of overlap. See Figure 1 below.
Figure 1. Forest plot summarizing complete-case analyses from systematically reviewed clinical trials of probiotics for the prevention of Clostridium difficile‐associated diarrhea (CDAD) in adults and children. Although only 4 of the 31 individual trials had statistically significant results, the pooled estimate shows a statistically and clinically signficant reduction in risk of CDAD for the studied probiotics, without statistically significant heterogeneity among the individual trials’ relative risk estimates. Note that the confidence inferval for the pooled estimate is entirely contained by all but two of the confidence intervals from the individual trials and that even the confidence intervals from these two exceptions largely contain the pooled estimate.
Reprint of Figure 3 from Joshua Z Goldenberg, Christina Yap, Lyubov Lytvyn, et al’s “ Probiotics for the prevention of Clostridium difficile‐associated diarrhea in adults and children”, published December 12, 2017 in “Cochrane Database of Systematic Reviews” by John Wiley and Sons. Copyright by John Wiley and Sons. Reprinted under one-time use license from John Wiley.
Summary
In conclusion, p values are an important component of determining whether an outcome can be deemed to be statistically significant, but this depends on the question under investigation, and is only one part of a more complete analysis. When appraising evidence for whether an intervention works, it is important to keep in mind that if one relies only on statistical inferences from individual studies, one is vulnerable to making unreliable assessments that substantially misstate the plausibility that an intervention does (or does not) have an effect. Statistical analysis cannot replace context-specific scientific judgment; both are needed to make reliable evidence appraisals.
A deeper dive into how to use p-values to assess the probability that the null hypothesis is true
A common misinterpretation of p-values is that they measure the probability that the null hypothesis is true, given the sample data. As stated above, the p-value, by itself, cannot speak to this probability, but if one is willing to supply a judgment on the prior probability that the null hypothesis is true, one can use that and the p-value to get a lower bound on the probability of interest. The compelling figure that accompanies Regina Nuzzo’s terrific Nature article on p-values and their shortcomings nicely illustrates such results for six combinations involving three example prior probabilities and p-values of 0.05 and 0.01.4 Table 1 shows posterior probabilities for those and other input combinations.
| Table 1. Plausible lower bound for the posterior (post-study) probability of the null hypothesis being true for a given prior (pre-study) probability and study p-value. Note that low p-values do not necessarily imply that the null hypothesis is unlikely to be true! |
|
P-value |
| Prior Probability for Null |
0.1000 |
0.0500 |
0.0100 |
0.0050 |
0.0010 |
0.0005 |
0.0001 |
| 5% |
3.2% |
2.1% |
0.7% |
0.4% |
0.1% |
0.1% |
0.0% |
| 10% |
6.5% |
4.3% |
1.4% |
0.8% |
0.2% |
0.1% |
0.0% |
| 25% |
17.3% |
12.0% |
4.0% |
2.3% |
0.6% |
0.3% |
0.1% |
| 50% |
38.5% |
28.9% |
11.1% |
6.7% |
1.8% |
1.0% |
0.2% |
| 75% |
65.3% |
55.0% |
27.3% |
17.8% |
5.3% |
3.0% |
0.7% |
| 90% |
84.9% |
78.6% |
53.0% |
39.3% |
14.5% |
8.5% |
2.2% |
| 95% |
92.2% |
88.6% |
70.4% |
57.8% |
26.3% |
16.4% |
4.5% |
| Note: Calculations use the Bayes Factor – e p ln(p), which is shown to be a lower bound for the Bayes Factor among an appealing set of candidates, thus resulting in a plausible “lower bound” for the posterior probability that the null hypothesis is true. For example, when the prior probability is 50%, a p-value of 0.05 implies that the null hypothesis retains at least a 28.9% probability of being true. |
The calculations used in that figure and in Table 1 for converting the two inputs, a prior probability for the null hypothesis and a p-value, into a posterior probability for the null hypothesis is simply an application of a much more general formula, one that has been known for over 200 years. This formula is simple to state and remember when expressed as odds. According to Bayes Theorem, Posteriors Odds equals Prior Odds multiplied by a term we call the Likelihood Ratio. The likelihood ratio is a ratio of two conditional probabilities for the observed data, with each computed under differing hypotheses.5 [Another very widely-used application of this general formula is when physicians use the results and the operational characteristics (e.g. the sensitivity and specificity) of clinical tests to inform medical diagnoses.6] The formula uses odds not in the way that they are defined in horse racing where long-shots have high odds, but in the way that statisticians define it, as the ratio of the probability of an event to the probability of the absence of that event. To a statistician, high odds mean high probability for the event. When the probability of an event P is greater than 0, the odds are P / (1 – P). For example, if the probability of an event is 0.75 (or 75%), then the odds would be 0.75 / ( 1 – 0.75 ) = 3. If one knows the odds O, then one can find the probability P, using the equation P = O / ( 1 + O ). For example, if the odds are 4:1, or 4, then the probability is 4/5 = 0.80, and if the odds are 1:4, or 0.25, then the probability is 0.25 / 1.25 = 0.20.
In order to use this long-known formula, one has to have a way to convert the p-value into a value to use for the “Likelihood Ratio” term, which in this context is called a Bayes Factor. For the Nature article, Nuzzo used a conversion proposed in the 1990s by Thomas Sellke, M. J. Bayarri, and James O. Berger and that they eventually published in the widely read American Statistician. That conversion has an appealing statistical motivation as the minimum possible value for the Bayesian Factor among a realistic set of candidates and thus it provides a useful plausible lower bound on the Bayesian Factor for p < 1 /e ≈ 0.368, where e is the Euler number, exp(1) ≈ 2.718,7 BayesFactor = – e * p * ln(p), where ln(p) is the natural logarithm of p. (For p ≥1/e, one can use BayesFactor=1.) For example, p=0.04 would result in a BayesFactor of -exp(1) * 0.04 * ln( 0.04 ), approximately 0.35. So, if one specified that the prior probability for the null hypothesis is 50%, a toss-up, that corresponds to a prior odds of 1, then the BayesFactor for a p-value of 0.04 converts that prior odds of 1 into a posterior odds of 0.35, which corresponds to a posterior probability of 26% for the null hypothesis, substantially higher than 4%. In the analogous setting of diagnostic medicine, consider a test result that moves a physician’s suspicion for whether the patient has a disease from a pre-test value of 50% up to a post-test value of 74%. Such a result would be considered useful, but it would not be considered definitive, something for clinicians to keep in mind when they see that a study’s p-value was just under 0.05!
Another notable conversion of the p-value into a Bayes Factor was, as far as I can tell, first reported in a pioneering 1963 article in the social sciences literature that was authored by illustrious Bayesian statisticians. 8 That same Bayes Factor formula can be found clearly presented in the second5 of Steven N. Goodman’s excellent two-part set of Annals of Internal Medicine articles concerning fallacious use of p-values in evidence-based medicine. That conversion involves statistics that have an approximately normal distribution and is thus applicable to most statistics in the medical literature. That conversion reports the minimum theoretically possible value for the Bayes Factor, BayesFactormin = exp( – Z2 / 2 ), where Z is the number of standard errors the test statistic is from the null value. (Z can be estimated in Microsoft Excel by using the formula Z = NORMSINV( p ) or Z = NORMSINV( p / 2 ). For example, a two-sided p-value of 0.04 corresponds to Z ≈ -2.054 and a BayesFactormin of exp( – (-2.054 * -2.054) / 2 ) ≈ 0.121. So, if the prior probability for the null hypothesis is 50%, a p-value of 0.04 would mean that, at the minimum, the null hypothesis has a posterior probability of 0.121 / 1.21 = 10.8% of being true, substantively higher than the 4% probability that the popular misinterpretation of p-values would yield. When that factor was introduced in the 1963 article, it was noted by the authors as not being one that would be realistically attained by any study, as it would involve an impossibly lucky guess for the best possible prior probability to use, but it is still useful mathematically because it results in a theoretical minimum for the posterior probability that the null hypothesis is true. In mathematics, we routinely use well-chose impractical scenarios to define the limits for what is practically possible. Given that decisionmakers want to know how probable the null hypothesis remains in light of the study data, it is helpful to know the minimum possible theoretical value for it. Table 2 shows these posterior probabilities for the same inputs used above in Table 1. Notably, a p-value of 0.05 may not even be enough to make the null hypothesis less likely than not!
|
|
P-value |
| Prior Probability for Null |
0.1000 |
0.0500 |
0.0100 |
0.0050 |
0.0010 |
0.0005 |
0.0001 |
| 5% |
1.3% |
0.8% |
0.2% |
0.1% |
0.0% |
0.0% |
0.0% |
| 10% |
2.8% |
1.6% |
0.4% |
0.2% |
0.0% |
0.0% |
0.0% |
| 25% |
7.9% |
4.7% |
1.2% |
0.6% |
0.1% |
0.1% |
0.0% |
| 50% |
20.5% |
12.8% |
3.5% |
1.9% |
0.4% |
0.2% |
0.1% |
| 75% |
43.7% |
30.5% |
9.8% |
5.5% |
1.3% |
0.7% |
0.2% |
| 90% |
69.9% |
56.9% |
24.6% |
14.9% |
3.9% |
2.1% |
0.5% |
| 95% |
83.1% |
73.6% |
40.8% |
27.0% |
7.8% |
4.3% |
1.0% |
| Note: For 2-sided p-values based on approximately normally distributed test statistics, using the mathematically lowest theoretically possible Bayesian Factor,5,8 thus ensuring the lowest possible value for the posterior probability for the null hypothesis. Although these lower bonds would never be attained in any realistic application, this table is useful in showing the smallest null hypothesis probability that is even theoretically possible. Note that even with a p-value of 0.05, the posterior probability for the null hypothesis may still be high. |
References
- Wasserstein RL, Lazar NA. The ASA Statement on p-Values: Context, Process, and Purpose. The American Statistician. 2016;70(2):129-133.
- Goodman SN. Toward evidence-based medical statistics. 1: The P value fallacy. Annals of Internal Medicine. 1999;130(12):995-1004.
- Goldenberg JZ, Yap C, Lytvyn L, et al. Probiotics for the prevention of Clostridium difficile-associated diarrhea in adults and children. Cochrane Database Syst Rev. 2017;12:CD006095.
- Nuzzo R. Statistical Errors. Nature. 2014;506(7487):150-152.
- Goodman SN. Toward evidence-based medical statistics. 2: The Bayes factor. Annals of Internal Medicine. 1999;130(12):1005-1013.
- Deeks JJ, Altman DG. Diagnostic tests 4: likelihood ratios. Bmj. 2004;329(7458):168-169.
- Sellke T, Bayarri MJ, Berger JO. Calibration of p values for testing precise null hypotheses. Am Stat. 2001;55(1):62-71.
- Edwards W, Lindman H, Savage LJ. Bayesian Statistical-Inference for Psychological-Research. Psychol Rev. 1963;70(3):193-242.
Can diet shape the effects of probiotics or prebiotics?
/in Consumer Blog, ISAPP Science Blog /by KCBy Prof. Maria Marco PhD, University of California – Davis and Prof. Kevin Whelan PhD, King’s College London
If you take any probiotic or prebiotic product off the shelf and give it to several different people to consume, you might find that each person experiences a different effect. One person may notice a dramatic reduction in gastrointestinal symptoms, for example, while another person may experience no benefit. On one level this is not surprising, since every person is unique. But as scientists, we are interested in finding out exactly what makes a person respond to a given probiotic or prebiotic to help healthcare providers know which products to recommend to which people.
Among factors that might impact someone’s response to a probiotic or prebiotic – such as baseline microbiota, medications, and host genetics – diet emerges as a top candidate. Ample evidence has emerged over the past ten years that diet has direct and important effects on the structure and function of the gut microbiome. Overall the human gut microbiome is shaped by habitual diet (that is, the types of foods consumed habitually over time), but the microbes can also can fluctuate in response to short-term dietary shifts. Different dietary patterns are associated with distinct gut microbiome capabilities. Since probiotics and prebiotics may then interact with gut microbes when consumed, it is plausible that probiotic activity and prebiotic-mediated gut microbiome modulation may be impacted by host diet.
A discussion group convened at ISAPP’s 2022 annual meeting brought together experts from academia and industry to address whether there is evidence to support the impact of diet on the health effects of probiotics and prebiotics. To answer this question, we looked at how many probiotic or prebiotic studies included data on subjects’ diets.
Our expert group agreed that diet should be included in the development of new human studies on probiotics and prebiotics, as well as other ‘-biotics’ and fermented foods. These data are urgently needed because although diet may be a main factor affecting outcomes of clinical trials for such products, it is currently a “hidden” factor.
We acknowledge there will be challenges in taking diet into account in future trials. For one, should researchers merely record subjects’ habitual dietary intake, or should they provide a prescribed diet for the duration of the trial? The dietary intervention (nutrient, food, or whole diet) must also be clearly defined, and researchers should carefully consider how to measure diet (e.g. using prospective or retrospective methods). In the nutrition field, it is well known that there are challenges and limitations in the ways dietary intake is recorded as well as the selection of dietary exclusion criteria. Hence, it is crucial that dietitians knowledgeable in dietary assessment and microbiome research contribute to the design of such trials.
If more probiotic and prebiotic trials begin to include measures of diet, perhaps we will get closer to understanding the precise factors that shape someone’s response to these products, ultimately allowing people to have more confidence that the product they consume will give them the benefits they expect.
Episode 8: The link between digestive symptoms, IBS and the gut microbiota: A gastroenterologist’s perspective
/in Podcast, Season One /by KCPodcast: Play in new window | Download
Subscribe: Apple Podcasts | Spotify | RSS
The Science, Microbes & Health Podcast
This podcast covers emerging topics and challenges in the science of probiotics, prebiotics, synbiotics, postbiotics and fermented foods. This is the podcast of The International Scientific Association for Probiotics and Prebiotic (ISAPP), a nonprofit scientific organization dedicated to advancing the science of these fields.
The link between digestive symptoms, IBS and the gut microbiota: A gastroenterologist’s perspective, with Prof. Eamonn Quigley
Episode summary:
In this episode, the ISAPP hosts focus their discussion around irritable bowel syndrome (IBS) with Prof. Eamonn Quigley, MD, of Weill Cornell Medical College. Prof. Quigley says patients are increasingly curious about the link between IBS and gut microbiota. He outlines what we know so far about the etiology of IBS, and the evidence for how gut microbiota may contribute to the condition as well as possible interventions that target the gut microbes.
Key topics from this episode:
The typical symptoms is abdominal pain associated with a disturbance in bowel function which could be diarrhea or constipation, or even alternating between them, depending on the patient.
Estimates say 5-10% of all people globally have IBS.
There is no clear cause for IBS identified to date. IBS has been linked to the gut-brain axis (as it often co-occurs with depression and anxiety), gut microbiota, diet, previous gastrointestinal infections (Salmonella, Shigella, Campylobacter infections), and antibiotic use. It is also more common in women.
Approaches have tended to focus on treatment of symptoms: for example, treating the pain or diarrhea. Diet has also become an essential part of IBS treatment. But overall quality of life for IBS patients is of crucial importance. The focus should not be only on treating symptoms but also on improving their quality of life.
Episode abbreviations and links:
FODMAP: fermentable oligosaccharides, disaccharides, monosaccharides and polyols (i.e. types of carbohydrates that are poorly absorbed in the small intestine).
EMA: European Medicines Agency (i.e. the European counterpart of the US Food and Drug Administration)
Study: Lactobacillus and bifidobacterium in irritable bowel syndrome: Symptom responses and relationship to cytokine profiles
CME course on digestion and gut microbiota: Android version, iOS version, web version
Additional resources:
I have IBS – should I have my microbiome tested? ISAPP blog
The Microbiome — Can it aid in the diagnosis and therapy of irritable bowel syndrome (IBS)? ISAPP blog
About Prof. Eamonn Quigley:
Eamonn M M Quigley MD FRCP FACP MACG FRCPI MWGO is David M Underwood Chair of Medicine in Digestive Disorders and Chief of the Division of Gastroenterology and Hepatology at Houston Methodist Hospital. A native of Cork, Ireland, he graduated in medicine from University College Cork. He trained in internal medicine in Glasgow, completed a two-year research fellowship at the Mayo Clinic and training in gastroenterology in Manchester, UK. He joined the University of Nebraska Medical Center in 1986 where he rose to become Chief of Gastroenterology and Hepatology. Returning to Cork in 1998 he served as Dean of the Medical School and a PI at the Alimentary Pharmabiotic Center. He served as president of the American College of Gastroenterology and the WGO and as editor-in-chief of the American Journal of Gastroenterology.
Interests include IBS, gastrointestinal motility and the role of gut microbiota in health and disease. He has authored over 1000 publications and has received awards and honorary titles world-wide. Married for over 40 years to Dr Una O’Sullivan they have 4 children and three grandchildren. Interests outside of medicine include literature, music and sport and rugby, in particular; Dr Quigley remains a passionate supporter of Munster and Irish rugby.
Episode 7: Evidence for probiotic use in pediatric populations
/in Podcast, Season One /by KCPodcast: Play in new window | Download
Subscribe: Apple Podcasts | Spotify | RSS
The Science, Microbes & Health Podcast
This podcast covers emerging topics and challenges in the science of probiotics, prebiotics, synbiotics, postbiotics and fermented foods. This is the podcast of The International Scientific Association for Probiotics and Prebiotic (ISAPP), a nonprofit scientific organization dedicated to advancing the science of these fields.
Evidence for probiotic use in pediatric populations, with Prof. Michael Cabana
Episode summary:
In this episode, the ISAPP hosts discuss probiotics for pediatric populations with Prof. Michael Cabana, MD, MPH, from Albert Einstein College of Medicine and The Children’s Hospital at Montefiore. Prof. Cabana starts by acknowledging the gap between the demand for probiotic interventions and the evidence that currently exists for their efficacy. He gives an overview of the challenges in designing trials on probiotic interventions for children, and summarizes what the evidence shows to date.
Key topics from this episode:
Episode links:
Additional resources:
ISAPP Digs Deeper into Evidence on Probiotics for Colic with New Meta-Analysis. ISAPP blog
Probiotics to Prevent Necrotizing Enterocolitis: Moving to Evidence-Based Use. ISAPP blog
About Prof. Michael Cabana, MD:
Prof. Michael Cabana, MD, MPH, is a Professor of Pediatrics & the Michael I. Cohen University Chair of Pediatrics at Albert Einstein College of Medicine, as well as Physician-in-Chief, The Children’s Hospital at Montefiore (CHAM). He is also a member of the United States Preventive Services Task Force USPSTF (here), a prestigious appointment for medical personnel to weigh evidence (risk vs. harms) on prevention interventions recommended in the United States. He is a clinical trialist (see the trials listed here), with a focus on allergy in children. He has also conducted trials using probiotic interventions. Prof. Cabana served on the ISAPP board of directors from 2008 to 2018. He has an MD from University of Pennsylvania, an MPH from Johns Hopkins, and an MA in business from Wharton Business School.
Dr. Cabana’s comments do not necessarily reflect the views of the USPSTF.
Human milk oligosaccharides as prebiotics to be discussed in upcoming ISAPP webinar
/in Consumer Blog, ISAPP Science Blog /by KCHuman milk oligosaccharides (HMOs), non-digestible carbohydrates found in breast milk, have beneficial effects on infant health by acting as substrates for immune-modulating bacteria in the intestinal tract. The past several years have brought an increase in our understanding of how HMOs confer health benefits, prompting the inclusion of synthetic HMOs in some infant formula products.
These topics will be covered in an upcoming webinar, “Human milk oligosaccharides: Prebiotics in a class of their own?”, with a presentation by Ardythe Morrow PhD, Professor of Pediatrics, Cincinnati Children’s Hospital Medical Center, University of Cincinnati College of Medicine. The webinar will provide an overview of what HMOs are, how they are breaking new ground with the types of health benefits they can provide to infants and the recent technological innovations that will facilitate their translation into new infant formulas.
Dr. Karen Scott, Rowett Institute, University of Aberdeen, and Dr. Margriet Schoterman, FrieslandCampina, will host the webinar. All are welcome to join this webinar, scheduled for Wednesday, Oct 19th, 2022, from 10-11 AM Eastern Daylight Time.
Registration is now closed. Please watch the recording of this webinar below.
Can Probiotics Cause Harm? The example of pregnancy
/in ISAPP Science Blog /by KCBy Prof. Dan Merenstein MD, Georgetown University School of Medicine, Washington DC, USA and Dr. Maria Carmen Collado, Institute of Agrochemistry and Food Technology-National Research Council (IATA-CSIC), Valencia, Spain
Limiting excessive weight gain and controlling blood pressure during pregnancy are important to prevent pre-eclampsia and other complications of pregnancy. Researchers have examined if there is a role for probiotics in maintaining a healthy pregnancy. A recent Cochrane review, which evaluated evidence on probiotics for preventing gestational diabetes (GDM), concluded, “Low-certainty evidence from six trials has not clearly identified the effect of probiotics on the risk of GDM. However, high-certainty evidence suggests there is an increased risk of pre-eclampsia with probiotic administration.” This was an unexpected conclusion, which raised concerns about probiotic safety. A close look at the basis for this statement is warranted to determine if certain strains of probiotics are contraindicated for pregnant women.
Most people familiar with probiotic science understand that giving anyone live bacteria carries some risk. The definition of probiotics is live microorganisms that, when administered in adequate amounts, confer a health benefit on the host. It is not live microorganisms that, when administered in adequate amounts, confer a health benefit on the host that outweighs potential adverse events. But clinicians understand that risk versus benefit must be considered for all interventions.
Many interventions associated with significant positive outcomes also are associated with some adverse events, some quite significant. For example, a recent United States Preventive Services Task Force report found that beta carotene, with or without vitamin A, was significantly associated with an increased risk of lung cancer and cardiovascular disease mortality. Aspirin kills thousands of people each year, with many more hospitalized with significant bleeds. While for an exercise doctors recommend all the time, biking, the CDC reports nearly 1,000 bicyclists die and over 130,000 are injured in crashes every year in the US.
Studies that led to the Cochrane conclusion
But let’s get back to trying to understand what made the Cochrane review come out with this warning about probiotics and pre-eclampsia. Turns out the conclusion was based on four randomized clinical trials which reported pre-eclampsia as an adverse event. All four studies were well done with low risk of bias per the Cochrane report.
Here is a summary of the four studies that collected preeclampsia data, included in the Cochrane review:
Callaway et al.(2019) studied a mixture of Lactobacillus rhamnosus (LGG) and Bifidobacterium animalis subspecies lactis BB-12 for the prevention of gestational diabetes The reported pre-eclampsia in the probiotic group was 19 (9.2%) participants compared to 10 (4.9%) in the placebo group, p-value=0.09. This was in an obese cohort, with an average BMI of both groups near 32 (kg/m2).
Lindsay et al. (2014) evaluated the effect of Lactobacillus salivarius UCC118 on maternal fasting glucose. They reported preeclampsia in 3/62 in the probiotic group versus 2/74 in the placebo group (p-value >0.366). Again, this was in an obese cohort with early pregnancy BMIs in the probiotic group, averaging 32.9 versus 34.1 in the placebo group.
Pellonpera et al. (2019) conducted a 4-arm study to determine if fish oil and or Lactobacillus rhamnosus HN001 and Bifidobacterium animalis ssp. lactis 420 could prevent gestational diabetes. In total there were 10 cases of pre-eclampsia among the four groups as shown below, (each group had about 95 total participants) and no significant differences between them, p value=0.80.
Okesene-Gafa et al. (2019) published in the American Journal of Obstetrics and Gynecology in 2019 looking at culturally tailored dietary intervention and or daily probiotic capsules containing lactobacillus rhamnosus GG and Bifidobacterium lactis BB12 impact pregnancy weight-gain and birthweight. (This was also an obese cohort with an average BMI of 38.8.) They found pregnancy induced hypertension in the probiotic group in 4/96 (4.2%) of women versus 2/93 (2.2%) in the placebo group (p value=0.31).
Is there a rationale for the preeclampsia warning?
The increased rate of preeclampsia in probiotic groups was only with studies using obese subjects. Importantly, obesity has been associated with a higher risk of preeclampsia (see here and here). A recent meta-analysis, which included 86 studies representing 20,328,777 pregnant women, showed that higher BMI is associated with adverse pregnancy outcomes, among them, gestational diabetes and preeclampsia. Furthermore, the adjusted risk of preeclampsia is estimated to be double for overweight mothers and almost triple for obese mothers, compared to normal weight mothers.
It has been reported that pro-inflammatory signals (TNF-alpha, IL6) produced in adipose tissue of obese individuals induces a proinflammatory state characterized by insulin resistance and altered endothelial function. The gut microbiota is also disrupted in these individuals, consistent with observations that report an altered gut microbiota composition in obese versus lean individuals (see here, and the effects on offspring here and here). This suggests that obese mothers may have an increased risk of adverse events, but still the evidence supports that the addition of certain strains of probiotics may exacerbate this risk. Furthermore, it is relevant to mention the accumulating data showing that during gestation in parallel to the physiological, immune and metabolic adaptations, gut microbiota changes over the pregnancy (see here, here, here and here) although little is known on the impact of pre-gestational BMI on gut microbiota changes during pregnancy. However, specific microbial shifts have been reported to be predictive of GDM and also, gut microbial differences in women with and without GDM have been reported (here and here) . It has been also reported that the gut microbiota shifts (in composition and activity metabolites) in women with preeclampsia (see here). Thus, it is quite possible that the women in these studies, obese women, react to gut-microbiota-related interventions differently than non-obese women and that their pre-pregnancy weight puts them at an increased risk of complications.
It is worth noting that the total number of cases cited in the Cochrane review supporting their conclusion was 31 cases of preeclampsia in 472 women who took probiotics versus 17 in 483 women in the placebo groups. Thus, 14 more women who experienced preeclampsia, 9 of whom came from one of the studies [“probiotics increase the risk of pre-eclampsia compared to placebo (RR 1.85, 95% CI 1.04 to 3.29; p-value=0.04; 4 studies, 955 women; high-certainty evidence”] This is not a very large number of subjects for such a strong conclusion. The authors don’t mention if this high-certainty evidence is in all women or just obese women. By combining four studies, in which none found a significant increase in preeclampsia, the authors did find significance. Is this a convincing number of subjects? The Cochrane author, Dr. Marloes Dekker Nitert replied to an inquiry from us that she believes that this difference makes it unethical to conduct further studies in pregnant women, stating, “I think that there now is a lack of clinical equipoise to do an RCT on a combination of Lactobacillus/Bifidobacterium.”
This is a strong statement but is consistent with their high-certainty of evidence statement. We acknowledge that something does appear to be going on. It is possible that certain populations react differentially to certain strains. Thus, maybe mild to morbidly obese women are a subgroup that needs closer monitoring during pregnancy and maybe even in non-pregnant settings, as they may react differently to probiotic interventions. Maybe it is just certain strains, as the Cochrane author was very clear in her email to state, “a combination of Lactobacillus/Bifidobacterium” and not generalize to all probiotics. We agree and in fact it is possible that different strains of Lactobacillus/Bifidobacterium will have different outcomes. Pregnancy is also a continuum and to think that giving an intervention during the first trimester is the same as during the third makes little scientific or clinical sense. Along these lines, one study showed the association of probiotic intake with different effects in early versus late pregnancy; an analysis that specifically focused on women in the third trimester of pregnancy found no association between probiotics and adverse fetal outcomes.
Conclusions
In summary, we must recognize that certain strains of probiotics may cause harm in certain populations. This reinforces the importance of diligent collection of adverse event data during all clinical trials. Although Cochrane is renowned to conduct analyses of the highest caliber, we wonder if four studies of 955 mostly obese women, in which 14 more in the probiotic group than the placebo group have a secondary outcome of harm, warrant the conclusion that there is “high-certainty evidence” that probiotics cause harm. This seems overstated based on our review of the literature. Should women and clinicians pay particular attention to this subgroup (obese pregnant women) and this outcome (preeclampsia, hypertension)? We think the answer is yes. But we do not conclude that all women at all stages of pregnancy need to refrain from probiotics. Fortunately, at the time of writing there appear to be 87 trials listed on clinicaltrials.gov looking at probiotics and pregnancy. As in many things the details still need to be further elucidated and we expect more clarification on this issue over the next 5-10 years.
A pediatrician’s perspective on c-section births and the gut microbiome
/in Consumer Blog, ISAPP Science Blog /by KCBy Prof. Hania Szajewska, MD, Medical University of Warsaw, Poland and Kristina Campbell, MSc, ISAPP Consulting Communications Director
The decision to have a Cesarean section (C-section) should always depend on whether this is the best choice for the mother and baby, and it is never made by pediatricians. However, pediatricians are often asked about the consequences of C-section delivery for a child later in life and whether potential C-section-related harms may be reduced.
The data show that delivery by C-section is now more common than ever globally. The World Health Organization estimates the C-section rate is around 21% of all births, and predicted to continue increasing. Although C-section rates are increasing both in developed and developing countries, Korea, Chile, Mexico, and Turkey have the highest rates in the world, with C-sections constituting 45% to 53% of all births. C-sections outnumber vaginal births in countries that include Dominican Republic, Brazil, Cyprus, Egypt, and Turkey.
Cesarean delivery is a medical procedure that can of course save an infant (or a mother) in a moment of danger, making birth less risky overall. But analyses have shown not all C-sections are initiated for safety reasons—some are driven by convenience and other non-medical factors. In areas with the highest C-section rates, only around half of the time are they required for life-saving reasons. Although the rate of medically necessary C-sections globally is difficult to establish, the WHO estimates it is between 10-15% of all births.
Non-essential C-sections would be perfectly reasonable if the health risks later in life were negligible. But are they? Scientific work in the past decade has shown that, in fact, there may be downsides to being born by C-section—and these health risks may manifest later in a child’s life.
By now, many observational studies have associated Cesarean births with an increased risk of various chronic health conditions that appear long after birth. C-section is associated with a higher risk of asthma and allergy, as well as obesity and type 2 diabetes. A systematic review and meta-analysis (incorporating 61 studies, which together included more than 20 million deliveries) also linked C-sections with autism spectrum disorders and attention deficit hyperactivity disorder (ADHD). Type 1 diabetes is also more prevalent in children born by c-section.
Since association is not the same as causation, scientists have looked at possible biological correlates of C-section and how they could be tied to future health problems. A leading hypothesis is that C-section deliveries cause health problems by disrupting the infant’s normal gut microbiota (i.e. the collection of microorganisms in specific ‘habitats’ on the infant’s body, such as the gut) within a critical time window for immune system development.
An altered microbiota in C-section births
One of the main clues about whether C-section births affect health via the microbiota is the consistent observation that infants born by C-section have a different collection of microorganisms in their digestive tracts and elsewhere on their bodies immediately after birth, compared with vaginally-born controls. Newborns delivered by C-section tend to harbor in their guts disease-causing microbes commonly found in hospitals (e.g. Enterococcus and Klebsiella), and lack strains of gut bacteria found in healthy children (e.g. Bacteroides species). Because it is known that gut microbiota are in close communication with the immune system, this difference in birth microbes may set the immune system up for later dysfunction.
However, an important confounding factor exists. Antibiotic administration is a recommended medical practice for C-section births in order to prevent infections. Antibiotics are potent disruptors of microbial communities – in this case the mother’s, or perhaps the infant’s if antibiotics are administered prior to umbilical cord clamping. It is not yet clear whether the timing of antibiotic administration can prevent such disruptions. (See conflicting evidence here and here; also see here.).
Gut microbiota disruption is associated with C-sections, but since C-section and antibiotics nearly always go together (with potential exposure of the infant to these drugs), it is not clear to what extent C-section and/or antibiotic treatments drive increased risk of chronic disease later in life. Antibiotic treatments within the first 2 years of life are independently associated with an increased risk of several conditions: childhood-onset asthma, allergic rhinitis, atopic dermatitis, celiac disease, overweight / obesity, and ADHD.
Options for microbiota ‘restoration’
If mechanistic studies continue to support the idea that the C-section-disrupted gut microbiota is the trigger for chronic diseases later in life, strategies could be proposed for ‘restoring’ or normalizing the infant gut microbiota after such births. Already some microbiota modifying interventions have been evaluated:
So far, probiotics, synbiotics, and microbiota ‘restoration’ are not sufficiently reliable solutions for correcting the microbiota disruptions that accompany C-section births. Further studies are needed to develop these approaches.
A leading strategy
At present, breastfeeding is the main strategy for supporting the infant gut microbiota after C-section for the greatest chance of avoiding negative health consequences. Breastfeeding has multiple benefits, but may be of increased importance after C-section birth. Mothers should be supported after giving birth by C-section to breastfeed the infant during this critical period of early life and immune system development.
What is a strain in microbiology and why does it matter?
/in ISAPP Science Blog /by KCBy Prof. Colin Hill, Microbiology Department and APC Microbiome Ireland, University College Cork, Ireland
At the recent ISAPP meeting in Sitges we had an excellent debate on the topic of ‘All probiotic effects must be considered strain-specific’. Notwithstanding which side of the debate prevailed, it does raise the question: what exactly is a strain? As a card-carrying microbiologist I should probably be able to simply define the term and give you a convincing answer, but I find that it is a surprisingly difficult concept to capture. It is unfortunately a little technical as a topic for a light-hearted blog, but here goes. Let me start by saying that the term ‘strain’ is important largely because we like to name things and then use those names when we share information, but that the concept of ‘strain’ may have no logical basis in nature where mutations and changes to a bacterial genome are constantly occurring events.
Let’s suppose I have a culture of Lactobacillus acidophilus growing in a test-tube, grown from a single colony. This clonal population is obviously a single strain that I will name strain Lb. acidophilus ISAPP2022. That was easy! I am aware of course that within this population there will almost certainly be a small number of individual cells with mutations (single nucleotide polymorphisms, or SNPs), cells that may have lost a plasmid, or cells that have undergone small genomic rearrangements. Nonetheless, because this genetic heterogeneity is unavoidable, I still consider this to be a pure strain. If I isolate an antibiotic resistant version of this strain by plating the strain on agar containing streptomycin and selecting a resistant colony I will now have an alternative clonal population all sharing a SNP (almost certainly in a gene encoding a ribosomal subunit). Even though there is a potentially very important genotypic and phenotypic difference I would not consider this to be a new strain, but rather it is a variant of Lb. acidophilus ISAPP2022. To help people in the lab or collaborators I might call this variant ISAPP2022SmR, or ISAPP2022-1. In my view, I could continue to make changes to ISAPP2022 and all of those individual clonal populations will still be variants of the original strain. So, the variant concept is that any change in the genome, no matter how small, creates a new variant. When I grow ISAPP2022 in my lab for many years, or share it with others around the word, it is my view that we are all working with the same strain, despite the fact that different variants will inevitably emerge over time and in different labs.
Where the strain concept becomes more difficult is when I isolate a bacterium from a novel source and I want to determine if it is the same strain as ISAPP2022. If the whole genome sequence (WGS) is a perfect match (100% average nucleotide identity or ANI) then both isolates are the same strain and both can be called ISAPP2022. If they have only a few SNPs then they are variants of the same strain. If the two isolates only share 95% ANI then they are obviously not the same strain and cannot even be considered as members of the same species (I am using a species ANI cut-off of 96% that I adopted from a recent paper in IJSEM.
Where it gets really tricky is when the ANI lies between 96% (so that we know that the isolates are both members of the same species) and 100% (where they are unequivocally the same strain). Where should we place the cut-off to define a strain? At what point is a threshold crossed and an isolate goes from being a variant to becoming a new strain? Should this be a mathematical decision based solely on ANI, or do we have to consider the functionality of the changes? If it is mathematical then we could simply choose a specific value, say 99.95% or 99.99% ANI, and declare anything below that value is a new strain. Remember that the 2Mb genome size of Lb. acidophilus would mean that two isolates sharing 99.99% ANI could differ by up to 200 SNPs. This could lead to a situation where an isolate with 199 SNPs compared to ISAPP2022 is considered a variant, but an isolate with 201 SNPs is a new strain (even though it only differs from the variant with 199 SNPs by two additional SNPs). This feels very unsatisfactory. But what about an isolate with only 50 SNPs, but one that has a very different phenotype to ISAPP2022 because the SNPs are located in important genes? Or what about an isolate with an additional plasmid, or missing a plasmid, or with a chromosomal deletion or insertion? I would argue we should not have a hard and fast cut-off based on SNPs alone, but we should continue to call all of these variants, and not define them as new strains.
So, by how much do two isolates have to differ before we no longer consider them as variants of one another, but as new strains? I will leave that question to taxonomists and philosophers since for me it falls into the territory of ‘how many angels can dance on the head of pin?’
All this may seem somewhat esoteric, but there are practical implications. Can we translate the findings from a clinical trial done with a specific variant of a strain to all other variants of the same strain? If Lactobacillus acidophilus ISAPP2022 has been shown to deliver a health benefit (and is therefore a probiotic), can we assume that Lb acidophilus ISAP2022-1 or any other variant will have the same effect? What if a variant has only one mutation, but that mutation eliminates an important phenotype required for the functionality of the original strain? I am afraid that at the end of all this verbiage I have simply rephrased the original debate topic from ‘All probiotic effects must be considered strain-specific’ to ‘All probiotic effects must be considered variant-specific’. Looks like we might be heading back to the debate stage in 2023!
Episode 6: Mechanisms of action for probiotics
/in Podcast, Season One /by KCPodcast: Play in new window | Download
Subscribe: Apple Podcasts | Spotify | RSS
The Science, Microbes & Health Podcast
This podcast covers emerging topics and challenges in the science of probiotics, prebiotics, synbiotics, postbiotics and fermented foods. This is the podcast of The International Scientific Association for Probiotics and Prebiotic (ISAPP), a nonprofit scientific organization dedicated to advancing the science of these fields.
Mechanisms of action for probiotics, with Prof. Sarah Lebeer
Episode summary:
In this episode, the ISAPP hosts speak with Prof. Sarah Lebeer of University of Antwerp, Belgium, to bring clarity to a commonly misunderstood topic: probiotic mechanisms of action. They discuss how probiotic mechanisms are often multi-factorial and difficult to unravel scientifically. Nevertheless, Prof. Lebeer describes five distinct mechanisms of action by which a probiotic may benefit a host.
See ISAPP’s other podcast episode on mechanisms of action, with Prof. Maria Marco: Why mechanistic research on probiotics is captivating and important.
Key topics from this episode:
Episode links:
The International Scientific Association for Probiotics and Prebiotics consensus statement on the scope and appropriate use of the term probiotic
Graphical summary of 5 main mechanisms of action for probiotics
(Image by Sarah Lebeer. Image copyright.)
Additional resources:
Current status of research on probiotic and prebiotic mechanisms of action. ISAPP blog
Importance of understanding probiotic mechanisms of action. ISAPP blog
About Prof. Sarah Lebeer:
Sarah Lebeer is a research professor at the Department of Bioscience Engineering of the University of Antwerp, Belgium. She has studied bioscience engineering, with a specialisation in cell and gene technology/food & health and obtained her Master at KU Leuven (Belgium). In 2008, she obtained a PhD degree with a topic on the mode of action of gastro-intestinal probiotics in inflammatory bowel diseases and a scholarship in the team of Prof. Jos Vanderleyden (KU Leuven). After a postdoc on the interaction between lactobacilli, viruses and mucosal immunology, in November 2011, she was offered a tenure track position at the University of Antwerp. Since then, she is leading the Laboratory for Applied Microbiology and Biotechnology of the ENdEMIC research group.
In 2020, she was awarded with an ERC Starting Grant that enables her to gain more in-depth knowledge of the evolutionary history and ecology of lactobacilli (https://www.lebeerlab.com). This rationale was also an important driving force to revise the Lactobacillus genus taxonomy with a large international consortium. Within the ERC project, Sarah has also launched the Isala citizen-science project to gain new insights in the role of vaginal lactobacilli for women’s health (https://isala.be). Since 2018, Sarah is an academic board member of the International Scientific Association on Probiotics and Prebiotics (www.isappscience.org). Communicating about beneficial microbes and probiotics for experts and laymen is an important inspiration for her daily work.
Episode 5: Prebiotics for animal health
/in Podcast, Season One /by KCPodcast: Play in new window | Download
Subscribe: Apple Podcasts | Spotify | RSS
The Science, Microbes & Health Podcast
This podcast covers emerging topics and challenges in the science of probiotics, prebiotics, synbiotics, postbiotics and fermented foods. This is the podcast of The International Scientific Association for Probiotics and Prebiotic (ISAPP), a nonprofit scientific organization dedicated to advancing the science of these fields.
Prebiotics for animal health, with Prof. George Fahey
Episode summary:
The hosts discuss prebiotics for animals with Prof. George Fahey, a prominent animal nutrition scientist who is currently Professor Emeritus at University of Illinois. Fahey explains how animal nutrition research relates to human nutrition research, and the changes in the field he has seen over the course of his long career. He describes the research on prebiotics for animal nutrition, covering both livestock and companion animals.
Key topics from this episode:
Episode links:
Expert consensus document: The International Scientific Association for Probiotics and Prebiotics (ISAPP) consensus statement on the definition and scope of prebiotics
The International Scientific Association for Probiotics and Prebiotics consensus statement on the scope and appropriate use of the term probiotic
Additional resources:
Are prebiotics good for dogs and cats? An animal gut health expert explains. ISAPP blog post
Using probiotics to support digestive health for dogs. ISAPP blog post
Prebiotics. ISAPP infographic
About Prof. George Fahey:
George C. Fahey, Jr. is Professor Emeritus of Animal Sciences and Nutritional Sciences at the University of Illinois at Urbana-Champaign. He served on the faculty since 1976 and held research, teaching, and administrative appointments. His research was in the area of carbohydrate nutrition of animals and humans. He published numerous books, book chapters, journal articles, and research abstracts.
He currently serves on two editorial boards, numerous GRAS expert panels, and is scientific advisor to both industry and governmental organizations. He retired from the University in 2010 but continues to serve on graduate student committees and departmental search committees. He owns Fahey Nutrition Consulting, Inc. that provides services to the human and pet food industries.
Bifidobacteria in the infant gut use human milk oligosaccharides: how does this lead to health benefits?
/in ISAPP Science Blog /by KCBy Martin Frederik Laursen, Technical University of Denmark, 2022 co-recipient of Glenn Gibson Early Career Research Prize
Breast milk is the ‘gold standard’ of infant nutrition, and recently scientists have zeroed in on human milk oligosaccharides (HMOs) as key components of human milk, which through specific interaction with bifidobacteria, may improve infant health. Clarifying mechanisms by which HMOs act in concert with bifidobacteria in the infant gut may lead to better nutritional products for infants.
Back in early 2016, I was in the middle of my PhD studies working on determinants of the infant gut microbiota composition in the Licht lab at the National Food Institute, Technical University of Denmark. I had been working with fecal samples from a Danish infant cohort study, called SKOT (Danish abbreviation for “Diet and well-being of young children”), investigating how the diet introduced in the complementary feeding period (as recorded by the researchers) influences the gut microbiota development 1,2. Around the same time, Henrik Munch Roager, PostDoc in the lab, was developing a liquid chromatography mass spectrometry (LC-MS)-based method for quantifying the aromatic amino acids (AAA) and their bacterially produced metabolites in fecal samples (the 3 AAAs and 16 derivatives thereof). These bacterially produced AAA metabolites were starting to receive attention because of their role in microbiota-host cross-talk and interaction with various receptors such as the Aryl Hydrocarbon Receptor (AhR) expressed in immune cells and important for controlling immune responses at mucosal surfaces 3,4. However, virtually nothing was known about bacterial metabolism of the AAAs in the gut in an early life context. Further, the fecal samples collected from the SKOT cohort were obtained in a period of life when infants are experiencing rapid dietary changes (e.g. cessation of breastfeeding and introduction of various new foods). Thus, we wondered whether the AAA metabolites would be affected by diet and whether these metabolites might contribute to the development of the infant’s immune system. Our initial results quickly guided us on the track of breastfeeding and bifidobacteria! Here is a summary of the story, published last year in Nature Microbiology5. (See the accompanying News & Views article here.)
We initially looked at the data from a subset of 59 infants, aged 9 months, from the SKOT cohort. Here we found that both the gut microbiome and the AAA metabolome were affected by breastfeeding status (breastfed versus weaned). It is well established that certain bifidobacteria dominate the bacterial gut community in breastfed infants due to their efficient utilization of HMOs – which are abundant components of human breastmilk 6. Our data showed the same, namely enrichment of Bifidobacterium in the breastfed infants, but also indicated that the abundance of specific AAA metabolites were dependent on breastfeeding.
Trying to connect the gut microbiome and AAA metabolome, we found striking correlations between the relative abundance of Bifidobacterium and specifically abundances of three aromatic amino acid catabolites – namely indolelactic acid (ILA), phenyllactic acid (PLA) and 4-hydroxyphenyllactic acid (4-OH-PLA), collectively aromatic lactic acids. These metabolites are formed in two enzymatic reactions (a transamination followed by a hydrogenation) of the aromatic amino acids tryptophan, phenylalanine and tyrosine. However, the genes involved in this pathway were not known for bifidobacteria. Digging deeper we discovered that not all Bifidobacterium species found in the infant’s gut correlated with these metabolites. This was only true for the Bifidobacterium species enriched in the breastfed infants (e.g. B. longum, B. bifidum and B. breve), but not post-weaning/adult type bifidobacteria such as B. adolescentis and B. catenulatum group.
We decided to go back to the lab and investigate these associations by culturing representative strains of the Bifidobacterium species found in the gut of these infants. Indeed, our results confirmed that Bifidobacterium species are able to produce aromatic lactic acids, and importantly that the ability to produce them was much stronger for the HMO-utilizing (e.g. B. longum, B. bifidum and B. breve) compared to the non-HMO utilizing bifidobacteria (e.g. B. adolescentis, B. animalis and B. catenulatum). Next, in a series of experiments we identified the genetic pathway in Bifidobacterium species responsible for production of the aromatic lactic acids and performed enzyme kinetic studies of the key enzyme, an aromatic lactate dehydrogenase (Aldh), catalyzing the last step of the conversion of aromatic amino acids into aromatic lactic acids. Thus, we were able to demonstrate the genetic and enzymatic basis for production of these metabolites in Bifidobacterium species.
To explore the temporal dynamics of Bifidobacteria and aromatic lactic acids and validate our findings in an early infancy context (a critical phase of immune system development), we recruited 25 infants (Copenhagen Infant Gut [CIG] cohort) from which we obtained feces from birth until six months of age. These data were instrumental for demonstrating the tight connection between specific Bifidobacterium species, HMO-utilization and production of aromatic lactic acids in the early infancy gut and further indicated that formula supplementation, pre-term delivery and antibiotics negatively influence the concentrations of these metabolites in early life.
Having established that HMO-utilizing Bifidobacterium species are key producers of aromatic lactic acids in the infant gut, we focused on the potential health implications of this. We were able to show that the capacity of early infancy feces to in vitro activate the AhR, depended on the abundance of aromatic lactic acid producing Bifidobacterium species and the concentrations of ILA (a known AhR agonist) in the fecal samples obtained from the CIG cohort. Further, using isolated human immune cells (ex vivo) we showed that ILA modulates cytokine responses in Th17 polarized cells – namely it increased IL-22 production in a dose and AhR-dependent manner. IL-22 is a cytokine important for protection of mucosal surfaces, e.g. it affects secretion of antimicrobial proteins, permeability and mucus production 7. Further, we tested ILA in LPS/INFγ induced monocytes (ex vivo), and found that ILA was able to decrease the production of the proinflammatory cytokine IL-12p70, in a manner dependent upon both AhR and the Hydroxycarboxylic Acid (HCA3) receptor, a receptor expressed in neutrophils, macrophages and monocytes and involved in mediation of anti-inflammatory processes 8,9. Overall, our data reveal potentially important ways in which bifidobacteria influence the infant’s developing immune system.
Figure 1 – HMO-utilizing Bifidobacterium species produce immuno-regulatory aromatic lactic acids in the infant gut.
Our study provided a novel link between HMO-utilizing Bifidobacterium species, production of aromatic lactic acids and immune-regulation in early life (Figure 1). This may explain previous observations that the relative abundance of bifidobacteria in the infant gut is inversely associated with development of asthma and allergic diseases 10–12 and our results, together with other recent findings13–15 are pointing towards aromatic lactic acids (especially ILA) as potentially important mediators of beneficial immune effects induced by HMO-utilizing Bifidobacterium species.
References
Episode 4: Weighing evidence for probiotic interventions: Perspectives of a primary care physician
/in Podcast, Season One /by KCPodcast: Play in new window | Download
Subscribe: Apple Podcasts | Spotify | RSS
The Science, Microbes & Health Podcast
This podcast covers emerging topics and challenges in the science of probiotics, prebiotics, synbiotics, postbiotics and fermented foods. This is the podcast of The International Scientific Association for Probiotics and Prebiotic (ISAPP), a nonprofit scientific organization dedicated to advancing the science of these fields.
Weighing evidence for probiotic interventions: Perspectives of a primary care physician, with Prof. Dan Merenstein, MD
Episode summary:
In this episode, the ISAPP host Prof. Dan Tancredi discusses evidence for probiotic interventions with Prof. Dan Merenstein, MD, a family medicine researcher based at Georgetown University. They discuss what it means to practice evidence-based medicine, and what kind of evidence clinicians should look for when deciding whether an intervention is appropriate. Prof. Merenstein shares how probiotic evidence has strengthened in the past few decades, and gives tips on what to look for in a probiotic intervention study.
Key topics from this episode:
Episode links:
CONSORT (Consolidated Standards of Reporting Trials) establishes a well-accepted, evidence-based, minimum set of recommendations for reporting randomized trials.
IPDMA (individual patient data meta-analysis) – see this Sung et al. paper on infantile colic and L. reuteri
BB12: Bifidobacterium animalis subsp. lactis BB-12, a well-studied probiotic. See this paper.
A priori: without prior knowledge
Additional resources:
The International Scientific Association for Probiotics and Prebiotics consensus statement on the scope and appropriate use of the term probiotic
About Prof. Dan Merenstein:
Dr. Daniel Merenstein is a Professor with tenure of Family Medicine at Georgetown University, where he also directs Family Medicine research. Dr. Merenstein has a secondary appointment in the undergraduate Department of Human Science, in the School of Nursing and Health Studies. Dr. Merenstein teaches two undergraduate classes, a research capstone and a seminar class on evaluating evidence based medical decisions. He has been funded by the NIH, USDA, Foundations and Industry, for grants over $100 million. Dr. Merenstein is the President of the board of directors of the International Scientific Association of Probiotics and Prebiotics.
The primary goal of Dr. Merenstein’s research is to provide answers to common clinical questions that lack evidence and improve patient care. Dr. Merenstein is a clinical trialist who has recruited over 2,100 participants for 10 probiotic trials since 2006. He is an expert on probiotics, antibiotic stewardship in outpatient settings and also conducts HIV research in a large women’s cohort. He sees patients in clinic one day a week.
Probiotics vs. prebiotics: Which to choose? And when?
/in Consumer Blog, ISAPP Science Blog /by KCBy Dr. Karen Scott, PhD, Rowett Institute, University of Aberdeen, Scotland
As consumers we are constantly bombarded with information on what we should eat to improve our health. Yet the information changes so fast that it sometimes seems that what was good for us last week should now be avoided at all costs!
Probiotics and prebiotics are not exempt from such confusing recommendations, and one area lacking clarity for many is which of them we should pick, and when. In this blog I will consider the relative merits of probiotics and prebiotics for the gut environment and health.
By definition, both probiotics and prebiotics should ‘confer a health benefit on the host’. Since an improvement in health can be either subjective (simply feeling better) or measurable (e.g. a lowering in blood pressure) it is clear that there is not a single way to define a ‘health benefit’. This was discussed nicely in a previous blog by Prof Colin Hill.
Although consumption of both probiotics and prebiotics should provide a health benefit, this does not mean that both need to act through the gut microbiota. Prebiotics definitively need to be selectively utilised by host microorganisms – they are food for our existing microbiota. However, depending on the site of action, this need not be the gut microbiota, and prebiotics targeting other microbial ecosystems in or on the body are being developed. Traditionally prebiotics have specifically been used to boost numbers of gut bacteria such as Bifidobacterium and the Lactobacilliaceae family, but new prebiotics targeting different members of the gut microbiota are also currently being researched.
Probiotics are live bacteria and despite a wealth of scientific evidence that specific probiotic bacterial strains confer specific health benefits, we often still do not know the exact mechanisms of action. This can make it difficult both to explain how or why they work, and to select new strains conferring similar health benefits. Many probiotics exert their effects within the gut environment, but they may or may not do this by interacting with the resident gut microbiota. For instance probiotics that reduce inflammation do so by interacting directly with cells in the mucosal immune system. Yet strains of lactobacilli (see here for what’s included in this group of bacteria) may do this by modulating cytokine production while Bifidobacterium strains induce tolerance acquisition. These very different mechanisms are one reason why mixtures containing several probiotic species or strains may in the end prove the most effective way to improve health. On the other hand, some probiotics do interact with the resident gut microbes: probiotics that act by inhibiting the growth of pathogenic bacteria clearly interact with other bacteria. Sometimes these may be potential disease-causing members of the resident microbiota, normally kept in check by other commensal microbes that themselves have become depleted due to some external impact, and some may be incoming pathogens. Such interactions can occur in the gut or elsewhere in the body.
This brings me back to the original question, and one I am frequently asked – should I take a probiotic or a prebiotic? The true and quick answer to this question is ‘it depends’! It depends why you are asking the question, and what you want to achieve. Let’s think about a few possible reasons for asking the question.
I want to improve the diversity of my microbiota. Should I take a prebiotic or a probiotic?
My first reaction was that there is an easy answer to this question – a prebiotic. Prebiotics are ‘food’ for your resident bacteria, so it follows that if you want to improve the diversity of your existing microbiota you should take a prebiotic. However, in reality this is too simplistic. Since prebiotics are selectively utilised by a few specific bacteria within the commensal microbiota to provide a health benefit, taking a prebiotic will boost the numbers of those specific bacteria. If the overall bacterial diversity is low, this may indeed improve the diversity. However, if the person asking the question already has a diverse microbiota, although taking one specific prebiotic may boost numbers of a specific bacterium, it may not change the overall diversity in a measurable way. In fact the best way to increase the overall diversity of your microbiota is to consume a diverse fibre-rich diet – in that way you are providing all sorts of different foods for the many different species of bacteria living in the gut, and this will increase the diversity of your microbiota. Of course, if you already consume a diverse fibre-rich diet your microbiota may already be very diverse, and any increased diversity may not be measurable.
I want to increase numbers of bifidobacteria in my microbiota. Should I take a prebiotic or a probiotic?
Again, I initially thought this was easy to answer – a prebiotic. There is a considerable amount of evidence that prebiotics based on fructo-oligosaccharides (FOS or inulin) boost numbers of bifidobacteria in the human gut. But this is only true as long as there are bifidobacteria present that can be targeted by consuming suitable prebiotics. Some scientific studies have shown that there are people who respond to prebiotic consumption and people who do not (categorised as responders and non-responders). This can be for two very different reasons. If an individual is devoid of all Bifidobacterium species completely, no amount of prebiotic will increase bifidobacteria numbers, so they would be a non-responder. In contrast if someone already has a large, diverse bifidobacteria population, a prebiotic may not make a meaningful impact on numbers – so they may also be a non-responder.
However, for those people who do not have any resident Bifidobacterium species, the only possible way to increase them would indeed be to consume a probiotic- specifically a probiotic containing one or several specific Bifidobacterium species. Consuming a suitable diet, or a prebiotic alongside the probiotic, may help retention of the consumed bifidobacteria, but this also depends on interactions with the host and resident microbiota.
I want to increase numbers of ‘specific bacterium x’ in my microbiota. Should I take a prebiotic or a probiotic?
The answer here overlaps with answer 2, and depends on the specific bacterium, and what products are available commercially, but the answer could be to take either, or a combination of both – i.e. a synbiotic.
If bacterium x is available as a probiotic, consuming that particular product could help. If bacterium x has been widely researched, and the specific compounds it uses for growth have been established, identifying and consuming products containing those compounds could boost numbers of bacterium x within the resident microbiota. Such research may already have identified combination products – synbiotics – that could also be available.
One caveat for the answers to questions 2 and 3 is that probiotics do not need to establish or alter the gut microbiota to have a beneficial effect on health. In fact, a healthy large intestine has a microbial population of around 1011-1012 bacterial cells per ml, or up to 1014 cells in total, while a standard pot of yogurt contains 1010 bacterial cells (108 cells/ml). Assuming every probiotic bacterial cell reaches the large intestine alive, they would be present in a ratio of 1: 10,000. This makes it difficult for them to find a specific niche to colonise, so consuming a probiotic may not “increase numbers of ‘specific bacterium x’ in my microbiota”, but this does not mean that the function of the probiotic within the gut ecosystem would not provide a health benefit. Many probiotics act without establishing in the microbiota.
I’ve been prescribed antibiotics. Should I take a prebiotic or a probiotic?
In this case the answer is clear cut – a probiotic.
There is a lot of evidence that consumption of probiotics can alleviate symptoms of, or reduce the duration of, antibiotic associated diarrhoea. From what we know about mechanisms of action, consumption of antibiotics kills many resident gut bacteria, reducing the overall bacterial population and providing an opportunity for harmful bacteria to become more dominant. Consuming certain probiotics can either help boost bacterial numbers in the large intestine, preventing the increased growth in pathogenic bacteria until the resident population recovers, or can increase production of short chain fatty acids, decreasing the colonic pH, preventing growth of harmful bacteria. Ideally probiotics would be taken alongside antibiotics, from day 1, to avoid the increase in numbers of the potentially harmful bacteria in the first place. This has been shown to be more effective. Consuming the probiotic alongside prebiotics that could help the resident microbiota recover more quickly may be even more effective. Even if you’ve already started the course of antibiotics, it’s not too late to start taking probiotics to reduce any side-effects. Always remember to complete taking the course of antibiotics as prescribed.
Putting all of this together to answer the initial question of whether it’s better to take probiotics or prebiotics, a better answer may in fact be take both to cover the different effects each has, maximising the benefit to health. There are specific times when probiotics are better, and other times when prebiotics are better, and consuming both together may make each more effective. In any case care has to be taken to consume a product that has been confirmed through robust studies to have the specific benefit that is required.
Episode 3: The science of fermented foods, part 2
/in Podcast, Season One /by KCPodcast: Play in new window | Download
Subscribe: Apple Podcasts | Spotify | RSS
The Science, Microbes & Health Podcast
This podcast covers emerging topics and challenges in the science of probiotics, prebiotics, synbiotics, postbiotics and fermented foods. This is the podcast of The International Scientific Association for Probiotics and Prebiotic (ISAPP), a nonprofit scientific organization dedicated to advancing the science of these fields.
The science of fermented foods, part 2, with Prof. Bob Hutkins
Episode summary:
Before listening to this episode, it’s recommended that you check out episode #1, The science of fermented foods, Part 1. In this episode, the hosts continue their discussion of fermented foods with Prof. Bob Hutkins, University of Nebraska – Lincoln. Prof. Hutkins elaborates on how the microbes associated with fermented foods may confer health benefits, as well as how food scientists choose strains for fermentation. He emphasizes how the live microbes in fermented foods differ from probiotics.
Key topics from this episode:
Episode links:
Microbiology and Technology of Fermented Foods, 2nd Ed., by Robert W. Hutkins
The International Scientific Association for Probiotics and Prebiotics (ISAPP) consensus statement on fermented foods
The International Scientific Association for Probiotics and Prebiotics consensus statement on the scope and appropriate use of the term probiotic
Gut-microbiota-targeted diets modulate human immune status, study by Stanford researchers
Additional resources:
Expert consensus document: The International Scientific Association for Probiotics and Prebiotics (ISAPP) consensus statement on the definition and scope of postbiotics.
Postbiotics. ISAPP infographic
Fermented foods. ISAPP infographic
What are fermented foods? ISAPP video
Do fermented foods contain probiotics? ISAPP blog post
How are probiotic foods and fermented foods different? ISAPP infographic
Are fermented foods probiotics? Webinar by Mary Ellen Sanders, PhD
About Prof. Bob Hutkins:
Bob Hutkins is the Khem Shahani Professor of Food Microbiology at the University of Nebraska. He received his Ph.D. from the University of Minnesota and was a postdoctoral fellow at Boston University School of Medicine. Prior to joining the University of Nebraska, he was a research scientist at Sanofi Bio Ingredients.
The Hutkins Lab studies bacteria important in human health and in fermented foods. His group is particularly interested in understanding factors affecting persistence and colonization of probiotic bacteria in the gastrointestinal tract and how prebiotics shift the intestinal microbiota and metabolic activities. The lab also conducts clinical studies using combinations of pro- and prebiotics (synbiotics) to enhance health outcomes. More recently we have developed metagenome-based models that can be used in personalized nutrition.
Professor Hutkins has published widely on probiotics, prebiotics, and fermented foods and is the author of the recently published 2nd edition of Microbiology and Technology of Fermented Foods.
Do polyphenols qualify as prebiotics? The latest scientific perspectives
/in ISAPP Science Blog, Consumer Blog /by KCKristina Campbell MSc, Science writer
When the ISAPP scientific consensus definition of ‘prebiotic’ was published in 2017, the co-authors on the paper included polyphenols as potential prebiotic substances. At the time, the available data on the effect of polyphenols on the gut microbiota were insufficient to show a true prebiotic effect.
An ISAPP webinar held in April 2022, aimed to give an update on the health effects of polyphenols and their mechanisms of action, along with how well polyphenols fit the prebiotic definition. Prof. Daniele Del Rio from University of Parma, Italy, and Prof. Yves Desjardins from Université Laval, Canada, presented the latest perspectives in the field.
What are polyphenols?
Polyphenols are a group of compounds found in plants, with over 6000 types identified to date. They can be divided into two main categories, flavonoids and non-flavonoids.
Polyphenols are absorbed in two different ways in the body. A very small fraction is absorbed in the small intestine, but 95% of them reach the lower gut and interact with gut microbiota. Although polyphenols have a special capacity to influence the activities of microorganisms, some resident microorganisms, in turn, can change the chemical structure of polyphenols through enzymatic action. These interactions produce a unique array of metabolites, which may be responsible for some of polyphenols’ prebiotic effects.
What are the health effects of polyphenols?
Epidemiological studies show that polyphenols in the diet are associated with many health benefits, including prevention of cardiovascular disease, certain cancers, and metabolic disease. These effects occur through various mechanisms. However, association is not proof of causation. So how good is the evidence that polyphenols can lead to health benefits?
Numerous human studies exist, but the most robust study to date for the health benefits of polyphenols is a randomized, controlled trial of over 20,000 adults, published in 2022, which showed supplementation with cocoa extract reduced death from cardiovascular events (although it did not reduce the number of cardiovascular events).
What are the mechanisms of action for polyphenols?
Polyphenols have multiple mechanisms of action. Del Rio focuses on the metabolites produced from dietary polyphenols called flavan-3-ols, which are found in red wine, grapes, tea, berries, chocolate and other foods. Along with colleagues, he showed that the metabolites produced in response to a polyphenol-rich food occur two ‘waves’: a small wave in the first 2 hours after ingestion, and a larger wave 5-35 hours after ingestion. The second wave is produced when flavan-3-ols reach the colon and interact with gut microbiota.
Work is ongoing to link these metabolites to specific health effects. Along these lines, Del Rio described a study showing how cranberry flavan-3-ol metabolites help defend against infectious Escherichia coli in a model system of bladder epithelial cells. These polyphenols are transformed by the gut microbiota into smaller compounds that are absorbed—so the health benefit comes not from the activity of polyphenols directly, but from the molecule(s) that the gut microbiota has produced from the polyphenols.
How else do polyphenols work? Ample evidence suggests polyphenols interact in different ways with gut microbes: they have direct antimicrobial effects, they affect quorum sensing, they compete with bacteria for some minerals, and/or they modify ecology, thereby affecting biofilm formation. Desjardins explained that these interactions may occur in parallel: for example, polyphenols may exert antimicrobial effects when they reach the colon, and at the same time, microorganisms in the gut begin to degrade them.
The mode of action of polyphenols Desjardins studies is the prebiotic mode of action—or as he describes it, “prebiotic with a twist”. A landmark paper from 2015 showed how cranberry polyphenols had protective effects on metabolism and obesity through the creation of mucin in the intestine, which formed a good niche for Akkermansia muciniphila, a keystone bacterial species for good metabolic health. Other polyphenols have since been shown to work the same way: by stimulating production of mucin, thereby providing ideal conditions for beneficial bacteria to grow. In this way, polyphenols appear to show small-scale effects comparable to the effects of probiotics, by inducing a host response that alters the bacterial niche.
Are the effects of polyphenols individual?
Del Rio offered some evidence that the health effects of polyphenols, via metabolites, is personalized: a study showed the existence of three distinct patterns of metabolite production in response to dietary polyphenols (ellagitannins). These may depend on the particular microbes of the gut and their ability to produce the relevant metabolites—so in essence, in each case the gut microbiota is equipped to produce a certain set of metabolites in response to polyphenols. More work is needed, however, to be able to personalize polyphenol intake.
Do polyphenols qualify as prebiotic substances?
Polyphenols clearly interact with gut microbiota to influence human health. The definition of a prebiotic is “a substrate that is selectively utilized by host microorganisms conferring a health benefit”. Given the available evidence that polyphenols are not metabolized or utilized by bacteria in all cases in the same direct way as carbohydrate prebiotics, Desjardins sees them as having a “prebiotic-like effect”. Rather, polyphenols are transformed into other biologically active molecules that ultimately provide health benefits to the host. These prebiotic-like properties of polyphenols are nicely summarized in a 2021 review paper and include decreasing inflammation, increasing bacteriocins and defensins, increasing gut barrier function (thereby reducing low-grade inflammation), modulating bile acids, and increasing gut immuno-globulins.
Overall, the speakers showed that polyphenols exert their health effects in several ways—and while the gut microbiota are important for their health effects, polyphenols, as a heterogenous group, may not strictly meet the criteria for prebiotics. Clearly, more research on polyphenols may reveal other mechanisms by which these important nutrients influence the gut microbiome and contribute to host health, and they may someday be regarded as prebiotics.
Episode 2: Why mechanistic research on probiotics is captivating and important
/in Podcast, Season One /by KCPodcast: Play in new window | Download
Subscribe: Apple Podcasts | Spotify | RSS
The Science, Microbes & Health Podcast
This podcast covers emerging topics and challenges in the science of probiotics, prebiotics, synbiotics, postbiotics and fermented foods. This is the podcast of The International Scientific Association for Probiotics and Prebiotic (ISAPP), a nonprofit scientific organization dedicated to advancing the science of these fields.
Why mechanistic research on probiotics is captivating and important, with Prof. Maria Marco
Episode summary:
In this episode, the ISAPP hosts discuss probiotic mechanisms of action with Prof. Maria Marco, University of California, Davis. Prof. Marco is a well-known probiotic researcher with special expertise in food-associated lactobacilli. Here she explains how studying probiotics in food science can lead to fundamental insights in biology. She shares why it’s important to understand probiotic mechanisms of action, and describes how scientists go about identifying which compounds or pathways are important for probiotic health effects.
Key topics from this episode:
Episode links:
The International Scientific Association for Probiotics and Prebiotics consensus statement on the scope and appropriate use of the term probiotic
Prof. Marco refers to two of her mentors, Willem De Vos and Michiel Kleerebezem
See this overview of Koch’s postulates
Additional resources:
Bacterial genes lead researchers to discover a new way that lactic acid bacteria can make energy and thrive in their environments, ISAPP blog post featuring recent work from Prof. Marco’s lab
About Prof. Maria Marco:
Maria Marco is a Professor in the Department of Food Science and Technology and Chair of the Food Science Graduate Group at the University of California, Davis. She received her PhD in microbiology from the University of California, Berkeley and then was a postdoc and project leader at NIZO Food Research, The Netherlands. Dr. Marco has 20 years’ experience investigating fermented foods, probiotics, and diet-dependent, host-microbe interactions in digestive tract. Her laboratory at UC Davis is broadly engaged in the study of food and intestinal microbiomes and the ecology and genetics of lactic acid bacteria.
Episode 1: The science of fermented foods, part 1
/in Podcast, Season One /by KCPodcast: Play in new window | Download
Subscribe: Apple Podcasts | Spotify | RSS
The Science, Microbes & Health Podcast
This podcast covers emerging topics and challenges in the science of probiotics, prebiotics, synbiotics, postbiotics and fermented foods. This is the podcast of The International Scientific Association for Probiotics and Prebiotic (ISAPP), a nonprofit scientific organization dedicated to advancing the science of these fields.
The science of fermented foods, part 1, with Prof. Bob Hutkins
Episode summary:
The hosts discuss fermented foods with Prof. Bob Hutkins, University of Nebraska – Lincoln. Prof. Hutkins wrote a popular textbook on fermented foods and has had a 40-year career in fermentation science. He shares why he ended up in fermentation science, as well as how fermented foods are made and how important live microbes are for their health benefits.
Key topics from this episode:
Episode links:
Microbiology and Technology of Fermented Foods, 2nd Ed., by Robert W. Hutkins
The International Scientific Association for Probiotics and Prebiotics (ISAPP) consensus statement on fermented foods
The International Scientific Association for Probiotics and Prebiotics consensus statement on the scope and appropriate use of the term probiotic
Synbiotics: Definitions, Characterization, and Assessment – ISAPP webinar featuring Prof. Bob Hutkins and Prof. Kelly Swanson
Additional resources:
Fermented foods. ISAPP infographic
What are fermented foods? ISAPP video
Do fermented foods contain probiotics? ISAPP blog post
How are probiotic foods and fermented foods different? ISAPP infographic
Are fermented foods probiotics? Webinar by Mary Ellen Sanders, PhD
About Prof. Bob Hutkins:
Bob Hutkins is the Khem Shahani Professor of Food Microbiology at the University of Nebraska. He received his Ph.D. from the University of Minnesota and was a postdoctoral fellow at Boston University School of Medicine. Prior to joining the University of Nebraska, he was a research scientist at Sanofi Bio Ingredients.
The Hutkins Lab studies bacteria important in human health and in fermented foods. His group is particularly interested in understanding factors affecting persistence and colonization of probiotic bacteria in the gastrointestinal tract and how prebiotics shift the intestinal microbiota and metabolic activities. The lab also conducts clinical studies using combinations of pro- and prebiotics (synbiotics) to enhance health outcomes. More recently we have developed metagenome-based models that can be used in personalized nutrition.
Professor Hutkins has published widely on probiotics, prebiotics, and fermented foods and is the author of the recently published 2nd edition of Microbiology and Technology of Fermented Foods.
The gut mycobiome and misinformation about Candida
/in ISAPP Science Blog /by KCBy Prof. Eamonn Quigley, MD, The Methodist Hospital and Weill Cornell School of Medicine, Houston
As a gastroenterologist, I frequently meet with patients who are adamant that a Candida infection is the cause of their ailments. Patients experiencing a range of symptoms, including digestive problems, sometimes believe they have an overgrowth of Candida in their gastrointestinal (GI) tract and want to know what to do about it. Their insistence is perhaps not surprising, given how many many websites and social media ‘gurus’ share lists of symptoms supposedly tied to Candida infections. Even cookbooks exist with recipes specifically tailored to “cure” someone of Candida infection through dietary changes. Some articles aim to counter the hype – for example, an article titled “Is gut Candida overgrowth actually real, and do Candida diets work?” Yet patients are too often confused about the evidence on Candida and other fungi in the GI tract. In a 2021 ISAPP presentation on the gut mycobiome, I provided a clinical perspective on fungal infections and the related evidence base.
Fungal infections do occur
Much of the misinformation I encounter on Candida infections focuses on selling a story that encourages people to blame Candida overgrowth as the cause of their symptoms and undertake expensive or complicated dietary and supplement regimens to “cure” the infection. This is not to say that fungal infections do not take place in the body. Fungal infections, from Candida or other fungi, frequently occur on the nails or skin. Patients taking oral or inhaled steroids may develop Candida infections in the oropharynx and esophagus. Immunocompromised patients also face a greater risk of Candidiasis and Candidemia—these include HIV patients; patients undergoing chemotherapy; transplant patients; and patients suffering from malnutrition.
Fungal infections are rare in the GI tract
Regardless, instances of documented Candida infection in the GI tract remain few in number. One study published in the 90s reported 10 patients hospitalized with severe diarrhea1. These patients suffered from chronic illness, underwent intense antimicrobial treatment or chemotherapy, and faced severe outcomes such as dehydration—and clinicians consistently identified the growth of Candida albicans in the patient fecal samples. Other studies on the matter lack the clinical evidence to conclude that fungal infections drive GI disease. A study examining small intestinal fungal overgrowth identified instances of fungal overgrowth among 150 patients with unexplained symptoms2. However, the lack of documentation of response to an antifungal treatment protocol makes it difficult to attribute the observed symptoms to the presence of fungal organisms.
The gut mycobiome in IBS
The gut microbiome has taken centre stage in common discourse about gut health. In line with this movement, my colleagues at Cork investigated the fungal members of the gut microbiome – that is, the gut mycobiome – in the guts of patients diagnosed with irritable bowel syndrome (IBS)3 to ascertain whether there was any correlation with symptoms. This effort revealed DNA sequences belonging to many fungal species. However, no significant differences in the number of fungal species were observed between IBS patients and volunteers. A smaller study done on a Dutch cohort, on the other hand, detected significantly reduced total fungal diversity among IBS patients4. So, it’s not yet clear whether mycobiome differences exist across populations with IBS.
Studying the gut mycobiome for further insights
The few studies that have examined the human gut mycobiome expose the need to answer basic questions about the fungal components of the gut microbiome. For instance, what is the gut mycobiome composition among people not suffering from GI-related symptoms? Efforts to answer these questions would require longitudinal sample collection to account for the high turnover of microbes in the GI tract. We would also need to perform stool measures not typically performed in the clinic to better correlate fungal overgrowth with GI-related symptoms. Overall, any gut mycobiome study requires careful and detailed experimental design.
We also have to consider where the gut mycobiome originates. A recent study in mSphere showed that the increased amount of DNA belonging to S. cerevisiae in stool samples coincided with the number of times subjects consumed bread and other fungi-rich foods5. S. cerevisiae also failed to grow in lab conditions mimicking the gut environment after 7 days of incubation. These findings suggest that the fungi identified in gut mycobiome profiles are not persistent gut colonizers, but transient members of the gut microbiome that come from the food we digest or our saliva.
A survey of the literature on the gut mycobiome and fungal infections in the GI tract highlights the need to conduct more studies on the role fungi play in gut and overall health. My clinical approach when I encounter someone claiming to have GI symptoms caused by Candida infection is a skeptical, yet empathetic response. Through proper communication of the evidence, we can investigate the origin of symptoms together and identify the best treatment methods for any GI-related disease, whether caused by fungal infections or not.
References
(1) Gupta, T. P.; Ehrinpreis, M. N. Candida-Associated Diarrhea in Hospitalized Patients. Gastroenterology 1990, 98 (3), 780–785. https://doi.org/10.1016/0016-5085(90)90303-i.
(2) Jacobs, C.; Coss Adame, E.; Attaluri, A.; Valestin, J.; Rao, S. S. C. Dysmotility and Proton Pump Inhibitor Use Are Independent Risk Factors for Small Intestinal Bacterial and/or Fungal Overgrowth. Aliment Pharmacol Ther 2013, 37 (11), 1103–1111. https://doi.org/10.1111/apt.12304.
(3) Das, A.; O’Herlihy, E.; Shanahan, F.; O’Toole, P. W.; Jeffery, I. B. The Fecal Mycobiome in Patients with Irritable Bowel Syndrome. Sci Rep 2021, 11 (1), 124. https://doi.org/10.1038/s41598-020-79478-6.
(4) Botschuijver, S.; Roeselers, G.; Levin, E.; Jonkers, D. M.; Welting, O.; Heinsbroek, S. E. M.; de Weerd, H. H.; Boekhout, T.; Fornai, M.; Masclee, A. A.; Schuren, F. H. J.; de Jonge, W. J.; Seppen, J.; van den Wijngaard, R. M. Intestinal Fungal Dysbiosis Is Associated With Visceral Hypersensitivity in Patients With Irritable Bowel Syndrome and Rats. Gastroenterology 2017, 153 (4), 1026–1039. https://doi.org/10.1053/j.gastro.2017.06.004.
(5) Auchtung, T. A.; Fofanova, T. Y.; Stewart, C. J.; Nash, A. K.; Wong, M. C.; Gesell, J. R.; Auchtung, J. M.; Ajami, N. J.; Petrosino, J. F. Investigating Colonization of the Healthy Adult Gastrointestinal Tract by Fungi. mSphere 2018, 3 (2), e00092-18. https://doi.org/10.1128/mSphere.00092-18.
New ISAPP Webinar: Fermented Foods and Health — Continuing Education Credit Available for Dietitians
/in News /by KCDietitians – along with many other nutritional professionals – often receive questions about consuming fermented foods for digestive health. But how strong is the evidence that fermented foods can improve digestive health?
ISAPP is pleased to work with Today’s Dietitian to offer a free webinar in which Hannah Holscher, PhD, RD, and Jennifer Burton, MS, RD, LDN will discuss the foundational elements of fermented foods, the role of microbes in fermentation, how they differ from probiotics and prebiotics, and how to incorporate fermented foods into client diets in an evidence-based manner. Participants will come away with a grasp of the scientific evidence that supports fermented food consumption. This activity is accredited by the Academy of Nutrition and Dietetics Commission on Dietetic Registration (CDR) for 1.0 CPEUs for dietitians.
The one-hour virtual event, titled “Fermented Foods and Health — Does Today’s Science Support Yesterday’s Tradition?”, was held April 20th, 2022, at 2:00 pm Eastern Time.
ISAPP and Today’s Dietitian also collaborated on a self-study activity titled “Evidence-based use of probiotics, prebiotics and fermented foods for digestive health”. This free activity, which provides more detail on the topic that the 1-hour webinar above, was approved by CDR to offer 2.0 CPEUs for dietitians and is available here through November 2023.
Improving the quality of microbiome studies – STORMS
/in ISAPP Science Blog /by KCBy Mary Ellen Sanders, PhD, ISAPP Executive Science Officer
In mid-March I attended the Gut Microbiota for Health annual meeting. I was fortunate to participate in a short workshop chaired by Dr. Geoff Preidis MD, PhD, a pediatric gastroenterologist from Baylor College of Medicine and Dr. Brendan Kelly MD, MSCE, an infectious disease physician and clinical epidemiologist from University of Pennsylvania. The topic of this workshop was “Designing microbiome trials – unique considerations.”
Dr. Preidis introduced the topic by recounting his effort (Preidis et al. 2020) to review evidence for probiotics for GI endpoints, including for his special interest, necrotizing enterocolitis (NEC). After a thorough review of available studies testing the ability of probiotics to prevent morbidity and mortality outcomes for premature neonates, he and the team found 63 randomized controlled trials that assessed close to 16,000 premature babies. Although the effect size for the different clinical endpoints was impressive and clinically meaningful, AGA was only able to give a conditional recommendation for probiotic use in this population.
Why? In part, because inadequate conduct or reporting of these studies led to reduced confidence in their conclusions. For example, proper approaches to mitigate selection bias must be reported. Some examples of selection bias include survival bias (where part of the target study population is more likely to die before they can be studied), convenience sampling (where members of the target study population are not selected at random), and loss to follow-up (when probability of dropping out is related to one of the factors being studied). These are important considerations that might influence microbiome results. If the publication on the trial does not clearly indicate how these potential biases were addressed, then the study cannot be judged as low risk of bias. It’s possible in such a study that bias is addressed correctly but reported incompletely. But the reader cannot ascertain this.
With an eye toward improving the quality and transparency of future studies that include microbiome endpoints, Dr. Preidis shared a paper by a multidisciplinary team of bioinformaticians, epidemiologists, biostatisticians, and microbiologists titled Strengthening The Organization and Reporting of Microbiome Studies (STORMS): A Reporting Checklist for Human Microbiome Research.
Dr. Preidis kindly agree to help the ISAPP community by answering a few questions about STORMS:
Dr. Preidis, why is the STORMS approach so important?
Before STORMS, we lacked consistent recommendations for how methods and results of human microbiome research should be reported. Part of the problem was the complex, multi-disciplinary nature of these studies (e.g., epidemiology, microbiology, genomics, bioinformatics). Inconsistent reporting negatively impacts the field because it renders studies difficult to replicate or compare to similar studies. STORMS is an important step toward gaining more useful information from human microbiome research.
One very practical outcome of this paper is a STORMS checklist, which is intended to help authors provide a complete and concise description of their study. How can we get journal editors and reviewers to request this checklist be submitted along with manuscripts for publication?
We can reach out to colleagues who serve on editorial boards to initiate discussions among the editors regarding how the STORMS checklist might benefit reviewers and readers of a specific journal.
How does this checklist differ from or augment the well-known CONSORT checklist?
Whereas the CONSORT checklist presents an evidence-based, minimum set of recommendations for reporting randomized trials, the STORMS checklist facilitates the reporting of a comprehensive array of observational and experimental study designs including cross-sectional, case-control, cohort studies, and randomized controlled trials. In addition to standard elements of study design, the STORMS checklist also addresses critical components that are unique to microbiome studies. These include details on the collection, handling, and preservation of specimens; laboratory efforts to mitigate batch effects; bioinformatics processing; handling of sparse, unusually distributed multi-dimensional data; and reporting of results containing very large numbers of microbial features.
How will papers reported using STORMS facilitate subsequent meta-analyses?
When included as a supplemental table to a manuscript, the STORMS checklist will facilitate comparative analysis of published results by ensuring that all key elements are reported completely and organized in a way that makes the work of systematic reviewers more efficient and more accurate.
I have been struck through the years of reading microbiome research that primary and secondary outcomes seem to be rarely stated up front. Or if such trials are registered, for example on clinicaltrials.gov, the paper does not necessarily focus on the pre-stated primary objectives. This approach runs the risk of researchers finding the one positive story to tell out of the plethora of data generated in microbiome studies. Will STORMS help researchers design more hypothesis driven studies?
Not necessarily. The STORMS checklist was not created to assess study or methodological rigor; rather, it aims to aid authors’ organization and ease the process of reviewer and reader assessment of how studies are conducted and analyzed. However, if investigators use this checklist in the planning phases of a study in conjunction with sound principles of study design, I believe it can help improve the quality of human microbiome studies – not just the writing and reporting of results.
Do you have any additional comments?
One of the strengths of the STORMS checklist is that it was developed by a multi-disciplinary team representing a consensus across a broad cross-section of the microbiome research community. Importantly, it remains a work in progress, with planned updates that will address evolving standards and technological processes. Anyone interested in joining the STORMS Consortium should visit the consortium website (www.stormsmicrobiome.org).
ISAPP’s Guiding Principles for the Definitions of ‘Biotics’
/in ISAPP Science Blog /by KCBy Mary Ellen Sanders, PhD, ISAPP Executive Science Officer
Articulating a definition for a scientific concept is a significant challenge. Inevitably, scientists have different perspectives on what falls inside and outside the bounds of a term. Prof. Glenn Gibson, ISAPP co-founder and longtime board member, recently published a paper that describes his path to coining the word ‘prebiotic’, with this observation: “One thing I have learned about definitions is that if you propose one, then be ready for it to be changed, dismissed or ignored!”
Mary Ellen Sanders with Glenn Gibson
Members of the ISAPP board, however, have remained steadfast in their belief that such definitions are worth creating. They are the basis for shared understanding and coordinated progress across a scientific field.
Developing the consensus definition papers on probiotics, prebiotics, synbiotics, postbiotics and fermented foods was demanding on the part of all involved. The objective of the panels that met to discuss these definitions was clear – to provide common ground for consistent use of this growing body of terms for all stakeholders. Although some disagreement among the broader scientific community exists about some of the definitions, ISAPP’s approach relied on important, underlying principles:
In my opinion, many published definitions, including previous ones for postbiotics (see supplementary table here), are untenable because they don’t recognize these principles. There may also be a tendency to rely on historical use of terms, rather than to describe what is justified by current scientific knowledge. A good example of this is provided by the first definition of probiotics, published in 1965. It was “substances secreted by one microorganism that stimulate another microorganism” (Lily and Stillwell, 1965), which is far from the current definition of “live microorganisms that, when administered in adequate amounts, confer a health benefit on the host (Hill et al. 2014).
If you’re looking for a concise summary of the five published ISAPP definitions, see here for our definitions infographic.
Additional reflections: I noted with a smile Glenn’s views on ISAPP, specifically on the appropriate pronunciation of the abbreviation ‘ISAPP’. “My only negative is that everyone involved in the organisation aside from 2 or 3 of us pronounce its acronym wrongly.” Most board members, including myself, have always pronounced this as ‘eye-sap’. Glenn opines, “The abbreviation is not eye-SAPP, it is ISAPP (with the ‘I’ – remarkably enough – being spoken as it is in the word ‘International’).” I wonder how he pronounces IBM?
Domestic horses from different geographical locations harbor antibiotic resistant gut bacteria, unlike their wild counterparts
/in ISAPP Science Blog /by KCBy Dr. Gabriel Vinderola, PhD, Associate Professor of Microbiology at the Faculty of Chemical Engineering from the National University of Litoral and Principal Researcher from CONICET at Dairy Products Institute (CONICET-UNL), Santa Fe, Argentina
It all started on the 12-hour ferry trip that links Turku with Stockholm during one of the last still warm- summer days of September 2016, when a group of scientists met: Seppo Salminen, Miguel Gueimonde, Carlos Gómez- Galllego and Akihito Endo (joining us virtually from Japan well before the pandemic made these virtual meetings so popular). One of the topics was the possibility of conducting a study comparing the gut microbiome of feral and domestic horses. We had no specific funding for the project but we agreed it would be worthwhile and all agreed to participate.
Misaki wild horses from Cape Toi’s Reserve, Japan. Photo courtesy of Seppo Salminen.
Domesticated horses live under different conditions compared with their wild ancestors. We hypothesized that the animals’ housing, regular veterinary care and feeds would lead to an altered microbiota compared to wild horses. The project was ambitious and challenging in several ways: we aimed at sequencing all microbes, not just bacteria, by using whole genome sequencing; sampling droppings from feral horses needed special permission from the parks or reserves where these horses were held; the project required shipping samples from different parts of the world to the same place where they would be processed; and this all had to be managed without specific financial support to cover the expenses. Curiosity, personal dedication and funding from each end fueled this project.
Little by little, samples of feces of feral and domestic horses were collected in Argentina, Finland, Spain, Russia and Japan. Fecal DNA was extracted in every sampling location and sent to Prof. Li Ang in China for whole genome sequencing and data analysis. A remarkable contribution was made by Prof. Ang and his team from the Zhengzhou University in China. In his words:
Cimarron wild horses from the State Park Ernesto Tornquist, Argentina. Photo courtesy of Seppo Salminen.
The fecal microbiome of 57 domestic and feral horses from five different locations on three continents were analyzed, observing geographical differences. A higher abundance of eukaryota (p < 0.05) and viruses (p < 0.05) and lower abundance of archaea (p < 0.05) were found in feral animals when compared with domestic ones. The abundance of genes coding for microbe-produced enzymes involved in the metabolism of carbohydrates was significantly higher (p < 0.05) in feral animals regardless of the geographic origin, which may reflect the fact that feral horses are exposed to a much more diverse natural vegetal diet than their domesticated counterparts. Differences in the fecal resistomes between both groups of animals were also observed. The domestic/captive horse microbiomes were enriched in genes conferring resistance to tetracycline, likely reflecting the use of this antibiotic in the management of these animals. Our data also showed an impoverishment of the fecal microbiome in domestic horses with diet, antibiotic exposure and hygiene being likely drivers, a fact that has been also reported for us, humans.
Almost 6 years passed since the results of those ideas discussed on board a ferry slowly galloped into the cover of the February edition of Nature Communications Biology. We hope this will be a starting point for more work that can help uncover the best ways to support equine health.
Mini-tutorial on statistical analysis: Correcting a common misinterpretation of p-values
/in ISAPP Science Blog /by KCDaniel Tancredi PhD, Professor of Pediatrics, UC Davis School of Medicine and Center for Healthcare Policy and Research, Sacramento, CA.
Decision makers frequently rely on p-values to decide whether and how to use a study to inform their decisions. Many misinterpret what a low p-value actually means, however. I will attempt to correct this common misinterpretation and explain how to use small p-values to evaluate whether a null hypothesis is plausible. I will show that a low p-value should be used in the same way that a clinician should use a valuable but imperfect clinical diagnostic test result; as one factor, but not the only one, on which to base a decision.
Typically, readers assume that if a p value is low, such as less than 0.05, and thus the test statistic is statistically significant at the conventional level, that there is a good chance that the study results are not “due to chance”. But it is not as simple as that. Let’s suppose that one has a p-value that was generated in a well-designed placebo-controlled randomized clinical trial. We will assume that the trial had a sample size that would provide 80% power to detect what the investigators considered to be the minimum clinical significant difference. The null hypothesis in a trial like this would predict that there is no difference between the placebo and intervention arms for the primary outcome. Once the data were analyzed and reported properly, the p-value was estimated to be just under 0.05. Does this p value less than 0.05 mean that the null hypothesis has no more than a 5% probability of being true? Does it even ensure that the null hypothesis is unlikely? If the calculated p-value was greater than 0.05 (let’s say p=0.10) would that be enough to ensure that the probability of the null hypothesis being true is definitely greater than 5%?
The answer to each of these questions is “no”! This is where, unfortunately, it begins to get (a little) complicated for many users of statistics. The p-value is calculated under the assumption that the null hypothesis is true, and so it does not and cannot measure the probability of that assumption being correct.1 Even though it is common to interpret a p-value as though it is an objective and sufficient statement about the probability that the null hypothesis is true, that is not the case. Statisticians have been trying to communicate this nuance for decades, including the issuance of a statement in 2016 on p-values by the American Statistical Association, a rare statement on statistical practice in that prominent organization’s long history1.
Using a p-value to calculate the probability that the null hypothesis is true
If one wants to use a p-value as one factor in a procedure that can produce a statement about the probability that the null hypothesis is true, one needs to supply an additional input that can be very difficult to obtain. This is a prior (or pre-study) probability, a quantitative estimate of the probability that the null hypothesis is true. This is based on a considered judgment of the state of existing knowledge, what is already known (outside the study results) about how and by how much the intervention may affect the outcomes being assessed.2 Of course, such a judgment can vary a great deal from one individual to the next, according to his/her ability to gather and appraise that knowledge. These judgments can also be influenced by other interests, including financial and ideological, how much of a stake one has in each of the various competing scientific explanations. Depending on the prior (or pre-experiment) probability for the null hypothesis, a p-value of 0.01 may not be enough to ensure that the null hypothesis has less than a 50% posterior (or after-experiment) probability of being true, whereas a p-value of 0.10 may be enough to make the posterior probability of the null hypothesis be comfortably under 5% (for those interested in taking this discussion further, the final section of this post illustrates this in more detail and with examples).
Determining whether an intervention works
Fundamentally, p-values cannot be used by themselves as though they are objective and reliable ways to make prudent decisions about whether an intervention works. Statisticians emphasize the necessity for the results of individual studies to be interpreted in a broader context, one that involves both statistical judgment and judgment on the underlying scientific plausibility of the hypothesized effects. It is well known that when sample sizes are very large, such as in many observational studies involving tens or even hundreds of thousands of observations, p-values can be very low, even for effect sizes whose confidence intervals are relatively narrow yet do not include any effects that would be of practical importance. In evidence-based medicine we typically face the opposite challenge, where small sample sizes and/or relatively infrequent outcome events result in p-values greater than 0.05 and 95% confidence intervals that are ambiguous because they include the null value (as is implied by p>0.05), but with outcomes that would be very important clinically. Thankfully, in my own career, it does seem to me to have become better appreciated that simply describing studies as positive or negative depending on which side of 0.05 the p-value falls is an unreliable method for evaluating evidence.
p-values and meta-analysis
Another thing to keep in mind is that even when a majority of individual studies that address the same research question may have p-values above 0.05, the meta-analysis of those study results can still indicate a statistically and clinically significant effect. As an example I will use a 2017 Cochrane review of the use of probiotics for the prevention of Clostridioides difficile‐associated diarrhea (CDAD) in adults and children.3 The overwhelming majority of studies, 17 of 21, were supposedly “negative” in that they have confidence intervals that include the null value, but the overall pooled estimate reports a statistically significant and clinically important range of effects. Also note that the overwhelming majority of the studies report confidence intervals that are consistent with the confidence interval for the overall pooled estimate, when one considers the degree of overlap. See Figure 1 below.
Figure 1. Forest plot summarizing complete-case analyses from systematically reviewed clinical trials of probiotics for the prevention of Clostridium difficile‐associated diarrhea (CDAD) in adults and children. Although only 4 of the 31 individual trials had statistically significant results, the pooled estimate shows a statistically and clinically signficant reduction in risk of CDAD for the studied probiotics, without statistically significant heterogeneity among the individual trials’ relative risk estimates. Note that the confidence inferval for the pooled estimate is entirely contained by all but two of the confidence intervals from the individual trials and that even the confidence intervals from these two exceptions largely contain the pooled estimate.
Reprint of Figure 3 from Joshua Z Goldenberg, Christina Yap, Lyubov Lytvyn, et al’s “ Probiotics for the prevention of Clostridium difficile‐associated diarrhea in adults and children”, published December 12, 2017 in “Cochrane Database of Systematic Reviews” by John Wiley and Sons. Copyright by John Wiley and Sons. Reprinted under one-time use license from John Wiley.
Summary
In conclusion, p values are an important component of determining whether an outcome can be deemed to be statistically significant, but this depends on the question under investigation, and is only one part of a more complete analysis. When appraising evidence for whether an intervention works, it is important to keep in mind that if one relies only on statistical inferences from individual studies, one is vulnerable to making unreliable assessments that substantially misstate the plausibility that an intervention does (or does not) have an effect. Statistical analysis cannot replace context-specific scientific judgment; both are needed to make reliable evidence appraisals.
A deeper dive into how to use p-values to assess the probability that the null hypothesis is true
A common misinterpretation of p-values is that they measure the probability that the null hypothesis is true, given the sample data. As stated above, the p-value, by itself, cannot speak to this probability, but if one is willing to supply a judgment on the prior probability that the null hypothesis is true, one can use that and the p-value to get a lower bound on the probability of interest. The compelling figure that accompanies Regina Nuzzo’s terrific Nature article on p-values and their shortcomings nicely illustrates such results for six combinations involving three example prior probabilities and p-values of 0.05 and 0.01.4 Table 1 shows posterior probabilities for those and other input combinations.
The calculations used in that figure and in Table 1 for converting the two inputs, a prior probability for the null hypothesis and a p-value, into a posterior probability for the null hypothesis is simply an application of a much more general formula, one that has been known for over 200 years. This formula is simple to state and remember when expressed as odds. According to Bayes Theorem, Posteriors Odds equals Prior Odds multiplied by a term we call the Likelihood Ratio. The likelihood ratio is a ratio of two conditional probabilities for the observed data, with each computed under differing hypotheses.5 [Another very widely-used application of this general formula is when physicians use the results and the operational characteristics (e.g. the sensitivity and specificity) of clinical tests to inform medical diagnoses.6] The formula uses odds not in the way that they are defined in horse racing where long-shots have high odds, but in the way that statisticians define it, as the ratio of the probability of an event to the probability of the absence of that event. To a statistician, high odds mean high probability for the event. When the probability of an event P is greater than 0, the odds are P / (1 – P). For example, if the probability of an event is 0.75 (or 75%), then the odds would be 0.75 / ( 1 – 0.75 ) = 3. If one knows the odds O, then one can find the probability P, using the equation P = O / ( 1 + O ). For example, if the odds are 4:1, or 4, then the probability is 4/5 = 0.80, and if the odds are 1:4, or 0.25, then the probability is 0.25 / 1.25 = 0.20.
In order to use this long-known formula, one has to have a way to convert the p-value into a value to use for the “Likelihood Ratio” term, which in this context is called a Bayes Factor. For the Nature article, Nuzzo used a conversion proposed in the 1990s by Thomas Sellke, M. J. Bayarri, and James O. Berger and that they eventually published in the widely read American Statistician. That conversion has an appealing statistical motivation as the minimum possible value for the Bayesian Factor among a realistic set of candidates and thus it provides a useful plausible lower bound on the Bayesian Factor for p < 1 /e ≈ 0.368, where e is the Euler number, exp(1) ≈ 2.718,7 BayesFactor = – e * p * ln(p), where ln(p) is the natural logarithm of p. (For p ≥1/e, one can use BayesFactor=1.) For example, p=0.04 would result in a BayesFactor of -exp(1) * 0.04 * ln( 0.04 ), approximately 0.35. So, if one specified that the prior probability for the null hypothesis is 50%, a toss-up, that corresponds to a prior odds of 1, then the BayesFactor for a p-value of 0.04 converts that prior odds of 1 into a posterior odds of 0.35, which corresponds to a posterior probability of 26% for the null hypothesis, substantially higher than 4%. In the analogous setting of diagnostic medicine, consider a test result that moves a physician’s suspicion for whether the patient has a disease from a pre-test value of 50% up to a post-test value of 74%. Such a result would be considered useful, but it would not be considered definitive, something for clinicians to keep in mind when they see that a study’s p-value was just under 0.05!
Another notable conversion of the p-value into a Bayes Factor was, as far as I can tell, first reported in a pioneering 1963 article in the social sciences literature that was authored by illustrious Bayesian statisticians. 8 That same Bayes Factor formula can be found clearly presented in the second5 of Steven N. Goodman’s excellent two-part set of Annals of Internal Medicine articles concerning fallacious use of p-values in evidence-based medicine. That conversion involves statistics that have an approximately normal distribution and is thus applicable to most statistics in the medical literature. That conversion reports the minimum theoretically possible value for the Bayes Factor, BayesFactormin = exp( – Z2 / 2 ), where Z is the number of standard errors the test statistic is from the null value. (Z can be estimated in Microsoft Excel by using the formula Z = NORMSINV( p ) or Z = NORMSINV( p / 2 ). For example, a two-sided p-value of 0.04 corresponds to Z ≈ -2.054 and a BayesFactormin of exp( – (-2.054 * -2.054) / 2 ) ≈ 0.121. So, if the prior probability for the null hypothesis is 50%, a p-value of 0.04 would mean that, at the minimum, the null hypothesis has a posterior probability of 0.121 / 1.21 = 10.8% of being true, substantively higher than the 4% probability that the popular misinterpretation of p-values would yield. When that factor was introduced in the 1963 article, it was noted by the authors as not being one that would be realistically attained by any study, as it would involve an impossibly lucky guess for the best possible prior probability to use, but it is still useful mathematically because it results in a theoretical minimum for the posterior probability that the null hypothesis is true. In mathematics, we routinely use well-chose impractical scenarios to define the limits for what is practically possible. Given that decisionmakers want to know how probable the null hypothesis remains in light of the study data, it is helpful to know the minimum possible theoretical value for it. Table 2 shows these posterior probabilities for the same inputs used above in Table 1. Notably, a p-value of 0.05 may not even be enough to make the null hypothesis less likely than not!
References
Bacterial genes lead researchers to discover a new way that lactic acid bacteria can make energy and thrive in their environments
/in ISAPP Science Blog /by KCLactic acid bacteria are an important group of bacteria associated with the human microbiome. Notably, they are also responsible for creating fermented foods such as sauerkraut, yogurt, and kefir. In the past two decades, culture-independent techniques have allowed scientists to sequence the genomes of these bacteria and discover more about their capabilities.
Researchers studying a type of lactic acid bacteria called Lactiplantibacillus plantarum found something unexpected: they contained genes for making energy in a way that had not been previously documented. Generally, living organisms obtain energy from their surroundings either by fermentation or respiration. L. plantarum have long been understood to obtain energy using fermentation, but the new genetic analysis found they had additional genes that were suited to respiration. Could they be using both fermentation and respiration?
ISAPP board member Prof. Maria Marco is a leading expert on lactic acid bacteria and their role in fermented foods and in human health. In her lab at University of California Davis, she decided to investigate why L. plantarum had genes equipping it for respiration. Her group recently published findings that show a new type of “hybrid” metabolism used by these lactobacilli.
Here is a Q&A with Prof. Marco about these exciting new findings.
What indicated to you that some of the genes in L. plantarum didn’t ‘belong’?
Organisms that use respiration normally require an external molecule that can accept electrons, such as oxygen. Interestingly, some microorganisms can also use solid electron acceptors located outside the cell, such as iron. This ability, called extracellular electron transfer, has been linked to proteins encoded by specific genes. L. plantarum had these genes, even though this species is known to use fermentation. We first learned about their potential function from Dr. Sam Light, now at the University of Chicago. Sam discovered a related pathway in the foodborne pathogen Listeria monocytogenes. Sam came across our research on L. plantarum because we previously published a paper showing that a couple of genes in this pathway are switched on in the mammalian digestive tract. We wondered what the proteins encoded by these genes were doing.
How did you set out to investigate the metabolism of these bacteria?
We investigated this hybrid metabolism in a variety of ways. Using genetic and biochemical approaches we studied the extent to which L. plantarum and other lactic acid bacteria are able to use terminal electron acceptors like iron. Our collaborators at Lawrence Berkeley National Lab and Rice University contributed vital expertise with their electrochemistry experiments, including making fermented kale juice in a bioelectrochemical reactor.
What did you find out?
We discovered a previously unknown method of energy metabolism in Lactiplantibacillus plantarum. This hybrid strategy blends features of respiration (a high NAD+/NADH ratio and use of a respiratory protein) with features of fermentation (use of endogenous electron acceptors and substrate-level phosphorylation).
We verified that this hybrid metabolism happens in different laboratory media and in kale juice fermentations. We also found that, in the complex nutritive environment of a kale juice fermentation, this hybrid metabolism increases the rate and extent of fermentation and increases acidification. Within the ecological context of the fermented food, this could give L. plantarum a fitness advantage in outcompeting other microorganisms. This could potentially be used to change the flavor and texture of fermented foods.
This discovery gives us a new understanding of the physiology and ecology of lactic acid bacteria.
Are there any indications about whether this energy-making strategy is shared by other lactic acid bacteria?
Some other fermentative lactic acid bacteria also contain the same genetic pathway. It is likely that we are just at the tip of the iceberg learning about the extent of this hybrid metabolism in lactobacilli and related bacteria.
Your finding means there is electron transfer during lactic acid bacteria metabolism. What does this add to previous knowledge about bacterially-produced ‘electricity’?
Certain soil and aquatic microbes have been the focus of research on bacterially-produced electricity. We found that by giving L. plantarum the right nutritive environment, it can produce current to the same level as some of those microbes. We believe there is potential to apply the findings from our studies to better inform food fermentation processes and to guide fermentations to generate new or improved products. Because strains of L. plantarum and related bacteria are also used as probiotics, this information may also be useful for understanding their molecular mechanisms of action in the human digestive tract.
How might this knowledge be applied in practice?
Our findings can lead to new technologies which use lactic acid bacteria to produce healthier and tastier fermented foods and beverages. Because this hybrid metabolism leads to efficient fermentation and a larger yield, it could also help minimize food waste. We plan to continue studying the diversity, expression, and regulation of this hybrid metabolism in the environments in which these bacteria are found.
Decoding a Probiotic Product Label
/in Consumer Blog /by KCBy Mary Ellen Sanders, PhD
Interested in knowing what’s in your probiotic product? Unfortunately, there are many ways that probiotic product labels can fall short.
First, not all items labeled as “probiotic” truly meet the scientific criteria for a probiotic product. See here for information on what qualifies as a probiotic. Some fermented foods are marketed today claiming to be ‘probiotic’. But these products often have undefined microbial content and lack any studies documenting health effects, criteria that are required for a probiotic. Instead, such products could legitimately be labeled as containing ‘live, active cultures’. Dietary supplement products formulated with untested microbes should similarly not be labeled as probiotics.
Also, a label might not provide adequate information on what types of microbes are contained in the product. Genus and species may be listed, but no strain designation. Maybe only “bifidobacteria” or “lactobacilli” are listed.
They might not disclose the potency of individual strains in the product. Some may provide a total count of colony forming units (cfu)/dose or serving, which in the case of a single strain product is informative. But in the case of a multi-strain product – products may contain a couple or up to 30 strains – you don’t know if equal amounts of all strains are included, or perhaps the bulk of the count is made up of the strain in the formulation that is least expensive to manufacture rather than the one that will make the probiotic more effective. Some products may provide one count for “Lactobacillus” and another count for “Bifidobacterium”, a slightly more informative approach than just total count, but still lacking in detail. Many challenges exist for multi-strain products, including the lack of robust methods to quantify different strains once combined, especially viable ones. This topic was one that was covered in an ISAPP webinar, Current issues in probiotic quality: An update for industry.
The label may state that the count is “at time of manufacture”, a number that is no doubt inadequate if you purchase the product close to the end of its shelf-life. Most probiotic strains suffer cell count decline over the course of shelf-life, with some strains more susceptible than others. This situation almost guarantees that by the pull-by date for a multi-strain product, the ratio of cfu of strains to each other is likely much different than at the time of formulation.
Surveys of probiotic product labels
Additionally, it is difficult for consumers to know what products are backed by studies documenting a health benefit. If a product is not labeled sufficiently, it is impossible to link it to evidence. Two studies surveyed commercial probiotic products in the Washington DC area, Retail Refrigerated Probiotic Foods and Their Association with Evidence of Health Benefits and More Information Needed on Probiotic Supplement Product Labels. Results showed that 45% of retail dietary supplement products did not provide strain designations and an equal number did not provide a measure of potency through the end of shelf-life. Only 35% of products could be linked (based on strain and dose) to evidence of a health benefit. Food products fared a bit better, with 49% of refrigerated probiotic food products being linked to evidence of a health benefit. One clear indication from this study was that if your food product discloses the strain designation, it is likely to have evidence of a health benefit. An overall conclusion was that product labeling – at least in this region – needs improvement.
Historical context on guidelines for probiotic product labels
According to the FAO/WHO 2002 Working Group on Guidelines for the Evaluation of Probiotics in Food (page 39 of this combined document), the following information should be on probiotic labels:
– Genus, species and strain designation for each probiotic strain in the product.
– Minimum viable numbers of each probiotic strain at the end of the shelf-life, typically expressed in colony forming units (or cfu).
– The suggested serving size (or dose) must deliver the effective dose of probiotics related to any health benefit communicated on the label.
– Health claim(s) (as allowed by law and substantiated by studies)
– Proper storage conditions
– Corporate contact details for consumer information
These principles are developed and reiterated in “Best Practices Guidelines for Probiotics” (2017) developed by the Council for Responsible Nutrition and IPA.
Additional information
ISAPP created an infographic to explain the information on a probiotic labels. Our example portrays an imaginary dietary supplement for sale in the United States. Labels on foods containing a probiotic or a probiotic produced in another country would be labeled differently from this example to comply with applicable regulations. For those interested in probiotic labels in the EU, see the infographic ISAPP created for a probiotic product in the European Union. Also of interest, USP.org created an infographic on “How to Read a Dietary Supplement Label” for U.S. products.
ISAPP begins its search for the next executive director
/in News /by KCBy Mary Ellen Sanders, PhD
When Glenn Gibson, Irene Lenior-Wijnkoop and I first kicked around the idea of an organization for scientists and dedicated to science of probiotics and prebiotics (in 1999), I don’t think any of us would have anticipated that in 2022 that organization would be celebrating its 20th anniversary. But I do think if we could have seen where we’ve come, we would have been pleased.
As I write, my mind is racing over so many highs (dinner at the National Academy of Sciences in DC and lunch at the top of the ski lift in Copper Mountain, Colorado) and lows (broken down and lost buses on the way to our gala event in Berkeley) of our gatherings and the many positive outcomes due to ISAPP bringing excellent industry, academic and government scientists together to discuss how to improve the field.
But my purpose with this blog is to look forward, not back. I want to let the ISAPP community know that ISAPP is searching for a new executive director to take over for me when I retire in June 2023. The position announcement is found here.
Together, we have accomplished a lot in ISAPP’s twenty-year existence: dozens of publications, including five highly-viewed scientific consensus definitions, interactions with U.S. regulators that helped them reduce barriers to human research on probiotics, networking among scientists leading to collaborative research, concise summaries of the ‘biotics’ family of substances and fermented foods in friendly language in our series of infographics and videos, developing continuing education materials, collaborating with scientific organizations on projects of mutual interest, and more. We have built a strong, collegial community, which is leading cutting-edge science on probiotics, prebiotics, synbiotics, postbiotics and fermented foods. These community members have many more goals and ambitions to realize in the years ahead. I am ready to pass the baton to someone who will build on the momentum of the past and extend ISAPP’s global activities. I look forward to what the future brings for our vibrant organization.