A randomized double-blind placebo controlled pilot study of probiotics in adolescents with severe obesity

Abstract

Purpose

The purpose of the study is to assess the effect of probiotic supplementation on gut microbiota and insulin resistance in adolescents with severe obesity.

Methods

Through a randomized, double blind, placebo-controlled, 12-week pilot clinical trial, 15 adolescents with severe obesity received either an oral probiotic ‘Visbiome®’ (n = 8) or placebo (n = 7). Anthropometry, fasting glucose, insulin, hs-CRP and stool for microbiome and calprotectin were collected at baseline (week 0) and 12 weeks after intervention.

Results

Among completers (n = 4 in each of the two groups), mean change in fasting glucose was significantly lower in the probiotic group (0 ± 4 mg/dL) as compared to the placebo group (6.3 ± 1.7 mg/dL) (p = 0.028). Gut microbial Firmicutes to Bacteroidetes (F/B) ratio had a greater decline from week 0 to week 12 in the probiotic group (mean 17.7 ± 25.1 to 2.39 ± 2.0, respectively) but was not statistically significant (p = 0.06) as compared to in the placebo group (mean 12.8 ± 18.2 to 6.9 ± 5.61, respectively) (p = 0.89). Weight and BMI (mean ± SD) trended to remain stable in the treatment group (-1.07 ± 6.1 kg and -0.3 ± 2.2 kg/m² respectively) as compared to the placebo group (3.9 ± 5.1 kg, 1.0 ± 1.6 kg/m²) but was not significant (p = 0.12 for weight and 0.38 for BMI). No significant change in the fasting insulin, HOMA-IR, or serum and stool inflammatory markers were noted between the two groups (p > 0.05). One participant in the treatment arm reported adverse effects of gastrointestinal intolerance.

Conclusion

Probiotic therapy with Visbiome® may improve the fasting glucose and possibly decrease the gut microbial F/B ratio as compared to placebo in adolescents with severe obesity. Future larger studies are required to confirm these findings.

U.S. Clinical Trial Registry number: NCT03109587

Supplementary Information

The online version contains supplementary material available at 10.1007/s40200-021-00855-7.

Introduction

Every year roughly 5000 youth are diagnosed with Type 2 Diabetes (T2D) in the US [1]. Pediatric T2D is rapidly progressive to complications and current therapeutic options fail to prevent the progression of disease [2]. Novel strategies to either reduce the incidence or halt the rapid progression of T2D in youth are urgently needed. Aside from the well-established modifiable risk factors of diet and physical activity, gut microbial dysbiosis is a newly emerging environmental risk factor that can modulate insulin sensitivity and risk for T2D [3]. Compared to lean, healthy counterparts, microbial dysbiosis in obesity and metabolic syndrome is characterized by decreased microbial diversity and richness [4, 5], increased Firmicutes and reduced Bacteroidetes in adults and children [6–8]. Dysbiosis leads to an increase in intestinal permeability and triggers the systemic low-grade inflammation in obesity and related disorders [9, 10]. Therefore, modulation of the gut microbiota may have great potential to treat obesity, insulin resistance and prevent progression to T2D.

Fecal microbiota transplantation from lean donors to obese individuals with metabolic syndrome has been shown to increase gut microbial diversity and improve insulin sensitivity [11] suggesting that microbial dysbiosis is potentially modifiable and may affect insulin sensitivity. In mice, probiotic treatment has shown promising outcome for the prevention and treatment of obesity and related diseases, even without modulation of caloric intake [12, 13]. In human adults with obesity and T2D, randomized controlled trials of probiotics have also shown improvement in metabolic syndrome and glucose metabolism [14–16]. However, microbiota studies in youth with obesity are limited. Randomized controlled trials in children with obesity have either assessed the effect of a single bacterial strain supplementation [17–20] or have not analyzed the effect on gut microbiota [18–24].

It remains unclear if the microbiota or insulin sensitivity in youth with obesity can be modified through probiotic therapy. Furthermore, the interpretation of microbiota data in animal studies and human clinical trials is complicated by the methodology used for microbiome analysis. Of the limited studies in the pediatric population, two randomized clinical trials that assessed gut microbiota changes by targeted RT-PCR approach [25] and 16S rRNA sequencing [26], did not find a change in gut microbiota after a multi-strain probiotic treatment. In this pilot study, we aimed to determine if an oral multi-strain probiotic ‘Visbiome®’ will change the gut microbiota, insulin resistance and inflammation in adolescents with severe obesity, by using metagenomic shotgun sequencing.

This combination of probiotic ‘Visbiome®’ was chosen because of glycemic benefits shown in humans and mice models. Supplementation of a single strain of Lactobacillus (L. casei and L. rhamnosus GG) has resulted in improved glycemic control in diabetic mice [27, 28]. Further, strains of Bifidobacteria and Lactobacilli promote growth of butyrate producers [29, 30] that may have protective effects against T2D [10]. ‘Visbiome®’ has been used in obese mouse models [31] and in women with gestational diabetes mellitus [32] resulting in beneficial effects on insulin resistance and inflammation.

As the prevalence of severe obesity in US children continues to increase [33, 34] and puberty poses a physiologic state of insulin resistance, our study population was focused on pubertal (Tanner stage ≥ 3) adolescents with BMI ≥ 99^th percentile (and > 120% of the 95^th percentile) and acanthosis on exam, who are clinically at high risk for development of T2D.

We hypothesize that treatment with the probiotic ‘Visbiome®’ will alter the gut microbiota as assessed by metagenomic shotgun sequencing, specifically decrease the Firmicutes to Bacteroidetes (F/B) ratio and decrease insulin resistance as compared to placebo in adolescents with severe obesity.

Methods

Study design

A randomized, double-blind, placebo-controlled pilot study with 12 weeks of intervention was conducted at Seattle Children’s Hospital, Seattle, WA, USA from June 2018 to July 2019. The study protocol was approved by Seattle Children’s Institutional Review Board (IRB). The clinical trial was registered at www.clinicaltrials.gov (NCT03109587). Written informed assent and consent was obtained from participants and their parent/guardian. A formal power analysis was not done as this was a pilot study. The initial study was planned to recruit 16 adolescents (n = 8 in each group), however one participant withdrew after consenting. Fifteen adolescents were then randomized to receive either placebo (n = 7) or probiotic (n = 8). Investigators and participants were blinded to the intervention. Blood and fecal samples were collected before starting treatment (week 0) and at the end of treatment while still on study medication (week 12). Weekly phone calls were made to ensure compliance and document any self-reported adverse effects (SAE). Participants were compensated with $50/visit and parents received $25/visit.

Two months after study initiation, one participant who had lactose intolerance, reported persistent abdominal bloating and diarrhea (SAE) thought to be due to study medication. The study was paused for three months to modify the protocol to exclude participants with ‘any food or gastrointestinal intolerance’, which was approved by the IRB. Existing participants were contacted and given the option to opt out if they met the new exclusion criteria; none left the study at this period.

Participants

Youth (aged ≥ 13 years) with severe obesity (BMI ≥ 99th percentile for age and > 120% of the 95^th percentile) followed in the endocrine clinic for 1 year prior, were identified through the Seattle Children’s Hospital clinical repository database and were assessed for eligibility (n = 160). They were contacted via a letter and phone call or approached at their regular endocrine clinic visit. Of these, 145 participants were excluded, as 78 did not meet inclusion criteria (either based on chart review, or after contacting them via phone or when approached at their clinic visit), 53 were unable to be reached via phone and one withdrew after consenting at the clinic visit. Fifteen participants who met the inclusion and exclusion criteria were consented at their endocrine clinic visit. They were scheduled for the study visit 1 (week 0) within 2 weeks of consenting.

Inclusion criteria

Male or female, age 13 to 19 years, pubertal Tanner stage ≥ 3, BMI ≥ 99^th percentile (and > 120% of the 95^th percentile) for age and acanthosis nigricans on physical exam.

Exclusion criteria

Based on medical chart review and/or self-reported history of: secondary diabetes (such as cystic fibrosis-related diabetes, pancreatitis, etc.), or monogenic diabetes; any severe underlying disorders, such as cancer, immunodeficiency and active inflammatory diseases requiring anti- inflammatory medication; current or previous treatment with insulin, metformin, antibiotics or probiotics (including any foods with added probiotics/synbiotics) within 3 months prior to recruitment; start of a new dietary intervention within 1 month prior to recruitment.

Study procedures

Study visit 1, SV1 (week 0)

The first study visit was conducted after a minimum of 8 h fasting. Anthropometric measurements of height and weight, vital signs (heart rate, blood pressure, respiratory rate, and oral temperature), medical history including a history of any major change in dietary and physical activity, family history of diabetes and ethnicity were obtained. A complete physical exam including pubertal Tanner staging was done by a pediatric endocrinologist.

Stool collection kits (Kendall Precision stool container labelled with participant study ID and visit number, Fisher Scientific brand plastic stool hat and disposable latex micro-touch surgical gloves) with detailed instructions to self-collect at home were given. Two stool kits for collection of 2 samples (one at the start and one at end of study period 12 weeks) were given. Participants were asked to obtain at least 30 mL of fecal sample (marked on stool container) and the date of collection was to be noted on the container. Pre-labelled envelopes were provided, to be sent to the Center of Microbiome Sciences and Therapeutics (CMiST) lab at University of Washington, Seattle, WA within 24 h of collection by overnight shipping at room temperature. Participants were strictly instructed to start medication only after the initial stool sample had been collected and date of starting the medication was documented. This was confirmed by phone follow up. Participants were taught the method of administration of study medication and asked to refrain from a new diet or physical activity intervention for the duration of the study.

Measurements of weight (digital scale) and height (stadiometer) were obtained. All measurements were taken to the nearest 0.1 kg or 0.1 cm and taken 3 times with an average of the three used as the result. Body Mass Index (BMI) was calculated using weight and height (kg/m²). BMI Z-score and percentiles were based on CDC growth charts for boys and girls aged 2–20 years [35].

Blood for (fasting) glucose, insulin, and high sensitivity CRP (hs-CRP) was obtained in containers for K2 EDTA and sodium fluoride EDTA plasma. Samples were inverted 8 times, placed on ice, centrifuged at 4 °C for 10 min at 1300 g. Aliquots of 0.5 mL in color coded micro tubes were stored at -70 °C until shipment to Northwest Lipid Research Laboratories at University of Washington. Fecal samples upon arrival to the CMiST lab were immediately aliquoted into 1 mL in a BSL2 sterile chamber and stored at -80 °C freezer until analyzed for microbiota composition and calprotectin.

Randomization and blinding

Using a computer-generated sequence, 1:1 randomization was performed by the hospital Research Pharmacy. The investigators, pharmacy, lab personnel and participants were blinded to the intervention until completion of the entire study.

Study medication

Participants were instructed to take 2 sachets of study medication daily at bedtime, dissolved in cold water, for 12 weeks. The probiotic Visbiome® (marketed as Vivomixx® in Europe) manufactured by ExeGi Pharma, LLC, is a mixture of Lactobacilli and Bifidobacteria strains and was provided by the drug company at no cost. Each sachet consists of total 900 billion of the following probiotic strains: Lactobacillus plantarum DSM 24730, Lactobacillus plantarum DSM 24731, Lactobacillus plantarum DSM 24735, Lactobacillus plantarum DSM 24801, Lactobacillus salivarius DSM 24800, Lactobacillus paracasei DSM 24737, Lactobacillus delbrueckii DSM 25998, Pediococcus pentosaceus DSM 24734, Bifidobacterium animalis DSM 24736, Bifidobacterium breve DSM 24732. Inactive ingredients were maltose and silicon dioxide. The placebo contained only maltose and silicon dioxide without any probiotic strains. Placebo and probiotic had the same appearance. Our study medication qualified for Investigational New Drug (IND) exemption by the US FDA and no funds were received from the pharmaceutical company for this study.

Study visit 2, SV2 (Week 12)

Following SV1, participants were contacted by weekly phone calls to encourage compliance and to troubleshoot problems or document any SAE. At the end of 12 weeks of medication, participants and their parents/caregivers were asked to bring all unused medication to the final study visit (SV2). The second stool specimen was collected during week 11–12 at home with the same instructions. Measures of anthropometry, physical exam and history of a major change in diet and physical activity were obtained. All participants confirmed that no new dietary or physical activity intervention was initiated for the duration of the study. Any SAE were documented. Compliance was assessed by counting left-over medication.

Microbiota analysis

Metagenomic sequencing was chosen for microbiota analysis as it has a higher resolution and sensitivity to detect changes at the microbial phyla and genera level as compared to 16S rRNA sequencing [36, 37].

Library preparation and sequencing

DNA were isolated from fecal samples using a QIAamp PowerFecal Pro DNA kit (QIAGEN) according to manufacture instructions. DNA sequencing libraries were prepared using a QIAseq FX DNA Library kit (QIAGEN) with 0.2–0.7 ng of purified DNA. The sequencing libraries were cleaned up using 1X Agencourt AMPure XP beads (Beckman Coulter). Libraries were sequenced on a NextSeq 500, with an average of 5.32 × 10⁷ reads per sample (range 4.84 × 10⁷ – 5.86 × 10⁷). Reads were adapter- and quality-trimmed using the QIAGEN CLC Genomics Workbench (version 11.0) and the Data QC. Filtered reads were mapped to a QIAGEN CLC curated microbial genome database containing 13,305 sequences from 1455 bacterial species using Microbial Genomics Pro Suite Module (version 2.2) to determine taxonomic composition. Specific CLC workflow information is available at http://resources.qiagenbioinformatics.com/tutorials/Taxonomic_Profiling.pdf. All taxa with < 1% relative abundance in all samples or > 1% relative abundance solely in extraction blanks were pooled into the “Other” category (Supplementary Table S1). The Bacteroidetes and Firmicutes taxa was calculated followed by log₁₀ transformation of the Firmicutes to Bacteroidetes (F/B) ratio. The F/B ratio between week 0 and week 12 was compared in the two groups. Alpha diversity (microbial richness) was determined using the Simpson Diversity Index and Beta diversity was calculated using the Bray–Curtis dissimilarity metric to compare between the probiotic and placebo groups.

Metabolic markers

Analysis was performed at Northwest Lipid Research Laboratories (NWLRL), University of Washington for insulin (two site immuno-enzymometric assay on a TOSOH 2000 auto-analyzer, Tosoh Bioscience) and glucose (Roche c501 Cobas auto-analyzer, Roche Diagnostics, North America) based on the hexokinase enzymatic measurement. Insulin assay has a sensitivity level of 0.5 uU/mL. A set of high, medium, and low insulin level control samples were included in each analytical batch to monitor the assay performance which resulted in an inter assay coefficient of variation (CV) of 2.8%, 2.5%, and 2.0% respectively. The fasting glucose assay has a sensitivity of 2 mg/dL and the inter-assay CV, determined on quality control samples with low, medium, and high levels of glucose, are 2.0%, 1.7%, and 1.3%, respectively. Insulin resistance was assessed by the Homoeostatic model assessment (HOMA) equation by Matthews et. al [38].

Inflammatory markers

High sensitivity C-reactive protein (hs-CRP) processed at NWLRL, University of Washington, was performed immunochemically using Siemens reagents (Siemens Healthcare Diagnostics. Inc, Newark, DE) on a Siemens BNII Nephelometer auto-analyzer. The assay sensitivity is 0.004 mg/dL and the inter-assay CVs on quality control samples with low, medium, and high CRP levels were 2.9%, 1.8%, and 1.4%, respectively. Fecal calprotectin was measured on Dynex DS2 analyzer, an automated EIA platform (ELISA kit and QUANTA Lite™ Calprotectin Extended Range reagent from INOVA Diagnostics, Inc. San Diego, USA). Per assay, calprotectin reference range is < 27 mg/kg (normal), 27–50 mg/kg (borderline) and > 50 mg/kg (abnormal). Values of < 27 were used as 27 and > 50 were used as 50 for the analysis.

Outcomes

Primary outcome

The change in gut microbiota composition, assessed by alpha diversity, beta diversity and the Firmicutes to Bacteroidetes (F/B) ratio between the placebo vs. probiotic groups.

Secondary outcome

The difference in mean change in metabolic markers (fasting glucose, insulin, HOMA-IR), inflammatory markers (hs-CRP, fecal calprotectin) and anthropometry between the two groups.

Statistical analysis

Data of all completers was analyzed as per protocol analysis. Statistical analyses were performed using the Graph Pad software version 8.0 for clinical metrics and in R (version 3.4.2) (R Core Team. R: A Language and Environment for Statistical Computing. Vienna, Austria: 2017. https://www.R-project.org/) for microbiological metrics. The mean change in metabolic and inflammatory markers was calculated by descriptive statistics. Continuous variables were presented as mean ± SD and categorical variables as proportions. Two-tailed Student’s t-test and Wilcoxon ranked sum test (for microbial diversities) were performed to compare the continuous variables between two groups. Chi-square tests were used to compare categorical variables between the two groups. PERMANOVA and homogeneity of variance were used to compare beta diversity metrics for microbial data. Wilcoxon signed rank sum tests with a Benjamini Hochberg correction for multiple comparisons was used to investigate individual taxa with significantly different relative abundances between week 0 vs. week 12 and to compare the F/B ratio between groups. P value of ≤ 0.05 was considered significant.

Results

Participant follow up and baseline characteristics

Of the 15 adolescents randomized to receive either placebo (n = 7) or probiotic (n = 8), three (42.8%) in the placebo group and two in the probiotic group (25%) were lost to follow up. One of 8 participants in the probiotic group had incomplete follow up measures due to inability to schedule for SV2 (only fecal sample and left-over medication received, no blood or anthropometry obtained). Further, another one of 8 participants (12.5%) in the probiotic group had persistent SAE thought to be due to study medication, which led to discontinuation and was excluded (Fig. 1).

Fig. 1.

Screening, randomization and follow up of study participants

All baseline characteristics were similar: age (15.7 ± 1.8 vs. 15.8 ± 1.8 years), weight (136.2 ± 26.5 vs. 134.9 ± 36.1 kg), BMI (45.9 ± 10.7 vs. 45.4 ± 12.3 kg/m2), BMI Z-score (2.7 ± 0.26 vs. 2.6 ± 0.28) for placebo and probiotic groups, respectively (Table 1). There was no statistically significant difference between the two groups regarding sex, ethnicity or family history of diabetes mellitus. Baseline metabolic labs were available for n = 4 for the placebo group and n = 5 for the probiotic group. Of these, none in the placebo group and two in the probiotic group had impaired fasting glucose of ≥ 100 mg/dL.

Table 1.

Baseline characteristics of all participants randomized to intervention at week 0

Baseline characteristics	Placebo (n = 7)	Probiotic (n = 8)	Total (n = 15)	P value
Age in years (mean ± SD)	15.7 ± 1.8	15.8 ± 1.8	15.86 ± 1.7	0.76
Sex % of female (n)	42.8 (3)	62.5 (5)	53.5 (8)	0.44
Ethnicity or racea % (n)
Caucasian	43 (3)	62.5 (5)	53.5 (8)	0.44
Hispanic	14 (1)	0	6.7 (1)
African American	14 (1)	0	6.7 (1)
Mixed	14 (1)	25 (2)	20 (3)
Middle Eastern	0	12.5 (1)	6.7 (1)
No-response	14 (1)	0	6.7 (1)
Baseline weight in kg (mean ± SD)	136.2 ± 26.5	134.9 ± 36.1	135.5 ± 30.8	0.94
Baseline BMI in kg/m2 (mean ± SD)	45.9 ± 10.7	45.4 ± 12.3	45.7 ± 11.2	0.46
Baseline BMI Z-score (mean ± SD)	2.7 ± 0.26	2.6 ± 0.28	2.7 ± 0.27	0.55
1st or 2nd degree family history of GDM or T2D % (n)	57 (4)	75 (6)	66.7 (10)	0.46
Obesity related co-morbidities % (n)	100 (7)	100 (8)	100 (15)
Polycystic Ovarian Syndrome (PCOS)	14 (1)	25 (2)	20 (3)	0.85
Hypertension	28 (2)	12.5 (1)	20 (3)	0.43
Obstructive Sleep Apnea (OSA)	28 (2)	37.5 (3)	33.3 (5)	0.71
Non-Alcoholic Fatty Liver Disease (NAFLD)	43 (3)	37.5 (3)	40 (6)	0.83
Dyslipidemia	57 (4)	62.5 (5)	60 (9)	0.83
bImpaired fasting glucose % (n)	0 (0)	25 (1)	12.5 (1)

There were no significant differences between the groups for all listed characteristics. P value is calculated by the two-tailed Student’s t-test (continuous variables) and chi-square test (categorical variables). ^aEthnicity/race was as reported by the patient or patient guardian/parent. ^bFrequency of impaired fasting glucose was only calculated for completers for blood analysis (n=4 for the placebo and n=4 for the probiotic group).

Compliance

Medication compliance as assessed by percentage of doses consumed in the placebo and probiotic group was similar: 85.7% (minimum 69 to maximum 97.6%) and 83.5% (minimum 64.2 to maximum 100%), respectively (Supplementary Figure S1).

Primary outcome

Fecal microbiota composition data was analyzed for those who completed the study in the placebo (n = 4) and probiotic (n = 5) groups. While there was no significant difference in the relative abundance at the genus or phylum level in both groups (Fig. 2A, B), there appeared a trend towards a higher relative abundance of Bacteroidetes and a lower relative abundance in Proteobacteria, Actinobacteria and Firmicutes phyla in the probiotic compared with the placebo group (Fig. 2B).

Fig. 2.

Proportion (relative abundance) of taxonomic composition of fecal microbiota at genus (A) and phylum (B) level at baseline (week 0) and at the end of intervention (week 12) in placebo (n = 4) vs. probiotic group (n = 5). The top 8 genera are listed in the legend for ease of viewing. Genera collapsed into the “other” category are listed in Table S1

The mean Firmicutes to Bacteroidetes (F/B) ratio had a greater decline in the probiotic group from week 0 to week 12 (17.7 ± 25.1 to 2.39 ± 2.0) but was not statistically significant (p = 0.06), as compared to the placebo group (12.8 ± 18.2 at week 0 to 6.9 ± 5.61 at week 12) (Fig. 3).

Fig. 3.

Log10 transformation of the Firmicutes: Bacteroidetes (F/B) ratio at week 0 (baseline) and at the end of treatment (week 12) in completers of placebo (n = 4) vs. probiotic group (n = 5). P value is calculated by the Wilcoxon ranked sum test with a Benjamini Hochberg correction. Data presented in boxes represents the interquartile region (25^th to 75^th percentile) and middle lines represent the medians

Alpha diversity (microbial richness) as measured by the Simpson index was not significantly different from baseline to after treatment in either of the groups at the genus (placebo: p = 0.65, probiotic: p = 1.0) or phylum (placebo: p = 0.65, probiotic: p = 0.60) levels (Fig. 4).

Fig. 4.

Alpha diversity (Simpson) index in placebo (n = 4) vs. probiotic (n = 5) group at baseline (week 0) and at the end of treatment (week 12) in completers at the genus (A) and phylum level (B). P value is calculated by the Wilcoxon ranked sum test. Data presented in boxes represents the interquartile region (25^th to 75^th percentile) and middle lines represent the medians. (Points are overlapping in all four columns in A and in week 12 probiotic group in B)

Beta diversity at the phylum level appeared different in the probiotic group from baseline to after treatment but was not statistically significant (PERMANOVA p = 0.29, homogeneity of variance p = 1.0) (Fig. 5B). There was no difference in beta diversity at the genus level (PERMANOVA p = 0.39, homogeneity of variance p = 1.0) (Fig. 5A).

Fig. 5.

Beta-Diversity index (in-between sample diversity) in placebo (n = 4) vs. probiotic group (n = 5) at baseline (week 0) and at the end of treatment (week 12) at the Phylum (A) and Genus (B) level. Each small dot represents a different sample, grouped as indicated on the right. Each large dot represents the centroid of each group. The distance between any two dots represents how dissimilar their respective microbial communities are. P value is calculated by the PERMANOVA and homogeneity of variance

Secondary outcomes

Metabolic and inflammatory markers, anthropometry: Mean change in fasting glucose was significantly lower in the probiotic group (mean 0 ± 4 mg/dL) as compared to the placebo group (mean 6.3 ± 1.7 mg/dL) (p = 0.028). The frequency of impaired fasting glucose increased from zero at week 0 to 75% (3 of 4 participants) at week 12 in the placebo group, as compared to a decrease from 25% (1 of 4 participants) at week 0 to zero (no participants) at week 12 in the probiotic group. There was no statistically significant difference in the change in insulin or HOMA-IR index between the groups (Table 2), though mean change in insulin and HOMA-IR after probiotic therapy was lower (7.8 ± 18.4 U/mL and 1.8 ± 3.8, respectively) as compared to after placebo (14.9 ± 13.9 U/mL and 4.2 ± 3.7, respectively). Similarly, no significant differences in the mean change in hs-CRP, fecal calprotectin, weight or BMI were noted between the two groups, except there tended to be an increase in the fecal calprotectin in the probiotic group (mean change 6.2 ± 49.6 mg/kg) as compared to the placebo group (mean change -6.3 ± 11.2 mg/kg) (Table 2).

Table 2.

Difference in secondary outcome measures in completers between baseline (week 0, SV1) and at the end of intervention (week 12, SV2)

Secondary outcome measures(mean ± SD)	Placebo group (n = 4)[2F, 2 M]			Probiotic group (n = 4)#[3F, 1 M]			*P value
	Week 0(SV1)	Week 12(SV2)	*Mean Change(SV2-SV1)	Week 0(SV1)	Week 12(SV2)	*Mean Change(SV2-SV1)
Metabolic markers
Fasting glucose (mg/dL)	94.2 ± 3.1	100.5 ± 4.8	6.3 ± 1.7	94.7 ± 5.5	94.7 ± 2.7	0 ± 4	0.028
Fasting insulin (U/ml)	33.8 ± 22.8	48.7 ± 34.1	14.9 ± 13.9	41.7 ± 20.1	49.5 ± 31.2	7.8 ± 18.4	0.56
HOMA-IR index	7.9 ± 5.1	12.1 ± 8.4	4.2 ± 3.7	9.9 ± 4.9	11.7 ± 7.5	1.8 ± 3.8	0.40
Inflammatory markers
Serum hs-CRP (mg/dL)	0.84 ± 0.55	0.59 ± 0.08	-0.25 ± 0.5	0.41 ± 1.8	0.31 ± 2.4	-0.1 ± 0.26	0.63
Fecal calprotectin (mg/kg)	31.5 ± 7.7	25.2 ± 3.5	-6.3 ± 11.2	56.4 ± 311	62.6 ± 52.71	6.2 ± 49.61	0.48
Anthropometry
Weight (kg)	131.7 ± 28.3	135.7 ± 30.1	3.9 ± 5.1	121.9 ± 15.1	120.8 ± 17.5	-1.07 ± 6.1	0.12
BMI (kg/m2)	44.2 ± 5.2	45.2 ± 5.8	1.0 ± 1.6	41.2 ± 4.1	40.9 ± 5.5	-0.3 ± 2.2	0.38
BMI Z-score	2.72 ± 0.3	2.75 ± 0.35	0.03 ± 0.06	2.64 ± 0.3	2.63 ± 0.32	-0.01 ± 0.1	0.51

p value which is significant are in boldface

*P value is calculated using unpaired t-test to compare the mean change (mean ± SD) in probiotic vs. placebo group. ^#In one participant of the probiotic group, blood and anthropometry were unable to be obtained, therefore n=4. ¹n=5; F=females, M=males.

Self-reported Adverse Effects (SAE)

One of 8 participants (12.5%) in the probiotic group reported persistent diarrhea and abdominal bloating leading to discomfort and opted to discontinue. This participant had self-reported lactose intolerance and dropped out of the study. No adverse effects were reported in the rest of the participants in either group.

Discussion

In this pilot study of ethnically diverse adolescents with severe obesity residing in a metropolitan city area, we determine that treatment with the probiotic Visbiome® is possibly effective in improving fasting glucose and may treat gut microbial dysbiosis by decreasing the F/B ratio more as compared to placebo. Our study allows novel insights into the effectiveness of a probiotic intervention in this population.

Bacteroidetes and Firmicutes constitute the majority of the human intestinal microbiota [39, 40]. Mouse models illustrate that higher Firmicutes and lower Bacteroidetes have an increased capacity to harvest energy from the diet, leading to obesity [41]. In humans, an association between the F/B ratio and obesity rates has been found, but causation has not been established. Obese adults have a fecal dysbiosis characterized by decreased B/F ratio by 16S RNA sequencing, irrespective of diet type [6, 7], and have an increase in the B/F ratio after bariatric surgery and weight loss as detected by metagenomic sequencing [42]. In contrast, a study by Schwiertz et al. [43] reported a higher Bacteroidetes and reduced Firmicutes in overweight and obese adults. A systematic review and meta-analysis of pediatric data showed that while studies in the pediatric population are scarce, a lower B/F ratio could be a risk factor for obesity [8]. Our results demonstrate that though the decrease in the F/B ratio after probiotic therapy as compared to placebo, was not statistically significant (p = 0.06) in this very small pilot clinical trial, it provides important preliminary data to plan targeted therapies for gut microbiota in obese adolescents.

To our knowledge one other study has shown a decrease in the F/B ratio in obese adolescents after a single strain probiotic [19]. Other studies in the pediatric population have shown no change in gut microbiota [25, 26] and most have not assessed microbiota composition [18–24] after intervention with probiotics. We observed that while there were no differences in the alpha (within-sample) diversity after probiotic intervention (Fig. 4), there appeared a change in the overall microbial taxonomic composition (Fig. 2B) at the phylum levels after probiotic treatment; a change which appeared absent in the placebo group.

Studies in adults have shown that decreased Firmicutes and increased Bacteroidetes is associated with improved insulin sensitivity, glucose metabolism [11]. Although we did not observe a difference in the change in insulin resistance (HOMA-IR) in the probiotic vs. placebo group, there was a statistically significant difference in the mean change in fasting glucose. A study by Stefanaki et. al. [25] used the same probiotic constituents (Vivomixx) and showed improvement in fasting glucose and hemoglobin A1c at 1 month after probiotic intervention, but this difference was not sustained at the end of 4 months due to suboptimal compliance. In contrast, participants of our study had higher compliance at the end of 12 weeks (mean doses consumed > 80%) as compared to 56% at the end of 4 months in the study by Stefanaki et al. Kassaian et al. [15] have also reported similar improvements in fasting glucose as well as in fasting insulin and HOMA-IR after synbiotic administration (probiotic + inulin base) with strains of Lactobacillus and Bifidobacterium. A combination of Lactobacilli and Bifidobacteria creates an ideal environment for butyrate producers [34, 35] which increase short chain fatty acid (SCFA) concentrations [10]. SCFAs stimulate L-cells in the intestine that produce GLP-1 [43] thus affecting glucose metabolism. Indeed, use of the VSL#3 probiotic which is a combination of Lactobacilli and Bifidobacteria, showed increase in GLP-1 in obese children with non-alcoholic steatohepatitis [21].

There was a trend towards an increase in the BMI and BMI Z-score in the placebo as compared to the probiotic treated group, though not statistically significant. This may explain the improvement in the fasting glucose after probiotic treatment as compared to after placebo. However, all study participants denied any new interventions to decrease weight for the duration of the study. Therefore, it may be possible that the BMI changes seen in the probiotic group are also due to the study medication itself. Although not statistically significant, there was an increase in fecal calprotectin in the probiotic group as compared to a decrease in the placebo group. It is possible the probiotic caused an increase in local intestinal inflammation, however there was no difference in the serum marker of inflammation as noted by the hs-CRP level. It is possible that the groups were just too small to detect a true difference, or a longer duration of intervention is required to see if these changes are sustainable. The lack of a structured lifestyle modification program could be another reason why we did not observe significant changes in anthropometry or serum inflammatory markers. We note that there were more females in the probiotic treated as compared to the placebo group. Since our sample size was too small for a sub-group analysis to detect a sex- based difference, future studies need to be adequately powered to delineate sex-based changes.

Overall, probiotic therapy was well tolerated except in one participant with lactose intolerance who experienced persistent gastrointestinal adverse effects. Thus, probiotics in individuals with food or gastrointestinal intolerance should be either contraindicated or used with caution.

One limitation of our study is the high attrition rate of 42% in the placebo and 37% in the treatment arm. Similar rates have been noted in previous studies in adolescent populations with obesity. In a systematic review of randomized control trials in children with obesity, attrition rates varied from 5 to 37% [44]. It is well described that older adolescents with obesity and T2D are more likely lost to follow up [45]. Moreover, taking 2 sachets of study medication daily may have been challenging for these subjects as compared to a once daily formulation, leading to higher attrition or lack of sustained interest. Thus, besides developing novel strategies to prevent T2D in youth, qualitative studies to identify reasons of poor follow up in youth with obesity are required. Lastly, despite the limitation of a small sample size in our study, we were able to notice an improvement in fasting glucose and trend towards change in gut microbiota composition in the probiotic treated group as compared to the placebo group. Future larger studies are required to confirm this finding.

Conclusion

Our pilot study demonstrates that in adolescents with severe obesity, a 12-week intervention with a probiotic (Visbiome®) may have an impact on the gut microbiota composition and possibly improves the fasting glucose level as compared to placebo. These results are promising and serve as a gateway for future studies for the development of novel therapeutic targets in this high-risk adolescent population.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Conceptualization, Christian L. Roth, Christiane Hampe and William R. DePaolo; Data curation, Arushi Verma and Maria T. Nelson; Formal analysis, Arushi Verma, Maria T. Nelson and Christian L. Roth; Funding acquisition, Arushi Verma, William R. DePaolo, Christiane Hampe and Christian L. Roth; Investigation, Arushi Verma, William R. DePaolo and Christian L. Roth; Methodology, Arushi Verma and Christian L. Roth; Project administration, Arushi Verma and Christian L. Roth; Resources, William R. DePaolo, Christiane Hampe and Christian L. Roth; Software, Maria T. Nelson; Supervision, Christian L. Roth; Validation, William R. DePaolo and Christian L. Roth; Visualization, Arushi Verma, William R. DePaolo and Christian L. Roth; Writing – original draft, Arushi Verma; Writing – review & editing, Arushi Verma, Maria T. Nelson, William R. DePaolo and Christian L. Roth.

Funding

This research was funded by the Office of Research Central, University of Washington (grant A118785 Royalty Research Fund, Pilot Study of Synbiotics in Pre-diabetic Adolescents).

Data availability

Data and materials are available upon request.

Code availability

Not applicable.

Declarations

Ethics approval

This study was approved by the institutional IRB (Seattle Children’s IRB).

Consent to participate

Informed consent and assent was obtained by all participants of the study.

Consent for publication

All authors consent for publication.

Conflict of interest/Competing interest

The authors declare no conflict of interest. The funders had no role in the design of the study, in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.