Clinical Response to Low FODMAP Diet in Children With IBS
Clinical Response to Low FODMAP Diet in Children With IBS
Children, aged 7–17 years, with paediatric Rome III-defined IBS determined by completion of the paediatric Rome III GI symptom questionnaire were eligible. Potential subjects were identified by screening referrals to tertiary paediatric gastroenterology care for chronic abdominal pain and via community newsletters and Internet advertisements. Parents and children were further screened by phone for inclusion and exclusion criteria and to establish current symptoms prior to enrolment as previously described. Study recruitment began in September 2011 and ended in December 2013. Baylor College of Medicine Institutional Review Board approval was obtained. The study was registered as ClinicalTrials.gov identifier NCT01339117.
Subjects completed a 7-day period in which they continued ingesting their habitual (baseline) diet (Figure 1). Following this baseline period, we employed a randomised, double-blind, crossover study design. Subjects received either a low FODMAP or typical American childhood diet (TACD) for 48 h. After 48 h on the first assigned diet, they returned to their habitual diet for 5 days. Following this 5-day washout period, they were crossed over to the other intervention diet for 48 h. The randomisation scheme was computer generated (www.randomization.com) with blocks of 10 without stratification and with access to the scheme only provided to the United States Department of Agriculture (USDA) Children's Nutrition Research Center (CNRC) research dietitian.
(Enlarge Image)
Figure 1.
Trial flow sheet. FODMAP, fermentable oligosaccharides, disaccharides, monosaccharides and polyols; TACD, typical American childhood diet.
A stool sample was collected while the subjects were on their habitual diet (baseline period) as previously described. Briefly, a stool 'hat' was placed on the toilet to capture the stool. The stool was immediately transferred to a sterile container and placed in the subjects' freezer. The stool then was transported on ice via courier to the investigators and stored at −80 °C.
FODMAPs were defined as previously described. The low FODMAP diet contained 0.15 g/kg/day (maximum 9 g/day) of FODMAPs. The TACD contained 0.7 g/kg/day (maximum 50 g/day) of FODMAPs. Subjects were not informed that FODMAP content was being altered between the two provided diets. The meals were prepared at the USDA CNRC, and we attempted to match the two study diets with the baseline diet in terms of daily number of calories, protein, fat and type of liquid/food consumed. Foods were selected from a general checklist of acceptable options that the child and a parent agreed could be consumed. When possible the same type of food or drink with different FODMAP content was provided during both dietary interventions. For example, lactose-free vs. lactose-containing milk or diet soda (with less fructose) vs. regular soda of the same brand were provided. At times, due to child preference, foods containing fats such as meats were increased to provide extra calories in cases where the FODMAP diet precluded adding more carbohydrate containing food. Children were provided a list of low FODMAP foods to consume if they remained hungry. The prepared foods were delivered to the subjects' residence for consumption. Following each diet, all food containers were returned to the CNRC and weighed to calculate the amount of food consumed during each dietary period. The Nutrition Data System for Research (University of Minnesota) version 2012 was used to analyse the food records.
The primary outcome measure (number of pain episodes) was captured through a Pain and Stool Diary over 24-h periods for the 7 days of the baseline period and for the 2 days of each dietary intervention. Abdominal pain location, severity and duration were captured over three 8-h time periods daily as previously described. Pain severity was measured on a 0–10 Likert scale with 0 being 'no pain at all' and 10 representing the 'worst pain you can imagine.' Stools were characterised using the modified Bristol stool form chart for children. Children were subtyped based on stool form.
Secondary outcome measures included associated daily GI symptoms (abdominal discomfort, bloating, flatus, nausea, heartburn) captured using a 0–10 Likert scale with 0 being 'none' and 10 being 'the worst.' A composite GI symptom score was calculated by summing the associated daily GI symptoms in a manner consistent with other investigators evaluating dietary interventions.
A food record was kept for 3 days during the baseline period and during each day of the intervention periods as previously described. The dietary components measured are shown in Table 1.
On the second day of each dietary intervention period, subjects collected hourly breath samples for hydrogen and methane. Samples were collected for ≥8 h, up to 15 h. Breath samples were collected using a standard collection kit (Kidsampler System; Quintron Instrument Company, Milwaukee, WI, USA) and were analysed for hydrogen, methane and carbon dioxide (for normalisation) as previously described (MicroLyzer Model SC; QuintTron Instrument Company). Three children did not complete the breath testing and were not included in the gas production analyses.
DNA extraction, bacterial 16S rRNA gene amplification and 454 sequencing of 16S rRNA gene libraries were performed at the Texas Children's Microbiome Center as previously described. In brief, DNA was extracted from stool samples using a commercial DNA extraction kit (MO-BIO PowerSoil DNA Isolation Kit; MO-BIO Laboratories, Carlsbad, CA, USA) following the modified protocols described by the Human Microbiome Project. Individual sequence libraries were generated using bar-coded primers targeting the V3–V5 region of the 16S rRNA gene. Following emulsion PCR, the PCR products were pooled and sequenced on the GS-FLX platform (454 Life Sciences/Roche, Branford, CT, USA).
The pooled sequence data were quality filtered, parsed by barcode and screened for chimaeras using the Quantitative Insights Into Microbial Ecology (QIIME) software package v 1.7.0. Sequences were screened for chimaeras using the ChimeraSlayer algorithm, and all potential chimaeras were excluded from downstream analysis. Sequences with lengths shorter than 200 bp, average quality scores <20, ambiguous base calls, or mismatches to their barcode or sequencing primer also were excluded from downstream analysis. The quality filtered sequences were assigned to operational taxonomic units (OTUs; sequences that share ≥97% similarity) using a closed reference approach in QIIME with the UCLUST algorithm and Greengenes reference database (version 13_5). In cases where more than one OTU was assigned to the same organism at the species level, re-evaluation of all individual sequences belonging to those OTUs was performed using the UCLUST algorithm to confirm that the correct taxonomic attribution was made.
An average of 5570 high-quality 16S rRNA gene sequences were generated per stool sample (range: 2757–13 194). Given the variation in library size and the potential for differences in sequencing depth to bias the calculation of diversity metrics, each library was randomly subsampled to contain 2700 sequences. All results presented here are based on these subsampled libraries. Alpha diversity metrics, including OTU richness, Shannon H and the reciprocal Simpson index (1/D) were calculated using QIIME. Beta diversity metrics, including UniFrac weighted or un-weighted, were calculated in QIIME.
Phylogenetic Investigation of Communities by Reconstruction of Unobserved States (PICRUSt) was used to generate functional metagenomic predictions by linking taxonomic information from the 16S rRNA gene sequences to Kyoto Encyclopedia of Genes and Genomes (KEGG) annotations of reference genomes. Orthologues of interest were evaluated by accessing their descriptive information at http://www.genome.jp/kegg/ko.html (date accessed: February 2, 2015).
Full sequence libraries were deposited in the National Center for Biotechnology Information Sequence Read Archive (272700).
The primary outcome measure was the number of daily abdominal pain episodes during each of the dietary intervention periods vs. the baseline period. Abdominal pain scores are frequently used as primary outcome measures in childhood irritable bowel syndrome trials. Secondary outcome measures included median pain severity, the composite GI symptom score, and breath hydrogen and methane production during each of the dietary intervention periods vs. the baseline period.
For gut microbiome biomarker analysis, Responders were classified as subjects who had a ≥50% decrease in the number of daily abdominal pain episodes during the low FODMAP dietary arm of the study and not to the TACD intervention. Nonresponders were classified as subjects who lacked ≥50% decrease in the frequency of abdominal pain episodes as compared to baseline during both the low FODMAP and TACD interventions. Subjects who had ≥ 50% decrease in frequency of abdominal pain episodes to both the low FODMAP and TACD interventions as well as those who had ≥50% decrease in frequency of abdominal pain episodes with the TACD intervention alone were classified as Placebo-responders. A ≥50% decrease was chosen per recommendations for outcomes in IBS studies and in parallel with our previous work.
Linear discriminant analysis effect size (LEfSe) (http://huttenhower.sph.harvard.edu/galaxy) was used to identify differences in taxa composition and KEGG orthologues between Responders and Nonresponders. LEfSe analysis includes a nonparametric factorial Kruskal–Wallis rank sum test to detect taxa with differential abundances. Linear discriminant analysis with bootstrapping more than 30 cycles was used to determine a linear discriminant analysis score to estimate the effect size between the two groups. Alpha values of 0.05 were used with a threshold on the logarithmic score of linear discriminant analysis being ≥2.0 as published and used by others in the field.
IBM Statistics (version 22; SPSS, Armonk, NY, USA) was used for statistical evaluation. Comparisons between dietary intervention periods and between each dietary intervention period and baseline used paired Wilcoxon (continuous variables) and McNemar's testing for categorical variables. Only those completing both dietary interventions were included in the analysis. Data from the second day of each intervention was used as was convention at the time of the study. The number of abdominal pain episodes at baseline was normalised to a per day value to allow direct comparison to the test diets. Data are presented as mean ± s.d. for parametric data or median (25–75% range) for nonparametric data unless otherwise specified. P < 0.05 were considered statistically significant.
On the basis of previous work demonstrating that children with IBS had on average 0.5 ± 0.4 episodes of abdominal pain per day and identifying a 50% change in the frequency of abdominal pain as being clinically significant, we had estimated that 33 children would be needed to complete the crossover trial with an α = 0.05 and power of 90% to detect a significant difference.
Methods
Subjects
Children, aged 7–17 years, with paediatric Rome III-defined IBS determined by completion of the paediatric Rome III GI symptom questionnaire were eligible. Potential subjects were identified by screening referrals to tertiary paediatric gastroenterology care for chronic abdominal pain and via community newsletters and Internet advertisements. Parents and children were further screened by phone for inclusion and exclusion criteria and to establish current symptoms prior to enrolment as previously described. Study recruitment began in September 2011 and ended in December 2013. Baylor College of Medicine Institutional Review Board approval was obtained. The study was registered as ClinicalTrials.gov identifier NCT01339117.
General Design
Subjects completed a 7-day period in which they continued ingesting their habitual (baseline) diet (Figure 1). Following this baseline period, we employed a randomised, double-blind, crossover study design. Subjects received either a low FODMAP or typical American childhood diet (TACD) for 48 h. After 48 h on the first assigned diet, they returned to their habitual diet for 5 days. Following this 5-day washout period, they were crossed over to the other intervention diet for 48 h. The randomisation scheme was computer generated (www.randomization.com) with blocks of 10 without stratification and with access to the scheme only provided to the United States Department of Agriculture (USDA) Children's Nutrition Research Center (CNRC) research dietitian.
(Enlarge Image)
Figure 1.
Trial flow sheet. FODMAP, fermentable oligosaccharides, disaccharides, monosaccharides and polyols; TACD, typical American childhood diet.
Stool Collection
A stool sample was collected while the subjects were on their habitual diet (baseline period) as previously described. Briefly, a stool 'hat' was placed on the toilet to capture the stool. The stool was immediately transferred to a sterile container and placed in the subjects' freezer. The stool then was transported on ice via courier to the investigators and stored at −80 °C.
Diets
FODMAPs were defined as previously described. The low FODMAP diet contained 0.15 g/kg/day (maximum 9 g/day) of FODMAPs. The TACD contained 0.7 g/kg/day (maximum 50 g/day) of FODMAPs. Subjects were not informed that FODMAP content was being altered between the two provided diets. The meals were prepared at the USDA CNRC, and we attempted to match the two study diets with the baseline diet in terms of daily number of calories, protein, fat and type of liquid/food consumed. Foods were selected from a general checklist of acceptable options that the child and a parent agreed could be consumed. When possible the same type of food or drink with different FODMAP content was provided during both dietary interventions. For example, lactose-free vs. lactose-containing milk or diet soda (with less fructose) vs. regular soda of the same brand were provided. At times, due to child preference, foods containing fats such as meats were increased to provide extra calories in cases where the FODMAP diet precluded adding more carbohydrate containing food. Children were provided a list of low FODMAP foods to consume if they remained hungry. The prepared foods were delivered to the subjects' residence for consumption. Following each diet, all food containers were returned to the CNRC and weighed to calculate the amount of food consumed during each dietary period. The Nutrition Data System for Research (University of Minnesota) version 2012 was used to analyse the food records.
Measures
The primary outcome measure (number of pain episodes) was captured through a Pain and Stool Diary over 24-h periods for the 7 days of the baseline period and for the 2 days of each dietary intervention. Abdominal pain location, severity and duration were captured over three 8-h time periods daily as previously described. Pain severity was measured on a 0–10 Likert scale with 0 being 'no pain at all' and 10 representing the 'worst pain you can imagine.' Stools were characterised using the modified Bristol stool form chart for children. Children were subtyped based on stool form.
Secondary outcome measures included associated daily GI symptoms (abdominal discomfort, bloating, flatus, nausea, heartburn) captured using a 0–10 Likert scale with 0 being 'none' and 10 being 'the worst.' A composite GI symptom score was calculated by summing the associated daily GI symptoms in a manner consistent with other investigators evaluating dietary interventions.
A food record was kept for 3 days during the baseline period and during each day of the intervention periods as previously described. The dietary components measured are shown in Table 1.
On the second day of each dietary intervention period, subjects collected hourly breath samples for hydrogen and methane. Samples were collected for ≥8 h, up to 15 h. Breath samples were collected using a standard collection kit (Kidsampler System; Quintron Instrument Company, Milwaukee, WI, USA) and were analysed for hydrogen, methane and carbon dioxide (for normalisation) as previously described (MicroLyzer Model SC; QuintTron Instrument Company). Three children did not complete the breath testing and were not included in the gas production analyses.
Microbiome Composition and Metabolic Capacity
DNA extraction, bacterial 16S rRNA gene amplification and 454 sequencing of 16S rRNA gene libraries were performed at the Texas Children's Microbiome Center as previously described. In brief, DNA was extracted from stool samples using a commercial DNA extraction kit (MO-BIO PowerSoil DNA Isolation Kit; MO-BIO Laboratories, Carlsbad, CA, USA) following the modified protocols described by the Human Microbiome Project. Individual sequence libraries were generated using bar-coded primers targeting the V3–V5 region of the 16S rRNA gene. Following emulsion PCR, the PCR products were pooled and sequenced on the GS-FLX platform (454 Life Sciences/Roche, Branford, CT, USA).
The pooled sequence data were quality filtered, parsed by barcode and screened for chimaeras using the Quantitative Insights Into Microbial Ecology (QIIME) software package v 1.7.0. Sequences were screened for chimaeras using the ChimeraSlayer algorithm, and all potential chimaeras were excluded from downstream analysis. Sequences with lengths shorter than 200 bp, average quality scores <20, ambiguous base calls, or mismatches to their barcode or sequencing primer also were excluded from downstream analysis. The quality filtered sequences were assigned to operational taxonomic units (OTUs; sequences that share ≥97% similarity) using a closed reference approach in QIIME with the UCLUST algorithm and Greengenes reference database (version 13_5). In cases where more than one OTU was assigned to the same organism at the species level, re-evaluation of all individual sequences belonging to those OTUs was performed using the UCLUST algorithm to confirm that the correct taxonomic attribution was made.
An average of 5570 high-quality 16S rRNA gene sequences were generated per stool sample (range: 2757–13 194). Given the variation in library size and the potential for differences in sequencing depth to bias the calculation of diversity metrics, each library was randomly subsampled to contain 2700 sequences. All results presented here are based on these subsampled libraries. Alpha diversity metrics, including OTU richness, Shannon H and the reciprocal Simpson index (1/D) were calculated using QIIME. Beta diversity metrics, including UniFrac weighted or un-weighted, were calculated in QIIME.
Phylogenetic Investigation of Communities by Reconstruction of Unobserved States (PICRUSt) was used to generate functional metagenomic predictions by linking taxonomic information from the 16S rRNA gene sequences to Kyoto Encyclopedia of Genes and Genomes (KEGG) annotations of reference genomes. Orthologues of interest were evaluated by accessing their descriptive information at http://www.genome.jp/kegg/ko.html (date accessed: February 2, 2015).
Full sequence libraries were deposited in the National Center for Biotechnology Information Sequence Read Archive (272700).
Statistical Analyses
The primary outcome measure was the number of daily abdominal pain episodes during each of the dietary intervention periods vs. the baseline period. Abdominal pain scores are frequently used as primary outcome measures in childhood irritable bowel syndrome trials. Secondary outcome measures included median pain severity, the composite GI symptom score, and breath hydrogen and methane production during each of the dietary intervention periods vs. the baseline period.
For gut microbiome biomarker analysis, Responders were classified as subjects who had a ≥50% decrease in the number of daily abdominal pain episodes during the low FODMAP dietary arm of the study and not to the TACD intervention. Nonresponders were classified as subjects who lacked ≥50% decrease in the frequency of abdominal pain episodes as compared to baseline during both the low FODMAP and TACD interventions. Subjects who had ≥ 50% decrease in frequency of abdominal pain episodes to both the low FODMAP and TACD interventions as well as those who had ≥50% decrease in frequency of abdominal pain episodes with the TACD intervention alone were classified as Placebo-responders. A ≥50% decrease was chosen per recommendations for outcomes in IBS studies and in parallel with our previous work.
Linear discriminant analysis effect size (LEfSe) (http://huttenhower.sph.harvard.edu/galaxy) was used to identify differences in taxa composition and KEGG orthologues between Responders and Nonresponders. LEfSe analysis includes a nonparametric factorial Kruskal–Wallis rank sum test to detect taxa with differential abundances. Linear discriminant analysis with bootstrapping more than 30 cycles was used to determine a linear discriminant analysis score to estimate the effect size between the two groups. Alpha values of 0.05 were used with a threshold on the logarithmic score of linear discriminant analysis being ≥2.0 as published and used by others in the field.
IBM Statistics (version 22; SPSS, Armonk, NY, USA) was used for statistical evaluation. Comparisons between dietary intervention periods and between each dietary intervention period and baseline used paired Wilcoxon (continuous variables) and McNemar's testing for categorical variables. Only those completing both dietary interventions were included in the analysis. Data from the second day of each intervention was used as was convention at the time of the study. The number of abdominal pain episodes at baseline was normalised to a per day value to allow direct comparison to the test diets. Data are presented as mean ± s.d. for parametric data or median (25–75% range) for nonparametric data unless otherwise specified. P < 0.05 were considered statistically significant.
On the basis of previous work demonstrating that children with IBS had on average 0.5 ± 0.4 episodes of abdominal pain per day and identifying a 50% change in the frequency of abdominal pain as being clinically significant, we had estimated that 33 children would be needed to complete the crossover trial with an α = 0.05 and power of 90% to detect a significant difference.