Secondary Logo

Journal Logo


Personalized Research on Diet in Ulcerative Colitis and Crohn's Disease: A Series of N-of-1 Diet Trials

Kaplan, Heather C. MD, MSCE1,2,3,*; Opipari-Arrigan, Lisa PhD2,3,4,*; Yang, Jiabei MS5; Schmid, Christopher H. PhD5; Schuler, Christine L. MD MPH3,6; Saeed, Shehzad A. MD7; Braly, Kimberly L. RDN8; Chang, Fandi BA, BS5; Murphy, Lauren BA2; Dodds, Cassandra M. MA2; Nuding, Mason BS9; Liu, Hao BS10; Pilley, Sheri BA11,12; Stone, Julie11,13; Woodward, Gisele MSW11,14; Yokois, Nancy MD15; Goyal, Alka MD16,17; Lee, Dale MD, MSCE8; Yeh, Ann Ming MD17; Lee, Peter MD18; Gold, Benjamin D. MD14; Molle-Rios, Zarela MD19; Zwiener, R. Jeff MD20; Ali, Sabina MD21; Chavannes, Mallory MD, MHSc22,23; Linville, Tiffany MD24; Patel, Ashish MD25,26; Ayers, Travis MD27; Bassett, Mikelle MD28; Boyle, Brendan MD29; Palomo, Pablo MD30; Verstraete, Sofia MD, MAS31; Dorsey, Jill MD, MS32; Kaplan, Jess L. MD33; Steiner, Steven J. MD34; Nguyen, Kaylie MS, PNP35; Burgis, Jennifer MD31; Suskind, David L. MD8;  for the ImproveCareNow Pediatric IBD Learning Health System

Author Information
The American Journal of Gastroenterology: June 2022 - Volume 117 - Issue 6 - p 902-917
doi: 10.14309/ajg.0000000000001800



Anti-inflammatory and immunosuppressive medications are the mainstays of treatment for pediatric inflammatory bowel disease (IBD) (1); however, patient and family interest is high in dietary strategies for managing IBD symptoms and disease activity. Between 36% and 50% of pediatric patients with IBD report using complementary and alternative therapies (2). In addition, online surveys of patients with IBD indicate frequent use of dietary therapy as either primary or adjunct treatment, with most patients reporting clinical benefit (3,4). Nevertheless, efficacy studies of diet in managing IBD symptoms and disease activity remain limited.

Formula-based exclusive enteral nutrition (EEN) is the most well-studied nutritional intervention for IBD. Studies have shown it to be as effective as steroids in reducing symptoms and inducing remission as steroids among pediatric patients with Crohn's disease (CD) (5). Despite its effectiveness, EEN is challenging for patients requiring abstinence from any nonformula foods for 6–8 weeks and typically necessitating a nasogastric feeding tube (6). Consequently, patients and families seek dietary options that are not strictly formula based. The Crohn's disease exclusion diet, a whole foods diet coupled with formula as partial enteral nutrition, has been shown to be comparable to EEN in inducing remission and decreasing inflammation in children with mild to moderate CD (7). In addition, there is emerging evidence examining particular whole foods (non–formula based) diets, such as the low fermentable oligo-, di-, and mono-saccharides and polyols diet and specific carbohydrate diet (SCD) (8–10).

The SCD was developed by Dr. Sydney Hass and popularized by Elaine Gottschall in her book Breaking the Vicious Cycle (11). SCD removes grains, including wheat, barley, corn, and rice, sugars except honey, and milk products except fully fermented yogurt and focuses on fruits, vegetables, fresh meat, beans, and healthy oils. Microbiome dysbiosis with high fat, high sugar, Western diets, and altered mucosal immunity associated with food additives has been implicated in the pathogenesis of IBD (12). As a grain-free, low-sugar, nutritionally complete diet, the SCD eliminates these purported triggers for dysbiosis, mucosal disruption, and immune dysregulation and therefore could play a role in altering the IBD disease course. Although more evidence is emerging about the SCD in adults with IBD, including data from the recent DINE-CD study (13), evidence of its effectiveness in children is limited to small prospective and retrospective case series (8,10,14,15) and a small, single-center randomized controlled trial (16). Given the burden of the lifestyle changes associated with SCD, more research is also needed about more liberalized versions of the diet.

We sought to add to the body of knowledge about the role of dietary therapy in children and adolescents with IBD. Because variation in environmental exposures, microbiomes, genetics, and food metabolism contributes to heterogeneous IBD phenotypes and may influence responses to diets (12,17), we chose an N-of-1 trial study design to examine the impact of diet on IBD symptoms and inflammation. N-of-1 methodology, part of the family of Single Case Designs (18), is particularly well suited for studying heterogeneous conditions. N-of-1 trials use a randomized, multiple crossover design, in which patients compare outcomes while testing different treatments, serving as their own control. Results from N-of-1 trials are often actionable for individuals and may be aggregated for population estimates of effectiveness (19). The Personalized Research on Diet in Ulcerative Colitis and Crohn's Disease (PRODUCE) study used a series of N-of-1 trials to compare SCD with a less restrictive, modified SCD (MSCD) and to compare both with the participant's baseline, usual diet (UD), to determine dietary effectiveness in reducing symptoms and inflammation in IBD. We hypothesized that individuals would vary in their response to dietary therapy.


The study design has been previously described (20). The Cincinnati Children's Hospital Medical Center Institutional Review Board (IRB 2017-0683) approved the study. All sites relied on this IRB except one that received local IRB approval.

Setting and participants

Patients aged 7–18 years with CD, ulcerative colitis (UC), or indeterminate colitis (IC) were recruited from 19 ImproveCareNow (ICN) sites. ICN is a learning health network dedicated to improving the care of children with IBD. ICN maintains a patient registry, collects outcomes and process data, and engages patients, families, and clinicians in research and quality improvement efforts (21). Children and adolescents with CD, UC, or IC who were enrolled in the ICN registry with evidence of active inflammation, without severe disease, and without recent medication additions or changes were eligible. Enrollment was allowed for patients who were newly diagnosed and not yet receiving immunosuppressive or other medications and those who had ongoing inflammation on current therapy. Detailed inclusion/exclusion criteria are provided in Supplementary Digital Content 1 (see Supplementary Table, and in a prior publication (20).

Study design

Participants were randomized to begin with either SCD or MSCD with a centralized, stratified, block randomization approach using a 1:1 allocation ratio. Stratification occurred within sites and according to disease type (UC/IC or CD). All patients started with at least a 1-week baseline (goal of 2-week baseline) on their UD and then alternated between SCD and MSCD for 4 periods, each 8 weeks long. The multiple crossovers are a key feature of N-of-1 trials to guard against secular trends that could confound treatment effects (22). N-of-1 trials were supported by the Eureka mobile health research platform configured for the PRODUCE study (23). Participants entered and had universal access to all raw patient-reported outcome (PRO) data through the Eureka mobile app. Clinicians had continuous data access to review raw participant data and final trial results (see mobile app screenshots, Supplementary Figure, Supplementary Digital Content 3, (20).

After randomization, a trained dietitian educated participants and had visits 2 weeks into the first and second diet period, phone contact at period mid-points, and clinic visits at the conclusion of the first diet period and the study's end. Flares leading to surgery or medication change (new medication, dose adjustment, and change in dosing interval), unintentional weight loss of ≥7.5%, or pregnancy resulted in discontinuation of the N-of-1 trial.


Patients followed a UD during the baseline period and had a 2-3-day transition from their UD to the initial treatment diet. SCD was defined as in Breaking the Vicious Cycle (11). MSCD was a more liberal version of the SCD that allowed for rice, oats, potatoes, maple syrup, and cocoa with weekly minimum and maximum intake requirements. Diets are detailed in Supplementary Digital Content 4 (see Supplementary Figure, and in a prior publication (20).

Data collection and outcomes

Demographic and select clinical data were obtained from the ImproveCareNow ICN2 Database. Primary study outcomes included PROs of daily stool frequency, daily stool consistency, weekly pain interference, weekly IBD symptoms (the focus of this article), and stool fecal calprotectin (baseline and once per diet period). A single primary outcome was not defined to allow patients and providers to review all measures and focus on those most important or relevant to their clinical situation. Measure specifications were published previously (20,24,25). IBD symptoms were measured using the Pediatric IBD Symptom Scale, a 4-item scale that assesses self-reported GI Symptoms in the past 7 days including loose stools, fecal urgency, bloody stools, and abdominal pain reported as standardized T-scores (population mean of 50, SD 10) (26). Modeled on the approach used in the DINE-CD study (13) for fecal calprotectin, participants collected stool samples at home following strict instructions and then shipped them overnight, on ice, to Seattle Children's Hospital for immediate processing. Nutritional status, adherence, and adverse events (AEs) were recorded by research coordinators and/or dietitians at key study time points. Reasons for early completion or withdrawal, provider assessment of clinical and laboratory response, and poststudy treatment decisions were recorded on study completion. Diet diary information, along with participant self-report of the frequency of consuming foods not on the diet, and dietitian assessment of family behaviors such as reading food labels, were used to support the dietitian's rating of participant diet adherence. Adherence was quantified from 1 (excellent) to 6 (nonadherent) using a novel, unvalidated scale developed for this study.

Statistical analysis

A completed diet period consisted of ≥6 weeks on the diet plus stool sample submission. Full completers completed the baseline and all 4 diet periods; early completers completed the baseline and first crossover (both diets); withdrawals did not complete the first crossover. Baseline characteristics of full completers, early completers, and withdrawals were compared using analysis of variance and Fisher exact test.

Individual N-of-1 trials

PROs for individuals were analyzed with Bayesian generalized linear models, adjusted for autocorrelation, using noninformative prior distributions. Measures from week 1 of each treatment period (excluding baseline) were discarded to avoid carryover effects. Missing observations were treated as parameters in the Bayesian model. We calculated posterior median treatment difference and 95% credible intervals (CrIs) and posterior probabilities of an improved and worsened response greater than set thresholds for: (i) SCD vs UD; (ii) MSCD vs UD; and (iii) SCD vs MSCD. A mean difference of 3 points in either direction was considered clinically meaningful (27). A diet was deemed superior to another diet if the probability of a clinically meaningful improvement was >50% and the probability that symptoms were worse was < 10%; otherwise, the response was determined to be equivocal. We plotted individual fecal calprotectin measurements by period and summarized percent change between diets and whether fecal calprotectin decreased to <250 μg/g (28,29). This threshold was selected to align with the DINE-CD study (13). Among those with a baseline fecal calprotectin >250 μg/g, responders were defined as those who decreased below this threshold. In secondary analysis, we also defined response as having both a decrease to <250 μg/g and >50% reduction from baseline.

Aggregate analysis

Bayesian generalized linear mixed models with noninformative prior distributions were used to aggregate results across individuals to obtain both average treatment effects and smoothed estimates of individuals (19,30,31). Separate models were first fit for full completers, early completers, and withdrawals. The separate models used all observed values, ignoring missing values, because imputing the missing values as model parameters did not affect final results and was computationally inefficient. To fit a model for all individuals, we imputed missing values in each set separately, using their own models, and then combined the imputed data sets. This approach assumes that the data are missing at random within a completion subgroup but allows data to be missing not at random across subgroups because the imputed values differ by subgroups that differ on their outcomes. We repeated this approach to create 5 complete imputed data sets and calculated the average treatment effect across all individuals in the study comparing UD vs SCD vs MSCD by combining posterior samples from the 5 imputed data sets using Rubin rules for multiple imputation (32). Data imputation enabled the aggregate results to appropriately reflect both those who completed and those who did not complete the study. PRO results were reported as posterior median treatment difference and 95% CrI and posterior probabilities of an improved and worsened response, as for the individual analysis. Fecal calprotectin results were reported as the posterior geometric mean and the corresponding 95% CrI for each diet and as a ratio of the posterior geometric means for the diet comparisons of interest. The geometric mean ratios correspond to the percentage change in fecal calprotectin.

See Supplementary Appendix (Supplementary Digital Content 5, that includes details of all models and power/sample size calculations. In brief, a sample size of 50 participants provided a power of >90% to detect a minimally important different of a 3-point change in PROMIS measures between the SCD and MSCD.


Study population

We enrolled 54 participants in the study (April 2018–December 2019). The study CONSORT diagram is provided in Supplementary Digital Content 6 ( Data from 2 patients identified as not having met inclusion/exclusion criteria after completing their N-of-1 trial (1 with a fistula and 1 with body mass index <fifth percentile at enrollment) were included in analyses. Twenty-one (39%) participants completed the entire N-of-1 trial (full completers), 9 (17%) completed a single crossover (early completers), and 24 (44%) withdrew before completing a single crossover.

Table 1 provides participant characteristics. Median participant age was 12.5 years (range 7–18 years). Most participants had CD (74%), were White (87%), and had commercial insurance (78%). Withdrawals (38%) and early completers (33%) had UC/IC more often than full completers (10%, P = 0.07) and were more often on aminosalicylates at enrollment.

Table 1.:
Participant characteristics

Table 2 includes information on early completers and withdrawals. The most common reasons for withdrawal or early completion were lack of clinical or laboratory response (n = 11), an AE (n = 11), and not desiring to continue the diet (n = 6). Fourteen (58%) withdrawals and 6 (67%) early completers either started a new medication or had a medication change on concluding trial participation.

Table 2.:
Reasons for withdrawal or early completion

Diet comparisons

Results are presented here for IBD symptoms and fecal calprotectin. Results for stool frequency, stool consistency, and pain interference appear in Supplementary Digital Content 7 ( Individual participant results are depicted graphically in Figure 1 for the Pediatric IBD Symptom Scale and in Figure 2 for fecal calprotectin. In addition, N-of-1 trial results are summarized by participant in Table 3 (full completers), Table 4 (early completers), and Table 5 (withdrawals). Figures depicting results from the analyses aggregating N-of-1 trial results across all participants (full completers, early completers, and withdrawals, n = 54) are included in Supplementary Digital Content 8 ( for IBD symptoms and Supplementary Digital Content 9 ( for fecal calprotectin.

Figure 1.:
Probability of symptomatic improvement in the Pediatric IBD Symptom Scale for individual N-of-1 trials. Probability of improvement for full completers, early completers, and withdrawals: (a) SCD vs UD, (b) MSCD vs UD, and (c) SCD vs MSCD. Each column represents the probabilities of benefit comparing one diet to the other for each individual ordered by disease type and extent of baseline symptoms (more to less). The sum of probabilities that the first diet is better than the second (black), that the diets are no different (dark gray), and that the first diet is worse than the second (light gray) is 1.00. aChild response was used in analysis, bparticipant randomized to begin with SCD but began with MSCD. CD, Crohn's disease; IBD, inflammatory bowel disease; IC, indeterminate colitis; MSCD, modified specific carbohydrate diet; SCD, specific carbohydrate diet; UC, ulcerative colitis; UD, usual diet.
Figure 2.:
Fecal calprotectin for individual N-of-1 trials. Individual fecal calprotectin for full completers (a), early completers (b), and withdrawals (c). Fecal calprotectin results for each diet period (baseline diet, first SCD, first MSCD, second SCD, and second MSCD) are shown for each participant. Within each group of participants, individuals are ordered by disease type and baseline fecal calprotectin (higher to lower). Fecal calprotectin units are μg/g. For withdrawals, participants with only 1 fecal calprotectin measurement are not included in the figure. aVertical axis for fecal calprotectin uses a logarithmic scale, bindicates that the participant was randomized to begin with SCD but began with MSCD. CD, Crohn's disease; IC, indeterminate colitis; MSCD, modified specific carbohydrate diet; SCD, specific carbohydrate diet; UC, ulcerative colitis.
Table 3.:
Individual results for full completers
Table 4.:
Individual results for early completers
Table 5.:
Individual results for withdrawals

Individual participant results


Most participants had no clinically meaningful difference between SCD and MSCD (Figure 1c). For IBD symptoms, one diet was superior relative to the other in 5 participants (3 favoring SCD and 2 favoring MSCD) (Tables 3–5). For fecal calprotectin, 7 full completers had >70% decreases in fecal calprotectin on one diet compared with the other; 3 favored SCD, and 4 favored MSCD (Table 3). Among early completers, 1 patient had results favoring SCD, and 1 favored MSCD (Table 4). Withdrawals did not have sufficient fecal calprotectin measurements on both SCD and MSCD to enable comparisons.

SCD and MSCD vs UD

Change in IBD symptoms on SCD vs UD varied by individual (Figure 1a and Tables 3–5). Overall, 22 participants (41%) were classified as responders, with a >50% probability of clinically meaningful improvement and a <10% probability of worsened symptoms on SCD compared with UD. More full completers were responders (67%) compared with early completers (33%) and withdrawals (21%). The magnitude of the individual inflammatory response on SCD varied (Figure 2). Among the 17 full completers whose fecal calprotectin was ≥250 μg/g at baseline, 11 (65%) saw reductions in fecal calprotectin to <250 μg/g on SCD, and all but 1 also had a decrease of >50% (Figure 2 and Table 2). No early completers or withdrawals saw fecal calprotectin reductions to <250 μg/g, but 4 participants (#24, 27, 32, and 45) had >70% reductions (Figure 2 and Tables 4 and 5).

Heterogeneity was also present in the individual probabilities of symptomatic improvement on MSCD vs UD (Figure 1b, Tables 3–5). Seventeen participants (31%) were responders for IBD symptoms. More full completers were responders (52%) compared with early completers (33%) and withdrawals (13%). Three individuals (1 full completer and 2 withdrawals) had clinically meaningful symptomatic worsening on MSCD compared with UD. Inflammatory responses to MSCD varied (Figure 2). Among 17 full completers whose initial fecal calprotectin was ≥250 μg/g, 9 (53%) had reduced fecal calprotectin to <250 μg/g and a >50% decrease from baseline (Table 3). Similar to SCD, no early completers or withdrawals had reductions to <250 μg/g (Tables 4 and 5).

Aggregate analysis


As expected, based on the consistency of the results across individual N-of-1 trials, on average, there was <1% probability of a clinically meaningful difference in IBD symptoms between SCD and MSCD across the pooled, imputed sample. The average treatment difference between SCD and MSCD was −0.3 (95% CrI −1.2, 0.75) (see Supplementary Digital Content 8, Similarly, there was no significant difference in the ratio of fecal calprotectin geometric means comparing SCD and MSCD (0.77, 95% CrI 0.51, 1.10) (see Supplementary Digital Content 9,

SCD and MSCD vs UD

Aggregating the data across the entire population, on average, there was a 62% probability that SCD was better than UD for reducing IBD symptoms, a 38% probability of no difference, and a <1% probability that UD was better than SCD. The ratio of the geometric means for fecal calprotectin comparing the SCD and UD across all participants was 0.50 (95% CrI 0.32, 0.76), signifying a 50% lower average fecal calprotectin on SCD compared with baseline/UD (see Supplementary Digital Content 9, Results were similar though weaker for MSCD. On average, there was a 45% probability that MSCD was better than UD, a 55% probability of no difference, and a <1% probability that UD was better for symptomatic improvement. The ratio of the fecal calprotectin geometric means comparing the MSCD and UD across all participants was 0.66 (95% CrI 0.44, 0.96), denoting a 34% lower average fecal calprotectin on the MSCD (see Supplementary Digital Content 9,


Table 6 summarizes the individual and aggregate analysis by diet comparisons.

Table 6.:
Summary of responders from individual analysis and posterior estimates from aggregate analysis


The PRODUCE study is the largest multicenter study of the SCD and MSCD in pediatric IBD. The N-of-1 study design facilitated comparisons of the SCD and MSCD within individuals, confirmed suspected heterogeneity across individuals, and generated aggregate effectiveness estimates for the sample. The N-of-1 trials were designed to allow for a robust comparison between SCD and MSCD if implemented as intended. With few exceptions, individuals with sufficient data for an adequate comparison showed no significant differences between the 2 diets. In addition, on average (across the pooled, imputed sample), there was no significant difference in IBD symptoms or fecal calprotectin between the SCD and MSCD. This could signify that the diets are equally effective (or ineffective). It is worth noting that the study as planned was not adequately powered to detect the 0.3-point difference in symptom scores between SCD and MSCD in this study; the large number of individuals who failed to complete the study reduced the power further.

The N-of-1 trial design was not optimized for a robust comparison between the intervention diets and UD for a number of reasons. First, the exposure to UD was not randomized (it always occurred as the first period and was not repeated). Second, participants provided a limited amount of baseline data on the UD. The average assessment of the participant's baseline status when using only 1 or 2 measurements is an imprecise measure of their true baseline and makes it difficult to establish the stability of the participants' symptoms and inflammation. If these measurements were randomly higher (or lower) than the true baseline, it could lead to greater (or smaller) observed improvements on each diet. Although bias induced by regression to the mean is mitigated by study inclusion not depending on baseline symptom levels, it is possible that individual differences may simply reflect natural variation that would have occurred in the absence of dietary changes. Even with these design limitations, the study provides interesting observations. Among full completers, >50% had meaningful symptomatic improvement and reduced fecal calprotectin compared with baseline (50% and 35% reduction in fecal calprotectin on the SCD and MSCD, respectively) on the SCD and/or MSCD. Despite limitations in causally attributing these improvements to dietary therapy, families and providers felt that they had adequate information from the N-of-1 trial to allow for an opportunity for shared decision making to inform treatment decisions. In fact, most responders continued either the SCD or MSCD after trial completion. High attrition (early completion and withdrawal) indicates a large subset of patients in whom the SCD and MSCD were ineffective or partially effective (e.g., resulted in improved symptoms without reductions in inflammation) compared with UD. Almost two-thirds of these participants either started a new medication or had a medication change on concluding their participation in the trial.

Directly comparing the SCD and MSCD is inherently valuable for families and clinicians considering dietary treatment. In nearly all cases in the individual N-of-1 trials and in the aggregate direct comparison of the SCD vs MSCD, there was no clinically meaningful difference between the 2 diets. Difficulty in maintaining adherence to rigid dietary protocols over time may make the greater flexibility of the MSCD a more appealing option that facilitates greater adherence over longer periods. Most patients continuing dietary therapy on N-of-1 trial conclusion chose the MSCD. Given this finding, future studies should include MSCD as one of the dietary therapies investigated.

The PRODUCE study used fecal calprotectin to assess inflammatory response to dietary therapy because it correlates significantly with endoscopic disease activity (33). One of the major criticisms of dietary therapies is that they seem to induce clinical, but not necessarily biological, remission (34). In PRODUCE, many of the early completers saw improvements in symptoms, but none saw reductions in fecal calprotectin <250 μg/g, supporting this observation. However, among full completers with a baseline fecal calprotectin ≥250 μg/g, 65% on SCD and 53% on MSCD had a decrease to <250 μg/g. These reductions suggest that whole foods and non–formula-based diets can potentially improve symptoms and inflammation in some individuals. This must be confirmed in studies where the design allows for a more robust comparison with UD or other standard therapies. In addition, future studies would benefit from endoscopically quantifying inflammation as a more definitive study endpoint.

Our study had a number of strengths. We learned a great deal about implementing a program to support dietary therapy across 19 diverse centers, and we have developed a number of enduring resources that can support education of clinical staff, patients, and families interested in dietary therapy ( Efficacy data on SCD and MSCD are limited, and clinicians and patients need a richer body of evidence to inform treatment choices (34). Using an N-of-1 approach, we generated efficacy data and individualized probabilistic estimates of effectiveness for each diet that enabled participants to make personalized decisions about using diet to manage IBD. This type of data would not have been available in a parallel group randomized controlled trial. In addition, we confirmed the suspected heterogeneity of treatment effects using an N-of-1 design supporting the notion that the inflammatory effects of food can vary between individuals (35).

This study also has important limitations. As previously described, 1–2 weeks of baseline data did not ensure robust diet comparisons of SCD and MSCD with UD. Patient, parent, and clinician stakeholders were engaged in the design of the protocol, and this limited period was chosen to honor their requests for a short baseline period to prioritize the goal of not delaying potential intervention for children with active disease. The PRODUCE study intentionally included patients with CD, UC, and IC to learn about variation in effects of diets based on disease type with the N-of-1 approach. However, because we enrolled mainly patients with CD, aggregate analyses, where all patients were pooled regardless of disease type, were more reflective of the majority of patients with CD and may not be as generalizable to patients with UC/IC. Looking at the N-of-1 trials grouped by disease type (as done in Figure 1) does suggest that there may be differences in responses based on CD vs UC/IC disease type. Future studies could consider limiting enrollment to patients with CD. Given the nature of the study, it was not possible for participants to be blinded to the dietary interventions; the unmasked nature of the study, and the subjective nature of PROs, may lead to bias. Furthermore, in an effort to be patient-centric and allow patients and providers to determine which PROs were most important to them, we did not select a single primary outcome a priori, which may have introduced unnecessary complexity and problems with interpretation of treatment effects that varied across outcomes. In addition, extensive validation of one of the key PRO measures used in the study (Pediatric IBD Symptom Scale) has not been completed. Although the measure has clear face validity as a composite measure of IBD symptoms and other measurement properties have been evaluated (e.g., known-groups validity and responsiveness to change over time), further validation is necessary to feel confident that this instrument is measuring its intended construct. In addition, because outcomes are greatly affected by adherence to the intervention diets and our assessment of adherence was limited and used an unvalidated scale, it is possible that we overestimated dietary adherence and this may have influenced our results. Finally, the study population was primarily white and insured, which limits the generalizability of our findings. Decreasing barriers to study participation for all ethnic, racial, and socioeconomic groups is imperative.

This work provides important lessons learned that should be considered in future studies examining efficacy of dietary interventions. It is crucial to note that it required more than 18 months to recruit 54 participants from 19 centers. It is not possible to determine whether slow recruitment related to the N-of-1 trial approach itself or to the inherent difficulty of studying major dietary change. The overall trial duration (34 weeks), prolonged intervention period length (8 weeks), complexity of the protocol, need for substantial lifestyle modification, and the financial commitment to adhering to a new diet may have precluded more robust enrollment. Skeptical clinicians may have also played a role. This raises questions about the feasibility of conducting future similar studies. Procedures that make study designs more tolerable for patients, such as truncated baseline diet periods, shorter duration of randomized periods, or restrictions on the types of interventions compared (e.g., not including repeated comparisons with UD), often undermine methodologic rigor. Continuing to engage all stakeholders in study design while also optimizing methodologic rigor is critical for future studies.

We decided to use an N-of-1 trial design because we expected heterogeneity in response to dietary interventions and because we desired to use a research approach where participants could receive individual benefit from participation (obtaining individualized results). However, some key N-of-1 design features were not well adapted to this research question. Interventions that take effect and wear off quickly are ideal for N-of-1 studies. Lifestyle interventions, like dietary change, require time before changes are fully evident and may require long periods for effects to recede. The 8-week treatment period was selected based on limited preliminary data, the experience of clinicians using SCD in clinical practice, and for practical reasons (to limit overall trial length). However, it is possible that this period length was not long enough to see effects. In addition, we assumed that a 1-week washout period was sufficient to eliminate carryover effects, but this may have been insufficient and could have resulted in misattributing effects (advantageous or disadvantageous) to the wrong intervention. Future application of N-of-1 trial methodology in gastrointestinal diseases may be better suited to interventions where the cause-effect relationship is more immediate.

In conclusion, the PRODUCE study used an N-of-1 approach to test the comparative effectiveness of the SCD and MSCD in children and adolescents with IBD. Our findings do not offer definitive evidence that dietary interventions are sufficient to treat all patients or preclude the need for medications. Instead, these results indicate that SCD and MSCD may be reasonable options for some patients to consider jointly with their providers. Providers should understand there are still substantial voids in knowledge and relay this to patients and families. Determining factors that predict success with dietary interventions (e.g., disease type, disease duration, and prior exposure to biologics), evaluating mucosal healing, and examining changes in the microbiome with dietary therapy are critical next steps to identifying which patients are most likely to benefit from dietary changes. In addition, researchers should apply robust comparative effectiveness designs that include dietary interventions as one of the interventions under investigation.


Guarantor of the article: Heather C. Kaplan, MD, MSCE.

Specific author contributions: H.C.K. and L.O.-A. conceived of the study idea and led the study design; they were responsible for the overall implementation of the study; they participated in analysis and interpretation of the results, led the drafting of the manuscript, and reviewed and revised the manuscript. C.H.S. and J.Y. contributed to the study design and were responsible for the analysis and interpretation of the results; they participated in drafting and critical revisions of the manuscript. H.L. and F.C. contributed to data analysis and critical revision of the manuscript. C.L.S. participated in interpretation of the results, led the drafting of the manuscript, and reviewed and revised the manuscript. S.A.S., D.L.S., J.B., K.L.B., G.W., J.S., S.P., and K.N. contributed to the study conception and design, study implementation, analysis and interpretation of the results, and drafting and critically revising the manuscript. L.M. and C.M.D. contributed to the study implementation, analysis and interpretation of the results, and drafting and critically revising the manuscript. M.N. contributed to data collection and study implementation and participated in critically revising the manuscript. N.Y., A.G., D.L., A.M.Y., P.L., B.D.G., Z.M.-R., R.J.Z., S.A., M.C., T.L., A.P., T.A., M.B., B.B., P.P., S.V., J.D., J.L.K., and S.J.S. contributed to the study design, data collection and study implementation, interpretation of the data for the manuscript, and critically revising the manuscript.

Financial support: This research was funded through a Patient-Centered Outcomes Research Institute (PCORI) Award (PPRND-1507-31321) and a grant from the Kenneth Rainin Foundation (2019-1250). The study sponsors had no role in the study design, collection, analysis, and interpretation of the data, or in the writing of the report.

Potential competing interests: The authors have no conflicts of interest to report, except for D.L.S. Since January 2021 (after completion of the PRODUCE study), D.L.S. has served as the cofounder and chief medical officer of Nimbal Health, a digital health platform providing in-between office care for patients with IBD with a specific focus on support for those using dietary therapy.

Trial identification #NCT03301311 (

Study Highlights


  • ✓ Diet may play an important role in the pathogenesis of inflammatory bowel disease (IBD).
  • ✓ Evidence on the efficacy of the specific carbohydrate diet (SCD), and liberalized forms of this diet, in improving symptoms and inflammation in pediatric IBD is limited.


  • ✓ No clinically meaningful difference between SCD and the modified version (MSCD) was found for most individuals.
  • ✓ SCD and MSCD might improve symptoms and inflammation relative to a usual diet in some children with IBD.
  • ✓ Recruitment and retention are critical challenges in extended dietary trials in pediatric IBD.


We are grateful to the patients and families that participated in the trial. We are extremely indebted to the site coinvestigators including Sheeja Abraham, MD; Rana Ammoury, MD; Orhan Atay, MD; Julie Bass, MD; Jeffrey Blumenthal, MD; Jessica Carnathan, APRN; Stanley Cohen, MD; Laura Cooke, MD; Fernando del Rosario, MD; Dana Dykes, MD; Maureen Egan, ANP; Jennifer Ezirike, MD; David Garcia, MD; Jose Garza, MD; Matthew Giefer, MD; Arieda Gjikopulli, MD; Bhaskar Gurram, MD; Hassan Hamandi, MD; Dyer Heintz, MD; Melvin B. Heyman, MD Jay Hochman, MD; Tatyana Hofmekler, MD; Marisa Izaguirre, MD; Michael Konikoff, MD; Erika Kutsch, DO; Sameer Lapsia, MD; Larry Glen Lewis, MD; Steven Liu, MD; Phuong Luu, MD; Ross Maltz, MD; Sana Mansoor, MD; Seth Marcus, MD; William Meyers, MD; Aminu Mohammed, MD; Nancy Nelson, MD; Dinesh Patel, MD; Edith Pilzer, MD; Sujal Rangwalla, DO; Larry Saripkin, MD; Olga Sherrod, MD; Mary Smith, PA; Kelly Summers, APRN; Margaret Talmadge, MD; V. Marc Tsou, MD; and Ghassan Wahbeh, MD. We would like to acknowledge the tremendous efforts of the site dietitians in making this study possible including: Caroline Adams; Elizabeth Bobo; Amy Eschberger; Audrey Fendley; Kristen Hami; Venus Kalami; Bailey Koch; Courtney Kolski; Anne Kristine; Abigail Lundin; Cathy Malone; Aimee Molineaux; Meike Orlick; Arlecia Phillips; Jaclyn Rogers; Emily Sutherland; Heather Twible; Cathryn Williams; and Briza York. We wish to also thank the site research coordinators including Kristin Ayers; Mohammad Aziz; Joshua Bolender; Jesse Burns; Kara Cooper; Bernadette Diez; David Drevno; Laura Eshee; Ling Fan; Joanna Filopoulos; Maddie Ford; Melissa Gindville; Samantha Gomez; Gretchen Gribble; Aditi Gupta; Shannon Henry; Jennifer Hoyer McCarthy; Grace Ji; Bianca Johnson; Brendan Klein; Kimberly Klipner; Stephanie Lammers; Lynnette Lee; Megan Nowlin, Mason Nuding; Omar Oquendo-Flores; Brenna Park-Egan; Morgan Rogers; Corey Schurman; Octavia Scott; Kimberly Shelly; Emmala Ryan Shonce; Emily Stekol; Becca Trombler; Mark Tufano; Gail Waltz; Marie Washek; Robyn Yano; Lina Yossef Salameh; and Melissa Zerofsky. We are also grateful to the patient partners: Catalina Berenblum and Alex Jofriet. Finally, we appreciate the support and enthusiasm of the entire ImproveCareNow network (full list of participating sites is available at


1. Zenlea T, Peppercorn MA. Immunosuppressive therapies for inflammatory bowel disease. World J Gastroenterol 2014;20:3146–52.
2. Wong AP, Clark AL, Garnett EA, et al. Use of complementary medicine in pediatric patients with inflammatory bowel disease: Results from a multicenter survey. J Pediatr Gastroenterol Nutr 2009;48:55–60.
3. de Vries JHM, Dijkhuizen M, Tap P, et al. Patient's dietary beliefs and behaviours in inflammatory bowel disease. Dig Dis 2019;37:131–9.
4. Suskind DL, Wahbeh G, Cohen SA, et al. Patients perceive clinical benefit with the specific carbohydrate diet for inflammatory bowel disease. Dig Dis Sci 2016;61:3255–60.
5. van Rheenen PF, Aloi M, Assa A, et al. The medical management of paediatric Crohn's disease: An ECCO-ESPGHAN guideline update. J Crohns Colitis 2021;15(2):171–194.
6. Van Limbergen J, Haskett J, Griffiths AM, et al. Toward enteral nutrition for the treatment of pediatric Crohn disease in Canada: A workshop to identify barriers and enablers. Can J Gastroenterol Hepatol 2015;29:351–6.
7. Levine A, Wine E, Assa A, et al. Crohn's disease exclusion diet plus partial enteral nutrition induces sustained remission in a randomized controlled trial. Gastroenterology 2019;157:440–50.e8.
8. Obih C, Wahbeh G, Lee D, et al. Specific carbohydrate diet for pediatric inflammatory bowel disease in clinical practice within an academic IBD center. Nutrition 2016;32:418–25.
9. Prince AC, Myers CE, Joyce T, et al. Fermentable carbohydrate restriction (low FODMAP diet) in clinical practice improves functional gastrointestinal symptoms in patients with inflammatory bowel disease. Inflamm Bowel Dis 2016;22:1129–36.
10. Suskind DL, Wahbeh G, Gregory N, et al. Nutritional therapy in pediatric Crohn disease: The specific carbohydrate diet. J Pediatr Gastroenterol Nutr 2014;58:87–91.
11. Gottschall E. Breaking the Vicious Cycle: Intestinal Health through Diet. Kirkton Press: Baltimore, Ontario, Canada, 1986–2014.
12. Lee D, Albenberg L, Compher C, et al. Diet in the pathogenesis and treatment of inflammatory bowel diseases. Gastroenterology 2015;148:1087–106.
13. Lewis JD, Sandler RS, Brotherton C, et al. A randomized trial comparing the specific carbohydrate diet to a mediterranean diet in adults with Crohn's disease. Gastroenterology 2021;161:837–52.e9.
14. Burgis JC, Nguyen K, Park KT, et al. Response to strict and liberalized specific carbohydrate diet in pediatric Crohn's disease. World J Gastroenterol 2016;22:2111–7.
15. Cohen SA, Gold BD, Oliva S, et al. Clinical and mucosal improvement with specific carbohydrate diet in pediatric Crohn disease. J Pediatr Gastroenterol Nutr 2014;59:516–21.
16. Suskind DL, Lee D, Kim YM, et al. The specific carbohydrate diet and diet modification as induction therapy for pediatric Crohn's disease: A randomized diet controlled trial. Nutrients 2020;12:3749.
17. Ananthakrishnan AN. Epidemiology and risk factors for IBD. Nat Rev Gastroenterol Hepatol 2015;12:205–17.
18. Perdices M, Tate RL. Single-subject designs as a tool for evidence-based clinical practice: Are they unrecognised and undervalued? Neuropsychol Rehabil 2009;19:904–27.
19. Zucker DR, Ruthazer R, Schmid CH. Individual (N-of-1) trials can be combined to give population comparative treatment effect estimates: Methodologic considerations. J Clin Epidemiol 2010;63:1312–23.
20. Kaplan HC, Opipari-Arrigan L, Schmid CH, et al. Evaluating the comparative effectiveness of two diets in pediatric inflammatory bowel disease: A study protocol for a series of N-of-1 trials. Healthcare (Basel) 2019;7:129.
21. ImproveCareNow. Who we are. ( Accessed March 2, 2013.
22. Kravitz RL, Duan N (eds.) and the DEcIDE Methods Center N-of-1 Guidance Panel (Duan N, Eslick I, Gabler NB, et al).Design and Implementation of N-of-1 Trials: A User's Guide. AHRQ Publication No. 13(14)-EHC122-EF. Agency for Healthcare Research and Quality: Rockville, MD, 2014. ( Accessed February 15, 2022.
23. Eureka. ( Accessed April 1, 2021.
24. Thia K, Faubion WA Jr, Loftus EV Jr, et al. Short CDAI: Development and validation of a shortened and simplified Crohn's disease activity index. Inflamm Bowel Dis 2011;17:105–11.
25. Turner D, Otley AR, Mack D, et al. Development, validation, and evaluation of a pediatric ulcerative colitis activity index: A prospective multicenter study. Gastroenterology 2007;133:423–32.
26. Schuchard J, Carle AC, Kappelman MD, et al. Interpreting patient-reported outcome scores: Pediatric inflammatory bowel disease as a use case. Acad Pediatr :1–9.
27. Thissen D, Liu Y, Magnus B, et al. Estimating minimally important difference (MID) in PROMIS pediatric measures using the scale-judgment method. Qual Life Res 2016;25:13–23.
28. Lin JF, Chen JM, Zuo JH, et al. Meta-analysis: Fecal calprotectin for assessment of inflammatory bowel disease activity. Inflamm Bowel Dis 2014;20:1407–15.
29. Waugh N, Cummins E, Royle P, et al. Faecal calprotectin testing for differentiating amongst inflammatory and non-inflammatory bowel diseases: Systematic review and economic evaluation. Health Technol Assess 2013;17:xv–xix, 1–211.
30. Moeyaert M, Rindskopf D, Onghena P, et al. Multilevel modeling of single-case data: A comparison of maximum likelihood and Bayesian estimation. Psychol Methods 2017;22:760–78.
31. Zucker DR, Schmid CH, McIntosh MW, et al. Combining single patient (N-of-1) trials to estimate population treatment effects and to evaluate individual patient responses to treatment. J Clin Epidemiol 1997;50:401–10.
32. Rubin DB. Multiple Imputation for Nonresponse in Surveys. John Wiley & Sons: Hoboken, NJ, 2004.
33. D'Haens G, Ferrante M, Vermeire S, et al. Fecal calprotectin is a surrogate marker for endoscopic lesions in inflammatory bowel disease. Inflamm Bowel Dis 2012;18:2218–24.
34. Turner D, Hanauer SB. Which diet for Crohn's disease? Food for thought on the specific carbohydrate diet, mediterranean diet, and beyond. Gastroenterology 2021;161:798–800.
35. Zmora N, Suez J, Elinav E. You are what you eat: Diet, health and the gut microbiota. Nat Rev Gastroenterol Hepatol 2019;16:35–56.

Supplemental Digital Content

© 2022 by The American College of Gastroenterology