Efficacy of Combination Antibiotic Therapy for Refractory Pediatric Inflammatory Bowel Disease

Abstract

Background

Recent studies have shown that oral combination antibiotics may improve disease course in refractory inflammatory bowel disease (IBD). Here, we describe the use of combination oral antibiotics as salvage therapy in refractory ulcerative colitis (UC), Crohn’s colitis, and IBD-unclassified (IBD-U) at a large pediatric IBD center.

Methods

Clinical response, disease activity indices, adverse events, and clinical outcomes were measured up to 1 year after antibiotic treatment in this retrospective cohort study of children with medically refractory IBD colitis.

Results

Sixty-three patients with refractory UC, Crohn’s colitis, and IBD-U (median age [interquartile range {IQR}], 15.3 [11.2–16.5] years; median disease duration [IQR], 1.2 [0.41–4.6] years) received a combination of 3 or 4 oral antibiotics (most commonly amoxicillin, metronidazole, and either doxycycline or ciprofloxacin) for a median (IQR) of 29 (21–58) days. Thirty-four patients (54%) were deemed corticosteroid-refractory or -dependent, with the majority (62/63) having a previous or present loss of response or primary nonresponse to anti–tumor necrosis factor alpha (anti-TNFα) therapy. Use of combination antibiotics led to a significant decrease in median Pediatric Ulcerative Colitis Activity Index (PUCAI) score (IQR) from 55 (40–65) to 10 (0–40; P < 0.0001) over 3 ± 1 weeks, with 25/63 (39.7%) patients achieving clinical remission (PUCAI <10 points). The clinical benefits of oral antibiotics were independent of anti-TNFα therapy optimization. Among children entering clinical remission (n = 25), only 1 patient required surgery at 1-year follow-up, vs 10 patients in the nonresponder group. Negative predictors of response to combination antibiotics were exposure to doxycycline (odds ratio [OR], 0.25; 95% CI, 0.08–0.76) and PUCAI ≥65 at baseline (OR, 0.2; 95% CI, 0.05–0.74).

Conclusions

Oral combination antibiotics appears to be an effective rescue and steroid-sparing therapy to induce remission in the short term in patients failing a biologic.

In this retrospective study, we describe the benefits of combination oral antibiotics as “salvage” therapy for refractory pediatric IBD colitis. Treatment with a combination of oral antibiotics led to clinical response in 63.5% and clinical remission in 39.7%.

INTRODUCTION

Inflammatory bowel disease (IBD) is thought to involve an inappropriate immune response to the intestinal microbiota in genetically susceptible individuals.¹ Host genetics explain only a minority of the variance in disease risk, suggesting the importance of environmental factors such as the gut microbiota.² Over the past decade, molecular analysis of the human intestinal microbiome using culture-independent DNA sequencing methods has accelerated our understanding of the alteration in microbiota composition and function that contributes to intestinal inflammation. With these advances, dysbiosis characterized by a decrease in community richness, reduced proportions of Bifidobacteria and Firmicutes, and increased proportions of Proteobacteria and Actinobacteria have been described in IBD.³

Although the role of the gut microbiota in IBD has been increasingly recognized, our current therapeutic strategies continue to target the immune system rather than the dysbiosis. Indeed, gut microbiota–targeted therapies including antibiotics, probiotics, prebiotics, and fecal microbial transplantation have generally failed to demonstrate consistent results, reinforcing the skepticism surrounding these approaches.⁴ However, in recent years, interest in the use of broad-spectrum antibiotics as salvage therapy in refractory colitis has been growing, with a few clinical studies suggesting a potential, but still unclear, role in these evolving strategies. Specifically, 2 Japanese randomized controlled trials using a combination of oral antibiotics (amoxicillin, tetracycline, and metronidazole [ATM]) against Fusobacterium varium for 2 weeks showed improvement in clinical and endoscopic remission rates when used in ulcerative colitis (UC) patients with chronic relapsing colitis.^{5, 6} Likewise, almost half of the 15 included children (7/15) with moderate to severe refractory UC responded to a 2–3 week course of an oral broad-spectrum antibiotic “cocktail” (including metronidazole, amoxicillin, doxycycline, and—in hospitalized patients—vancomycin) in a pediatric cohort study by Turner and colleagues.⁷ More recently, another pediatric case series showed similar benefit when using the same antibiotic cocktail in children with refractory UC, reporting a 38% clinical remission rate (3/8 children).⁸ Although current adult guidelines^{9, 10} recommend the use of antibiotics only if infection is considered, or immediately before surgery, the recently published pediatric guidelines for management of acute severe colitis (ASC) specify that a short course of antibiotics may be considered in selected patients refractory to conventional therapies while preparing for colectomy.¹¹

On the basis of this preliminary evidence,⁷ we have been utilizing combination antibiotics clinically for our pediatric patients with refractory UC, IBD-unclassified (IBD-U), or Crohn’s colitis. Here, we report our experience utilizing a combination of 3 or 4 antibiotics in children with moderately to severely active IBD-colitis refractory to corticosteroids and/or failing a biologic at our institution since the year 2014. The primary aim of this retrospective study was to determine the impact of this therapy on clinical response and remission rates. The secondary aim was to describe adverse events associated with antibiotic use in this population.

METHODS

Patient Population and Study Design

We reviewed the medical records of all consecutive children aged 2–18 years with colonic IBD including colonic and ileocolonic Crohn’s disease (CD), UC, and IBD-U treated with a combination of at least 3 oral antibiotics for moderately to severely active disease, as similarly used in the previous literature,⁷ at the Children’s Hospital of Philadelphia (CHOP) between November 2014 and October 2017. Disease activity was defined by the Pediatric Ulcerative Colitis Activity Index (PUCAI) and retrospectively calculated by chart review in all, including patients with ileocolonic CD, as the primary burden of their disease was colonic at time of antibiotic initiation. Inflammatory bowel disease was confirmed according to accepted pediatric criteria, including upper and lower endoscopy with biopsies and small bowel imaging. Medical refractoriness was defined by corticosteroid resistance, as shown by no or partial response to more than 7 days of high-dose corticosteroids (≥1 mg/kg/d prednisone equivalent) or primary nonresponse or loss of response to a biologic. Exclusion criteria included concomitant infectious colitis, initiation of corticosteroid therapy within 7 days of initiating combination antibiotics, colonic disease limited to the rectum, and current ileostomy or previous total or subtotal colectomy.

The choices of antibiotic regimen and duration of therapy were made by the provider and were generally based upon the antibiotic cocktail used by Turner and colleagues.⁷ In the Turner study, children over the age of 7 years were prescribed triple therapy (orally prescribed) with amoxicillin 50 mg/kg divided by 3 (up to 500 mg three times a day), metronidazole 15 mg/kg divided by 3 (up to 250 mg three times a day), and doxycycline 4 mg/kg divided by 2 (up to 100 mg twice a day). Doxycycline was substituted for ciprofloxacin 20 mg/kg divided by 2 (up to 250 mg twice a day) in children 7 years and younger. Patients with a known allergy to 1 of the drugs were treated with oral gentamycin instead of the allergenic drug. Vancomycin could be added as the fourth medication (250 mg, or 125 mg in those younger than 8 years, four times a day) in those younger than 8 years) in hospitalized children. The antibiotic cocktail was typically prescribed for 3 ± 1 weeks; however, shorter duration could be seen in patients with lack of response or disease progression, whereas longer duration was observed when antibiotics were used as bridge therapy to a second-line biologic.

Disease activity was measured by the PUCAI based on historical information obtained from the electronic medical record. The PUCAI has been validated in pediatric UC and has been proven to perform well also in the acute colitis setting.^{12, 13} Biomarkers of disease activity, including hemoglobin, albumin, C-reactive protein (CRP), erythrocyte sedimentation rate (ESR), and fecal calprotectin, when available, were also analyzed. Normal values for albumin, CRP, and hemoglobin were determined by reference ranges based on age and sex. Data were collected at the time of initiation of antibiotics, 3 ± 1 weeks after initiation of antibiotics, at termination of antibiotics (if treatment continued beyond 3 ± 1 weeks), and 6 and 12 months after termination of antibiotics. Information regarding demographics, current and past IBD therapies, history of Clostridium difficile colitis before and during study duration, results of therapeutic drug monitoring, simultaneous change in maintenance therapy including dose or interval optimization, surgeries, and adverse events was also collected.

Primary and Secondary Outcomes

The primary outcomes were clinical response and remission rates 3 ± 1 weeks after initiation of combination antibiotics. This time point was chosen based on the paper published by Turner and colleagues,⁷ which utilized a 3-week treatment regimen. Because there was variation in the duration of treatment in our cohort, if a patient was treated with combination antibiotic therapy for more than 3 weeks, outcome data were collected 3 ± 1 weeks after initiation of combination antibiotic therapy (primary outcome) and at termination (secondary outcome). A decrease in PUCAI of at least 20 points was used to define clinical response, whereas remission was defined by a PUCAI score of less than 10. Additional secondary outcomes were steroid-free remission and prevention of surgical intervention and adverse events throughout the study period (1-year follow-up post-treatment). Steroid-free remission was defined as inactive disease and no current corticosteroid therapy or corticosteroid discontinuation for more than 3 months in steroid-dependent patients.

The last PUCAI and laboratory values recorded before treatment escalation or surgical intervention in children not responding to the antibiotic regimen were recorded for the primary time point (3 ± 1 weeks) to avoid attributing the response from treatment escalation to antibiotic treatment.

Statistical Analysis

Study data were collected from the electronic medical record and managed using Research Electronic Data Capture (REDCap). All statistical tests were performed in SAS, version 9.4. Descriptive summary data were reported as median (interquartile range [IQR]) for variables not normally distributed, unless otherwise stated. To compare outcomes before and after the intervention, paired t tests or nonparametric Wilcoxon signed-rank tests were used on the entire cohort, along with a subset of the cohort with respect to whether anti–tumor necrosis factor alpha (anti-TNFα) maintenance therapy was optimized at the same time as starting combination antibiotic therapy or not. A linear mixed effects model using an autoregressive correlation structure to account for repeated measures over time was used to observe the time-dependent effects of combination antibiotics on PUCAI and concomitant optimization of anti-TNFα maintenance therapy. Specifically, we divided patients into 2 subgroups according to whether anti-TNFα therapy was optimized during antibiotic therapy. Optimization was defined by achievement of the therapeutic through level (TL) by the time of termination in patients found to have a subtherapeutic TL at time of antibiotic initiation. Because TLs were not available for all children, also included in the optimization group were any patients with an increase in dose/kg and/or interval shortening or any patients exposed to antibiotics during the induction phase (n = 5). In the nonoptimization group, patients who were transitioned to vedolizumab were also included, as vedolizumab is associated with a longer onset of action.^14–16 This model included main effects for antibiotics and anti-TNFα therapy optimization and a factor for the interaction between antibiotics and anti-TNFα therapy optimization. Results from the linear mixed effects model included beta estimates with standard error and P values for the estimates.

Finally, univariate and multivariable logistic regression models were used to determine the association between predictive factors and achievement of clinical remission (PUCAI <10) with combination antibiotic therapy. Unadjusted logistic regression was used to obtain the corresponding odds ratio (OR) for each predictor. Multivariable logistic regression was then modeled to associate univariately significant predictors with response to therapy. The following variables were selected: IBD subtype, age at time of antibiotic initiation, disease activity as defined by PUCAI and CRP at baseline, antibiotic prescribed, exposure to corticosteroids, anti-TNFα failure, and previous history of C. difficile. Statistical significance was determined at a 2-sided α level of 0.05 (P ≤ 0.05) for all tests.

Ethical Considerations

The study protocol was approved by the Institutional Review Board at CHOP, Philadelphia, Pennsylvania.

RESULTS

Demographics and Baseline Characteristics

Eighty-six patients were treated with combination antibiotic therapy for refractory IBD colitis during the study period. Twenty-three patients were excluded: 9 because of insufficient data, 6 due to initiation of corticosteroids at the same time as the antibiotic intervention, 2 had an ileostomy, and 6 were found to have a concomitant infectious enteritis. Therefore, a total of 63 patients were included in this study.

The demographics and baseline characteristics of the 63 included patients are shown in Table 1. Twenty-seven (43%) had a diagnosis of colonic or ileocolonic CD, 23 (36.5%) had UC, and 13 (21%) were classified as IBD-U, with extensive colonic disease present in 30 (73.4%) patients with UC or IBD-U. All patients had moderately to severely active disease at initiation of antibiotics, as demonstrated by a median PUCAI (IQR) of 55 (40–65), with 21 (33%) patients being hospitalized with ASC, as defined by a PUCAI of at least 65,¹⁷ and 17 (27%) having been referred for surgical intervention (colectomy or diverting ileostomy). Thirty-four (54%) patients were already being treated with systemic corticosteroids, with 27 of these patients (79%) deemed to be corticosteroid-dependent (defined by failure to taper off corticosteroids within 3 months of starting corticosteroids, without recurrent disease, or if relapse occurred within 3 months of stopping corticosteroids¹⁷) or having shown no or partial response to more than 7 days of high-dose corticosteroids (≥1 mg/kg/d prednisone equivalent). Median continuous corticosteroid exposure duration before start of antibiotics for corticosteroid-dependent and -refractory patients (IQR) was 67 (45–131) and 11 (10–15) days, respectively. All patients in the cohort, except 1, had either failed or were losing response to anti-TNFα therapy at the time of antibiotic initiation.

TABLE 1.

Patient Characteristics at Time of Initiation of Antibiotics (n = 63)

Characteristics
Male, No. (%)	33 (52.4)
Age at start of antibiotics, median (IQR), y	15.3 (11.2–16.5)
IBD subtype, No. (%)
CD	27 (42.8)
L2 (colonic)	18 (66.7)
L3 (ileocolonic)	9 (33.3)
UC	23 (36.5)
IBD-U	13 (20.6)
E2 (left-sided colitis)	6 (16.7)
E3 (extensive colitis)	2 (5.6)
E4 (pancolitis)	28 (77.8)
Age at diagnosis, median (IQR), y	11.5 (7.7–14.6)
<6, No. (%)	9 (14.3)
6–9, No. (%)	15 (23.8)
10–13, No. (%)	20 (31.8)
14–17, No. (%)	19 (30.2)
Disease duration, median (IQR), y	1.2 (0.4–4.6)
Previously failed biological therapy, No. (%)	23 (36.5)
Type of biological therapy, No.
Anti-TNFα	23
1 anti-TNFα failed	14
≥ 2 anti-TNFα failed	9
Vedolizumab	2
Ustekinumab	1
Current medication, No. (%)
Systemic steroids	34 (54.0)
Corticosteroid dependence	27 (79.4)
Corticosteroid refractoriness	7 (20.5)
Budesonide	14 (22.2)
Azathioprine/6-mercaptopurine	4 (6.3)
In combination with a biologic	3
Methotrexate	30 (47.6)
In combination with a biologic	26
5-aminosalicylic acid (oral and/or rectal therapy)	13 (20.6)
Enteral nutrition therapy	1 (1.6)
Biological therapy	57 (90.4)
Anti-TNFα	48 (76.2)
Infliximab	39 (61.9)
Adalimumab	9 (14.3)
Vedolizumab	9 (14.3)
Othera	2 (3.2)
History of C. difficile infection, No. (%)	21 (33.3)
Hospitalization at time of initiation, No. (%)	21 (33.3)
Surgery referral, No. (%)	17 (27.0)
Duration of follow-up, median (IQR), d	244 (128–390)

^aOne patient on tofacitinib and 1 patient on canakinumab.

The various combinations of antibiotics prescribed are shown in Figure 1. The most common antibiotic regimen was amoxicillin, metronidazole, and either doxycycline or ciprofloxacin. The median duration of treatment (IQR) was 29 (21–58) days.

FIGURE 1.

Combination antibiotic regimens.

Clinical Response and Remission

Significant improvement in disease activity was demonstrated after initiating antibiotics, with a median PUCAI (IQR) of 55 (40–65) at baseline decreasing to 10 (0–40; P < 0.0001) 3 ± 1 weeks after initiating antibiotics. Forty (63.5%) patients experienced a clinical response, with a change in PUCAI of ≥20 points, and 25 (39.7%) patients entered clinical remission (PUCAI <10 points), including 6/11 who achieved corticosteroid-free remission. An additional 5 patients showing significant response at week 3 entered clinical remission the following month while using combination antibiotic therapy as a bridge to a second-line biologic (4 on vedolizumab and 1 on ustekinumab). A subset analysis including only patients with ASC (n = 26, including 23 patients hospitalized) also showed significant benefits with use of antibiotics. In this medically refractory population, in which all patients were affected by nonresponse to a biologic (24 on anti-TNFα therapy and 2 on vedolizumab) and 69% (18/26) were deemed to be corticosteroid-dependent (n = 11) or -refractory (n = 7), the median PUCAI (IQR) decreased from 65 (60–70) at baseline to 35 (10–65) at 3 ± 1 weeks (P < 0.0001). Significant response to antibiotics was observed in inpatients for whom daily PUCAI was available (n = 23) as early as day 3 (mean [range], 5 [3–13]). A total of 7 (26.9%) patients achieved clinical remission, of whom only 1 required surgery (diverting ileostomy for IBD-U) 8 months after antibiotic treatment. In comparison, 8 patients required surgery among those who failed combination antibiotics in this subgroup.

Biomarkers of Disease Activity

From baseline to 3 ± 1 weeks after initiation of therapy, significant decreases in CRP (1.1 to 0 mg/dL, P < 0.0001) and ESR (38 to 21 mm/h, P < 0.0001) were observed, as shown in Table 2. Significant improvement in hemoglobin was also observed (10.7 to 11.6 g/dL, P = 0.03); however, it should be noted that 13/63 patients received a blood transfusion between the 2 time points. No significant change in albumin was observed with antibiotic therapy.

TABLE 2.

Outcomes Over Treatment Duration (n = 63)

	Baseline	3 ± 1 wka	P
Disease activity (n = 63)
PUCAI, median (IQR)	55 (40–65)	10 (0–40)	<0.0001
Clinical response (PUCAI ∆ ≥20), No. (%)	N/A	40 (63.5)
Clinical remission (PUCAI <10), No. (%)	N/A	25 (39.7)
Steroid-free remission, No. (No. weaning)	N/A	6/34 (17)
Laboratory biomarkers, median (IQR)
Albumin (n = 41), g/dL	3.7 (3.1–4.2)	3.7 (3.4–4.2)	0.06
Hemoglobin (n = 47), g/dL	10.7 (9.8–11.8)	11.6 (10–12.5)	0.03
CRP (n = 40), mg/dL	1.1 (0–2.2)	0 (0–1.0)	<0.0001
ESR (n = 23), mm/h	38 (27–50)	21 (13–35)	<0.0001

^aThe last PUCAI and laboratory values recorded before treatment escalation or surgical intervention in children not responding to the antibiotic cocktail were carried forward through the primary time point of 3 ± 1 weeks.

Subanalysis: Patients on Anti-TNFα Therapy

The majority (48, 76.2%) of patients were on anti-TNFα therapy at the time of initiation of antibiotics. The trough level at the time of initiation (associated with the last anti-TNFα dose given before starting antibiotic therapy) was available in 37/48 (77%). Twenty-three (62.2%) patients had a therapeutic level (defined as an infliximab TL ≥5 ug/L and an adalimumab TL ≥7.5 ug/L¹⁸), whereas 14 (37.8%) had a nontherapeutic level (including 3 with neutralizing antibodies). Eleven patients (23%) did not have a TL available. Overall, 14 (29.2%) patients had their maintenance therapy optimized empirically or based on target drug monitoring, whereas 34 (70.8%) did not (Supplementary Table 1). Results of the linear mixed effects model accounting for the potential impact of maintenance therapy optimization, which could have occurred during the period of antibiotic therapy, on PUCAI change did not show a statistically significant interaction effect between these 2 interventions (P = 0.15), as shown in Figure 2.

FIGURE 2.

Response to antibiotic combination is independent of anti-TNFα optimization.

Surgical Intervention

At the time of antibiotic initiation, 17 (27%) patients had been referred to surgery for medically refractory colitis. However, only 9/17 (52.9%) required surgery by the end of follow-up (median [IQR], 244 [128–390] days), in addition to 2 other patients who had no documentation of surgery referral at the time of antibiotic initiation. Of these 11 patients, 5 underwent diverting ileostomy (4 with colonic CD and 1 with IBD-U), and 6 with UC underwent subtotal colectomy. Of the 8 patients for whom surgery was avoided, 3 were transitioned to a different class of biologic, 1 was switched from infliximab to adalimumab for secondary loss of response, 2 had their maintenance therapy optimized, and 2 remained on their primary maintenance therapy by the end of 1-year follow-up.

Medical Therapy and Surgery in Patients Entering Clinical Remission

Among the 25 patients entering clinical remission at 3 ± 1 weeks, 20 (80%) were on anti-TNFα therapy at the time of initiation of antibiotics. At 1-year follow-up, successful rescue of anti-TNFα therapy occurred in 13 (65%), whereas 6 were transitioned to another biologic class (vedolizumab or ustekinumab) and 1 discontinued therapy because of intolerance. Among patients who remained on anti-TNFα therapy, 7 were characterized by secondary loss of response (LOR) despite therapeutic TL, 3 achieved a therapeutic TL with anti-TNFα optimization by the end of the antibiotic course, and 3 had no available TL and no optimization of their primary therapy. Among patients on concomitant corticosteroid therapy entering clinical remission (n = 11), 6 (54.5%) were corticosteroid-free by termination of antibiotics, with an additional 4 patients subsequently weaned off corticosteroids at 1-year follow-up. As previously mentioned, 1 patient required surgery 8 months after antibiotic treatment. During the study period, among patients who had previously responded to a course of antibiotics (n = 40), 19 children received a second course of combination antibiotic for a disease flare, with 15 (78.9%) showing a similar positive response. Among these patients, 14 were transitioned to vedolizumab, with antibiotics being used successfully as a bridge therapy in 12 of them until full efficacy of the biologic. A second course of antibiotics was used in 5 additional patients on vedolizumab while awaiting insurance approval for escalation of maintenance therapy to every 4 weeks, showing benefit in 3 patients.

Adverse Events

No serious drug-related toxicities were observed during the study period (Supplementary Table 3). Among all patients receiving combination antibiotics, nausea, abdominal pain, and headache were the most commonly reported adverse events, occurring in 10, 5, and 4 patients, respectively. Paresthesia occurred in 1 patient, who was subsequently diagnosed with anti-TNFα-induced neuropathy. One patient developed a vaginal yeast infection during treatment. Notably, there were no reported cases of C. difficile infection during the course of antibiotics. During the follow-up period, 2 patients tested positive for C. difficile within 1 month of treatment completion while 4 additional patients were found to be positive more than 6 months after antibiotic discontinuation, of whom 2 had a history of previous C. difficile infection. It should be noted that the rate observed in our cohort (9.5%) was comparable to the incidence reported in IBD populations (5.1%–16.7%) in the current literature.^{19, 20}

Predictors of Clinical Remission

Using univariate logistic regression, PUCAI ≥65 at baseline (OR, 0.2; 95% confidence interval [CI], 0.05–0.74; P = 0.016) and exposure to doxycycline (OR, 0.25; 95% CI, 0.08–0.76; P = 0.015) were found to decrease the odds of clinical remission with combination antibiotic therapy. In a multivariate logistic regression, both factors remained significant (P = 0.352) (Supplementary Table 4).

DISCUSSION

In this single-center retrospective study, we describe the benefits of combination oral antibiotics as “salvage” therapy for refractory colitis, consistent with the findings previously published in 2 smaller pediatric case series.^{7, 8} Our cohort consisted of patients with either UC, IBD-U, or colonic or ileocolonic CD and previous or current LOR to anti-TNFα therapy at the time of antibiotic initiation. In this medically refractory population, use of a combination of 3 or 4 oral antibiotics led to clinical response in 63.5% and clinical remission in 39.7% without frequent side effects. In addition to improvement in clinical response and remission rates, we showed independent improvements in biomarkers of disease. Furthermore, in terms of long-term outcomes, children entering complete remission were free of surgery, with the exception of 1 patient with IBD-U, who underwent diverting ileostomy several months after combination antibiotic treatment.

Our cohort included children with severely refractory disease already losing response to potent immunosuppressive therapy, yet our results have shown that combination antibiotics may serve as an effective rescue therapy in this population. Indeed, while one-third of patients entering remission required transition to a second-line biologic, response to primary therapy was successfully recaptured and maintained in the majority, as shown in long-term follow-up. Importantly, our study showed that combination antibiotic use provided a benefit independent of concomitant anti-TNFα optimization and therefore supports the concept of microbial modulation as an alternative or adjunctive approach to immunosuppressive therapy in reducing intestinal inflammation.

It is noteworthy that a significant number of patients in our study were corticosteroid-dependent or -refractory, and that combination antibiotic therapy was not only effective in these children but also facilitated tapering and withdrawal of corticosteroids. Combination antibiotic therapy therefore appears to be a valuable steroid-sparing option for short-term control of active disease in refractory pediatric colitis. As such, we have also demonstrated that this therapy could be positioned as an effective bridging strategy while transitioning patients to medications that have delayed onset of demonstrable efficacy, such as vedolizumab. Additionally, a significant number of children who responded to antibiotic therapy showed a similar positive response to a second course of treatment, suggesting again that dysbiosis may play a causative role in perpetuating inflammation in a population subset. Considering microbial modulatory interventions involving diet, antibiotics, and pre- or probiotics over or in combination with immunosuppressive therapy may benefit this group of patients.

Multiple possible etiologies of efficacy of combination antibiotic therapy have been suggested in the literature, including targeting an unknown but specific pathogen in IBD, preventing immune system stimulation by translocation of bacteria, altering the composition and biomass of the luminal microbiome, and providing an antibiotic(s) with anti-inflammatory properties. It is interesting that previous clinical trials using narrow- as opposed to broad-spectrum antibiotics in UC failed to show beneficial effects, suggesting that a therapeutic strategy capable of more drastically reducing bacterial load may show greater treatment efficacy.^{21, 22}

Interestingly, in our study, doxycycline exposure (OR, 0.25; 95% CI, 0.08–0.76) and severe disease activity (PUCAI ≥ 65) at baseline (OR, 0.2; 95% CI, 0.05–0.74) were found to be independently associated with a decrease in the likelihood of achieving remission. The selection of antibiotics in our cohort was based on the anti–F. varium protocol modified by Turner et al.⁷ The original cocktail used in the adult Japanese randomized controlled trial (amoxicillin, metronidazole, and tetracycline) was based on previous observations showing a high prevalence of F. varium in mucosal biopsies from UC patients,²³ which were replicated in animal models.⁶ However, the lack of significant reduction of F. varium in mucosal biopsies when compared with placebo after antimicrobial treatment has raised questions regarding the rationale behind the selection of this antibiotic cocktail.²⁴ Therefore, it may be that other antibiotic regimens are more effective than the ones proposed here or that other pathogenic organisms play a critical role, particularly in patients with ASC, for whom, although significant benefits were shown as early as day 3 in our study, a lower rate of clinical remission was found (26.9% vs 39.7%). Certainly, further studies characterizing the microbiota associated with this highly refractory population are needed to develop better microbial-based strategies and select the best candidates for this approach. Until then, given the favorable risk–benefit profile, we believe that a short course of combination antibiotics may reasonably be considered in patients with ASC, some of whom have no remaining options aside from colectomy.

The limitations of this study include the retrospective and observational study design without a comparative control group. However, the size of our cohort likely balances some of these limitations and allows for improved generalization and clinical application of this intervention. Additionally, the predominance of anti-TNFα failure in this refractory patient population represents a group of children with IBD who are in need of novel medical therapies.

CONCLUSIONS

In conclusion, we showed that the use of combination antibiotics is an effective and relatively fast-acting therapeutic option in pediatric patients with moderately to severely refractory IBD-colitis, including UC, IBD-U, and Crohn’s colitis. Taken together, we have demonstrated that this intervention may be positioned not only as an effective rescue and steroid-sparing therapy in patients failing biologics but also as a successful bridge to alternative therapy with longer onset of action. Although we found a lower rate of response in children with ASC, we believe that, in agreement with the recently published guidelines for management of pediatric UC,¹¹ a short course of combination antibiotics may reasonably be considered as salvage therapy in this refractory population left with limited therapeutic options. Clinical response should be assessed frequently and therapy discontinued if no improvement is documented within 1 week. Further studies leading to an understanding of the changes in the composition and functions of the gut microbiome in responders and nonresponders to combination antibiotic therapy are needed to develop better antimicrobial-based strategies.

Author contributions: No writing assistance was provided, and all views expressed in this manuscript are those of the authors. The authors jointly drafted and edited the manuscript.

Supported by: This work was supported in part by National Institutes of Health grant K23 DK109136.

Conflicts of interest: L. A. received research funding from Seres Therapeutics. R. N. B. has served as a consultant for AbbVie, Celgene, Janssen, Pfizer and Eli Lilly. J. B., A. K., N. H., P. M., A. B. G. and J. R. K. as well as Rachel Rogers report no potential conflicts of interest.