This observational prospective study was performed in a single center at a pediatric hospital from April 2017 to May 2018. Children referred to the Emergency Department and/or who were admitted to the Pediatric Ward and received antibiotic treatment were consecutively enrolled. Exclusion criteria were the presence of diarrhea or vomit at recruitment, known immunodeficiency, food allergy, chronic gastrointestinal, mental or neurological diseases, treatment with acid inhibitors, or previous gastrointestinal surgery. At recruitment, we collected each patient's demographic data; site of infection; dose, duration, and kind of antibiotic; and previous AAD or other gastrointestinal symptoms reported in the last 12 months. Two weeks after the onset of antibiotic treatment, we interviewed the parents regarding the occurrence of fever or new infections and of gastrointestinal symptoms such as abdominal pain, constipation or diarrhea (considering day of onset and duration, number of stools, stool investigations, and results), and the eventual use of probiotics (specifying strain, period of intake, and indication from health care practitioners). If gastrointestinal symptoms persisted, we repeated the recall every 2 weeks until they resolved.

The primary objective of the study was to measure the incidence of AAD, abdominal pain, and constipation in inpatient children. The secondary aims were to assess the duration and possible risk factors related to antibiotic treatment and the efficacy of specific probiotic strains.

AAD was defined as the presence of 3 or more loose or liquid bowel movements per day during antibiotic treatment (early AAD [7]) or within 14 days from the antibiotics course excluding other etiologies [56]. We used the term ‘protected penicillins’ to indicate the combination of penicillins plus clavulanate or sulbactam or tazobactam molecules.

We considered antibiotic-associated abdominal pain (AAAP) as the presence of abdominal pain not reduced by defecation and not related to other recognizable conditions [16]. We also recorded fussiness and persisting crying without any other obvious causes in infants and younger children [17].

We considered antibiotic-associated constipation (AAC) the presence during antibiotic treatment and the period of recall of at least 2 of the following: difficult or painful evacuation, hard or voluminous stools, and need to use a laxative or enema [18].

The choice of different antibiotic treatment and probiotic products was not established a priori but was determined by clinicians based on the clinical presentation, experience, and attitude.

Informed consent was obtained by all parents of the enrolled children and the study was approved by our local institutional review board. This was an independent and spontaneous study in all stages of its design and conduction; data collection, management, analysis, and interpretation; and preparation, review, and approval of the paper. No pharmacological or probiotic manufacturer was ever involved in any way or at any stage.

All data were collected on an Microsoft Excel 2007 (Microsoft, Redmond, WA, USA) spreadsheet and the statistical analysis was performed by an external statistician blinded to the study design and outcomes using Medcalc v.18.6 (MedCalc Software bvba, Ostend, Belgium) and R v.3.5.0 software. The kinds of probiotics and antibiotics and their indications were divided into categories and analyzed. For probiotics, we considered ‘assumed’ if the product was taken at least 24 hours prior to the gastrointestinal manifestations [19]. We compared each strain of probiotics in terms of assumption or no assumption. To determine significant differences between two dichotomous variables, such as a risk evaluation of site of infection, antibiotic type, probiotic type, sex, and previous gastrointestinal medical history, we determined the relative risk (RR) with 95% confidence intervals (CI) and, when indicated, the number needed to treat (NNT) [20]. We analyzed mean±standard deviations and medians of age, day of onset, and duration of gastrointestinal symptoms and used a t-test to define significant variations in duration [21]. Significance was considered at values of p<0.05.

Based on the observational design of our study, a proper power size calculation was not performed. However, based on previous pediatric studies of AAD and probiotics, we arbitrarily decided to reach a similar sample size of inpatient children [711].

At the first recall, 62/289 (21.4%) children reported diarrhea; 3 were excluded because of positive rotavirus or norovirus results on stool tests during the follow-up period. Hence, AAD was found in 59/289 (20.4%) children, of whom 47 (79.7%) were admitted to the hospital and 12 (20.3%) were discharged from the A&E. The median AAD duration was 5 days (mean, 4.3±3.8 days): In 22/59 (37.3%) cases, it lasted less than 3 days; in 32/59 (54.2%) cases, it lasted 3–7 days; and in 5/59 (8.5%) cases, it lasted more than 1 week, with the longest duration being 20 days. The onset of AAD occurred after a median of 5 days (mean, 6.2±4.2) since starting antibiotic treatment; in 48/59 (81.4%) cases, the antibiotic treatment was still ongoing when the diarrhea occurred (early AAD) (Table 1). We reported a mean 5.2±1.6 bowel movements per day in children with AAD.

AAAP was reported in 27/289 (9.3%) children; 21 episodes (77.8%) occurred in children admitted to the hospital and 6 (22.2%) in children discharged from the A&E. The median AAAP duration was 4 days (mean, 5.4±3.9 days); the onset of AAAP occurred after a median 5 days (mean, 6.4±5.4 days) since starting antibiotic treatment. AAC was reported in 23/289 (8.0%) children; the median duration of AAC was 3 days (mean, 4.6±3.1 days) and the onset occurred after a median of 3 days (mean, 5.0±4.3 days) from antibiotic treatment.

None of the patients required hospitalization for the antibiotic-associated gastrointestinal symptoms.

The incidence and characteristics of the gastrointestinal manifestations by the different antibiotics are shown in Table 2.

The most frequently prescribed antibiotics were protected penicillins (in 194/289 children [67.1%]), cephalosporins (in 70/289 children [24.0%]), and macrolides (in 64/289 children [22.1%]). Fifty-four patients (18.7%) assumed 2 different antibiotic classes, and 2 children took 3 different antibiotics. The mean duration of antibiotic treatment was 10.7±6.5 days (median, 10 days). In 46/289 children (15.9%), a previous antibiotic treatment was reported within 2 weeks prior to recruitment: among these children, 23 assumed penicillins, 5 amoxicillin, 11 macrolides, and 7 cephalosporins (of which 1 was associated with penicillins, 1 with macrolides, and 1 with sulfamethoxazole-trimethoprim). In this group of patients, the antibiotic was stopped by the time of referral in 20/46 (43.5%) of cases at a mean 6.4±3.3 days before recruitment.

Of the 194 children assuming protected penicillins, 38 reported AAD (19.6%), 23 AAAP (11.9%) and 18 AAC (9.3%). Of the 70 assuming cephalosporins, 15 reported AAD (21.4%), 2 AAAP (2.9%), and 4 AAC (5.7%). Of the 64 assuming macrolides, 16 reported AAD (25.0%), 4 AAAP (6.2%), and 1 AAC (1.6%). None of these three classes of antibiotics reached a statistically significant increase in gastrointestinal symptoms or duration compared to the other classes of antibiotics.

Overall, 134/289 (46.4%) children were given intravenous antibiotics (3 were administered 2 different classes). In this group, AAD was significantly more frequent compared to oral antibiotic treatment (36/134 vs. 23/155, 26.9% vs. 14.8%; RR, 1.81; 95% CI, 1.13–2.89). In particular, 23/77 (29.9%) children administered protected penicillins intravenously reported AAD compared to 15/117 (12.8%) of those administered the equivalent oral drug (RR, 2.33; 95% CI, 1.30–4.18). None of the other classes of antibiotics reached a statistically significant difference in gastrointestinal symptom incidence or duration.

The indications for antibiotic treatment were acute infections (231 children [79.9%]) or surgical intervention (58 cases [20.1%]). Risk factors for gastrointestinal symptoms according to patient characteristics are shown in Table 3. We recognized the following as additional risk factors for AAD:

AAAP was significantly associated with:

The following were additional risk factors for AAC:

Neither sex nor family history of AAD or functional gastrointestinal disorders (FGIDs) was associated with a significant increase in gastrointestinal symptoms related to antibiotic therapy.

In our population, 234/289 (81.0%) children assumed a probiotic product to prevent or treat a gastrointestinal symptom, with a mean supplementation of 13.1±6.9 days. Probiotic intake started after a median of 2 days (mean, 3.9±3.4) since beginning the antibiotic treatment. The total numbers of children assuming the different probiotic strains (pre, during, or post-occurrence of gastrointestinal symptoms) are shown in Fig. 2. Among them, 42/234 (17.9%) assumed 2 different probiotic products with different strains. In general, 212/234 (90.6%) children reported a daily intake of the probiotic product according to the dose and method of administration shown on the package leaflet. No increased incidence of gastrointestinal symptoms was found in the 22 children who reported decreased compliance with the probiotic instructions. Among children who received probiotics, 8/234 (3.4%) were on probiotic supplementation at recruitment because of a previous antibiotic therapy; in the other 226/234 (96.6%), the probiotic product was assumed after enrollment, at the beginning of treatment, or during or even after the antibiotic treatment for concomitant gastrointestinal symptoms. Physicians recommended probiotics in 177/289 (61.2%) cases; the most indicated strains were LGG ATCC 53103 (in 110/177 children) and L. reuteri DSM 17938 (in 37/177). On the other hand, 57/289 children (19.7%) assumed probiotics without a medical prescription, and the most independently chosen probiotics were LGG (in 23 children), Bacillus clausii (in 11), Saccharomyces boulardii (in 4), and other strains (in 19).

Gastrointestinal manifestations according to the reported assumed probiotic strain are shown in Table 4. Prophylaxis with a probiotic significantly reduced AAD and AAAP but not AAC. Compared to the group of patients who did not assume probiotics, AAD decreased from 33/79 (41.8%) to 26/210 (12.4%) (RR, 0.30; 95% CI, 0.19-0.46; and NNT, 3.4; 95% CI, 2.55–5.12), with a delayed onset (from mean 4.6 to 8.1 days) (p=0.03), without any significant reduction in duration (from mean 4.9 to 3.6 days). In our population, the most effective strains for the prevention of AAD were: LGG ATCC 53103, with occurrence of AAD in 18/117 (15.4%) children (RR, 0.37; 95% CI, 0.22–0.61; and NNT, 3.8; 95% CI, 2.61–6.93); L. reuteri DSM 17938, with AAD in 6/41 (14.6%) (RR, 0.35; 95% CI, 0.16-0.77; and NNT, 3.7; 95% CI, 2.27–9.85) and different multi-strain formulations, with AAD in 1/25 (4%) children (RR, 0.10; 95% CI, 0.01–0.67; and NNT, 2.6; 95% CI, 1.74–5.57).

Compared to the group who did not assume probiotics as prophylaxis, AAAP decreased from 12/64 (18.8%) to 15/255 (6.7%) (RR, 0.36; 95% CI, 0.18–0.72; and NNT, 8.3; 95% CI, 5.00–24.25) with no significant reduction in duration (from 4.7 to 5.5 days) but with a significant delay in symptom onset (from 2.7 to 8.3 days) (p=0.04). The most efficient probiotics for the prevention of AAAP were LGG ATCC 53103, with AAAP in 9/129 children (7%) (RR, 0.37; 95% CI, 0.17–0.84; and NNT, 8.5; 95% CI, 4.78–38.60) and L. reuteri DSM 17938, with 2/45 (4.4%) AAAP (RR, 0.24; 95% CI, 0.06–1.01; and NNT, 7; 95% CI, 3.73–54.53).

Probiotic use did not significantly reduce AAC incidence (16/231 vs. 7/58 [6.9% vs. 12.1%]; RR, 0.57; 95% CI, 0.25–1.33) or duration (from 2.7 to 5.4 days) (p=0.06).

We reported a 70% lower incidence of AAD in the probiotic group similar to the 52% decrease reported in the most recent ESPGHAN recommendations (based on 21 randomized controlled trials, n=3,255) [5]. The same paper also made a strong recommendation for LGG and Saccharomyces boulardii. We can partly confirm this statement: We obtained a 63% decrease with LGG ATCC 53103 (as the most commonly used probiotic class, with 57% of total assumptions and 70% of the probiotics indicated by the physician), but we have not accumulated a sufficient sample size of Saccharomyces boulardii, which is less commonly used in our country. However, we report a similar 65% decrease in AAD for L. reuteri DSM 17938, a strain with a minor level of evidence [2627]. The best preventive result for AAD in our study (1/25, with a 90% decrease) was obtained by different products containing a mixture of at least two different probiotic strains. However, the heterogeneity of this group and the small sample size prevented a precise analysis and general conclusions about their efficacy [5]. Specific probiotic strains have previously shown a decrease in duration of diarrhea of approximately 24 hours [26]; we obtained a comparable decrease (from 4.9 to 3.6 days), but the difference was not significant. We also found a few days' delay in the onset of AAD in the group that used probiotics, especially L. reuteri DSM 17938.

We found an incidence of 9.3% for abdominal pain and 8% for constipation reported during or soon after antibiotic treatment. However, data assessing the correlation between antibiotic treatment and abdominal pain or constipation in children are very limited in the literature. A Swedish cohort study of 2700 adolescents found that broad-spectrum antibiotics, or at least 2 antibiotic cycles in the first 2 years of life, were associated with a significant increase in abdominal pain in female teenagers, reaching an incidence similar to our AAAP rate [28]. Another study suggested a consistent increase in FGIDs (including irritable bowel syndrome and functional abdominal pain) after both viral and bacterial gastrointestinal infections [29]. Previous studies showed a significant increase in FGIDs 6 months after bacterial infections (36% vs. 11% in controls) [30] but not after rotavirus infections [31]. Like infections, antibiotic treatments perturb the microbiota, leading to dysregulation of neuroimmune functions and triggering of inflammation or alterations in gastrointestinal motility and behavior [32]. An individual predisposition to an antibiotic-related effect could explain our finding of a 7-fold higher risk of post-antibiotic abdominal pain in our group of children who reported the same symptoms in the previous year. We found no significant sex-based differences in the incidence of AAAP or AAC despite a longer AAAP duration among girls (7.1 vs. 4.1 days). Different antibiotic class, administration route, and infection site did not significantly affect AAAP and AAC. Our children admitted for surgical intervention reported an increased rate of constipation as previously reported in literature [33]; this could be explained by different factors including bedding, anesthetic agents, and stress as previously suggested by other authors [1834]. We also found that children reporting constipation in the past year were more prone to experience recurrence under antibiotic treatment and have significantly more AAAP.

Probiotics as prophylaxis overall reduced the AAAP incidence of 64%, particularly L. reuteri DSM 17938 and, even more significantly, LGG ATCC 53103. Several meta-analyses and systematic reviews indicated a significant decrease in abdominal pain–related FGIDs with these probiotic strains: L. reuteri for functional abdominal pain and infantile colic and LGG for irritable bowel syndrome [2632343536]. Conversely, probiotics were not effective for constipation as already reported [183234].

The major strength of our study lies in its prospective design, relatively large group of inpatient children, standardized recall, and analysis of different possible risk and protective factors for AAD and other gastrointestinal symptoms. Moreover, this was a spontaneous real-life study with a comparative analysis of different antibiotic treatments and probiotic strains. We are aware of some limitations of our results. First, it was an open observational study with symptoms mostly based on parental reports. Furthermore, we had a heterogeneous subgroup of children and treatments with a limited sample size to gather evidences on some classes of antibiotics and probiotics.

In conclusion, AAD occurred in 1/5 of our inpatient children, with a significant increase seen in children younger than 3 years, with lower respiratory or urinary tract infections, in whom intravenous antibiotics were administered, and who reported previous AAD episodes. The incidence of abdominal pain was 9%, a value that was significantly increased in children >6 years old and in those with recurrent abdominal pain or constipation in the previous year. In 8% of children, constipation was reported with an increased risk in children having surgery or previous constipation. LGG and L. reuteri DSM 17938 proved to be an effective method of preventing AAD and abdominal pain and should be considered, particularly for at-risk children.