Thirty male elite cyclists aged 19–40 y volunteered to participate in the study, which was approved by the Helsinki Ethics Committee of Ziv Medical Center, Zefat, Israel (# 0075–15–Ziv). Before the study began, its purpose and objectives were carefully explained to the participants, before they signed informed consents form. Three cyclists of the experimental group dropped out during the first month of the study (see below). Participants’ characteristics (n = 27) are shown in Table 1. Inclusion criteria necessitated that all cyclists competed at an elite or category 1 level competitions and continued with their normal training routine throughout the study duration. The cyclists were not limited in their training capacity due to illnesses or any other medical condition. The participants did not consume antibiotics or probiotic supplements, medications or ergogenic supplements in the 3 months preceding the study, and throughout the study duration. The study took place during the fall and early mild Mediterranean winter weather conditions.
Study design and protocol
The study followed a randomized, double-blind, two-arm, placebo-controlled trial design (see Fig. 1). Participants were randomized (as explained below), and underwent two session of laboratory tests, before and following 90 d of probiotic / placebo intervention. The cyclists were instructed to rest and refrain from strenuous activity for at least 24 h before the scheduled laboratory tests. The participants filled an online questionnaire to assess the frequency and severity of their GI symptoms prior to, during and after training and competitions.
During their first visit, the cyclists were randomly assigned into two groups: experimental (E) group (n = 11) and control (C) group (n = 16). E group participants received a 90 d supply of probiotic supplement capsules (see below), while C group received the same number of placebo capsules identical in shape and color to the probiotic capsules. All participants were instructed to begin consuming one capsule per day on the following day. During the first visit, baseline values were also recorded. Cyclists underwent anthropometric measurements and venous blood sampling for resting inflammatory markers analysis. The participants then underwent a series of tests which included a cardio-pulmonary exercise test (CPET) during which maximal oxygen consumption (VO2max), and the ventilatory threshold (VTh) were determined. Following 3 h rest, the participants performed a time-to-fatigue (TTF) test at 85 % of their maximal power (POmax) attained during the VO2max test. All tests were performed on a constant-power cycle ergometer. Following the 90 d supplementation/placebo period, the participants reported to the laboratory for a second series of tests identical to those performed during their first laboratory visit.
To assure compliance (consumption of the supplement/placebo) weekly reminders were made to the cyclists by personal phone calls and by text and e-mail messages. Compliance was 91 %, with 3 dropouts from the E group due to discontinuation of training.
Supplemented product composition
The probiotic supplement contained about 15 billion colony forming units (CFU) of a probiotic blend consisting of 5 strains: at least (≥) 4.3 × 109 CFU Lactobacillus helveticus Lafti L10 (28.6 %), ≥4.3 × 109 CFU Bifidobacterium animalis ssp. lactis Lafti B94 (28.6 %), ≥3.9 × 109 CFU Enterococcus faecium R0026 (25.7 %), ≥2.1 × 109 CFU Bifidobacterium longum R0175 (14.3 %) and ≥0.4 × 109 CFU Bacillus subtilis R0179 (2.8 %). Bacteria viability tests were carried out by the manufacturer and the marketing company (Altman Inc. Israel) and clinical documentation from fecal samples presented binding sites were reached .The sensorially identical placebo capsules contained the excipients only (potato starch, magnesium stearate, ascorbic acid and white vegetable powder) without the bacteria, which was specially produced for this study by Lallemand Health Solutions Inc. (Montreal, Quebec, Canada) and was identified using random codes for blinding.
Body mass (weight), height and body composition were measured before each visit. The cyclists were weighed barefoot, dressed in light underwear using a Shekel model H151-8 scale (Shekel Scales Ltd., Kibbutz Beit Keshet, Israel). Body composition was assessed using Skyndex Electronic Skinfold Caliper (Caldwell, Justiss & Co., Inc., Fayetteville, AR, USA), measuring 4 skinfolds (triceps, biceps, subscapularis, iliac crest) in triplicates and the average of each skin fold was used to calculate body density  and percent body fat was calculated using the Siri equation . Body mass index (BMI) was calculated as body mass (kg) / height2 (m). All anthropometric measurements were carried out by the same researcher.
Personal and GI symptoms questionnaire
The online questionnaire was based on Peters et al.  questionnaire which was specifically chosen for its unique relevance to endurance athletes. It was administered using “Qualtrics online survey solutions”  and consisted of questions referring to socioeconomic status, training, medication, and GI symptoms. GI symptoms prevalence was evaluated during non-exercise periods (e.g., rest), training, competition, and during the first 2 h recovery from training or competition. GI symptoms were classified into UGI tract symptoms (nausea, belching, heartburn, chest pain and vomiting) and the LGI tract (cramps, bloating, diarrhea, flatulence, urge to defecate and defecation). Symptoms incidence was categorized by percentage with a slider questionnaire (0-100 %). Participants indicated the use of liquid or solid food (water, thirst quencher, energy drink, solid food, and/or a homemade product) 2 h before training or competition and during training or competition, categorized as mentioned above. The product names and types of the thirst quenchers, energy drinks and solid foods used had to be indicated, whereas the ingredients of homemade products were also listed. The cyclists were asked about using medication (both for general use, for sports-related symptoms, and for GI symptoms during exercise). The questionnaire was modified and translated from its original English version  to the cyclists’ native tongue. Test-retest evaluation of the modified questionnaire was carried out on 12 cyclists who did not participate in the study to assure its consistency with repeated completion after 10 d. Internal reliability values were tested with Pearson correlation and were 0.72, 0.85, 0.72, 0.62 and 0.62 for GI symptoms at rest, at training, at competition, after training and after competition, respectively.
All exercise tests were performed in an air-conditioned room (24 ± 2 °C, 45 ± 7 % RH) using the same constant-power cycle ergometer (Ergoline Ergometer 100, Bitz, Germany). Saddle height and handlebar reach were measured and documented for each cyclist. Cyclists used their personal pedals and riding shoes and were tested wearing respiratory apparatus and headgear. HR was monitored continuously using a Polar M400 (Polar Electro Finland Oy, Kempele, Finland) telemetry system and was averaged every 5 s during rest and exercise. VO2max and other cardio-pulmonary variable (e.g., VTh) were measured using Metalyzer 3B (Cortex Biophysik GmbH, Leipzig, Germany) metabolic cart and were determined following personalized graded exercise protocol. The flowmeter and CO2 and O2 analyzers were calibrated before each test following the manufacture calibration procedures. After a 10–15 min warmup, the test was conducted beginning at a power output (PO) of ≈ 100 W with personally selected cadence (mean ± SD 90 ± 8 RPM), PO was raised every minute by 25 W until the cyclist reached a volitional exhaustion, his RER (VCO2/VO2) values were at least 1.15, or his VO2 readings did not increase (plateaued) for 3 consecutive 20 s intervals, while PO was raised, or the cyclists asked to stop the test. Typical test duration was 8–13 min. HR was recorded every 20 s and rate of perceived exertion (RPE, scale 6 to 20) was recorded every minute. VTh was determined graphically as the point at which ventilation (VE) starts to dramatically increase despite the steady rise in PO and VO2, with the ventilatory equivalents method (VE/VCO2 to VE/VO2 proportion vs. PO) and was identified at PO in which VE/VO2 rose while VE/VCO2 was unchanged or decreased [31, 32].
Time to fatigue (TTF) test
TTF test was performed 3 h following the conclusion of the VO2max test, using the same cycle ergometer with identical individual settings. After 10 min warmup at 50 % of POmax, the cyclists rested for 2 min, thereafter, commenced the TTF test at an intensity of 85 % of POmax. The cyclists were instructed to maintain pedaling cadence at 90–100 RPM, HR was recorded every 15 s and RPE was recorded every minute throughout the test and at the conclusion of the TTF test. TTF was determined when the cyclists’ cadence was lower than 55 RPM. The cyclists neither saw the elapsed time, nor did they receive external encouragement throughout the TTF test.
Inflammatory markers analysis
Resting venous blood samples were collected into EDTA tubes (BD Vacutainer, Plymouth, UK). Then centrifuged at 1000 g for 15 min at cold (4 °C) centrifuge, the serum was then transferred by Pasteur pipet into 0.5–0.6 ml Eppendorf tubes and was frozen at -40 °C for future analyses. Analyses of C-reactive protein (CRP), IL-6 and TNFα were performed using ELISA kits in accordance with the Quantikine Colorimetric Sandwich ELISA protocol (Minneapolis, MN, USA).
The table-One R package was used to generate results presented as the means ± SD for all variables compared between the intervention groups E vs. C. The data exhibited normal distribution using boxplot analysis and the t-tests were used to compare intervention to placebo. For sensitivity analysis the comparisons were carried-out also with a non-parametric Mann-Whitney test. The non-parametric results were virtually similar to the reported t-test. Effect in each group was calculated as number of SEs below or above the mean group value (Figs. 2, 3 and 4). No corrections for multiple comparisons were done as this was an exploratory pilot study with a small number of participants.
Since the participants’ characteristics analysis revealed difference in training hours during the study period, this was corrected by analysis of covariance (ANCOVA) performed on the main outcomes calculated as delta (Δ) changes from baseline. The adjusted effect size of each dimension (performance, inflammation and GI symptoms) is presented in separate forests-plots to allow graphical evaluation of both general trend of the effects and significant group differences. All analyses were performed using the CRAN R-Project basic, tableOne and Forest plot packages. Significance level was set at p < 0.05.