The study involved a randomised, double-blind, placebo-controlled parallel-groups design consisting of an initial 2-wk pre-supplement (baseline) period, 11.0 ± 0.4 wk (mean ± SD) of supplementation and a 2-wk post-supplement follow-up period. Subjects were recruited and randomly allocated using a computer generated list to either probiotic or placebo treatment based on gender, age and maximal oxygen uptake (VO2max) by a researcher independent from the study team. Participants and the study team were blinded to the interventions until the completion of the statistical analysis as the probiotic and placebo were manufactured to be identical in packaging, encapsulation and taste.
Cyclists and triathletes from Canberra, Australia and its surrounding regions were contacted via email alerts and during competition to participate in the study by the main researcher. One hundred and nine cyclists volunteered. Subjects were required to declare their use of other dietary and/or ergogenic aids that may have influenced underlying immune function and/or exercise performance. All subjects on immuno-modulatory medications were excluded. Inclusion in the study was dependent on the subjects not taking antibiotics or supplements/foods containing probiotics for at least one month prior to and during the study period. Subjects were also required to have a maximal oxygen uptake (VO2max) of at least 45 ml/kg/min for women and 50 ml/kg/min for men. Ninety-nine subjects met the study requirements.
Subjects consumed one capsule daily of either the probiotic or placebo supplement. The probiotic capsule contained a minimum of one billion (109) colony-forming units of Lactobacillus fermentum VRI-003 PCC® (Probiomics Ltd, Sydney, Australia). This dose was chosen on the basis of commercial viability and is consistent with other probiotic studies showing efficacy for URTI and GI illness . The placebo supplement consisted of microcrystalline cellulose. Subjects were able to consume the supplement at any time with or without food. Subjects returned the bottles following supplementation and capsules were counted to verify the degree of compliance. All subjects completed a four day food diary during the study that incorporated two week days and a weekend to allow adjustment for the effect of dietary food intake on microflora. Verbal and written guidelines were provided to subjects to ensure foods were recorded accurately. Detailed descriptions including brand name, packaging, method of preparation and quantity were recorded. Subjects were asked to maintain a normal diet beyond the instruction to refrain from eating probiotic-enriched yoghurt and probiotic or prebiotic enriched foods or supplements. All records were reviewed by a dietician. Total energy (kJ), carbohydrate (g), fat (g), protein (g) and fibre (g) were assessed by using FoodWorks professional edition software package (version 3.0, Xyris Software, Brisbane, Australia).
This study was conducted according to the guidelines of the Declaration of Helsinki and all procedures involving human subjects were approved by the Human Research Ethics Committees of the Australian Institute of Sport and Griffith University. All participants provided written informed consent.
Subjects recorded symptoms of GI, URT and lower respiratory illness on a daily illness log over the study period, as previously described . Briefly, symptoms of GI illness included nausea, vomiting, diarrhoea, abdominal pain, abdominal bloating, flatulence, stomach "rumbles" and loss of appetite. URTI symptoms included throat soreness, sneezing, a blocked or runny nose and cough. Lower respiratory illness symptoms included coughing with chest congestion and/or wheezing. Two or more symptoms on at least two consecutive days were defined during the analysis as an episode of illness. Symptoms separated by only one day were counted as the same episode. The severity of symptoms were self-rated as mild, moderate or severe based on the impact of the symptoms on training for that day, with mild symptoms resulting in no change to training, moderate symptoms necessitating a reduction in training volume and/or intensity, and severe symptoms leading to a total cessation of training on that day. The duration and mean severity of each episode were calculated. Subjects were also asked to record all medications consumed during the study, including antibiotics, anti-inflammatories, pain killers, decongestants and anti-histamines. Dietary supplements that might have influenced underlying immune function or exercise performance were also recorded.
Training and performance measures
Subjects recorded information on all types of physical activity undertaken during the study. For each training session subjects recorded training distance (km), duration (min) and intensity (scored on a 1-5 scale: 1, easy; 5, maximal). At the start and end of the study subjects undertook an incremental performance test to determine peak power output, VO2max and the acute post-exercise cytokine response. The test was performed on a cycle ergometer (Excalibur Sport, Lode NV Groningen, the Netherlands) as previously described .
Saliva and blood samples from all subjects, and faecal samples from a cohort of 20 subjects from each group (10 male and 10 female), were obtained pre- and post-supplementation. Saliva was collected using an eye spear (Defries Industries, Victoria, Australia). The eye spear was placed between the cheek and teeth for up to five minutes, centrifuged for 5 minutes at ~800 g and frozen at -80°C until analysis. Albumin concentration and osmolality was assessed to control for changes in salivary flow rate and hydration. All saliva samples were taken at the same time of the day to control for diurnal variation. Blood samples were taken prior to and immediately after the acute exercise challenge undertaken at the start and end of supplementation. Pre-exercise post-prandial samples were used to determine whether supplementation enhanced resting plasma cytokine concentrations. Blood samples taken immediately after the exercise challenge were used to determine if supplementation ameliorated the acute post-exercise plasma cytokine response. Blood samples were drawn directly into K3EDTA tubes (Greiner Bio-one; Frickenhausen, Germany) from an antecubital vein immediately before and after the exercise test to exhaustion. Plasma was separated by centrifugation at ~800 g for 5 min and stored frozen at -80°C until analysis. Faecal samples were collected in a sealable plastic bag and frozen immediately in a portable -20°C freezer (Waeco Pacific Pty Ltd).
Measures of mucosal immunity
Lactoferrin, lysozyme and SIgA concentrations were measured spectrophotometrically by enzyme linked immunosorbant assay (ELISA) using commercial kits (lactoferrin - EMD Chemicals, New Jersey, USA, lysozyme - Sapphire Bioscience Redfern Australia, SIgA - Salimetrics, IgA -Salimetrics, Philadelphia, USA). Albumin concentration was measured by immunoturbidimetric assay on a Hitachi 911 Chemistry Analyser (Roche). Osmolality was measured on a Model 3320 Osmometer (Advanced Instruments Inc) as per the manufacturer's instructions. The inter-assay variability for high and low controls in the lactoferrin, lysozyme and SIgA assays are included in Table three. Variability was acceptable at < 10% for the low and high positive controls.
Measures of systemic immunity
The cytokines analysed were granulocyte macrophage-colony stimulating factor, (GM-CSF), interleukin (IL)-1RA, IL-6, IL-8, IL-10, tumor necrosis factor (TNF)-α and interferon gamma (INF-γ). The concentration of plasma cytokines were measured on a Bio-Plex Suspension Array System (Bio-Rad Laboratories Pty Ltd; Hercules, CA, USA). The samples were analysed on custom manufactured Multiplex Cytokine Kits (Bio-Rad Laboratories Pty Ltd; Hercules, CA, USA). Plates were read using the Bio-Plex Suspension Array System (Bio-Rad Laboratories Pty Ltd; Hercules, CA, USA). A full blood count including white cell count and differential were analysed on a haematology analyser (Advia, GMI, MI, USA). Results from each assay were accepted if the positive controls were within two standard deviations from their established mean concentration. The CVs for the low and high controls ranged from 4.5% to 12%.
DNA was extracted according to Abell et al.  and quantified using Quant-iT™ Picogreen (Invitrogen). Microbiome diversity was examined using a universal bacteria 16S rRNA primer set (907f - 1392rgc). The amplified product was analysed by denaturing gradient gel electrophoresis (DGGE). The PCR and DGGE gel conditions followed the protocol of Abell et al.  with the exception that a 35% - 70% denaturing gradient was used. Dominant DGGE bands were extracted from the gels and sequenced for putative identification. DNA was extracted from the DGGE bands using a modification of the method described by Boom et al. . The DGGE bands were excised from the gels using a x-tracta (Geneworks, Hindmarsh, SA, AU) and transferred to a 1.7 ml tube and submerged in 100 μl of diffusion buffer (0.5M ammonium acetate; 10 mM magnesium acetate; 1 mM EDTA pH 8.0; 0.1% SDS) followed by incubation at 50°C for 30 min. The extracted DNA was then re-amplified using the primers 907f and 1392r without the GC-clamp and sequenced in both directions by capillary separation on an AB 3730xl sequencer. The sequences were then quality checked and assembled using the software packed ChromasPro version 1.41 (Technelysium Pty Ltd). The complete sequences were then putatively identified using Basic Local Alignment Search Tool (BLAST)  and GenBank. DGGE banding patterns were analysed to estimate bacterial diversity for each specimen using GelCompar II version 6.0 (Applied Maths, Inc., Texas, USA) software package and the normalised banding patterns were further analysed with Primer6 version 6.1.12 and Permanova+ addition version 1.02 (PRIMER-E Ltd, Plymouth, UK).
Quantitative real-time PCRs (QPCR) were performed in reaction volumes of 20 μl containing 1X IQ™ SYBR® Green Supermix (Bio-Rad Laboratories, Hercules, CA) and 0.2 mg/ml BSA using a Chromo-4 thermocycler (Bio-Rad Laboratories, Hercules, CA). Results were analysed with the Opticon Monitor 3 software (ver. 3.1) (Bio-Rad Laboratories, Hercules, CA). All assays were performed in duplicates with primer pair specific thermocycler programs followed by melt curve analysis to assure specificity of primers. The program for detection of the total bacterial population  consisted of an initial 4 min, 95°C hot start followed by 35 cycles of 95°C for 20 s, 60°C for 20 s and 72°C for 45 s with fluorescent acquisition after each cycle. The Lactobacillus[27, 28] and the Bifidobacterium group was detected using the following parameters, 95°C hot start for 4 min then 35 cycles of 95°C for 20 s, 58°C for 20 s, 72°C for 30 s and 80°C for 30 s followed by acquisition. The Clostridium coccoides spp.  and Bacteroides fragilis  assays had identical parameters, 95°C hot start for 4 min then 35 cycles of 95°C for 20 s, 58°C for 20 s and 72°C for 30 s followed by acquisition. The PCR parameters for the detection of Escherichia coli were as follows: 94°C hot start for 4 min followed by 35 cycles of 94°C for 30 s, 20 s at 60°C and 45 s at 72°C with fluorescent acquisition after each cycle.
Our analytical approach centred on the practical/clinical significance of probiotic supplementation rather than statistical significance alone. We made inferences about the true (large-sample) effect of the supplement on a given symptom based on the uncertainty in the effect in relation to the smallest clinically important values [33, 34]. A full description of the approach utilized is available in additional file 1.
Descriptive statistics of all measures are presented as mean ± standard deviation. The effects of supplementation on illness symptoms were more precise for the post-only analysis (rather than traditional pre-post analysis) and are reported here. All symptoms were analysed per 100 days and a proportion of each subject's symptom scores in the shoulder periods was assigned to the subject's accumulated scores to account for the start of supplementation and the monitoring period following the end of supplementation. The number of symptom episodes of a given symptom per 100 days, total number of days of the symptom per 100 days, and total load of the symptom per 100 days (sum of the product of symptom severity and number of days of the symptom per 100 days) were analysed as ratios: the mean of the probiotic group divided by the mean of the placebo group during treatment. Ratios of 1.20 and 1/1.20 ( = 0.83) were chosen as smallest clinically important differences. Effects on symptom severity and on all training-related measures were analysed as differences rather than ratios of means. Pre-specified sub-group analysis included examining the data by gender, age and training load. The smallest important effects for these measures were derived by standardisation: 0.20 of the pooled between-subject standard deviation in the control group and probiotic groups . Differences between group means of subject characteristics were assessed with a modification  of Cohen's scale  for standardised effects (small, 0.20-0.60; moderate, 0.60-1.20; large, > 1.20). Confidence limits for the effects on symptom scores and training measures were obtained with bootstrapping. These analyses were performed with the Statistical Analysis System (Version 9.1, SAS Institute, Cary, NC).
Effects of the probiotic on measures of immunology and enteric microflora (Q-PCR data) are presented for pre-post analyses. These measures were log-transformed before analysis to permit the effect of the treatment to be properly analyzed as factors or percents, and magnitudes of effects were determined by standardisation of the log-transformed variable. We used the t statistic for independent samples with unequal variances. Baseline values of the dependent variable were included as a covariate in these analyses to account for regression to the mean. We also investigated the extent to which bacterial counts accounted for symptom scores in this subsample of subjects. In these analyses the log-transformed bacterial count or the pre-post change in the log-transformed count was the covariate, and the dependent variable was rank-transformed.
A sample size of 80 subjects was required for identifying substantial changes in the incidence of illness . We assumed a rate of upper respiratory tract illness symptoms of 60% in the placebo group, with sufficient power (86% at an alpha-level of 0.05) to detect a 50% reduction in symptoms.