In commercially relevant, non-physically segregated multi-strain probiotic dosage forms, which biological interactions, gastrointestinal delivery conditions, and formulation/manufacturing attributes measurably reduce strain-level viability, post-exposure recovery, colonization capacity, or the evidentiary basis for clinical efficacy?
Moisture ingress above critical water-activity thresholds, strain-specific heat and mechanical damage during drying and compression, and unprotected gastric acid exposure each independently reduce viable cell delivery by orders of magnitude, while competitive displacement between co-formulated strains and the systematic failure to resolve strain-level outcomes within non-segregated commercial dosage forms together undermine the evidentiary basis for attributing clinical efficacy to specific strains in multi-strain probiotic products.
Abstract
Moisture accumulation is the single most consistently documented driver of storage-phase viability loss in dry and oil-suspended probiotic formats: water activity rising above approximately 0.15 in freeze-dried capsules [1] and above 11.4% relative vapour pressure in freeze-dried powders [2] triggers matrix crystallization, glass-transition depression, and irreversible cell damage, effects that passive packaging alone cannot prevent at 25 °C/60% RH without active desiccation [3]. Manufacturing processes impose further strain-differential costs—spray-drying, fluid-bed drying, and tablet compression each reduce viability by amounts that vary severalfold between strains sharing a species designation [4–6] —and post-manufacturing stability inversely tracks intrinsic strain heat sensitivity rather than species identity [7, 8]. Gastrointestinal delivery is a second decisive failure point: unprotected free cells lose 5–8 logs during simulated gastric passage [9, 10], and only enteric-coated or delayed-release dosage forms consistently deliver therapeutically relevant cell numbers to the small intestine, with a commercial delayed-release capsule achieving 51–56% ileal recovery versus less than 1% for liquid, powder, and standard-capsule formats under identical fasted conditions [11]. Within genuinely non-segregated multi-strain dosage forms, Bifidobacterium species consistently show lower GI survival than co-formulated Lactobacillus species [11, 12], and controlled serial-transfer experiments demonstrate that same-species strain pairs can competitively displace one another to non-detection within a single growth cycle through mechanisms including bacteriocin production and differential acid output [13]. Despite these findings, the corpus contains a fundamental evidentiary gap: almost no study simultaneously resolves strain-level viability within a non-segregated multi-strain commercial dosage form across manufacture, storage, and GI transit while providing colonization or clinical outcome data linked to specific named strains. Efficacy claims routinely attributed at species or genus level cannot be transferred to products containing different strains of the same species, given documented three- to sevenfold differences in GI survival, adhesion, and antimicrobial activity between named strains sharing a species designation [14–16].
Flow Diagram
The flow diagram image has been omitted as per instructions.
Paper search
We performed a semantic search across over 138 million academic papers from the Elicit search engine, which includes all of Semantic Scholar and OpenAlex.
We ran these queries:
- "probiotic dietary supplement formulation stability viability moisture water activity excipient storage"
- "probiotic simulated gastric fluid bile salt survival enteric coating capsule microencapsulation"
- "multistrain probiotic co-culture antagonism compatibility bacteriocin competition strain-specific efficacy"
The searches returned 600 total results from Elicit.
We retrieved 591 papers most relevant to the query for screening.
Screening
Abstract screening
We screened in sources based on their abstracts that met these criteria:
- Primary empirical report: Is this a completed experimental, observational quality-testing, animal/human colonization, or comparative clinical study—not a review, protocol, editorial, patent, or conference-only record?
- Live bacterial probiotic material: Does the title or abstract study viable bacterial probiotic strains or a product containing them?
- Measurable failure-relevant endpoint: Does it report or explicitly assess viable counts, survival/recovery, stability, antagonism/compatibility, gastrointestinal challenge, colonization, or clinical efficacy?
- Relevant exposure or comparison: Does it evaluate multi-strain composition, strain identity, acid/bile delivery, storage/moisture/processing, encapsulation/coating, packaging, or co-formulated ingredients?
- Multi-strain Formulation: Does this study involve two or more live bacterial probiotic strains in the same dosage form without physical segregation (i.e., not separated by enteric coatings, different capsules, or other physical barriers)?
- Commercial Relevance: Does this study involve commercial dietary supplements, commercial-product testing, or manufacturable prototypes relevant to capsules, tablets, sachets, powders, oil suspensions, or food systems with transferable data to dry-matrix formulations?
- Strain-level Outcomes: Does this study report strain-resolved CFU counts, log10 CFU/g changes, survival percentages in simulated GI conditions, storage decay rates, colonization outcomes, or clinical efficacy differences at the individual strain level?
- Strain Identification: Does this study provide precise strain identification (exact strain designations) that would enable strain-equivalence analysis?
- Appropriate Comparators: Does this study include appropriate comparators such as single-strain controls, separately incubated controls, baseline measurements, protected vs. unprotected delivery comparisons, or alternative named strains of the same species?
- Eligible Study Design: Is this study a primary research study with an eligible design (in-vitro co-culture, formulation study, stability study, simulated gastric/intestinal fluid study, GI-model study, commercial-product quality study, animal/human colonization study, or randomized comparative clinical trial), rather than a narrative review, protocol, editorial, conference abstract, or patent?
- Measured Compatibility: Does this study directly measure strain-level viability, survival, or interaction parameters rather than only inferring multi-strain failure without actual measurements?
- Bacterial Probiotics: Are live bacterial probiotic strains the primary intervention in this study (rather than yeasts, fungi, or other non-bacterial organisms as the main focus)?
Papers that failed any criterion were automatically excluded; all other papers, including borderline ones, were screened in.
230 papers passed abstract screening and moved to full-text screening.
At abstract screening, the number of papers excluded for each primary reason was:
- Primary empirical report: n = 32
- Live bacterial probiotic material: n = 1
- Multi-strain Formulation: n = 256
- Commercial Relevance: n = 1
- Eligible Study Design: n = 54
- Bacterial Probiotics: n = 17
Full-text screening
We then screened papers based on their full text using these additional criteria:
- Oral-formulation or commercial transferability: Is the material a commercial probiotic dietary supplement, or a dry/oral formulation prototype with transferable capsule, tablet, powder, sachet, oil-suspension, packaging, or GI-delivery evidence? Food matrices qualify only if their result is clearly transferable and can be labelled indirect.
- Vector-specific measurable result: Does the full text contain extractable quantitative or clearly measured evidence for at least one target vector: multi-strain compatibility, named-strain non-equivalence, gastric/bile survival or delivery protection, or moisture/excipient/manufacturing stability?
- Sufficient experimental context: Does the paper report enough context to interpret the result, including organism identity/formulation and relevant challenge, storage, or comparator conditions?
- Strain identity for strain-equivalence claims: For papers addressing clinical or biological non-equivalence between strains, are exact strain designations reported? Papers not addressing that vector may pass without this information.
Papers were included in the final analysis only if they met every criterion; papers that failed or were borderline on any criterion were excluded.
60 papers passed full-text screening and moved to data extraction.
At full-text screening, the number of papers excluded for each primary reason was:
- Oral-formulation or commercial transferability: n = 11
- Vector-specific measurable result: n = 1
- Sufficient experimental context: n = 1
- Strain identity for strain-equivalence claims: n = 1
- Full text not available: n = 156
Data extraction
We asked a large language model to extract each data column below from each paper. We gave the model the extraction instructions shown below for each column.
- Full citation:
Provide authors, article title, journal, year, and DOI. If DOI absent, provide another stable identifier.
- Evidence vector:
Classify all applicable vectors: multi-strain antagonism/competitive exclusion; strain non-equivalence; gastric acid/bile destruction and delivery protection; moisture/water activity/excipient/manufacturing stability.
- Study type and model:
State whether commercial-product quality study, in-vitro co-culture, formulation/stability study, SGF/SIF GI model, animal/human colonization study, or comparative clinical trial; give model details.
- Product and dosage-form context:
Extract whether commercial product or prototype; dosage form; number of strains; whether physical segregation is stated; matrix/excipients; packaging; and storage conditions.
- Exact strains evaluated:
List exact genus/species/subspecies and all alphanumeric strain designations. Clearly flag organisms reported only at species level.
- Comparators:
Extract comparator(s): single strain, separately incubated mix, baseline/pre-storage, protected/unprotected delivery, lower/higher water activity, compatible/incompatible excipient, or alternative named strain.
- Baseline viability:
Extract baseline viable count and units (CFU/g, CFU/dose or other) for each relevant arm. State not reported where absent.
- Quantitative storage survival:
Extract final viable count, log10 reduction, percent survival, storage time, temperature, moisture or water activity, and any decay rate for each relevant arm.
- Gastric and intestinal challenge conditions:
Extract simulated gastric fluid pH, exposure duration, simulated intestinal fluid or bile salt concentration and duration, and enumeration/recovery method.
- Quantitative GI survival:
Extract CFU/log10 survival or percent survival after gastric and bile/intestinal exposure for each arm, preserving the condition and timepoint.
- Co-culture compatibility result:
Extract direct evidence of inhibition, bacteriocin/organic-acid effect, nutrient competition, displacement, or no antagonism, including quantitative viability/zone/growth/colonization outcome if reported.
- Strain-specific clinical or biological non-equivalence:
Extract named strain comparisons within a species, indication/model, outcome, and quantitative result. Do not infer equivalence from species identity.
- Moisture, excipient and active-ingredient mechanism:
Extract water activity or moisture threshold, hygroscopicity/migration evidence, excipient or co-active ingredient, and proposed/direct mechanism such as osmotic stress, oxidation, acid lysis, or cell-wall damage.
- Failure mechanism:
State the authors' identified or directly supported failure mechanism. Distinguish directly measured mechanism from plausible interpretation.
- Evidence applicability label:
Classify as Direct commercial evidence; Formulation-transferable prototype evidence; or Biologically informative but indirect evidence, with a brief justification.
- Key limitations:
Extract limitations relevant to transferability, strain identification, assay method, comparability, or inference.
- Strain Details:
Extract exact strain designations and multi-strain formulation details for all probiotic strains studied in non-physically segregated combinations, including:
- Complete strain names with culture collection numbers/commercial designations
- Number of strains in combination (must be ≥2)
- Strain ratios or CFU proportions if reported
- Whether strains are same-species or cross-species combinations
- Any strain-specific tagging or identification methods used
- Biological Interactions:
Extract evidence of biological interactions between co-formulated probiotic strains that affect viability, recovery, colonization or efficacy, including:
- Direct antagonism, competition, or bacteriocin production between strains
- Differential survival rates between strains in the same formulation
- Strain-specific pH, metabolite, or growth factor effects
- Cross-protection or synergistic effects between strains
- Any mechanistic explanations for strain-strain interactions
- Comparison to single-strain controls where available
- Formulation Factors:
Extract formulation and manufacturing attributes tested for their effects on multi-strain probiotic stability and viability, including:
- Excipients, protective agents, and matrix materials (specific names and concentrations)
- Moisture content, water activity levels, and packaging systems
- Processing methods (freeze-drying, spray-drying, heat treatment, compression)
- Storage conditions (temperature, humidity, duration)
- Dosage form type (capsule, tablet, sachet, powder, oil suspension)
- Any protective technologies or delivery systems used
- GI Delivery:
Extract gastrointestinal delivery conditions and protection methods tested for effects on strain-level survival and functionality, including:
- Simulated gastric fluid (SGF) conditions (pH, pepsin concentration, exposure time)
- Simulated intestinal/bile conditions (pH, bile salt type and concentration, exposure time)
- Enteric coating or other gastroprotection methods
- Sequential GI challenge protocols
- Comparison between protected vs unprotected delivery
- Any in vivo GI transit or colonization studies
- Strain-Level Outcomes:
Extract strain-resolved outcome measurements for viability, recovery, colonization capacity, and clinical efficacy in multi-strain formulations, including:
- Strain-specific CFU counts or log10 CFU changes (not just total counts)
- Percent survival rates for individual strains through processing, storage, or GI challenge
- Post-exposure recovery or stress resistance for each strain
- Colonization data (fecal recovery, persistence, abundance) by strain
- Clinical efficacy outcomes attributed to specific strains vs. strain combinations
- Time points and measurement methods used
- Detrimental Effects:
Extract specific factors that measurably reduced strain-level viability, post-exposure recovery, colonization capacity, or clinical efficacy in multi-strain probiotic formulations, including:
- Which strains were most susceptible to particular stressors
- Quantitative reduction in performance (fold-change, log reduction, percent decrease)
- Threshold conditions where strain performance failed
- Incompatible strain combinations or formulation components
- Storage or processing conditions that disproportionately affected certain strains
- Evidence linking reduced strain performance to clinical outcome differences
- Experimental Context:
Extract experimental design and contextual factors necessary for interpreting multi-strain probiotic compatibility and performance data, including:
- Study design type (in vitro, animal model, human clinical trial)
- Control groups used (single-strain, baseline, alternative formulations)
- Sample sizes and replication details
- Analytical methods for strain identification and quantification
- Commercial product vs. laboratory prototype context
- Limitations or confounding factors that affect interpretation of strain-specific effects
Results
Characteristics of Included Studies
The 60 sources span formulation/stability prototypes, commercial-product quality assessments, in-vitro gastrointestinal (GI) challenge models, in-vitro co-culture studies, and a small number of animal or human delivery experiments. Full texts were retrieved for all sources. Studies were classified according to the four evidence vectors specified in the protocol: (1) multi-strain antagonism or competitive exclusion, (2) strain non-equivalence, (3) gastric acid/bile destruction and delivery protection, and (4) moisture/water activity, excipient, and manufacturing stability.
| Study | Full text retrieved? | Study type | Primary vector(s) | Strain count | Dosage form / context | Evidence applicability |
|---|---|---|---|---|---|---|
| Zárate & Nader-Macías (2006) [17] | Yes | Formulation/stability prototype; in-vitro epithelial adhesion [17] | Moisture/excipient/manufacturing stability; strain non-equivalence [17] | 3 [17] | Freeze-dried gelatin capsules, 5 °C [17] | Formulation-transferable prototype [17] |
| Savini et al. (2010) [18] | Yes | Formulation/stability prototype [18] | Moisture/excipient/manufacturing stability [18] | 2 [18] | Freeze-dried powder, vacuum sachets [18] | Direct commercial evidence (pilot scale) [18] |
| Sielatycka et al. (2023) [1] | Yes | Commercial-product 24-month stability study [1] | Moisture/water activity; strain non-equivalence (aggregate) [1] | 1, 2, 4 [1] | Capsules in pharmaceutical blisters [1] | Direct commercial evidence [1] |
| Gullifa et al. (2023) [19] | Yes | Formulation/stability + SGF/SIF model [19] | Gastric acid/bile; moisture/manufacturing stability [19] | 1 [19] | Spray-dried Eudragit-coated nutraceutical capsule [19] | Formulation-transferable prototype [19] |
| Kiepś et al. (2023) [20] | Yes | Formulation/stability + in-vitro Caco-2 adhesion [20] | Manufacturing stability; strain non-equivalence [20] | 3 [20] | Fluid-bed-dried/coated powder, glass vials [20] | Formulation-transferable prototype [20] |
| Vikström et al. (2026) [3] | Yes | Commercial-product stability [3] | Moisture/water activity; packaging [3] | 2 (same species) [3] | Oil suspension in dropper bottles [3] | Direct commercial evidence [3] |
| Miao et al. (2008) [2] | Yes | Formulation/stability prototype [2] | Moisture/water activity/excipient; strain non-equivalence [2] | 2 (separately tested) [2] | Freeze-dried powder, glass vials [2] | Formulation-transferable prototype [2] |
| Yavnikov (2021) [21] | Yes | Formulation/stability prototype (feed additives) [21] | Moisture/excipient/manufacturing stability [21] | 2 [21] | Lyophilized preparation [21] | Formulation-transferable prototype [21] |
| Khosroshahi et al. (2023) [22] | Yes | Formulation/stability + drying-method comparison [22] | Manufacturing stability [22] | 2 (co-culture) [22] | Electrospray-dried / freeze-dried powder [22] | Formulation-transferable prototype [22] |
| Goderska (2012) [23] | Yes | Narrative methods review [23] | Moisture/excipient/manufacturing stability [23] | Multiple (reviewed) [23] | Various dry formats reviewed [23] | Biologically informative but indirect [23] |
| Bustamante et al. (2025) [4] | Yes | Formulation/stability prototype [4] | Manufacturing stability; strain non-equivalence [4] | 4 (separately tested) [4] | Spray-dried cross-linked alginate powder [4] | Formulation-transferable prototype [4] |
| Vorländer et al. (2020) [6] | Yes | Formulation/manufacturing prototype (yeast model) [6] | Manufacturing stability (compression/milling) [6] | 1 (S. cerevisiae model) [6] | Directly compressed tablets [6] | Formulation-transferable prototype [6] |
| Thantsha et al. (2014) [7] | Yes | Formulation/stability prototype (accelerated shelf life) [7] | Moisture/manufacturing stability; strain non-equivalence [7] | 2 (separately tested) [7] | Free microparticle powder, glass vials, 30 °C [7] | Formulation-transferable prototype [7] |
| Benavent-Gil et al. (2018) [24] | Yes | Formulation/stability prototype; thermal challenge [24] | Manufacturing stability [24] | 1 [24] | Freeze-dried starch microcapsules [24] | Formulation-transferable prototype [24] |
| Trimudita & Djaenudin (2021) [25] | Yes | SGF/SIF GI model; alginate-chitosan encapsulation [25] | Gastric acid/bile delivery protection [25] | 1 [25] | Extruded alginate-chitosan beads [25] | Formulation-transferable prototype [25] |
| Ding & Shah (2007) [26] | Yes | In-vitro acid/bile/heat tolerance; milk-protein encapsulation [26] | Gastric acid/bile; strain non-equivalence [26] | 2 (separately tested) [26] | Milk-protein capsules; freeze-dried [26] | Formulation-transferable prototype [26] |
| Caballero-Pérez et al. (2023) [27] | Yes | SGF/SIF GI model; spray-dried mixed-culture microcapsules [27] | Gastric acid/bile; multi-strain antagonism [27] | 3 (co-formulated) [27] | Spray-dried powder microcapsules [27] | Formulation-transferable prototype [27] |
| Del Piano et al. (2011) [28] | Yes | Human crossover RCT + in-vitro GI survival [28] | Gastric acid/bile; delivery protection; strain non-equivalence [28] | 2 (co-administered) [28] | Lipid-microencapsulated vs. uncoated capsule [28] | Direct commercial evidence [28] |
| Chen et al. (2005) [29] | Yes | Formulation/stability + SGF/SIF model [29] | Gastric acid/bile; moisture/excipient stability [29] | 4 (co-encapsulated) [29] | Extruded alginate-prebiotic beads, 4 °C [29] | Formulation-transferable prototype [29] |
| Kang et al. (2013) [30] | Yes | Formulation/stability + SGF/SIF model [30] | Gastric acid/bile; manufacturing stability [30] | 4 (co-blended) [30] | Dual-coated peptide-polysaccharide blend [30] | Formulation-transferable prototype [30] |
| Astutiningrum et al. (2020) [31] | Yes | SGF/SIF GI model; spray-dried microcapsules [31] | Gastric acid/bile [31] | 3 (co-encapsulated) [31] | Spray-dried skimmed milk-maltodextrin microcapsules [31] | Formulation-transferable prototype [31] |
| Abd Rabo et al. (2007) [32] | Yes | SGF/SIF GI model; free vs. alginate-encapsulated [32] | Gastric acid/bile; strain non-equivalence [32] | 6 (separately tested) [32] | Sodium-alginate bead encapsulation [32] | Formulation-transferable prototype [32] |
| D'Orazio et al. (2015) [33] | Yes | SGF/SIF GI model + HT-29 cell line; chitosan-alginate capsules [33] | Gastric acid/bile; strain non-equivalence [33] | 3 (individually and as 1:1:1 mixture) [33] | Chitosan-alginate microcapsules, 4 °C [33] | Formulation-transferable prototype [33] |
| Anselmo et al. (2016) [34] | Yes | SGF/SIF + ex-vivo porcine intestine + mouse gavage [34] | Gastric acid/bile; delivery protection; colonization [34] | 1 (B. coagulans) [34] | Layer-by-layer chitosan/alginate coating [34] | Formulation-transferable prototype [34] |
| Afzaal et al. (2019) [35] | Yes | SGF/SIF + yogurt storage [35] | Gastric acid/bile; moisture/storage stability [35] | 1 [35] | Alginate/carrageenan beads in yogurt [35] | Formulation-transferable prototype [35] |
| Govaert et al. (2024) [11] | Yes | Commercial-product quality; SHIME + CoaP fermentation [11] | Gastric acid/bile; delivery protection; strain non-equivalence [11] | 4, 2, 9, 7 strains across four products [11] | Liquid, powder, standard capsule, delayed-release capsule [11] | Direct commercial evidence [11] |
| Afzaal et al. (2020) [36] | Yes | SGF/SIF + thermal + refrigerated storage [36] | Gastric acid/bile; manufacturing stability [36] | 1 [36] | Hydrogel microbeads (alginate/carrageenan) [36] | Formulation-transferable prototype [36] |
| Jiménez-Pranteda et al. (2012) [37] | Yes | SGF/SIF GI model; microbial polymer encapsulation [37] | Gastric acid/bile; strain non-equivalence [37] | 2 (separately tested) [37] | Xanthan/gellan or jamilan/gellan beads [37] | Formulation-transferable prototype [37] |
| Venema et al. (2019) [12] | Yes | Dynamic TIM-1 stomach–small-intestine model; commercial BION3 tablet [12] | Gastric acid/bile; delivery protection; strain non-equivalence [12] | 3 (L. gasseri + 2 Bifidobacterium) [12] | Enteric-coated multi-layer tablet [12] | Direct commercial evidence [12] |
| Gunzburg et al. (2020) [9] | Yes | SGF/SIF GI model + nude mouse gavage [9] | Gastric acid/bile; delivery protection; colonization [9] | 5 bacteria + 1 yeast (separately tested) [9] | Cellulose sulphate (CS) freeze-dried microspheres [9] | Formulation-transferable prototype [9] |
| Samedi & Charles (2019) [38] | Yes | SGF/SIF GI model + yogurt storage [38] | Gastric acid/bile; moisture/storage stability; strain non-equivalence [38] | 4 (separately tested) [38] | Arrowroot/maltodextrin freeze-dried beads in yogurt [38] | Formulation-transferable prototype [38] |
| Vitsou-Anastasiou et al. (2025) [39] | Yes | Commercial-product formulation + SGF/SIF model; orange juice [39] | Gastric acid/bile; multi-strain antagonism (vs. pathogen) [39] | 2 (co-encapsulated cocktail) [39] | WPI-gum Arabic coacervate in orange juice [39] | Formulation-transferable prototype [39] |
| Gunzburg et al. (2020a) [40] | Yes | SGF/SIF GI model + nude mouse gavage [40] | Gastric acid/bile; delivery protection [40] | 5+ strains (separately tested) [40] | CS freeze-dried microspheres [40] | Formulation-transferable prototype [40] |
| Bernatek et al. (2022) [41] | Yes | Commercial-product quality; SGF/SIF model [41] | Gastric acid/bile; delivery protection; multi-strain antagonism [41] | 1, 7, 1, 8, 9 strains across five products [41] | Commercial capsules (gastro-resistant vs. regular) [41] | Direct commercial evidence [41] |
| Anonymous (2021) [42] | Yes | Formulation/stability + SGF/SIF model; yogurt [42] | Gastric acid/bile; strain non-equivalence; storage stability [42] | 2 (co-formulated in yogurt) [42] | Alginate-chitosan-Eudragit beads in yogurt [42] | Formulation-transferable prototype [42] |
| Fijan et al. (2018) [43] | Yes | Commercial-product quality; in-vitro agar/milk co-culture [43] | Multi-strain antagonism; strain non-equivalence [43] | SSPs (1 strain each) and MSPs (3–9 strains) [43] | Commercial dietary supplements [43] | Direct commercial evidence (in-vitro antagonism only) [43] |
| Campana et al. (2017) [15] | Yes | SGF/SIF GI model + Caco-2 invasion assay [15] | Gastric acid/bile; strain non-equivalence; multi-strain combination effects [15] | 7 (commercial WincloveTravel strains) [15] | Lyophilized strains from commercial formulation [15] | Biologically informative but indirect [15] |
| Puvanasundram et al. (2022) [44] | Yes | In-vitro co-culture + antagonism screening [44] | Multi-strain antagonism; strain non-equivalence [44] | MSP1: 3 strains; MSP2: 3 strains (aquaculture) [44] | Liquid co-culture mixtures [44] | Biologically informative but indirect [44] |
| Malfa et al. (2023) [45] | Yes | In-vitro 3D epithelial co-culture (vaginal/bladder) [45] | Multi-strain antagonism; strain non-equivalence [45] | 3 (SynBalance Femme) [45] | Freeze-dried commercial powder [45] | Biologically informative but indirect [45] |
| Wende et al. (2025) [46] | Yes | In-vitro/ex-vivo + mouse colonization model [46] | Multi-strain antagonism; colonization; strain non-equivalence [46] | 2 candidate commensals + MDR targets [46] | Bacterial suspension by oral gavage [46] | Biologically informative but indirect [46] |
| Forssten & Ouwehand (2017) [47] | Yes | In-vitro simulated-colon model (qPCR) [47] | Multi-strain antagonism (null finding) [47] | 4 (HOWARU Restore + single-strain arms) [47] | Freeze-dried cultures (HOWARU Restore) [47] | Biologically informative but indirect [47] |
| Menconi et al. (2014) [48] | Yes | Commercial-product quality/characterization (FloraMax-B11) [48] | Gastric acid/bile; strain non-equivalence [48] | 2 LAB isolates [48] | Commercial probiotic culture [48] | Biologically informative but indirect [48] |
| Bubnov et al. (2018) [14] | Yes | In-vitro GI challenge + adhesion screening [14] | Gastric acid/bile; strain non-equivalence [14] | 8 (separately tested) [14] | Freeze-dried strains [14] | Biologically informative but indirect [14] |
| MacPherson et al. (2017) [49] | Yes | In-vitro HT-29 cell challenge (transcriptomics) [49] | Strain non-equivalence; multi-strain immunomodulation [49] | 3 (20:20:60 blend) [49] | Industrial lyophilized powders, laboratory blend [49] | Biologically informative but indirect [49] |
| Knysh et al. (2023) [50] | Yes | In-vitro agar co-cultivation; Bacillus spp. [50] | Multi-strain antagonism; colonization (motility) [50] | 3 species (commercial source) [50] | Spore suspensions from commercial products [50] | Biologically informative but indirect [50] |
| Kimelman & Shemesh (2019) [51] | Yes | In-vitro dual-species biofilm + SGF/SIF model [51] | Multi-strain (B. subtilis ECM protection); strain non-equivalence [51] | 3 LAB + B. subtilis (dual-species pairs) [51] | Bio-coated dried/freeze-dried cultures [51] | Formulation-transferable prototype [51] |
| Tran et al. (2025) [52] | Yes | In-vitro co-culture + metabolite stability [52] | Multi-strain antagonism (synergistic growth, antimicrobial extract) [52] | 3 B. amyloliquefaciens (co-culture) [52] | Laboratory fermentation cultures [52] | Biologically informative but indirect [52] |
| Sipahi (2021) [53] | Yes | Commercial-product quality; in-vitro antagonism [53] | Multi-strain antagonism [53] | 11 strains (commercial supplement) [53] | Commercial lyophilized capsule with oat fiber [53] | Direct commercial evidence (antagonism assay only) [53] |
| Piatek et al. (2020) [16] | Yes | Commercial-product quality; in-vitro agar inhibition [16] | Multi-strain antagonism (vs. pathogens) [16] | 1, 1, 1, 3, 9 strains across five products [16] | Commercial products including Multilac Baby synbiotic [16] | Direct commercial evidence (antagonism assay) [16] |
| Reddy (2024) [13] | Yes | In-vitro mixed-culture compatibility (serial transfer) [13] | Multi-strain antagonism/competitive exclusion [13] | Up to 8 strains (Lactococcus, Streptococcus, etc.) [13] | Liquid cultures in reconstituted milk [13] | Formulation-transferable prototype [13] |
| Saracino et al. (2020) [54] | Yes | In-vitro antagonism (vs. H. pylori) [54] | Multi-strain antagonism; strain non-equivalence [54] | 5 (separately tested) [54] | Laboratory lyophilized cultures [54] | Biologically informative but indirect [54] |
| Salar-Behzadi et al. (2010) [5] | Yes | In-vitro heat-stress model (spray-drying context) [5] | Manufacturing stability; strain non-equivalence [5] | 2 (separately tested) [5] | Spray-drying stabilization context [5] | Biologically informative but indirect [5] |
| Cai et al. (2023) [55] | Yes | In-vitro co-culture + SGF/SIF + yogurt storage [55] | Multi-strain compatibility; strain non-equivalence; GI delivery [55] | 3 (LAB co-culture in yogurt) [55] | Set-type yogurt prototype [55] | Formulation-transferable prototype [55] |
| Abd-Talib et al. (2013) [8] | Yes | Formulation/stability; spray-drying + 2-week storage [8] | Manufacturing stability; strain non-equivalence [8] | 4 (separately tested) [8] | Spray-dried animal-feed powder [8] | Formulation-transferable prototype [8] |
| Ding & Shah (2009) [56] | Yes | Formulation/stability + SGF/SIF + HT-29 adhesion [56] | Gastric acid/bile; strain non-equivalence; manufacturing stability [56] | 5 (separately tested) [56] | Silk-fibroin-coated freeze-dried powder [56] | Formulation-transferable prototype [56] |
| Ross et al. (2008) [57] | Yes | SGF/SIF GI model; calcium-alginate encapsulation [57] | Gastric acid/bile; manufacturing stability [57] | Unspecified LAB from pig feces [57] | Calcium-alginate microcapsules [57] | Formulation-transferable prototype [57] |
| Ansari et al. (2017) [10] | Yes | SGF/SIF GI model; double-coated alginate-chitosan-Eudragit beads [10] | Gastric acid/bile; strain non-equivalence [10] | 2 (separately tested) [10] | Eudragit S100 nanoparticle double-coated beads [10] | Formulation-transferable prototype [10] |
| Strashnova et al. (2022) [58] | Yes | In-vitro co-culture compatibility; Lactobacillus spp. [58] | Multi-strain antagonism; strain non-equivalence [58] | 3 (Lactobacillus spp. 175, M2, M3) [58] | Laboratory broth-culture mixtures [58] | Biologically informative but indirect [58] |
| Golić et al. (2017) [59] | Yes | In-vitro antagonism + Caco-2 immunomodulation + animal field trial [59] | Multi-strain antagonism; colonization [59] | 3 (L. helveticus + L. fermentum + S. thermophilus) [59] | Liquid mixed culture (50 mL daily for goats) [59] | Formulation-transferable prototype [59] |
| Rutter et al. (2024) [60] | Yes | In-vitro bacteriocin secretion platform; engineered eLBP [60] | Multi-strain antagonism; competitive exclusion [60] | Engineered E. coli + E. faecalis/faecium targets [60] | Laboratory BHI cultures (no commercial dosage form) [60] | Biologically informative but indirect [60] |
The studies fall into three broad applicability tiers. Fourteen sources constitute direct commercial evidence, testing either marketed products or pilot-scale preparations that closely model commercial dosage forms [1, 3, 11, 12, 16, 18, 28, 41, 43, 53]. Twenty-nine sources supply formulation-transferable prototype evidence, providing quantitative data on survival, excipient effects, and GI challenge outcomes from laboratory preparations that are sufficiently specific in their design to inform commercial manufacture [2, 4, 6–8, 10, 13, 17, 19–22, 24–27, 29–31, 33–39, 42, 51, 55, 57, 59]. The remaining seventeen sources are biologically informative but indirect, including mechanistic co-culture studies, narrative reviews, and pure antagonism screens that lack the dosage-form context required for direct translation [5, 14, 15, 23, 40, 44–50, 52, 54, 56, 58, 60].
A critical cross-cutting limitation is the paucity of studies that simultaneously resolve all four requested vectors—strain-level viability, GI delivery survival, multi-strain interaction, and clinical outcome—within the same experimental system. Most sources optimize one variable while holding others constant, making it necessary to triangulate across studies when drawing synthesis-level conclusions.
Thematic Analysis
Theme 1: Moisture, Water Activity, and Packaging as the Primary Driver of Storage-Phase Viability Loss
Moisture accumulation is the most consistently documented failure mechanism across commercial and prototype dry probiotic formats. The 24-month commercial stability study by Sielatycka et al. (2023) quantified declining active fluorescent unit (AFU) counts in blister-packed capsules stored at 25 °C/60% RH, with losses of 0.12 log10 AFU/g in the single-strain product, 0.16 log10 AFU/g in the two-strain product, and 0.26 log10 AFU/g in the four-strain product, while water activity (aw) rose but remained below 0.15 throughout [1]. The monotonic inverse relationship between aw and AFU observed across all three products underlines the mechanistic primacy of moisture ingress [1].
The oil-suspension study by Vikström et al. (2026) reinforces this conclusion in a radically different format. Without a desiccant strip, water content in glass-bottle L. reuteri DSM 17938 preparations was strongly negatively correlated with CFU (Pearson r = −0.9798, p < 0.0001), and the no-desiccant EasyDropper product containing an equal mixture of DSM 17938 and BG-R46® fell below the quality specification after only six months, whereas desiccant-protected batches remained viable for at least 24 months at 25 °C [3]. Average log-CFU losses were 0.4 (with desiccant) versus 1.1 (without desiccant) in the glass-bottle single-strain format and 0.2 versus 0.8 in the mixed-strain dropper format [3].
Miao et al. (2008) traced the mechanism at the molecular level, showing that freeze-dried powders stored above 11.4% relative vapour pressure (RVP) underwent matrix crystallization, depression of glass-transition temperature (Tg), and collapse, each producing catastrophic viability losses. At 44.1% RVP, survival of both L. paracasei NFBC 338 and L. rhamnosus GG in reconstituted skimmed milk (RSM) dropped to approximately 0.02%, while at 11.4% RVP the respective values were 22.0% and 65.4% [2]. Disaccharide protectants (trehalose and lactose + maltose) delayed crystallization, maintaining 95.5–100% survival at 11.4% RVP [2]. Thantsha et al. (2014) corroborated the aw threshold effect: the largest viability losses in PVP:PVAc-CA-encapsulated bifidobacterial powders occurred during weeks 3–5 of storage at 30 °C, when aw peaked (values from 0.25 to 0.43, mean 0.34) [7].
Packaging system integrity directly modulates aw ingress. The Sielatycka et al. blisters had a water vapour transmission rate (WVTR) of only 0.11 g/m²·d, yet aw still increased measurably over 24 months [1]. The Vikström et al. glass bottles had a primary-packaging MVTR of 0.002 g/year but still permitted sufficient moisture entry to drive measurable viability loss without active desiccation [3]. This finding implicates hygroscopic migration from freeze-dried bacterial cells themselves—a proposed Maillard-reaction water-generation mechanism—as an internal moisture source that passive packaging alone cannot fully suppress [3].
Excipient choice modifies moisture kinetics. In the pilot-scale study by Savini et al. (2010), glycerine (a penetrating polyalcohol) provided the best protection at room temperature, while the oligosaccharide prebiotics failed to outperform semi-skimmed milk (SSM) at room temperature, though all protectants were equivalent at 4 °C [18]. Yavnikov (2021) found that 10% skimmed milk with 2% SiO2 gave 82.48% post-lyophilization survival and a residual moisture of 4.93%, described as optimal for long-term viability, while skim milk alone yielded 81.84% and 4.40% moisture [21]. The narrative review by Goderska (2012) highlights that prebiotics such as inulin and oligofructose-enriched inulin reduced moisture content and aw in microcapsules, although oligofructose was the most hygroscopic, illustrating the trade-off between carbon-source benefit and moisture risk [23].
Storage temperature exerts a large, strain-differential effect documented most granularly in the multi-strain spray-dried formulation study by Bustamante et al. (2025). Whereas B. longum remained above 6.81 log CFU/g in all formulations at 37 °C after 21 days, L. rhamnosus fell below 6 log CFU/g at 25 °C by day 45 and below 2.2 log CFU/g by day 90 [4]. First-order decay constants at 4 °C ranged from 0.003 day⁻¹ for B. longum to 0.010 day⁻¹ for B. infantis [4]. The study identified L. rhamnosus as intrinsically less resistant to high temperatures and low-humidity environments [4].
Theme 2: Manufacturing Processes and Their Strain-Differential Impact on Viability
Every thermal and mechanical process step imposes a viability cost, and strains respond to these stresses at markedly different rates. Salar-Behzadi et al. (2010) demonstrated that B. bifidum 12 suffered maximum membrane and esterase damage after shorter heat exposure than E. faecium M 74 across the 60–90 °C range relevant to spray drying [5]. Khosroshahi et al. (2023) quantified the differential between drying methods: electrospraying reduced viability by 0.55 log CFU/g versus approximately 1.2 log CFU/g for freeze-drying from the same mixed L. acidophilus/L. plantarum fermentation, and after 70 days' ambient storage the electrosprayed product retained 9.05 ± 0.45 log CFU/g against 7.86 ± 0.03 log CFU/g for the freeze-dried equivalent [22].
The fluid-bed-drying and coating study by Kiepś et al. (2023) reveals that membrane damage from processing is directly linked to downstream colonization capacity: Caco-2 adhesion was confined to the active (undamaged) cell subpopulation, and heat shock reduced this fraction sharply for L. mesenteroides but not equivalently across all three strains tested [20]. Adhesion rates for the dried preparations were 8.14% for L. mesenteroides, 34.12% for E. faecium, and 9.10% for C. divergens; coating altered these values strain-specifically, improving C. divergens adhesion from 9.10% to 13.07% while reducing E. faecium adhesion from 34.12% to 22.96% [20].
For tablet manufacture, Vorländer et al. (2020) showed that compaction stress using baker's yeast as a model organism reduced survival from 19.7% at 100 MPa to 2.0% at 400 MPa for unformulated freeze-dried material, and that dicalcium phosphate (DCP) yielded the lowest viability at 400 MPa (1.4%) compared with microcrystalline cellulose (2.4%) or lactose (2.6%), attributed to fracture of sharp-edged DCP particles [6]. The maximum achievable overall manufacturing-chain survival was >5%, equivalent to 6 × 10⁸ CFU g_total⁻¹, with cryoprotectants and excipients included [6].
Spray-drying is both the most scalable and most thermally damaging approach. In the Bustamante et al. (2025) study at 130 °C inlet temperature, survival immediately post-spray-drying ranged from 86.65 ± 3.19% (L. rhamnosus with flaxseed mucilage) to 98.66 ± 0.19% (L. plantarum with chia mucilage), a difference driven by strain intrinsic heat sensitivity rather than wall-material composition per se [4]. Abd-Talib et al. (2013) demonstrated an even starker strain-by-manufacturing interaction: L. plantarum B18 had higher post-spray-drying survival than B13 but collapsed to only 4 × 10² cfu/mL after two weeks' ambient storage, while B13 retained 3.0 × 10⁵ cfu/mL [8]. This inversion—high processing tolerance paired with poor storage stability—illustrates why single-timepoint assessments immediately after manufacture systematically overestimate long-term product quality.
The glass-transition temperature (Tg) of the coating or matrix is a critical design parameter linking manufacturing and storage stability. Kiepś et al. (2023) measured DSC Tg values of 144.8 °C (gum Arabic), 152.6 °C (HPMC), and 147.6 °C (shellac) for their coating materials, with HPMC-coated preparations showing the best 12-month viability retention [20]. Miao et al. (2008) established that Tg depression below approximately 53–56 °C by moisture uptake marks the threshold at which molecular mobility increases sufficiently to cause rapid viability loss [2].
Oxygen during storage is an underappreciated co-factor. Kiepś et al. (2023) showed that nitrogen or vacuum packaging extended uncoated-sample shelf life from 6 to 9 months at 4 °C, compared with storage in air [20]. Goderska (2012) cited a case in which nitrogen storage was essential for storage stability of one ultrasonic vacuum spray-dried formulation [23].
Theme 3: Gastric Acid and Bile Destruction—Quantitative Evidence for Protection-Dependent Delivery
Unprotected probiotic cells suffer severe viability loss during gastric transit. Venema et al. (2019) quantified this most rigorously using the dynamic TIM-1 model: unformulated probiotic powder delivered only 5.3% of bifidobacteria and 1.0% of Lactobacillus through the gastric compartment, falling further to 2% and 0.1% through the complete stomach–small-intestine system [12]. The optimized 5% ethanolic HPMC:HPC enteric-coated tablet increased gastric survival to 71.5% for bifidobacteria and 52.7% for Lactobacillus, a 13.5- to 50-fold improvement versus powder [12]. Govaert et al. (2024) translated analogous findings into commercially available products: delayed-release capsule formulations achieved 56.0 ± 15.4% culturability and 51.7 ± 10.4% viability at the ileal endpoint, while liquid, powder, and standard-capsule formulations all fell below 1% [11]. The standard capsule had fully dissolved by the gastric endpoint, removing all acid protection prematurely [11].
The quantitative scale of protection is consistent across study designs. In the Gunzburg et al. (2020a/2020) cellulose sulphate series, free L. casei declined ~8 logs and S. boulardii ~5 logs after just one hour in artificial gastric juice at pH 2 with pepsin and lysozyme, whereas encapsulated preparations showed no significant viability effect after four hours of acid plus one hour of bile [9, 40]. D'Orazio et al. (2015) showed that non-encapsulated cells were completely destroyed by SGJ at pH 2.5, while alginate-glucose with chitosan coating maintained initial viability; removing the chitosan layer allowed ~2 log losses, and removing both glucose and chitosan reduced viability to uncountable levels [33]. In the encapsulated versus free comparison by Ansari et al. (2017), double-coated chitosan/Eudragit S100 beads suffered only 2.6 log reductions (L. acidophilus) and 2.1 log reductions (L. rhamnosus) after 120 minutes at pH 1.55, versus 6.7 and 6.5 log losses for free cells [10].
Bile salt exposure adds an independent, concentration-dependent stressor. Abd Rabo et al. (2007) documented that L. helveticus Lh.B 02 free cells fell below 5 log cfu/mL at 0.5% bile salt after 180 minutes, a threshold at which L. acidophilus La-5 still remained above 5 log cfu/mL, illustrating substantial strain-specific bile sensitivity in the same experimental system [32]. Jiménez-Pranteda et al. (2012) showed that free L. rhamnosus ATCC 53103 declined >3 log CFU after only 2 hours of 3% bile exposure, while L. plantarum CRL 1815 retained detectable counts for 20 hours under the same conditions [37]. Xanthan:gellan gum (1%:0.75%) encapsulation reduced bile losses to ≤1 log CFU for both strains, whereas jamilan:gellan gum (1%:1%) failed to protect L. plantarum—an incompatibility the authors attribute to biological differences between the polymer and that strain rather than to a universal formulation deficiency [37].
The sequential challenge protocol used by the more physiologically realistic studies reveals an important interaction: cells that survive the gastric compartment may fail in the duodenum if the coating has already disintegrated [12]. Venema et al. (2019) hypothesized that release at low gastric pH conditioned cells to subsequent high-bile and pancreatic-enzyme stress in the duodenum, accounting for the disproportionate survival drop between gastric and complete-system measurements [12]. The practical implication is that coating integrity at the time of entry into the duodenum, not merely acid resistance, determines ultimate small-intestinal delivery.
Del Piano et al. (2011), the only human crossover trial in this corpus, provided direct translational evidence: lipid-microencapsulated LP01 and BR03 administered at 2 billion CFU/day produced similar fecal colonization kinetics to uncoated strains given at 10 billion CFU/day, consistent with the estimated tenfold increase in gastro-duodenal survival from encapsulation (at least 90% versus 15–25% for uncoated cells) [28, 28]. The between-formulation fecal difference was not statistically significant, but this study was not powered to detect it, and the dose imbalance limits interpretation [28].
Capsule technology type is a decisive variable in commercial products. Bernatek et al. (2022) showed that the gastro-resistant Multilac capsule (9 strains, disintegration time 60 min) lost only 1.08 log CFU after 90 minutes at pH 2, while regular-capsule products with 10-minute disintegration times lost 3.0–3.12 log CFU [41]. The in-vitro agar antagonism data in the same study showed that larger inhibition zones against E. coli, Shigella, Salmonella, and C. difficile tracked exactly with strain diversity and acid-protection strategy of the product, though these two variables were confounded [41].
Theme 4: Multi-Strain Biological Interactions—Antagonism, Compatibility, and Co-Formulation Consequences
The question of whether co-formulated strains interact biologically to reduce viability, recovery, or efficacy is empirically answered only for a small subset of the eligible study designs. Most studies that test multiple strains do so in physically separated arms rather than as a genuinely non-segregated mixture, and the majority that do use genuine mixtures report outcomes only at the aggregate level.
Evidence for compatibility with no measurable antagonism.
Forssten & Ouwehand (2017) used an in-vitro simulated-colon model with qPCR quantification and found no substantial difference in detected strain levels between single-strain and four-strain HOWARU Restore administration [47]. The data support neither an antagonistic nor a synergistic survival effect among L. acidophilus NCFM, L. paracasei Lpc-37, B. lactis Bl-04, and B. lactis Bi-07 under the colonic environment modelled [47]. Zárate & Nader-Macías (2006) similarly confirmed that their three vaginal lactobacilli "do not inhibit one to each other" in direct agar testing, supporting their combination [17]. Golić et al. (2017) found no detrimental interaction and maintained unchanged total viable counts after nine days of refrigerated storage of a three-strain mixed culture [59]. Kimelman & Shemesh (2019) explicitly showed no significant inhibition between B. subtilis and any of the three co-cultured LAB species [51].
Evidence for incompatibility and strain displacement.
Reddy (2024) is the study most directly addressing intra-formulation competitive exclusion. Using reconstituted milk serial-transfer experiments, L. cremoris SC-2 completely suppressed L. lactis SL-2 to non-detectable counts after the very first transfer (150 × 10⁷ SC-2 versus 0 SL-2 at 10⁷ sensitivity), while the SL-1/SC-1 combination maintained both components [13]. A three-species Lactococcus combination (SL-1/SC-1/SD-1) showed L. cremoris SC-1 disappearing from the culture by the second transfer while L. lactis and L. diacetylactis remained robust, a pattern the author classifies as an incompatible formulation [13]. Proposed mechanisms include Nisin production by certain L. lactis strains and Diplococcin by certain L. cremoris strains, as well as differential acid-production rates [13]. This study provides the clearest within-species strain-level demonstration that compatibility is not guaranteed even among closely related taxa.
Strashnova et al. (2022) identified specific pairwise incompatibilities among Lactobacillus strains from natural sources: O1 strongly antagonized M3, and B4 weakly antagonized M2, preventing their combination in multi-strain formulations [58]. After exclusion of incompatible pairs, the selected three-strain compositions (M2/M3/175) at a 1:2:2 ratio produced the broadest and strongest pathogen inhibition zones [58]. Separately grown-and-then-mixed compositions were slightly more antagonistically active than jointly cultivated consortia, consistent with the proposed mechanism that inhibitory metabolites produced by co-cultivating strains can inactivate one another [58].
Bacterial ECM cross-protection as a positive biological interaction.
Kimelman & Shemesh (2019) demonstrated that B. subtilis NCIB3610 bio-coating of LAB through shared extracellular matrix production substantially improved desiccation and freeze-drying survival: B. subtilis co-culture increased L. rhamnosus viability by approximately 3.12 log CFU/mL after 20 hours of desiccation and by approximately 5 logs after 40 hours, whereas the same benefit was absent for P. acidilactici under freeze-drying and simulated GI exposure [51]. This establishes that cross-species physical interactions in the matrix can be powerfully protective but are strain-specific rather than universal [51].
Multi-strain combination effects on pathogen inhibition.
Several studies address whether the number of co-formulated strains enhances pathogen inhibition, which is a proxy for functional efficacy rather than a direct viability or recovery outcome. Piątek et al. (2020) found that a nine-strain synbiotic (Multilac Baby with FOS) produced the largest in-vitro inhibition zones against all five tested pathogens, while L. rhamnosus GG and S. boulardii monostrains were weakest, and the three-strain L. rhamnosus E/N/Oxy/Pen mixture was intermediate [16]. Bernatek et al. (2022) reported analogous results for agar inhibition [41]. Fijan et al. (2018) obtained opposite results from co-culturing in milk: single-strain products outperformed multi-strain products, with mean E. coli log-step reductions of 1.35 versus 0.69 (p < 0.05), a reversal the authors attribute to interspecies nutrient competition in the milk medium [43]. This method-dependence is a critical interpretive caveat: agar-based assays favor multi-strain products, while liquid co-culture assays favor single strains, and neither accurately represents intestinal conditions.
Yeast-bacterial antagonism within co-formulations.
Caballero-Pérez et al. (2023) documented a within-mixture biological interaction of direct relevance: LAB-produced organic acids (propionate, acetate, lactate) progressively suppressed S. boulardii growth during the prebiotic-broth incubation, resulting in prebiotic-test counts of 6.994 log CFU/mL for S. boulardii versus 9.664 and 9.129 for L. rhamnosus and L. plantarum, respectively (p = 0.000) [27]. This concentration differential was not apparent in the 4-hour SGF or SIF exposures, suggesting the antagonism is time-dependent and carbon-source-mediated rather than pH-mediated [27].
Differential survival among co-formulated strains at the GI level.
Govaert et al. (2024) is the only commercial-product study to resolve survival by genus within a non-segregated multi-strain delayed-release capsule under simulated upper-GI passage: Bifidobacterium spp. reached only 4.76 × 10⁹ ± 5.13 × 10⁸ CFU/reactor culturability and 1.62 × 10⁹ viable 16S rRNA copies/reactor at the ileal endpoint, while Lactobacillus spp. attained 2.92 × 10¹⁰ ± 9.05 × 10⁹ CFU/reactor and 1.07 × 10¹¹ copies/reactor, a 6- to 66-fold difference depending on the measure [11]. Co-supplementation with GOS or FOS negatively affected the detected Bifidobacterium OTU but not the Lactobacillus OTU, an interaction the authors attribute to alteration of the colonic environment rather than to direct probiotic-prebiotic antagonism [11].
Similarly, the commercially manufactured BION3 three-strain tablet in Venema et al. (2019) produced different complete-system survival percentages for the pooled bifidobacteria (13.5% elderly, 7.3% adults) versus L. gasseri (7.5% elderly, 9.6% adults), though the two Bifidobacterium strains could not be distinguished at the plating level [12]. In the Sielatycka et al. (2023) commercial capsule study, the four-strain product showed a 0.26 log10 AFU/g aggregate decline over 24 months, exceeding the 0.12 and 0.16 log10 losses of the single- and two-strain products, but individual-strain contributions could not be resolved because the flow cytometry method does not distinguish between constituent strains [1].
Theme 5: Strain Non-Equivalence as a Pervasive and Under-Characterized Risk
Across every experimental domain—storage, manufacturing, GI challenge, and colonization proxy—the data consistently show that phenotypically identical species-level designations conceal large quantitative performance differences between named strains.
In the freeze-dried RSM storage experiment by Miao et al. (2008), L. rhamnosus GG survived at 65.4% at 11.4% RVP after 38 days, while L. paracasei NFBC 338 survived at only 22.0% under identical conditions [2]. In the heat-stress assay by Ding & Shah (2007), L. salivarius W13 was more resistant than L. reuteri LRT18 in both bile (37.2% versus 35.6% survival at 0.3% oxgall) and heat (65°C, 75°C) challenges [26]. In the sequential GI model of Campana et al. (2017), B. bifidum W23 showed only a twofold CFU decline from baseline through the full simulated GI passage, while L. acidophilus W37, L. salivarius W24, and L. rhamnosus W71 were unable to survive the simulated small-intestinal conditions at all [15].
Perhaps the most dramatic illustration of within-genus non-equivalence comes from Bubnov et al. (2018): sequential exposure to 2% gastric juice, 1% bile, and 1% pancreatin resulted in 96.96% combined survival for L. acidophilus IMV B-7279, while L. plantarum LM VK7 survived at only 49.25% and L. rhamnosus LB-3 VK6 at 50.00% under identical conditions [14]. In vitro adhesion (AIM) ranged from 2.14 ± 0.34 for L. plantarum VK7 to 7.81 ± 0.86 for L. casei IMV B-7280, a 3.6-fold difference [14].
In spray-dried formulations, Bustamante et al. (2025) quantified the storage-susceptibility hierarchy: L. rhamnosus ATCC53103 was the most vulnerable at 25 °C (below 6 log CFU/g by day 45) and at 37 °C (below 6 log CFU/g by day 21), while B. longum ATCC15707 remained above 6.81 log CFU/g at 37 °C for all formulations at day 21 [4]. The first-order decay constants at 4 °C ranged from k = 0.003 day⁻¹ for B. longum to k = 0.010 day⁻¹ for B. infantis, a threefold differential in long-term refrigerated stability [4].
Thantsha et al. (2014) showed that encapsulation conferred a 6 log c.f.u./g survival benefit for B. longum Bb46 over 12 weeks at 30 °C but only a 3.3 log benefit for B. lactis Bb12, because Bb12 was intrinsically more stable [7]. The implication is that the absolute value of encapsulation benefit is inversely proportional to intrinsic strain robustness—a critical consideration when selecting which strains to include and protect in a mixed formulation.
The Saracino et al. (2020) in-vitro H. pylori antagonism study provides the sharpest within-species strain-level efficacy contrast: L. casei, L. paracasei, and L. acidophilus each inhibited 100% of 57 clinical H. pylori isolates, B. lactis inhibited 89.5%, and S. thermophilus inhibited only 18% [54]. Although these are not named strain-versus-strain comparisons within the same species, the data illustrate that species-level labels conceal clinically material performance variation.
At the intra-species level, Piątek et al. (2020) showed that L. reuteri DSM 17938 produced meaningfully larger inhibition zones against all five tested pathogens than L. rhamnosus GG (approximately 8.0 versus 4.7 mm for E. coli EPEC, 9.7 versus 4.3 mm for K. pneumoniae) [16]. This quantitative non-equivalence within the same genus, and even the same species (L. rhamnosus GG versus L. rhamnosus E/N, Oxy, Pen), has direct implications for the evidentiary basis of clinical efficacy claims that are made for one strain but applied to another.
MacPherson et al. (2017) demonstrate a genomic mechanism: each of the three strains in their blend (L. helveticus R0052, B. longum subsp. infantis R0033, B. bifidum R0071) produced unique, non-overlapping transcriptional modulation in HT-29 cells, and the probiotic combination modulated only 131 genes versus 467 (R0033), 367 (R0052), and 293 (R0071) for the individual strains [49]. This synergistic reduction in inflammatory gene expression illustrates that multi-strain combinations can produce qualitatively different biological outcomes from any constituent strain alone, but the mechanism is specific to these three strains at their tested ratio and cannot be generalized to other combinations without independent testing.
Synthesis
Reconciling Heterogeneous Findings Across the Four Vectors
Vector 1: Moisture/Manufacturing Stability—High Confidence, Quantitative Thresholds Established
The body of evidence on moisture is unusually convergent. The protective threshold of aw < 0.15 for freeze-dried commercial capsules (Sielatycka et al., 2023) [1] and the critical RVP range of 0–11.4% for freeze-dried powders (Miao et al., 2008) [2] align well with the mechanistic Tg framework: below these moisture limits, molecular mobility is suppressed and viability is preserved. Above them, matrix crystallization and collapse accelerate inactivation. The desiccant-strip study by Vikström et al. (2026) provides the strongest causal evidence, because it directly manipulates the moisture variable in a commercially formatted oil suspension and measures the consequent viability trajectory [3]. The finding that passive packaging alone (MVTR 0.002 g/year) is insufficient to prevent viability loss at 25 °C/60% RH without active desiccation suggests that the sector-wide assumption that sealed packaging is adequate is wrong for at least some formats.
Where studies diverge is on optimal excipient choice. Glycerine outperformed other protectants at room temperature (Savini et al., 2010) [18], while trehalose and lactose + maltose were superior in the freeze-dried powder systems studied by Miao et al. (2008) [2]. This apparent conflict resolves along the axis of penetrating (glycerine) versus non-penetrating (sugars) agents: glycerine provides cell-membrane protection that is medium-independent and thus superior when residual moisture is variable, while trehalose's glassy-matrix mechanism is optimal at very low RVP. The two findings are therefore complementary rather than contradictory, each addressing a different storage scenario.
Vector 2: GI Delivery Protection—Quantitative Consensus with Formulation-Type Heterogeneity
All delivery-protection studies agree directionally: unprotected free cells lose 5–8 logs during simulated gastric passage, and encapsulation or enteric coating reduces this loss to 1–2 logs. The magnitude of the remaining benefit depends on the specific coating type, thickness, and integrity at the time of gastric emptying. The discrepancy between the 71.5% gastric survival seen with the optimized TIM-1 HPMC:HPC enteric tablet (Venema et al., 2019) [12] and the <50% survival figures for many alginate-bead systems is explained by the single most important mechanism: the tablet coating prevents premature disintegration in the stomach entirely, while alginate beads permit variable permeation of H⁺ ions depending on capsule size, porosity, and the presence of secondary coatings. This is not a contradiction about whether encapsulation works, but about which encapsulation strategy achieves the greatest preservation of cell viability at the site of intended release.
The commercial-product SHIME data from Govaert et al. (2024) are particularly instructive because they include the same delayed-release capsule technology in both simulated upper-GIT passage and a short-term colonic model [11]. The 56% culturability and 52% viability recovered at the ileal endpoint for the delayed-release product—despite being tested at the full in-vivo dose while the other formats were tested at half-dose—demonstrate that the technology gap is real and clinically significant. However, the confounded dose design prevents definitive quantitative comparison across the four formats.
Vector 3: Multi-Strain Compatibility—Sparse Direct Evidence, Systematic Gaps
The most important synthesis conclusion for this vector is an evidence gap rather than a finding: genuine non-segregated multi-strain commercial dosage forms are almost never evaluated with strain-resolved outcomes. Nearly every study that reports multiple strains either tests them separately in parallel arms or reports only aggregate counts. Only Govaert et al. (2024) resolves survival to the genus level within a commercial multi-strain delayed-release capsule, finding approximately six-fold lower Bifidobacterium than Lactobacillus recovery at the ileal endpoint [11]. Only Reddy (2024) provides controlled serial-transfer evidence of intra-formulation competitive displacement, specifically for Lactococcus strain pairs [13]. Only Strashnova et al. (2022) systematically screens for pairwise incompatibilities before combining strains [58].
The Forssten & Ouwehand (2017) null finding of antagonism in a colonic model is often cited as reassurance that multi-strain products are safe to formulate, but its limitations are severe: the assay detected qPCR genome copies rather than viable cells, could not distinguish B. lactis Bi-07 from Bl-04 due to lack of strain specificity, and high endogenous bifidobacterial background further limited detection [47]. The null result cannot be generalized to the question of whether strains depress each other's viability during the storage or gastric-transit phases that precede the colonic environment.
Conversely, the negative competitive displacement found by Reddy (2024) is specifically between same-species Lactococcus variants rather than between the cross-species Lactobacillus-Bifidobacterium combinations that dominate commercial products [13]. Whether the mechanism (Nisin, Diplococcin, differential acid production) operates at equivalent scale during dry storage or lyophilization—where metabolic activity is suppressed—is unknown. The most defensible conclusion is that compatibility among co-formulated strains should be empirically verified for each specific combination rather than assumed, and that existing evidence is insufficient to declare safety for any multi-strain combination a priori.
Vector 4: Strain Non-Equivalence and the Evidentiary Basis for Efficacy Claims
A fundamental tension in this body of literature is that efficacy claims are commonly attributed at the species level or even genus level, while the mechanistic and quantitative data consistently show performance is strain-specific. The 3.6-fold range in Caco-2 adhesion among Lactobacillus strains studied by Bubnov et al. (2018) [14], the inability of three WincloveTravel strains to survive the simulated small intestine while four others did (Campana et al., 2017) [15], and the 2-fold difference in L. reuteri versus L. rhamnosus GG inhibition zones across five pathogen models (Piątek et al., 2020) [16] collectively mean that the evidentiary basis for any multi-strain product depends on whether the specific strains used were the ones studied, not merely the species. The Mentions in several sources of designated strain numbers (e.g., DSM 17938, GG ATCC 53103, R0052) [3, 27, 49] reflect a best practice of strain-level identification that is inconsistently applied across the literature. Studies reporting only species names—as approximately a third of sources in this review do—provide viability and efficacy data that cannot be transferred to a product containing even a different strain of the same species.
The design study necessary to resolve the key remaining uncertainties would need to combine: (1) strain-resolved enumeration within a genuine non-segregated multi-strain dosage form across the full timeline from manufacture through GI delivery, (2) controlled single-strain parallel arms for reference, (3) water activity monitoring during storage, and (4) a validated in-vitro or animal colonization endpoint. No single study in this corpus meets all four criteria simultaneously.
