This review was developed and executed according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (48). To identify all relevant studies examining the effect of isolation and confinement on the human gut microbiota the databases PubMed, Web of Science, Cochrane Library (Wiley) and EBM-Reviews (Ovid) Cochrane Library were searched on September 2^nd 2022. The protocol of this systematic review is registered on the PROSPERO platform with the registration number CRD42022357589. The full search strategy was conducted in assistance with a specialized librarian and is documented in the supporting information (Supplementary Text S1). It consists of three models: isolation condition, human gut microbiota and exclusion of animals. For the search, a very specific search term was chosen to represent the isolation conditions as best as possible. Broader search terms were also tested, but not all relevant studies were found.

Inclusion criteria were established based on the five PICOS dimensions, i.e., participants, interventions, comparator, outcome and study design (49).

Participants: Healthy adults, regardless of sex, age, or weight status, who had been under isolation conditions for at least 28 days were included.

Intervention: Isolation in an environment with constant / reduced antigen exposure. Such isolation conditions are found (a) in space missions; (b) in isolation simulations, such as MARS-500, SIRIUS and Lunar Palace-1 or in bed-rest studies; (c) in extreme environmental conditions, such as in Antarctica, Arctic and Siberia. The intake of probiotics was allowed. Studies that focused on the use of antibiotics for preventing infectious and inflammatory diseases in humans were excluded.

Comparator: Studies with or without control groups met eligibility criteria.

Outcome: Assessment of the microbiota of the human GI tract.

Study design: Randomized controlled trials or non-randomized controlled trials with any publication date and written in English, German and Russian. Only original articles were included.

To identify eligible studies, the search results of the databases were combined, and the duplicates were removed. Two authors (BK and CS) independently screened titles and abstracts to identify relevant trials. Full-text articles were evaluated regarding their eligibility (BK, CS), with uncertainties being discussed between the authors (<3% cases). A third author (IM) was involved if the discrepancy persisted.

Throughout the decades, the methods of microbial analysis changed from cultivation and cell counting to molecular-based approaches like next-generation sequencing, micro-arrays or quantitative polymerase chain reaction. Due to the huge diversity between the method procedures and the associated heterogeneity of the outcomes, the results were assessed separately from each other. The studies were classified into two groups according to microbiota analysis method:

Group 1–Cultivation-based approaches for microbiota analysis.

Group 2–Molecular-based approaches for microbiota analysis.

Additionally, subgroups were created to provide a more homogenous summary of findings.

Subgroup 1–Isolation caused by space missions.

Subgroup 2–Isolation caused by spaceflight- or gravity simulators, e.g., MARS-500, bioregenerative life support systems (BLSS) or bed rest-studies.

Subgroup 3–Isolation in a natural, earth-bound habitat, e.g., Antarctica.

Certain experiments may have resulted in multiple publications concerning the GI microbiota. As the outcomes may differ in detail and description, all publications are listed in the tables. However, a summary of these studies is provided in the text and data evaluation sections.

The following information was extracted from each included article for groups 1 and 2: study characteristics, conditions of isolation, methods of GI microbiota analysis and outcomes. Each study’s characteristics are reported using the original data and summarized in tabular form. Characteristics across the studies are presented as mean, minimum and maximum for sample size, age, body mass index (BMI) and study length.

Primary outcomes concerning the GI microbiota were alpha diversity (richness and biodiversity), community structure/beta diversity and significant shifts in the abundance of individual microbial taxa (according to the current version of the International Code of Nomenclature of Prokaryotes (50)). Significant differences for alpha diversity, community structure, and taxonomy abundances at both phylum and genus level were summarized for isolation/in-mission and post-isolation/post-mission. We also described conclusions on a pre/post comparison if this was possible. For the microbial abundance outcomes in studies using molecular based methods which were not specifically referred to in text, no effect was presumed and written as unchanged (↔). If data was only presented graphically, the abundance shifts were extracted as best as conceivable. The graphical representation of relative abundance shifts was limited to the phylum level to avoid over- or underreporting of data. Non-isolated control subjects were not considered in further analysis. The data were analyzed separately for groups and subgroups. Finally, data was summarized across the habitats. Secondary outcomes including anthropometric, clinical, behavioral, psychological changes were also retrieved.

For the included studies, a risk of bias assessment was performed using Risk of Bias In Non-randomized Studies of Interventions (ROBINS-I) tool (51). As only non-randomized isolation interventions were included in this systematic review, we chose the Cochrane tool as this tool views each study as an attempt to emulate a hypothetical pragmatic randomized trial and covers seven distinct domains through which bias might be introduced. In the first two domains, issues related to confounding and selection of participants are addressed before the interventions to be compared (“baseline”), while the third domain discusses intervention classification. In the remaining four domains, the following issues are addressed after the start of interventions: biases due to deviations from intended interventions, missing data, measurement of outcomes, and selection of the reported result.

The rating ranged between “Low risk,” “Moderate risk,” “Serious risk” and “Critical risk” of bias. The authors declare that “Low risk” corresponds to the risk of bias in a high-quality trial, however, due to the limited number of studies found, no study was excluded for risk of bias.

A detailed overview of the characteristics for the single trials is presented in Table 1 for cultivation methods and Table 2 for molecular-based approaches. The characteristics across the studies are summarized in the text.

The 19 included studies ranged from 1964 to 2021 and data was mainly published by Asian researchers (n = 15), otherwise from American (n = 3) or by multiple institutions working together (n = 1). Shifts in microbial abundance as a consequence of residency in space (n = 4) or long-term confinement in natural, terrestrial habitats (n = 2) have been reported. However, the alterations of the human GI microbiota by isolation was mainly studied in spaceflight- and gravity simulation facilities (n = 13 via 9 different units). There are several publications on two of these experimental units: In the case of the Mars-500 experiment, 3 publications were identified (44, 52, 53) that differed partly regarding sample size, methodology and microbiota classification. Likewise, in case of the Chinese Lunar Palace 1 experiment, 3 publications (46, 54, 55) were found showing similar differences between each other. We thus examined a total of 142 participants being mainly men (exact number unclear). Most of the studies were conducted exclusively in men (9 studies), in 5 studies both sexes were included while in 5 studies sex was not reported. The median number of subjects involved in an intervention was 4, ranging from 1 to 82, covering the ages between 21 and 50 years (median age 30.4 years). Due to the natural or artificially created extreme conditions, all study subjects were of healthy condition and predominantly of normal weight (median BMI 22.6 kg/m²). The isolation intervention lasted on average 120 days (covering 30–520 days) and was accompanied by pre- and/or post-intervention measurements in >65% (n = 13). In 10 experiments, the diet was very tightly controlled and based on a typical space diet. Furthermore, 4 studies investigated the possibility of maintaining the GI microbiota composition by providing the volunteers with supportive probiotics or prebiotics for either regular or intermittent intake. Although all research groups collected stool samples to study the microbiota composition, no study reported stool frequency and/or consistency.

A detailed overview of the in-mission outcomes for the single trials is presented in Table 3 for cultivation-based methods and Table 4 for molecular-based approaches. The outcomes across the studies are in the text and summarized in Figure 2. Further outcomes of pre- and post-mission are provided in the Supplements (Supplementary Table S1, S2).

In the cultivation studies, changes in Lactobacillus, Bifidobacterium, (potentially pathogenic) Escherichia, spore-forming bacteria, Proteus, Bacteroides, Enterococcus, Klebsiella, Citrobacter, Shigella, Corynebacterium and, in simulated emergency situations, changes in Clostridium perfringens were detected. One study found that the Bacteroides were the most numerous and stable, whereas severe fluctuations occurred in the transit microbiota (mentioned Staphylococcus and yeast) (56). Another study also reported alterations, which, however, tended to return normal after the end of the confinement (58). Finally, one study examined the aerobic and anaerobic flora in four young men over a six-week period. As the subjects had been on the space-diet for a longer period of time, differences in both the aerobic and anaerobic flora began to appear. Seven new types of organisms became prevalent, which had not been described before, mostly gas-forming, black slime-producing, and proteolytic organisms. It was assumed that the specialized space-diet (not further reported) was a contributing factor (59). Due to the different cultivation conditions, the cultivated bacterial groups observed were heterogeneous between studies.

The introduction of molecular-based approaches led to a better assessment of the microbiota allowing for a comprehensive description of microbial communities. Out of 13 studies, alpha diversity was reported to be reduced (richness: n = 3; biodiversity n = 3), to remain stable (richness: n = 4; biodiversity n = 4) or to increase (richness: n = 4; biodiversity n = 4). Richness and biodiversity were not reported separately in 3 studies (40, 56, 60). More dissimilarities were reported in microbial community structure (n = 7) with a tendency towards convergence observed repeatedly (54, 60). However, in a study that reported the differences between control subjects, the scale of microbiota changes in microbial diversity in the test subject during isolation was relatively small (25).

At the phylum level abundances predominantly remained stable following the confinement; in all but 4 of the 13 included studies (n = 14 participants) shifts in abundances were observed.

Changes reported at genus (Table 5) and species level (Supplementary Table S3) were not consistent across studies. Even more, Bacteroides (spp.), Eubacterium (spp.), Faecalibacterium (spp.), Lactobacillus (spp.), Prevotella (spp.), Alistipes (spp.), Blautia (spp.), Lachnospira, Ruminococcus (spp.), Bifidobacterium (spp.) and Clostridium (spp.) were reported to be either increasing or decreasing in their relative abundances in-mission. Across all studies, abundance shifts were predominantly associated with the phyla Bacillota (67.5%) and Bacteroidota (20.5%, Figure 2). Regardless of the taxonomic ranking, two studies reported no shifts at all (25, 55). The proportions of increasing and decreasing phyla decrease in Figures 2B,C due to the diminishing number of reporting studies.

No isolation type provided an indication that the intensity of a change is related to the length of the confinement.

Due to the heterogeneity in outcome comparisons and microbiota analysis techniques used, it was not feasible to summarize the findings clearly in figures. This led to the breakdown into groups and subgroups, as mentioned in the methods section.

Studies in space and simulation units were more similar in both genera and species profiles, while in the natural habitats the shifts were reported to be more diverse (Table 5; Figure 3). However, 8 shifts in genera or genus-specific species were shared by all three habitats. Space and simulation experiments shared shifts in 11 genera or genus-specific species.

There was a high degree of heterogeneity throughout all habitats during the intervention. It has been observed that neither species that exhibited abundance shifts across all three habitats showed the same trend. Only some Streptococcus species shifted rather consistently in abundance during the intervention (Supplementary Table S3).

Besides GI microbiota responses to extreme environmental conditions, significant reductions in body mass (25, 54, 55) as well as energy intake (64) were found. In line with these results are blood values for metabolism. Li et al. (46) and Hao et al. (54) reported that the participants’ health parameters (heart rate, blood pressure and BMI) were within the normal range throughout the study.

Furthermore, significant changes in several physiological parameters were observed, accompanied by an increase in inflammatory parameters (37, 53), changes in urinary metabolites (65), a decrease in mineral bone density and in muscle turnover (65), changes in respiratory minute volume (56), telomere elongation (25) as well as disrupted glucose metabolism (53), affected antibiotic resistance genes (40), genome instability, DNA methylation in immune and oxidative stress-related pathways (25) and finally a reduction in 25-hydroxyvitamin D levels (64). However, none of the participants demonstrated symptoms of diseases associated with considerable changes in the composition of the microbiota (44). After the isolation, 25-hydroxyvitamin D levels (64), but also mean telomere length and global gene expression converged to pre-flight levels, with an increased number of short telomeres observed and the expression of some genes still disturbed (25). However, it remains unclear if these changes are in relation to confinement or in association to microbiota shifts.

In contrast, some psychological parameters were reported to be affected, e.g., an increase in the abundance of Faecalibacterium spp. correlating negatively with mood (46). Scores of stress and anxiety seemed to be lower (63), except when in the situation of absence of natural light, where the incidence of anxiety and depression increased (55). After the isolation period of 340 days, a cognitive decline was observed (25).

However, many studies reporting different aspects of human biology and health outcomes due to isolation and confinement have been published elsewhere (42, 66) which is why only the changes that have been reported in the context of our investigation are mentioned here.

The risk of bias for the studies included in the review conducted via the ROBINS-I checklist is presented in Table 6 for all studies. Among the 19 non-RCTs, 1 was considered to have critical risk of bias due to confounding bias and 4 were at serious risk of bias. It seems that the older studies have this increased risk. Only one study was included providing a low risk of bias. A moderate risk of bias was commonly attributed to bias in selection of the reported results, which is why the highest number of studies provides a moderate risk (68%).

Overall, this systematic review has several strengths and limitations. A clear strength is the methodological approach taken according to PRISMA and Cochrane criteria. In order to obtain as broad as possible a knowledge of the current data situation, a very specific search term was used which was superior to broader search terms; however, only 19 articles could be included in the analysis. Despite clear eligibility criteria, the heterogeneity of the studies was high at the methodology (starting from sample preparation and processing) and descriptive levels. To counter this problem, subgroup analyses were performed which reduced heterogeneity to some degree. Despite differences in analysis techniques, habitats, study designs and frameworks for well-conducted studies, all studies were rather highly controlled, which is also reflected in the risk of bias. Here, the ROBINS-I tool for assessing risk of bias in non-randomized studies of interventions, recommended by the Cochrane Handbook, was used.

One of the main issues in the studies reviewed was the extremely low sample sizes, making it challenging to conduct quantitative analyses at the individual study level. Additionally, there were inconsistencies in the study protocols and characteristics, making comparisons between studies difficult. To address these limitations, future studies should aim to increase their sample sizes to enhance statistical power. Furthermore, most of the studies did neither analyze immunological/biochemical parameters in parallel to the microbiota analysis or, the data were published separately and not reported in context. To better understand the effects of long-term isolation on the human GI microbiota, researchers should consider internal and external factors, such as nutrition, genetics, and the immune system. This will enable a clearer differentiation between the effects of isolation and those stemming from other variables. Finally, this systematic review is the first of its kind, providing new insights into the effects of isolation on the human GI microbiota.