The shoulder microbiome: a systematic review and meta analysis

Abstract

Background

The shoulder microbiome is an emerging field in orthopedic research. Large synovial joints which were typically considered to be sterile now have strong evidence demonstrating the presence of native organisms colonizing the joint. Many of the conditions that surgeons see and treat in the shoulder, including osteoarthritis, rotator cuff tears, and adhesive capsulitis, have unclear etiology. The shoulder microbiome is a potential source of pathology in dysbiosis states, or source of protection against pathology in normobiosis states. The purpose of this review is to characterize the published research detailing the microbiome of the native shoulder joint.

Methods

We conducted a systematic review and meta-analysis of data searches on Medline, Embase, Cochrane Central Register of Controlled Trials, and SCOPUS. The following search terms were used with various permutations; shoulder, skin, microbiome, infection, colonization, commensal, microbiota, flora. The terms ‘shoulder’ and ‘shoulder AND skin’ were combined with the other 6 terms for a total of 12 searches per database. Two independent reviewers conducted the review with a third reviewer available to resolve differences of interpretation. Studies were eligible if they were human studies of patients undergoing shoulder surgery (or surgical preparation therapy) with a shoulder that had not previously been operated on, where microbial samples were reported. Studies were excluded if they were systematic reviews and meta-analyses, animal studies, cadaveric studies, studies of patients under the age of 18 and studies including postoperative shoulders.

Results

The search methodology yielded 47 eligible studies for analysis, with a total of 3283 patients. Native shoulder sampling yielded positive organisms in 50.8% of all samples (95% confidence interval [CI] = 41.8%-59.8%, I² = 97.9%, P = .000). The predominate organism reported was C. acnes, with 29% of studies specific for C. acnes, and a total of 52 other organisms detected in the native shoulder. Skin preparation solutions reduced the skin colonization rate of 55.7% (95% CI = 32.7%-78.6%, I² = 98.2%, P = .000) to 43.5% (95% CI = 29.5%-57.5%, I² = 98.2%, P = .000). Despite the high rates of colonization detected under sterile conditions, of 42 studies and 3083 patients reporting postoperative outcomes, the infection rate was 1.8% (95% CI = 0.5%-3.2%, I² = 0.0%, P = .612).

Conclusion

There is strong evidence to support the existence of a native shoulder microbiome. Further research is required to characterize this microbiome and correlate it to disease and health states.

The human shoulder is a unique set of articulations that are structured to provide humans with the perfect balance between mobility and stability for upper-limb function.⁷⁴ The corollary of this is that the shoulder is susceptible to degeneration and imbalance with resultant clinically important deficits in function.^8,25,56 The major syndromes of shoulder degeneration including rotator cuff tears, osteoarthritis and adhesive capsulitis, are well-described, but their etiology remains unclear.^28,65 An emerging avenue of inquiry is the shoulder microbiome.^60,62 The terms ‘microbiome’ and ‘microbiota’ have been used interchangeably, they refer to all of the organisms present in a specific habitat, including bacteria, fungi, viruses, yeast, and mites.² The term ‘microbiome’ specifically refers to the genetic material (DNA or RNA) and genome of the microbial community. Humans are intrinsically linked to their microbiome; the human body consists of approximately 3 x 10¹³ human cells, and the number of organisms on the human skin and mucous membranes numbers 3.8 × 10¹³.⁷⁵ The human skin has over three million micro-organisms per square centimeter and is fertile ground for the controlled growth of bacteria, most of which are considered to be beneficial to the host.^12,49,64 The skin surface bacteria exist in a state of flux, there are competing communities of bacteria that can be approximately grouped into different categories. Commensal bacteria can be subdivided into resident and transient commensals, which are part of a balanced microbiota, whereas pathogenic bacteria by definition cause harm.⁴⁹ Deep to the skin, the conventional idea that the tissues of the musculoskeletal system exist within a sterile environment has been challenged by studies demonstrating the presence of organisms in the microbiome of the native spine⁵² and the shoulder.⁵¹ Current understanding of the microbiome has been aided by the evolution of Next-Generation Sequencing (NGS) technology which can sequence entire bacterial genomes from samples of tissues or fluid.⁴² NGS is an unbiased and useful tool in the assessment of microorganisms in specimens, however it cannot distinguish specimens that are live-colonizing organisms from those that are dead transient organisms.² The microbiome has a diverse composition, and fungal colonization is readily detected with culture and NGS;⁴ however, due to the predominance of bacterial results reported in the shoulder literature, we chose to keep the focus of the studies in this review specific to bacterial colonization of the shoulder.

For the practice of shoulder surgery, it is critical to be able to define which organisms are pathogenic and where these organisms reside; surgical site infections are estimated to add a cost of $3000 to $29,000 and increase the length of hospital stay by 7 days on average per infection.⁷⁵ The periprosthetic infection for shoulder arthroplasty has devastating consequences for the patient, the surgeon, and the health-care system.^34,37 Tens of thousands of shoulder arthroplasty infections are potentially caused by Cutibacterium acnes every year.²⁶ The etiology of shoulder surgical infection is difficult to study as infection is a relatively uncommon occurrence after shoulder surgery.¹² This may lead to the failure of even large-scale randomized studies to demonstrate a meaningful difference between treatment groups.⁹

The composition of the shoulder microbiome has potential implications on the health and disease states of the shoulder joint. The shoulder microbiome is not proven as an entity in the deep tissues or well-defined in the literature; the purpose of this study is to systematically review published articles pertaining to the shoulder microbiome and its composition. We aim to establish what is currently known of the composition of the shoulder microbiome, as well as identifying any trends in terms of the microbiome and shoulder degeneration. The aim of this meta-analysis is to identify what evidence there is for a shoulder microbiome, and identify the prevalence of clinically significant infection in the presence of positive microbiome sampling.

Materials and methods

Preferred Reporting Items for Systematic Review and Meta-Analyses (PRISMA)

This systematic review was performed using the PRISMA framework (Fig. 1).

Figure 1.

PRISMA flowchart of the search process for this systematic review. PRISMA, Preferred Reporting Items for Systematic Review and Meta-Analyses.

Search strategy

Two independent researchers reviewed the results of searches on the following databases; Medline, Embase, Cochrane Central Register of Controlled Trials, and SCOPUS. The following search terms were used with various permutations; shoulder, skin, microbiome, infection, colonization, commensal, microbiota, flora. The terms ‘shoulder’ and ‘shoulder AND skin’ were combined with the other 6 terms for a total of 12 searches per database.

Inclusion criteria were selected for human studies, patients undergoing shoulder surgery (or surgical preparation therapy) with a shoulder that had not previously been operated on, where microbial samples were reported. Exclusion criteria included systematic reviews and meta-analyses, animal studies, cadaveric studies, studies of patients under the age of 18 and studies including postoperative shoulders. The initial search was on January 1, 2022, and the final search was on October 1, 2023.

Data extraction

Data extraction was performed by 2 independent reviewers (RN and XC), with a third reviewer (GS) consulted for any differences of opinion. The reviewers collected the following data: methods (first author’s name, publication year, study design, sample size, mean duration of follow-up), participants (number of participants, age, gender), interventions (surgical procedure), and outcomes (primary outcome: organism detection and secondary outcome: clinically significant postoperative infections).

Statistical analysis

This meta-analysis was performed according to the guidelines from the Quality Reporting of Meta-Analyses Group and the Meta-analysis of Observational Studies in Epidemiology Group. The meta-analysis was reported in accordance with AMSTAR (Assessing the Methodological Quality of Systematic Reviews) and PRISMA guidelines. Data were separated into categories according to the sampling region; skin swabs and surface biopsies were analyzed separately to surgical wounds with deeper tissue sampling. We conducted a meta-analysis of proportions for the available organism results. Meta-analysis results were presented and interpreted with the aid of forest plots. The Chi-squared statistic was used to measure heterogeneity between studies. We used a fixed-effects model and random-effects model according to homogeneity (I2 < 50%) and heterogeneity (I2 > 50%) respectively. Sensitivity analysis for outcomes was performed using the leave-one-out approach to assess the robustness of the results. Subgroup analysis was not performed due to the potential bias of included studies. We calculated weighted means by dividing the total number of organism positive samples by the total sample size. The pooled mean infection rate was assessed according to the total number of patients who sustained a clinically significant infection divided by the overall number of patients undergoing surgery with studies reporting infection rates. Statistical significance was set at 5% (α = 0.05). Statistical analysis was performed using STATA version 15 (StataCorp LLC, College Station, TX, USA).

Results

Study selection

The search methodology yielded 6623 results, which were further screened to 122 full-text articles which were read in entirety, and yielded 47 eligible studies for analysis (Fig. 1).^{1,3,5,7,10,11,13,14,}17, 18, 19^,22, 23, 24^{,27,33,36,38,41,43,}53, 54, 55^,57, 58, 59^{,61,63,67,71,72}

Study characteristics

A total of 3283 patients were found in the eligible studies. Three studies reported on the culture and NGS results for patients, while 44 studies reported on culture results alone. All randomized studies included were assessed as having a low risk of bias.⁶⁶

Microbial results

Figure 2 demonstrates the overall positive pooled effect of the studies, indicating that native shoulder sampling yields positive organisms in 50.8% of all samples (95% confidence interval [CI] = 41.8%-59.8%; I² = 97.9%, P = .000). Of these native shoulder sampling studies, subgroup analysis is demonstrated in Figure 3 where skin-only sampling organism results are compared to surgical wound organism sampling results (I² = 90.0%, P = .000). There was a selection bias in organism type identified in the studies; 29% of included studies were specific for Cutibacterium acnes (C. acnes). Of the remaining 71% of studies, 34% reported the presence of C. acnes, and 37% reported the presence of alternate organisms (Fig. 4). The 52 organisms other than C. acnes identified on and in the native human shoulder are listed in Table I. There was a difference in the rates of organism identification in patients who were sampled with and without skin preparation solutions as demonstrated in Figure 5. All of the included studies regarding skin preparation were designed to specifically target C. acnes, and used the resultant rates of C. acnes eradication as a primary outcome measure. Patients who had sampling without the use of skin preparation solutions had a skin colonization rate of 55.7% (95% CI = 32.7%-78.6%, I² = 98.2%, P = .000). Patients who had sampling after the use of skin preparation solutions had a colonization rate of 43.5% (95% CI = 29.5%-57.5%, I² = 98.2%, P = .000). When looking at clinically significant infections, there were 42 studies and 3083 patients who contributed to the reported pooled prevalence. The overall infection rate was 1.8% (95% CI = 0.5%-3.2%, I² = 0.0%, P = .612) and came from positive cases in 5 studies (Fig. 6). One patient sustained a deep C. acnes infection following reverse total shoulder arthroplasty 6 months postoperatively, and required a 2-stage revision. This sampling was performed after the patient received a standard combination of intravenous cefazolin, vancomycin, and topical 2% chlorhexidine gluconate and 70% isopropyl alcohol.⁷¹ One patient sustained a superficial infection with C. acnes and Staphylococcus aureus after acromioclavicular joint stabilization, which required repeat surgery. This patient had a similar combination of cefazolin and topical 2% chlorhexidine gluconate and 70% isopropyl alcohol.⁵⁰ Two patients sustained a deep infection with C. acnes requiring revision surgery after an open shoulder procedure via either a deltopectoral or anterolateral approach to the shoulder. The patients had no intravenous antibiotics before sampling and had skin preparation with an alcoholic preparation solution.²³ Five patients in one study sustained deep infections requiring surgical débridement. The procedures were open shoulder procedures via either a deltopectoral or superior approach. All patients had topical skin preparation at the time of surgery with 0.5% chlorhexidine gluconate and 70% isopropyl alcohol. Three of these patients were treated with preoperative benzoyl peroxide (BPO) and 2 of these patients used soap wash in the preoperative period only. Of the BPO patients, two had positive deep cultures for Staphylococcus aureus, and the third patient had S epidermidis and C. acnes. The 2 soap patients both had deep specimen growth of C. acnes only. One patient undergoing shoulder arthroplasty via an open approach had a deep infection with C. acnes which required repeat surgery. This patient had preoperative skin preparation with 2% chlorhexidine gluconate and 70% isopropyl alcohol but was not randomized to have topical BPO, in a study which had a treatment group with BPO topical application.³ The comparative use of traditional culture methods and NGS was observed in three studies (Fig. 7). The relative risk of positive organism results with NGS was 1.33 (95% CI = 1.03-1.71, I² = 67.4%, P = .046).

Figure 2.

Forest plot of native shoulder specimen organisms for all eligible studies. CI, confidence interval.

Figure 3.

Forest plot comparing superficial against deep sampling methods for native shoulder specimen organisms for all eligible studies. CI, confidence interval.

Figure 4.

Pie graph demonstrating the breakdown of native shoulder specimen organism results.

Table I.

All organisms reported in the native shoulder in searched literature.

All other organisms identified in the native shoulder
Acinetobacter iwoffii	Micrococcus luteus
Actinomyces	Micrococcus species
Bacillus cereus	Moraxella
Bacillus magaterium	Morganella morganii
Bacillus species	Mycobacterium
Beta-hemolytic Streptococcus	Oxalobacteraceae unclassified
Bifidobacterium	Pantoea septica
Brachybacterium muris	Peptostreptococcus
Burkholderia species	Pseudomonas species
Clostridium perfringens	Ralstonia
Coagulase-negative Staphylococcus	Roseomonas
Corynebacterium afermentans	Rothia
Corynebacterium tuberculostearicum	Scopulariopsis
Corynebacteruim species	Staphlococcus aureus
Cutibacterium avidum	Staphylococcus capitus
Dermabacter hominis	Staphylococcus caprae
Dermacoccus species	Staphylococcus cohnii
Diptheroids	Staphylococcus cristatus
Enterobacter aerogenes	Staphylococcus epidermidis
Enterococcus faecalis	Staphylococcus hominis
Erythrobacter acea	Staphylococcus lugdensis
Escherichia coli	Staphylococcus saccharolyticus
Eubacterium	Staphylococcus saprophyticus
Granulicatella	Staphylococcus warneri
Kocuria varians	Streptococcus mitis
Lactobacillus	Streptococcus viridans

Figure 5.

Forest plot comparing studies which obtained native shoulder specimens with skin preparation and without skin preparation. CI, confidence interval.

Figure 6.

Forest plot of subgroup of studies which had positive findings for clinically significant infections in patients who had sampling of the native shoulder organisms. CI, confidence interval.

Figure 7.

Comparison forest plot of studies which obtained native shoulder specimens and then processed the specimens with traditional culture and next generation gequencing. CI, confidence interval.

Discussion

The studies that characterize the bacteria of the deep and superficial shoulder microbiome are summarized in Figure 2, the organisms which we found in these studies are listed in Table I. This represents the most complete record in the literature characterizing the breadth of the native human shoulder microbiome. There is significant heterogeneity in these studies in terms of methodology; they vary by type of preparation before surgery, by procedure, type of sample, sample size, culture or genetic sequencing methods. The study results we reported in terms of bacterial prevalence are concomitantly variable. The findings in this study indicate that there is a microbiome in the sterile surface environment of the shoulder and there is a possible microbiome in the deep environment of the shoulder, although this has not been proven. Our study found that NGS detected more organisms in the native shoulder microbiome than traditional culture methods (Fig. 7), and organism yield was less when deeper sampling was performed (Fig. 3) and after the application of skin preparation solutions (Fig. 5). Debate continues to circulate regarding whether positive bacterial cultures and ribosomal ribonucleic acid sequencing from the shoulder joint are the result of occult bacterial infection, colonization of the joint, or contamination during the surgical approach.

The most commonly found organism in the shoulder microbiome is Cutibacterium acnes. C. acnes is a unique pathogen⁶⁷ which is associated with a higher rate of deep postsurgical infections in the shoulder than any other organism.^46,50,58 C. acnes is considered to be a marker organism which is responsible for 51%-56% of all infections after rotator cuff surgery.⁷⁵ C. acnes avoids recognition by host immune cells, which increases the risk of occult periprosthetic infection occurring.¹⁵ It promotes bacterial adhesion by the formation of a biofilm, and has virulence factors that trigger an inflammatory response.^40,73 C. acnes can be classified by phylotypes into type IA1, IA2, IB, IC, II and III, which display different degrees of inflammation and virulence.^21,30,44 Phylotype IB of C. acnes has the lowest proinflammatory potential, and phylotype III has the highest proinflammatory potential.¹⁵ Phylotype IB is the most common pathogen in orthopedic infections.⁴⁴

The colonization of the shoulder with C. acnes is greater for chest and back than the shoulder and axilla due to higher numbers of sebaceous glands on the chest and back.³¹ Increased sebum is associated with increased levels of Cutibacterium species including C. acnes.⁴⁹ Male patients have repeatedly been shown to produce more positive culture results for C. acnes than women.^35,47,48,64 Males are typically more hairy than females and a higher prevalence of hairy skin is associated with a higher number of sebaceous glands, which may account for this increased prevalence of C. acnes in males.^16,45,48 Males are also more likely to sustain a deep postoperative infection with C. acnes, which is likely related to the higher baseline burden of C. acnes in men.¹⁶ C. acnes has been linked to increased testosterone, which is theorized to increase output of sebum from the pilosebaceous glands, which facilitates C. acnes colonization.^44,68 There are lower rates of cultured C. acnes in patients of Asian descent, which has been hypothesized as being due to Asian people having a relatively lower follicular density.²⁹ This variation needs to be taken into account when analyzing literature from Asia and Western countries.¹⁹

C. acnes is indolent in nature and therefore presents a unique diagnostic challenge for Shoulder Surgeons. C. acnes does not share the same typical clinical presentation of pathogens such as Staphylococcus aureus, it is more likely to present as vague pain, shoulder stiffness or component loosening.^1,32,46 This subtle presentation is often misdiagnosed or unsuspected, with associated poor outcomes.⁷⁰ C. acnes has been implicated in the development of autoimmune conditions such as sarcoidosis, SAPHO syndrome (synovitis, acne, pustulosis, osteitis) and chronic recurrent multifocal osteomyelitis. There is a positive correlation between the microbial community of C. acnes swabs taken from the face and the shoulder indicating a variable effect of C. acnes based on location.⁴⁹ Successful antibiotic therapy for patients suffering from SAPHO syndrome confirm that C. acnes has a role in the development of the condition, which is characterized by an immune response to C. acnes or other colonizing organisms.⁶

The studies included in this review support the theory that there is a deep shoulder microbiome, and also lend support to the theory that this microbiome has significance for the health of the shoulder. Hsu et al (2020) obtained skin and deep tissue cultures from 25 patients undergoing primary shoulder arthroplasty and 27 patients undergoing revision shoulder arthroplasty. They found no significant difference in the C. acnes load of the skin swabs from patients undergoing primary arthroplasty compared to revision arthroplasty (P = .512). DNA extraction, sequencing and subtyping was performed on the C. acnes organisms that were cultured from the shoulder swabs. There was a significantly higher prevalence of C. acnes subtype A in revision shoulder arthroplasty patients (36.9%) compared to the primary shoulder arthroplasty group (16%, P =.002).²⁰ This supports the theory that an abundance of certain subtypes of C. acnes, or ‘dysbiosis’ can be associated with an increased risk of disease in organ systems. Certain bacterial species may be opportunistic in the setting of dysbiosis, and the restoration of normobiosis can be protective against pathological bacterial subtypes.²⁰ Our study did not differentiate between C. acnes phylotypes due to a low number of studies reporting on the phylotypes and therefore cannot comment on whether clinically significant infections were related to certain C. acnes phylotypes.

Patzer et al performed a prospective, randomized, C. acnes detection study on 115 patients undergoing shoulder arthroscopy. The study was designed such that perioperative antibiotics were given after the skin incision so as to avoid attrition of the bacterial load in the samples. Patients were randomized by sealed envelope to either a subacromial or glenohumeral joint space approach as the first arthroscopic portal, and first area of sampling. There was a significantly higher intraoperative sample positive C. acnes rate for the glenohumeral joint (11 of 58 patients positive, 18.9%) compared to the subacromial space (2 of 57 patients positive, 3.5%, P = .016). The authors concluded that the significantly different culture results in each compartment, separated by the intact rotator cuff, goes against the contamination-by-approach theory, and supports the presence of a deep shoulder microbiome.⁴⁷ The proposed model of surgically driven inoculation is that the surgical blade disrupts the dermis and releases bacteria onto all instruments, allowing the surgeon to spread the bacterium to the deeper tissue layers.^39,40 Variations of this model posit skin retraction, soft-tissue handling, and patient sweating during surgery as methods of bacterial inoculation.⁶⁹ Our study presents deep shoulder specimen data without a true indication of whether there has been partial or complete contamination of the microbiome by approach. A sophisticated research protocol will need to be implemented to definitively answer the question of the composition of the native deep shoulder microbiome with the elimination of the possibility of contamination.

Our study found only 11 clinically significant infections in 5 studies, (Fig. 6) despite consistently high levels of bacterial positive samples obtained from shoulder surgery wounds across all studies. This cannot be taken as evidence in support of the presence of a deep shoulder microbiome; these infections may be from contamination-by-approach, and the low numbers of infection may be a result of commensal bacterial phylotypes or other host factors which are undefined. When true infection does occur, etiologic theories include dysbiosis, a low microbial efficacy of antiseptic agents, a failure of the antiseptic agents to reach pathogens in deeper skin layers, glands or follicles, and the recontamination of wounds following antisepsis by means of direct contact or aerosols. The implication from this review is that surgical sites that are meticulously prepped are clean, but never truly sterile due to the presence of the microbiome in the deep tissues.⁷⁵

Conclusion

The shoulder microbiome is described piecemeal in the literature with heterogenous study designs and variable sampling protocols. NGS is an emerging way to identify the shoulder microbiome, but we have not yet ascertained the complete picture of the shoulder microbiome and its implications on shoulder health and disease. The overarching question of the veracity of deep shoulder sampling results remains. This review was specific to bacterial species; there are further layers of the microbiome in other taxonomic groups that also need to be explored. Further high-quality research and technology is required to elucidate the full nature of shoulder microbiome in different shoulder pathology populations and identify correlations between disease states and the composition of the microbiota. This correlation should extend to patient-reported outcome measures as well as advanced imaging of the shoulder joint, as the information from these tools will be additive to that of the shoulder microbiome, and will aid in completing our understanding of the human shoulder and its various states of disease, and health.

Disclaimers:

Funding: The first author received a grant from the Australian Orthopaedic Association Research Foundation (AOARF).

Conflicts of interest: Dr. Rajpal S Narulla is supported by an Australian Government Research Training Program Scholarship; there was no specific grant for this study or direct funding for any data collection, data analysis, or the preparation or editing of the manuscript. All the other authors, their immediate families, and any research foundations with which they are affiliated have not received any financial payments or other benefits from any commercial entity related to the subject of this article.