Sensitivity of Frozen Section Histology for Identifying Propionibacterium acnes Infections in Revision Shoulder Arthroplasty

Grosso, Matthew J. BS; Frangiamore, Salvatore J. MD; Ricchetti, Eric T. MD; Bauer, Thomas W. MD, PhD; Iannotti, Joseph P. MD, PhD

Journal of Bone & Joint Surgery - American Volume:
doi: 10.2106/JBJS.M.00258
Scientific Articles
Abstract

Background: Propionibacterium acnes is a clinically relevant pathogen with total shoulder arthroplasty. The purpose of this study was to determine the sensitivity of frozen section histology in identifying patients with Propionibacterium acnes infection during revision total shoulder arthroplasty and investigate various diagnostic thresholds of acute inflammation that may improve frozen section performance.

Methods: We reviewed the results of forty-five patients who underwent revision total shoulder arthroplasty. Patients were divided into the non-infection group (n = 15), the Propionibacterium acnes infection group (n = 18), and the other infection group (n = 12). Routine preoperative testing was performed and intraoperative tissue culture and frozen section histology were collected for each patient. The histologic diagnosis was determined by one pathologist for each of the four different thresholds. The absolute maximum polymorphonuclear leukocyte concentration was used to construct a receiver operating characteristics curve to determine a new potential optimal threshold.

Results: Using the current thresholds for grading frozen section histology, the sensitivity was lower for the Propionibacterium acnes infection group (50%) compared with the other infection group (67%). The specificity of frozen section was 100%. Using a receiver operating characteristics curve, an optimized threshold was found at a total of ten polymorphonuclear leukocytes in five high-power fields (400×). Using this threshold, the sensitivity of frozen section for Propionibacterium acnes was increased to 72%, and the specificity remained at 100%.

Conclusions: Using current histopathology grading systems, frozen sections were specific but showed low sensitivity with respect to the Propionibacterium acnes infection. A new threshold value of a total of ten or more polymorphonuclear leukocytes in five high-power fields may increase the sensitivity of frozen section, with minimal impact on specificity.

Level of Evidence: Diagnostic Level III. See Instructions for Authors for a complete description of levels of evidence.

Author Information

1Department of Orthopedic Surgery, Building A-41 (M.J.G., S.J.F., E.T.R., and J.P.I.), and Department of Anatomic Pathology, Building L-25 (T.W.B.), Cleveland Clinic, 9500 Euclid Avenue, Cleveland, OH 44195. E-mail address for M.J. Grosso: mgrosso13@gmail.com. E-mail address for S.J. Frangiamore:frangis@ccf.org. E-mail address for E.T. Ricchetti: ricchee@ccf.org. E-mail address for J.P. Iannotti: iannotj@ccf.org. E-mail address for T.W. Bauer: bauert@ccf.org

Article Outline

Infection in prosthetic joint surgery is a serious complication that is associated with both diagnostic and therapeutic challenges. These challenges are particularly evident in the shoulder, where established algorithms for the diagnosis of lower-extremity periprosthetic joint infection have been less effective. For example, the sensitivities in the shoulder are 42% for C-reactive protein (CRP) and 16% for an erythrocyte sedimentation rate (ESR), whereas the sensitivities in the lower extremity are 88% for CRP and 75% for ESR1,2. The discrepancy in the accuracy of these tests between the lower extremity and the shoulder is likely related to the indolent nature of the common shoulder pathogen, Propionibacterium acnes.

P. acnes is a relatively slow-growing, anaerobic, gram-positive organism, and has been identified as the most common pathogen with infection in revision shoulder arthroplasty3. It is considered part of the skin flora on most healthy adults, and it has particular affinity for areas of the shoulder, such as the axilla, in comparison with the hip and knee4. Because of its slow-growing nature, the organism often requires longer culture times (seven to fourteen days) compared with most other pathogenic bacteria, and can be missed by microbiology laboratories that discard culture plates after a few days5. Despite its indolent nature and identification as part of normal skin flora, shoulder periprosthetic joint infection caused by P. acnes can lead to devastating complications, similar to more aggressive organisms such as Staphylococcus aureus3,6,7. These infections often lead to costly and complicated surgical and medical interventions with compromised clinical outcomes.

Frozen section histology of periprosthetic tissues has been recommended for patients undergoing revision hip or knee arthroplasty for whom a diagnosis of periprosthetic joint infection has not been established or has not been excluded8,9. In the appropriate setting, frozen section can be used as a tool for identifying infection and determining surgical and medical management, including extent of removal of components and cement, one or two-stage revision, and length and type of postoperative antibiotics10. In revision hip and knee arthroplasty, frozen section has a relatively high sensitivity (77% to 95%) and specificity (92% to 96%)11,12, but criteria for frozen section diagnosis around the hip and knee have been based largely on S. aureus infections. Our previous retrospective study suggests lower sensitivity for frozen sections to diagnose periprosthetic joint infection of the shoulder compared with the lower extremity13. In addition, no studies have evaluated the performance of frozen section for P. acnes-specific infections in revision shoulder arthroplasty. We hypothesize that low virulent organisms like P. acnes induce a less vigorous inflammatory reaction, characterized by a lower tissue concentration of neutrophils, than higher virulent organisms such as S. aureus, such that the threshold for frozen section diagnosis in the shoulder might be lower than that in the hip or knee. The purposes of this study were to determine the sensitivity and specificity of frozen section histology in identifying patients with P. acnes infection during revision shoulder arthroplasty, and to investigate various morphologic thresholds that may improve frozen section performance in diagnosis of this particular infection.

Back to Top | Article Outline

Materials and Methods

With the approval of the institutional review board, our patient population was retrospectively selected from a database of all patients who underwent revision shoulder arthroplasty by two surgeons (J.P.I. and E.T.R.) at one academic institution between 2005 and 2012. All cases with periprosthetic joint infection were identified and were compared with a group of aseptic revisions. There is no gold standard to establish the diagnosis of infection in shoulder arthroplasty, and controversy exists even for patients with positive intraoperative cultures because of the difficulty in distinguishing true infection from contamination, especially for P. acnes6,13-15. Therefore, we created a spectrum of infection categories based on criteria for likelihood of infection. These criteria were most consistent with the existing shoulder literature6,16-18 (Table I). Patients were excluded if they were lacking preoperative or intraoperative data that prevented them from clearly fitting into one of these categories. For determination of sensitivity and specificity of frozen section analysis, and to be consistent with current literature, patients were considered infected if they met criteria for the probable or definite infection categories6,16-18. Patients were considered not infected if they met the criteria for the category with no evidence of infection. Infected patients were subsequently divided into P. acnes infections and infections from other organisms to create three groups: (1) the non-infection group, (2) the P. acnes infection group, and (3) the other infection group for final analysis (Table II). The groupings allowed us to examine the characteristics of P. acnes infections, while comparing them with the non-infected group and the other infection group.

The non-infection group included all fifteen patients who underwent revision shoulder arthroplasty from September 2011 to September 2012 who met the criteria for the category with no evidence of infection (Table II). All patients during this period of time had intraoperative samples submitted for frozen section and culture, no matter the preoperative indications. No patients were excluded. This allowed an accurate assessment of frozen section specificity by eliminating possible selection bias.

The P. acnes infection group consisted of all eighteen patients from 2005 to 2012 who fit the criteria for definite infection (twelve patients) or probable infection (six patients) (Table II), and for whom the only organism that grew on tissue or aspirate culture was P. acnes. Two patients were excluded because ESR, CRP, and frozen section were not obtained.

The other infection group consisted of all twelve patients from 2005 to 2012 who fit the criteria for definite infection (eight patients) or probable infection (four patients) (Table II), and for whom any organism besides P. acnes grew on culture. Patients with multiple-organism infections were placed in this category, even if one of the organisms was P. acnes. Two patients were excluded because ESR, CRP, and frozen section were not obtained.

Forty-five patients were included in this study across the three groups. The electronic medical records of all forty-five patients were reviewed for preoperative and intraoperative cultures and preoperative ESR and CRP. Because of the slow-growing nature of P. acnes, anaerobic tissue and fluid cultures for the shoulder are incubated for fourteen days at our institution. Information was also collected on patient demographics and operative procedure.

The intraoperative frozen section histology specimens for the forty-five patients were collected for reanalysis. All tissue samples that were evaluated were deep periprosthetic specimens, obtained from along the interfaces of the glenoid or humeral components or adjacent to the implants. These samples included intramedullary tissue if a humeral stem was removed, tissue off the glenoid surface and glenoid holes if a glenoid component was removed, and deep capsular tissue or deep bursal or pseudocapsular tissue in a cuff-deficient revision. None of the analyzed samples were from a superficial location. One senior anatomic pathologist (T.W.B.) analyzed all specimens and used four different thresholds to diagnose infection or non-infection (Table III)8,19. The total number of polymorphonuclear leukocytes in the five high-power fields (400×) with the highest concentration of polymorphonuclear leukocytes was determined. The anatomic pathologist was blinded to infection category during analysis of each specimen.

We compared the sensitivity of frozen section histology for the P. acnes infection group and the other infection group for each of the four diagnostic thresholds (Table III). In addition, a receiver operating characteristics (ROC) curve based on the total number of polymorphonuclear leukocytes in the five high-power fields with the highest concentration of polymorphonuclear leukocytes was utilized to determine a potential new optimal threshold. We also compared the means and sensitivities of ESR and CRP between these two groups. Institutional cutoff values were used to determine a positive ESR (>10 mm/h) and CRP (>1 mg/dL).

Chi-square analysis was used to compare sensitivities of the various tests. Student t tests were used to compare means. A p value of <0.05 was considered to be significant.

Back to Top | Article Outline
Source of Funding

There was no external funding source for this study.

Back to Top | Article Outline

Results

Preoperative Testing

The mean ESR and CRP for the non-infection, P. acnes infection, and other infection groups are reported in Table IV. For the eighteen patients in the P. acnes infection group, the sensitivity was 61% (eleven patients) for ESR and 33% (six patients) for CRP. For the twelve patients in the other infection group, the sensitivity was 83% (ten patients) for ESR and 58% (seven patients) for CRP. For the thirteen patients in both groups, the specificity was 85% (eleven patients) for both ESR and CRP.

Back to Top | Article Outline
Frozen Section Analysis

Differences in the sensitivity of frozen section histology between the P. acnes infection group and other infection group are reported in Table V. On the basis of our current institutional guidelines, for the eighteen patients in the P. acnes infection group, the sensitivity was 50% (nine patients), and for the twelve patients in the other infection group, the sensitivity was 67% (eight patients) (p = 0.15). Specificity was 100% (0 false positives).

There was minimal variation in the frozen section sensitivity for the P. acnes infection group between the previously established guidelines (Table V). An absolute polymorphonuclear leukocyte concentration of twenty-three or more in ten high-power fields19 was the most sensitive guideline for P. acnes (56%), and the American Academy of Orthopaedic Surgeons (AAOS) Guideline of ten polymorphonuclear leukocytes per high-power field for five or more fields was the least sensitive guideline (39%). There were no false positive interpretations using any of the thresholds, yielding a specificity of 100% using all four guidelines (Table V).

The total number of polymorphonuclear leukocytes in five high-power fields was used to create an ROC curve (Fig. 1). The majority of specimens (60% [nine patients]) in the fifteen patients in the non-infection group had zero polymorphonuclear leukocytes in five high-power fields, and the highest number of polymorphonuclear leukocytes was six. The ROC indicates that any cutoff from seven to ten polymorphonuclear leukocytes will improve sensitivity of frozen section histology for P. acnes infections (72%), while maintaining 100% specificity (Fig. 1).

Back to Top | Article Outline

Discussion

Identifying infection in revision total shoulder arthroplasty is critical for determining appropriate medical and surgical management. The slow-growing P. acnes is a unique, but common, shoulder pathogen. Evidence suggests that the indolent nature of this organism inhibits the ability of common diagnostic tests to identify infection. Intraoperative frozen section histology is particularly important because of the potential immediate impact on medical and surgical management. The objectives of this study were to calculate the sensitivity and specificity of frozen section histology for patients with a P. acnes infection based on histologic guidelines that have been advocated for diagnosing periprosthetic infections of the hip and knee, and then to calculate a new threshold that might improve the sensitivity and specificity of frozen section diagnosis of periprosthetic shoulder infections. Our results showed that the guidelines recommended by the AAOS Practice Guidelines Committee for Diagnosis of Periprosthetic Joint Infections of the Hip and Knee8 had high specificity (no false positives) but relatively low sensitivity when applied to revision shoulder arthroplasty. Our ROC curve analysis suggests that a new threshold of a total of ten or more polymorphonuclear leukocytes in five high-power fields may help increase the sensitivity of frozen sections to diagnose the commonly indolent periprosthetic infections of the shoulder with minimal impact on specificity.

In revision shoulder arthroplasty, results from frozen section histology can change medical and surgical management. Although the performance of frozen section in the lower-extremity joints is well studied11,12, few studies have evaluated the sensitivity in the shoulder, and no studies have examined the performance of frozen section in P. acnes infections. Our data support a low sensitivity of frozen section in P. acnes infections (50%) using the diagnostic thresholds currently recommended for revision hip and knee arthroplasty, and highlight the need for improved tests for identifying infection with this pathogen. On the basis of our data as well as previous studies, ESR, CRP, and preoperative aspirate cultures also have lower sensitivity for diagnosing periprosthetic shoulder infections than when those tests are applied to periprosthetic knee or hip infections.

Poor frozen section performance can impact surgical and postoperative medical management. For example, a patient with negative preoperative signs of infection (normal ESR, CRP, negative aspirate) will be assumed to have aseptic failure and, without evidence of acute inflammation on frozen section, will undergo a one-stage revision. If positive tissue culture occurs late, as is common for P. acnes, antibiotic therapy may be considerably delayed, usually after a patient’s discharge. At our institution, treatment for later positive culture results includes six weeks of intravenous antibiotics, followed by potential extended oral suppressive antibiotic therapy. This situation is not uncommon. Indeed, for our eighteen patients with P. acnes infection, 39% (seven patients) underwent a one-stage procedure followed by delayed medical management (Table II). The length of delay before initiation of intravenous antibiotic therapy was based on P. acnes culture times and ranged from four to eight days (mean, 5.2 days), but can be even longer for this slow-growing organism5. For one patient, culture time took twenty-one days, but intravenous antibiotics were initiated earlier because of gross intraoperative signs of infection. The long-term impact of this potential delay in treatment is not known, nor is the potential difference in long-term outcome between this approach and a two-stage reimplantation that would be performed if the diagnosis of infection was confirmed at the time of surgery.

Our new proposed threshold of ten polymorphonuclear leukocytes per five high-power fields calculated from the ROC curve resulted in the frozen section for a P. acnes infection having an increased sensitivity (73%) compared with its previous sensitivity (50%). We believe that use of this new criterion has utility for intraoperative decision-making that ideally will lead to improved outcomes. The presence of a positive frozen section, depending on the other associated clinical data, may lead the treating surgeon to maintain antibiotic therapy postoperatively until culture results are final to avoid a delay in treatment, or it may be the final piece of evidence to convince a surgeon to perform a two-stage exchange. In addition, by maintaining a high specificity (100%) with this new threshold, the risk of these potential additional treatments (extended antibiotic therapy and/or two-stage exchange) being unnecessary remains low. Finally, it should be noted that the predictive value of any laboratory test, including frozen section interpretation, reflects in part the prevalence of a condition. For example, although our specificity of 100% reflects no false-positive results, that result might be different if the population undergoing frozen sections was different.

We chose to create categories of infection based on likelihood because of the lack of consensus in defining infection in shoulder arthroplasty and the controversy that exists in labeling P. acnes positive cultures as genuine or a contaminant. For this study, we chose to only include those patients with a high probability of having true infections to reduce the probability that the lower sensitivity of frozen section in the P. acnes infection group was a result of increased false-positive infections. Therefore, we included patients who were classified as definite infection and probable infection. As there is no gold-standard criterion for infection, it was reasonable and most consistent with the shoulder literature for us to use these categories6,16-18. We further analyzed the performance of frozen section in patients with an extremely high likelihood of infection (definite infection only). We again found an increased sensitivity in the P. acnes infection group between our current institutional criteria (66%) and the new threshold criteria (83%) (data not shown).

Because this was a retrospective study, we recommend caution in using this new threshold to decide one-stage or two-stage revision with no other supporting criteria. A prospective study with a larger number of both infected and non-infected cases is needed. The risks and benefits of a two-stage revision also need to be balanced with a better understanding of the natural history of P. acnes in patients with positive revision shoulder arthroplasty. For example, in a previous study with a two-year follow-up, we found a low recurrence rate of infection for patients with at least one positive intraoperative P. acnes culture and no other signs of infection who did not undergo a two-stage revision13. However, the patient population in that study would fit mostly into the possible infection category (Table I) and, therefore, could not be appropriately matched with the P. acnes infection group used in this study. This study may help to further define this category of patients. The presence of acute inflammation in periprosthetic tissue can support the diagnosis of infection in cases with one unexpected positive culture. If the frozen section (and permanent histology) does not show acute inflammation, then it further supports that a single unexpected positive culture after revision may not require further treatment.

Periprosthetic infection with P. acnes following shoulder replacement surgery can be a devastating complication3,6,7. Early identification of a joint infection is essential in guiding surgical and medical management. Frozen section histology can be a helpful tool for identifying infection intraoperatively. We found a high specificity of frozen section, but a low sensitivity for those patients with P. acnes infection when using currently recommended diagnostic thresholds. We improved sensitivity by utilizing a new histologic threshold value of ten polymorphonuclear leukocytes per five high-power fields, while maintaining 100% specificity.

Investigation performed at the Cleveland Clinic, Cleveland, Ohio

A commentary by Pierre Mansat, MD, PhD, is linked to the online version of this article at jbjs.org.

Back to Top | Article Outline

References

1. Berbari E Mabry T Tsaras G Spangehl M Erwin PJ Murad MH Steckelberg J Osmon D. Inflammatory blood laboratory levels as markers of prosthetic joint infection: a systematic review and meta-analysis. J Bone Joint Surg Am. 2010 Sep 1;92(11):2102–9.
2. Piper KE Fernandez-Sampedro M Steckelberg KE Mandrekar JN Karau MJ Steckelberg JM Berbari EF Osmon DR Hanssen AD Lewallen DG Cofield RH Sperling JW Sanchez-Sotelo J Huddleston PM Dekutoski MB Yaszemski M Currier B Patel R. C-reactive protein, erythrocyte sedimentation rate and orthopedic implant infection. PLoS One. 2010;5(2):e9358. Epub 2010 Feb 22.
3. Sperling JW Kozak TK Hanssen AD Cofield RH. Infection after shoulder arthroplasty. Clin Orthop Relat Res. 2001 Jan;(382):206–16.
4. Patel A Calfee RP Plante M Fischer SA Green A. Propionibacterium acnes colonization of the human shoulder. J Shoulder Elbow Surg. 2009 Nov-Dec;18(6):897–902. Epub 2009 Apr 11.
5. Brook I Frazier EH. Infections caused by Propionibacterium species. Rev Infect Dis. 1991 Sep-Oct;13(5):819–22.
6. Dodson CC Craig EV Cordasco FA Dines DM Dines JS Dicarlo E Brause BD Warren RF. Propionibacterium acnes infection after shoulder arthroplasty: a diagnostic challenge. J Shoulder Elbow Surg. 2010 Mar;19(2):303–7. Epub 2009 Nov 1.
7. Cofield RH Edgerton BC. Total shoulder arthroplasty: complications and revision surgery. Instr Course Lect. 1990;39:449–62.
8. Della Valle C Parvizi J Bauer TW Dicesare PE Evans RP Segreti J Spangehl M Watters WC 3rd Keith M Turkelson CM Wies JL Sluka P Hitchcock K; American Academy of Orthopaedic Surgeons. Diagnosis of periprosthetic joint infections of the hip and knee. J Am Acad Orthop Surg. 2010 Dec;18(12):760–70.
9. Bauer TW Parvizi J Kobayashi N Krebs V. Diagnosis of periprosthetic infection. J Bone Joint Surg Am. 2006 Apr;88(4):869–82.
10. Bauer TW Hayashi R. The role of the pathologist in diagnosing periprosthetic infection. Surg Path Clin. 2012 Mar;5(1):67–77.
11. Wong YC Lee QJ Wai YL Ng WF. Intraoperative frozen section for detecting active infection in failed hip and knee arthroplasties. J Arthroplasty. 2005 Dec;20(8):1015–20.
12. Lonner JH Desai P Dicesare PE Steiner G Zuckerman JD. The reliability of analysis of intraoperative frozen sections for identifying active infection during revision hip or knee arthroplasty. J Bone Joint Surg Am. 1996 Oct;78(10):1553–8.
13. Grosso MJ Sabesan VJ Ho JC Ricchetti ET Iannotti JP. Reinfection rates after 1-stage revision shoulder arthroplasty for patients with unexpected positive intraoperative cultures. J Shoulder Elbow Surg. 2012 Jun;21(6):754–8. Epub 2012 Feb 3.
14. Kelly JD 2nd Hobgood ER. Positive culture rate in revision shoulder arthroplasty. Clin Orthop Relat Res. 2009 Sep;467(9):2343–8. Epub 2009 May 12.
15. Topolski MS Chin PYK Sperling JW Cofield RH. Revision shoulder arthroplasty with positive intraoperative cultures: the value of preoperative studies and intraoperative histology. J Shoulder Elbow Surg. 2006 Jul-Aug;15(4):402–6.
16. Zeller V Ghorbani A Strady C Leonard P Mamoudy P Desplaces N. Propionibacterium acnes: an agent of prosthetic joint infection and colonization. J Infect. 2007 Aug;55(2):119–24. Epub 2007 Apr 5.
17. Trampuz A Osmon DR Hanssen AD Steckelberg JM Patel R. Molecular and antibiofilm approaches to prosthetic joint infection. Clin Orthop Relat Res. 2003 Sep;(414):69–88.
18. Piper KE Jacobson MJ Cofield RH Sperling JW Sanchez-Sotelo J Osmon DR McDowell A Patrick S Steckelberg JM Mandrekar JN Fernandez Sampedro M Patel R. Microbiologic diagnosis of prosthetic shoulder infection by use of implant sonication. J Clin Microbiol. 2009 Jun;47(6):1878–84. Epub 2009 Mar 4.
19. Morawietz L Tiddens O Mueller M Tohtz S Gansukh T Schroeder JH Perka C Krenn V. Twenty-three neutrophil granulocytes in 10 high-power fields is the best histopathological threshold to differentiate between aseptic and septic endoprosthesis loosening. Histopathology. 2009 Jun;54(7):847–53.

Disclosure: None of the authors received payments or services, either directly or indirectly (i.e., via his or her institution), from a third party in support of any aspect of this work. One or more of the authors, or his or her institution, has had a financial relationship, in the thirty-six months prior to submission of this work, with an entity in the biomedical arena that could be perceived to influence or have the potential to influence what is written in this work. No author has had any other relationships, or has engaged in any other activities, that could be perceived to influence or have the potential to influence what is written in this work. The complete Disclosures of Potential Conflicts of Interest submitted by authors are always provided with the online version of the article.

Copyright 2014 by The Journal of Bone and Joint Surgery, Incorporated