The Investigational Review Board (IRB) at our institution approved this study (IRB #38891).
Patients
This study included all 193 revision shoulder arthroplasties that were performed at our center by either of two attending surgeons from May 2006 to January 2011 for the treatment of stiffness, pain, or loosening of existing arthroplasty components. Cases of revision of previously placed methylmethacrylate spacers, drainage of postoperative hematomas, treatment of periprosthetic fractures, and treatment of acute joint sepsis or cellulitis were excluded so that our analysis could be confined to cases of revision arthroplasty without obvious clinical evidence of infection.
The medical records of each patient were carefully analyzed to determine the results of intraoperative bacterial cultures along with the factors known to the surgeon before and during surgery that might correlate with the results of cultures for Propionibacterium acnes and/or other organisms (see Appendix).
Bacterial Cultures
In each case, tissue culture specimen samples (average, 3.8; range, one to ten) were obtained from tissues representative of the different aspects of the surgical field. Specimens were carefully handled with sterile instruments and were immediately placed in sterile specimen containers to minimize any risk of contamination. All specimens were processed by the laboratory within one hour after surgery in a class-2 laminar flow biological safety cabinet. Fluid and homogenized tissue specimens were inoculated onto the following microbiological media: blood agar (trypticase soy agar with 5% sheep blood), chocolate agar, Brucella agar (with blood, hemin, and vitamin K), and brain-heart infusion broth. All media, with the exception of the Brucella agar, were incubated at 37°C with 5% CO2 for twenty-eight days. Brucella agar plates were incubated anaerobically at 37°C for twenty-eight days. Plates were sealed in a manner that allowed sterile aeration without desiccation. Media were examined daily for growth visually but were only opened if growth was noted, again with great care being taken to avoid any risk of contamination. All bacteria that were isolated received a full species-level identification by means of 16S rDNA sequencing, as described previously47,62.
Statistical Analysis
The two primary outcome variables for this study were the final results of the bacterial cultures for (1) Propionibacterium acnes and (2) other organisms.
Potential prognostic factors for the culture results were grouped with respect to the sequence in which they would become known to the surgeon: Stage 1 included preoperative characteristics (demographic characteristics, comorbidities, blood laboratory tests, and radiographic findings), Stage 2 included intraoperative characteristics (prosthetic loosening, presence of a membrane, and presence of cloudy fluid), and Stage 3 included the findings on histological analysis of tissue sent for pathological examination because it appeared to the surgeon to be suspicious for infection.
Descriptive statistics are presented as the mean and the standard deviation for continuous variables and as the number and percentage of surgical procedures with a given feature for categorical variables. Univariate logistic regression models were run to estimate the unadjusted odds ratios (OR) and 95% confidence intervals (CI) and the p values for each factor with respect to (1) the risk of Propionibacterium acnes and (2) the risk of other organisms. Variables related to the judgment of the individual surgeon, such as decisions regarding implant removal and replacement, were excluded from the univariate and multivariate analyses.
Subsequently, multivariate models were developed to predict (1) the risk of Propionibacterium acnes and (2) the risk of other organisms. A cutoff of p < 0.2 from the univariate models was used as a filter for determining which factors to consider in the multivariate analysis. The multivariate models were built with use of the three stages of variables described above. Thus, Stage-1 variables with a p value of <0.2 in the univariate analysis were considered for inclusion. Next, additional variables with a p value of <0.2 were selected from Stage 2, and, finally, variables with a p value of <0.2 were selected from Stage 3. Variables were selected with use of the forward stepwise selection technique, with p < 0.05 as the criterion for inclusion in the model. Variables selected in previous stages were retained in the model. At the end of each stage, interactions were tested for variables in the model and were added if p < 0.01. Alternative multivariate models were constructed by repeating the three-stage variables selection process, but at each stage we used backward elimination variable selection (p > 0.05 for exclusion) instead of forward stepwise selection. We calculated the area under the receiver operating characteristic curve (AUC), sensitivity, specificity, positive predictive value, and negative predictive value for the model completed after each stage. A specificity and sensitivity combination was selected to achieve a specificity of approximately 0.50 and the best possible sensitivity, with the belief that this balance of sensitivity and specificity would be of most use to surgeons in estimating the prognosis for positive cultures for each patient. The Pearson correlation coefficient (r) was used to describe association among predictor variables. The level of significance was set at p < 0.05. Calculations were carried out in R (version 2.12.0; R Development Core Team, Vienna, Austria).
Source of Funding
The DePuy/Douglas T. Harryman II Endowed Chair for Shoulder Research funded this study.
Of the 193 cases of shoulder arthroplasty revision, 108 (56%) were associated with positive cultures (see Appendix). Forty-four positive cultures demonstrated growth of Propionibacterium acnes alone and thirty-one demonstrated growth of Propionibacterium acnes and other organisms so that of the positive cultures, 69% identified Propionibacterium acnes. Thirty-three cultures demonstrated growth of only other organisms, and eighty-five cultures were negative. Male sex was associated with a 600% increase in the risk of a positive culture for Propionibacterium acnes (p < 0.001) but not for other organisms (Figs. 1-A and 1-B). Cultures for Propionibacterium acnes became positive three to twenty-seven days after surgery. Forty-five percent of the positive Propionibacterium acnes cultures were positive at one week, with 86% being positive at two weeks, 97% being positive at three weeks, and 100% being positive at four weeks (the maximum incubation time for cultures in this study) (see Appendix). The number of specimens obtained for culture per shoulder increased over the time span of this study (r = 0.38) and was correlated with some of the variables that were significantly associated with positive Propionibacterium acnes cultures in the univariate analysis, such as cloudy fluid (r = 0.31), humeral loosening (r = 0.26), and membrane formation (r = 0.26). The OR for the risk of Propionibacterium acnes in relation to the number of culture specimens obtained was 1.24 per culture (p = 0.01), and the OR for risk of other organisms was 1.35 per culture (p < 0.001) (Figs. 1-A and 1-B). In other words, the risk of a positive culture for Propionibacterium acnes was approximately 25% larger with every additional sample, and the risk of a positive culture for other microorganisms was 35% larger with every additional sample. In the shoulders that had positive cultures for Propionibacterium acnes, not all specimens were positive; on average, 2.4 of the 4.3 cultures per procedure were positive for Propionibacterium acnes and 1.9 were culture-negative.
The risks of positive cultures for Propionibacterium acnes and for other organisms were related (OR = 1.92; p = 0.04, chi-squared test). The risk of positive cultures for other organisms was higher among patients in whom Propionibacterium acnes was present (43%) than among those in whom Propionibacterium acnes was absent (28%). Conversely, the risk of Propionibacterium acnes was higher among those in whom other organisms were present (49%) than among those in whom other organisms were absent (34%).
Most patients had normal values for C-reactive protein (CRP) (0 to 10 mg/L), erythrocyte sedimentation rate (0 to 15 mm/hr), and white blood-cell count (WBC) (4300 to 10,000/μL). Twenty (17%) of 119 CRP values, twenty-four (20%) of 119 erythrocyte sedimentation rate values, and fourteen (8%) of 183 WBC values were elevated. In the group of 108 patients with positive cultures, only nine (13%) of seventy CRP values, twelve (17%) of seventy-one erythrocyte sedimentation rate values, and nine (9%) of 103 WBC values were elevated. The odds of a positive culture were not significantly related to the WBC, ESR, or CRP values (Figs. 1-A and 1-B).
One hundred and thirty-four shoulders had at least one procedure prior to the arthroplasty for with the revision was performed. Fifty-nine shoulders had no surgical procedures prior to the arthroplasty for which the revision was performed. In this series, the odds of a positive culture were not significantly related to the number of prior surgical procedures on the shoulder, to rheumatoid arthritis, or to other comorbidities (Figs. 1-A and 1-B).
Eighty-four shoulders had both a glenoid and a humeral component in place at the time of revision, whereas 109 had only a humeral component. Of the eighty-four glenoid components, fifty-two were loose on radiographs, fifty-nine were loose at the time of surgery, and thirty-six showed implant wear at the time of surgery. Of the 193 humeral components, twenty-three were loose on radiographs and twenty-nine were loose at the time of surgery. Humeral component loosening and humeral osteolysis on radiographs were associated with threefold and tenfold increases in the risk of a positive Propionibacterium acnes culture, respectively (p < 0.01) (Fig. 1-A). The surgical findings of glenoid wear and membrane formation were each associated with fourfold increases in the prognosis for a positive culture for Propionibacterium acnes (p < 0.001). The presence of cloudy fluid was associated with a twelvefold increase in the risk of a positive Propionibacterium acnes culture (p < 0.05).
The prognosis for a positive culture for other organisms (with or without Propionibacterium acnes) was increased at a significant level for smokers, shoulders with radiographic evidence of glenoid loosening, shoulders with surgical findings of a loose glenoid component, and the formation of a membrane (Fig. 1-B).
The odds of a positive culture were not significantly increased by histopathological findings of acute inflammation, chronic inflammation, or foreign-body reaction. Positive cultures for Propionibacterium acnes and positive cultures for other organisms were more than twice as likely in cases in which antibiotics had been withheld prior to the harvesting of specimens at the time of surgery as compared with cases in which antibiotics had not been withheld (p < 0.05) (Figs. 1-A and 1-B).
Multivariate models for a positive Propionibacterium acnes culture are shown in Table I. The preoperative observations of male sex and radiographic evidence of humeral osteolysis and the intraoperative observations of cloudy fluid or a membrane were significant and independent factors predictive of a positive Propionibacterium acnes culture in the multivariate model. For a choice of specificity of approximately 0.50, the model produced a sensitivity of 0.87, meaning that this model would have predicted almost 90% of the positive Propionibacterium acnes cultures and would have predicted approximately half of the negative cultures (see Appendix).
Multivariate models for a positive culture for organisms other than Propionibacterium acnes are shown in Table II. Procedure date, diabetes, smoking, glenoid loosening, and histopathological studies showing chronic inflammation were each significant and independent characteristics predictive of organisms other than Propionibacterium acnes.
To our knowledge, this is the first study to statistically analyze the preoperative and intraoperative factors associated with positive culture results for Propionibacterium acnes and for other organisms in a large series of patients undergoing revision shoulder arthroplasty for causes other than obvious infection. This investigation demonstrates that male sex, osteolysis, membrane formation, and cloudy fluid are significant and independent prognostic factors for a positive culture for Propionibacterium acnes in tissue sampled at the time of revision shoulder arthroplasty for the treatment of pain, stiffness, and/or prosthetic loosening.
We emphasize that this study concerns risk factors for positive cultures in revision shoulder arthroplasties, recognizing that robust criteria for establishing the diagnosis of low-grade chronic infection in cases of failed shoulder arthroplasty have yet to be established6,8,16,20-22,24,26,41,44,45,47,49,53,63-65.
The results of the present study indicate that the likelihood of a positive culture at the time of revision shoulder arthroplasty for the treatment of pain, stiffness, or loosening is high (56% in our series). This high rate of positive cultures may be due to our culture protocol, which involved withholding preoperative antibiotics, using both aerobic and anaerobic media, and observing the cultures for twenty-eight days; it is of note that 55% of the positive cultures in our study would have been missed with an observation time limited to one week. The high rate of positive cultures may also be related to the submission of multiple tissue samples for culture. Importantly, in the shoulders that cultured positive for Propionibacterium acnes, only an average of 2.4 of 4.3 cultures per procedure were positive; if a smaller number of tissue samples had been obtained for the shoulders in this series, some of the positive cultures would likely have been missed.
Our univariate statistical analysis showed that the prognosis for a positive Propionibacterium acnes culture was increased sixfold for male patients (p < 0.001), tenfold for shoulders with humeral osteolysis on radiographs (p < 0.01), twelvefold for shoulders with cloudy joint fluid (p < 0.05), fourfold each for shoulders with glenoid component wear (p < 0.001) and membrane formation (p < 0.001), and threefold for shoulders with radiographic evidence of humeral component loosening (p < 0.01). Our multivariate analysis showed that male sex, radiographic evidence of humeral component loosening, membrane formation, and cloudy fluid were each independent and significant predictors of positive Propionibacterium acnes cultures.
The results of the present study should be viewed in light of certain limitations. We confined our analysis to the objective results of cultures rather than applying an arbitrary definition of “infection,” recognizing the subtle and varying clinical presentations that confound the establishment of rigorous diagnostic criteria for periprosthetic infection with Propionibacterium acnes6,8,16,20-22,24,26,39,41,45,47,49,53,63-65. We emphasize that a positive culture is not the equivalent of an infection any more than a negative culture is the equivalent of a sterile surgical field. For example, Jawa et al.19 chose to use an antibiotic-loaded cement spacer for the treatment of four cases of “infection” after shoulder arthroplasty although there was no growth on culture in those cases. Our observation that, among shoulders that demonstrated a positive culture for Propionibacterium acnes, an average of 1.9 of the 4.3 cultures per procedure were negative points to the fact that tissue cultures represent samples of a nonhomogeneous environment and therefore are at risk for being falsely negative if the involved areas are not sampled. As this was not a prospective study, there was no standardized protocol for withholding antibiotics prior to surgery, for the number of cultures obtained in each case, or for obtaining histopathological examination; rather, these decisions were made on the basis of the surgeon’s judgment in each case. Finally, we could not determine the role that the positive cultures had in the production of the shoulder pain, stiffness, or component loosening for which the revision surgery was performed.
This study demonstrates a significant relationship between preoperative and intraoperative characteristics and the prognosis for culturing of Propionibacterium acnes from tissue samples harvested from a large series of shoulder arthroplasties that were revised because of pain, stiffness, or component loosening. Surgeons should be aware that the likelihood of a positive culture for Propionibacterium acnes is high in cases of revision shoulder arthroplasty, particularly in male patients with radiographic findings of humeral loosening and osteolysis, surgical findings of glenoid wear, and a membrane around the implant. Shoulders with these risk factors may merit treatment with more aggressive surgical and antibiotic strategies than those that would be used for shoulders without these risk factors.
As the incidence of infection following shoulder arthroplasty is in excess of one of every 100 cases3-7 and as the majority of the reported infections following shoulder arthroplasty were associated with a positive culture for Propionibacterium acnes3,6-38,40,45, the clinical relevance of this study is apparent. Because Propionibacterium acnes does not usually produce the typical clinical or laboratory manifestations of infection13,15,19,21-24,26,27,39,42,44,46 and because the results of cultures performed at the time of surgery may take weeks to finalize8,39,46,47,49-52,60, the surgeon performing a revision shoulder arthroplasty must make decisions regarding prosthesis retention or exchange and the immediate institution of antibiotic therapy intraoperatively. Propionibacterium acnes has a known propensity to form a biofilm, a bacteria-containing polysaccharide matrix attached to the prosthesis39,49,50,59,60,66,67. If a prosthesis coated with a biofilm is not removed, the organisms may not be eradicated; if a new prosthesis is inserted in the presence of Propionibacterium acnes without immediate antibiotic prophylaxis, a biofilm may form on the revision implant67.
Propionibacterium acnes is a normal component of the skin microbiome, inhabiting the normal sebaceous glands and hair bulbs of the chest where the skin incision is made for shoulder arthroplasty, especially in males22,68-70. Propionibacterium acnes is not eliminated by the usual preoperative antibiotics71 or by standard surgical skin preparations72,73 and is commonly introduced into surgical wounds56,63,69,74,75, where it can become a part of durable biofilms on orthopaedic implants39,40,49,50,59,60,66,67,76 in which it is protected from the action of antibiotics and host immune response49 and is difficult to eradicate18. This knowledge suggests a strategy for minimizing the risk of Propionibacterium acnes infection, including using prophylactic antibiotics with maximum effectiveness against Propionibacterium acnes, optimizing skin preparation73, discarding the skin knife and using a new one for deep dissection74, frequent irrigation of the surgical field with generous volumes of antibiotic-containing saline solution to minimize the inoculum, removal of all potentially colonized foreign material, insertion of new implants with fresh gloves, and the use of nasal oxygen for twenty-four hours after surgery77.
Propionibacterium acnes is commonly cultured from tissues obtained at the time of surgical revision of painful, stiff, or loose shoulder implants15,20,23,26,33,54 in the absence of the usual clinical manifestations of infection and in the absence of abnormal blood tests13,15,19,21-24,26,27,39,42,44,46. These symptoms may be delayed by months or longer after the index procedure because of the slow doubling times of this organism20. In cases with an increased prognosis for positive cultures, the surgeon should have a high index of suspicion and may consider informing the patient of the possibility of infection, withholding antibiotics until multiple tissue cultures can be performed (and observed for twenty-eight days), removing extant implants, performing reimplantation of the humeral component with allograft soaked in antibiotics, and administering Propionibacterium acnes-specific antibiotics60 to be continued for twenty-eight days until the culture results are final.