Does exchange arthroplasty of an infected shoulder prosthesis provide better eradication rate and better functional outcome, compared to a permanent spacer or resection arthroplasty? a systematic review

Periprosthetic joint infection (PJI) is a devastating complication following hemiarthroplasty and total arthroplasty of the shoulder. It complicates 0 to 3.9 % of primary anatomical shoulder arthroplasty [1‐3], and 2 to 18.8 % of reverse shoulder arthroplasty [4‐6]. The surgical management of such cases is varied and includes either a single- or two-stage exchange arthroplasty [7, 8], resection arthroplasty with or without a permanent spacer [9, 10], arthroscopic lavage or open debridement with implant retention [11], arthrodesis [12] or amputation [13].

With no clear management guidelines, there are ongoing discussions regarding the advantages of each treatment strategy. The aim of treatment is to eradicate infection and prevent recurrence, with the challenge of optimising function of the joint in light of potential soft tissue compromise, bone loss, and large dead space in the subacromial region [14].

Unlike PJI following total hip and knee arthroplasty, the two-staged exchange is not considered the ‘gold-standard’ of treatment, with numerous prospective and retrospective cohort studies demonstrating favorable results of other treatment options [7, 9, 15, 16], however no definitive comparison of their outcome has been analysed in a randomized control study.

In this comprehensive systematic review of the literature, we aim to determine which, if any, treatment option provides better infection eradication rates, as well as enabling high functional outcomes for established periprosthetic shoulder infections. We hypothesized that single- or two-stage revision surgery results in better infection eradication rate and functional outcome than other treatment modalities.

Studies reporting the eradication rate of infection and/or functional outcomes following revision procedures to the shoulder as a result of periprosthetic infections, published from January 2000 to June 2014, were reviewed.

The criteria for inclusion of studies were as follows: (a) Studies with an abstract or written fully in English; (b) Studies reporting the results of infection eradication following a single-stage arthroplasty, two-stage arthroplasty, and resection arthroplasty with or without a permanent spacer (excluding debridement with implant retention, arthrodesis and amputation); (c) Studies reporting the functional results following a single-stage arthroplasty, two-stage arthroplasty, and resection arthroplasty with or without a permanent spacer; (d) Studies relating to delayed or chronic (6 weeks or later) stages of disease; (e) Study design was either a randomised controlled trial; comparative prospective study; prospective study with historical controls; prospective case series with no comparison group; comparative retrospective study; retrospective study with historical control group; or retrospective study with no control group; (f) Only the longest follow-up and largest patient series was included if more than one paper by the same author(s) was retrieved and if the patient cohort was deemed to be similar and the follow-ups were found to overlap; (g) The study cohort had to include 5 or more cases, even if treated with different surgical procedures; (h) The follow-up had to be of a minimum 6 months; (i) If reporting infection recurrence rate following treatment, the following variables had to be reported: number of patients treated; type of treatment; number of recurrent infections; and (j) If reporting functional outcome following the treatment, the following variables had to be reported: number of patients treated; type of treatment; pre- and/or post-operative functional scores (including Neer and Constant scores), as well as other measures such as range of motion and activity of daily living. No studies were excluded based upon the indication of the primary procedure (i.e. proximal humeral fracture, osteoarthritis, rotator cuff arthropathy, rheumatoid arthritis, or avascular necrosis).

International databases were searched systematically as previously described by Romano et al. [17]. Databases included: EMBASE; PubMed/Medline; Medline Daily Update; Medline In-Process and other non-indexed citations; Google Scholar; SCOPUS; CINAHL; Cochrane Central Register of Controlled Trials and Cochrane Database of Systematic Reviews; NHS Health Technology Assessment; http://​www.​google.​com; and http://​www.​yahoo.​com. Keywords were used alone or in various combinations to identify relevant papers: shoulder; infection; arthroplasty; prosthesis; shoulder replacement; prosthetic joint infection; periprosthetic joint infection; exchange arthroplasty; one-stage; single-stage; two-stage; resection arthroplasty; and permanent spacer.

Figure 1 illustrates the systematic exclusion of papers in this review. Previously published criteria to assess the quality of studies in systematic reviews, was utilized in this paper [18‐20] however the quality score was not used as an exclusion criterion. This included evidence of: (a) Patient cohort demographics (age and sex, indications for index shoulder procedure, isolated pathogens); (b) Description of the treatment modality (indications, length of antibiotic therapy, interim period length between stages, implants types); (c) Reported outcomes (frequency of recurrent infections, number of patients lost to follow-up).

Four investigators, DAG, AV, SS and CLR, searched and reviewed the literature independently, and then compared and combined their lists to complete the literature search. Any discrepancies were solved by reclassification as mutually agreed. Initial inclusion was based upon the studies title and abstract, but the latter stages of the review process excluded papers based on the inclusion criteria.

Statistical analysis was performed using t-tests when analyzing two treatment modalities directly, and ANOVA (one-way analysis of variance) to compare all four treatment groups. In addition Fisher’s Exact test was used to compare bivariate data of means. A p-value of <0.05 was deemed to be statistically significant.

The systematic review included eight original articles relating to resection arthroplasty (n = 83), 6 on single-stage exchange (n = 75), 13 on two-stage exchange (n = 142) and 8 papers on permanent spacer (n = 68).

All studies reported on a remarkably limited series of patients and presented material was often diverse, reporting a variety of treatment modalities. The range in the number of cases was from 2 to 35 for a single-stage exchange, 2 to 19 for a two-stage exchange, 5 to 21 for resection arthroplasty, and 1 to 15 for permanent spacers. The mean duration of follow-up was 46.8 months (standard deviation (SD) 17.6) for single-stage exchange, 37.9 months (SD 12.8) after two-stage exchange, 39.8 months (SD 20.8) after resection arthroplasty, and 31.0 months (SD 9.8) after permanent spacer implant (Table 1).

The database search retrieved no randomized controlled trials with the majority of papers describing a retrospective case series without a control group. The quality of included studies is shown in Table 2, with a mean 74.89 % (SD 11.82) of all 11 variables reported in the studies (min 54.55 %, max 90.91 %). Specifically for a single-stage exchange 78.79 % (SD 13.69) of the variables were reported, 74.13 % (SD 8.17) reported for a two-stage, 64.94 % (SD 8.18) reported for resection arthroplasty, and 73.73 % (SD 9.72) reported for a permanent spacer. Only 8 studies (38.10 %) gave a detailed description of their criteria for selecting each treatment modality. Other relevant variables such length of antibiotic therapy (47.62 %) and duration of interim period between the two stages of a two-staged exchange (46.15 %) were poorly reported.

The highest rate of infection eradication was observed following the use of a permanent spacer (95.6 %), then a single-stage exchange (94.7 %), two-stage exchange (90.8 %) and resection arthroplasty (86.7 %) (Table 1). However this difference was not statistically significant (p = 0.650) when comparing all treatment options together, or when compared individually to one another (Table 3).

Regarding functional outcome, it should be noted not all papers reported this data and a variety of evaluation measures were used; the Constant-Murley score was the most commonly and consistently reported scoring system (Table 4). However, very few studies provided pre- and post-operative functional scores and this makes analysis of data particularly challenging (Table 5). In the only three studies that reported pre-operative Constant score, this was, on average, significantly better in patients undergoing single-stage revision (score 37, SD 3), compared to those treated with a two-stage procedure (score 16, SD 1) (p < 0.0001). Coste et al. [21] reported also pre-operative Constant score in 10 patients undergoing resection arthroplasty and in only 3 patients treated with a permanent spacer, showing a better pre-operative function in patients treated with single-stage exchange compared to both the other treatment modalities.

The poorly reported pre-operative Constant scores and the different baseline in different cohorts of patients should be taken into consideration when comparing post-operative function. In fact, considering post-operative Constant scores, patients treated with single-stage exchange (mean score 51, SD 13) or a two-stage revision (44, SD 9) seem to perform better than those undergoing resection arthroplasty (32, SD 7) or a permanent spacer implant (31, SD 9). However, when considering the average difference between pre- and post-operative Constant score, no statistical difference can be seen any more (Table 6).

The failure to demonstrate a superiority of one treatment in eradicating infection differs from recently reported systematic reviews regarding established hip and knee periprosthetic infections. Romano et al. demonstrated a superiority of two-stage exchange for the treatment of periprosthetic total knee infection compared to a single-stage exchange [17], which was also evident following a two-stage exchange for periprosthetic hip infection [22]. The present study included a much lower number of published papers and reported cases of infected shoulder prosthesis compared to the aforementioned systematic reviews. The relatively low number of patients reported may explain, at least in part, the inability to demonstrate a statistical difference between infection eradication rates.

Another potential bias of our analysis of the data is patient’s selection in the different studies considered. In this regard, it should be noted that the vast majority of studies failed to describe the indications for their treatments. We can only speculate that specific host and pathogen characteristics influenced the surgeon to perform a specific operation over another, but on review of the study cohorts, we did not find data to support a difference in age, severity of infection, type of microorganism or duration of infection in patients treated according to different treatment modalities. This may suggest that the choice largely relies on a surgeon or hospital-based routine, rather than on an established protocol and consistent case-by-case evaluation.

Similarly, it is unclear from the included studies the type of antibiotic regime that is adopted as part of their pre- and post-operative management, and only 44 % of these studies documented the length of the antibiotic treatment [Table 2]. The combination of appropriate targeted antibacterial therapy based upon confirmed cultures; together with an initial radical wound debridement is paramount to eradicating the infection [23]. No standard antibiotic regime exists and varies widely [24], but the importance of this omission in some studies must be appreciated.

We did not identify any publications relating to eradication and functional rates following these procedures using national joint registry (NJR) data. We believe the data from NJR is fundamentally flawed for the surveillance of infection procedures, as it is currently impossible to identify which treatment modality has the highest infection eradication rate, or has the best post-operative functional outcome as the British NJR, for example, seeks to collect pre-operative Oxford functional scores but not post-operative scores [25]. We believe the adoption of the NJR in this way will be extremely useful and may be fundamental in guiding future treatment.

Furthermore, patients treated with a single-stage procedure had, in the few studies that reported this data, a better average pre-operative function than those receiving a two-stage revision. Similar findings regarding pre-operative function were retrieved when comparing single-stage revision with permanent spacer or resection arthroplasty, in the only paper that reported those values [21]. If we assume pre-operative joint function as a rough indicator of the overall status of the patient, we may speculate that according to these findings, patients treated with a single-stage procedure may have had a less severe condition that those treated with a two-stage exchange or other treatment modalities.

Concerning more closely functional results, although a single-stage exchange arthroplasty would theoretically be associated with a functional advantage, compared to a permanent spacer or a resection arthroplasty, we did not find data to support this hypothesis, due to the lack of pre-operative data. However, a two-stage exchange may be hindered by the soft tissue insult, in particular rotator cuff insufficiency, as a result of the initial aggressive debridement and subsequent re-implantation at a later stage [14, 26], even if the presence of the interim spacer permits constant expansion of the soft tissues [3]; however, when comparing pre- and post-operative Constant scores, the difference was not statistically different when comparing single- and two-stage exchange. Coste et al. [21] noted from their series that greater preservation of function occurred with shorter delays in diagnosis and definitive treatment.

In conclusion, if surgical aims are to eradicate infection and prevent recurrence, whilst optimizing the function of the joint [14], the result of this systematic review failed to dictate when one treatment should be used over another; moreover, it should be noted that the surgical decision in a given patient relies also on a number of other variables, that were not analyzed in this review.

Furthermore, we acknowledge the following limitations of this study, which reflects the quality of the included studies and the available information:

Also, we did not distinguish between recurrent and new infections, as this distinction was not made in most of the papers. The criteria for differentiating between recurrent and new infections is weakly supported in the literature, and somewhat artificial, and we believe this distinction is unreliable. The results of cultural examination in periprosthetic infection are too unpredictable, especially after previous antibiotic treatments.

Our findings should still be regarded as preliminary, since sample sizes are small and further confirmation is required when more data becomes available for review. Clearly there is a need for a large, multi-center, prospective study to establish the superiority of one surgical treatment over another.

This systematic review failed to demonstrate a clear difference in infection eradication between a single- or two-stage exchange arthroplasty, use of permanent spacer, or resection arthroplasty, for established periprosthetic shoulder infection. Moreover, functional improvements, poorly reported in the majority of studies, were not shown to be significantly different between treatment modalities.