Kidney Injury and Mortality Are Minimal Following Two-Stage Treatment of Periprosthetic Joint Infection

Abstract

Background:

The current gold standard treatment for chronic periprosthetic joint infection (PJI) is a two-stage surgery with antibiotics between implant resection and reimplantation. Recently, claims have been reported of nephrotoxic antibiotic exposure and acute kidney injury (AKI) during the two-stage PJI treatment. This study evaluated kidney function and mortality before, during, and after two-stage revisions for PJI using a standardized protocol.

Methods:

Clinical data from 138 hips and knees undergoing two-stage treatment for PJI were retrospectively reviewed. Standardized perioperative clinical protocols were used for all patients, including hypotensive anesthesia and tranexamic acid to minimize operative blood loss, aggressive fluid resuscitation, medical optimization, and six weeks of intravenous antibiotics prior to reimplantation. The incidence of AKI, serum creatinine (Cr), estimated glomerular filtration rate (eGFR), blood urea nitrogen (BUN), and overall mortality were collected from a statewide healthcare system. P-values ≤ 0.05 were considered statistically significant.

Results:

Overall, the incidence of AKI was 9.4% (13 of 138), which was 4.3% in patients who did not have pre-existing kidney disease. There were no significant differences observed in mean serum Cr (1.10, 1.12, 1.13 mg/dL), eGFR (78.6, 77.7, 74.8 mL/minute/1.73 m²), or BUN levels (19.8, 18.9, 19.0 mg/dL) across the pre-resection, the inter-stage, and post-reimplantation periods, respectively (P ≥ 0.432; power ≥ 85.3%). Mortality was zero within 90 days of resection and 1.4% (two of 138) within one year of resection; both deaths were due to cardiac events unrelated to kidney function. Kaplan-Meier survivorship estimates were 98% at two years and 86% at five years post-resection.

Conclusion:

In this single-institution series using a standardized protocol, mean kidney function metrics remained relatively stable throughout the two-stage treatment. Furthermore, mortality after the two-stage approach was negligible and unlikely to be associated with the treatment itself. Still, comparative studies of the one-stage technique remain warranted.

Introduction

Periprosthetic joint infection (PJI) is a challenging complication following total joint arthroplasty (TJA) and is associated with considerable morbidity and mortality [1,2]. While the incidence of PJI after primary TJA is 1 to 2%, the overall number of patients requiring revision for PJI is expected to increase as primary TJA volumes rise [3]. For chronic PJI, the two-stage exchange remains the gold standard treatment option, involving explantation of components, extensive debridement, placement of an antibiotic-loaded spacer, administration of intravenous antibiotics, and later reimplantation once infection is eradicated. While success rates vary based on the criteria used [4], the success rate of the two-stage approach related to reinfection has been reported up to 93% at a mean follow-up of 6.5 years [5,6]. However, the two-stage treatment requires at least two operations (sometimes more) and prolonged antibiotic exposure, which may increase patient risk.

Among these risks are acute kidney injury (AKI), which has emerged as a particular concern of the two-stage revision for PJI. Nephrotoxic antibiotics, administered intravenously and via antibiotic-loaded temporary spacers, have been implicated in AKI during two-stage revision. The reported incidence of AKI varies widely from 0 to 33% [7,8], with a recent meta-analysis estimating a pooled rate of 16.6% across 14 studies [9]. The development of AKI is multifactorial, influenced by antibiotic selection, spacer composition, patient comorbidities, and concomitant medications [10,11], along with inadequate perioperative surgical and medical management, including excessive blood loss and suboptimal fluid resuscitation. Beyond renal complications, the burden of two major procedures also raises concern for the possibility of increased mortality compared to primary or aseptic revision arthroplasty [12–14] along with increased morbidity during the inter-stage period.

While the single-stage revision has gained attention as a potential alternative with favorable outcomes in selected populations, the two-stage approach remains the most commonly utilized treatment in the United States [9,15,16]. Given the uncertainty surrounding systemic risks, particularly kidney dysfunction and mortality, further investigation is warranted. The present study aimed to evaluate the following outcomes in a specialized PJI treatment institution with optimized surgical and perioperative protocols: (1) the incidence of AKI (meeting a dichotomous endpoint based on established definitions), (2) kidney function metrics, and (3) overall mortality throughout the two-stage treatment course.

Methods

After institutional review board approval, 160 patients who underwent two-stage surgical treatment for PJI from August 30, 2010, to June 22, 2022, were retrospectively reviewed. In total, 138 cases (66 hips; 72 knees) were available for data collection and analysis. Inclusion criteria for this study were patients who met Musculoskeletal Infection Society criteria for PJI [17] and underwent a planned two-stage exchange. A total of 22 cases were excluded due to native joint resections, joints that were not reimplanted by the same surgeon or at the same institution because patients relocated, and resections performed in the setting of a pre-existing fracture.

Two-Stage Treatment

An evidence-based contemporary standardized protocol for two-stage treatment of hip and knee PJI was used in all cases. The resection procedure consisted of component and hardware removal, aggressive synovectomy and debridement, irrigation with nine liters of pulsatile lavage followed by a dilute povidone iodine soak for three minutes, and the insertion of an antibiotic-impregnated spacer. For hips, one of two articulating antibiotic spacers, either a molded Prostalac Hip System (DePuy Orthopedics, Inc., Warsaw, Indiana, USA) or a prefabricated Remedy Hip System (OsteoRemedies, Memphis, Tennessee, USA) was used. For knees, an articulating or static spacer was implanted as appropriate with high-dose antibiotic cement. Indications for static spacers were at the discretion of the surgeon and included substantial bone loss and insufficient collateral ligament integrity. In cases where a static spacer was deemed appropriate, high-dose antibiotic-laden cement was molded to the distal femur and proximal tibia. Intramedullary dowels were used with all static spacers and, depending on surgeon preference, were formed around a central reinforcing Steinmann pin or nail or constructed without a central metal pin, utilizing a cement gun nozzle lubricated with normal saline to facilitate cement removal once cured. Spacers and dowels utilized high-dose antibiotic cement and dosing appropriate for the cement type. For a Cobalt (Enovis, Lewisville, Texas) and Palacos (Heraeus Medical, Yardley, Pennsylvania) cement batch, two grams of vancomycin and zero to two grams tobramycin were used, and for Simplex P (Stryker, Mahwah, New Jersey) cement, three to four grams of vancomycin and 3.6 grams tobramycin were utilized. In Simplex P, one cubic centimeter of methylene blue per cement batch was added to provide color to facilitate visualization upon removal at reimplantation. In indicated cases of multi-organism and recalcitrant infections, an antifungal (50 micrograms amphotericin B) was added to the cement. In total, 89.1% (123 of 138) of cases used a prefabricated articulating spacer.

A minimum of five sets of cultures were obtained at the start of resection, with samples taken at the fascial layer if indicated and within the joint. Immediately after resection surgery, patients were seen by an infectious disease specialist who coordinated inpatient and a minimum of six weeks of outpatient intravenous (IV) antibiotic treatment. Antibiotics were tailored to the sensitivities of organisms grown in culture, or in cases with no growth, either a broad-spectrum dual-antibiotic regimen was used, or antibiotics were chosen based on the patient’s prior infection history. Erythrocyte sedimentation rate (ESR) and C-reactive protein (CRP) were obtained at four weeks to establish a post-resection baseline and comparison value for pre-resection lab values. After completing six weeks of IV antibiotic therapy, patients were given a two-week antibiotic holiday before being re-evaluated for reimplantation. At the office evaluation before reimplantation, ESR and CRP were rechecked for a decreasing trend to normalization, compared with the ESR and CRP obtained at four weeks. If levels were persistently elevated, joint aspiration was performed by musculoskeletal radiology before reimplantation, or if unsuccessful, was performed by the surgeon at the start of the intended reimplantation procedure. If the aspiration yielded a total nucleated cell count greater than 3,000 or neutrophils/bands greater than 90%, reimplantation was deferred, and repeat insertion of an antibiotic-cement spacer affixed with high-dose antibiotic cement was performed. A minimum of three cultures were obtained on entrance into the joint space before spacer removal during reimplantation procedures.

All culture-positive infections were treated with six weeks of targeted IV antibiotics, and culture-negative infections were treated with broad-spectrum dual antibiotics cefepime and vancomycin for six weeks. Broad-spectrum high-dose dual antibiotics were used in all antibiotic cement spacers; typically, 3.0 grams of vancomycin and 3.6 grams of tobramycin. All patients were followed by an infectious disease specialist for the duration of IV antibiotics prior to reimplantation. Culture-negative infections were followed more closely with weekly labs to measure vancomycin levels.

Patients were medically optimized before the resection, except in more acute presentations, and in the inter-stage before reimplantation per standardized protocols developed and implemented by an internal medicine specialist whose practice focuses solely on perioperative medical management of total joint arthroplasty patients. Medical optimization included, but was not limited to, postoperative fluid resuscitation, perioperative glucose control, nutrition, smoking cessation, anemia correction, and multiorgan system optimization. Specifically, this included control of diabetes (hemoglobin A1c less than 8.0), correction of post-resection anemia (hemoglobin greater than 10.0 before reimplantation), correction of malnutrition based on screening laboratory values, and appropriate control of hypertension. Patients were screened for nasal colonization with methicillin-sensitive Staphylococcus aureus (MSSA) or methicillin-resistant Staphylococcus aureus (MRSA) and treated with nasal mupirocin before reimplantation. All patients were administered IV cefazolin within one hour before the start of surgical procedures or cefazolin plus vancomycin in the presence of MRSA colonization. After reimplantation, patients were placed on prophylactic IV antibiotics until discharge and discharged on oral antibiotics for a minimum of seven days until intraoperative cultures were final. Oral cephalosporins, either cephalexin or cefadroxil, were used in most patients after reimplantation unless the offending organism was MRSA, in which case oral Bactrim double strength was used. Chronic antibiotic suppression was not routinely used after reimplantation. Patients were seen and evaluated after reimplantation at four weeks, four months, and annually thereafter with additional clinic visits scheduled with priority as needed.

Cohort Characteristics

Study variables were manually collected from manual chart reviews within a statewide healthcare system. The cohort was 51% women and 83% American Society of Anesthesiologists Physical Status (ASA-PS) classification III or IV with a mean age and body mass index (BMI) of 67 years (range, 44 to 88) and 32 (range, 17 to 60), respectively. Using the McPherson Classification of PJI [17], the majority of hips (79%) and knees (83%) were categorized as III-A-1 or worse (Table 1). Pre-existing kidney disease and diabetes were documented in 18.1 and 32.6% of all patients. In addition, non-steroidal anti-inflammatory drugs (NSAIDs) were used in most cases at all time intervals (61% pre-resection, 77% during the inter-stage, and 83% post-reimplantation); however, the use of NSAIDs did not have a significant effect on mean kidney function metrics at any time point (P ≥ 0.127). Similarly, the specific use of intravenous vancomycin during the inter-stage period was 52.9%, however, did not have a significant effect on mean kidney function metrics compared to all other antibiotics grouped collectively (P ≥ 0.192).

Table 1.

McPherson Staging System.

Classification	Knees (N)	Hips (N)	Knees (%)	Hips (%)
I A 1	2	3	3	5
I A 2	0	1	0	2
I B 1	0	1	0	2
I B 2	1	1	1	2
I B 3	1	0	1	0
II A 2	2	4	3	6
II B 1	2	0	3	0
II B 2	1	1	1	2
II C 1	2	1	3	2
II C 2	1	2	1	3
III A 2	14	14	19	21
III A 3	4	2	6	3
III B 2	23	24	32	36
III B 3	6	5	8	8
III C 1	1	0	1	0
III C 2	11	5	15	8
III C 3	1	2	1	3
Total	72	66	100	100

Outcomes

Manual chart review in the electronic medical record was performed to collect data on the kidney function metrics of serum creatinine (Cr), estimated glomerular filtration rate (eGFR, CKD-EPI 2021) [18], and blood urea nitrogen (BUN), from routine laboratory values obtained at three time points: pre-resection (within one month of resection), during the inter-stage period (after resection date and before reimplantation date), and post-reimplantation (after reimplantation date, up to two years). If multiple laboratory values were available within a given interval, the statistical mean for that interval was used for analysis. Acute kidney injury was defined as a serum Cr level ≥ 1.5 times baseline or an increase of ≥ 0.3 mg/dL, as previously described [12].

Mortality data including the cause and date of death were documented from a statewide healthcare system to calculate survivorship estimates.

Data Analyses

All data analyses were performed in Minitab Statistical Software 22.3.1 (State College, Pennsylvania, USA). Outliers were evaluated with Dixon’s r22 ratio tests. Normality of continuous data was assessed with Anderson-Darling normality tests. For normally distributed data, two groups were compared with paired or unpaired two-sample t-tests depending on the data assumptions, while three groups or more were compared with analyses of variance (ANOVA) tests. Categorical data were compared with Chi-square tests with Fisher’s exact tests P-values used for 2×2 comparison tables only. The means were reported with standard deviations, and 95% confidence intervals (CI) were reported where appropriate. P-values ≤ 0.05 were considered statistically significant.

Kaplan-Meier methodology was used to calculate survivorship estimates, censoring patients at latest clinical follow-up or death. Confidence intervals related to survivorship estimates were calculated at the 95% level.

Results

Overall, the incidence of AKI was 9.4% (13 of 138), which was 4.3% in patients who did not have pre-existing kidney disease. Of the 25 patients who had pre-existing kidney disease, three (12.0%) developed AKI during the inter-stage period. However, no patients required dialysis during the treatment course.

For the full cohort regardless of pre-existing kidney disease, no significant changes were observed in mean serum Cr, eGFR, or BUN levels between pre-resection, during the inter-stage period, or post-reimplantation (Figure 1, P ≥ 0.432; power ≥ 85.3%). The mean serum creatinine levels from pre-resection to post-reimplantation increased by 0.039 mg/dL (95% confidence interval [CI], −0.29 to 0.21, P = 0.954). The mean eGFR values from pre-resection to post-reimplantation decreased by −3.8 mL/min/1.73m² (95% confidence interval (CI), −9.5 to 2.0, P = 0.432). The mean BUN levels from pre-resection to post-resection decreased by −0.86 mg/dL (95% CI, −2.1 to 3.8, Figure 1; P=0.764). Results were consistent within both hip and knee cohorts (P ≥ 0.436).

Figure 1.

There were no significant differences observed in kidney function metrics of (A) serum creatinine (Cr), (B) estimated glomerular filtration rate (eGFR), or (C) blood urea nitrogen (BUN) levels between pre-resection, the inter-stage, or post-reimplantation periods (P ≥ 0.432; power ≥ 85.3%).

In addition, 8.7% of cases (12 of 138) required a repeat debridement and/or an additional spacer placement because of suspected ongoing infection, resulting in longer antibiotic exposure. However, only one hip case of the subset of 12 that required an additional irrigation or spacer placement developed AKI during the two-stage treatment course. Of the 13 patients who developed AKI, five were culture-negative infections, which was not associated with a higher incidence of AKI (P = 0.757), although these patients received more antibiotics throughout the treatment course than patients who had culture-positive infections.

Mortality was zero within 90 days of resection, and 1.4% (two of 138) within one year of resection. Both deaths were due to cardiac events unrelated to kidney function. Overall mortality was 18.8% (26 of 138). There were three of 26 deaths (11.5%) related to kidney function at a mean of 6.3 years after resection (range, five to nine). There were eight of 26 total deaths that occurred without a known cause identified in the EMR at a mean of 6.8 years after resection (range, four to 10). Survivorship free from all-cause mortality was 98% at two years (95% CI, 96 to 100) and 86% at five years (95% CI, 78 to 95) (Figure 2).

Figure 2:

Kaplan-Meier survivorship free from mortality was 98% at two years and 86% at five years post-resection.

Discussion

The present study evaluated kidney function and mortality in 138 patients who underwent two-stage revision for PJI. The overall incidence of AKI was 9.4%, which was 4.3% when evaluating patients who did not have pre-existing kidney disease. For patients who had pre-existing kidney disease, the incidence of AKI was 12%, which may be clinically meaningful although the subgroup was relatively small (three of 25). Importantly, mean serum creatinine, eGFR, and BUN values in the full cohort remained relatively unchanged across the two-stage treatment course, indicating preserved kidney function for most patients using these standardized perioperative protocols. The results reported in this study do not support the conclusions of some recent studies that suggest a much greater incidence of AKI for the two-stage approach to treating hip and knee PJI. A recent meta-analysis demonstrated a pooled AKI incidence of 16.6%, with 1.4% requiring dialysis [9]. Similarly, a systematic review reported AKI in up to 21% of patients following antibiotic spacer placement, with a host of identified risk factors including age [8,19,20], hypoalbuminemia [9], pre-existing chronic kidney disease [8,12,21–23], obesity [20,23], and medications such as ACE inhibitors [24], and postoperative NSAIDs [25]. Dagneaux et al. was the first study to report that a dose of vancomycin or aminoglycoside used in the spacer more than 3.6 grams per batch of cement increased the risk of AKI nearly two-fold [21]. Additionally, Valenzuela et al. conducted a randomized control trial (RCT) that demonstrated the placement of an antibiotic cement spacer was an independent risk factor for developing AKI in the two-stage procedure (odds ratio 7.48) [12]. While the present study’s primary objective was not to determine the primary risk factors for AKI, we found that NSAID use at any time point was not associated with the risk of AKI. The data presented here provide further evidence for the continued use of the two-stage revision for PJI [15,25–28], especially with a perioperative medical program that optimizes medical comorbidities. Medical optimization and implementation of a perioperative, multidisciplinary protocol have previously been associated with a reduced risk of AKI [20]. These data suggest that AKI development may be more closely related to patient selection and perioperative management than to the two-stage revision itself.

Regarding mortality, this study demonstrated zero mortality within 90 days, 1.4% within one year, and 18.8% overall mortality with a mean time to death after resection of 6.0 years (range, 0.5 to 11.8). Furthermore, survivorship free from all-cause mortality was 98% at two years and 86% at five years. These results compare favorably to previous studies. Aftab et al. reported a mortality rate range from 4 to 8% after one year for two-stage revision [29]. Additionally, Wildeman et al., in a national database study, reported mortality rates at one, five, and 10 years of approximately 5, 20, and 45%, respectively [30]. The low mortality rates in our series at early follow-up (90 days, one year, and two years) may reflect the improved perioperative protocols and comprehensive multidisciplinary management for this specific cohort. Although overall mortality for our cohort was 18.8%, only three patient deaths were directly attributable to renal causes at a minimum of 5.0 years after resection and therefore unlikely to be related to the two-stage revision and associated antibiotics. This finding emphasizes that while protecting renal function is crucial, long-term survival may be driven more by overall systemic health and comorbidity burden than by the two-stage treatment course itself.

This study had potential limitations. The retrospective review of single institutional database should be acknowledged as a potential limitation. Additionally, the patient cohort was relatively small compared to some larger meta-analysis studies; however, data collected for this study were much more granular in comparison to other meta-analyses and national databases. In addition, this study consisted of patients from a single institution, therefore limiting its overall generalizability. Furthermore, there was no control group because all patients underwent the same surgical protocol, which limited dose-response analyses related to antibiotic exposure.

There were also some strengths to this study. A strict protocol was used for all patients improving the reliability of the study data. The outcome measures are well-researched, measurable, and easily attainable in a standard medical facility. Additionally, this was one of the few studies analyzing two-stage revision data from a single institution with a modern perioperative protocol with robust patient optimization throughout the two-stage treatment course.

Conclusion

In this single-institution series using a standardized protocol, mean kidney function metrics remained relatively stable throughout the two-stage treatment. Furthermore, mortality throughout the two-stage treatment course was negligible and unlikely to be associated with the treatment itself. Comparative studies evaluating one-stage and two-stage treatment remain warranted.