From 1993 to 2004, 235 consecutive revision total knee arthroplasties were performed in 225 patients. Patients with diabetes mellitus (twenty), peripheral arterial occlusive disease (three), psoriasis (one), any previous lower-extremity infection (three), osteomyelitis (one), or a malignant tumor (one) and those being treated with immunosuppressive agents (one), undergoing repeat revision (ten), or lost to follow-up (six) were excluded from the study. The remaining 179 patients (183 revision total knee arthroplasties) were enrolled in the study. The primary reason for all of the primary total knee arthroplasties was osteoarthritis. All of the revisions were first-time revisions. All of the patients were determined preoperatively and intraoperatively to be without infection (i.e., they had no preoperative symptoms or signs of infection, a normal preoperative erythrocyte sedimentation rate and C-reactive protein level, and negative intraoperative cultures).
This study was approved by our institutional review board and was registered with our institutional review board with the register number 96-01-01A. Prior to the operation, all of the patients provided written informed consent. Vancomycin-impregnated cement (1 g of vancomycin in 40 g of Simplex-P cement [Stryker, Limerick, Ireland]) was used to fix the components in patients with odd chart numbers (Group 2), whereas pure Simplex-P cement without impregnated antibiotics was used to fix the components in patients with even chart numbers (Group 1). In four patients (eight knees) undergoing bilateral revision total knee arthroplasty, vancomycin-impregnated cement was used on one side (Group 2) and not on the other (Group 1). In all of the bilateral cases, the two revision arthroplasties were performed separately under different anesthetics; the time between the operations ranged from six to twenty-four months.
Group 1 comprised ninety knees. The mean age of the patients was seventy years (range, forty-nine to eighty-nine years) at the time of the index arthroplasty. Fifty-five of the knees were in men, and thirty-five were in women. There were forty-five left knees and forty-five right knees. Group 2 consisted of ninety-three knees in patients with an average age of seventy-one years (range, fifty-five to ninety years) at the time of the index arthroplasty. Sixty-one of the knees were in men, and thirty-two were in women. There were fifty-one left knees and forty-two right knees.
In Group 1, forty-nine primary total knee arthroplasties were cementless, thirty-three were performed with cement, and eight were done with hybrid fixation. In Group 2, fifty-four primary total knee arthroplasties were cementless, twenty-nine were done with cement, and ten were done with hybrid fixation. No antibiotic-impregnated cement was used in any of the 183 primary total knee arthroplasties. In Group 1, the indication for revision was loosening in sixty-nine knees and severe wear of the prosthesis in twenty-one knees. In Group 2, the indication for revision was loosening in sixty-four knees and severe wear of the prosthesis in twenty-nine knees. All of the operations were done in standard operating rooms without ultraviolet light, laminar flow, or isolation space suits. All of the revision total knee arthroplasties were performed through a midline incision and a medial parapatellar approach, with an Insall snip29 done in eighty-five knees. No quadriceps turndown or tibial tubercle osteotomy was done in this series. The NexGen revision prosthesis system (Zimmer, Warsaw, Indiana) was used in all knees. Extension stems (fully cemented and 75 or 100 mm in length) were used for both the femoral and the tibial component, and at least two augments (and as many as four) were used for the femoral components. Tibial augments were used in seven knees in Group 1 and in ten of Group 2. All of the components were fixed with three 120-g packs of Simplex-P cement. All of the procedures were performed by the same surgeon (F.-Y.C.). Although the patients were blinded to whether or not the cement was impregnated with vancomycin, the surgeon was not. However, the use of antibiotics was randomly determined, as described above.
With the numbers available, there were no significant differences between the two groups with regard to age, sex, side of the operation, reason for revision, time between the primary and revision procedures, body mass index, score according to the American Society of Anesthesiologists (ASA) Physical Status Grading System30, preoperative or postoperative Hospital for Special Surgery (HSS) knee score19, tourniquet time, operative time, length of hospital stay, number of days with drainage (i.e., time until the wound and drainage-tube holes became dry), or period of follow-up (p > 0.05) (Table I).
Each patient had intravenous bolus injections of cefazolin (500 mg) and gentamicin (80 mg) before the operation. Intravenous injections of cefazolin (500 mg every six hours) and gentamicin (80 mg every twelve hours) were given for thirty-six hours after the operation, and then oral cefazolin (500 mg every six hours) was given for another seven days. A drain was used for forty-eight hours and then removed. All of the wounds and drainage-tube holes healed within five days after the operation, and no prolonged drainage was noted in any knee. Continuous passive motion was used from the third postoperative day to the day of discharge. Immediate full weight-bearing on the operatively treated knee was allowed from the second postoperative day, and crutches were used as needed. All of the patients were discharged when they had good wound-healing, and the mean hospital stay was thirteen days (range, nine to seventeen days). The patient demographics, findings on the preoperative physical examination, preoperative radiographic findings, preoperative laboratory data, findings of the preoperative evaluation to clear the patient for anesthesia, type of anesthesia, operative time, tourniquet time, estimated blood loss, postoperative transfusions, wound condition, and range of motion of the knee were all recorded for each patient.
Patients were examined at three weeks, eight weeks, and six months postoperatively and every six months thereafter. The average duration of follow-up was eighty-nine months (range, thirty-six to 156 months). Radiographic evaluations were performed at every follow-up visit, and a functional evaluation was performed with use of the HSS score starting on the third postoperative visit. The infections were classified as superficial or deep16 and also according to the system for classification of infections described by Tsukayama et al.25,28,31. All deep infections were confirmed by laboratory tests (erythrocyte sedimentation rate and C-reactive protein level) and joint fluid culture.
We estimated the sample size with the assumption of a 10% infection rate in the control group, on the basis of our previous experience around 1990, and a 1% infection rate in the treatment group was considered clinically relevant. We estimated that a sample size of eighty-eight patients in each group would be required to provide a statistical power of 80% to detect a difference of 9% in the infection rate between the two groups with a 5% level of significance in a two-sided test. Therefore, we planned to recruit at least ninety subjects for each group in this study. All patients were evaluated for all response variables, which included demographic variables and important outcomes. Continuous response variables are represented as the mean and standard deviation, and discrete variables are given as percentages. The Fisher exact test was used to assess differences between the two groups with respect to each discrete variable, and a Student t test was used to assess differences between the groups with respect to each continuous variable. The p value was set at 0.05 for each test before the analysis. Exploratory Cox regression analysis was used to assess and adjust for any possible risk factors for infection.
Source of Funding
There was no external funding source for this study.
No infection was noted in Group 2 (treated with vancomycin-impregnated cement). Six deep infections and one superficial infection were noted in Group 1 (the control group). The differences between the groups with regard to the rate of deep infection and the overall rate of infection (both superficial and deep) were significant (p = 0.0130 and 0.0062, respectively) (Table I). In Group 1, one (1%) of the knees had an early superficial wound infection (with coagulase-negative Staphylococcus), which was noted twenty-one days after the revision surgery and was treated with wound débridement and intravenous antibiotics for two weeks and oral antibiotics for another two weeks. In Group 1, six knees (7%) had a deep infection: two of these infections were type I (acute postoperative), and four were type II (late chronic) according to the system of Tsukayama et al.25,28,31. All of these infections developed within two years (range, twenty-four days to eighteen months) following the revision total knee arthroplasty. None of the patients with an infection were smokers. No tibial augment had been used in the six patients with a deep infection, whereas two femoral augments had been used in five of these patients and four femoral augments had been used in one. Culture results showed that the pathogen was methicillin-resistant Staphylococcus aureus in three knees, coagulase-negative Staphylococcus in two, and group-B Streptococcus in one. None of the infections developed in a patient treated with a bilateral procedure. With the numbers available, age, sex, side of the lesion, reason for the revision, time between the primary and revision procedures, body mass index, ASA grade, tourniquet time, operative time, hospital stay, HSS score, and period of follow-up were not found to be significant factors affecting the development of infection (p > 0.05) (Tables II and III). The results of the univariate Cox regression analysis with time-to-infection as an outcome are shown in Table III. Vancomycin was the only significant factor that reduced the risk of infection.
No component loosening or other complications developed following the revision total knee arthroplasties in either group during the follow-up period. Excluding the infected knees, the knees in Group 1 had a mean postoperative HSS score of 85 points (range, 81 to 92 points) and those in Group 2 had a mean score of 87 points (range, 82 to 93 points) (p = 0.623).