We performed a retrospective cohort study using prospectively collected data from our arthroplasty database; patients were not recalled for this study. We identified 9245 patients undergoing primary hip or knee arthroplasty between January 2001 and April 2006. Of these, 7739 patients underwent unilateral joint arthroplasty, and 1496 had simultaneous bilateral surgery. Primary hip arthroplasty was performed in 5060 patients, including 4635 unilateral surgeries, and 425 simultaneous bilateral hip replacements. Primary knee arthroplasty was performed in 4185 patients, including 3114 unilateral procedures and 1071 bilateral total knee replacements. There were 3882 males with a mean age of 62 years (range, 15–94 years) and 5363 females with a mean age of 66 years (range, 14–97 years). The minimum followup was 12 months (mean, 43 months; range 12–76 months). We did not have information regarding the number of patients lost to followup before the 12-month minimum followup.
The protocols for infection prophylaxis were the same throughout the study period. This included administration of intravenous antibiotic within 1 hour of the arthroplasty and for 24 hours postoperatively. We performed all arthroplasties in operating rooms equipped with vertical laminar flow with all members of the surgical team wearing helmet aspirator suits. Neuraxial anesthesia was used for all surgeries unless contraindicated. We placed patients in the supine position for surgery. Skin preparation included the use of alcohol and iodine lavage plus an Incise adhesive drape (3M™ Ioban™, St. Paul, MN). We routinely used double gloves for all procedures.
We performed hip arthroplasty through a modified anterolateral approach using uncemented components in all cases. Knee arthroplasty was performed under tourniquet using a medial parapatellar arthrotomy approach. We used cemented fixation for all knees. No drains were used in any of the cases.
Postoperative wound management consisted of application of a sterile dressing, which was placed over the incision in the operating room and usually kept for 48 hours. We then inspected the wound and changed the sterile gauze twice daily until the patient was discharged from the hospital.
The prophylactic anticoagulation regimen consisted of administration of warfarin on postoperative Day 1 and continuing for 6 weeks, aiming for an international normalized ratio of 1.5 to 2.0. Unless contraindicated, patients also received 1000 units of intravenous heparin at the time of dislocation of the hip or before inflation of the tourniquet during knee arthroplasty. Patients’ comorbid conditions or postoperative cardiovascular complications altered the chemoprophylaxis protocol in some cases, which included administration of intravenous or subcutaneous heparin for patients with arrhythmia or cardiac disease.
We followed patients at 6 weeks, 6 months, 2 years, and 5 years after surgery. They are encouraged to contact their surgeon as needed or in the presence of fever, wound drainage, or any unexpected adverse event. We prescribed antibiotic prophylaxis before dental procedures (600 mg clindamycin orally) to all patients at their first followup, which was recommended during the first 2 years after arthroplasty.
The diagnosis of deep periprosthetic infection was made if at least three of the following five criteria were present: (1) abnormal serology (erythrocyte sedimentation rate > 30 mm/hour; C-reactive protein > 1 mg/dL); (2) strong clinical and radiographic suspicion for periprosthetic infection such as periosteal elevation, focal osteolysis, hot and swollen joint, draining sinus, (3) positive joint aspiration culture; (4) evidence of purulence during the subsequent surgical intervention; and (5) a positive intraoperative culture [11
We performed descriptive analysis using univariate statistics on a large number of variables (Table ). Patients were included only once in the analysis even though some patients received multiple surgeries. All continuous variables are reported as means and standard deviations. We compared potential predictive variables between the two groups of patients (with or without deep periprosthetic infection) using parametric (t statistics) or nonparametric statistics (Wilcoxon) as appropriate. We reported categorical variables as frequency distribution and comparisons were made using parametric (chi square statistics) and nonparametric statistics (Fisher’s exact). We kept variables significant at p = 0.10 in the final statistical model using stepwise logistic regression. Multivariate analysis then was performed using selection stepwise logistic regression after adjusting for the potential confounders to determine noteworthy predictors of periprosthetic infection. Confounders included the following: race, physical status score (ASA), body mass index, rheumatoid arthritis, venous thromboembolism, anemia, hypercholesterolemia, dementia, knee arthroplasty, simultaneous bilateral surgery, operative time, hospital length, postoperative creatinine, allogenic transfusion, postoperative atrial fibrillation, postoperative myocardial infarction, postoperative urinary tract infection, postoperative wound drainage, and postoperative hematoma. Variables were considered clinically important predictors of periprosthetic infection with a p < 0.05 after adjusting for the potential confounders. All analyses performed were two-tailed and at an alpha level of 0.05. All results have been reported as 95% confidence intervals and p values.