The prevalence of obesity is increasing in industrialized countries1,2, including among patients referred to orthopaedic units. In one orthopaedic unit in the United States, the proportion of obese patients increased from 30% to 52% from 1990 to 20053.
Earlier studies have yielded contradictory results regarding the effect of obesity on the incidence of periprosthetic infection following hip and knee replacement4-8. The results of studies that employed appropriate multivariate analyses suggest that an increase in the risk of infection may be limited to morbidly obese patients8,9.
Diabetes commonly occurs in conjunction with obesity and is responsible for a substantial part of the long-term adverse health outcomes of obesity1,2. Diabetes is considered a risk factor for surgical site infections in general10, and it has been reported to increase the risk of infection following joint replacement4,8,11-13. Thus, diabetes could partially explain the obesity-related risk of postoperative infection. Some studies, however, have found no effect of diabetes on the periprosthetic infection rate5,6,14, and the combined effects of obesity and diabetes are not known.
The purpose of the present study was to analyze the effects of obesity and diabetes on the periprosthetic infection rate in a large population-based series of contemporary primary hip and knee replacements performed to treat osteoarthritis at one hospital that specializes in joint replacement surgery.
Our publicly funded tertiary-care center performed 8775 primary hip and knee replacement procedures between September 1, 2002, and January 31, 2008, to treat primary osteoarthritis. All publicly funded joint replacement surgery in the hospital district (which has a population of approximately 470,000) is centralized to this one hospital, so the data represent an unselected series of consecutive patients with end-stage osteoarthritis of the hip or knee.
Patients who had undergone open surgery of the involved joint prior to the hip or knee replacement and patients who had undergone both hip and knee replacement during the same anesthesia session were excluded. The remaining 7181 primary hip and knee replacement operations (involving 8083 joints in 6372 patients) were analyzed.
Patient demographics, preoperative clinical state of health, and operative details were collected prospectively in a joint replacement database. Preoperative data were recorded by an orthopaedic surgeon at an outpatient visit a few weeks prior to surgery. Operative data were recorded in the operating room by an assistant nurse and were confirmed by the operating surgeon. Supplementary operative details were collected from the hospital administrative database.
Operative Details
All operations were performed in a modern operating room with positive pressure, ultra-clean air filtration, and vertical laminar air flow. The surgeons and other sterile personnel wore indicator double gloves and ventilated surgical helmets. Senior surgeons, most of whom performed more than 200 joint replacements annually, operated on the majority of the patients (Table I).
Spinal anesthesia was used in 98.1% of the cases (7018 of 7152 for which data regarding anesthesia were available). A 3.0-g bolus of intravenous cefuroxime was administered as antibiotic prophylaxis in 98.0% (7036) of the cases, no antibiotic prophylaxis was used in three cases, and other intravenous antibiotics were administered in the remaining cases. Antibiotic-impregnated cement (antibiotic-containing Simplex [Stryker, Kalamazoo, Michigan] or PALACOS with Gentamicin [Heraeus Medical, Wehrheim, Germany]) was used for fixation of cemented hip and knee prostheses in all except eighteen (0.3%) of the cases. A closed suction drain was used in 6394 cases (89.0%) and was removed on the first postoperative day. A four-week course of enoxaparin was administered as thromboprophylaxis in 6533 cases (91.0%). The majority of patients (6611, 92.1%) arrived at the hospital on the day of operation, and the median duration of the postoperative stay was three days.
If a patient was known to have diabetes, the plasma glucose level was measured four to six times per day with use of a bedside capillary blood glucose meter. If a patient had an abnormal glucose level preoperatively but had not been diagnosed as diabetic, the plasma glucose level was measured two to four times per day, depending on the degree of hyperglycemia. Short-acting insulin was administered to keep the blood glucose level below 8.0 mmol/L (144 mg/dL) in a fasting state and below 10.0 mmol/L (180 mg/dL) in a non-fasting state.
Use of Diabetes Medication
In Finland, patients with diabetes are authorized to receive diabetes medication free of charge after they have undergone an evaluation by the Social Insurance Institution (SII). The criteria for reimbursement of the cost of diabetes medication in a patient with classical symptoms of untreated diabetes are a fasting plasma glucose level of ≥7.0 mmol/L (126 mg/dL) or a random plasma glucose level of ≥11.1 mmol/L (200 mg/dL), and the criterion in an asymptomatic patient is fasting plasma glucose levels of ≥7.0 mmol/L in repetitive measurements15. In overweight patients (body mass index [BMI], ≥25 kg/m2) with diabetes, a six-month lifestyle intervention is required. During the study period, patients with type-2 diabetes also had to be treated for six months before becoming eligible for reimbursement. For the present study, the Drug Reimbursement Register of the SII was searched to identify patients who were authorized to receive reimbursement for diabetes medication.
The SII has registered prescriptions in the Drug Prescription Register since 1994. This register covers all prescriptions, regardless of the payer, so purchases of diabetes medication can be reliably identified. This register was also searched for data on all purchases of such agents by patients in the study. The search included all drugs in category A10 (drugs used in diabetes) according to the Anatomical Therapeutic Chemical classification system developed by the World Health Organization: insulins and analogues, blood glucose-lowering drugs excluding insulins, and other drugs used in diabetes16. The searches of the SII registers were carried out for all patients.
Preoperative Glucose Level
The plasma glucose level of each patient was measured for a mean of six weeks (range, zero to twelve weeks) prior to surgery as part of the standard preoperative routine. The glucose level represented a fasting state in 815 patients and a random sample in 4645 patients; the preoperative glucose level in 1721 patients was not available because the samples were analyzed outside our university hospital laboratory. A glucose value of <6.1 mmol/L (110 mg/dL), which is the normal fasting level according to the criteria of the World Health Organization17, was classified as normal. Values that were ≥6.1 mmol/L were divided into two groups of equal size, using the median of these higher values (6.8 mmol/L [122 mg/dL]) as the cutoff point.
Outcome
The primary outcome of interest was the occurrence of periprosthetic joint infection during the first postoperative year. Data on postoperative infection were gathered from the hospital infection register, which conducts prospective, active surveillance for nosocomial infections following the National Nosocomial Infections Surveillance System methodology18,19 and the Centers for Disease Control and Prevention criteria20. Both deep incisional infections and organ or space infections were classified as periprosthetic joint infections because making a distinction between these types of infections can be difficult, especially in the knee19.
Statistical Analysis
All statistical analyses were performed on the basis of the number of joint replacement procedures (unilateral or simultaneous-bilateral arthroplasties). Preliminary analyses indicated that the effects of obesity and diabetes on the periprosthetic joint infection rate after hip replacement were comparable to their effects on the rate after knee replacement, and these two procedures were therefore combined to maximize the statistical power of the analyses. Secondary analyses of the hip and knee replacement subgroups were also performed.
Analyses of the infection rate according to obesity were performed by grouping patients into five categories according to the BMI in kg/m2: <25 (normal), 25 to 29 (overweight), 30 to 34 (obese), 35 to 39 (severe obesity), and ≥40 (morbid obesity). Patients were considered to have a diagnosis of diabetes preoperatively if they had been authorized to receive reimbursement for diabetes medications, or had purchased such medications, before the surgery. These patients were compared with patients who were not eligible for reimbursement and had not made any such purchases at any time during the study period (the nondiabetic group) as well as with patients who became eligible for reimbursement or began making purchases only after the surgery. In addition, the patients with diabetes were categorized according to the types of medications purchased during the six months prior to surgery: no purchases, monotherapy with a single oral agent, a combination of two or more oral agents, insulin monotherapy, or a combination of oral agents and insulin. The preoperative plasma glucose level was categorized as described above.
Confidence intervals for the infection rate were calculated with use of the Wilson method21. The chi-square test was used for univariate comparisons of categorical variables, and the Mann-Whitney U test or the Kruskal-Wallis test was used for continuous variables. The association between obesity, diabetes, or hyperglycemia and the one-year periprosthetic joint infection rate was analyzed with use of binary logistic regression analysis. The regression analyses were first run without adjustment (univariate analysis) and then with adjustment for age, sex, American Society of Anesthesiologists (ASA) risk score22, arthroplasty site (hip or knee), BMI, and diabetic status. The results of the regression analyses are presented as odds ratios (ORs) with accompanying 95% confidence intervals (CIs). A p value of <0.05 was considered significant.
Source of Funding
No external funding was received for this study.
Patients
Patient demographics, diabetic status, and operative data for the hip and knee replacements are presented in Table I. Altogether, 885 operations (12.3%) were performed on patients with diabetes that had been diagnosed at the time of surgery. The patient was registered as eligible for reimbursement of the cost of diabetes medications at the time of surgery in 716 cases (10.0%) and as having purchased diabetes medications prior to surgery in 874 cases (12.2%). On average, the reimbursement began 6.4 years (range, 0.1 to 41.4 years) before the surgery.
Diabetes was more prevalent in the knee replacement subgroup than in the hip replacement subgroup (Table I), but the duration of the disease did not differ between these subgroups. The prevalence of diabetes increased from 6% (sixty-five of 1105) in the normal weight group to 26% (fifty-one of 193) in the morbidly obese group (p < 0.001), and the prevalence increased from 10% (eighty-nine of 880) in 2003 to 14% (248 of 1750) in 2007 (p = 0.015). No temporal change in the mean BMI was observed.
Uncemented hip prostheses were used more frequently in morbidly obese patients (54% [twenty of thirty-seven]) than in patients with a normal BMI (31% [217 of 700], p < 0.001). Hybrid knee prostheses were used more frequently in morbidly obese patients (11% [seventeen of 156]) than in patients with a normal BMI (6% [25 of 405], p = 0.052). The duration of the operation was approximately ten minutes longer in the morbidly obese group than in the other BMI categories (median duration, 112 compared with 100 to 105 minutes [p < 0.001] for unilateral operations and 218.5 compared with 200 to 209 minutes [p = 0.194] for bilateral operations). Diabetes was not associated with the type of fixation or with the duration of the operation. Resident surgeons operated on relatively more patients with diabetes (14% [161 of 1121]) compared with senior orthopaedic surgeons (12% [724 of 6060]; p < 0.024), and they operated on relatively more obese patients (46% [427 of 938] compared with 39% [1960 of 5015]; p = 0.006). Senior orthopaedic surgeons performed 160 of the 193 operations on morbidly obese patients.
Infections
Fifty-two periprosthetic joint infections (0.72%; 95% CI, 0.55% to 0.95%) occurred during the first postoperative year. The one-year infection rate was 0.64% (95% CI, 0.42% to 0.98%) following hip replacement and 0.79% (95% CI, 0.56% to 1.12%) following knee replacement (p = 0.459).
Staphylococcus aureus accounted for 33% of the periprosthetic joint infections, followed by coagulase-negative staphylococci (21%) and beta hemolytic streptococci (10%). The pathogen remained unknown in ten infections (19%).
Effects of Obesity, Diabetes, and Hyperglycemia
The effects of obesity, diabetes, and the type of diabetes medication on the periprosthetic joint infection rate are presented in Table II. The patients who developed a periprosthetic joint infection had a higher median BMI (31.5 compared with 28.7 kg/m2; p = 0.001), but only morbid obesity was associated with an increased periprosthetic joint infection rate in the multivariate analysis. The infection rate increased from 0.37% (95% CI, 0.15% to 0.96%) in patients with a normal BMI to 4.66% (95% CI, 2.47% to 8.62%) in the morbidly obese group (adjusted OR, 6.4; 95% CI, 1.7 to 24.6). A periprosthetic joint infection developed after three (8.11%; 95% CI, 2.80% to 21.30%) of thirty-seven hip replacements performed in morbidly obese patients (adjusted OR, 30.79; 95% CI, 4.98 to 190.44) and after six (3.85%; 95% CI, 1.77% to 8.14%) of 156 knee replacements in such patients (adjusted OR, 8.06; 95% CI, 1.61 to 40.37).
Diabetes diagnosed at the time of surgery approximately doubled the risk of periprosthetic joint infection (Table II). The effect of diabetes on the infection rate was greater in hip replacements than in knee replacements. The infection rate following hip replacement was 2.19% (95% CI, 1.07% to 4.46%) in patients with a preoperative diagnosis of diabetes compared with 0.48% (95% CI, 0.28% to 0.80%) in patients without a preoperative diagnosis of diabetes (adjusted OR, 3.49; 95% CI, 1.06 to 11.47). The infection rate following knee replacement was 1.59% (95% CI, 0.84% to 2.99%) in patients with a preoperative diagnosis of diabetes compared with 0.66% (95% CI, 0.43% to 0.99%) in patients without a preoperative diagnosis of diabetes (adjusted OR, 1.85; 95% CI, 0.75 to 4.58). Diabetes diagnosed after surgery was not associated with an increased incidence of periprosthetic joint infection (Table II).
In obese patients, diabetes tended to increase the risk of infection (Fig. 1), but the difference was not significant in multivariate comparisons between patients with and without a preoperative diagnosis of diabetes in the individual BMI groups. The adjusted OR was 3.08 (95% CI, 0.84 to 11.28) for a BMI of 25 to 29 kg/m2, 1.55 (95% CI, 0.45 to 5.35) for 30 to 34 kg/m2, 9.73 (95% CI, 0.67 to 142.44) for 35 to 39 kg/m2, and 2.49 (95% CI, 0.58 to 10.80) for ≥40 kg/m2. There were no periprosthetic joint infections following the sixty-five joint replacements in diabetic patients with a BMI of <25 kg/m2, whereas the highest infection rate was observed in morbidly obese patients (five of fifty-one; 9.8% [95% CI, 4.26% to 20.98%]).
Preoperative Hyperglycemia
Preoperative hyperglycemia was not significantly associated with periprosthetic joint infection following hip or knee replacement in the multivariate analysis. When patients without a diagnosis of diabetes at the time of surgery were analyzed separately, the periprosthetic joint infection rate was higher in patients with a glucose level of ≥6.9 mmol/L (1.15%; 95% CI, 0.56% to 2.35%) compared with <6.9 mmol/L (0.28%; 95% CI, 0.15% to 0.53%; p = 0.002). This difference was seen across all overweight and obese BMI groups (Fig. 2). However, after adjustment in the multivariate analysis, the difference did not reach significance (adjusted OR, 3.25; 95% CI, 0.96 to 11.04). A similar association was observed in the subgroup of patients in whom the glucose measurement represented a random sample, whereas no significant differences among the glucose categories were observed in the subgroup of patients in whom the glucose measurement represented a fasting state. In both subgroups, however, the infection rate was lowest when the plasma glucose level was <6.1 mmol/L. One hundred and fifty-eight (26%) of the 610 patients with a plasma glucose level of ≥6.9 mmol/L who had not been diagnosed as diabetic were diagnosed as diabetic after surgery. Because the preoperative glucose level was available for only two patients who had been diagnosed with diabetes prior to the surgery and who subsequently developed a periprosthetic joint infection, the effect of glucose control on the infection rate in diabetic patients could not be analyzed.
Other Factors Associated with Periprosthetic Joint Infection in Obese and Diabetic Patients
In patients with a preoperative diagnosis of diabetes, periprosthetic joint infection was associated with male sex (68.8% [eleven of sixteen] of diabetic patients with infections were male compared with 40.2% [349 of 869] of diabetic patients without infections; p = 0.021) and with a higher BMI (median, 34.3 kg/m2 compared with 27.9 kg/m2; p = 0.017). Age, duration of the operation, perioperative blood loss, ASA risk score, bilateral procedures, surgeon experience level (orthopaedic surgeon or resident), and fixation method had no significant effect on the infection rate. The rate of periprosthetic joint infection varied according to the timing of the diabetes diagnosis (as reflected by the time at which reimbursement for diabetes medications was approved relative to the time of the surgery) (Table II). Use of a biguanide medication (e.g., metformin) was associated with an increased infection rate in the univariate analysis (OR, 3.31; 95% CI, 1.27 to 8.68) but not in the multivariate analysis (adjusted OR, 2.31; 95% CI, 0.84 to 6.36). Other diabetes medications had no significant effect on the infection rate.
In obese patients, periprosthetic joint infection was associated with male sex; 50.0% (13) of 26 obese patients with an infection were male compared with 31.0% (732) of 2361 obese patients without an infection (p = 0.038). Periprosthetic joint infection in obese patients was also associated with a higher ASA risk score (median, 3 compared with 2; p = 0.002) and with diabetes. Age, duration of the surgery, perioperative blood loss, proportion of bilateral surgeries, and surgeon experience were similar in obese patients with and without periprosthetic joint infection. The proportion of hybrid and uncemented prostheses tended to be higher in morbidly obese patients who developed a periprosthetic joint infection (33% [three of nine] and 22% [two of nine], respectively) than in those who did not (14% [twenty-five of 184] and 10% [eighteen of 184]; p = 0.059).
Sensitivity Analyses
The association of diabetes with periprosthetic joint infection was similar when the definition of diabetes was based on eligibility for reimbursement of the cost of diabetes medications (adjusted OR, 2.44; 95% CI, 1.18 to 5.06) compared with when it was based on purchases of such medications (adjusted OR, 2.22; 95% CI, 1.18 to 4.97). The adjusted odds ratios for patients with morbid obesity and with diabetes remained essentially unchanged when the subgroup with cemented prostheses and the subgroup with unilateral operations were analyzed separately, and also when the definition of periprosthetic joint infection was restricted to an organ or space infection (n = 40) and to a culture-positive infection. When operations performed by orthopaedic surgeons were analyzed separately, the adjusted ORs increased for patients with diabetes (2.99; 95% CI, 1.35 to 6.60) and for patients with morbid obesity (10.08; 95% CI, 1.87 to 54.24), probably reflecting residual confounding related to the complexity of surgery or to other factors that could not be measured.
Our study confirmed that both morbid obesity and diabetes (independent of obesity) increased the risk of periprosthetic joint infection following primary hip and knee replacement performed to treat osteoarthritis. Moreover, there was a trend toward an association between the preoperative plasma glucose level and the periprosthetic joint infection rate in patients without a prior diagnosis of diabetes.
The main strengths of our study are the use of a large population-based clinical series, which allowed analysis of the independent effects of obesity and diabetes, and the use of infection data based on prospective active surveillance performed with accepted diagnostic criteria and methodology. Ideally, the effects of obesity, diabetes, and glycemic control on the periprosthetic joint infection rate should have been analyzed in a prospective setting. However, current infection rates are low, which would make collecting sufficient data for such a study in a timely acceptable fashion very difficult. Although the data on infections were based on prospective surveillance, it is possible that some infections went undetected13,19, but it is unlikely that this would have altered the key findings of our study.
The two-fold increase in the infection rate in patients with a diagnosis of diabetes compared with nondiabetic patients is comparable with the findings of several earlier studies11,23,24 although even higher odds ratios have been reported in other studies8,12. Still other studies have not shown an increased risk in diabetic patients5,14,25, possibly because of the small number of patients with an infection.
Factors that may affect the risk of periprosthetic joint infection in diabetic patients include the type of hyperglycemia treatment, glycemic control, occurrence of diabetes-related complications, and postoperative hyperglycemia. Meding et al.25 reported that insulin-dependent diabetes was associated with an increased periprosthetic joint infection rate (3.4% compared with 0% in patients not receiving insulin [p < 0.05] and 0.7% in patients without a diagnosis of diabetes), corroborating earlier studies26,27. In contrast, the type of diabetes treatment had no effect on the risk of periprosthetic joint infection in diabetic patients in the multivariate analyses in our study.
Marchant et al.23 reviewed a large register-based series of hip and knee replacements and reported that the incidence of wound infection during the immediate postoperative hospitalization did not differ between patients without diabetes and those with controlled diabetes. However, the infection rate was 2.3 times higher in the group of patients with poorly controlled diabetes. Similarly, Pedersen et al.24 reported a two-fold increase in the rate of revision hip replacement due to infection in patients with diabetes-related complications. Patients with uncomplicated diabetes had a revision rate that was similar to those without diabetes. Because of the register-based nature of those studies, the effects of obesity, the glucose level, and the glycosylated hemoglobin level could not be analyzed. Moreover, the proportion of patients with poorly controlled diabetes was only 3% in the former study23, although others have reported that almost half of diabetic patients do not meet the treatment targets28,29. This suggests that the patients with uncontrolled diabetes in the study by Marchant et al. represented those with the poorest glycemic control.
Interestingly, diabetic patients in our study whose diagnosis was within one year prior to surgery had the greatest elevation in the infection rate. This finding is in accordance with a large Danish register-based study in which diabetic patients who had undergone primary hip replacement had higher odds of requiring revision arthroplasty because of deep infection if the diabetes diagnosis had been made less than five years before the surgery24. Poor glycemic control is a possible explanation, but we were unable to test this hypothesis in the present study.
Obesity does not appear to play a major role as a predictor of periprosthetic joint infection as long as the BMI remains below 40 kg/m2. Ignoring the effects of concomitant diabetes may explain the inconsistency among earlier studies focusing on the association between obesity and periprosthetic joint infection4,5,7,11,12,14,27. In the present study, it is noteworthy that the preoperative glucose level in obese patients without a preoperative diagnosis of diabetes appeared to be associated with the periprosthetic joint infection rate (Fig. 2). As the hyperglycemic patients in this group probably represent previously undiagnosed diabetics, identifying them preoperatively would not only be valuable for achieving safer joint replacement surgery but would also be important for their overall long-term prognosis.
In accordance with earlier studies8,9,30,31, morbidly obese patients had a high periprosthetic joint infection rate after both hip and knee replacement. Wound-related complications are common in this patient group30, but greater technical difficulty and longer duration of the operation, poor vascularization of fat tissue, and comorbidities are also possible causative mechanisms. Diabetes in particular appeared to play a prominent role (Fig. 1); in fact, the periprosthetic joint infection rate was close to 10% in morbidly obese patients with diabetes, which raises the question of whether performing joint replacements in patients who are both diabetic and morbidly obese is appropriate.
Since bariatric surgery leads to the resolution of diabetes in a considerable proportion of patients (by correcting insulin resistance)32-34 and also reduces musculoskeletal pain35, it might be a suitable first-line intervention for some morbidly obese patients with osteoarthritis. At a minimum, screening for diabetes and optimizing glucose control preoperatively would appear to be reasonable in this patient group.
The laboratory data used in this study were collected as part of the preoperative routine rather than for scientific purposes, which created some limitations. First, glucose data were not available for 24% of the patients because the blood tests were performed in outside laboratories. Second, we were unable to identify patients who had diabetes at the time of the surgery but had not previously been diagnosed as diabetic by the end of the study period, because fasting blood plasma samples were not routinely collected and an oral glucose tolerance test (the gold standard for diagnosing diabetes) was not performed. Although the 12% prevalence of diabetes in our study cohort is similar to that reported in a population-based study from the same geographical area36, it is known that up to one-half of diabetes cases are undiagnosed36,37. Diabetic patients who received that diagnosis only after surgery and those who remained undiagnosed would have been analyzed as part of the nondiabetic group, and this would have decreased the reported odds ratios for the effects of diabetes. Third, the effect of glycemic control on the periprosthetic joint infection rate could not be analyzed because the glycosylated hemoglobin level was not measured routinely, even in patients known to be diabetic.
Given the epidemic of obesity and diabetes, it is clear that these two conditions cannot be ignored in orthopaedic practice3. Potentially modifiable risk factors, including diabetes and diabetic control, should be targeted preoperatively in order to reduce the periprosthetic joint infection rate. Screening obese patients for diabetes and treating hyperglycemia in known diabetics perioperatively and postoperatively appear to be reasonable even though evidence showing the value of such practices in preventing periprosthetic joint infection is lacking38.