Journal of the New Zealand Medical Association, 25-January-2008, Vol 121 No 1268
The influence of steroid injections on the incidence of infection following total knee arthroplasty
Geoffrey Horne, Peter Devane, Andrew Davidson, Kathryn Adams, Gordon Purdie
The initial management of patients with arthritis of the knee should be non-operative. Joint aspiration and intra-articular steroid injection are techniques that may be used to improve synovitis in these patients.1 However, total knee arthroplasty is the most effective treatment for the relief of pain in severe arthritis of the knee.1,2
The most severe complications of TKA are wound and joint infection which occur in about 0.5–2% of operations performed.1–6 The mortality rate for prosthetic joint infection may be as high as 2.5%.7 Infection may occur early in the postoperative period, in which case the joint may be salvaged by exploration and synovectomy, or it may be a late, deep joint phenomenon where complete removal and revision of the joint in a two step process is indicated.1,5,7
Risk factors that have been identified as being related to a higher risk of surgical site infection or deep wound infection include obesity, diabetes mellitus, malignancy, smoking, renal failure, urinary tract infection, rheumatoid arthritis, previous knee surgery, and prolonged postoperative drainage for longer than 6 days.1,2,4,5,8,9
Recent research has demonstrated an increase in deep infection rate and wound healing problems after intra-articular steroid hip injection (IASHI). prior to arthroplasty.10,11 These findings support the outcome of a systematic, comprehensive survey of the expert opinion and current practice of orthopaedic surgeons in Ontario, Canada, that reveal that a substantial number of surgeons believe that the infection rate related to THA may be increased after intra-articular steroid hip injection.12
To date, published studies that assess the effect of intra-articular steroid injections on the risk of infection following total arthroplasty in knees, have offered differing results.13,14
For example, Joshy et al, in a recent retrospective cohort study found no significant difference in the numbers of patients who received intra-articular steroid injection between the infection and non infection groups.13 They concluded that infection following total knee replacement is due to multiple factors and that the use of intra-articular steroids does not alter the incidence of deep infections following total knee arthroplasty.
Conversely, in a more recently published retrospective study, Papavasiliou et al reviewed 231 patients who had undergone total knee replacement over a period of 2.5 years.14 These researchers concluded that the decision to administer intra-articular steroids to a patient who may be a candidate for total knee replacement should not be taken lightly because of a risk of postoperative deep infection.
In the light of these recent differences in research findings, this case-control study aimed to assess the influence of preoperative intra-articular corticosteroid injection on the incidence of postoperative wound healing problems and infection in patients undergoing total knee joint replacement surgery at two university-affiliated tertiary hospitals in Wellington, New Zealand.
It was hypothesised that a corticosteroid injection into the knee joint at any time prior to the operation would increase the risk of wound healing problems and infection of the joint following surgery, with resultant worsened knee symptoms as reported by patients on Oxford Knee Scores. We also aimed to assess whether the outcome was influenced by the person who administered the injection, for example, whether it was the patient’s general practitioner, orthopaedic surgeon or rheumatologist.
As well as the rate of wound healing problems and infections, quality-of-life measures have been shown to be valuable in assessing the success of joint replacement surgery.15 Oxford scores are considered a simple and effective method for assessing quality of life.16
The Oxford Knee Score (OKS) contains 12 questions, each of which has five options graded 1 to 5. The scores are summed. Higher scores indicate more severe symptoms from the knee in question.
The study was a retrospective, case control design. The two groups for the study were a control group of patients who had undergone total knee joint replacement without subsequent infective complications, and a group of patients who had either had postoperative wound healing problems or revision surgery for deep infection at any stage following initial surgery.
A list of all patients who underwent primary total knee arthroplasty at Hutt Hospital, Wellington between July 1999 and December 2002, and at Wellington Public Hospital between January 1992 and December 2002 was obtained from the computerised medical records systems at the two hospitals.
From these lists it was determined which patients either had a subsequent revision surgery on the same knee that was replaced, or were readmitted within 6 months with wound healing problems, that is, proven infection, but not requiring re-operation. Further data was then gathered, finding all patients who underwent revision to a total knee joint arthroplasty between January 2003 and December 2004 at each of the two hospitals.
The clinical records of all patients who had either been readmitted within 6 months for wound healing problems or had revision knee joint surgery were then reviewed to determine which cases were in fact due to suspected infection and which were due to other problems such as mechanical failure.
Patients who had either a deep joint infection requiring revision surgery, or who had a wound healing problem with suspected infection requiring readmission were included in the study group.
The group of patients from Wellington hospital who had had total knee arthroplasty between 1992 and 2002 and not experienced a subsequent infection was much larger than was required (1200), so a random sample of 165 cases was taken from this sample to make it more manageable for comparison with the study group. It was calculated that we would need sample sizes of 152 controls and 38 infected to achieve a power of 80%, given an estimated odds ratio of 3.0. This estimate was conservatively based on a previous study that had found a relative risk of 4.0 in hips injected with steroids.10
A reply-paid letter was then sent to all patients, outlining the purpose of the study and asking patients whether they had ever had an injection into their knee joint, at any stage before their total knee arthroplasty (Appendix 1).
The patients were asked simply whether they received any injection before surgery. They were not asked to identify the reason for the injection. We also asked patients to indicate whether that injection was performed by a rheumatologist, general practitioner or orthopaedic surgeon; how many injections they received and how long before the operation the last injection was given. The patients were also asked to indicate whether, at the time of surgery, they had diabetes or they smoked cigarettes. They were also asked to complete a validated knee assessment tool, the OKS (Appendix 2).
In all, 392 patients were contacted. 352 were controls and 40 were infection cases. The reply rate was 65% overall after 4 weeks.
Two groups were thus produced:
Of the non-responders we contacted the general practitioner of all of the infection cases and a sample of one-third of all the control cases, to see if any had changed address or were deceased. To improve the reply rate, especially for the infection cases, those still living at the same address were contacted again by mail.
The reply rates for this second round of letters increased the overall reply rates to 69.6% for controls and 77.5%% for study cases after one week, at which time we finalised our data set. After removing patients who were deceased, or who failed to reply adequately we were left with sample sizes of 219 controls and 29 study cases.
The data was then analysed statistically using EpiInfo (Centers for Disease Control & Prevention, Atlanta, GA, USA) for an association between preoperative intra-articular steroid injections and wound or joint infection following operation.
Secondary relationships to be assessed were between each of smoking, diabetes and infection; between infection and Oxford Knee Score; and between injection by each of rheumatologists, general practitioners, and orthopaedic surgeons and infection rate following surgery.
Finally we looked at whether the time before the operation that the injection was given, and the numbers of injections given, influenced infection rates.
The overall reply rate for this study was 70.4%. The infection case reply rate was 77.5% and for non-infection controls it was 69.6%. The difference between rates of reply was not significant (Table 1).
Table 1. Reply rates by hospital and overall
For all respondents, 13 patients were deceased or unable to reply. A further 15 controls were discarded due to an incomplete OKS or steroid injection questionnaire. One patient filled in only the OKS, and 35 filled out the injection questionnaire but did not fully complete the OKS. Patients who adequately filled in one of the two sections were included in the study.
In all, 32.8% of patients had received a steroid injection at some stage before knee replacement surgery. Of all patients, 7.3% claimed to be smokers at the time of their surgery and 8.2% had diabetes. The average OKS was 26.03 for all subjects. The national mean for OKSs at 6 months post-primary TKA was 23.22 in 2003.
Most injections were delivered by general practitioners (GPs) who gave 37%, with orthopaedic surgeons giving 35% and rheumatologists administering 22%. The remainder reported more than one injection, given by different doctors. The average number of injections (of all those who received any injections) was 2.23 with a range of 1–15. The mean time before surgery from the last injection was 16 months, with a range of 1 month to 45 years. A telephone call to the respondent verified the 45-year interval. 79% of patients claimed that their last injection was within 12 months of surgery.
The rates of pre-operative injections in each of the control and study groups were 32.0% (17/28) and 39.3% (70/219) respectively with an odds ratio of 1.38 (95%CI 0.55–3.31). There was no significant difference detected (p= 0.44).
The rates of diabetes were 7.3% and 14.3% in the control and infection groups respectively (OR 2.09; 95%CI 0.47–7.22) and smoking occurred at rates of 6.9% and 10.7% (OR 1.62; 95%CI 0.28–6.33). The rates of reported diabetes and smoking were not found to be significantly different between the two groups (p=0.21, p=0.46 respectively) (Table 2).
Table 2. Rates of injection, diabetes, and smoking in the infection and control groups
The administrators of steroid injections were relatively mixed for the control group and mostly orthopaedic surgeons for the infection group, however the difference between the two groups was not significant (Figure 1).
Figure 1. Injection administrators
The mean OKS was 25.3 for the control group (95%CI 23.8–26.8) and 31.9 for the infection group (95%CI 26.9–36.8). There was a significant difference in this measure between the two groups (p=0.014).
Table 3. Mean Oxford Knee Score (OKS) for control and infection groups
This case control study aimed to examine the effect of pre-operative steroid injections on the rate of postoperative wound healing problems and joint infection in the knees of patients undergoing total knee arthroplasty. It was hypothesised that the results would mirror those of previous studies that have found higher rates of post surgical problems in hips that have been injected with steroids.
The results to this study indicated that there was no increased risk of postoperative wound healing problems or infection attributable to patients receiving an injection into their knee before surgery. While the response rate to the letters was satisfactory, due to the rarity of postoperative infection, trial size of the infection group was based on feasibility.
There was no detectable increased risk of wound or joint problems related to either diabetes or cigarette smoking. Only 7.3% of patients claimed to have been cigarette smokers at the time of surgery. The expected number of cigarette smokers in a population of New Zealanders 50 years of age or older, would be 10–15%.17
It was decided not to use time before operation and number of injections given as measures since the replies were often non-specific and the data was skewed by the occasional extreme case. As an example, one patient reported that his last injection was 45 years before the operation. This patient had a steroid injection into his knee following a rugby injury as a young adult. His knee stiffened afterwards, so he determined never to have another injection into the knee joint.
The overall mean OKS (26.03, 95%CI 24.55–27.52) was slightly above the 2003 national mean for TKA recipients of 23.22. However, the national mean figure was for patients reporting 6 months after surgery whereas many patients in the study population would have been reporting at a much longer interval following their operations. This could account for the knee symptoms of the study group being of less severity than the national average.
A significant difference was found in OKS between the control group and the infection group, with the latter reporting more severe knee symptoms. This is not surprising, as those postoperative complications are likely to have caused a less satisfactory state of the knee.
The person who administered injections into the knee was thought to be a possible important variable by the authors of studies on hip replacement surgery. This study did not find a significant difference in the degree of safety with which general practitioners, rheumatologists, and orthopaedic surgeons administered injections into the knee joint.
Some weaknesses were identified during the course of this study. Patients who indicated that they had an injection into their knee before surgery may not in fact have been injected with steroids, and the accuracy of recall (particularly with an ageing study group) must be taken into consideration. However the fact that an injection had been administered would tend to over-estimate the influence of steroid injection if only local anaesthetic were injected, and since there was no association found it is extremely unlikely that steroid injection has an influence on infection rates.
The strengths of this study included the high patient response rate and the simplicity of the questionnaire which allowed patients to provide useful information. It also provided a virtually exhaustive database of infection and wound healing cases for the last 10 years, in the area studied.
Our results did not indicate that intra-articular steroid injection was contraindicated in patients who are candidates for total knee replacement surgery, although patients who had infective complications following TKA did have a less favourable long-term knee function.
This is a small, retrospective study in a limited number of patients, undertaken primarily to assess long-term outcomes of total knee arthroplasty, comparing those who had prior intra-articular injections, to those who did not.
This study suggests the outcome for patients who had intra-articular injections prior to TKA, is no less successful than for those who did not have injections. The findings indicate that if injections into the knee do have an effect, then it is much smaller than that seen in hip replacement surgery.
The recent published research on infection rate relationships to prior intra-articular injections and total knee arthroplasty have all been retrospective studies. For this reason, and because the fewer than expected infected cases presented for this study, it is recommended that a larger, prospective, multicentre study be undertaken, that could more confidently determine these relationships.
Competing interests: None known.
Author information: Geoffrey Horne, Orthopaedic Surgeon1; Peter Devane, Orthopaedic Surgeon1; Kathryn Adams, Orthopaedic Research Nurse1; Andrew Davidson, Medical Student2; Gordon Purdie, Biostatistician2
Acknowledgements: We thank the New Zealand Orthopaedics Association for their support in this project as well as George Serbiesku at Wellington Hospital and Pauline Barclay at Hutt Hospital for their assistance with data collection.
Correspondence: Kathryn Adams, Orthopaedic Research Nurse, Orthopaedic Department, Wellington Hospital, Riddiford Street, Wellington, New Zealand. Fax: +64 (0)4 3855831; email Kathryn.email@example.com
We are currently conducting a study looking at whether there may be an association between injections into the knee before surgery and problems with wound healing after knee replacement surgery.
We are trying to establish whether patients had an injection into their knee at any time before the knee operation (into the same knee that was operated on). You have been identified from hospital records at Hutt Hospital or Wellington Hospital as having had a total knee replacement operation between 1999 and 2002.
We would greatly appreciate it if you could answer the following questions and fill out the attached questionnaire, returning the completed forms to us in the envelope provided.
Andrew Davidson; Medical Student, University of Otago
Professor JG Horne; Orthopaedics Department, Wellington Hospital
1) Did you have an injection into your knee from a GP, rheumatologist or orthopaedic surgeon, at any time before you had a knee replacement operation?
c) How long before your operation was the last injection?
2 – At the time of surgery were you:
b) A smoker?
3) – Enclosed you will find a knee scoring chart. Please fill this out with regards to the current condition of your knee.
issue | Search journal |
Archived issues | Classifieds
| Hotline (free ads)
Subscribe | Contribute | Advertise | Contact Us | Copyright | Other Journals