Does the Type of Ventilation Affect the Risk for Infections After Hip Replacements?

Overview

Background: During hip replacement surgery, there is a risk that bacteria in the operating room can cause an infection. To try and reduce this risk, some operating rooms use a special system called laminar airflow (LAF), which reduces the number of bacteria in the air. However, it's not clear if LAF is better than the older system, called turbulent airflow (TAF), for preventing infections. Aim: The aim of this study is to compare the two airflow systems and see if LAF is better at preventing infections after hip replacement surgery. Methods: Information from a database containing all hip replacement surgeries done in Denmark between 2010 and 2020 is examined. The number of infections that occur in surgeries done with LAF, which reduces the number of bacteria in the air during surgery, is compared to the number of infections that occur in surgeries done with TAF. To make the results more credible, the data from the hip register was combined with data from the bacterial cultures taken during surgery. Use and relevance: Infections after hip replacement surgery can be very serious and expensive to treat. Hospitals need to choose the best airflow system to help prevent these infections. This study is important because it gives more accurate information about which system is better at preventing infections and can help hospitals make better choices when they are designing or renovating operating rooms.

Rationale and background Prosthetic joint infection (PJI) is one of the most feared complication after total hip arthroplasty (THA), with increased mortality and morbidity, as well as a significant cost for society. Since the 1960s, in addition to systemic prophylactic antibiotics and antibiotic-laden bone cement, efforts have been made to make the air in the operating rooms (OR) as clean as possible with filtering and replacement of the air . This has led to the official current PJI incidence of 0.57-0.80 % in the first year after surgery according to the Danish Hip Arthroplasty Register (DHR) and the Swedish Hip Arthroplasty Register (SHAR), even though the true incidence of PJI probably is about 40 % higher, due to under-registration in the registers following revision surgery . The risk of PJI is potentially affected by contamination in the OR, patient related risk factors and prosthesis factors. The prosthesis factors include type of implant and implant fixation method. Cemented implants have a lower overall revision rate, although the results are not unanimous when it comes to revision due to PJI. The risk of contamination in the OR is affected by the cleanliness of the air. There are two different types of ventilation systems used to achieve a clean environment in an OR. The first is turbulent airflow (TAF), where the air is continuously replaced with filtered ultra-clean air that trickles down from the ventilation shaft. The second is laminar airflow (LAF) where the ultra-clean air flows in a single direction, with the goal to prevent particles carrying bacteria, that are floating in the air, to land on the patient or sterile equipment. LAF decreases the bacterial load significantly in the OR, measured in CFUs. However, the number of CFUs has not been found to correlate with the incidence of PJI in real world data, where most studies show no advantage of LAF compared to TAF. Only one major observational study shows that LAF decreases the risk of infection in a subgroup of LAF . All other register studies show no difference between LAF and TAF or a higher risk for PJI with LAF. There are no randomized, controlled trials (RCTs) comparing modern-day ventilation systems when it comes to the PJI incidence in THAs. There are no randomized, controlled trials (RCTs) comparing modern-day ventilation systems when it comes to the PJI incidence in THAs. However, in 1982, an RCT was conducted by Lidwell et al., comparing different ultraclean air flow systems, including LAF, with standard operating room ventilation at the time. In this study, there was a significant difference in PJI incidence with 2.2 % PJIs in the standard OR group and 1.0 % in the ultraclean group. However, these standard ORs had much higher bacterial contamination, with a median of 164 CFUs/m3 , compared to modern-day TAF ORs, that have a median ranging from 4,5-22 CFU/m3. The ultraclean group had a median of 2-10 CFUs/m3, thus it resembles modern-day TAF ORs, both in bacterial load and PJI rate . Due to additional costs in building and maintaining LAF systems, a public analysis was conducted in Denmark in 2011 to compare the cost-effectiveness of LAF and TAF. When considering the estimated annual costs, which included operational expenses, depreciation, and interests, LAF ORs were estimated to cost 508,732 DKK per year, whereas TAF ORs were projected to cost 304,530 DKK per year. As a result, there was an annual cost difference of 204,202 DKK (€27,000) per unit, with LAF being 67 % more expensive than TAF. A PJI was estimated to cost the society around 204,000 DKK in the same year . An issue in the previous studies on PJI in LAF and TAF are that they are epidemiological and rely solely on the PJI diagnosis reported to the different registers. However, diagnosing a PJI is not always evident in a clinical setting. It has been shown that the DHR is lacking both registration of revisions and later confirmed PJI , and a lack of registered PJI revisions has also been observed in the Swedish hip arthroplasty register (SHAR) . The PJI diagnosis in the DHR can be greatly improved, if combined with the Danish Microbiology Database (MiBa) . When validating the diagnosis PJI in DHR, the sensitivity was found to be only 67 %. When using an algorithm including data from MiBa, sensitivity increased to 90 %, without providing more false positives. A simplified algorithm utilizing MiBa data exhibited no statistically significant difference when compared to the more extensive algorithm employed in the validation study by Gundtoft et al.. This simplified algorithm, which doesn't require access to medical records, will be applied in the criteria for the primary outcome. Additionally, previous studies show that one positive bacterial sample at a clinically aseptic revision increases the risk of an occult infection, diagnosed later at a re-revision. This will be investigated as well. Thus, using these new and well validated methods, the investigators aim to estimate the risk of PJI with LAF versus TAF ventilation better than has been done in previous studies. Data sources The Danish Civil Registration System (Centrale Person Register, CPR) contains information including vital status and time of emigration on all Danish residents . All Danish residents and citizens are assigned a unique and permanent individual identification number (CPR number) at birth or upon immigration. It enables an unambiguous linkage between different registers and allows an individual follow-up over time . The Danish National Patient Register (NPR) collects data of all Danish residents contact to a public hospital, including date, performed examinations or surgical procedures and diagnosis, classified by International Classification of Diseases (ICD) . The data includes both outpatient and inpatient contacts and is linked with the patient's CPR number . The Danish Hip Arthroplasty Register (DHR) collects data including type of OR ventilation from all THAs performed in Denmark, including primary THAs as well as revisions . It is mandatory for all surgeons performing THAs to report to the register, which results in high completeness. The data is previously validated . In the case of a revision, the surgeon reports the indication immediately after surgery . The revisions performed within 90 and 365 days are located in the DHR, which also contains information from the NPR about revisions. The Danish Microbiology Database (MiBa) automatically collects all microbiology results from all departments of clinical microbiology in Denmark since 2010, which are then stored electronically using the patient's CPR number as patient identifier . The register completeness is previously validated . The data will more specifically be extracted from the Healthcare-Associated Infections Database - HAIBA, a part of MiBa. The Danish National Prescription Register (DNPR) collects detailed information on redeemed prescriptions in Denmark since 1995. It is previously validated and has a high degree of completeness and is frequently used in epidemiological research . Income Statistics Register - The register contains income data on more than 160 variables including salaries and savings of people with a Danish income and is widely used in research. Each individual is identified with an unambiguous CPR number . Danish Education Register - The register has a high coverage on education level, both among people born in Denmark and immigrants and is widely used in research. Each individual is identified with an unambiguous CPR number . Variables The investigators will look at several variables and adjust for these where needed. The variables obtained from DHR is body mass index (BMI), age, gender, time of surgery, diagnosis, type of prosthesis and fixation method, whether cement is used and if it contains antibiotics, previous operations on the same hip and ASA score. ASA score and BMI are only available from 2017 and onwards in DHR. From NPR, concomitant diagnoses will be obtained to produce an Elixhauser comorbidity score, and socioeconomic factors, such as income and education will be obtained from the Danish Income Statistics Register and Education Register. Elaboration of the methods part: A follow-up time of 90 and 365 days respectively is chosen, since 90 days is the recommended surveillance period in the United States and is traditionally the limit between early and late infections, whereas it is important to have a follow up of a year not to miss up to 20 % of the infections. At 1 year, 80-90 % of the surgically requiring infections that will occur in the lifetime of the hip prosthesis have taken place. After 1 year, the infection rate drastically decreases and is less likely to be due to contamination during surgery. The definition of PJI is based on the studies by Gundtoft et al. If, where it has been concluded, that the validity of the PJI diagnosis is greatly improved when using these criteria, which are validated specifically for DHR. The original study by Gundtoft et al. used an extensive algorithm to determine if a PJI was present. However, there was no statistically significant difference between the extensive algorithm and a simplified algorithm, that defines a PJI as two or more positive identical bacterial samples. The investigators definition is in line with international consensus and does not differ greatly from the definition made by the European Bone and Joint Infection Society (EBJIS) , although erythrocyte sedimentation rate, aspiration of joint fluid, white blood cell count in joint fluid and histological examination of intraoperative tissue biopsies are not routinely used in Denmark. Thus, these criteria are not used. Sinus tract communication with the hip joint or an opening to the hip joint is expected to cause the surgeon to register the diagnosis "infection" in DHR upon revision. Data for this analysis will be extracted from DHR and HAIBA. For the revision surgery to be registered in HAIBA, at least three biopsies at the revision surgery are required, hence this is a part of the criteria. Aseptic loosening is a separate endpoint, since it has been shown that aseptic loosening might in fact turn out to be a PJI and is thus a point of interest. A sensitivity analysis is made for a) and b), since Milandt et al. have shown that having one positive intraoperative bacterial culture increases the risk of an occult infection, that is later diagnosed upon re-revision. When there was only one unexpected positive culture (unexpected meaning not registered as PJI in DHR), first-time revisions were associated with an increased relative risk of 2.63 for subsequent re-revision specifically for PJI. However, this association was not observed for the patients with two or more unexpected positive cultures, which were much more likely to receive antibiotics and be treated as PJI, after the microbiology results were obtained. This is most likely, because one positive culture has been interpreted as contamination. A sensitivity analysis is made for c), since DHR only have ASA and BMI registered starting in 2017 and thus this cannot be adjusted for prior in the data prior to 2017. EBJIS have proposed a term called PJI-likely which include 1 positive unexpected positive culture. PJI-likely is based on international consensus; however, the definition relies on analyses that are not commonly performed in routine practice in Denmark, as stated in the description of the primary endpoint. Since it is not possible to use these criteria in Danish register studies, the investigators have instead decided to do sensitivity analyses based on the validation of infections in the DHR by Gundtoft et al. and the findings of Milandt et al.. Statistical analysis: Descriptive results for continuous variables will be shown as mean(SD)/median(range) depending on the distribution of the variables. Categorical variables will be presented using frequencies and percentages. Persons operated in LAF and TAF ORs are compared using t-tests/Wilcoxon-tests and chi-square tests, respectively. This is supplemented with cumulative incidence plots showing risk of infection (primary outcome) for LAF and TAF, regarding revision due to other causes and death as competing events. The two groups are compared using Grays test. To fulfill the requirements at Statistics Denmark about data on individual persons, a smooth curve is used to make the plot rather than the statistical correct step-function. Information at time 90 and 365 days after the operation will be presented. This type of plot will be used for all endpoints. For the primary endpoint, two Cox regression analyses will be performed with the time to PJI as outcome. One model will only cover the first 90 days of follow-up, the other will cover all 365 days of follow-up. Censoring is made at time of revision due to other causes than PJI, death, emigration, and end of follow-up (365 days after operation), whichever comes first. Bilateral THA will be included in the model as two observations with inclusion of a time-dependent variable conditioning on the status of the other hip . A robust sandwich covariance matrix estimate is used to account for the intrapersonal dependence. The model will include the following potential confounders: Education, age, comorbidities, gender, previous hip surgery, primary diagnosis, type of cement, type of fixation, year of surgery, income and duration of surgery. ASA and BMI will be adjusted for in sensitivity analysis c). A direct acyclic graph is used when determining the variables needed to adjust for. The linearity assumption for continuous confounders is checked by including these as cubic splines with 3 to 7 knots and using the model with lowest AIC, as suggested by Harrell . The proportional hazards assumption for the Cox-regression models is assessed using plots for the cumulative sum of martingale residuals. If the assumption is violated for the exposure (LAF/TAF), the time-axis will be split into smaller parts in which the assumption is fulfilled, and a separate Cox regression is made for each part. If the proportional hazard assumption is violated for confounders, time-dependent covariates is used for these. Similar Cox-regression models will be used to analyze both secondary endpoints, as well as for the sensitivity analysis of the primary endpoint. The assumptions of proportional hazards for these endpoints are checked as described for the model for the primary endpoint. All tests will be two-sided and a p-value of \<0.05 will be considered statistically significant. Statistical analysis will be conducted using SAS version 9.4 (SAS Institute Inc.). Missingness Thanks to the Danish registers, there is no loss to follow-up for the primary and secondary endpoints, nor for the endpoints of the sensitivity analyses. Information for most of the confounders is based om mandatory registration in DHR, thus a very limited amount of missing data in the confounders is expected (less than 5 % with missing information about one/several confounders). The missingness is not likely to be associated with the exposure, thus the complete case analysis is expected to be valid . If the missingness is found to be more pronounced, multiple imputation will be used to create a full analysis data set. The imputation will use 100 samples. Multiple imputation will be made using R (The R Foundation). Results from analysis on the full analysis data set as well as for the complete case analysis will be presented.