Operating room ventilation and the risk of revision due to infection after total hip arthroplasty: assessment of validated data in the Norwegian Arthroplasty Register

This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.


Introduction
Infection after total hip arthroplasty (THA) is devastating for the patients and generate high public costs [2]. The air in the operating room (OR) is considered a potential source of contamination and subsequently a risk factor for surgical site infection (SSI) [3][4][5][6] due to airborne bacteria and other viable microorganisms (colony-forming units (CFU)) shed from the surgical staff or the patient itself [7,8]. The amount of CFU in the OR may be altered by staff behavior such as the number of personnel, door openings, physical movement and the use of other preventive measures such as impermeable gowns and space helmets [9][10][11]. Previous studies have postulated that the density of CFU is correlated with the rate of postoperative infection [7,[12][13][14][15], but the findings are controversial as the isolated effect of air cleanliness is hard to assess [16]. Other studies show that air contamination is not directly associated with wound contamination and periprosthetic joint infection (PJI) [17,18].
Unidirectional airflow (UDF or UDAF) systems (formerly known as laminar airflow (LAF) systems) have been used during ultraclean surgery since the late 1960's, as they were thought to reduce the incidence of SSI by reducing the CFU density [14]. UDF systems work by sending parallel, filtered air streams with constant velocity directly on to the surgical field to intentionally displace and reduce the flow of less clean air from the rest of the OR to the surgical field. This is in contrast to the conventional ventilation (CV) systems, which utilizes the dilution principle. CV systems supplies turbulent air in order to dilute airborne contamination, mixing polluted air with clean air, and are often termed turbulent-and/or mixing ventilation systems [19].
UDF as a prophylactic measure against SSI has been supported ever since Lidwell and colleagues published their randomized, clinical trial in the 1980's [20]. For THA and total knee arthroplasty (TKA), they found lower risk of "deep joint sepsis" after arthroplasty performed in ultra-clean air (CFU<10/m 3 ) (UCA) with a relative risk (RR) of 0.4 compared to a control group with non-ultra-clean air. The study has been criticized for having methodological weaknesses [16,21], but both historic and recent reevaluations of the study confirm the validity of the findings [22][23][24]. Subsequent observational studies from the same decade, controlled for antibiotic prophylaxis, found no convincing influence of OR ventilation on the rate of SSI [25,26]. More recently, studies from surveillance registries have suggested that LAF actually may increase the risk of infection after arthroplasty [27][28][29]. Two recent systematic reviews, based partly on these registry studies, conclude that UDF systems should not be installed in new ORs [21,30]. One of the reviews includes an observational study from the Norwegian Arthroplasty register (NAR), not studying the effect of UDF specifically, using ventilation only as an adjustment variable in the study of infection trend [31]. In a previous validation study, we found 12% misreporting of ventilation data to the NAR, questioning the validity of studies based exclusively on ventilation data reported from surgeons or surgical departments [1]. In addition, there are numerous different configurations of UDF systems and when studying their effect on the rate of postoperative infections, it is important to know the dissimilarities between the different UDF systems and that these have evolved over the decades.
Our aim in the present study was to assess the association between validated, factual OR ventilation systems and the risk of revision due to deep infection after primary THA.

Methods
Since its inception in 1987, the NAR has registered data on primary-and revision THAs in Norway.
The register form is filled in by the surgeon immediately after surgery, containing information on patient identity, date of operation, indication for surgery and other surgery-related factors. In addition, certain patient-related factors like sex, age and comorbidity are registered. Primary THA and any subsequent revisions are linked through a unique person identity number that follows each citizen from birth to death. Revision is defined as removal or exchange of prosthesis parts, whereas revision cause, i.e. deep infection, is determined by the surgeon based on perioperative assessments and clinical evaluation. Cases of revision due to infection, is thus reported to the NAR before the culturing of peroperative tissue samples is ready. The data is validated, with 97% completeness of reporting of primary THAs, 93% reporting of revisions, and 100% coverage of Norwegian hospitals [32].
The factual OR ventilation on each hospital was validated and either confirmed or corrected in a previous study [1]. To be included as a UDF system, it had to be verified that the system had been installed with a multistage HEPA-filtered, unidirectional diffuser array. Based on technical data collected, the following classification of ventilation systems was established for further analyses: small, low volume, unidirectional, vertical flow systems (lvUDVF -volume flow rate (VFR) (m 3 /h) <10,000 and canopy size (m 2 ) <10); large, high volume, unidirectional, vertical flow systems (hvUDVF -VFR >=10,000 and canopy size >=10) and unidirectional horizontal flow (UDHF). We did not have complete data on the volume of each OR, so we were not able to calculate the exact air changes per hour (ACH). As the ACH also might be dependent on other factors, we did not include ACH in the definition of the different UDF systems. The CV systems included in this study was verified to fulfill the requirement of multi-stage HEPA-filtered air with 20 air changes per hour (ACH) and positive pressurization [33].
The period of inclusion was 2005-2015, primarily due to the fact that the patients American Society of Anaesthesiologists (ASA)-class, a risk factor associated with infection [34], was only reported to the NAR from 2005 and onwards. All patients during this period received systemic, antibiotic prophylaxis.
A separate survey confirmed negligible use of space suits and/or helmets. Three of the hospitals used space suits in very short periods of time, but discontinued the use due to loss of spatial awareness.
We had validated ventilation data on 40 out of 62 public hospitals reporting THAs to the NAR in the inclusion period [1]. Out of 60,298 THAs performed in these 40 hospitals, 2,046 were performed in a period of ventilation system exchange or update, and were excluded. 4,313 THAs performed in UDVF ventilation were excluded due to lack of detailed information on certain ventilation covariates from parts of the inclusion period, essential for our main analyses, or due to the current UDVF system not fulfilling the defined criteria for lvUDVF or hvUDVF. In addition, 2,647 THAs were excluded due to missing patient-or procedure covariates. Hence, 51,292 THAs were eligible for analyses.

Statistics
The association between OR ventilation and revision due to infection, was estimated by Cox regression analyses. Relative risk (RR), as a measure of hazard rate ratios, was calculated with 95% confidence intervals (95% CI). End-point was date of revision due to deep infection. Further, we calculated adjusted 4-year survival rates, as well as Kaplan Meier 4-year survival rates, and cumulative survival curves with OR ventilation as strata. In the multivariate analyses, we adjusted for sex, age at primary surgery, indication for primary THA, ASA-class, method of fixation, modularity of the prosthesis, and duration of surgery. Year of primary THA was adjusted for to adjust for unknown time dependent confounding. We did additional analyses with 1 and 2-year follow-up. Further, we performed additional assessments adjusting for spatial orientation of the wound in the OR, whether the wound was oriented upwards or to the side, based on an evaluation of patient positioning and surgical approach, as a potential risk factor. We performed the analyses in concordance with the guidelines for statistical analyses of arthroplasty register data [35]. P-values of less than 0.05 and non-overlapping 95% CI were considered statistically significant.
Statistical analyses were performed using SPSS version 24 (SPSS Inc., 2004) and R (R Foundation for Statistical Computing, 2014). The study was performed in accordance with the RECORD and STROBE statement.

Ethics
The registration of data and further assessment was performed confidentially on patient consent and according to Norwegian and EU data protection rules.

Results
Among the 51,292 eligible THAs, 575 (1.1%) had been revised due to infection. Demographics and distribution of risk factors in the different ventilation groups is presented in Table I. All patients received systemic antibiotic prophylaxis and all cemented THAs had antibiotic loaded bone cement.
The distribution of the risk factors was similar for the four ventilation groups, except for more uncemented THAs in the two hospitals using UDHF (one rural-and one regional hospital). In the remaining three ventilation groups, rural-, regional-, university-, and specialized elective hospitals were evenly represented. During the study period, four hospitals converted from CV to hvUDVF and  Table II. Sex, age, ASA class, and duration of surgery were associated with risk of revision due to infection.
Assessing the UDF group as one big entity, primary THAs performed in ORs with such unclassified UDF had similar risk of revision due to infection compared to CV (RR=0.9, 95% CI: 0.7-1.2). The risk of revision due to infection after THAs performed in ORs with lvUDVF and UDHF was similar to those performed in CV (Table III, Figure 2). THAs performed in ORs with hvUDVF had a lower risk of revision due to infection than those performed in CV (Table III, Figure 2). No UDF system was associated with higher risk of revision due to infection after THA compared to CV.
Adjusting for wound spatial orientation and reducing follow up time to 1 year and 2 years had only minor influences on the results. We did not have complete data on the spatial volume of all ORs in order to calculate the exact ACH, but adjusting for operating room volume in analyses of the available ORs had negligible impact on the results.

Discussion
The risk of revision due to infection after primary THA performed in ORs with hvUDVF was 20% lower than after THA performed in CV, whereas THA performed in ORs with lvUDVF or UDHF had similar risk of revision due to infection as THA performed in CV. No UDF system was associated with higher risk of revision due to infection after THA compared to CV.
Recent registry studies as well as systematic reviews and meta-analyses are questioning the effect of LAF/UDF as a prophylactic measure against postoperative infection, as they for arthroplasty suggest an increased risk of SSI and revision due to infection [21,[27][28][29][30][31]. This is in contrast to the results from our study on validated ventilation data. Recent WHO-guidelines, although conditional, recommends not to use UDF systems to reduce the risk of SSI in arthroplasty [36]. The WHO recommendation is based partly on a few observational studies with some methodological issues; no UDF system differentiation or definition based on technical specifications, limited documentation of validation on the UDF systems, and limited information on coverage or completeness of reporting of the end point SSI or revision due to infection [1,22,23,37]. Some of these studies had only six months to one year follow-up, and others had no systematic post discharge surveillance. This has been a point of debate as low-grade infections caused by airborne contaminants might be excluded as they may present at a much later stage [22,23]. Coagulase negative staphylococci (CoNS) are the most common bacteria causing revision of infected THA [38]. Since CoNS are regarded as commensal bacteria and since CoNS also have been shown to be the most common bacteria causing late infection [38], this may support that direct contamination from primary surgery is the most common mechanism of THA infection, even in infection more than 2 years after primary THA. Hematogenous seeding of CoNS is possible, but less likely to occur as this requires substantial bacteremia [39,40]. We studied the effect of OR ventilation with four years follow up, comparable to the Lidwell studies [22,41].
One possible explanation for the reported contrary effect of UDF, could be improper positioning and movement of personnel, theatre lamps etc. in the airflow [19,42,43], thereby abolishing the preventive effect by creating more turbulence. This might especially be the case in the boundary areas due to insufficient size of the protected UDF zone. Studies have shown impact of canopy size on bacterial counts in the surgical area [44,45], where the minimum size of an UDF ceiling distribution system has been recommended to be at least 320 x 320 cm for ultraclean surgery [46].
We studied the effect of canopy size on infection risk by defining cut-off for canopy size in accordance to this recommendation.
In addition, the potentially lower tissue temperature and bacterial impingement danger [47] due to disruption of the wounds own protective, thermal plume, is also claimed to disturb the effectiveness of the UDF [27,48,49]. One recent study identifies the use of UDF as a significant risk factor for hypothermia [50], and thereby subsequently a risk factor for infection of the wound [51]. To counteract these issues, forced air warming (FAW) systems have been used. These are also thought to disturb the laminar airflow [52,53]. However, recent reviews conclude that the evidence for this is sparse [54][55][56]. A recent experimental study found that the disturbing effect of FAW is counteracted by sufficient air velocity in the UDF systems [57]. As the air velocity of different UDF systems is adjustable, we could not use it as a constant adjustment variable in our analyses. It was therefore indirectly assessed by studying the VFR, varying in the range of 1,000-5,000 m 3 /h in older, low volume systems and 10,000-20,000 m 3 /h in newer, high volume systems. The latter being necessary to create velocities in the desired minimum range of 0.3 -0.38 m/s [58] through large area canopies.
UDF-systems are able to create lower CFU concentrations than CV systems both in air and close to the operation site. This is shown both in Computational Fluid Dynamics (CFD)-studies [59] and in experimental studies [44,[60][61][62][63][64][65][66]. Studies have also shown an association between the CFU concentration and SSI [7,[12][13][14][15], but the question still remains whether other risk factors such as the patient's immunological status, bacterial virulence, antibiotic prophylaxis, surgical technique etc. are indeed much more important. The latter is supported by studies showing that SSI after elective orthopedic surgery is more frequently caused by endogenous transmission than previously assumed [67,68]. Also indicating the patients' skin commensals as source of infection, are studies on the bacteriology of infected shoulder arthroplasty and postoperative infections after spine surgery, showing a high proportion of Propionibacterium (Cutibacterium) acnes [69][70][71]. This is a species known to be abundant in sebaceous glands of the skin in such regions, and as the bacteriology of infected total hip arthroplasties is different, dominated by staphylococci, this might indicate that the patient is their own source of infection. If the cleanliness of the air in the OR is the same during different types of prosthetic surgery and a significant source for postoperative infection; why does the bacteriological spectrum of infections vary between different regions of the body? This questions the extent of air cleanliness importance. Despite this, and with increasing antibiotic resistance taken into account [72], it seems logical to reduce the peroperative, bacterial load to a minimum. This will only be more important in an era with increasing microbial resistance to antibiotics [73].
Our finding of a 20% lower risk of revision due to infection after THA performed in hvUDVF compared to CV is minute, considering also that the incidence of revision due to infection is only around 1%.
However, UDF systems can create cleaner air, and taking our results into account, it seems erroneous to discontinue the use of large, high volume, vertical UDF systems in the operating rooms of the future. Technological development and multidisciplinary cooperation with focus on correct implementation and function of the ventilation systems, should be encouraged [23,65,66].
Our study is based on data from the NAR with large number of THAs, with good quality, coverage and completeness [74][75][76]. This gives us a unique opportunity to study relatively rare events, like deep infection after THA, with detailed information on surgery-and patient related confounders. Other register studies on OR ventilation are criticized for not doing a thorough adjustment of antibiotic prophylaxis, for using surgeon or surgical department reported data on ventilation, for not differentiating the UDF systems on technical specifications and for having a limited follow-up time [22]. All of our cases received systemic, antibiotic prophylaxis and the multivariate analysis were conducted on basis of validated ventilation data. Further, we did sub analyses on canopy size and VFR with 4-year follow up. All this adds strength to our study and makes it a substantial contribution to new knowledge on the field. This study will only suggest the association between OR ventilation and revision due to deep infection after primary THA. There will be unknown confounding like human behavioral factors in the OR, incorrect implementation and maintenance of the ventilation systems, and other factors potentially disturbing the UDF. We have no information on patient warming systems, use of surgical drapes, number of personnel in the room, number of door openings etc., but we have no reason to believe that this would be different between the four ventilation groups in our study. In addition, revision due to infection may be underreported [77][78][79], but as the underreporting of revision is similar between the hospitals, this will add minimal selection bias and subsequent impact on our results [32].
There has been an increase in the share of hvUDVF systems the last 20 years (figure 1). This increase is parallel to the reported, increased risk of revision due to infection after THA [80,81]. This will necessarily be a time dependent confounder in our analyses and we have addressed this by adjusting for year of primary surgery as a continuous variable in the analysis.
Only two of the included hospitals used UDHF. In addition, these two hospitals had a higher share of uncemented THAs. This may add selection bias, but the type of fixation was adjusted for.
The modularity of the prosthesis may affect the incidence of reported revision due to infection. Nonmodular/monoblock THAs (i.e. Charnley prostheses) were used by some hospitals until 2014. They do not contain modular parts, and hence, infections of such THAs treated with debridement, antibiotics and implant retention (DAIR) were therefore not reported to the NAR until 2011 from when all DAIRs were reported regardless of component exchange or not. This is in contrast to modular THAs, which contains removable components exchanged during a DAIR-procedure. Hence, these debridements were defined as revisions throughout the study period, and were subsequently reported to the NAR as such. This will potentially lead to an underreporting of revision due to infection after THA with monoblock prostheses, and was addressed by adjusting for modularity. 40 of 62 public hospitals were included. Most of the excluded hospitals performed primary THAs throughout the whole study period, as did most of the included ones, and with a completeness of reporting of more than 97% [75]. Time trends of reporting are therefore not thought to affect the findings. The reporting of primary THAs was similar in the two groups (included/excluded) and the distribution of hospital types in the two groups was also similar (rural hospitals, regional-/university hospitals, specialized elective hospitals). We therefore believe that the impact of selection bias is minimal.

Conclusions
Unclassified, unidirectional airflow (UDF) ventilation assessed as one big entity did not influence the risk of revision due to infection after primary THA compared to conventional, turbulent, mixing ventilation (CV). When differentiating the UDF systems on technical specifications, however, primary THAs performed in ORs with large, high volume, unidirectional, vertical flow (hvUDVF) ventilation systems had a lower risk of revision due to infection compared to ORs with CV. Considering also that UDF systems can create lower particle-and microbial load than CV systems, our findings support the use of hvUDVF systems for all ultraclean surgery in the future.

Acknowledgements
All authors have approved the final article. HL, EL and HD conceived and planned the study. HL collected the data. HL and HD performed the analyses. HL wrote the manuscript. All authors contributed in interpretation of the analyses and critical revision of the manuscript.
We would like to thank the NAR contact surgeons and engineers at the hospitals for contributing to data collection. We also thank Norwegian surgeons for persistently and thoroughly reporting primary THAs to the NAR.

Funding
This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Conflicts of interest
None declared.