Emperor’s New Clothes: Is Particle Disease Really Infected Particle Disease?

Abstract

Aseptic loosening remains the most significant long-term complication of total hip replacement. The current paradigm points to an inflammatory response to wear particles as its main trigger. Recently, there have been increasing numbers of positive bacterial isolates reported among patients with clinically absent infection. This paper reviews existing evidence on possible involvement of bacteria and microbial-associated molecular patterns in the pathology of so-called “aseptic loosening.”

Aseptic loosening (AL) accounts for up to 70% of revision operations after total hip arthroplasty (THA). Final clinical presentations include pain and disability due to early implant migration, late failure due to periprosthetic osteolysis, cement mantle debonding, and loss of fixation due to bone loss. The current paradigm points to an inflammatory and osteolytic granulomatous response to wear particles as the main trigger for AL, with possible contributions from increased intra-articular pressure, compromised initial osseointegration, and persistent micromotion.¹

By definition, AL occurs in the absence of clinical or laboratory signs of bacterial infection. Therefore, septic and AL was distinguished mainly by the clinical evolution of signs and symptoms preoperatively and results of cultures of fluid or material from the peri-prosthetic tissues postoperatively. However, there has been an increasing number of positive bacterial isolates among patients with clinically absent infection² Moreover, bacteria have been detected with microscopy or other methods in cases originally thought to be aseptic but later proved to be truly septic, and the bacteriology was falsely negative.

This paper reviews existing evidence on possible involvement of bacteria in the pathology of so-called “aseptic loosening.” The importance of appropriately understanding the pathophysiology of AL has important implications, as the treatment of patients with AL is completely different from those suffering from septic loosening. Thus, the change in paradigm of AL would potentially have dramatic consequences for millions of patients worldwide.

THE EXCLUSION OF INFECTION

The exclusion of infection in the revision total hip setting is difficult. As there is no single completely accurate diagnostic method, physicians rely on a combination of clinical suspicion, serological and microbiological studies and more recently, molecular techniques for the diagnosis.^3,4 Currently used microbiologic diagnostic tools rely on isolation of a pathogen by culture for diagnosis. This may fail to identify some of the microorganisms present on the implant and in the periprosthetic tissues and thus underestimates the incidence of bacterial presence in revision total hip patients.^5–7 The rate of periprosthetic hip infections, misdiagnosed as AL, remains unclear.^8,9 By way of comparison, the rate of culture-negative clinically apparent infections (without an identified pathogen) varies between 7% and 15% in different studies.^10,11 This suggests that even with fulfilled diagnostic criteria, microbiological studies are not able to completely rule in or rule out the presence of infection. This effect might be even more pronounced in the absence of clinical manifestations of infection.

POSSIBLE REASONS FOR FALSE-NEGATIVE CULTURE RESULTS

There are numerous causes for false-negative culture results in the hip revision setting, leading to the false impression of AL. First, biofilm formation on the implant decreases the number of planktonic bacteria available for subsequent culture.^12–15 Additionally, sessile bacteria enter a different metabolic state and multiply much slower and even have different antibiotic resistance profiles.^12,16 Second, the use of prophylactic antibiotics may reduce the detection rates of bacteria when using routine culture techniques.^16,17 This point is debatable, since recent research, including a single randomized clinical trial, presented contradicting data.^18,19 However, unfavorable environmental conditions (such as failure to adhere to strict anaerobic bacteriological practice) could hamper the culture results more than the presence of antibiotic in the culture medium.^5,6,20 Third, slow-growing pathogens (e.g., small colony variants of staphylococci) may not be identified during the standard incubation period of 5 days.²¹ One-fourth of the coagulase-negative staphylococci isolated from patients with confirmed septic loosening were not detected until the second week of culture.²² A large proportion of orthopedic pathogens are first to grow on culture media only beyond the 10th day.^22,23 In one study, prolonged microbiological culture for 2 weeks yielded proof of periprosthetic joint infection in a significant proportion of patients who would otherwise remain misdiagnosed as AL (26.4% of the entire study cohort) if the culture duration had been only 1 week.²² Therefore, some researchers recommend incubation for 2 weeks to improve sensitivity.^24–26 Fourth, standard culture media support only a limited group of microorganisms. This media would be inappropriate for the detection of fastidious or rare pathogens.^7,27,28 Finally, inappropriate sampling, for example, retrieving very small tissue specimens (under 1 cm³) or retrieving specimens from a nonrepresentative place can also contribute to negative culture results.²⁹ Atkins et al.³⁰ suggested in their prospective study that at least three positive samples were necessary to predict infection, regardless of the pathogen isolated.

LIMITATIONS OF CONVENTIONAL CULTURES

Negative culture results do not necessarily rule out periprosthetic infection. The sensitivity of fluid cultures range from 50% to 93%.^31–35 The sensitivity of the periprosthetic tissue cultures, widely regarded as a standard reference for infection diagnosis, remains unclear. The study by Neut et al. clearly describes the limits of routine culture methods. Among 22 patients (including 11 total hip replacement patients) undergoing revision surgery for implant-related infection, bacterial growth was observed in less then half of the patients with periprosthetic tissue cultures, 64% with prolonged culturing, and 86% with extensive culturing of biomaterial surface scrapings. The authors also highlighted the importance of extensive, biomaterial-based cultures, as opposed to routine tissue cultures—the former method displaying multiple strains and/or subtypes, the latter indicating mostly only a single strain.³⁶

CULTURE-INDEPENDENT METHODS

The limited reliability of fluid/tissue cultures to rule out infection has stimulated research into culture-independent techniques for the determination of implant-associated pathogens. Tunney et al. tried to improve the pathogen detection rate by examining both the periprosthetic tissue samples and removed prostheses. In addition to routine cultures, they immediately transferred explanted prostheses to an anaerobic atmosphere and subjected the implants to mild ultrasonication to dislodge adherent bacteria. The authors cultured quantifiable numbers of bacteria after sonication in 26 of the 120 implants examined. Tissue cultures were positive in only 5 of 26 corresponding samples, and negative in all other samples. Thus, sonicate cultures were more often positive than standard cultures. It is unlikely that these results represented mere contamination: No bacteria were isolated from 20 sterilized implants that the authors used as control samples and processed in the same manner. Additionally, the authors performed quantitative tissue pathology. Inflammatory cells were present in all available (18/26) samples from sonicate-positive patients. Interestingly, inflammatory cells were also present in 87% of periprosthetic tissue culture-negative patients. The authors suggested that bacteria that were not isolated by the techniques of culture might have infected the implants. It was interesting that only 2 of the 18 patients with positive sonicate cultures were positive by standard cultures of joint fluid or periprosthetic tissue. They also suggested that the increased detection of bacteria from prostheses by ultrasonication followed by culture could improve postoperative antibiotic therapy and should reduce the need for further revision.⁶

In another study by the same authors that was based on the same patient population, the detection rates of bacterial infection of hip prostheses by culture versus culture-independent methods were compared. Interestingly, many culture-negative patients were positive by the other techniques. Four per cent of the cohort was tissue-sample culture positive; however, 22% were sonicate-sample culture positive. With the use of immunofluorescence microscopy with a monoclonal antibody specific for Propionibacterium acnes and a polyclonal antiserum specific for Staphylococcus spp., 63% of the samples were positive. The bacteria were present either as single cells or in aggregates of up to 300 bacterial cells. Even greater positivity rate (72%) was observed for broad-range PCR with universal primers for bacterial 16S rRNA gene.⁵

The above two seminal reports highlight the importance of sonication and other advanced processing and diagnostic techniques in determining the presence of bacterial byproducts around joint implants. Unfortunately, the studies of 120 unselected hip revisions did not report clinical follow-up, which could translate the positive microbiological results into either infections that would subsequently become symptomatic, or those that would not become clinically symptomatic and therefore are of unknown significance. The authors also did not determine the value of sonicate culture, immunofluorescence microscopy, and 16S rRNA PCR specifically in the population of patients diagnosed with AL. Additionally, the authors did not sequence the DNA—which might confirm the relationship between the results of samples with both positive culture and PCR, and which could give clues as to the origin of PCR positive results in samples with negative cultures. However, one should note that efficient, cost-effective methods of DNA sequencing were not readily available at the time of conducting many of those studies.

There is also a comparable body of evidence with other joint replacements, for example, in a study by Mariani et al.³⁷ 16S rRNA PCR suggested the presence of bacteria in 32 out of 50 symptomatic total knees, compared to only 15 diagnosed by culture. There are, however, two limitations of this study—first is the lack of sequencing of the DNA, which could confirm the congruence of the samples with both positive cultures and PCR, and could give clues as to the origin of PCR positive results in samples with negative cultures. Second, the authors found 16S rRNA positivity in the case of a culture-proven infection due to a Candida sp.; however, Candida are not known to harbor this particular molecular fingerprint, as noted by Trampuz et al.³⁸

In a study of 10 consecutive revisions (five clinically infected and five because of AL), Dempsey et al. found bacteria in all 10 specimens both by culture and PCR. Additionally, a diverse range of bacterial species was found in both infected and noninfected cases, with no significant differences between those two groups in terms of organisms cultured. Even more surprisingly, the most predominant species were members of the Lysobacter genus, and not Staphylococci or Streptococci.⁷

Fenollar et al. analyzed 525 bone and joint samples with 16S rRNA gene PCR amplification followed by sequencing of all amplified products. There were false negatives both in culture (13%) and PCR groups (7%). Unusual organisms were detected only in the patients with mixed infections. The authors hypothesized that the “unusual” organisms are difficult to isolate and they may have low pathogenic potential, although no conclusive evidence in support of this claim is presented. They recommended treating the results of PCR as an assay complementary to culture. While based on a large number of patients, this study provided no internal control for the validity of the results (e.g., if a single bacteria strain was detected by culture, no further analyses were performed) and surprisingly, only one culture per patient was analyzed with clinical outcome as the defining criteria for infection, which could leave the false impression of lower rates of infection, especially when slowly growing pathogens are present.³⁹

Other studies have demonstrated the presence of bacteria in cases when conventional culture was negative.²⁷ Also in a laboratory, controlled environment, conventional culture techniques yield less results than direct DNA extraction, especially in the detection of the strict anaerobes.⁴⁰

POLYMICROBIAL INFECTIONS AND ASEPTIC LOOSENING

Culture-independent molecular-based studies not only give more positive results but also show greater diversity of bacteria than culture-based methods.^9,41,42 In diagnosed infections, the molecular techniques generally yield comparable pathogens to culture; however, greater diversity of bacteria compared to the culture method is seen in patients with AL.^41,43 This suggests that AL may be associated with polymicrobial population shifts to cause subclinical infection.⁴⁴ With standard techniques we may only be culturing the planktonic bacteria, disrupted from the biofilm surface, while the true “culprits” of inflammatory reaction remain undetected, protected within biofilm structures. Molecular methods also allow the recovery of organisms that are difficult to culture. They might have low pathogenic potential, which is responsible for delayed loosening with few—if any—clinical symptoms of infection, and are therefore easily undiagnosed. In those situations, PCR might become complementary to culture, but its value remains to be determined.^39,40,45

CULTURE-INDEPENDENT METHODS—LIMITATIONS

In multiple studies cited previously, simple polymerase chain reaction (PCR) assays were used for identification of pathogens and for culture validation. With the use of 16S ribosomal RNA, which is highly conserved sequence among all bacteria, or with the species-specific primers, high rates of false-negative and -positive results were noted. Additionally, some studies failed to demonstrate any benefit of these more sophisticated techniques over traditional culture methods.²⁶ False negatives were attributed to inefficient DNA extraction from the sample (especially from Gram-positive bacteria), loss of DNA during sample processing, or the presence of PCR inhibitors in the sample.^36,38 Multiple developments have been proposed, including real-time monitoring of amplified DNA; DNA amplification, that is, multiple displacement amplification; following the PCR with cloning and sequencing; using genus-focused primers and the use of elaborate laboratory techniques, such as PCR-based ESI-TOF-MS (electron spray ionization time-of-flight mass spectrometry).^{15,39,43,46,47}

The significance of the results of culture-independent diagnostic methods is controversial and raises the question of appropriate validation and clinical significance. In one study, bacterial DNA was extracted even from implants taken from patients undergoing routine removal of osteosynthesis material without any evidence of infection. The identified DNA, however, could not have been directly linked to living bacteria because the methods used in that study were not able to discriminate between living and dead cells. Nevertheless, the assumption that the detected DNA originated from living bacteria was reasonable, since free DNA would have rapidly degraded on the implant during the 6 months of being incorporated in the body, as would have dead or dying bacteria.⁴⁰

With the use of PCR-based techniques, false-negative results can be observed due to the presence of polymerase inhibitors in the sample, low amounts of DNA due to the failure to extract from thick-walled gram-positive cocci or due to the losses incurred at the purification step.³⁷ Low amounts of DNA can also be encountered with small populations of bacteria, and amplification can result in carryover of the contaminating DNA and false-positive results.⁴⁸

The use of genus-focused PCR primers is useful in the case of known bacteria; however, when applied to the general population, its sensitivity of 71% and specificity of 49% as compared to traditional culture methods seem unacceptably low.⁴⁷ Also, the use of sophisticated culture systems (e.g., automated blood culture systems) can lead to an increased rate of false-positive results, without improved pathogen detection.⁴⁹ Lastly, histology, usually used in culture-negative cases to confirm the infectious origin of loosening, is frequently negative in infections caused by slow-growing bacteria, for example, Propionibacterium acnes.²³

THE ROLE OF ANTIBIOTICS

There is also evidence that antibiotics reduce the rate of AL and periprosthetic inflammation.⁵⁰ There are two possibilities explaining those phenomena. Firstly, the antibiotic prophylaxis simply kills the bacteria that could cause low-grade infection.⁵⁰ Second, some antibiotics, for example, cephalothin and erythromycin, exhibit a direct inhibitory effect on the matrix metalloproteinase, which leads to the reduction of tissue destruction around the loose total hip components.^51,52 However, laboratory studies failed to demonstrate an inhibitory effect of gentamicin on the metalloproteinases.⁵¹ Gentamicin is widely used as a local antibiotic, which was also shown in registry studies to diminish the frequency of septic and AL.⁵⁰ The protective properties of antibiotics against AL might simply reflect partially treated yet undiagnosed infections, subsequently recorded as AL. The finding that more frequent dosage of antibiotics with a short serum half-life in addition to gentamicin bone cement yields lower infection and AL rates than less frequent application might indirectly support this theory.⁵⁰ Still, this issue should best be solved by a large randomized trial to try to prove a causative relationship between antibiotic dosage and AL. One large randomized trial compared the effects of antibiotics when administered systematically versus addition to the bone cement. There was a trend for better outcomes over the 10-year observation period with local dosage; however, this effect did not reach statistical significance, despite a large sample size.⁵³ In a study by Havelin et al.,⁵⁴ the rate of revision for AL was lower in patients who had received antibiotic-containing bone cement, compared to patients who had received regular cement. In a registry study by Engesaeter et al., combined antibiotic delivery reduced the revision rates for both septic and aseptic implant loosening. The authors explained this phenomenon by the fact that antibiotic prophylaxis prevents unrecognized low-grade infections, misclassified as AL.⁵⁰

Byrne et al. noted a high incidence of intraoperative contamination despite standard prophylaxis. However, this was not reflected by a similar rate of postoperative infection. This may be due to a small bacterial inoculum in each case or may be due to the therapeutic effect of perioperative intravenous antibiotic prophylaxis.⁵⁵

Alternatively, in a study by Neut et al., a clear inhibition zone was observed around a gentamicin-loaded bone cement fragment, despite the fact that the antibiotic-loaded bone cement had remained in situ for more than 5 years in a patient ultimately revised for infection. Therefore, it would appear that the biofilm formed in vivo protects the bacteria against gentamicin.³⁶ Biofilm formation was also observed on the surface of antibiotic-loaded bone cement spacers used for treating infection during two-stage revision surgery⁵⁶; however, the presence of biofilm was not confirmed with scanning electron microscopy.⁵⁷ On the other hand, the inability to detect biofilm with scanning electron microscopy did not necessarily mean that biofilm was absent from the surface of the spacers. This is due to the fact that only a portion of the spacer was examined.

MICROBIAL-ASSOCIATED MOLECULAR PATTERNS (MAMPs) AS TRIGGERS FOR OR MODULATORS OF AL

Multiple studies suggest that MAMPs, either present on the implant⁵⁸ or in the periprosthetic tissue at the time of implantation or accumulated in the effective joint space after the implantation,⁵⁹ adhere to generated wear particles and cause immunostimulation,⁶⁰ which leads to secretion of proinflammatory and pro-osteolytic substances. Lipopolysaccharide (LPS) is the most studied MAMP; however, this family also includes peptidoglycan, lipoteichoic acid, and teichoic acid. MAMPs induce immune reactions comparable to LPS.⁶¹ Interestingly, MAMPs are found in human tissues, without signs of bacterial infection and are thought to be responsible for developing some chronic disorders.^62,63 Even small amounts of endotoxin present on biomaterials can elicit a substantial inflammatory reaction, over-reaching the biological effects of the biomaterials themselves.⁶⁴

Ragab et al. proved the presence of significant levels of adherent LPS on commonly used preparations of titanium particles as well as on titanium and titanium-alloy implant surfaces. They also developed a protocol that removed more than 99% of the adherent LPS from the titanium particles without detectably affecting their size or shape.⁵⁸ Based on their work, multiple basic studies were developed to investigate the role of MAMPs in AL.^59,65–68

The hypothesis that MAMPs contribute to AL around implants relies on the presence of MAMPs in the periprosthetic milieu either at the time of implantation, or accumulation of MAMPs afterwards. There are multiple possible sources of MAMPs in the periprosthetic tissues. The most studied are the biofilms on the prosthetic or bone surface, undetected by regular microbiological methods. Interestingly, in a study by Sierra et al., the percentage of positive sonication cultures was significantly higher in patients with severe osteolysis, compared to patients with lower degrees of lysis, as assessed by the Paprosky and Engh classifications. All of the patients were considered initially to be suffering from AL, based on the fact that all 52 patients were without the symptoms or signs of infection, had negative histology for infection, and at least five out of six periprosthetic tissue cultures were negative.⁶⁹

A second source of MAMPs at the prosthesis–bone interface are surface contaminants that remain on sterilized implants.⁵⁸ Contamination during implant manufacturing has been described,⁶⁵ as well as failure of autoclaving to remove the MAMPs.^70,71 The contaminants evoke TLR- or TNF-mediated immune responses, which do not occur once the contaminants are removed or are inactivated.⁷² In human studies, a high failure rate due to impaired osseointegration was attributed to LPS-contaminated acetabular cups—this fact could be interpreted as an argument for the MAMPs theory, although those results were never published in a peer-reviewed journal.⁷³ However, an interesting study would be to prove that excellent osseointegration of the same cups can occur once LPS is removed—that is, impaired osseointegration is not the result of mechanical causes or another inherent property of this cup model.

With the possibility of MAMPs contamination, another version of MAMPs theory becomes evident, that is, bacterial substances do not cause AL by attaching to the wear particles and causing bone resorption, but rather by impairing initial osseointegration. This latter point has been proven to predispose patients to AL.⁷⁴ Bonsignore et al. proved in a mouse model that LPS-coated debris on sterile implants inhibits bone-to-implant contact and biomechanical pullout strength. When analyzed in knock-out mice, it became apparent that LPS acted through its primary receptor, TLR-4, leading to impaired osteoblast differentiation, without affecting the early stages of osteogenesis (attachment, spreading, or growth).⁶⁵ Harder et al. measured ex vivo whole blood activity, and concluded that LPS-coated implants evoke an inflammatory response by either TLR or TNF-α activation.

LPS is also normally present in large quantities in the intestinal lumen. Functionally intact gut epithelial barrier prevents absorption of all but minute quantities of LPS. However, a functionally deficient mucosal barrier may allow increased endotoxin absorption.⁷⁵ Increased MAMPs permeability can even occur after ingesting a single high-fat meal.⁷⁶ MAMPs were also detected in bone marrow, synovial membrane, and fluid of native rheumatoid joints before joint replacement. Their role in primary disease pathogenesis, as well in AL after joint replacement, still remains to be clarified.⁷⁷ Lastly, elective total hip replacement causes almost 10-fold increase in serum endotoxin levels, with highly variable timing of peak values between the 1st and the 3rd day postoperatively.⁷⁸

MAMPs ligate TLRs and subsequently promote innate immune responses, potentially leading to granuloma formation, osteolysis, and AL.⁷⁹ The variation in host innate immune response to wear particles, attributed to patients’ genetics, could heavily influence the dynamics of AL, though they are still incompletely defined.⁸⁰ A concomitant mechanism of LPS modulation could be direct inhibition of late osteoblast differentiation.⁶⁵

In order to determine the presence of endotoxin in human serum or plasma, highly sensitive bioassays are necessary—LPS has a very short serum half-life and are poorly immunogenic despite strong potency in eliciting immune reaction.⁸¹ The assay to detect LPS in tissues has been developed only recently by Nalepka and Greenfield.⁸² They detected LPS in periprosthetic tissue from patients with AL after total joint replacement for inflammatory arthritis. However, little to no LPS has been detected in a control group with the underlying diagnosis of osteoarthritis. The authors, active proponents of MAMPs hypothesis, could not explain this difference, although they present several possible theories on the subject.⁵⁹

WEAK LINKS

In one study involving 21 hip and knee femoral components, sonicate cultures from aseptically failed implants did not yield more organisms than from negative controls (sterilized prostheses). However, the authors themselves had noted the limitations inherent to their study—first, pre-/perioperative antibiotics may have been delivered, and the relatively high detection limit (50 colony forming units) set by their study protocol.⁸³

Bjerkan et al. reported that 16S rRNA gene real-time PCR did not identify more cases of septic prosthetic loosening than did culture of periprosthetic biopsies obtained in a highly standardized manner.⁸⁴ This is very important in light of the fact that most of the studies do not give specific details of tissue harvesting, which is crucial information.

The weak links in the evidence underlying the MAMPs theory should be noted as well. First, if the bacteria were causing MAMPs production, then it would be logical that antibiotics should reduce the rate of AL. However, if MAMPs contamination, and not the presence of bacteria, was responsible for development of the AL, what do the antibiotics reduce the AL rate? Second, in an animal model, locally applied endotoxin stimulated bone resorption similar to that in experiments with wear particles. Endotoxin on the surface of implants and particles appeared to be inactivated in situ. A clean implant surface did not absorb endotoxin. Those results suggested that endotoxin adhering to orthopedic implants was not a major cause for concern⁸⁵; however, other studies, one using a metallic implant in situ⁸⁶ and another, in a murine calvarial model without a metallic implant,⁶⁸ found conflicting results. Lastly, in all the human studies, the presence of MAMPs might be purely coincidental, without a causative relationship for AL.

OUTLOOK

The particle disease theory of AL serves as a framework for our understanding of complex processes involved in the biological response to THA, or any implant within the body, for that matter. However, the insufficiencies and simplifications necessary to present broad research concepts for the sake of practical application may have resulted in the omission of truly critical factors influencing the course of events involved in AL.

Some of the registry studies support the theory that the addition of antibiotics to the bone cement improves the loosening-free survival of hip implants.⁵⁰ This effect is more pronounced with longer follow-up periods.^50,87 Additional studies with long-term data are necessary to verify those claims, which support the theory that many of the cases that we consider as AL may be in fact low-grade infections.

Other causative mechanisms of AL of hip implants remain under investigation namely implant micromotion, eccentric mechanical loads, fatigue failure at bone-implant or bone–cement interface, and increased synovial fluid pressures. There is still a need for a well-designed clinical study to answer the question whether bacterial biofilms could induce loosening. Such a sufficiently powered study would include direct identification of bacteria on the surface of implants, compared to multiple culturing techniques, PCR-based techniques with DNA sequencing, and ultrasonication. Ideally long-term follow-up of perhaps 10 years or more would enable detailed analysis and correlation of the laboratory, clinical, and radiological findings.

CONCLUSION

The outcomes of revision operations for AL are often guarded, even with modern surgical techniques. This has led some authors to postulate that there is more than the simple response to wear particles underlying the presumptive diagnosis of AL and that the original more simplistic theory may incompletely guide our clinical practice.^88–90 Direct detection of bacteria on the implant surface in culture-negative patients has caused many surgeons to abandon the term “aseptic loosening” in the favor of the term “undiagnosed septic loosening.” In the words of Nelson et al., “The absence of evidence clearly was not evidence of absence.”⁸