Total hip arthroplasty (THA) and total knee arthroplasty (TKA) can significantly improve a patient’s quality of life and are cost-effective orthopedic surgeries [1–3]. Prosthetic joint infection (PJI) is a major complication following THA and TKA, with an incidence of 0.57% to 1.7% [4, 5]. Infections that occur within 90 days of arthroplasty are classified as early onset PJI, whereas infections that occur several months to several years after arthroplasty are considered delayed or late onset PJI [6]. Risk factors associated with PJI include comorbidities such as obesity, diabetes mellitus, neoplasm, immunosuppression, postoperative complications, revision arthroplasty and a history of PJI [6]. Gram-positive cocci, especially staphylococci (60%), gram-negative bacteria (6%), anaerobes (4%), fungi (1%) and polymicrobial (20%) account for PJI pathogens, and 7% of cases are culture-negative [6, 7]. Mycobacteria, such as rapidly growing mycobacteria (RGM) and Mycobacterium tuberculosis (MTB), are rare causes of PJI [8]. When PJI is suspected, synovial fluid or periprosthetic tissue samples are mainly obtained for bacterial culture before giving empirical antibiotics directed towards common pathogens. Routine bacterial cultures are not ideal techniques for mycobacterial isolation, and can potentially lead to false negatives and delayed treatment. It is important to differentiate between mycobacterial and non-mycobacterial joint infections early, as the treatment of mycobacterial PJI is significantly different from that of non-mycobacterial infections. Comparisons between the clinical and treatment outcomes of mycobacterial and non-mycobacterial PJI have never been studied. This study aimed to compare the clinical manifestations, infection etiologies, treatments and long-term outcomes of mycobacterial PJI with those of non-mycobacterial PJI, and to identify the factors predictive of the outcome of mycobacterial PJI at 6 and 12 months after diagnosis.

Of 178 subjects, 16 (9%) had PJI caused by mycobacterial infection. Mycobacterial PJI in this study was classified into 2 categories: 1) PJI caused by RGM namely M. fortuitum, M. abscessus and M. peregrinum; 2) PJI caused by MTB. The pathogenesis of the two types of mycobacterial PJI is completely different. RGM PJI are likely associated with intra-operative contamination [10] or disseminated infection in immunocompromised patients [11], whereas a local reactivation of dormant foci or hematogenous spreading is a feasible etiology for MTB PJI. RGM were the leading etiologic agents of mycobacterial PJI. Thus, this study focuses on the clinical characteristics and factor associated with RGM PJI. Of the 16 subjects diagnosed with mycobacterial PJI, 7 subjects had at least 1 of the following conditions: systemic lupus erythematosus (SLE), chronic corticosteroid therapy, chronic liver disease, diabetes mellitus, leukemia or chronic kidney disease. These conditions may be predisposing factors for mycobacterial infection [12–14]. However, the present study found no comorbidities that were significantly associated with mycobacterial PJI similar to several previously published studies [10, 15–17].

Early-onset PJI was significantly associated with RGM PJI in this study. Most of RGM PJI cases had an early-onset infection, which was markedly faster than the onsets reported in a previous study. Eid et al. [8] found the median onset of RGM PJI to be 312 weeks after prosthesis implantation. The shorter onset of RGM PJI in our study may relate to intra-operative contamination. However, this retrospective study is not able to find the source of RGM in our patients. A previous study discovered a cluster of genetically identical early-onset M. fortuitum PJIs, but a common source of these infections was not identified [10]. An association between an early-onset RGM PJI and intra-operative contamination therefore remains controversial. Five subjects with MTB PJI showed delayed to late onset of infection similar to a previous systematic review [14]. Distant or local foci of active tuberculosis were not recognized in the MTB subjects.

Periprosthetic tissue and synovial fluid cultures as well as histopathology are warranted in the diagnosis of mycobacterial PJI. All cases of mycobacterial PJI in this study were diagnosed using periprosthetic tissue cultures, including 3 subjects with mycobacterial culture-positive synovial fluid. Previous case reports have noted that periprosthetic tissue cultures have a high sensitivity in the diagnosis of mycobacterial PJI [16–20]. In the isolation of mycobacteria from periprosthetic tissue, several critical points should be considered. RGM were able to grow on routine bacterial culture media such as blood agar and MacConkey agar [10, 15, 17, 18]. However, RGM culture growth may be delayed at least 5–7 days, a time frame in which routine specimens may have already been discarded. Moreover, RGM manifests as weakly gram-positive bacilli on culture and has occasionally been found to resemble corynebacteria [21]. In contrast, MTB is hard to isolate on routine bacterial cultures. Special solid or liquid media for mycobacteria and a prolonged incubation time are required to isolate mycobacteria [21]. As such, when mycobacterial PJI is suspected, such as those PJIs with negative cultures, periprosthetic tissue samples should be submitted for mycobacterial culture in addition to another attempted bacterial isolation.

The mycobacterial PJI group received inappropriate initial antibiotics significantly more often than patients with non-mycobacterial PJI. A low index of suspicion and negative routine cultures could be the reason patients in the mycobacterial PJI group received inappropriate initial antibiotics. Antibiotic susceptibility testing (AST), combination antimicrobial therapy and a prolonged course of treatment are indicated in the management of mycobacterial PJI. RGM are usually resistant to traditional first-line anti-tuberculosis drugs. Yang et al. [22] showed that almost all M. fortuitum group, M. chelonae and M. abscessus were susceptible to amikacin; > 60%–80% of isolates were susceptible to clarithromycin, fluoroquinolones, and imipenem. The efficacies of cefoxitin, doxycycline and tobramycin against the 3 RGM were unsatisfactory [22]. However, there is no standardized treatment protocol or study correlating in vitro pathogen susceptibility and the clinical response of specific antibiotics directed towards RGM [11]. The majority of our patients with RGM PJI exhibited a satisfactory response with empirical treatment using amikacin in combination with cefoxitin, followed by combined fluoroquinolone and clarithromycin regimens. In this study, the treatment regimens in 5 subjects with MTB PJI were individualized because of drug intolerance, drug hypersensitivity and resistant organisms complicated treatment. This study revealed that subjects with mycobacterial PJI received a long duration of maintenance antimicrobial therapy, which is necessary to prevent recurrent infection.

Debridement and implant removal are required in the treatment of PJI. Removal of the infected prosthesis was a factor associated with a significantly increased favorable outcome rate 6 and 12 months after infection diagnosis in this study. Ninety-one subjects (52.4%) underwent two-stage surgery, which involves debridement and removal of all infected prosthetic components, insertion of antibiotic cement spacer, followed by timely re-implantation of a new prosthesis [9]. Two-stage surgery was associated with successful treatment of both bacterial [23] and fungal [24] PJI. The role of antibiotic cement spacer in this setting is to deliver high concentration of antibiotics directly to the infected site and to maintain limb length by preventing soft tissue contracture [25]. However, the optimal surgical strategy for treating mycobacterial PJI is still unknown. A case series evaluating M. fortuitum PJI found that the removal of the infected prosthesis combined with a prolonged course of antibiotics generally led to good outcomes [26]. A previous study evaluating MTB PJI of the knee concluded that late onset of infection (> 6–8 weeks), presence of a draining sinus, concomitant bacterial infection, osteolysis or prosthesis instability indicated the need to remove the infected prosthesis [27]. Obesity may diminish treatment success, as BMI was a factor significantly associated with an unfavorable outcome in this study.

The present study had some limitations. The retrospective design prevented us from assessing several variables, such as HIV status, a history of previous TB, previous antibiotic treatment from outside hospitals and overall therapeutic compliance. The prevalence of mycobacterial PJI and culture-proven non-mycobacterial PJI might be underestimated owing to the culture-negative PJI group. A number of subjects with culture-negative PJI received empirical antimicrobial therapy prior to obtaining their synovial tissues for culture, which hampered the sensitivity of pathogen identification. Patients with PJI, the definite favorable outcome can be judged after a follow-up of at least 2 years. The favorable outcome was followed up only within 12 months in this study. The 6-month and 12-month outcomes were not completely evaluated in 6 of 16 mycobacterial PJIs owing to a transfer out of our hospital or loss to follow-up. Future long-term prospective cohort studies should be conducted to more accurately identify factors associated with successful and unsuccessful treatment outcomes.

Mycobacteria, particularly RGM are important pathogens causing PJI. The clinical characteristics of mycobacterial PJI are indistinguishable from those of culture-proven non-mycobacterial PJI. RGM usually cause an early onset infection. Additional cultures for mycobacteria are an optimal front-line diagnostic protocol for those with suspected PJI with an early onset of infection, a negative routine culture or refractory to antibiotics. Removal of the infected implant significantly improves the chance of a favorable outcome.