Introduction

Candida albicans is the most common pathogen involved in bloodstream infections [1], with current investigations raising concerns on increasing numbers of multidrug-resistant strains as a potentially new global health threat [2,3,4]. Despite its importance, limited remains known on spondylodiscitis caused by Candida spp. [5]. Preliminary findings indicate that Candida spondylodiscitis affects high-risk cohorts, including those with prior IV drug abuse, obesity, and diabetes, oftentimes resulting in fatal outcomes [6, 7].

Importantly, diagnostical and therapeutical approaches among this uniquely challenging cohort remain controversial, given limited patient numbers in existing investigations [5]. In fact, to the best of our knowledge, all present investigations are case reports or smaller case series. Moreover, these limited reports on Candida spondylodiscitis have short-term follow-up only [8, 9], while limiting their analysis to certain spinal levels [10], the presence of instrumentation [11, 12], immunosuppressive patients [13], or certain Candida strains [14,15,16]. This is critical, as it reduces the number of cases in an already limited cohort of patients.

Given its increasing importance, controversial approaches to diagnosis and treatment, and limited patient numbers in existing reports, this review is the first to systematically summarize all cases of Candida spondylodiscitis to date. In addition to analyzing baseline demographics, and identifying potential patients at risk, we aimed to summarize current diagnostical and therapeutic management, together with the outcome and prognostic factors.

Methods

This systematic review was based on PRISMA guidelines [17, 18] and registered in the PROSPERO [19] International prospective register of systematic reviews (CRD42022365441). Data collection was performed from October 17, 2022, through November 15, 2022. A re-run prior to the final analysis was performed to include any studies that may have been published during the data collection phase.

We analyzed a total of five databases (PubMed, Web of Science, Embase, Scopus, and OVID Medline) using the search syntax “albicans” [Title/Abstract] AND (“spondylodiscitis” [Title/Abstract]) with variations depending on the unique syntax of each database (Supplementary Table 1). If available, we also included articles from the “Similar articles”-tools from PubMed to increase the scope of articles screened [20]. Articles that were not available as a full-text English version were excluded unless an English abstract was available yielding sufficient data for analysis.

We included articles reporting microbiology-confirmed cases of Candida spondylodiscitis of any segment of the spine. No restrictions were made based on the year of publication, status of immunocompetency, or length of follow-up. PICO criteria [21] as a mean of evidence-based analysis were followed throughout: we included articles reporting patients with confirmed Candida spondylodiscitis who underwent diagnostic and therapeutic management. No comparator or control group was given and the primary outcome of interest was overall survival (Supplementary Table 2). Articles were excluded if they (1) were non-original articles, (2) were solely reporting osteomyelitis without involvement of the disc, (3) were not involved human subjects, and (4) reported spondylodiscitis caused by other infections agent(s). After removal of duplicates, titles and abstracts were screened for inclusion eligibility, and if deemed suitable analyzed as a full text.

We collected baseline demographics and information on past history, including age, sex, immunocompetency, presence of IV drug abuse, duration and types of clinical symptoms, previous spinal surgeries, instrumentation status, and affected spinal segment. Comorbidities were analyzed based on the age-adjusted Charlson Comorbidity Index (CCI) [22, 23]. In terms of diagnostical workup, we collected data items on CT-guided biopsy prior to initiating antifungal treatment, leukocyte count, erythrocyte sedimentation rate (ESR), and C-reactive protein (CRP), as well as microbiology results. Lastly, we gained data on the treatment regimen including antifungal agents and operative procedures.

Outcome parameters included length of hospital stay (LOS), perioperative complications, number of revisions, final recovery status, and mortality. Partial recovery was defined as continued clinical symptoms or permanent neurologic damage, whereas full recovery was defined as resolution of all initial symptoms at the most recent follow-up.

All results were presented as (1) overall cohort, (2) separated by Candida albicans versus non-albicans, and (3) based on survival at the last follow-up. Categorical variables were presented as absolute and relative values, and continuous variables as median and interquartile ranges. For comparison between groups, we used Fisher’s exact test for binary data to compare frequencies of side effects in two groups and Wilcoxon’s rank sum test to compare continuous data. A p-value of < 0.05 was considered significant. Kaplan-Meier survivorship analyses were performed, and differences in survival calculated based on log-rank tests.

Results

After removing 382 duplicates from the initial 1038 studies identified in the search process, 656 articles were analyzed for title and abstract. Following further detailed exclusion criteria, we analyzed 92 articles as full texts, of which 72 were finally included (Fig. 1).

Fig. 1
figure 1

PRISMA flow chart

Full size image

The final cohort consisted of 89 patients with a median age of 61 (Table 1, Table 2). Median age-adjusted CCI was 3, 15% had a history of IV drug use, and 23% of patients were immunocompromised. Nearly all patients presented with back or neck pain (99%), whereas fever and weakness were present in 19% only. Symptoms lasted for an average of nine weeks until the final diagnosis was made. Median leukocyte count was 8×109/mL, CRP was 3 mg/dL, and ESR was 65 mm/h. CT-guided biopsy was performed in 53%, with Candida albicans (60%) identified as the most common pathogen. Empiric antibiotic treatment prior to definite diagnosis was administered in 41% of cases. Antifungal monotherapy was given in 58%. The most commonly used antifungal agents included fluconazole (68%), amphotericin B (38%), and echinocandins (26%). The median length of antifungal treatment was six months. Surgical intervention was performed in 68%, including 34% undergoing instrumented discectomy. At a median follow-up of 12 months, 3% developed sepsis, 6% underwent revision, and 12% died of disease. The two year survivorship free of death was 80% (95% CI, 62 to 98%), and the two year survivorship free of revision was 94% (95% CI, 84 to 100%).

Table 1 Overview of 89 patients with Candida spondylodiscitis identified in 72 studies
Full size table
Table 2 Baseline demographics and treatment outcome of 89 patients with Candida spondylodiscitis
Full size table

Among 85 patients with information on the Candida strain, 53 were for Candida albicans (62%) and 32 for Candida non-albicans (38%; Table 3). Both groups did not show statistically significant differences in baseline demographics, prior history, antifungal therapy, and surgical intervention. In contrast, patients with Candida albicans had a significantly higher leukocyte count (p = 0.039) and trended towards significantly increased ESR (p = 0.054) compared to those affected by non-albicans Candida spondylodiscitis. No statistically significant difference was noted in median follow-up, LOS, sepsis, and revision rate, as well as with respect to final recovery status and death. Likewise, the two year survivorship free of death was comparable between albicans (91%; 95% CI, 50 to 100%) non-albicans Candida (82%; 95% CI, 64 to 100%) spondylodiscitis (p = 0.68).

Table 3 Demographics and outcome stratified by Candida albicans versus non-albicans Candida strain
Full size table

In total, 76 patients had a known survival status at the last follow-up, with 67 surviving and 9 dying by disease (Table 4). Younger age (p = 0.042) and longer length of antifungal therapy (p = 0.061) were predictive of survival, whereas outcome did not differ based on Candida strain (p = 0.74) and affected spinal level (p = 0.44).

Table 4 Patient demographics stratified by survival following Candida spondylodiscitis
Full size table

Discussion

Limited remains known on Candida spondylodiscitis outside of case reports and smaller case series. As such, this systematic review analyzed 89 patients treated for spondylodiscitis at a median follow-up of 12 months. Our results demonstrated one in five patients affected by Candida spondylodiscitis to die within two years. Importantly, Candida albicans and non-albicans were similar in their prognosis, whereas younger age and prolonged antifungal treatment were associated with increased survival.

Baseline clinical and demographic factors

Knowledge of baseline demographics among patients affected by Candida spondylodiscitis is important, as it may allow to identify potential patients at risk. First reports date back to the 1970s [87], and Candida spondylodiscitis has since then been associated with immunocompromised patients, and those with other significant comorbidities including diabetes, obesity, or IV drug abuse [88,89,90]. In this systematic review, we identified most affected patients to be males in their sixties, with one in four demonstrating signs of immunocompromise, and another 15% reporting previous IV drug use. While these findings partially reflect the aforementioned historical literature, one must acknowledge that the majority of patients had no clearly attributable risk factors. Prospective multicentre studies will be necessary to draw final conclusions, based on comparison with spondylodiscitis caused by other pathogens.

Diagnostic challenges of Candida spondylodiscitis

The diagnosis of Candida spondylodiscitis remains challenging, especially as symptom onset is subacute, and oftentimes gradually progressive over weeks to months [91,92,93]. In fact, 99% of patients had back or neck pain during initial clinical assessment, while less than 20% had systematic signs of infection (fever). Correspondingly, and similar to previous findings, CRP was only mildly elevated [94,95,96]. Although initial identification of Candida spondylodiscitis may therefore remain difficult, a number of factors distinguish it from a bacterial entity, once the diagnosis of spondylodiscitis is suspected. Foremost, the lumbar spine is more commonly affected (50%) [97], symptom onset is less acute, and Candida spondylodiscitis rarely causes chills [98, 99]. This differentiation is important, as it may allow for an earlier antifungal treatment initiation.

Antifungal treatment and surgical considerations as the mainstay of therapy

Pathogen eradication is the primary goal in treatment of Candida spondylodiscitis, with its success depending on rates of antifungal disc penetration and antimicrobial resistances [76]. The Infectious Diseases Society of America (IDSA) recommends fluconazole (400 mg; 6 mg/kg daily) for six to 12 months in case of Candida osteomyelitis but does not specify for spondylodiscitis [100]. Although our findings partially reflect current IDSA recommendations on Candida osteomyelitis, treatment still varied significantly among the 89 patients. We believe this heterogeneity to be attributable to possible treatment resistances requiring changes in antifungal medications [101], as well as contradictory recommendations on antifungal length, ranging from a few weeks to over a year [12, 24, 53]. Finally, different approaches on empiric treatment may have been a contributing factor [29, 54, 61]. Despite these discrepancies, antimicrobial therapy remains the most important part of the treatment of spondylodiscitis [5] with surgery only being recommended in cases of spinal instability, a mass effect due to abscess, or neurologic deterioration [10, 102, 103]. We believe the prolongated symptom onset prior to definite antifungal treatment, combined with possible extensive local disease progression, and possibly unclear preoperative diagnosis to explain the high rates of surgery among our patients (68%). Despite an increase in antifungal resistance, large-scale surveys of pathogenic yeasts isolated from blood cultures suggest that clinical consequences may not be expected from this trend [104]. Further, Candida spp. may be subject to a confirmation bias owing to advances in testing susceptibility [105]. Surgical intervention was most commonly pursued if spinal instability was suspected or management solely based on antifungal agents deemed insufficient.

Prognostic factors in Candida spondylodiscitis

The prognosis in Candida spondylodiscitis remained poor, with only 77% achieving full recovery, and 12% dying at a median follow-up of one year. The two year calculated survivorship free of death was even as low as 80%, predicting one in five patients to die within two years of diagnosis. Importantly, Candida strain was not predictive of death, as previously described in periprosthetic joint infections [16, 40]. In contrast, younger age at diagnosis and longer antifungal treatment were predictive of survival, with the latter being a promising alternative for future research. Of note, only 6% of patients were in need of revision surgery. This is important, as spinal revision surgery is known to significantly compromise functional recovery [77, 86, 106].

Limitations

This systematic review had limitations that in return were attributable to the weaknesses of its included studies. Foremost, most articles included five cases or less, limiting the overall number and generalizability. In addition, not all information could be included for every patient. In specific, details on secondary diseases were not provided in the majority of articles, and follow-up defined inconsistently. Additionally, some data such as clinical examination results and laboratory results were inconsistently reported among cases. This may have compromised the generalizability of our results. We have nevertheless included these data points to enhance to informative value of our tables, even when data could not be included in our statistical analysis. Finally, we were not able to provide regression models to precise potential risk and outcome factors, as the high proportion of missing information would have corrupted the analysis.

In conclusion, patients affected by Candida spondylodiscitis tend to be males in their sixties, present with local rather than systematic symptoms, and have a poor short-term prognosis with one in five dying within two years of diagnosis. This first systematic review on Candida spondylodiscitis might help physicians in identifying patients at risk and when providing a prognosis. Consensus meetings will be necessary to determine an optimal future treatment approach.