Undiagnosed invasive candidiasis: incorporating non-culture diagnostics into rational prophylactic and preemptive antifungal strategies

Abstract

The insensitivity of blood cultures for diagnosing invasive candidiasis fuels prophylactic and preemptive antifungal treatment. Assays like serum β-D-glucan or mannan/anti-mannan detection can identify blood culture-negative invasive candidiasis, but their roles in guiding antifungal therapy are undefined. We propose that non-culture tests can be incorporated into rational management strategies, based on clinical setting. As an example, β-D-glucan sensitivity/specificity for blood culture-negative, deep-seated candidiasis is approximately 60/75%. In intensive care units with <1 or 3% invasive candidiasis rates, positive/negative predictive values are <2/>99% and 6/98%, respectively. With pre-test likelihoods of 10 and 33%, positive/negative predictive values are 20/94% and 54/79%, respectively. Based on these data, negative and positive β-D-glucan results likely will be most useful for discontinuing prophylaxis among low-risk to moderate-risk patients (pre-test likelihoods ∼3–10%), and triggering preemptive therapy among moderate-risk to high-risk patients (pre-test likelihoods ∼10–25%), respectively. In extremely high-risk patients, universal prophylaxis is likely to be the best strategy.

Keywords::

• anti-mannan antibody
• Candida
• candidemia
• candidiasis
• deep-seated candidiasis
• Fungitell
• intra-abdominal candidiasis
• mannan
• β-D-glucan

Clinicians have access to a broad armamentarium of antifungal drugs with good activity against Candida species. Nevertheless, mortality rates among patients with candidemia and other types of invasive candidiasis are 25–40% [1,2]. In part, poor outcomes stem from difficulties in diagnosing invasive candidiasis. The gold standard diagnostic tests (microbiologic cultures of blood or other samples collected under sterile conditions) have sensitivity of only 50% and turn positive late in the disease [3]. The combination of inadequate diagnostics and poor outcomes of invasive candidiasis has fuelled widespread use of prophylactic and preemptive antifungal therapy [4]. The benefits of these practices are uncertain. Moreover, they are often implemented in an ad hoc manner, with implications for institutional ecology, antifungal resistance and pharmacy budgets.

Non-culture diagnostics such as β-D-glucan and mannan/anti-mannan antibody assays are directed against constituents of Candida cells rather than viable organisms [3]. Used judiciously as adjuncts to cultures, these tests may identify more patients with invasive candidiasis, at earlier stages of disease. However, there is uncertainty about the performance of non-culture diagnostics, particularly in at-risk patient populations such as intensive care unit (ICU) residents [5]. As such, their precise roles in the management of invasive candidiasis are undefined. We propose that non-culture tests can be incorporated into rational antifungal strategies, provided consideration is given to the type of candidiasis being diagnosed, strengths and limitations of each test and clinical setting.

Invasive candidiasis encompasses candidemia, candidemia complicated by deep-seated candidiasis (i.e., infection of submucosal tissues or organs) and deep-seated candidiasis in the absence of candidemia (which can result from direct inoculation or prior hematogenous seeding) [3]. In considering the challenges of diagnosing invasive candidiasis, it is important to understand that cultures are quite sensitive at recovering viable Candida. The limit of detection in blood, for example, is ≤1 colony forming unit/ml, which is superior to PCR [6,7]. Bloodstream burdens of Candida glabrata are typically lower than other species, which may delay time to culture positivity [3]. Nevertheless, blood cultures should be positive during most episodes of active candidemia. Blood cultures will be negative in cases of intermittent candidemia, deep-seated candidiasis that persists after bloodstream sterilization and deep-seated candidiasis resulting from direct inoculation of Candida. Fungal-selective media may improve the sensitivity of blood cultures and shorten time to positivity [8], but they are unlikely to significantly improve identification of the preceding entities. Tissue cultures are limited by the need for invasive procedures and uneven distributions of viable organisms.

β-D-Glucan and mannan are cell wall constituents of Candida spp. and several other fungi. A serum β-D-glucan assay (Fungitell; Associates of Cape Cod, East Falmouth, MA, USA) is approved by the US FDA as an adjunct to cultures for diagnosing invasive fungal infections. In meta-analyses, the overall sensitivity and specificity of β-D-glucan testing for invasive candidiasis was approximately 75 and 80%, respectively [9,10]. In general, specificity and positive predictive values (PPV) are lower in ICU patients. Causes of false-positivity include systemic bacterial infections, fungal colonization, disruptions of gastrointestinal mucosa, receipt of broad-spectrum antibiotics like pipercillin-tazobactam and hemodialysis [5]. Test performance is improved if two consecutive results are positive [3]. Mannan and anti-mannan IgG assays (Platelia, BioRad, Marnes-la-Coquette, France) are employed widely in Europe. In a meta-analysis, sensitivity/specificity of mannan and anti-mannan were 58/93% and 59/83%, respectively. Values for a combined mannan/anti-mannan assay were 83/86% [11].

The overwhelming majority of studies have investigated non-culture diagnostics during candidemia. Non-culture tests are unlikely to be more sensitive than blood cultures for diagnosing candidemia that is not complicated by deep-seated candidiasis, but may shorten turn-around times. To improve significantly upon the sensitivity of cultures, non-culture tests must diagnose blood culture-negative, deep-seated candidiasis. Assays may accomplish this goal by detecting Candida cellular constituents that are released from deep tissue sites or that persist in blood after viable cells are eliminated [12]. Two recent studies assessed the Fungitell β-D-glucan assay among prospectively enrolled patients with deep-seated candidiasis. In the first, sensitivities of contemporaneously collected blood cultures and β-D-glucan serum samples were 21 and 67%, respectively [13]. Among all patients with invasive candidiasis, sensitivity/specificity of β-D-glucan was 56/73%. In the second study, sensitivity/specificity of two consecutive positive β-D-glucan results for intra-abdominal candidiasis was 65/78% [14]. The sensitivity of blood culture was only 7%. The interpretation of specificity in these studies was complicated by the fact that negative controls were also at risk for invasive candidiasis. Therefore, it is impossible to know if positive results for controls were false-positives or true-positives that were missed by cultures.

It is clear that clinicians must accept a level of uncertainty in using non-culture tests to diagnose invasive candidiasis. Indeed, the assays are best viewed as Bayesian biomarkers that assign a probability of invasive candidiasis, which is shaped by the pre-test likelihood of the disease and sensitivity and specificity of the test. The heterogeneity of patients at-risk for invasive candidiasis further complicates interpretation of results. Among patients in ICUs with invasive candidiasis prevalence rates of less than 1 and 3% [15,16], for example, PPV/negative predictive values (PPV/NPV) of a non-culture test would be <2/>99% and 6/98%, respectively (assuming sensitivity/specificity of 60/75%, as described for β-D-glucan in the deep-seated candidiasis studies). An increase in pre-test likelihood to 10%, as with the use of clinical prediction scores that consider risk factors for invasive candidiasis among ICU patients [15], improves PPV to 20%, but decreases NPV to 94%. In high-risk settings with pre-test likelihood of 33%, such as among ICU patients with necrotizing pancreatitis or undergoing repeat gastrointestinal surgery [14], PPV and NPV would be 54 and 79%, respectively. Therefore, at low pre-test likelihoods of invasive candidiasis, PPVs of non-culture tests are poor while NPVs are excellent. As pre-test likelihood increases, PPVs improve at the expense of NPVs.

In prophylactic antifungal strategies, treatment is started in all patients in a given clinical setting. In settings of low pre-test likelihood (such as a general ICU population), negative non-culture results make invasive candidiasis extremely unlikely, and can be used to justify discontinuing unnecessary therapy. As the pre-test likelihood of invasive candidiasis increases in particular subgroups, a negative test may still leave a high enough probability of invasive candidiasis that clinicians will not be comfortable stopping an antifungal agent. In the preceding paragraph, NPVs in moderate-risk (10% pre-test likelihood) and extremely high-risk (33% pre-test likelihood) ICU populations were estimated to be 94 and 79%, respectively. For a discontinuation strategy to work in the first setting, clinicians must be willing to stop treatment based on negative result and accept a 6% chance the patient has invasive candidiasis. In the second setting, NPV is almost certainly too low for a discontinuation strategy to be viable. In practical terms, therefore, antifungal discontinuation is likely to be most acceptable when pre-test likelihoods of invasive candidiasis range from approximately 3 to 10% (i.e., low-risk to moderate-risk ICU patients). At pre-test likelihoods less than approximately 3%, the probability of invasive candidiasis is sufficiently low that negative results have little added value.

In preemptive strategies, a positive non-culture test result is used to initiate antifungal treatment. Studies have shown that antifungal prophylaxis is beneficial in preventing deep-seated candidiasis in ICUs with baseline rates of invasive fungal infections ranging from approximately 15 to 35% [17–21]. These data suggest that the threshold PPV to trigger preemptive therapy may fall within this range. To achieve these PPVs, non-culture testing is best targeted to settings in which pre-test likelihood of invasive candidiasis is greater than or equal to approximately 10% (i.e., ICU patients at moderate-risk or greater). However, at a certain pre-test likelihood, the probability of invasive candidiasis is sufficiently high and NPV becomes sufficiently low that universal antifungal therapy may be justified. The pre-test likelihood that should trigger universal prophylaxis is not established, but the practice is generally recommended for extremely high-risk populations like ICU patients with necrotizing pancreatitis or complicated gastrointestinal surgery [22]. Therefore, preemptive strategies built upon non-culture tests are likely to be most useful for moderate-risk to high-risk ICU patients, but not highest-risk patients.

Prophylactic and preemptive antifungal regimens against invasive candidiasis are not validated in multicenter studies, and optimal approaches for incorporating non-culture diagnostics into management strategies are undefined [4]. In summary, we predict that non-culture testing will be shown to be most helpful in guiding therapy as follows: in low-risk to moderate-risk settings (pre-test likelihood ∼3–10%), negative results can be used to discontinue antifungal prophylaxis; in moderate-risk to high-risk settings (pre-test likelihood ∼10–25 or 35%), positive results can be used to trigger preemptive therapy. In very low-risk settings (pre-test likelihood ∼<3%), there is likely to be no role for routine non-culture testing or prophylactic/preemptive therapy. In highest-risk settings (pre-test likelihood ∼ >25 or 35%), the optimal strategy is likely to be universal prophylaxis without non-culture testing.

Our proposals are not presented as treatment guidelines, but rather as starting points for discussion. We have focused on tests that are currently employed by clinicians. However, the principles we present are applicable to other assays as they enter the clinic, or as further data become available on test performance in various populations.

Financial & competing interests disclosure

CJ Clancy has received research support for investigator-initiated research projects from Pfizer, Merck and CSL Behring. MH Nguyen has received research support for investigator-initiated research projects from Pfizer, Merck, Astellas and Vira-Cor IBT. The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

No writing assistance was utilized in the production of this manuscript.