Candidemia in critically ill immunocompromised patients: report of a retrospective multicenter cohort study

Candidemia represents 10% of nosocomial infections in hospitalized patients and is associated with mortality described to be as high as 40% [1‐4]. Underlying immune defect, solid or hematological malignancy, may predispose to candidemia which develops during the clinical course of these conditions in 1.8% of cases [5].

Despite being widely studied, several areas of uncertainty remain. First, diagnosis of candidemia is often delayed as consequences of time required to obtain blood cultures positivity. Although several studies suggested benefit of early initiation of antifungal therapy on patients survival [6‐8], evidence supporting benefits of preemptive treatment in high-risk critically ill patients is lacking [9, 10]. Antifungal resistance among documented Candida species is growing, and resistance to fluconazole and echinocandins has been described in 20% and 6.5% of Candida, respectively [4]. Last, patients’ severity and comorbidities, such as immunosuppression, are known risk factor for candidemia [11] which may participate in the observed grim prognosis of candidemia.

Half of the patients with candidemia are critically ill [1, 4], mortality in patients with hemodynamic instability reaching 70% [7]. Underlying malignancy, either hematological or solid tumors, is frequently associated with candidemia, half of the patients with candidemia having underlying malignancy [4]. These patients may have specific risk factors for candidemia such as catheter or neutropenia, risk factors for fluconazole-resistant species such as prophylaxis, or specific acute condition such as typhlitis [12]. Nonetheless, immunocompromised critically ill patients with candidemia have been poorly described except in specific subgroups such as HIV infected patients [12] or organ transplant recipients [13].

The primary objective of this study was to assess outcome of immunocompromised critically ill patients with candidemia. Our secondary objectives were to describe clinical features of these patients and fungal ecology and to identify prognostic factors in this setting.

The observed poor prognosis is concordant with previous studies in this field. Hence, overall mortality after candidemia has been reported to be up to 40% in the general population [1‐4], rising to 50% in critically ill patients [1, 4] and 70% in patients with septic shock [7]. Mortality of onco-hematological patients with candidemia was reported to be of 40% [4, 17, 18]. Recently, Lortholary et al. [4] reported in a large prospective cohort of candidemia, a mortality of 50% for onco-hematological patients admitted in ICU with candidemia. Our data are in line with these reports, suggesting a high mortality associated with both underlying immune defect and underlying comorbidities.

Interestingly, in this study, most of the variables associated with outcome were surrogate of patients’ severity. These data are concordant with previous studies that reported higher mortality of critically ill patients with candidemia [1, 4]. In this line, Kollef et al. [7], reported a 69% mortality in patients with septic shock and candidemia. Although neutropenia has been associated with poor outcome in the general population of patients with candidemia [8], we were unable to detect such an effect in this study. The only exception is the peculiar population of allogeneic stem cell transplant recipients. These data are concordant with the poor prognosis of critically ill allogeneic stem cell transplant recipients, in whom a mortality of 51% was described, rising to 71% when mechanical ventilation was needed [19]. In this cohort, allogeneic stem cell transplant recipients with candidemia had a hospital mortality of 88%. This high mortality may reflect cumulative impact of mortality risk in this subgroup, direct influence of candidemia and the fact that candidemia may be a surrogate marker of severity or underlying immune defect in these patients [20, 21]. This later is further underlined by the lack of influence of candidemia on outcome after adjustment for confounders and when compared to a control group of patients without candidemia [16].

Last, several findings specific to candidemia and its management are to be noted. First, Candida specie and susceptibility had no influence on outcome. Some of the previous studies reported Candida species, namely C. glabrata or C. parapsilosis to be associated with better prognosis [4, 8, 22]. This association is, however, inconstantly reported in the literature [4, 5], and our results do not support such association. Our study may, however, lack statistical power to detect such an effect. Similarly, we were unable to detect the influence of management strategies on outcome. Neither initial therapy nor catheter withdrawal was associated with outcome while having been demonstrated factors associated with survival for patients with candidemia by previous studies [7, 8, 23], and being recommended by guidelines [6]. This may be explained by the homogenous management strategies in this study leading to high rate of catheter removal and systemic initiation of antifungal treatment the day of Candida identification. Nonetheless, the delay from onset of symptoms to antifungal therapy or catheter removal remained unacceptably long with a median of 3 days, reflecting time to culture positivity. Hence, despite rate of antifungal prophylaxis in this population (18%) and preemptive therapies, antifungal therapy remains dictated in this setting by culture positivity. This may reflect either a lack of clinical vignette specific enough to lead to adequate preemptive therapy initiation or failure to identify these vignettes. Development and validation of strategy that may allow reduction of this delay may be required. In this line, extension and validation of preemptive strategy in immunocompromised patients, excluded from recent trials [9, 10], validation of biomarkers driven strategies in this setting might deserve to be evaluated.

Our study has several limitations. First, the observational design and lack of a control group preclude any causality inference in this setting. In addition, this study was performed in only three centers, with high volume and experience of immunocompromised patients. This could explain the homogeneity in terms of patients cares and might constitute a selection bias. In this line, rate of hematological malignancies was high, and over-represented, in line with case mix in participating centers. Whether this may partly explain the lack of influence of underlying disease may deserve to be assessed by additional studies. In this line, the study period extends over a decade and changes in practices may have influenced our findings. We described two different groups of patients with candidemia, namely patients with ICU acquired candidemia and patient admitted in ICU for candidemia. If mixing these two groups might be misleading, we demonstrated similar features and outcomes of them. Moreover, despite the relatively large sample size in line with an uncommon disease, our negative findings might be related to the lack of statistical power. Thus, the lack of influence of Candida specie or Candida susceptibility and the absence of influence of management strategy may reflect the lack of statistical power rather than the lack of influence. Last, although we failed to observe an association between candidemia and mortality, this post hoc analysis was limited by selection bias in the control group. Since comparability across groups cannot be ensured after adjustment, these results are to be interpreted cautiously. Nonetheless, this post hoc analysis, although hypothesis generating, may suggest a lack of increased mortality in patients with candidemia after adjustment of case mix and patients’ severity that may deserve to be explored in future studies.