Risks, diagnosis and outcomes of invasive fungal infections in haematopoietic stem cell transplant recipients


Dr K. A. Marr, Division of Infectious Diseases, Oregon Health and Science University, 3181 SW Sam Jackson Park Rd, Portland, OR 97239, USA. E-mail: kmarr@fhcrc.org


Invasive fungal infections (IFIs) continue to cause considerable morbidity and mortality in haematopoietic stem cell transplant (HSCT) recipients. This review focuses on the risks for, and diagnosis of, IFIs (candidiasis, aspergillosis and other mould infections), and factors that affect current outcomes. Diagnosis of IFI is difficult, with the sensitivity of the gold standard tests (culture and histopathology) often <50%. Therefore, physicians rely on a constellation of clinical signs, radiography, culture, histopathology and adjunctive tests to establish diagnosis. HSCT recipients often have multiple co-morbidities, and understanding the current outcomes and prognostic variables is therefore important for overall management. This paper reviews historical trends and current data.

Yeasts and moulds are endogenous colonizers and ubiquitous inhabitants of the environment; the healthy human immune system appears well equipped to defend against invasive infection. However, in immunosuppressed patients, many yeasts and moulds become opportunistic pathogens. As a result, an eclectic list of fungi has been documented to cause invasive fungal infections (IFI) in haematopoietic stem cell transplantation (HSCT) recipients. However, the most common causes of IFI, and those that will be discussed in this review, can be divided into three groups: (i) Candida species [yeasts that disseminate systemically from mucosal surfaces to cause invasive candidiasis (IC)]; (ii) Aspergillus species (environmental moulds that cause invasive aspergillosis (IA); and (iii) other moulds (Zygomycetes, Fusarium species and Scedosporium species) that are increasingly reported as important pathogens in HSCT recipients. This review focuses on the risks, diagnosis, and outcomes of these IFIs. Many new antifungal drugs have been introduced in recent years; this review does not attempt to discuss therapy. For reviews on antifungal therapy in adults and children, the reader is referred to recently published manuscripts (Kauffman, 2006; Blyth et al, 2007).

Epidemiology and risks for IFI

Factors that are important in contributing to HSCT recipients’ risk for IFI include recent-past and present immune function, impacted by cellular deficiencies and dysfunction associated with administration of pharmacologic agents (Fukuda et al, 2003), use of indwelling devices, such as intravascular (IV) catheters, and breakdown of physical barriers that normally impair systemic invasion (e.g. translocation of Candida species across the intestinal mucosa) (Pfaller & Diekema, 2007), and exposure to inhaled fungi during periods of risk. In addition, new data are indicating the importance of underlying innate immunity of allogeneic HSCT donors in conferring risks to the recipients (Bochud et al, 2006), and uncovering ‘biological’ variables, such as iron overload, which impact risks for infection (Altes et al, 2004, 2007; Miceli et al, 2006; Armand et al, 2007). Risks specific to each group of fungi are discussed below and summarized in Table I.

Table I.   Factors reported to increase risk for IFI.
CandidiasisInvasive mould infections*
  1. *Most data recovered from studies specific to aspergillosis.

  2. HSCT, haematotpoietic stem cell transplantation; GVHD, graft-versus-host disease; TBI, total body irradation; CMV, cytomegalovirus; HLA, human leucocyte antigen; URD, unrelated donor; BM, bone marrow; PBSC, peripheral blood stem cells; LAF, laminar air flow; IFI, invasive fungal infections.

 Older age Underlying disease: haematological malignancy in other than first remission,  aplastic anemia, myelodysplastic syndrome and multiple myeloma
 Underlying disease not aplastic anaemia Older age
  CMV donor+/recipient+
  Iron overload pre/post-transplant
  Known colonization and/or pretransplant superficial infection (e.g. onychomycosis)
Treatment/HSCT complication
 Indwelling catheters Type of transplant (allo > auto)
 GI-tract colonization Type of donor (HLA match/URD)
 Severe GI tract mucositis: associated with  conditioning or GVHD Stem cells: cord blood > BM or PBSC
 High dose TBI Specific conditioning regimens and stem cell
 Protracted neutropenia Product manipulation (e.g. CD34 selection)
  Severe acute and chronic GVHD and specific therapies (e.g. high dose,  prolonged corticosteroids)
  Development of CMV disease
  Transplant environmental conditions (e.g. building construction, summer and lack of LAF)
  Respiratory virus infections (parainfluenza 3 and respiratory syncytial virus)
  Secondary neutropenia and other factors that delay neutrophil and T cell engraftment

Candida species are a common inhabitant of the human gastrointestinal (GI) tract and skin. Breaks in the integrity of skin and GI tract mucosal barriers such as occurs during mucositis, may lead to IC, particularly in the context of immunosuppression (Pappas, 2006). Older studies, performed prior to widespread prophylactic use of azole drugs, showed that underlying patient age, acute graft-versus-host disease (GVHD), donor human leucocyte antigen (HLA) mismatch, duration of neutropenia, underlying disease and specific conditioning agents (such as high dose total body irradiation) contribute to patient risks for IC (Goodrich et al, 1991). Use of prophylactic azoles had decreased the attack rate and 1-year cumulative incidence of IC, although recent studies have reported a consistent low incidence of ‘breakthrough’ infection, especially in the highest-risk patient population. Most recent studies report incidence rates in 5% range, with risk factors including (GI)-tract colonization, cytomegalovirus (CMV) disease, and prior episodes of bacteraemia (Marr et al, 2000). Risks associated with GI tract breakdown are one of the most important distinguishing features of this patient population; molecular studies have confirmed that colonizing isolates of C. glabrata and C. krusei can cause invasive infection in the setting of mucositis and GVHD (Redding et al, 2004; Westbrook et al, 2007). A notable exception is infection with C. parapsilosis, which more frequently is associated with skin colonization and contamination of IV infusates, although it also colonizes the GI tract, especially in preterm infants (Clark et al, 2004; Posteraro et al, 2004; Kaufman et al, 2006). All pathogenic Candida species can cause disease in this patient population; risks for specific infection include use of prior or concurrent antifungal therapies that dictate colonization rates, and patient age, which appears to also influence colonization. For instance, C. glabrata more frequently causes infection in older patients, while C. parapsilosis more frequently infects younger patients (Malani et al, 2005).

Aspergillus spp. are ubiquitous environmental moulds. Many hundreds of conidia (spores) are inhaled daily, and in most people, they are cleared without clinical consequence (Hope et al, 2005). However, a depressed immune system may present risks for IA, which may be accompanied by systemic dissemination (Morgan et al, 2005). The relative contribution of exposure during periods of risk- or ‘reactivation’ of dormant conidia during periods of risk – is not well defined. In recipients of HSCT, the incidence of IA increased during the 1990’s (Marr et al, 2002a), and although recent evidence suggests that the incidence has at least stabilized, infection is still common (Slavin et al, 1995; Marr et al, 2002a; Martino et al, 2002; Pagano et al, 2006). It is of note that some centres report a much higher incidence than others, suggesting geographical influence in either the exposure to Aspergillus species, differences in host risks, or significant diagnostic bias (Perfect et al, 2001; Morgan et al, 2005). The time-interval between transplantation and the presentation of IA has increased, largely associated with shifts in transplant practices, such as use of peripheral blood rather than bone marrow (BM) as stem cell source, and non-myleoablative conditioning regimens (Grow et al, 2002; Marr et al, 2002a,b; Fukuda et al, 2003; Wald et al, 1997; Upton et al, 2007). Clinicians must therefore have a higher suspicion of IA longer after transplant, especially in patients who have long-term immunosuppression associated with GVHD (Wald et al, 1997; Martino et al, 2002; Upton et al, 2007).

Both autologous and allogeneic HSCT recipients have risks for IA, although GVHD and associated therapies both increase and extend duration of risks in the allogeneic population (Wald et al, 1997; Marr et al, 2002b; Upton et al, 2007). Somewhat surprising is the finding that non-myeloablative HSCT recipients have been reported to have an equivalent, if not higher incidence of IA (Martino et al, 2001; Junghanss et al, 2002; Daly et al, 2003; Fukuda et al, 2003); this observation serves to emphasize the importance of GVHD (and therapies) in driving the current epidemiology of infection. Specific risks have been identified and are summarized in Table I.

Of particular interest are the results of recent studies suggesting that specific polymorphisms in ‘innate immunity’ genes, such as IL10, TNFRSF1A and TLR4, are associated with relative risks for IA in HSCT recipients and haematology patients (Bochud et al, 2006; Sainz et al, 2007a,b; Seo et al, 2005). These observations complement a number of other studies that serve to identify more ‘biological’ explanations for infection risks after HSCT; for instance, iron overload increasingly appears to be a risk factor in patients who develop mould infections (Altes et al, 2004, 2007; Miceli et al, 2006).

Zygomycetes, Fusarium spp. and Scedosporium spp. are emerging as important mould infections in HSCT recipients (Pagano et al, 2006). There are few studies examining risks for specific mould infections. In one study, infections caused by organisms from the class Zygomycetes represented 20% of culture-confirmed mould infections in HSCT recipients (Kontoyiannis et al, 2005), and in another study, Zygomycetes constituted 4·2% of invasive mould infections (Marr et al, 2002a). Other studies suggest that the incidence rates may be increasing (Imhof et al, 2004).

Zygomycetes are ubiquitous in the soil. After inhalation of spores or penetration through the skin, and in the presence of appropriate host risk factors, these moulds become angioinvasive, with the potential to disseminate systemically. Within the class Zygomycetes are three orders: Mucorales, Motierellales and Entomophthorales. Rhizopus spp. and Mucor spp. from the Mucorales order are the most common Zygomycetes encountered, and infections are often termed mucormycosis (Roden et al, 2005). Risks for zygomycosis in HSCT recipients largely overlap with those of aspergillosis, and include prolonged neutropenia, receipt of HLA mismatched and unrelated donor (URD) transplant, corticosteroids, severe GVHD and iron overload (Gaziev et al, 1996; Roden et al, 2005). Diabetes is a risk factor for zygomycosis independent of immune-status and HSCT, possibly in part because of abnormal iron metabolism as part of diabetic disease (Kontoyiannis et al, 2005; Roden et al, 2005; Kontoyiannis & Lewis, 2006). Recent studies have noted increasing numbers of zygomycosis cases presenting during prophylaxis or treatment of IA with voriconazole (Imhof et al, 2004; Kontoyiannis et al, 2005; Trifilio et al, 2007). Whether this increased frequency during IA prophylaxis is the result of selective pressure from azole use, diagnostic bias, or changes in hosts are not clear.

Fusarium spp. are an infrequent, but serious cause of IFIs in HSCT recipients. These organisms are plant pathogens found in the soil, but they are also waterborne and in biofilms (Elvers et al, 1998; Nucci, 2003). Transmission to HSCT recipients has been documented through water sources, such as in patients taking showers (Anaissie et al, 2001). One recent study that analysed the genetic diversity of multiple Fusarium oxysporum isolates in the US found that a recently dispersed clonal lineage was responsible for over 70% of clinical infections; of these, many clinical isolates were found to match isolates found in hospital water supplies (O'Donnell et al, 2004). These data confirm the importance of nosocomial water transmission of this organism. Another portal of entry is at areas of skin breakdown, onychomycosis (Pavlovic & Bulajic, 2006) or through central venous catheters (Nucci et al, 2004). Receipt of an HLA-mismatched or URD transplant and the underlying diagnosis of multiple myeloma appear to be risk factors for fusariosis (Nucci et al, 2004). Similar to aspergillosis, fusariosis presents early after transplant, which appears to be related to previous prolonged neutropenia (Nucci et al, 2004), and later after transplant, when GVHD, steroid receipt, and T-cell depletion are major risks (Nucci et al, 2004). The clinical presentation of infection may be different compared with aspergillosis, as more patients with fusariosis have multiple skin lesions and positive blood cultures (Boutati & Anaissie, 1997).

In immunocompromised hosts, other potentially ‘emerging’ moulds include Scedosporium prolificans and S. apiospermum. These organisms have a high degree of geographical restriction; disseminated infection with S. prolificans has been reported most frequently in Spain and Australia (Berenguer et al, 1997; Slavin, 2002). They produce disease resembling aspergillosis (Perfect et al, 2003; Bhatti et al, 2006), and have been associated with similar risk factors of neutropenia and GVHD (Marr et al, 2002a; Husain et al, 2005).


An enormous challenge in the diagnosis of IFI is the inadequate sensitivities of the current ‘gold standard’ diagnostic methods (e.g. blood cultures for Candida spp., or culture of a mould from a sterile site). An ideal diagnostic test for IFIs would have very high sensitivity for early disease; current diagnostic methods have not reached this goal. Diagnosis is therefore based on a combination of the symptoms and signs, histopathology and mycological growth, and the detection of fungal nucleic acids or wall components, such as galactomannan (GM) or β-glucan. However, to obtain a realistic post-test probability of the presence of an IFI, these tests can only be undertaken and interpreted by the physician in the context of the individual patient's risk for IFI. Risk assessment is therefore a crucial step in the diagnostic process.

Symptoms and signs

Whilst there are some pathognomonic clinical signs for specific infections, such as the papular rash in acute candidemia, the symptoms and signs of IFI are notoriously vague, and are easy to discount or miss. However, early detection by biopsy of a small skin lesion for instance may well be life-saving if IFI is diagnosed and intervention started early. Detailed attention to the review of the systems and physical examination is particularly important in those HSCT recipients who have a higher risk of IFI, even many months after transplant.

Invasive candidiasis may present with the non-specific symptoms of fever of unknown origin, or a sepsis syndrome (Pappas, 2006). The classic picture of acute IC with a discrete erythematous or haemorrhagic palpable rash, consistent with small vessel vasculitis may not occur, or may occur in the absence of positive blood cultures (Pappas, 2006). IC is a metastatic disease, and may present with infection in any organ system-for instance meningitis or endopthalmitis (Pappas, 2006).

Chronic disseminated candidiasis, or hepatosplenic candidiasis, is a different clinical entity compared with acute candidemia. During neutropenia and mucosal breakdown, Candida species, especially C. albicans, can invade into the portal vasculature and disseminate to the liver and/or spleen. In this setting, positive blood cultures are infrequent. Right upper quadrant pain and/or increased serum liver alkaline phosphatase levels develop during neutrophil recovery. A computed tomography (CT) scan can show multiple lesions throughout the liver and spleen. These lesions have actually been shown to decrease in appearance during subsequent periods of neutropenia (Pestalozzi et al, 1997); they are largely thought to be a product of the inflammatory response. Because recurrence of malignancy or other infections (including mould infections) is part of the differential diagnosis of these lesions, in the absence of positive blood cultures, diagnostic biopsy of these lesions may be helpful, although the sensitivity for diagnosis of IFI is not high. Frequently, these infections are accompanied by a large amount of chronic inflammation, and a prolonged febrile course can result despite appropriate antifungal therapy.

The clinical syndromes associated with the invasive mould infections in HSCT recipients are generally similar, often vague and usually progress rapidly. Conidia in the airways initiate infection, with germination into hyphal forms that can spread by local invasion, with the potential for haematogenous dissemination. Alternatively, moulds may invade through breaks in the skin. The clinical syndrome associated with pulmonary disease caused by moulds may include fever, chest pain, cough and haemoptysis. Respiratory failure can develop with progressive growth of mould and inflammation, which can result in necrotic debris and pulmonary infarction. Recently, it was suggested that a significant amount of clinical manifestations result from an immune reconstitution syndrome that develops concurrent with resolution of neutropenia (Miceli et al, 2007). Aspergillus lung infection may also present with highly aggressive Aspergillus tracheobronchitis (ATB), which, although not as common as pulmonary aspergillosis, is associated with a high mortality (Denning, 1995; Tasci et al, 2006). ATB presents in a variety of ways, including obstructive tracheobronchitis, where patients present with thick mucus plugs, and pseudomembranous ATB (Denning, 1995). Although these infections are more common in patients after lung transplantation and in the setting of advanced acquired immunodeficiency syndrome, they can occur in the haematology population as well. Rarely, fusariosis presents as obstructive bronchitis (Nucci & Anaissie, 2006).

Sino-orbital-cerebral mould infection presents with nasal stuffiness, headache, peri-orbital erythema or oedema, or non-specifically, with fever. As the disease progresses, often very rapidly, the meninges and neural tissue become involved, and proptosis and cranial nerve palsies become evident (Kontoyiannis & Lewis, 2006). As a result, even very mild swelling or erythema on the face of a neutropenic patient should be regarded with considerable alarm (Kontoyiannis & Lewis, 2006). Detectable black necrotic lesions are only observed in the nostrils and sinuses of only 50% of patients with early zygomycosis, so absence should not be used to exclude disease (Yohai et al, 1994).

As moulds are angioinvasive and cause end-artery embolization, papules or necrotic lesions on the skin may be painless. As a result, the patient is often unaware of their presence and so these lesions are easily missed unless actively searched for. Disseminated aspergillosis is typical of this, and often initially presents with a single small (centimetre-sized) non-painful dark papule. Skin lesions of disseminated Fusarium spp. may be painless but more often resemble Pseudomonas infection (ecthyma gangrenosum) with a dark purple/black painful necrotic centre, or have a target lesion with erythema surrounding a central papule, or even present as bullae (Nucci & Anaissie, 2002).

Moulds, particularly Zygomycetes and Fusarium spp., may also invade through broken areas of the skin, initially establishing locally necrotic infection, even causing fasciitis with, or without dissemination (Nucci & Anaissie, 2002; Kontoyiannis & Lewis, 2006). It is very important to note that whilst biopsy of new skin lesion may reveal an IFI diagnosis very quickly, a negative biopsy does not exclude the diagnosis of disseminated mould disease.

Mould infection of the GI tract is probably an under-recognized entity. This can occur from primary invasion or as a result of disseminated disease, and often presents as a GI bleed. In one study, only 25% of Zygomycetes intestinal infections were diagnosed premortem (Roden et al, 2005). Of note, patients with GI-tract IFI frequently have concomitant gut GVHD, which has placed them at high risk for IFI, but paradoxically confuses the clinical picture and the possibility of invasive GI mould infection is overlooked. These organisms can cause colitis in the setting of neutropenia as well, and should be considered in the differential diagnosis of ‘neutropenic colitis’ that is unresponsive to broad spectrum antibacterial therapy.

Whilst specific syndromes have been described, it is important to emphasize that often the only presenting symptom of mould infection is fever, or new development of a pulmonary nodule. Different antifungal susceptibilities of the most frequent fungal pathogens underscores the importance of establishing a microbe-specific diagnosis, using the tools reviewed below.

Diagnostic studies


Radiology plays an important role in diagnosis and follow-up of fungal disease. CT and/or magnetic resonance imaging (MRI) scans are pivotal for the diagnosis of hepato-splenic candidiasis (Kontoyiannis et al, 2000). Lesions are usually detected with the return of neutrophils and accumulating inflammatory response, so the diagnosis cannot be excluded based on evaluation during neutropenia (Kontoyiannis et al, 2000). Echocardiography is an important component of diagnosis of candidal endocarditis, or vegetations on the end of a central venous catheter (Pappas, 2006).

Chest radiography is frequently too insensitive to reliably diagnose IA. A chest CT scan is therefore an important diagnostic tool during febrile illness in the transplanted and neutropenic populations (Kuhlman et al, 1985). The radiographic presentation in neutropenic patients has been best described (Kuhlman et al, 1985). Infection usually starts as a nodular opacity with surrounding attenuation, or ‘halo sign’. This sign is not specific to IA, but appears early; as such, detection has been shown to be associated with better outcomes (Caillot et al, 2001a; Greene et al, 2007). IA lesions typically become larger during neutrophil engraftment and/or during the first 10 d of therapy (Caillot et al, 2001b), and they eventually cavitate, producing what is sometimes referred to as the ‘air crescent sign’ (Caillot et al, 2001b). This progression of signs however may not occur and the authors have seen IA presenting as isolated nodules, or dense lobar infiltrates. More recent studies have focused on the progression of radiographic abnormalities in patients who develop disease during the postengraftment period after allogeneic HSCT. One study documented the appearance of bronchopneumonia-type aspergillosis more frequently during this late period (Kojima et al, 2005).

Diagnosis of invasive pulmonary fusariosis, zygomycosis and scedosporiosis by CT scan is similar to IA, including a variety of features such as nodules (with or without the halo sign), ground glass opacities or even pulmonary infarction secondary to angioinvasion (Nucci, 2003; Kontoyiannis & Lewis, 2006; Nucci & Anaissie, 2006). However, there are few studies to guide our understanding of the natural history of non-Aspergillus pulmonary nodules. One recent study suggested that multiple (>10) pulmonary nodules detected on CT scan favours the diagnosis of pulmonary zygomycosis over IA (Chamilos et al, 2005). Another study suggested that zygomycosis has a predilection for the upper lobes (McAdams et al, 1997). All mould pulmonary infections may present with radiographic abnormalities very similar to aspergillosis; nodular lesions that increase in size during neutrophil recovery, halos representing surrounding inflammation and haemorrhage, and cavity formation is a typical course (Fig 1).

Figure 1.

 Radiographic images of documented invasive infection with Cunninghamella species in a HSCT recipient. The image shown in panel (B) was obtained 1 week after that shown in panel (A). The patient was treated with liposomal amphotericin B [initiated by CT finding in panel (A)], and ultimately was ‘cured’ of infection after lobectomy.

Radiographic diagnosis of rhino-cerebral mould infection may be difficult in the early stages, when the radiological signs are non-specific or absent. Because of the rapid and devastating nature of the disease, it is one of the few instances when frequent (daily) repeat CT or MRI scans may be pivotal to management (Kontoyiannis & Lewis, 2006). Mucosal thickening in the sinuses is notoriously difficult to interpret, particularly in the presence of mucositis, or after nasogastric tube placement. However, a totally opacified sinus, bony erosion, and sinus air-fluid levels in a neutropenic patient should be treated as highly suspicious of invasive mould infection (Kontoyiannis & Lewis, 2006). Orbital MRI and CT scans may be normal in the early stages of orbital extension, although MRI may have a higher sensitivity (Fatterpekar et al, 1999). Extraocular muscle thickening is highly suggestive of orbital invasion.


Whilst identification of a mould or yeast infection by culture or histopathology is considered the ‘gold standard’ for diagnosis of an IFI, these tests have frustratingly low sensitivities. Candida spp. are detected in blood cultures during candidemia between 40% and 70% of the time (Berenguer et al, 1993; Borst et al, 2001). Positive blood culture with Candida spp. should never be ignored. Fusarium spp. and Scedosporium spp. grow in blood cultures, and in one study Fusarium spp. was detected in blood cultures from 40·5% of patients with fusariosis (Nucci & Anaissie, 2006), whilst Scedosporium was detected in blood cultures of 80% of patients with invasive disease (Maertens et al, 2000). Aspergillus spp. and Zygomycetes are almost never recovered from the blood cultures.

The likelihood of growing fungi from sterile sites is often poor. However, all efforts should be made to culture fungi, as identification to the species level can guide appropriate therapy. Multiple studies have suggested that there exists inherently ‘resistant’ or difficult to treat Aspergillus species, which include A. ustus, A. lentulus and A. terreus (Balajee et al, 2006; Panackal et al, 2006). This has been studied in most depth for A. terreus; results of a multi-centre, retrospective study suggested poor outcomes in HSCT patients with A. terreus, that was at least partly associated with resistance to amphotericin B (Steinbach et al, 2004). Aspergillus species are currently determined by growth morphology (Hope et al, 2005), however, efforts are underway to establish improved rapid species-level identification.

Conidia are continually inhaled into the airways and Aspergillus spp. are one of the most common microbiology laboratory contaminants (Hope et al, 2005). A common question in HSCT patients concerns the predictive value for an IFI when a mould is isolated from a patients’ sputum. This has been under-studied, but one report suggested that a sputum positive for Aspergillus spp. in an HSCT patient has a predictive value for IA of 79% (Horvath & Dummer, 1996). An approach to this problem is to determine if the patient has evidence of disease, when treatment is warranted. If there is no evidence of active IFI, but the risk of the patient acquiring IFI from the ‘colonizing’ mould is high, giving prophylaxis along with ongoing evaluation for evidence of disease is probably sensible. However, growth of A. niger from the GI tract has been noted to have a low predictive value of disease, even in the highest-risk allogeneic HSCT population (Wald et al, 1997).

Histopathological diagnosis

Fungi may be identified by morphological features on histopathology. Understanding the basics of fungal stains is important for the physician to interpret the results of pathology studies. Fungal-specific stains include Gomori-silver stain (also known as Grocott's Methenamine Silver (GMS), Grocott or silver stain) and periodic acid schiff (PAS) stains (Hope et al, 2005). PAS has the benefit of highlighting the background host cellular detail, such as the relationship of the mould to blood vessels. GMS stain is much more sensitive in detecting small numbers of hyphae compared with PAS, but GMS does not stain the host tissue (Hope et al, 2005). Ideally, both stains should be used. Fluorescent dyes, such as calcofluor white, bind to beta-glucans in the fungal cell walls (Hope et al, 2005). Fluorescent dyes are very helpful for quick diagnosis on fresh tissue, such as frozen tissue sections and bronchial alveolar lavage (BAL) fluid. However, their sensitivity and specificity is unpredictable, as it depends on fungal load and on the type, quality and depth of the tissue specimen (Hope et al, 2005).

Candida spp. may be identified on histopathological analysis as budding yeasts; Candida albicans can appear filamentous, with pseudohyphae production in tissue sections. Aspergillus spp. are observed as septate acute-angle branching hyphae (Larone, 1995). The Zygomycetes are identified as wide filamentous fungi that are ‘ribbon like’ and aseptate (Larone, 1995; Hope et al, 2005). However, despite these classic descriptions, even highly experienced histopathologists often cannot identify the type of fungus, particularly with moulds, as the architectural integrity of the organism is difficult to preserve in pathology specimens (Hope et al, 2005). Aspergillus spp. are not distinguishable from other filamentous fungi, such as Fusarium spp. or Scedosporium spp. (Hope et al, 2005). Moreover, the pseudohyphae of Candida may be confused with filamentous fungi, particularly in less optimal tissue samples. Septa may be difficult to find and therefore distinguishing between Aspergillus spp. and Zygomycetes is frequently impossible (Hope et al, 2005). Most often, the crucial role of the histopathologist is to identify the presence of fungi, rather than the type. However, and very importantly, failure to identify fungi in a pathology specimen does not exclude an IFI.

Detection of fungal wall components and fungal nucleic acid

Galactomannan is a carbohydrate component of Aspergillus spp. cell wall that is released during hyphal growth and its detection in body fluids has become a useful aid in the diagnosis of IA (Marr et al, 2004; Mennink-Kersten et al, 2004; Pfeiffer et al, 2006). GM is measured using a double sandwich enzyme immunoassay (EIA), with the results reported as an ‘index’, or ratio of optical density of the sample compared with threshold controls provided in the kit (Marr et al, 2004). Utility of the GM EIA has been assayed in the clinical setting by sampling serum, BAL fluid, cerebrospinal fluid and pleural fluid. In HSCT recipients, the sensitivity, specificity and predictive values of the serum GM EIA have been reported to be variable, with particular discrepancy in sensitivities. One study reported a sensitivity and specificity of 89% and 92% in HSCT recipients (Marr et al, 2004). The GM assay has some important caveats, some of which probably explain the reported variability in performance. False negative results have been observed with concomitant use of anti-fungal agents, presumably because the level of GM is related to fungal burden (Marr et al, 2005). Patients who receive certain β-lactam antibiotics, such as piperacillin–tazobactam, and children, frequently have false positive GM EIAs (Machetti et al, 2005); several studies have noted that GM- or a molecule that reacts with the EIA antibody is found in food products, milk, and other fluids used for performing BAL, specifically, plasmalyte (Ansorg et al, 1997; Gangneux et al, 2002; Hage et al, 2007). It has also been noted that certain bacteria present in the GI tract (e.g. Bifidobacterium spp.) present cross-reactive epitopes, potentially explaining false positivity that occurs in the presence of GI tract mucositis (Mennink-Kersten et al, 2005). Kit to kit variation in the results have been demonstrated, particularly with interpreting results at the low end of positive: indices 0·5–0·7 (Upton et al, 2005). However, despite these caveats, the GM EIA is valuable. The reproducibly high negative predictive value may be especially important. For instance, one study showed that regular screening with pulmonary CT and GM-reduced empiric anti-fungal therapy, without altering clinical outcomes (Maertens et al, 2005).

The beta-d-glucan test detects β-glucans in the cell wall of numerous yeasts and moulds, with the exception of Cryptococcus spp. and Zygomycetes (Ostrosky-Zeichner et al, 2005). Early studies evaluating the test that is marketed in the United States suggest a sensitivity and specificity of 62% and 94% respectively, for diagnosing IFIs (Ostrosky-Zeichner et al, 2005). Further analysis is required before definitive performance characteristics can be determined. However, this test also shows great potential for screening or aiding diagnosis of IFIs.

The detection of nucleic acids by polymerase chain reaction (PCR) is under investigation. However, the major problem is the comparison of PCR results between centres, as often different primers, protocols and reagents are employed. As a result there are widely reported sensitivities and specificities of PCR detection of fungi in blood and tissue specimens (Buchheidt et al, 2004; Bialek et al, 2005; Rickerts et al, 2006; White et al, 2006). A recent study examined the sensitivity and specificity of the combination of GM EIA and PCR together in BAL fluid for the diagnosis of IA (Musher et al, 2004). This study found a combined sensitivity and specificity better than individual tests alone. However it also demonstrated that a negative value for one of the tests, but positive value for the other, may still be predictive of disease (Musher et al, 2004). PCR is also being used for pre-emptive screening of IA, with promising results; the results of one study that has been reported in preliminary form suggests potentially better outcomes when liposomal amphotericin B was administered based on serial PCR screening rather than by conventional means (Hebart et al, 2004; Halliday et al, 2006).

The above discussion emphasizes that the diagnosis of IFIs typically falls within a spectrum of certainty. Indeed, this is why consensus groups have established precise definitions of possible, probable and proven IA, using a constellation of clinical and laboratory findings (Ascioglu et al, 2002). These definitions are not meant to guide clinical care, but to provide consistency in definitions and probability of disease for clinical study purposes.

Outcomes and prognostic factors

Until very recently, the diagnosis of an IFI in a HSCT recipient carried with it an extremely poor prognosis, with mortalities estimated at 60–80% for mould infections. However, outcomes for some infections appear to be improving, probably associated with earlier diagnosis, changes in transplant practices, and less toxic treatment options (Upton et al, 2007).

One difficulty in discussing outcomes is the fact that true ‘attributable’ mortality is somewhat elusive; patients who develop these infections typically are quite ill with multiple comorbidities that drive ultimate outcomes. The attributable mortality in patients with haematological malignancies is not well described in the era of azole prophylaxis and, in most candidemia studies, haematology patients represent a small sub-population of the study cohort. However, in one recent study specific to patients with haematological malignancies the estimated attributable mortality was 30–40% (Pagano et al, 2006). Mortality appeared to be related to underlying disease, and it varied according to the species of Candida (Pagano et al, 2006). Infections with C. tropicalis and C. albicans tend to have the highest associated deaths, while infections caused by C. parapsilosis less frequently were associated with death (Pagano et al, 2006). Not surprisingly, the condition of the patient at the time of infection, as measured by Acute Physiology and Chronic Health Evaluation II scores, is an important predictor of outcome (Pappas et al, 2003; Cheng et al, 2006).

Outcomes of IA, and prognostic variables have been intensively studied in HSCT recipients, although many studies have been performed in single centres only. In one recent study, attributable mortality had dropped from 60–70% to 40% (Pagano et al, 2006). Another recent single-centre study showed a marked increase in the probability of survival in IA patients between 1990 and 2004 (Fig 2) (Upton et al, 2007). In the most recently evaluated period (2002–2004), survival 1 year after diagnosis of IA was 33%, however, the attributable mortality was estimated to be 17% and 23%, 30 and 90 d after diagnosis respectively (Upton et al, 2007). Distinction between the actual survival and the attributable mortality is important, as HSCT recipients have multiple co-morbidities. Many studies in the past have evaluated only survival after IFI, rather than attributable mortality, or relapse-free outcomes, and so comparisons across studies are difficult. None the less, the most recent studies (Pagano et al, 2006; Upton et al, 2007) clearly show trends to improved outcomes of this infection in the most recent years.

Figure 2.

 Kaplan–Meier probability of survival after diagnosis of invasive aspergillosis in haematopoietic stem cell transplant recipients in one institution, according to the year of IA diagnosis. The ’year of diagnosis’ is divided into five 3-year time-periods: 1990–1992, 1993–1995, 1996–1998, 1999–2001 and 2002–2004. Survival for diagnosis years 2002–2004 is better compared with that for 1990–2001 (P< 0·0001). Source: Upton et al, (2007).

Factors that have been found to decrease the risk of death after diagnosis of IA include receipt of peripheral blood as a stem cell source and reduced intensity, or non-myelaoablative conditioning therapy (Upton et al, 2007). Interestingly, the survival advantage associated with receipt of reduced intensity transplant was lost in multivariate modelling with inclusion of severe organ dysfunction (liver and kidney) as variables; presumably this occurred because co-morbidities have such a strong influence on survival, and are ‘linked’ to the type of conditioning therapy administered (Upton et al, 2007). A sub-analysis to determine the impact of antifungal drugs on attributable mortality confirmed a protective effect of voriconazole compared with prior therapies, in the cohort limited to patients who did not have severe organ dysfunction (Upton et al, 2007). Factors associated with outcomes in patients with IFIs are reviewed in Table II.

Table II.   Summary of reported survival from IFI after HSCT and factors associated with outcomes.
OrganismCandidemiaInvasive aspergillosisZygomycetes, Fusarium and Scedosporium
  1. Source: (Nucci et al, 2003, 2004; Husain et al, 2005; Cheng et al, 2006; Cordonnier et al, 2006; Wald et al, 1997; Roden et al, 2005; Pagano et al, 2006; Upton et al, 2007).

  2. APACHE II, acute physiology and chronic health evaluation II; HCST, haematopoietic stem cell transplantation; HLA, human leucocyte antigen; GVHD, graft-versus-host disease; CMV, cytomegalovirus; IA, invasive aspergillosis.

Survival 3 months after diagnosis60–70%44%Fusarium 13–21%
Zygomycetes 21–33%
Scedosporium 35–40%
Factors associated with outcomesUnderlying malignancy
Candida species
Shock, APACHE II score
Delayed diagnosis
Timing of diagnosis after HSCT
HLA mismatch
Source of stem cells
Conditioning regimen
Underlying liver, kidney function
Severity of GVHD
Neutrophil, monocyte count
CMV seropositivity
Amount and duration of corticosteroid administration
Disseminated IA, fungal load
Presence of pleural effusion
Pulmonary function prior to transplant
 Site of infection
 Species isolated
 (Cunninghamella worse) Fusariosis
 Disseminated infection

Outcomes of IA are also associated with extent of extra-pulmonary dissemination. In multivariate models, dissemination beyond the lungs has predictably been associated with high risks for death (Ribaud et al, 1999; Patterson et al, 2000; Lin et al, 2001). In a recent series of patients with central nervous system aspergillosis treated with voriconazole, outcomes appeared improved compared with previous experience with amphotericin B therapy (Schwartz et al, 2005, 2007). The precise reason for improved outcomes is unclear, and may involve multiple variables, including changes in antifungal therapy and/or hosts.

Infection with other moulds continue to have a dreadful prognosis, but because they are less frequent, our understanding of prognostic variables is more limited. In one recent study of IFI in patients with haematological malignancies, the highest rates of death occurred in people with zygomycosis (64%), followed by fusariosis (53%), and then aspergillosis (42%) and candidemia (33%) (Pagano et al, 2006). Zygomycoses outcomes appear variable and related to the form of disease and underlying host risk factors. In a retrospective review of 929 published cases of zygomycosis (Roden et al, 2005), the overall mortality was 44% for patients with underlying diabetes, 35% in patients with no underlying condition, and 66% in patients with underlying malignancies. Mortality also varied according to the site of infection: 96% of patients with disseminated disease died, 85% with GI infection died, and 76% with pulmonary infection died. In multivariate analysis, infection caused by Cunninghamella species and disseminated disease were independently associated with increased risks for death (Roden et al, 2005). Recent non-comparative studies of outcomes suggest more favourable outcome estimates with use of new antifungal drugs, with success approximating 60% at 12 weeks; however, it is difficult to determine whether this represents changes in hosts, host mix, and/or therapy – or whether the trends will continue (van Burik et al, 2006).

In one retrospective study examining fusariosis in HSCT recipients, the median duration of survival after diagnosis was 13 d, with only 13% of patients remaining alive 90 d after diagnosis (Nucci et al, 2004). The death rates of patients with early, late and very late fusariosis were 92%, 72% and 74% respectively (Nucci et al, 2004). Unlike IA, the type of HSCT (allogeneic or autologous) did not appear to make a difference in associated mortality (Nucci et al, 2004). Univariate analysis demonstrated that the predictors of death from fusariosis were GVHD and neutropenia, with the latter being the only significant prognostic variable in multivariate analysis (Nucci et al, 2004). In another study, independent predictors of poor outcomes included persistent neutropenia and use of corticosteroids (Nucci et al, 2003). The actuarial survival rate of patients without neutropenia or receipt of corticosteroids was 67% compared with 30% in patients who recovered from neutropenia but were receiving corticosteroids, and 4% in patients with persistent neutropenia only. None of the patients who had both risk factors survived (Nucci et al, 2003). It is notable that few studies have been published that evaluated risk factors for, or outcomes of invasive fusariosis in HSCT centres; for this reason, it is difficult to say how applicable the results are across geographically distinct transplant centres.

The mortality associated with Scedosporium spp. infections has been reported to approximate 70%. (Perfect et al, 2003). In a retrospective review of 80 cases of scedosporiosis in BM and solid organ transplant recipients, disseminated infection was associated with death, whereas adjunctive surgery was associated with survival (Husain et al, 2005). Receipt of voriconazole was associated with a trend towards better survival (Husain et al, 2005). As is the case for all non-controlled studies, it is difficult to determine how much the changes in therapy impacted outcomes, as opposed to (or in addition to) time-dependent changes in diagnostics and hosts.

The future

Earlier diagnosis of IFIs, and ‘tailoring’ of therapies, with a reduction in the amount of potentially toxic empirical anti-fungal agents administered are major goals in the management of HSCT recipients. Use of adjunctive diagnostic tests, such as the GM EIA, the 1–3 β-glucan assay and PCR technologies, may enable development of early diagnostic algorithms and pre-emptive therapy to more effectively prevent disseminated fungal disease. While preventative strategies need to be improved, there are already signs that outcomes are better than those historically documented. Changes in transplant practices, early diagnoses, and less toxic antifungal therapies will probably continue to influence the evolving epidemiology and outcomes of infection.


Dr Marr is supported by NIH grants AI054736, AI051468 and AI067710.