Impaired gut function as risk factor for invasive candidiasis in neutropenic patients


N. M. A. Blijlevens, Department of Haematology, University Medical Centre, St. Radboud Nijmegen, P.O. Box 9101, NL-6500 HB Nijmegen, the Netherlands. E-mail:

Invasive candidiasis is a major cause of infectious complications related to cytotoxic treatment. While neutropenia is known to be an important risk factor, the damage to the epithelium of the intestinal tract has been largely ignored. When the gut is colonized by Candida species, mucosal barrier injury provides a portal of entry for the yeast. Methods for determining changes in gut function such as impaired gut integrity and increased gut permeability are discussed and have shown that there is a relationship between mucosal barrier injury and colonization on the one hand and the onset of candidiasis on the other. This information can be used to select patients at high risk for invasive disease, which includes those treated for acute leukaemia with a regimen containing anthracyclines or high-dose cytarabine as well as those receiving a haematopoietic stem cell transplant after total body irradiation (TBI). Such high-risk patients are the ones most likely to benefit from antifungal prophylaxis and would normally not require any additional measures for prevention.


Invasive fungal diseases remain a major cause of morbidity and mortality in neutropenic patients. Historically, patients with haematological malignancies, primarily leukaemia, were the first ones noted to be at an increased risk for invasive mycoses. The advent of haematopoietic stem cell transplantation (HSCT) further augmented the number of patients at risk and, hence, the incidence of and mortality due to invasive fungal diseases. Candidiasis remains the most commonly encountered fungal infection in neutropenic patients, with Candida albicans being most frequently involved (Jarvis and Martone, 1992). Infections due to non-albicans Candida species also occur with variable frequency in patients treated for haematological malignancies, partly because of the selective pressure brought about by the widespread use of fluconazole for prophylaxis (Kanda et al, 2000). Many immunocompromised patients have not one but several predisposing factors that contribute to an increased risk of developing opportunistic Candida infections. Disruption of the mucocutaneous barriers and defects of the immune system caused by the cytoreductive therapy, together with the colonization of the mucosal surfaces by yeasts, are the major risk factors predisposing neutropenic patients to candidiasis.


Candida colonization of the oral cavity and gut

The mucosal surfaces of healthy individuals are less likely to be colonized with C. albicans than the mucosa of hospitalized patients (2–37% versus 80%) (Odds, 1987). Persistence of Candida species and adherence to mucosal surfaces appears to be a prerequisite for local mucosal infection and, subsequently, invasive disease, as regular surveillance cultures of haematological patients have shown that colonization invariably precedes infection. In a retrospective study of 424 neutropenic patients only one of 198 patients free of Candida colonization developed invasive candidiasis, whereas 1·2% (2/170) of patients with one colonized site and 18 out of 56 (32%) patients with multiple non-contiguous colonized sites developed systemic candidiasis (Martino et al, 1989). Isolation of Candida species of the digestive tract is usually accepted as indicative of Candida colonization but there are no internationally accepted criteria for defining this. A practical definition of colonization is to require the same yeast to be recovered from two consecutive specimens from the same site or from two different sites at the same time. The burden of Candida species in terms of the number of colony-forming units (cfu) in each ml of saliva or g of faeces is also a risk factor for invasive disease. None of the 71 neutropenic patients with < 10 cfu/ml Candida species orally or < 103 cfu/g of faeces developed candidaemia, whereas 7 out of 37 patients with higher counts either orally or in faeces and 12 out of 14 patients with higher counts did (unpublished observations). Similarly, patients with oral candidiasis have been shown to harbour more than 400 cfu/ml saliva, whereas carriers of C. albicans had less yeast in their saliva (Epstein et al, 1980). While these data suggest a role for surveillance cultures for predicting disease, colonization with C. albicans does not correlate well with disease. In contrast, the absence of colonization is highly predictive of the absence of disease (Sandford et al, 1980; Laverdiere et al, 2000). The utility of surveillance cultures has been a subject of much debate and most clinicians do not order these cultures as they are viewed as not being cost-effective. This is as a result partly of not being able to reliably detect patients at risk and the unwillingness to accept that the absence of colonization effectively rules out the risk of candidiasis. However, the prevalence of Candida species colonization is high enough among patients treated with intensive cytotoxic drugs or those prepared for an HSCT that is possible to rule out those patients at low risk of invasive candidiasis based on negative surveillance cultures, thereby selecting, by default, those at greater risk.

Colonization with non-albicans Candida species

In contrast to other Candida species, colonization with Candida tropicalis is highly predictive of disease (Sandford et al, 1980; Laverdiere et al, 2000). Other non-albicans Candida species are being isolated with increased frequency, probably as a result of the widespread use of fluconazole. A surveillance study by the European Organization for Research and Treatment of Cancer (EORTC) documented 249 candidaemias out of 270 episodes of fungaemia (92%) in cancer patients (Viscoli et al, 1999). Breakthrough candidaemias accounted for 40% of the episodes in leukaemia patients and antifungal prophylaxis was significantly associated with non-albicans candidaemia as in 65% (50/77) non-albicans strains of Candida were isolated. Moreover, 67% of the C. tropicalis and 50% of the C. krusei strains were isolated from breakthrough episodes. Whether or not these species were already present on the mucosal surfaces before exposure to fluconazole and were only revealed afterwards is not clear. However, by the end of a fluconazole placebo-controlled randomized prophylaxis study involving 266 patients treated with either intensive cytotoxic therapy for acute leukaemia or an autologous marrow transplantation colonization with non-albicans Candida species increased equally in both arms (e.g. placebo: 8% to 18%, fluconazole: 7% to 21%) and C. glabrata was the most common non-albicans species encountered in both groups without causing invasive disease. Another retrospective analysis of breakthrough versus non-breakthrough candidaemia also failed to establish previous antifungal therapy as a risk factor, although more candidaemias were caused by C. glabrata and C. krusei in patients who developed breakthrough candidaemia on fluconazole than those on amphotericin-B therapy. (Uzun et al, 2001).

Bacteraemia as a risk factor of Candida colonization

Treatment with broad-spectrum antibiotics, especially those with anaerobic activity, cause sustained increase of yeast colonization of the gastrointestinal tract (Samonis et al, 1993). A single centre study of 341 treated patients with malignant haematology disorders showed that disseminated candidiasis occurred predominately in patients with high-level Candida colonization (> 10 colonies per sample at two or more body sites) and 9 out of 10 with a systemic Candida infection had microbiologically proven bacteraemia within the preceding week (Guiot et al, 1994). Only 17% of 46 patients who did not have microbiologically documented bacteraemia versus 43% of 35 patients with a bacterial infection had high-level colonization with Candida (Guiot et al, 1994). These results suggest an association between the occurrence of candidiasis and both the onset of a bacterial infection and the degree of Candida colonization. In another study, 210/249 (84%) of the patients received broad-spectrum antibacterial agents either for prophylaxis or therapy on the day of the first positive Candida blood culture (Viscoli et al, 1999). Antifungal prophylaxis might itself lead to a higher rate of bacteraemia in febrile neutropenic patients, as the rate of bacteraemia was investigated in 3002 febrile neutropenic patients, enrolled in four different trials of the International Antimicrobial Therapy Cooperative Group (IATCG) of the EORTC, and patients receiving absorbable and non-absorbable antifungal agents had a significantly higher rate of bacteraemia than patients who did not (Viscoli et al, 2001). However, the authors offered no explanation of how receipt of absorbable antifungal agents might have an impact on the bacteraemia rate and it may be that the occurrence of bacteraemia and the use of antifungal prophylaxis simply indicates a population of neutropenic patients at greater risk of developing infectious complications. Neither were the data concerning degree of Candida colonization reported, but had they been available the attributable risk could have been estimated in relation to the rate of bacteraemia.

Gut toxicity

The effect of TBI on invasive candidiasis

Animal studies in mice treated with antibiotics showed high numbers of viable Candida in the stomach, lower counts in the upper small intestines and a steady progressively increase along the intestinal tract (Wiesner et al, 2001). The adherence of Candida to the jejunal mucosa was optimal at neutral pH (Mehentee & Hay, 1989). TBI induced achlorhydria and, as with chemotherapy (methotrexate, 5-fluoruracil), adherence of C. albicans to gastrointestinal mucosa was increased, especially in the upper small intestine (Sandovsky-Losica & Segal, 1989, 1990). After intensive treatment, Candida disseminated to various visceral organs of mice that had been inoculated orally with the yeast (Sandovsky-Losica et al, 1992). A report of 665 recipients of an allogeneic, syngeneic or autologous HSCT also showed that the TBI that formed part of the preparative regimen increased the incidence of invasive candidiasis (Verfaillie et al, 1991) and that the dose exceeding 12 Gray further augmented the risk (Goodrich et al, 1991).

The effect of cytotoxic therapy on invasive candidiasis

Not only TBI but also intensive chemotherapy increases the risk for invasive candidiasis. A randomized double-blind trial of prophylaxis comparing oral fluconazole (400 mg/d) with placebo-identified patients undergoing either induction therapy with a regimen of cytarabine plus an anthracycline or high-dose cytarabine and those receiving an autologous HSCT without a haematopoietic growth factor as those who most benefited from fluconazole prophylaxis. Remission–induction cytotoxic therapy and a high colonization index were independent predictors of definite or probable invasive candidiasis (Rotstein et al, 1999). Bow et al (1997) evaluated the effect of remission–induction therapy on d-xylose absorption in 110 adults with acute myeloid leukaemia. Serum concentrations of d-xylose were obtained 1 h after imbibing 5 g of the sugar at baseline and weekly for 4 weeks until neutrophil recovery. Significantly less d-xylose was absorbed in week 2 and 3, especially after high-dose cytarabine- or idarubicin-containing regimens, and the onset of bacteraemia and candidaemia was significantly correlated with the lowest d-xylose absorption. The 10 patients who developed hepatosplenic candidiasis had lower absorption of d-xylose at baseline, a shorter time to the onset of malabsorption and significantly lower serum levels in week 2 (Bow et al, 1997). The progressive damage to the functional integrity of the intestinal epithelial surface as measured by the malabsorption of d-xylose (median time of onset 14 d) correlated with a risk for systemic candidiasis occurring at a median of 12 d after intensive cytotoxic therapy (idarubicin–etoposide–carboplatin) in 43 patients (Bow et al, 1998).

The effect of cytotoxic therapy on gut mucosal barrier

In the period of reduced d-xylose absorption, high toxicity grades (National Cancer Institute of Canada Modified Common Toxicity Criteria) were observed for diarrhoea, nausea, stomatitis, oesophagitis, dysphagia and vomiting, but only in a minority of the patients (Bow et al, 1998). The peak daily output of diarrhoea was measured prospectively in another study of 202 recipients of an autologous or allogeneic HSCT following ablative therapy with high-dose regimens, but no association was found between this and the rate of bacteraemia or transplant-related mortality (Rapoport et al, 1999). However, higher oral mucositis peak values correlated well with increased transplant-related mortality and bacteraemia, especially when TBI had been given as part of the preparative regimen. Incorporating TBI into the intensive regimen for autologous HSCT recipients induces severe diarrhoea and these patients tend to experience more episodes of bacteraemia (Callum et al, 1991). A prospective study of 429 HSCT recipients showed clearly that the nature of the myeloablative regimen was the most significant determinant of oral mucositis (Wardley et al, 2000). However, in contrast to the mouth, the symptoms and physical signs of gut damage by cytotoxic therapy are not specific enough and are often masked by the administration of morphine to relieve pain. There is also no validated system for scoring gut damage and the several ‘common toxicity criteria’ that exist are inappropriate for accurately describing the course and severity of intestinal mucosal damage. Several investigators, e.g. Bow et al (1998), used non-invasive methods to assess gastro-intestinal permeability in vivo which is an accepted surrogate for gut disease if not gut damage (Bjarnason et al, 1995).

Determining gut mucosal barrier injury

The principle features of intestinal injury induced by cytotoxic therapy are alterations in permeability and loss of epithelial surface. Intestinal mucosal barrier injury (MBI) after intensive treatment is a complex pathobiological process representing far more than simply a toxicological side-effect. MBI is considered to consist of at least four successive phases known as (1) an inflammatory phase, (2) an epithelial phase, (3) an ulcerative-bacteriological phase, and (4) a healing phase (Sonis, 1998; Blijlevens et al, 2000). Proinflammatory cytokines are immediately released after cytotoxic therapy not only by the immune effector cells but also by epithelial cells (inflammatory phase) (Xun et al, 1994). These cytokines are capable of modulating the so-called ‘tight junctions’ of the gut, which are dynamic structures exerting physiological control over the flow of solutes through paracellular spaces and, hence, important for permeability. Electron microscopy of the upper gut showed open ‘tight junctions’ 3 d after chemotherapy (i.e. during the epithelial phase) and increased apoptosis in the intestinal crypts with a corresponding decrease of villus length, mitotic count per crypt and enterocyte height (Keefe et al, 2000). Apoptosis precedes hypoplastic villus atrophy and the consequent cessation in cell renewal allows ulcers to evolve (ulcerative-bacteriological phase).

Permeability can be measured by means of urinary excretion of test substances or by detecting their presence in the blood. Disaccharides (lactulose), monosaccharides [l-rhamnose, mannitol, 3-ortho-methylglucose (3-OMG)], various polymers of polyethylene glycol or 51Cr-labelled ethylenediaminetetraacetic acid (51Cr-EDTA) have proved useful for different intestinal diseases (Bjarnason et al, 1995). The main disadvantage of 51Cr-EDTA is that it is radioactive and necessitates the patients leaving the ward for the duration of the test. Factors such as bowel transit time, gastric emptying and renal function interfere with the sugar test when only one sugar is used. When two different sugar probes are used they affect equally the pre- and post-mucosal determinants, allowing the urinary excretion ratio to function as an index of intestinal permeability. Given the enormous gut damage induced by cytotoxic therapy, it is striking how few reports have been published on this subject. One of the first reports by Parrilli et al (1982) using only lactulose permeation in nine patients treated with MOPP (mechlorethamine, vincristine, procarbazine, prednisone) or CVP (cyclophosphamide, vincristine, prednisone) regimens showed a marked increase of urinary lactulose excretion after only 2 d of chemotherapy. Maximum excretion took place after 10 d (Parrilli et al, 1982). Others used different assays; 51Cr-EDTA (Selby et al, 1984; Johansson and Ekman, 1997), lactulose/mannitol (Pledger et al, 1988; Fegan et al, 1990; Daenen et al, 1991), lactulose/rhamnose (Keefe et al, 1997) and lactulose/rhamnose/3-OMG/xylose (Parrilli et al, 1989) in different patient populations. Irrespective of the test used, the same general pattern of disturbed permeability is apparent, with an early increase in permeability reaching a maximum 10–14 d after starting cytotoxic therapy, followed by recovery during the following 3–4 weeks. Therefore, a combination of various sugars with different means of epithelial transport should be able to detect changes in permeability and integrity of the gut. To this end, a combination of lactulose to indicate paracellular transport, i.e. open tight junctions, 3-OMG to measure ATP-dependent transcellular transport, d-xylose to detect carrier-mediated transcellular transport and l-rhamnose to monitor passive transcellular diffusion seems to provide the best assay of gut function. If this is shown to be the case, this multisugar test could be used to indicate whether invasive infectious complications after intensive cytotoxic therapy will occur and when (Bow et al, 1997).

Neutropenic enterocolitis and invasive candidiasis

Neutropenic enterocolitis, also called typhlitis, is the most severe manifestation of gut MBI (Blijlevens et al, 2000). All the factors that contribute to the development of MBI, e.g. cytotoxic therapy-induced neutropenia, damaged mucosal membranes and altered gut microflora as a result of antibiotic exposure, are present. Bacteraemia due to Pseudomonas species, staphylococci, enterococci and Clostridia species have all been associated with typhlitis. Candidaemia might also represent typhlitis, albeit in a less pronounced form, because of the favourable environment resulting from mucosal damage and the extensive use of antibacterial prophylaxis or therapy (Micozzi et al, 1996; Seipelt et al, 1998). In a 18-year post-mortem review of fungi, Candida species were cultured from blood in 19/36 (53%) cases of typhlitis (Katz et al, 1990). After intensification of cytotoxic therapy, the incidence of bacteraemia was also higher in courses complicated by neutropenic enterocolitis (Vlasveld et al, 1991; Gomez et al, 1998). Girmenia et al (1999) used a dot immunobinding assay for Candida mannoprotein antigen and detected the antigen in 12 out of 20 patients suffering from neutropenic enterocolitis, all of whom were colonized at the time by Candida species (Girmenia et al, 1999). Antigenaemia was absent in all five subjects whose stools were free of Candida. Patients treated for leukaemia who developed neutropenic enterocolitis had significantly lower d-xylose absorption in week 2 after starting chemotherapy and time to onset of malabsorption was nearly a week earlier than in patients who did not develop typhlitis (Bow et al, 1997). It was also noted that neutropenic enterocolitis was strongly correlated with the development of candidaemia.

Selecting high-risk patients

The best way of managing invasive candidiasis is to prevent it, but the results of antifungal prophylaxis studies are inconclusive for several reasons. Differences in the definition of invasive fungal disease biased the evaluation of fungal-related mortality. The meta-analysis of antifungal prophylaxis undertaken by Gotzsche and Johansen (1997) was flawed because they considered oesophageal candidiasis alongside candidaemia, lung infection and microscopically confirmed deep tissue infection as invasive infection, whereas oesophageal candidiasis is more like oropharyngeal and vulvovaginal candidiasis and local skin infections or candiduria, and is therefore a superficial infection (Rex et al, 2000). This might explain why Gotzsche and Johanssen (1997) estimated that 73 patients would need to be treated in order to prevent one case of invasive fungal disease. The low prevalence of invasive candidiasis is also borne out by others, with the incidence ranging from one case per year to an average of four cases per year (Viscoli et al, 1999). A more recent meta-analysis of 16 oral fluconazole prophylaxis studies in neutropenic patients concluded that fluconazole seemed to be effective only in studies in which the incidence of systemic candidiasis was higher than 15% (Kanda et al, 2000). Therefore, the efficacy of antifungal prophylaxis in reducing the mortality attributed to invasive candidiasis can only be determined in studies that have selected high-risk patients. Other problems with prospective, randomized clinical trials of therapies for invasive fungal infections have been highlighted, such as difficulties with timely diagnosis, high costs because of slow enrolment, design flaws and variable spectrum of severity of illness and underlying disease (Rex et al, 2001).

An important consequence of MBI is the likelihood of drug levels declining because of malabsorption or when patients are unable to swallow the drugs (Michallet et al, 1998; Prentice and Donnelly, 2001). This can be overcome by choosing to continue therapy parenterally when MBI occurs.

Hence, prophylaxis should be reserved for a high-risk group of patients (Prentice et al, 2000) for which antifungal prophylaxis has been shown to lower the mortality related to invasive candidiasis (Goodman et al, 1992; Slavin et al, 1995; Rotstein et al, 1999). Such patients should be selected on the basis of two criteria, namely Candida colonization of the gut and impaired gut integrity (Fig 1). Candida colonization is relatively straightforward to determine whereas defining impaired gut integrity, using permeability tests, is still under investigation. This can be partly overcome by selecting those patients most likely to develop impaired gut integrity, i.e. recipients of either allogeneic or autologous HSCT receiving TBI as part of the ablative therapy (Goodrich et al, 1991; Verfaillie et al, 1991) and patients receiving the first course of remission–induction therapy for acute myeloid leukaemia with a regimen containing an anthracycline or high-dose cytarabine (Bow et al, 1997; Rotstein et al, 1999).

Figure 1.

Patients at high risk of developing invasive candidiasis are those who are colonized and who also suffer from gut mucosal barrier injury (MBI).