During the past several decades, there has been a steady increase in the frequency of opportunistic invasive fungal infections (IFIs) in immunocompromised patients. However, there is substantial controversy concerning optimal diagnostic criteria for these IFIs. Therefore, members of the European Organization for Research and Treatment of Cancer/Invasive Fungal Infections Cooperative Group and the National Institute of Allergy and Infectious Diseases Mycoses Study Group formed a consensus committee to develop standard definitions for IFIs for clinical research. On the basis of a review of literature and an international consensus, a set of research-oriented definitions for the IFIs most often seen and studied in immunocompromised patients with cancer is proposed. Three levels of probability are proposed: “proven,” “probable,” and “possible.” The definitions are intended for use in the context of clinical and/or epidemiological research, not for clinical decision making.
Opportunistic invasive fungal infections (IFIs) are a major cause of morbidity and mortality in immunocompromised patients. However, there still remains much uncertainty and controversy regarding the best methods for establishing the diagnosis of most IFIs. Practicing physicians approach this uncertainty by treating suspected cases empirically, whereas those who review cases for research purposes tend to accept only cases in which the diagnosis is certain. This disparity of approaches is particularly apparent in the conduct of clinical trials designed to show that a new drug exhibits sufficient efficacy.
These difficulties are not unique to the study of IFIs, and wide practice variations are known to exist in all areas of medicine [1, 2]. The uncertainty in disease definition is thought to be a key contributor to these variations . Strategies to minimize such uncertainties have resulted in movements such as evidence-based medicine  and practice guidelines . In studies in which there is no assurance that homogeneous populations are being evaluated, the selection of study subjects may be biased and, therefore, their findings cannot be used to make generalizations about cause, epidemiology, prognosis, treatment, or prevention . Typically, a set of characteristic abnormalities is used for diagnosis of diseases that do not have pathognomonic signs. Such classification and diagnostic criteria have proven to be extremely useful in areas that involve rheumatic diseases, endocarditis, and psychiatric diseases [6–8]. Also, subdivision by certainty of diagnosis is useful in many situations in which definite criteria cannot be applied to all cases . A series of estimates of probability (e.g., definite, proven, suspected, presumptive, and probable) is also a part of all of these systems, which is also evident from the literature on IFIs . Unfortunately, even these terms may take a range of meanings, and there is thus wide variation in their interpretation . The resulting array of descriptive phrases makes it difficult to pool data from multiple centers and thus hinders drug development, impedes the pace of clinical research, and perpetuates confusion in the literature. Although there are reference standards for diagnosing IFIs, these usually involve use of invasive procedures to obtain tissue specimens for culture and histological examination. Unfortunately, these procedures are not always feasible. Therefore, clinicians caring for such patients may rely on a combination of less-specific clinical, laboratory, and radiological data. Indeed, situations that present significant diagnostic uncertainty are the norm for most IFIs, and diagnostic criteria with perfect sensitivity and specificity do not exist for any IFI or, indeed, for many other diseases [12–16].
In an effort to standardize the definitions of IFIs for clinical research, the Invasive Fungal Infections Cooperative Group (IFICG) of the European Organization for Research and Treatment of Cancer (EORTC) convened a committee of EORTC/IFICG members with the goal of defining and classifying the IFIs that are commonly seen and studied in immunocompromised patients with cancer. At a later stage, members of the Mycoses Study Group (MSG) of the National Institute of Allergy and Infectious Diseases (NIAID) subsequently joined this consensus effort. This document is the final result of the combined efforts of this consensus committee, on the basis of several rounds of discussion that culminated in the approval of the document.
The committee sought to develop definitions for clinical researchers that were based on published information, deemed clinically applicable by experienced researchers, and practical within the context of the types of objectively verifiable data generated in daily clinical practice. Of importance, these guidelines should not be taken as strict rules for making or excluding the diagnosis of an IFI in clinical settings.
Defining the problem. A systematic review of the literature for an explicit identification of major problems related to heterogeneity of immunocompromised patients with cancer who have IFIs was undertaken and is described elsewhere . Pneumocystis infections were not considered. In brief, the abstracts of 7086 articles published from 1985 through 1997 were screened. Of these, 173 articles were finally selected because they were reports exclusively regarding clinical research on immunocompromised patients with cancer or recipients of hematopoietic stem cell transplants who also had deep-tissue fungal infections. The minimum diagnostic criteria used to include patients in the study were extracted from definitions devised by the investigators. Likewise, the criteria used to express different degrees of diagnostic probability were summarized, as were the terms most often used to express these levels of uncertainty.
Construction and function of the consensus committee. The IFICG, 1 of the 16 cooperative groups of the EORTC, created a task force in March 1997 under the chairmanship of Dr. Ben E. de Pauw. This initial committee consisted of 12 members of the IFICG from 8 different countries. Members of NIAID MSG were asked to join this committee in 1998.
The committee began with the information from the literature review, and each committee member was asked to provide comments on the terms and the construction they preferred. These comments led to revisions of the draft document in a cycle that was repeated many times until a final consensus was reached. During this process, the consensus committee also met face-to-face twice to discuss the proposal. In addition, the proposal was discussed in 3 group meetings of EORTC/IFICG, where it was open to all members of IFICG for discussion.
The definitions. We propose definitions for a new classification based on the level of certainty for the diagnosis of IFIs (tables 1 and 2). This proposal includes both diagnostic criteria for proven IFIs and also classification criteria for probable and possible diseases that are intended to promote a more uniform description of the patients when various research endeavors are reported.
The committee chose the terms “proven,” “probable,” and “possible” to express disease certainty. Although other terms could be used, our literature review showed a clear trend among investigators favoring the use of these terms. These verbal expressions of subjective probability correspond to a reasonably accurate numerical estimate [12, 14]. Three elements form the basis of the proposed definitions: host factors, clinical manifestations, and mycological results.
Host factors. We restricted the scope of the definitions to patients with cancer, whether treated or not, and to recipients of hematopoietic stem cell transplants who were suspected of having an IFI. However, because these patients do not present a single uniform entity, additional host factors need to be considered when confronted with IFIs that cannot be proven. Those additional factors (2) reflect the literature and the opinions of the consensus committee. The criteria for proven IFIs are likely valid for all host groups, not just patients with cancer and recipients of hematopoietic stem cell transplants.
Mycological evidence. Mycological evidence begins with the specimen. Thus, specimens obtained from normally sterile but clinically abnormal sites were rated to be more reliable than were those obtained from adjacent normal sites or sites normally colonized with resident commensal flora. These specimens were considered necessary for proving IFIs. Hence, the mycological evidence acquired by means of either direct examination or culture of specimens from sites that may be colonized (e.g., sputum, bronchoalveolar lavage fluid, or sinus aspirate) were thought only to support the diagnosis, not prove it. Similarly, with the sole exception of Cryptococcus neoformans, indirect tests to detect antigen were considered to be suggestive but not conclusive. Thus, the nature and quality of the specimen and the use of direct and indirect mycological techniques were incorporated into each of the criteria.
Clinical features. An attempt was made to distinguish between evidence of abnormality of an organ or organ system that was consistent with an IFI from evidence that could be associated with another infective process. For example, evidence of abnormal appearance by radiological or other imaging was given a much higher rating than were other, less specific signs, such as pleural rub. Symptoms and some other clinical features were also regarded as less specific and only supportive. Thus, 2 levels of evidence (major and minor) were incorporated into the concept of clinical features of IFIs.
The consensus for many of the definitions was determined after extensive debate. In many cases, that debate is instructive and elements of it are reviewed here.
Proven category of infections. The “proven” category consists of criteria that allow IFIs to be diagnosed with certainty and that differentiates between deep-tissue infections and fungemia. There was general agreement among committee members that the highest level of certainty in diagnosing an invasive fungal infectious disease is attained by establishing the presence of fungi in tissue by biopsy or a needle aspirate. However, the committee also agreed that demonstration of infection either by culture or histological examination was sufficient to be able to distinguish molds from yeasts and that both, although desirable, were not strictly necessary.
A branching septate mold in tissue is most commonly Aspergillus species. However, other organisms, including hyaline and dematiaceous molds, may be morphologically indistinguishable from Aspergillus species. A Fontana-Masson stain may help to distinguish hyaline from dematiaceous organisms. By comparison, the histological appearance of the ribbon-like broad coenocytic (sparsely septated) hyphal structures of Zygomycetes is usually readily distinguishable from Aspergillus species and other septated molds. The histological appearance of the endemic dimorphic fungi—Histoplasma capsulatum, as small intracellular budding yeasts; Coccidioides immitis, as spherules; Paracoccidioides brasiliensis, as large yeasts with multiple daughter yeasts in a “pilot wheel configuration”; and Blastomyces dermatitidis, as thick-walled broad-based budding yeasts—is sufficiently distinctive as to permit a definitive diagnosis as proven fungal infection caused by 1 of these pathogens. Whenever culture is possible, a specific diagnosis to the species level should be provided.
We propose that either typical microscopic findings or the detection of antigen in CSF be accepted as diagnostic for cryptococcosis in an immunocompromised host in the appropriate clinical setting . It is, however, important to be aware of the rare but definite causes of a false-positive cryptococcal antigen result, including infection caused by Trichosporon species , infection with Stomatococcus mucilaginosis , presence of circulating rheumatoid factor , and concomitant malignancy .
For the purposes of these definitions, the committee preferred the term “fungemia” to “bloodstream fungal infections,” because this avoids the impression that isolating a fungus from culture of blood signifies an infection that is confined to the bloodstream. For Fusarium species and Penicillium marneffei, fungemia, more likely than not, represents deep-tissue infection. Culture of Aspergillus species and other Penicillium species from the blood may represent serious disease but is more likely to represent specimen contamination; therefore, it is not taken as proof of diseases for purposes of clinical research.
Probable and possible categories of infection. Patients in these categories present enough information suggesting an IFI to warrant some form of empirical antifungal therapy. These patients are frequently characterized by being febrile, despite receipt of broad-spectrum antibiotics, and they may have a potential focus of infection. A definitive tissue diagnosis for radiologically demonstrable lesions, if present, is not considered feasible. Many independent reviewers would dismiss these cases for lack of convincing evidence. The problem of uncertainty cannot be understated or disregarded as if it does not exist, because both clinicians and researchers are regularly confronted with it. Rather, we sought to incorporate uncertainty into our quest for diagnosis and translate it into probabilities as accurately as possible [15, 16to be considered “probable,” each of the 3 elements of host factor, clinical features, and mycological evidence has to be present . By contrast, a patient who has at least 1 criterion from the host factors category but who does not have clinical features or mycological evidence has a case that can be classified only as “possible.” Because this is the least specific category, but one often used in clinical practice to treat patients empirically, we do not recommend its use in clinical trials of antifungal agents. Rather, it might be considered in studies of empirical treatment, epidemiological studies, and studies of health economics.
Aspergillosis. Because culture has such a poor sensitivity in the diagnosis of invasive aspergillosis, reliance on culture alone results in substantial underdiagnosis. On the other hand, cultures that yield Aspergillus species do not always reflect invasive disease, because colonization can occur in immunocompromised patients, and false-positive results that result from environmental contamination are occasionally a problem. Thus, the committee strongly supported the concept of a proven mold infection on the basis of the findings of histopathology or microscopy without necessarily requiring culture confirmation. The committee also proposed use of Aspergillus antigen testing as a finding that would support a probable diagnosis. The detection of Aspergillus antigenemia has been shown in experimental assays to correlate with clinical diagnosis and response to antifungal therapy [24, 25]. However, the clinical utility of these assays has been limited, in part because of their lack of widespread availability. Recently, a sandwich ELISA technique that uses a monoclonal antibody to galactomannan has been developed . This assay (licensed by Sanofi Diagnostics Pasteur to Bio-Rad Laboratories) has been used most extensively in western Europe. Recent prospective studies of hematology patients have demonstrated a sensitivity and specificity of ã90% with this method . Of note, there have been false-positive reactions, with recommendations from the manufacturer to consider the test a true-positive result only when ã1 sample is positive. The use of PCR to detect invasive fungal pathogens, including Aspergillus species, has been reported, but false-positive results can occur, and a standardized commercial method is not available . Thus, the committee decided that at the present time, the routine use of PCR in the diagnosis of invasive aspergillosis cannot be recommended.
Candidemia. Although other presentations are possible and are recognized by the definitions, candidemia is taken as a key sign of disseminated candidiasis. However, significant controversy surrounds the interpretation of positive blood culture results for candidemia . In reviewing these controversies, the committee recognized that not all groups of patients with candidemia have the same risk of clinically significant widespread dissemination. The principal classification factor is the presence or absence of neutropenia. Patients with neutropenia have a much higher rate of provable visceral dissemination and a much higher mortality rate . Although similar diagnostic criteria are applied to the 2 groups, patients with and without neutropenia cannot rationally be aggregated for purposes of data analysis. Also of note, candidemia, rather than being too nonspecific, is instead a marker (although insensitive) of deeply invasive candidiasis—that is, culture of blood samples often yields negative results in the presence of deep visceral candidiasis .
The second major subdivision of patients with candidemia revolves around the presence or absence of a central venous catheter. Broad dismissal of positive results of culture of blood from patients who have a central venous catheter in place is inappropriate. However, it is certainly possible that specimens drawn through a catheter would become contaminated during the collection process. The committee considered addressing this with requirements for specific numbers of blood cultures but thought that this was impractical. Although increasing numbers of positive blood culture results have been correlated with the likelihood of significant invasion , final blood culture data may not exist at the time a patient is considered for eligibility in a trial. In a related problem, the relevance of the site from which the culture-positive blood specimen was obtained was extensively discussed. Cultures obtained via peripheral venipuncture or through a newly inserted catheter would seem preferable on the general grounds that the catheters, rather than the patient, may be the cause of a culture-positive specimen. However, it is also clear that catheters may represent a major intravascular nidus of infection and that sampling via the catheter may provide early indication of a problem that has yet to reach a stage at which it can be detected in a more dilute specimen obtained from a remote site. In addition, records on the precise site at which a blood specimen was obtained are not consistently available or accurate. A comprehensive review of 155 episodes of catheter-related candidemia in patients with cancer found that, regardless of whether blood cultures were obtained from a central catheter or peripheral site, the frequency of autopsy-proven candidiasis was the same . Thus, the committee elected to establish no specific requirements for the site from which the specimen was obtained. To resolve all of these challenges, heavy reliance was placed on the requirement for concomitant signs of infection.
Patients with candidemia who have temporally related signs of infection are always considered to have a significant, proven infection. Given that the signs and symptoms are incorporated into the definition of proven candidemia, the category of probable candidemia relies on host factors and microbiological factors. Patients with neutropenia, patients with graft-versus-host disease, or patients receiving corticosteroids who have candidemia without signs of infection are considered to have probable disease. But, for example, patients who have neutropenia and candidemia, who do not have signs of infection, and who do not also meet ⩾1 of the other “host factors” criteria are considered to have a disease process that cannot reasonably be studied within the guidelines of this document. Although such patients should probably still receive some form of therapy , their relevance to a clinical trial is uncertain. The committee decided not to incorporate and define a “possible” category for candidemia. The definition for chronic disseminated candidiasis (hepatosplenic candidiasis) likewise required a modification because of the infrequency of microbiological support for this condition. For this diagnosis, the committee decided not to define a “possible” category, and for chronic disseminated candidiasis only, a microbiological criterion is not required for a “probable” diagnosis in a suitable host.
Other forms of invasive candidiasis. Other forms of invasive candidiasis are handled in 2 ways. First, any situation in which a biopsy of a normally sterile site shows Candida species by culture or histopathologic examination will qualify as proven candidiasis. Because of the many possible permutations (e.g., candidemia, candidemia with hepatosplenic involvement, hepatosplenic involvement without candidemia), the definitions do not specifically contain a list of possible relevant categories. Rather, these are all grouped under the general title of “proven invasive candidiasis.”
More challenging are the scenarios in which Candida species are isolated from such specimens as urine, sputum, or wound drainage. A very conservative approach was taken here. Although a rational physician might well chose to give antifungal therapy to a patient with fever and Candida species in the urine (or sputum, or wound drainage), this scenario is simply too ill-defined for study in a clinical trial.
Comparison with earlier NIAID MSG definitions. The proposed definitions in this manuscript differ from the earlier NIAID MSG definitions by focusing on the specific issues of oncology and hematopoietic stem cell transplantation. They recognize that the interplay of host factors and clinical manifestations may enhance the diagnostic probability of an IFI. The committee recognized that not all patients with neutropenia have the same risk for development of IFIs . For example, recovering Aspergillus species from the respiratory secretions of a patient with profound neutropenia who has acute leukemia or a patient receiving high dosages of corticosteroids carries much more significance for the development of invasive aspergillosis than it does when Aspergillus species are recovered from a patient with lung cancer and transient immunosuppression . The new system now also incorporates the use of galactomannan antigen in defining invasive aspergillosis.
Limitations of these definitions for clinical practice. These definitions should not be used to guide clinical practice. There are frequent clinical situations in the “possible” category in which therapy is warranted on empirical grounds. The objective of this project was to develop definitions that would identify reasonably homogeneous groups of patients for clinical research, as well as to foster international collaboration, design of clinical trials, and interpretation of new therapeutic interventions for management of IFIs in patients receiving cancer chemotherapy and those undergoing hematopoietic stem cell transplantation.
We acknowledge the contributions of P. Ljungman, M. Nucci, J. E. Patterson, G. Petrikkos, M. A. Pfaller, M. G. Rinaldi, and C. Viscoli, who took part in the consensus committee meetings or sent us comments.