Epidemiology and sites of involvement of invasive fungal infections in patients with haematological malignancies: a 20-year autopsy study
Article first published online: 4 APR 2013
© 2013 Blackwell Verlag GmbH
Volume 56, Issue 6, pages 638–645, November 2013
How to Cite
Lewis, R. E., Cahyame-Zuniga, L., Leventakos, K., Chamilos, G., Ben-Ami, R., Tamboli, P., Tarrand, J., Bodey, G. P., Luna, M. and Kontoyiannis, D. P. (2013), Epidemiology and sites of involvement of invasive fungal infections in patients with haematological malignancies: a 20-year autopsy study. Mycoses, 56: 638–645. doi: 10.1111/myc.12081
- Issue published online: 22 OCT 2013
- Article first published online: 4 APR 2013
- Manuscript Accepted: 17 MAR 2013
- Manuscript Revised: 13 MAR 2013
- Manuscript Received: 11 JAN 2013
Autopsy studies remain an essential tool for understanding the patterns of fungal disease not detected ante mortem with current diagnostic approaches. We collected data concerning the microbiological trends, patient clinical characteristics and sites of involvement for invasive fungal infections (IFIs) identified at autopsy in a single large cancer treatment centre over a 20-year period (1989–2008). The autopsy rate and IFI prevalence both declined significantly during the study period. The prevalence of Aspergillus spp. decreased significantly from the first 15 years of the study (from 0.12 to 0.14 cases per 100 autopsies to 0.07 in 2004–2008; P = 0.04), with only Mucorales accounting for a greater proportion of IFIs over the duration of the study period (0.06 to 0.2 cases per 100 autopsies, P = 0.04). After 2003, moulds accounted for the majority of infections identified at autopsy in the spleen, kidney, heart and gastrointestinal tract. Despite a trend of decreasing prevalence from 1989 to 2004, invasive candidiasis increased in prevalence during later periods 2004–2008 (0.02–0.05 per 100 autopsies) with decreasing kidney, heart and spleen involvement. Despite a declining autopsy rate, these data suggest a decreasing prevalence overall of IFIs with changing patterns of dissemination in patients with haematological malignancies.
Invasive fungal infections (IFIs) remain an important cause of death in patients with leukaemia and recipients of haematopoietic stem cell transplantation (HSCT).[1-3] The epidemiology of IFIs has shifted over the past two decades, paralleling advances in treatment and transplantation for haematological malignancies, earlier IFI diagnosis and the introduction of new antifungal agents into clinical practice.[4-6] Since the 1990s, invasive aspergillosis has been the predominant IFI in patients with haematological malignancies,[1, 7] coinciding with the introduction and widespread use of fluconazole prophylaxis to reduce mortality associated with invasive candidiasis.
More recently, several cancer treatment centres have observed an increase in the prevalence of uncommon, but difficult-to-treat moulds such as Mucorales, Fusarium spp. and Phaeohyphomycetes.[3, 4, 6, 9, 10] The increase in these previously uncommon moulds has coincided with increasing antifungal resistance among Candida species[2, 11] and possibly also Aspergillus species.[12-14] Whether the emergence of these opportunistic fungi is simply a result of selection pressure induced by the widespread use of broad-spectrum antifungal agents, or a phenomenon driven by changes in chemotherapy and transplantation approaches, earlier diagnosis and changing demographics of at-risk populations is unknown.
Given the limited utility of current diagnostic approaches, autopsy series remain a key source of information for understanding the changing epidemiology of IFI in immunocompromised patient populations. Moreover, autopsy series provide a unique opportunity to explore trends of organ involvement by IFI. This may be especially relevant considering the pharmacokinetic limitations of some of the newer antifungal agents that have low or undetectable concentrations in some organs that are a common site of metastatic seeding with Candida or moulds.
In a previous study, we reported epidemiological and microbiological characteristics of IFIs identified in the autopsy examination of patients with haematological malignancies at our institution during the period from 1989 to 2003. In this study, we expanded our previous observations by examining patterns of organ involvement by IFIs as well as fungal species and immunosuppression-specific patterns associated with fungal dissemination over a 20-year period. The objective was to gain insight into how temporal trends in immunosuppression risk and antifungal exposure influence the epidemiology of IFI at autopsy in haematological malignancy patients.
Patients and study setting
Patients with haematological malignancies were identified who underwent autopsy examination at The University of Texas M. D. Anderson Cancer Center from January 1989, through August 2008. Autopsy and medical records were reviewed for demographic and cancer treatment information, including: the type and status of the underlying malignancy; the type and date of HSCT (if applicable); risk factors for IFIs [e.g. severe neutropenia, Grade III–IV graft-vs.-host disease (GvHD), receipt of a significant dose of corticosteroids]; human immunodeficiency virus infection status; the presence of intercurrent bacterial or viral infections; and the type of antifungal prophylaxis administered. In addition, data were collected on the fungal species identified in cultures from sterile sites, histopathological characteristics of organ involvement by IFIs, whether IFI contributed to death, and whether IFI was suspected ante mortem.
The EORTC/MSG criteria were applied for the ante mortem diagnosis of IFIs. A diagnosis of disseminated IFI required the involvement of two or more non-contiguous organs at autopsy. Mixed IFI was defined as the presence of more than one fungal morphotype (e.g. yeast and moulds) by histopathological examination, or the growth of two or more fungal pathogens in cultures drawn from a sterile site. Severe neutropenia was defined as a neutrophil count <100 mm−3 for more than 10 days. Significant corticosteroid use was defined as the use of a systemic corticosteroid at a cumulative dose equivalent to ≥600 mg of prednisone during the month prior to diagnosis of IFI. The date of death was considered the date of diagnosis if the infection was not detected ante mortem. IFI was considered to have contributed to patient death by the pathologist if there was significant involvement of a major organ (lung, heart, central nervous system) by the fungal pathogen.
Fungi that grew in culture were identified with the use of standard morphological criteria. In the case of mould infections where culture was negative, but with histopathology consistent with Aspergillus, these cases were recorded as culture-negative hyalohyphomycetes presumed to be Aspergillus. Similarly, in cases of yeast infection where culture was negative, but there was histopathological evidence of invasive yeast in tissue, the infection was recorded as culture-negative invasive candidiasis.
Trends in the prevalence and clinical characteristics of IFIs compared data from four 5-year periods (1989–1993; 1994–1998; 1999–2003 and 2004–2008) using the chi-square test for trend. Bivariate analysis was performed for demographic and clinical risk factors to screen for association with patterns of IFI organ involvement. Continuous variables were compared using anova with Tukey's test for differences. All P values <0.05 were considered significant. Statistical analysis was performed using SPSS Version 20, (IBM, Armonk, NY, USA).
Risk factors for IFIs
A total of 371 IFIs were identified by culture or histopathology in 1213 autopsies (31%) over the 20-year study period. The autopsy rate in our institution declined consistently from 0.63 autopsies per 100 deaths in 1989–1993 to 0.06 in 2004–2008 (P < 0.001; Table 1). The prevalence of IFIs at autopsy was stable during the first 15 years of the study (0.30–0.32 per 100 autopsies), but declined significantly during the last 5 years of the study to 0.19 cases per 100 autopsies (P < 0.001).
|Characteristic||1989–1993 n = 145 (%)||1994–1998 n = 86 (%)||1999–2003 n = 81 (%)||2004–2008 n = 59 (%)||P valuea|
|Autopsies per 100 deaths||0.63||0.35||0.27||0.06||<0.001|
|Prevalence of all invasive fungal infections (IFIs; per 100 autopsies)||0.32||0.30||0.31||0.19||<0.001|
|Male sex||87 (59)||54 (86)||40 (81)||40 (81)||0.18|
|Median age in years (range)||46 (15–87)||47 (2–83)||51 (19–77)||51 (3–79)||0.15|
|Acute myelogenous leukaemia||59 (41)||42 (49)||29 (36)||18 (31)||0.13|
|Myelodysplastic syndrome||8 (0.5)||5 (6)||7 (9)||2 (3)||0.62|
|Acute lymphoblastic leukaemia||23 (16)||16 (19)||19 (23)||5 (8)||0.13|
|Chronic myelogenous leukaemia/lymphoma||25 (17)||4 (5)||5 (6)||7 (12)||0.01|
|Non-Hodgkin's lymphoma||15 (10)||9 (10)||9 (11)||11 (19)||0.37|
|Chronic lymphocytic leukaemia||8 (6)||3 (3)||9 (11)||7 (12)||0.11|
|Other malignancies||7 (5)||7 (8)||5 (6)||9 (15)||0.08|
|Allogeneic haematopoietic stem cell transplantation (HSCT)||43 (30)||30 (35)||25 (31)||28 (47)||0.09|
|Severe neutropenia||130 (90)||70 (81)||57 (70)||21 (36)||<0.001|
|High-dose corticosteroids||28 (19)||56 (65)||41 (50)||50 (84)||<0.001|
|Autologous HSCT||7 (5)||2 (2)||4 (5)||2 (3)||0.77|
|Active malignancy||109 (75)||67 (78)||69 (85)||43 (73)||0.289|
|Active graft-vs.-host disease (Grade III–IV)||23 (16)||16 (19)||15 (19)||22 (37)||0.006|
|IFI evident only at autopsy||122 (84)||57 (66)||5 (68)||29 (49)||<0.001|
|IFI cause of death||111 (77)||70 (81)||59 (73)||29 (49)||<0.001|
|IFI breakthrough on antifungals||15 (10)||61 (71)||52 (64)||33 (56)||<0.001|
|Concomitant CMV infection||10 (7)||11 (13)||3 (4)||5 (8)||0.17|
|Concomitant bacterial infection||37 (26)||57 (66)||37 (46)||38 (64)||<0.001|
Several important changes in the demographic and clinical characteristics of patients with IFIs were observed over the 20-year study period (Table 1). A majority of autopsy subjects had acute myelogenous leukaemia or myelodysplastic syndrome, which represented between 40% and 50% of the malignancies associated with IFI. The frequency of patients with chronic myelogenous leukaemia or lymphoma decreased continuously during the first 15 years of the study period, but increased modestly during the final 5 years (P = 0.01). The percentage of patients with non-Hodgkin's lymphoma or chronic lymphocytic leukaemia also increased over the study period, but this trend was not significant. The vast majority of patients had evidence of active malignancy at autopsy (75–85%) that was constant during 20-year period. The number of autopsied patients who had received an allogeneic HSCT also increased during the study period from 30% to 47% (P = 0.08). Relatively fewer patients received autologous transplantation, ranging from 2% to 5%.
The prevalence of severe neutropenia as a predisposing risk factor for IFIs prior to patient death declined over the 20 year study period from 90% of autopsy cases in 1989–1993 to 44% in 2004–2008, P < 0.001; Table 1. In contrast, the percentage of patients exposed to high-dose corticosteroids increased (21% to 81% (P < 0.001) as did the prevalence of grade III–IV GVHD after HSCT (16–37%, P = 0.006).
Antemortem IFI diagnosis improved during the study from 16% in 1989–1993 to 51% in 2004–2008, (P < 0.001). The rate of breakthrough infections declined from 1994 to 2008 (71–56%, P < 0.001). Most IFIs during later periods of the study were associated with concomitant bacterial infection (64%). Notably, death attributed to the IFI remained at as stable rate during the first 15 years of the autopsy records (70–80%), but decreased to 49% in 2004–2008, (P < 0.001).
Prevalence of IFIs
The prevalence of various fungal pathogens identified at autopsy in patients with haematological malignancies changed significantly over the 20 years of autopsy records (Fig. 1). Aspergillus or presumed Aspergillus spp. (culture negative hyalohyphomycetes) accounted for the majority of infections during all the periods of the study, but declined after 2004 from 0.14 cases per 100 autopsies to 0.06, (P = 0.01). Similarly, the prevalence of Candida infections decreased from 0.10 cases per 100 autopsies to 0.02, but rebounded in 2004–2008 to 0.05/100 autopsies (P = 0.01). Concurrent Aspergillus and Candida infections also decreased during the study period (P = 0.02). Fusarium infections were 10–50-fold less common than Aspergillus infections and decreased from 0.008 cases per 100 autopsies to 0.003 per 100 autopsies in 2004–2008, (P = 0.08). Mucormycosis was the only mould infection whose prevalence increased over the study period, from 0.006 cases per 100 autopsies in 1989–1993 to 0.018 cases in 2004–2008 (P = 0.04).
Other fungal infections including Pneumocystis jiroveci (eight cases alone, two mixed with Candida), histoplasmosis (three cases), Cryptococcus neoformans (two cases) and phaeohyphomycosis (five cases alone, two mixed with other fungal pathogens) were detected sporadically at low rates in autopsy patients over the 20-year study period.
Aetiology of IFIs
Most mould infections reported at autopsy as aspergillosis were based on histopathology only, without definitive culture-based identification (Table 2). Among microbiologically documented infections at autopsy, the percentage of infections attributable to A. fumigatus increased over the study period, whereas infections due to other species such as A. flavus, A. terreus and A. niger decreased, although the small numbers limit analysis of the trends. Microbiologically documented Fusarium spp. infections remained relatively constant over the 20-year survey. However, cultures of Mucorales increased fourfold over the 20 year study period, (P = 0.04).
|Characteristic||1989–1993 n = 79 (%)a||1994–1998 n = 65 (%)a||1999–2003 n = 62 (%)a||2004–2008 n = 28 (%)a||χ2 trend P value|
|Culture negativeb||54 (68)||40 (62)||30 (48)||13 (46)||0.01|
|Aspergillus fumigatus||4 (5)||5 (8)||8 (13)||6 (21)||0.01|
|A. terreus||4 (5)||5 (8)||7 (11)||0 (0)||0.99|
|A. flavus||8 (10)||3 (5)||5 (8)||1 (4)||0.35|
|Otherc||1 (1)||0 (0)||1 (2)||1 (4)||0.40|
|Fusarium||4 (5)||2 (3)||4 (6)||1 (4)||0.96|
|Mucorales||4 (5)||10 (15)||7 (11)||6 (21)||0.04|
Most yeast infections (55%) during the first 5 years of the autopsy survey were based on histopathological evidence of invasion without accompanying culture information. However, histopathological identification lacking culture decreased during the study period representing only 5% of cases by 2004/2008, (P < 0.001). Among monomicrobial culture-documented infections (Table 3), C. albicans (35%) accounted for the majority of invasive infections, followed by C. glabrata (24%), C. tropicalis (15%), C. krusei (13%) and C. parapsilosis (3%). Multiple Candida infections ranged between 3% and 15% of all autopsy cases with documented yeast infection whereas non-Candida yeast and yeast-like species (i.e. Trichosporon, Rhodoturula, Saccharomyces cerevesiae) occurred in 4–10% of cases during the 20 year period. Interestingly, infections caused by Candida species with variable (C. glabrata) or non-susceptibility (C. krusei) to fluconazole decreased in the final 5 years of the study, whereas C. albicans and C. tropicalis infections increased.
|Characteristic||1989–1993 n = 66 (%)a||1994–1998 n = 32 (%)a||1999–2003 n = 29 (%)a||2004–2008 n = 19 (%)a||χ2 trend P value|
|C. albicans||11 (17)||7 (22)||8 (28)||5 (26)||0.22|
|C. glabrata||5 (8)||6 (19)||7 (24)||3 (16)||0.09|
|C. tropicalis||6 (9)||2 (6)||2 (7)||3 (16)||0.62|
|C. krusei||1 (2)||5 (16)||5 (17)||1 (5)||0.11|
|C. parapsilosis||1 (2)||0 (0)||1 (3)||1 (5)||0.30|
|C. lusitanaie||0 (0)||0 (0)||0 (0)||1 (5)||0.07|
|C. guillermondii||0 (0)||1 (3)||1 (3)||0 (0)||0.53|
|Mixed Candida||3 (5)||4 (13)||1 (3)||3 (16)||0.27|
|Histopathology only||37 (56)||7 (22)||3 (10)||1 (5)||<0.001|
|Other non-Candida yeastb||2 (4)||0 (0)||1 (3)||1 (5)||0.46|
The pattern of organ involvement by IFIs differed depending on the fungal pathogen and type of underlying immunosuppression. Candida spp. were frequently detected by both culture and histopathology in the lung (79%), blood (37%), gastrointestinal tract (35%), kidney (34%), liver (20%) and spleen (19%). Patterns of organ involvement did not differ significantly, however, among the isolated species. Patients with persistent neutropenia were more likely to have invasion of the kidney (P = 0.02) and heart (P = 0.02) compared with non-neutropenic patients. High-dose corticosteroid therapy did not appear to predispose to a specific pattern of organ involvement.
The lungs were the most common site of infection for moulds, occurring in more than 90% of all infections. Aspergillus infections most frequently affected the lung (92%), central nervous system (25%), heart (24%), kidney (15%) and gastrointestinal tract (15%). Aspergillus spp. were rarely (4%) isolated from blood cultures, and nearly all of the positive cultures were caused by A. terreus (60%) or A. flavus (40%). Compared with Aspergillus spp., Mucorales were more likely to be associated with invasion of the sinuses (23% vs. 5%, P = 0.007). Fusarium spp. were isolated frequently from the heart (63%), kidney (50%), spleen (50%) and bloodstream (40%).
We also compared patterns of organ dissemination over the study period for the four most common monomicrobial infections detected at autopsy among patients with haematological malignancies. Significant reductions in Candida dissemination to the spleen, kidney, heart, gastrointestinal tract and liver were observed over the 20 year study period, although Candida spp. dissemination to the liver rebounded back to a percentage observed in earlier periods of the study by 2004–2008 (Fig. 2). After 2003, moulds accounted for the majority of infections identified at autopsy in four of these five organs including the spleen, kidney, heart and gastrointestinal tract.
To our knowledge, this is the largest single-institution study of autopsy proven IFI in patients with haematological malignancies spanning two decades. Collectively, these autopsy data support the findings of recent epidemiological surveys that have documented a declining prevalence of IFIs and associated mortality in this high-risk population.[3, 17, 18] The improved outcomes of IFIs over the last 5–10 years have coincided with the introduction of new diagnostic tools, as well as more effective and less-toxic therapies for aspergillosis and candidiasis, which have made long-term antifungal prophylaxis and treatment more feasible. However, these trends were observed in a background of declining autopsy rates over the 20-year span of the study, consistent with the global trends of the vanishing ‘non-forensic autopsy’ in contemporary medicine.[18, 19] Multiple factors have been cited for the decline in autopsy rates, including public preferences, requirement for informed consent, concerns for limiting an institutional medical liability and the cost reimbursement for performing autopsies. Therefore, a large proportion of IFIs in the later years of our study, particularly those caused by cryptic pathogens associated with fatal outcomes, may have been under-represented in our analysis.
This study also reflects the progress achieved with an earlier diagnosis of IFIs in haematological malignancy patients. In the first 5 years of the study, 84% of the IFIs were evident only at autopsy and did not meet the European Organisation for Research and Treatment of Cancer/Mycoses Study Group criteria for ante mortem diagnosis of proven infection.[16, 20] By 2004–2008, this number had decreased to 49% of cases (P < 0.001). Improvements in ante mortem diagnosis of IFIs corresponded to the introduction of improved culture methods for fungi[21, 22] in our institution as well as the routine use of the Aspergillus ELISA galactomannan assay. However, our autopsy data also revealed that 5 of 11 (45%) patients with proven aspergillosis had repeatedly negative galactomannan test results prior to death – thus underscoring the importance of autopsy evidence for evaluating the performance of new diagnostic tests.
We also documented major shifts in the patterns of underlying immunosuppression associated with IFI in haematological malignancy patients over the 20-year study period. In the first 5 years of the study, severe neutropenia (polymorphonuclear neutrophil < 100 cells mm−3) was a predisposing condition in 90% of subjects, but declined to 44% by 2004–2008, P < 0.001. However, the use of high-dose corticosteroids increased during the study from 21% in 1989–1993, to 81% of patients in 2004–2008, P < 0.001. The shift from neutropenia to corticosteroid therapy as the predominant risk factor for IFIs in this population is consistent with the increased use of non-myeloablative conditioning for HSCT recipients, as well as targeted therapies or immunobiologicals for salvage chemotherapy in patients with haematological malignancies.[24, 25] In animal infection models and to some degree humans, the pathogenesis of invasive pulmonary aspergillosis differs considerably when infection is established in the setting of neutropenia as compared with high-dose corticosteroid therapy. In neutropenic patients, tissue necrosis is caused principally by haemorrhagic infarctions caused by unimpeded Aspergillus hyphae growth in the absence of a host immune response. Corticosteroid-treated hosts, however, are more likely to have tissue damage and necrosis caused by a defective, but exuberant inflammatory reaction to Aspergillus hyphae in the lung, which could theoretically alter classic patterns of dissemination. Therefore, the changes in the predominant underlying immunosuppression likely contributed to the changing prevalence and pattern of IFIs observed at autopsy.
Although invasive moulds continue to be the predominant IFIs in haematological malignancy patients, the prevalence of Aspergillus infections decreased substantially in the last 5 years whereas the frequency of Mucorales infections increased slightly. The increase in mucormycosis relative to aspergillosis in this population has been partially attributed to the increased use of echinocandins and voriconazole, which have good activity against Aspergillus spp., but limited or no activity against Mucorales. However, decreased early mortality due to aspergillosis may allow patients to survive longer and accumulate risk factors (i.e. hyperglycaemia, iron overload) or increased environmental exposures that may favour the development of mucormycosis.[4-6, 28] Nevertheless, the increase in mucormycosis is concerning in light of the higher mortality rates in patients infected with non-Aspergillus moulds including mucormycosis and fusariosis.
More than half of the invasive mould infections in this autopsy survey were disseminated, accounting for the involvement of almost every organ in the autopsy examination. Beyond the sinopulmonary tract and central nervous system, moulds frequently disseminated to unusual sites such as the heart, gastrointestinal tract, liver, spleen and kidneys, which are considered to be common sites for dissemination of Candida infections.[18, 29] Indeed, our data suggest that over the last 10 years of the study, moulds were a more common cause of hepatosplenic lesions and infections involving the heart and kidneys than yeast.
The changes in invasive candidiasis at autopsy over the 20 year study period mirror the changing epidemiology that has been described in multiple studies,[1, 3, 11, 30, 31] namely a decrease in disseminated and hepatosplenic infections following the introduction and widespread use of fluconazole prophylaxis in the haematologic malignancy population. Candida invasion of the lung was frequently reported at autopsy in our patients despite the rare clinical occurrence of Candida pneumonia. It is not clear whether this dissemination to the lung represents true infection represents true infection, or is an artifact of respiratory colonisation or post-mortem seeding. Even if invasion was described by histopathology, this is not necessarily diagnostic of true invasive candidiasis, which primarily seeds the organs via the bloodstream, as inflammatory damage with yeast filamentation and microinvasion can be occasionally observed in the lungs similar to the pathology observed with oesophageal candidiasis.
Interestingly, invasive infections with generally less virulent, fluconazole non-susceptible species such as C. glabrata and C. krusei decreased during the final 5 years of this study, offset by corresponding increases in C. albicans and C. tropicalis infections. This trend was consistent with culture-based surveillance studies of candidemia performed at our institution and others that identified C. tropicalis as a common Candida spp. associated with breakthrough infection in haematological malignancy patients on echinocandin therapy.[30, 33, 34]
In summary, IFIs remain a common infection in patients with haematological malignancies that are frequently disseminated and still underdiagnosed ante mortem. Although the prevalence of aspergillosis has decreased significantly over the last 5 years, non-Aspergillus moulds such as Mucorales, as well as mixed infections have remained stable or slightly increased accounting for a greater percentage of infections. Therefore, empiric or pre-emptive approaches to antifungal therapy for this population should be adapted to this changing epidemiology, as well as enhancing efforts towards their earlier ante mortem diagnosis through molecular methods. Finally, it is important to reverse the declining trend of medical autopsy, or we risk losing one of our most important definitive tools for understanding the epidemiology of fungal disease in this highly vulnerable population.
No financial support was sought for this study. None of the authors have disclosures or potential conflicts of interest related to this work. Dimitrios Kontoyiannis wishes to acknowledge his support through the Francis King Black Endowed Professorship.
- 16Revised definitions of invasive fungal disease from 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 (EORTC/MSG) Consensus Group. Clin Infect Dis 2008; 46: 1813–21., , et al.
- 30Candidemia in patients with hematologic malignancies in the era of new antifungal agents (2001–2007): stable incidence but changing epidemiology of a still frequently lethal infection. Cancer 2009; 115: 4745–52., , et al.