Disclaimer: Early release articles are not considered as final versions. Any changes will be reflected in the online version in the month the article is officially released.
Author affiliation: City St. George’s Hospital and St. George’s University, London, UK (L. Brown); Boehringer Ingelheim Pharmaceuticals, Inc., Ridgefield, Connecticut, USA (B. Jonat); Tropical Medicine Research Center for Talaromycosis, Pham Ngoc Thach University of Medicine, Ho Chi Minh City, Vietnam (V.T. Ly, N.T. Hoa, T. Le); Hospital for Tropical Diseases, Ho Chi Minh City (V.T. Ly, N.L.N. Tung); Oxford University Clinical Research Unit, Ho Chi Minh City (P.S. Lam, N.T. Thanh, N.T. Hoa); National Hospital for Tropical Diseases, Hanoi, Vietnam (D.T.N. Bich); Saskatchewan Cancer Agency, Regina, Saskatchewan, Canada (V.P. Thao); Duke University Medical Center, Durham, North Carolina, USA (N.T.M. Thu, T. Le)
Talaromycosis (formerly penicilliosis) is an invasive fungal disease caused by Talaromyces marneffei, a dimorphic fungus endemic to Southeast Asia. First discovered in captive wild bamboo rats (Rhizomys sinensis) in Vietnam in 1956, talaromycosis remained rare in humans until the onset of the HIV epidemic in the 1980s, when it rapidly emerged as a leading HIV-associated opportunistic infection and cause of HIV-associated death in Southeast Asia (1,2). The highest reported disease burden is in northern Thailand, Vietnam, and southern China, where talaromycosis accounts for 4%–20% of HIV-related hospital admissions and the mortality rate, despite antifungal treatment, is between 15% and 30% (2–7). Outside of those hyperendemic regions, talaromycosis is likely underdiagnosed, and the true burden of disease across Southeast Asia remains unknown (7,8). Although most cases (≈90%) are associated with advanced HIV disease, the incidence of talaromycosis is rising among persons with other immunocompromising conditions, such as primary immunodeficiency, malignancy, solid organ or stem cell transplant, and autoimmune or inflammatory conditions that require immunomodulating therapy (9–11). Cases are increasingly reported outside of talaromycosis-endemic regions because of increased migration and international travel (8).
A lack of understanding of disease reservoirs and exposure risk factors for talaromycosis has hampered disease prevention. The bamboo rat is the only known enzootic reservoir of T. marneffei. Surveys of rodents in talaromycosis-endemic regions have identified all 4 species of bamboo rat in Southeast Asia (R. sinensis, R. pruinosis, R. sumatrensis, and Cannomys badius) (Appendix Figure 1) as asymptomatic carriers of T. marneffei; reported prevalence ranges from 10% to 100% (12–15). The geographic distribution of bamboo rats follows the distribution of human talaromycosis (Appendix Figure 2), and T. marneffei isolates from bamboo rats have been shown to share similar genotypes to those infecting humans in the same region (16,17), suggesting the potential for bamboo rat–to–human transmission. However, epidemiologic evidence of direct bamboo rat–to–human transmission is lacking. A case–control study conducted in Chiang Mai, a rural and mountainous region of Thailand where bamboo rats are endemic, demonstrated no link between exposure or consumption of bamboo rats and human talaromycosis. Instead, agricultural work and soil exposure, particularly during the rainy season, were independent risk factors for talaromycosis (18). Because T. marneffei has been isolated more frequently in soil samples collected from bamboo rat burrows than from non–bamboo rat habitats (16,19), we hypothesize that exposure to soil associated with bamboo rats could be the primary driving factor for talaromycosis, rather than direct exposure to bamboo rats.
Aside from soil exposure, weather factors have a substantial effect on talaromycosis incidence. During the monsoon season, the incidence of talaromycosis rises 30% in Vietnam, 50% in Thailand, and 73% in southern China from that observed during the dry season and has been associated with increased temperatures and humidity, which provide favorable environmental conditions for T. marneffei (4,6,20,21). Talaromycosis is also known to occur in geographic hot spots in mountainous areas of Southeast Asia, such as the northern provinces of Chiang Mai and Chiang Rai in Thailand, the southern provinces of Guangdong and Guangxi in China, and the northeastern states of Manipur and Assam in India (22). Vietnam presents a unique opportunity to study the geographic and exposure risk factors for talaromycosis because talaromycosis occurs across diverse climatic and geographic ranges, from the temperate, subtropical regions in the north to the tropical monsoon regions in the south (Figure, panel A). Three quarters of the land in Vietnam is composed of highlands (defined as hills and mountains >400 m above sea level), which include the Northeast and Northwest regions in the north and the Central Highland region in the south. The remaining one quarter of the land consists of lowlands, which include the Red River Delta in the north, the coastline, and the Mekong Delta in the south. Talaromycosis is seen in both rural and urban settings in Vietnam, where risk factors for infection might be very different than in the rural settings of Chiang Mai, Thailand (18). This study aimed to evaluate the behavioral, environmental, and geographic risk factors for talaromycosis in the diverse climatic and geographic setting of the hyperendemic country of Vietnam.
Study Design and Participants
We conducted a multicenter, matched, case–control study recruiting participants with advanced HIV disease from the inpatient and outpatient departments of the National Hospital for Tropical Diseases (Hanoi, Vietnam) and the Hospital for Tropical Diseases (Ho Chi Minh City, Vietnam), during January 2013–July 2016. Those hospitals were estimated to receive 40% of all HIV inpatient referrals in Vietnam, which has a catchment population of 72 million. We predetermined risk factors of interest (n = 13) after a comprehensive literature review and included behavioral, environmental, and geographic risk factors (Table 1).
We consecutively recruited participants from a population of adults with HIV >18 years of age with a CD4 cell count <100 cells/µL who had culture-proven talaromycosis (defined as a compatible clinical syndrome and isolation of T. marneffei in blood, skin lesion, bone marrow, lymph node, or other specimens as clinically indicated). We consecutively recruited control-patients who came from the same at-risk population but had no clinical suspicion for or microbiological evidence of talaromycosis. To reduce the potential confounding effect, we matched cases to controls at a 1:2 ratio according to age (+5 years), sex, and CD4 cell count +50 cells/µL. When a CD4 count was not performed or available, we matched participants by absolute lymphocyte count (+330 cells/µL, calculated on the basis of an assumed 15% CD4-positive lymphocyte count, 50/0.15 = 330) or HIV disease at stage 3 or 4 according to World Health Organization (WHO) clinical staging. We recruited control-patients within the same week as the corresponding case-patients from the same site.
Sample Size Estimation
We calculated the sample size estimate for the matched case–control study according to Parker and Bregman (23). The target sample size of 200 in the case group and 400 in the control group was chosen to allow 90% power to detect an odds ratio (OR) of >1.9 for a risk factor, assuming a proportion of risk exposure of 0.2.
Data Collection
After participant recruitment, trained study nurses conducted face-to-face interviews using a standardized questionnaire of prespecified exposure variables (Table 1). To control for assessment bias, study nurses were blinded to the assignment of case and control and the study hypotheses. A photograph of the 4 species of bamboo rats (Appendix Figure 1) was shown to all participants to ensure that the history of exposure was correctly elicited. We assigned participants to 8 municipal regions based on current residential address. We categorized data from Ho Chi Minh City as a separate region because of the high number of participants recruited within the city limits. We performed a posthoc follow-up of participants residing in lowland regions (Ho Chi Minh City and the Mekong Delta) to investigate the risk posed by previous travel to or previous residence in highland regions that had been identified through the geographic mapping to be talaromycosis risk regions. The posthoc follow-up consisted of phone interviews conducted by research nurses who were blinded to the participant status as case or control.
Statistical Analysis
We used univariate and multivariable conditional logistic regressions to evaluate risk factors for talaromycosis using a complete case analysis. We included all 13 predefined exposure variables (Table 1) in the regression model. The Central Highlands and some provinces in the adjacent Southeast and South Central Coast regions in southern Vietnam consist of hills and mountainous terrain >400 meters in elevation (Figure, panel A). To investigate whether participants residing in those highland regions were at greater risk for talaromycosis than were residents of the lowland regions of Ho Chi Minh City and the Mekong Delta, we performed conditional logistic regression with the region as the only covariate. We adjusted pairwise comparisons between regions for multiple comparisons using a parametric single-step method (24). Participants recruited in northern Vietnam were excluded from this geographic analysis because of the later start date, which meant a substantially smaller number of patients were recruited with too diverse geographic distribution to enable robust statistical analysis in northern regions. We estimated the number of talaromycosis cases recruited per total HIV population in each region using HIV prevalence data obtained from the Vietnam Ministry of Health Administration of HIV/AIDS Control from 2007 to evaluate for any bias in referral pattern (25).
We visualized maps of the number of cases, controls, and case-to-control ratios in different geographic regions in southern Vietnam using the Quantum Geographic Information System version 1.5 (https://qgis.org). We performed all statistical analyses by using R version 3.2.1 (The R Project for Statistical Computing, https://www.r-project.org).
Ethics
This case–control study was a substudy of the Itraconazole and Amphotericin B for Talaromycosis clinical trial that was approved by the Vietnam Ministry of Health (protocol 781-Vietnam MOH), by the Oxford University Tropical Research Ethics Committee (protocol OxTREC 12–09), and by the ethical and scientific committees of the 2 study sites in Vietnam. All participants provided written consent before study enrollment.
Characteristics of Study Participants
During January 2013–July 2016, a total of 236 cases of talaromycosis were diagnosed across the 2 centers, of which 205 cases were recruited into the study and matched to 405 controls. Of the 31 cases screened but not recruited (5.1%), 25 died or were discharged before the talaromycosis diagnosis was confirmed (80.6%), and the remaining 6 declined to participate. Participants were mostly men (75% [456/610]), consistent with the HIV-infected population in Vietnam. Median age was 34 years (interquartile range [IQR] 31–38 years), and median CD4 count was 17 cells/μL (IQR 7–36). Case-patients and control-patients were similar in age, sex, and absolute lymphocyte counts (Table 2). CD4 count was significantly lower in case-patients than in control-patients. According to WHO HIV disease staging, all case-patients were classified as stage 4 (severe), whereas control-patients were mostly stage 3 or 4 (moderate or severe). Most participants were inpatient (100% [205/205] of case-patients and 91% [369/405] of control-patients). Other opportunistic infections, most commonly tuberculosis, pneumonia (caused by bacteria or Pneumocystis), oral/esophageal candidiasis, or cryptococcosis, were diagnosed in controls. Concurrent opportunistic infections were common among case-patients with talaromycosis (Table 2).
Analysis of Behavior and Environmental Risk Factors
Data were 100% complete for all 610 participants for 12 of 13 covariables (Table 1). Data were incomplete for only 1 area, fluconazole prophylaxis, in which data were missing for 14 (2%) participants (Table 3). In the univariate analysis, patients with talaromycosis were significantly more likely to work in outdoor settings, live within 100 meters of or have direct occupational contact with a tropical plant (bamboo, sugar cane, or rice), or have direct occupational contact with a highland plant (rubber, tea, coffee) and farm animals (cattle, swine, poultry). In the multivariable analysis, exposure to tropical plants (OR 1.73 [95% CI 1.10–2.73]; p = 0.017) and exposure to farm animals (OR 2.07 [95% CI 1.20–3.55]; p = 0.009) were the only independent risk factors for talaromycosis (Table 3).
Mapping of Cases and Controls and Geographic Risk Analysis
Geographic data were available for 204 (99.5%) of 205 case-patients and 403 (99.5%) of 405 control-patients (Figure, panels B, C). We performed mapping and geographic analysis only for southern Vietnam because recruitment in southern Vietnam began earlier and more case-patients (86% [175/205]) and control-patients (86% [348/405]) were enrolled in southern Vietnam to provide robust data for these analyses. The case-to-control ratio demonstrated regions with a higher or lower than expected number of cases relative to controls; 0.5 was the expected case-to-control ratio, given that 2 controls were recruited for each case (Figure, panel D). We found that case-to-control ratios were highest in the Central Highlands (2.18), followed by the surrounding provinces in the South-Central Coast (1.20) and the Southeast region (0.87). Case-to-control ratios were the lowest in Ho Chi Minh City (0.32) and the Mekong Delta (0.25), where the number of persons recruited was the highest (Figure, panel D). Conditional logistic regression demonstrated that participants in the highland regions in the Central Highlands, Southeast, and South-Central Coast were significantly more likely to develop talaromycosis than those residing in the lowlands of Ho Chi Minh City and Mekong Delta (multiplicity-corrected p<0.05 for all pairwise comparisons) (Table 4). The point estimate of the risk for talaromycosis was higher in the Central Highlands than in the Southeast lowland region, suggesting a risk exposure relationship for closer proximity to the highest risk Central Highland region, but this difference did not reach statistical significance (OR 3.32; p = 0.06). In the southern regions of Vietnam, the highest relative number of talaromycosis cases was in the Central Highlands (Table 5).
To investigate whether previous travel to or residence in any of the 3 identified risk regions was associated with the risk for talaromycosis, we performed a posthoc follow-up interview of participants residing in the lowlands of Ho Chi Minh City and the Mekong Delta. A total of 69 (90%) of 77 case-patients and 82 (32%) of 254 control-patients responded to the blinded posthoc follow-up phone interview. Participants who had lived or traveled to a high-risk region for >3 days were at significantly higher risk for talaromycosis (55/69 [81%] case-patients vs. 47/82 [57%] control-patients; OR 2.90 [95% CI 1.33–6.59]; p = 0.005).
This large case–control study investigated the behavioral, environmental, and geographic exposure risk factors for talaromycosis in susceptible persons living in a highly endemic region, Vietnam. Participants living within 100 meters of or having direct exposure to tropical plants (bamboo, sugar cane, or rice) or to farm animals (cattle, swine, or poultry) had 2-fold higher odds for talaromycosis. Our findings are consistent with the results of a previous smaller case–control study from Chiang Mai, Thailand (n = 180), which also found a 2-fold increase in the odds of talaromycosis with recent exposure to animals or plants (18). The strength of those associations is also similar across the 2 studies, despite the difference in study settings (urban vs. rural) and an 8-fold larger sample size in our study. Although direct exposure to soil was not identified as a risk factor in our study, soil is a known reservoir for other dimorphic fungal pathogens including Coccidioides spp., Paracoccidioides spp., Histoplasma capsulatum, and Sporothrix schenckii (26–28). Specifically, soil contaminated with animal droppings (bird or bat) is a known reservoir for Histoplasma spp., whereas soil mixed with decaying plant materials has been identified as a reservoir for Sporothrix spp. (29,30). Soil contaminated with farmed animal excreta or decaying tropical plant substrate is likely also favorable to T. marneffei’s growth, which supports the observed risk for talaromycosis in persons exposed to or living in proximity to animal and agricultural farming activities.
Direct contact with or consumption of bamboo rats was not associated with talaromycosis in our study, which is consistent with the previous case–control study from Thailand (18). Early efforts to isolate T. marneffei from the soil within bamboo rat burrows yielded some success (14,31,32). An environmental survey in Guangdong Province, China, demonstrated a higher prevalence of T. marneffei in soil collected from bamboo rat burrows (8.2% [15/184]) than in soil from sites not associated with bamboo rats (2.0% [1/50]) (19). T. marneffei was isolated in bamboo rat stool on the surface of the soil and around the bamboo roots found deep in the bamboo rat burrows (19). In Thailand, T. marneffei DNA was found in soil around a bat cave and elephant camp (33). However, the bamboo rat is the only known nonhuman host of T. marneffei (16,19). T. marneffei is found in the liver, spleen, and lungs of healthy-appearing bamboo rats, suggesting that bamboo rats are asymptomatic carriers of T. marneffei and serve as a zoonotic reservoir (19). Bamboo rats construct extensive burrow systems for foraging among bamboo crops by digging 1–2 meters into the ground. Although their primary food source is bamboo roots and shoots, bamboo rats are known to feed on sugar cane and cassava roots. In our study, exposure to tropical plants was an independent risk factor for talaromycosis, possibly because of their association with the bamboo rat environment. By burrowing into the soil and feeding on a variety of tropical vegetation, bamboo rats likely enhance T. marneffei transmission. Soil disturbance, caused by bamboo rat foraging, heavy rainfall, or agricultural activity (e.g., tillage, plowing, livestock grazing), causes aerosolization of T. marneffei conidia, the infective form inhaled by bamboo rats and humans (4,6,20,21). The findings of our study suggest that exposure to soil associated with bamboo rats (and possibly other animals) and activities that disturb soil are the main drivers for the acquisition of talaromycosis, rather than direct zoonotic transmission.
Geographic mapping was very informative and demonstrated that the Central Highlands and the adjacent Southeast and South-Central Coast regions are high-risk regions for talaromycosis, compared with the lowlands of Ho Chi Minh City and the Mekong Delta regions in Vietnam. Assuming the referral patterns of patients with HIV were similar among regions, talaromycosis cases per 100,000 persons with HIV were 8–10 times higher in the Southeast and Central Highlands regions than in the Mekong Delta and were 2–3 times higher than in Ho Chi Minh City. Highland regions are the known natural habitat of bamboo rats, and previous studies outside of Vietnam have demonstrated that human talaromycosis follows the geographic distribution of bamboo rats (16,17). Further interrogation of participants from the lowlands of Ho Chi Minh City and Mekong Delta revealed that previous residence in or travel to highland regions increased the odds of talaromycosis 3-fold. Our study demonstrates an epidemiologic link between human talaromycosis and residence in or travel to highland regions in a highly endemic country. Further research is needed to determine whether localized geographic hotspots of talaromycosis exist within those highland regions. Our data can inform prevention strategies for talaromycosis, including patient education and the consideration of primary itraconazole prophylaxis in high-risk persons considering travel to or living in regional hotspots.
The first limitation of our study is that 31 persons with talaromycosis were screened but not enrolled in our study, largely because of discharge or death. Those participants might have exhibited different risk factors from those enrolled in the study, which is a potential source of selection bias; however, the number is very small (5.1%). Second, control-patients who had concurrent bacteremia could potentially have been misclassified because bacteria can outgrow T. marneffei in culture, leading to false-negative results. CD4 count was only available for 30% of participants, and therefore matching was performed on the basis of absolute lymphocyte count and WHO HIV disease staging, which are suboptimal measures of immune susceptibility. Some characteristics of case-patients and control-patients, such as median CD4 cell count, WHO HIV stage, and proportion of inpatients, differed significantly. However, our control population consisted of clinically relevant persons with advanced HIV disease (CD4 <200 cells/mL) and other opportunistic infections, which are representative of the population at risk for talaromycosis. Although interviewers were blinded to the classification of case and control, observer bias is possible, because 40%–70% of persons with talaromycosis display characteristic umbilicated skin lesions (6,34,35). However, that possibility was mitigated because interviewers were blinded to study hypotheses to minimize observer bias. The differential referral pattern between cases and controls in different regions might have skewed the geographic results, but that skewing was unavoidable because of limited diagnostic capacities outside of major referral centers in Vietnam. The number of cases recruited in Ho Chi Minh City and Hanoi was imbalanced because enrollment started earlier in Ho Chi Minh City, resulting in overrepresentation of residents from southern Vietnam. Finally, potential for recall bias is inherent to a case–control study design. Further studies are needed to uncover more specific geographic risk factors and risk behaviors and to establish the causal link between environmental exposure and the development of talaromycosis.
In conclusion, this study demonstrates an epidemiologic link between human talaromycosis and geospatial proximity or travel to highland regions in a hyperendemic country, thus informing disease education and prevention strategies. Results from this large case–control study validate the previous findings that disturbance of soil associated with tropical plants and farmed animals, such as through agricultural or construction activity, increases the risk for talaromycosis.
We thank Noel Weiss for his advice on the study design. We thank the patients for their participation in this study. We thank the physicians at the study sites for their assistance with study recruitment: at the Hospital for Tropical Diseases in Ho Chi Minh City, Nguyen Huu Chi, Ngo Thi Kim Cuc, Nguyen Le Nhu Tung, Nguyen Thanh Liem, Le Duc Vinh, and Ly Quoc Cong; at the National Hospital for Tropical Diseases in Hanoi, Nguyen Tien Lam, Ta Thi Dieu Ngan, and Nguyen Kim Thu. We thank nurses Phan Thi Hong Dao and Hoang Suong Nguyet Anh for their assistance with the interviews. We thank Tran Phuong Thuy for the coordination of the project and Vo Thi Hoang Anh for assistance with Appendix Figure 1 and for obtaining photographers’ permission for the use of the bamboo rat photos. Finally, we thank Pierce Gardner and Marc Foca for their mentorship of Brian Jonat on the project.
This study was jointly funded by the University of Washington Center for AIDS Research (National Institutes of Health grant P30 AI027757 through a pilot international award to T.L.), by the Stony Brook Medicine International Research Fellowship (to B.J.), and by the National Institutes of Health (grants R01 AI181764-01 to T.L., R01AI177098-02, R01-AI143409-05, and U01AI169358-03 to N.T.H. and T.L.). L.B. is funded by the National Institute for Health Research and received fellowship grants from the Federation of European Microbiological Societies, Dowager Countess Eleanor Peel Trust, the British Society for Medical Mycology, and the Wingate Foundation.
T.L. conceptualized the study, acquired funding, and constructed the methodology. Data were curated by T.L., V.T.L., P.S.L., N.T.T., and D.T.N.B. Formal analysis and visualization were performed by T.L., B.J., and L.B. B.J. and L.B. were project administrators. T.L. and N.T.H. provided supervision. B.J. and L.B. wrote the original draft. V.T.L., P.S.L., N.T.T., D.N.T.B, V.P.T., N.T.M.T., N.T.H., and T.L. reviewed and edited the manuscript. All authors contributed to the revision of the paper for intellectual content and approved the final manuscript.
