In August 2017, hurricane Harvey and historic scale of flooding devastated the Houston metropolitan area and adjacent counties. A survey amongst immunocompromised Houston area residents revealed that almost half of them engaged in home clean-up and mould remediation activities, often with no or suboptimal personal protective equipment.¹ Although this observation raised the concern of extensive mould exposure of patients at risk for invasive mould infections (IMIs), prior research by our group at the University of Texas MD Anderson Cancer Center (MDACC) found no institution-wide increase in culture-documented IMIs after the hurricane.² As an increased use of voriconazole and amphotericin B was seen at MDACC in the 12 month- period following the hurricane,² there might have been a lower threshold for initiation of mould-active antifungal treatment or prophylaxis in high-risk patients and/or an increased incidence of infection events not meeting the conventional IMI definitions.

In order to provide a more comprehensive characterization of hurricane Harvey’s impact on the IMI landscape, the U.S. Centers for Disease Control and Prevention (CDC) developed a modified clinical IMI (mcIMI) case definition based on an expanded set of host and clinical features combined with therapeutic-intent antifungal drug prescription. We herein applied this less-restrictive mcIMI case definition specifically to MDACC patients residing in Harvey-affected Texas counties.

Four-hundred-and-four MDACC patients with mould-positive cultures between September 2016 (12 months pre-Harvey) and August 2018 (12 months post-Harvey) were identified using the institutional microbiology laboratory information system (Supplementary Data Set). A single mould genus was isolated from a single material in 329 out of these 404 patients. Aspergillus was the most commonly identified genus (n = 153, including 10 patients with multiple Aspergillus species). Nineteen patients had a single genus isolated from multiple respiratory samples and 8 patients had the same mould/genus isolated from multiple sites or materials. Thirty-one patients had two or more genera isolated from a single site or material. A total of 17 patients had multiple mould genera recovered from multiple respiratory samples or multiple sampling sites (Supplementary Data Set).

One-hundred-and-seven out of the 404 unique cases represented patients living in Harvey-affected areas ( Fig. 2). Four out of these 107 patients were excluded from further analysis after chart review. Three patients had a mould-positive culture within the study period, but their finally determined DOI was more than 12 months prior to hurricane Harvey. In addition, one case of “sterile hyphae” recovered from a skin lesion was later identified as a proven yeast infection with no evidence of an IMI event and was excluded from analysis (Fig. 2). After exclusions, 103 cases remained in the final analysis.

Notably, these 103 cases were not distributed evenly between the pre- and post-Harvey period. Instead, the number of patients with mould-positive cultures from Harvey-affected counties almost doubled from 36 pre-Harvey to 67 post-Harvey (Fig. 2), resulting in a significantly increased incidence of positive cultures after the hurricane (Fig. 3A, p < 0.01). Thirty-four out of the 67 patients with mould-positive cultures post-Harvey (51%) were adjudicated as having probable/proven IMIs (n = 11) or mcIMI (n = 23), compared to 23 (mc)IMI cases pre-Harvey (9 probable/proven IMIs and 14 mcIMI cases). The difference in incidence rates of probable/proven IMIs or mcIMI cases pre- and post-Harvey did not reach significance (Fig. 3B, p = 0.15– 0.18), whereas the incidence rate of patients with mould-positive cultures not meeting the IMI criteria significantly increased post-Harvey (Fig. 3C, p < 0.01). Of note, 76% of mould-positive cultures in patients not meeting the IMI criteria were obtained from respiratory samples, suggesting increased asymptomatic colonization of respiratory epithelia.

Patients with probable/proven IMIs were more likely to have active leukaemia/myelodysplastic syndrome (p < 0.001) and/or severe neutropenia < 500/μL (p < 0.001) and to receive corticosteroids (p < 0.01) or other immunosuppressive therapies (p < 0.001) compared with mcIMI cases and patients not meeting the IMI definitions (Table 2). Patients with proven/probable IMIs had higher rates of extrapulmonary or disseminated mycoses and had a higher proportion of Mucorales or Fusarium spp. recovered as the causative agent than patients with mcIMIs (Table 2). Compared to patients not meeting the IMI definitions, a higher proportion of patients with probable/proven IMIs or mcIMIs received systemic antifungal therapy, such as liposomal amphotericin B or broad-spectrum triazoles (p < 0.001) (Table 2). Similarly, patients meeting at least the mcIMI definition more often required hospitalization (p < 0.001) or ICU admission (p = 0.03) and had higher in-hospital mortality (p = 0.05) than patients with mould-positive cultures but no clinical correlate (Table 2).

Comparing all patients from Harvey-affected counties with mould-positive cultures before and after the hurricane, no significant differences in demographics, predisposing factors, and outcomes were seen, except for a higher percentage of patients with a history of lymphopenia in the post-Harvey cohort (Table S1, p < 0.01). Restricting the comparison to patients with IMI or mcIMI events, no significant differences in baseline characteristics and predisposing factors were found between the pre- and post-Harvey cohorts (Table 3). Nonetheless, the percentage of hospitalized patients requiring ICU admission in the course of their (mc)IMI treatment increased from 30% pre-Harvey to 56% post-Harvey (Table 3, p = 0.06). Likewise, the percentage of patients with (mc)IMI dying in-hospital rose from 17% pre- Harvey to 50% (p = 0.01), whereas 42-day mortality (30% pre-Harvey versus 45% post-Harvey, p = 0.26) and survival curves (log-rank test, p = 0.18) did not significantly differ depending on the DOI (Fig. 4).

Considering all patients regardless of their place of residence, our previous study, using the conventional and rather restrictive EORTC/MSG diagnostic criteria, revealed no significant changes to the institutional mould infection landscape after hurricane Harvey.² However, institution-wide data, including patients from non-affected areas, may “dilute” the trends and lack the granularity to identify subtle changes in IMI epidemiology in patient from areas impacted by floodwater damage. Therefore, the present study was uniquely restricted to patients from Harvey-affected counties, that is, counties qualifying for disaster assistance.³ In this cohort, we indeed found a significantly increased number of mould-positive cultures in the year following the hurricane compared with pre-Harvey data. However, although the incidence rates of proven/probable IMIs and mcIMIs slightly increased after the hurricane, this trend did not reach statistical significance. Instead, 49% of the mould-positive cultures in patients from Harvey-affected counties were not associated with clinical correlates meeting either the conventional EORTC/MSG definition or the mcIMI definition. As most mould-positive cultures without a clinical correlate were obtained from respiratory materials (76%), we hypothesize that the increase in mould-positive cultures post-Harvey is primarily due to asymptomatic colonization of respiratory epithelia.

Historic data providing a clear link between residential mould exposure in post-disaster settings, airway colonization, and IMI events are scarce. Despite high levels of mould infestation immediately following hurricanes Katrina and Rita in 2005,^5–6 there has been no evidence of elevated IMI incidence rates in exposed patient cohorts;^6–7 however, transient asymptomatic airway colonization with Mucorales was seen in some residents of floodwater-damaged buildings.⁸ A related observation was made for Aspergillus- and Basidiomycetes-positive sputum cultures after a tsunami in East Japan.⁹ In contrast, our breakdown of positive cultures did not reveal major shifts or a selective predominance of causative genera post-Harvey (Fig. 3A), which might be due to the much larger catchment area of our patients compared to the cited studies.

Of note, floodwater-damaged buildings can remain a source of increased exposure to pathogenic moulds even after mould remediation activities.¹⁰ Therefore, long-term surveillance programs are warranted in Harvey-affected areas for both, IMIs in immunocom-promised populations and non-infectious respiratory hypersensitivity syndromes (e. g., mould-associated asthma) that were seen after previous geo-meteorological disasters.⁷ As discussed previously,² there might also be a risk for delayed emergence of unusual mould pathogens such as the dimorphic fungus Coccidioides immitis in flooded areas.¹¹

In addition to increased numbers of mould-positive cultures, we observed a signal of worse outcomes in patients from Harvey-affected counties developing (mc)IMIs after the hurricane. While 42-day all-cause mortality did not significantly differ depending on the DOI, patients developing (mc)IMIs post-Harvey had significantly higher in-hospital mortality and tended to have higher ICU admission rates than patients with IMIs pre-Harvey. Although univariate analyses ruled out a significant impact of many important confounders (e. g., causative pathogens, sites of infection, cytopenia, underlying cancer diagnoses, and immunosuppressive therapies), the power of these analyses was limited, and meaningful multivariate analysis was not feasible due to the small sample size. Nonetheless, the observed trend toward worse outcomes in patients developing IMIs after hurricane Harvey is intriguing and the many dynamic and interrelated factors that could contribute to this observation deserve further study. One the one hand, it would be conceivable that increased colonization driven by extensive exposures increases the risk for severe IMI manifestations due to the high fungal burden.¹² On the other hand, residential exposure to moulds commonly found after water intrusion can trigger alterations in mould-reactive immune responses, especially elevated type–2 T-helper cell responses^13–14 that are considered non-protective and might contribute to immune pathology.¹⁵ Furthermore, floodwater-damaged housing can be a reservoir of moulds producing mycotoxins¹⁰ that have immunosuppressive properties and were shown to modulate host responses to invasive infection;¹⁶ however, the clinical significance of this hypothesis remains to be established.

Our retrospective monocentric study has several limitations. While the FEMA assistance level³ provides an at-large surrogate of a county’s devastation by the hurricane and subsequent flooding, our study design did not facilitate correlation of the patients’ individual risk for mould exposure (e.g., participation in mould remediation or home reconstruction activities) and the incidence of mould-positive cultures or IMIs. Furthermore, the denominators used for incidence density calculations (Fig. 2) were based on the institution-wide patient census since data restricted to patients living in Harvey-affected counties were not available. However, as there was no evidence for a significant shift in institutional patient catchment areas during the 2-year study period, this limitation likely has a minor impact on the validity of our analyses and conclusions. Similarly, the institutional laboratory information system did not facilitate a determination of the total number of mould cultures ordered from patients residing in Harvey-affected counties. We have previously reported that neither the number of mould cultures ordered institution-wide nor their positivity rate were significantly different before and after the hurricane.² However, in the absence of a culture census for patients from affected counties, our data do not allow us to determine whether the significantly higher post-Harvey incidence of positive cultures in patients from affected areas is driven by an increased number of cultures ordered, an increased culture positivity rate, or a combination of both. Furthermore, the relevance of individual pathogens as colonizers versus contaminants can be difficult to distinguish in the absence of a clinical correlate. For example, Aspergillus niger is known as a common colonizer of respiratory epithelia¹⁷ and it is also a common contaminant at the MDACC Microbiology Laboratory. In addition, the uncommon saprophytic moulds are common colonizers and rarely true pathogens, even in high-risk cancer patients.¹⁸ In order to examine these potential confounders in a sufficiently powered analysis and to evaluate the generalizability of our findings to other patient populations including patients with non-cancer-related predisposing factors for IMIs (e. g., patients with metabolic disorders such as diabetes mellitus), multi-centre data would be needed. Similarly, the mcIMI definition itself remains to be studied in multi-centre settings.

In summary, despite limitations, our unique study provides significant insights into the epidemiology of mould-positive cultures and IMI events after a devastating hurricane causing widespread flooding. Employing both, conventional EORTC/MSG definitions and a broader mcIMI case definition that considers therapeutic-intent antifungal drug prescription, our results corroborate the previously published observation that hurricane Harvey did not cause significant changes in IMI incidence and aetiological mould genera at MDACC.² The increased recovery of moulds from – predominantly respiratory – cultures in patients living in Harvey-affected counties likely reflects increased airway colonization and points to a need for long-term surveillance efforts, including non-infectious mould-associated diseases.¹⁹ Lastly, we found that increased detection of moulds was a marker of poor outcomes of IMI events in patients from Harvey-affected counties. Altogether, these results emphasize the importance of risk awareness, enhanced mould prevention strategies,²⁰ and improved clinical management of IMIs in high-risk patients after geo-meteorological disasters.