Background
In December 1994 and January 1997, articles in MMWR described a cluster of 10* infants from Cleveland, Ohio, with acute idiopathic pulmonary hemorrhage, also referred to as pulmonary hemosiderosis (1,2). The children resided in seven contiguous postal tracts and had had one or more hemorrhagic episodes, resulting in one death, during January 1993-December 1994. Preliminary results of a CDC case-control study (2) indicated that hemorrhage was associated with 1) major household water damage during the 6 months before illness and 2) increased levels of measurable household fungi, including the toxin-producing mold S. chartarum (syn. S. atra).

These findings and the observation that tricothecene mycotoxins were produced in the laboratory by some S. chartarum isolates recovered from the homes of study subjects have been published and referenced in peer-reviewed scientific literature (3-9). The hypothesis from the findings of the investigation was that infant pulmonary hemorrhage may be caused by exposure to potent mycotoxins produced by S. chartarum or other fungi growing in moist household environments (4,5). The findings also were cited in environmental health guidelines (10,11), congressional testimony (12), and the popular media (13-16), and have been debated among industrial hygienists and other occupational and environmental health scientists (17-21). Despite caution that "further research is needed to determine...causal[ity] (4)," the findings have influenced closure of public buildings, cleanup and remediation, and litigation (16,22-28).

In June 1997, a CDC scientific task force, in a review of the agency's response to the problem, advised the CDC director that concerns about the role of S. chartarum in pulmonary hemorrhage needed to be addressed. In response, CDC convened a multidisciplinary internal group of senior scientists (working group) and sought the individual opinions of outside experts. The working group and the outside experts conducted separate reviews of the Cleveland investigation. The working group reviewed background literature, internal CDC documents, and published CDC reports; examined the data set; and interviewed the principal investigators. The external experts reviewed relevant literature, including internal CDC documents and the working group report, and invited additional consultants to address specific topics. The working group and the external consultants each concluded that further work is needed to better describe the clinical problem, its public health impact, and the factors that put infants at risk (29,30).

Case Identification
The reviewers had concerns about the characterization of the clinical problem as "hemosiderosis." The acute presentation in all 10 cases, the narrow age distribution (6 weeks to 6 months), and the absence of iron deficiency suggest that the illness described in the cluster of cases in Cleveland (1,3) is clinically distinct from idiopathic pulmonary hemosiderosis (IPH), the condition to which this cluster was linked (31). Hemosiderosis (i.e., hemosiderin-laden macrophages in the interstitium and alveolar spaces of the lung) is a pathologic finding indicative of pulmonary bleeding of any type, not a unique characteristic of a specific disease, etiology, or pathophysiologic process (32,33). Therefore, in referring to the cluster of cases in Cleveland, the working group defined that cluster as AIPH in infants. From the limited clinical and historic information available to the reviewers on cases added to the Cleveland series since the original cluster (D. Dearborn, Case Western Reserve Department of Pediatrics, personal communication, September 1999), the external consultants concluded that some of these additional cases (6), including several identified in a retrospective review of sudden infant death syndrome cases (2), do not conform to the clinical patterns of cases in the original cluster. Both groups of reviewers recognized limitations that precluded drawing conclusions about clinical or etiologic ties to IPH.

Association of Household AIPH with Water Damage and Fungi
Both groups of reviewers concluded that the available evidence does not substantiate the reported epidemiologic associations---between household water damage and AIPH (3) or between household fungi and AIPH (4)---or any inferences regarding causality. The interpretation of water damage and its association with AIPH was considered to have been hampered by the limited descriptive information, by the lack of standard criteria for water damage, and by the absence of a standard protocol for inspecting and recording information from home to home. Similarly, assessment of exposure to fungi or mycotoxin also was difficult to interpret because the methods did not distinguish between contamination and clinically meaningful exposure. No isolates or serologic evidence of exposure to fungi or mycotoxin were obtained in individual case-infants.

Evaluation of Analysis Methods
Three factors, considered together, contributed to the groups' conclusions that S. chartarum was not clearly associated with AIPH:

1. The working group found that the reported odds ratio (OR) of 9.8 for a change of 10 colony-forming units (CFU) per m³ (4) was statistically unstable and potentially inflated. The estimate was very sensitive to at least three influential steps or strategies in the analysis. First, the mean airborne S. chartarum concentrations (CFU/m³) for each household were calculated incorrectly. Substituting the corrected means reduced the OR by 44% to 5.5. Second, the mean S. chartarum value (CFU/m³) was imputed in one case home.^† The sample was collected many months after sampling in the other case homes and, along with all other household samples collected at the same time, produced unusually heavy growth of non-Stachybotrys fungi, suggesting important differences in sampling technique, laboratory procedure, or environmental conditions at the time of the sampling. Exclusion of this household from the analysis^§ and correcting the means reduced the OR to 1.9. Third, matching on age in a small data set created an unstable OR. Subject age would not be expected to influence concurrent measurements of airborne fungi and did not correlate with the mean S. chartarum CFU/m³. Therefore, the strategy to match cases and controls based on age was unnecessary and potentially misleading. Analysis without the matching variable reduced the OR from 9.8 to 1.5.
2. Although the methods specified that sampling be done in a blinded manner (4), one investigator correctly inferred the identity of many case homes and wanted to be certain to identify culturable fungi in these homes if they were present. As a result, the investigator collected twice the number of air samples from case homes as were collected from control homes. In addition, investigators used aggressive, nonstandardized methods to generate artificial aerosols for sampling (e.g., vacuuming carpets and pounding on furnace ducts and furniture [4]), increasing the potential for differential exposure assessments of cases and controls if sampling were conducted in an unblinded manner.
3. Among homes classified as water damaged, the presence of any culturable airborne S. chartarum was identified in similar percentages of case and control homes (four of eight compared with three of seven) (CDC, unpublished data, February 1997). Although the numbers were small, this provided little evidence of a difference in the presence of airborne S. chartarum between water-damaged case and control homes. If the classifications of water damage were correct, this would suggest that water damage, or an unrecognized correlate of water damage, may be confounding any perceived association with S. chartarum.

Overall, the reviewers concluded that on the basis of these limitations the evidence from these studies was not of sufficient quality to support an association between S. chartarum and AIPH. In addition, the reviewers noted that evidence from other sources supporting a causal role of S. chartarum in AIPH is limited. First, AIPH is not consistent with historic accounts of animal and human illness caused by S. chartarum or related toxigenic fungi. Second, clusters of AIPH have not been reported in other flood-prone areas where growth of S. chartarum or other toxigenic fungi might be favored. Third, the mold-disease association observed in the Cleveland investigation was not observed in the investigation of a similar cluster in Chicago (34; CDC, unpublished data, May 1997).

Reported by: Office of the Director, CDC.

Editorial Note:
On the basis of the findings and conclusions in the reports of the CDC internal working group and the individual opinions of the external consultants, CDC advises that conclusions regarding the possible association between cases of ^§ The working group's reported reanalysis used the value originally coded in the laboratory record (0 CFU/m³). The result was identical to that obtained by excluding the household from the analysis. pulmonary hemorrhage/hemosiderosis in infants in Cleveland and household water damage or exposure to S. chartarum are not substantiated adequately by the scientific evidence produced in the CDC investigation (2-4). Serious shortcomings in the collection, analysis, and reporting of data resulted in inflated measures of association and restricted interpretation of the reports. The associations should be considered not proven; the etiology of AIPH is unresolved.

As a result of the reviews, CDC will implement the following:

1. CDC will continue to investigate cases of AIPH in infants, particularly when clusters of cases can be identified.

2. CDC will continue to consider possible associations between AIPH and many possible etiologies, including household water damage or exposure to environmental hydrophilic fungi/molds such as S. chartarum. Standardized protocols will be recommended for data collection and environmental assessment.

3. CDC will assist in implementation of surveillance for individual cases or clusters of cases of AIPH in infants.

Reported by: Office of the Director, CDC.