Insufficient Demonstration of Long-Term Stability of Aspergillus Galactomannan

LETTER

The report in the Journal of Clinical Microbiology by Wheat et al. (1) investigates the stability of the galactomannan index (GMI) (the ratio of the optical densities of patient samples to the mean optical density of control samples), following storage using the Platelia immunoassay to detect invasive aspergillosis. The authors state that the literature relating to this is “scant.” However, several recent publications have reported a lack of reproducibility of GM detection (2–6). Our observations also demonstrated a significant reduction in GMI during the storage of serum (7), as have two later publications (8, 9).

Wheat et al. performed a prospective evaluation of the stability at −20°C of GMI in serum and bronchoalveolar lavage (BAL) fluid using pooled samples. The authors provide no clinical information, even though it is critical to know the diagnoses of the patients, as a decline in GMI may be diagnosis dependent and influenced by serum albumin concentration (9). Pooling of samples is undesirable, potentially introducing confounding factors from one patient into all samples in a pool. Furthermore, testing of each pool, rather than of six independent samples, in triplicate gives a very small data set, precluding statistical analysis and making any meaningful conclusions on GMI stability unconvincing.

In the retrospective long-term evaluation, the sample size was also small. Moreover, it is important to establish whether these specimens came from different patients, as repeated samples from the same patient can mask the loss of GMI signals (7). Even though the authors claim that their data support the stability of GMI during storage, more than half of the samples showed signal loss after storage, with a ≥20% reduction in GMI in 47% of BAL fluid and 20% of serum samples. The lack of statistical significance may be due to a lack of statistical power (high type II error). In fact, if the Wheat et al. study had been powered for the reduction of GMI by at least 20%, a statistically significant reduction would have been observed.

Wheat et al. attempt several explanations for the discrepancy between their findings and those of other studies. They suggest that assay “lot-to-lot variation” can explain GMI signal loss. If this is true, then it is a major concern with the assay and not a weakness of our study (7). However, the significant GMI signal loss that we observed was with repeat testing using the same kit. They suggest that the “experience of the operators” may be a factor, even though this possibility is excluded in our paper by several lines of evidence, including statistical demonstration of assay repeatability of the threshold control. Wheat et al. also suggest that the inclusion of samples that were initially GMI negative may be “less informative or accurate” for GMI decline, but this is not supported by our data and ignores the fact that 67% of the GMI-positive cases showed signal loss.

In summary, the conclusion of the Wheat et al. study—that GMI shows long-term stability following storage—is speculative in view of its unsatisfactory design and lack of statistical power and is contradicted by the authors' own data as well as by data from larger studies that all show significant GMI signal loss.