Prognostic value of serum galactomannan in mixed ICU patients: a retrospective observational study

Little is known about serum galactomannan (GM) testing in (mostly non-neutropenic) mixed intensive care unit (ICU) patients. The aim of this study was to look for the incidence of invasive aspergillosis (IA) in critically ill patients, to validate previously reported GM thresholds, and to evaluate the prognostic value of GM.

This was a retrospective study of 474 GM samples in 160 patients from the start of January 2003 until the start of February 2004. GM tests were ordered because of a clinical suspicion of IA or on a regular basis in immune compromised patients. The number of samples per patient was 3 ± 2.6. We used the criteria of the European Organisation for Research and Treatment of Cancer (EORTC) to define proven IA, probable IA, and possible IA. The number of positive samples, with GM optical density (OD) > 0.5 was 230 (48.5%).

In our study population, five (3%) patients had proven IA, 11 (7%) had probable, 27 (17.5%) had possible, and 116 (72.5%) had no IA. We could not identify a GM threshold for IA with analysis of receiver operating characteristics (ROC) curves: with a sensitivity of (56.3%, 50%, 50%, 37.5%), specificity (38.2%, 67.5%, 68.8%, 72.9%), NPV (88.7%, 91.8%, 92.5%, 91.3%) and PPV (9.2%, 12.9%, 15.1%, 13.3%) for a cut-off of OD > 0.5, > 0.8, > 1.1 and > 1.5 respectively. IA was associated with high mortality of 87.5% and 100% in patients with probable and proven IA respectively. Patients with IA had a significant increase of GM during their stay (GMdelta 0.7 ± 1.5 vs -0.2 ± 1.5, P = 0.027). The overall ICU mortality was 41.9% and the hospital mortality was 58.1%. Patients who died in the ICU and in the hospital had higher APACHE- -II, SAPS-II and SOFA scores (P < 0.0001) and also a significant increase in GM during their stay with 0.27 ± 1.26 (ICU non-survivors) and 0.11 ± 1.55 (hospital non-survivors) compared to a decrease in GM -0.43 ± 1.7 (P = 0.004) and -0.48 ± 1.51 (P = 0.017) in ICU and hospital survivors respectively. Non-survivors also had higher mean GM values but this was not statistically significant. There was a trend towards higher GM values in patients treated with piperacillin/tazobactam (n = 34), but this did not reach statistical significance. Neutropenic patients (n = 31) showed an increase in GM during their stay 0.32 ± 1.3 vs a decrease with -0.43 ± 1.7 in non-neutropenic patients (P = 0.07). Patients on total parenteral nutrition (n = 125) had higher maximal GM levels (1.55 ± 1.94 vs 0.88 ± 1.25, P = 0.058). Patients who were mechanically ventilated had significantly higher mean (P = 0.038) and maximal (P = 0.007) GM levels.

We found a high incidence of proven and probable IA in a group of mixed ICU patients (10%) and the presence of IA was associated with a high mortality. The serum GM antigen detection test may not be useful in the diagnosis of IA in mixed ICU patients, according to the results of the ROC analysis. We could not define a useful threshold.