In the Journals

Portable mass spectrometry technique accurately detects C. difficile in unprocessed stool

Field asymmetric ion mobility spectrometry, a portable mass spectrometry instrument that provides rapid chemical composition analysis of gaseous mixtures, accurately distinguishes between Clostridium difficile-positive and -negative stool samples, according to recent study data.

In order to ascertain the diagnostic accuracy of field asymmetric ion mobility spectrometry (FAIMS) (Lonestar, Owlstone) for direct detection of C. difficile in unprocessed stool samples, Marije K. Bomers, MD, from VU University Medical Center in The Netherlands, and colleagues used FAIMS to analyze the headspace of 213 stool samples collected between March and July 2013 at two Amsterdam hospitals. All samples were analyzed by direct toxin enzyme immunoassay and anaerobic culture as reference standard, and 71 were determined to be positive for C. difficile. The study was then conducted in three phases.

Marije K. Bomers, MD

Marije K. Bomers

The training phase used 90 samples to develop classification algorithms; the test phase used 45 samples to test the algorithms for performance; and the blinded validation phase used 78 samples to evaluate the diagnostic accuracy of the best performing algorithm compared with the reference standard.

Out of three algorithms, a Random Forest classification algorithm performed most accurately in the training and test phases. Median predicted probability was 0.17 (IQR = 0.2) for negative samples (n = 81) and 0.94 (IQR = 0.44) for positive samples (n = 45; P < .001). The area under the receiver operating characteristic curve (C-statistic) was 0.91 (95% CI, 0.86-0.97). In the validation phase, median predicted probability was 0.18 (IQR = 0.18) for negative samples (n = 43) and 0.82 (IQR = 0.5) for positive samples (n = 26; P < .001). C-statistic was 0.86 (95% CI, 0.75-0.97). Diagnostic accuracy further improved when analysis was restricted to specimens tested within 1 week of sampling; median predicted probability was 0.19 (IQR = 0.18) for negative samples (n = 50) and 0.94 (IQR = 0.38) for positive samples (n = 26; P < .001) and C-statistic was 0.93 (95% CI, 0.85-1). A 0.32 predicted probability cutoff value for these samples corresponded to 92.3% (95% CI, 77.4-89.6) sensitivity and 86% (95% CI, 78.3-89.3) specificity. Discriminatory accuracy was further improved for even fresher samples.

“FAIMS analysis of stool samples can accurately discriminate C. difficile-positive from –negative samples,” the researchers concluded. “Moreover, it is easy to use, quick, and accurate. It has the potential of a very useful tool in the diagnosis of C. difficile infections, not only in the laboratory, but even on hospital wards.” – by Adam Leitenberger

Disclosure: The researchers report no relevant financial disclosures. 

Photo credit: Anita Edridge

Field asymmetric ion mobility spectrometry, a portable mass spectrometry instrument that provides rapid chemical composition analysis of gaseous mixtures, accurately distinguishes between Clostridium difficile-positive and -negative stool samples, according to recent study data.

In order to ascertain the diagnostic accuracy of field asymmetric ion mobility spectrometry (FAIMS) (Lonestar, Owlstone) for direct detection of C. difficile in unprocessed stool samples, Marije K. Bomers, MD, from VU University Medical Center in The Netherlands, and colleagues used FAIMS to analyze the headspace of 213 stool samples collected between March and July 2013 at two Amsterdam hospitals. All samples were analyzed by direct toxin enzyme immunoassay and anaerobic culture as reference standard, and 71 were determined to be positive for C. difficile. The study was then conducted in three phases.

Marije K. Bomers, MD

Marije K. Bomers

The training phase used 90 samples to develop classification algorithms; the test phase used 45 samples to test the algorithms for performance; and the blinded validation phase used 78 samples to evaluate the diagnostic accuracy of the best performing algorithm compared with the reference standard.

Out of three algorithms, a Random Forest classification algorithm performed most accurately in the training and test phases. Median predicted probability was 0.17 (IQR = 0.2) for negative samples (n = 81) and 0.94 (IQR = 0.44) for positive samples (n = 45; P < .001). The area under the receiver operating characteristic curve (C-statistic) was 0.91 (95% CI, 0.86-0.97). In the validation phase, median predicted probability was 0.18 (IQR = 0.18) for negative samples (n = 43) and 0.82 (IQR = 0.5) for positive samples (n = 26; P < .001). C-statistic was 0.86 (95% CI, 0.75-0.97). Diagnostic accuracy further improved when analysis was restricted to specimens tested within 1 week of sampling; median predicted probability was 0.19 (IQR = 0.18) for negative samples (n = 50) and 0.94 (IQR = 0.38) for positive samples (n = 26; P < .001) and C-statistic was 0.93 (95% CI, 0.85-1). A 0.32 predicted probability cutoff value for these samples corresponded to 92.3% (95% CI, 77.4-89.6) sensitivity and 86% (95% CI, 78.3-89.3) specificity. Discriminatory accuracy was further improved for even fresher samples.

“FAIMS analysis of stool samples can accurately discriminate C. difficile-positive from –negative samples,” the researchers concluded. “Moreover, it is easy to use, quick, and accurate. It has the potential of a very useful tool in the diagnosis of C. difficile infections, not only in the laboratory, but even on hospital wards.” – by Adam Leitenberger

Disclosure: The researchers report no relevant financial disclosures. 

Photo credit: Anita Edridge