Comparison of the performance of two spontaneous sedimentation techniques for the diagnosis of human intestinal parasites in the absence of a gold standard.

Performance evaluation of diagnostic tests is critical in the search for accurate diagnoses. A gold standard test is usually absent in parasitology, thus rendering satisfactory assessment of diagnostic accuracy difficult. Moreover, reliability (assessed by the study of repeatability) is a rarely studied characteristic of diagnostic tests. This study compared and evaluated the performance (repeatability, concordance and accuracy) of the spontaneous sedimentation technique (SST) and the Paratest for the diagnosis of Giardia lamblia, Entamoeba histolytica complex, Blastocystis spp., Ascaris lumbricoides, hookworm, Trichuris trichiura and Calodium hepaticum. Fecal samples of 143 individuals were separated into three replicates for each test. Concordance and homogeneity of the results between replicates of each test and between tests were evaluated. Proportions of positives, sensitivity and specificity were estimated using a Bayesian Latent Class Model. High repeatability of both tests was found for the detection of intestinal parasites, except for Blastocystis spp. and hookworm. Concordance between tests was generally high (concordance correlation coefficient, 0.72-0.88), except for Blastocystis spp., hookworm and T. trichiura. The Paratest detected more cases of Blastocystis spp. and fewer of hookworm than the SST. The tests were quite discordant in the detection of T. trichiura. A low sensitivity (39.4-49.2% for SST, 35.8-53.8% for Paratest) and a high specificity (93.2-97.2%) were found for both tests. The Paratest presented a slightly higher sensitivity for the diagnosis of Blastocystis spp. (53.8%), and SST did so for hookworm (49.2%). This is the first study on repeatability and accuracy (using a Bayesian approach) of two spontaneous sedimentation techniques. These results suggest underdiagnosis of little dense parasitic forms due to technical limitations in both tests. We conclude that the combined study of repeatability, concordance and accuracy is a key strategy for better evaluation of the performance of tests and is also useful for the identification of technical limitations.