IIF using HEp-2 cells is recommended as the gold standard reference method to detect the presence of ANAs [18], [19], [20]. The scope of our research, which processed and analyzed more than 10,000 fluorescence tests in a routine laboratory setting, was (1) to compare the positive and negative results between the automated and conventional visual approaches, (2) to build the cutoff LIU for different patterns as well as validate and compare the custom cutoff with an automated autotiter system in two different clinical laboratories according to the end-point titer, (3) to evaluate the performance of NOVA View in patients with a clear diagnosis of CTD, and (4) to assess the automatic ANA pattern recognition system and compare the pattern from HEp-2 cell slides produced by different manufacturers.
In this study, high-quality digital images led to excellent consistency between the automated IIF analysis and the manual microscopy results, as positive and negative agreement levels between the two methods were 98.1% and 95.9%, respectively (κ=0.973). Similar to the results in our study, many studies have examined agreement between automated and conventional ANA IIF analyses. Regardless of the platform on which they are based, such as EUROPattern, AKLIDES software or NOVA View, automated interpretation systems all exhibit reliable discrimination between positive and negative results [11], [13], [16], [21], [22], [23].
The ANA titer is an important concern in a routine clinical laboratory. There is evidence to indicate that a high ANA titer is more related to autoimmune diseases [2], [24], [25]. However, with respect to clinical efficacy, the end-point titer of ANA is not commonly manipulated in clinical laboratories in routine work. The automated system, which has the ability to provide the recommended autotiter according to a quantitative reading of the immunofluorescence intensity, may provide an objective value-added report.
Our results showed a significant association between the LIU value and the end-point titer at a dilution of 1:80, which was consistent with the report of S. Schouwers et al. [17], [25]. However, we further proved that the custom cutoff was superior to the preset cutoff in NOVA View (local vs. preset cutoff: 57.4% vs. 43.7%, respectively, p<0.01) in our laboratory. However, when the test validation was performed in the other laboratory, the exact prediction was approximately the same (p=0.11), whereas the acceptable accuracy increased with the preset cutoff value (p<0.01). Therefore, it is important to establish a local cutoff to increase the accuracy of the value-added reporting. Nevertheless, the preset cutoff was more credible than the findings originating from other laboratories without a custom cutoff.
In addition, the median ANA titers for the patients with a clear diagnosis of SLE, UCTD, SjS and RA were 1:640, 1:640, 1:640 and 1:320, respectively, which showed that these patients had a relatively increased ANA titer compared with disease control group with the median ANA titers as negative. However, because of the insufficient sample size, the local cutoff for the basic pattern of nuclear dots was not established; this will be investigated further in a future study.
Another vital component of the automated digital immunofluorescence system is pattern recognition. In our study, NOVA View correctly identified 61.9% of the samples, which has the similar performance that reported by Bizzaro et al. [14]. Optimal performance in pattern recognition was found for the centromere type (92.6%), whereas the lowest performance was for the nuclear dot pattern (27.3%). The most common patterns of speckled and homogeneous were identified in 62.7% and 57.4% of cases, respectively. Such situations may occur, in part, because some investigators tend to recognize a dense granular pattern as homogeneous. In addition, the computer-aided system, which recognizes a pattern based on an algorithm, may be affected by cytoplasmic fluorescence with a mixed pattern. Nevertheless, the automated pattern-determining system was not satisfactory. Thus, confirmation of digital immunofluorescence images by expert technicians is suggested to be an essential prerequisite of high-quality ANA reports.
As there existed heterogeneity in the performance of various HEp-2 assay kits, what is the difference ratio between INOVA and some other commercial kits? To our knowledge, most literatures focused on reporting positive or negative rate on single devices [11], [12], [26], [27], [28], even less data attempted to compare patterns between different assay kits [9], [29]. Therefore, we further evaluated the discrepancy caused by the HEp-2 assay kits in our clinical laboratory. In our research, 1611 ANA-positive serum samples were processed using two kits, and 76 samples (4.7%) were found to have different pattern results; in these cases, a third commercial kit for ANA detection was used to further confirm the ANA pattern. The principal problem remains the ability to distinguish between the speckled and homogeneous patterns. Another key issue was with regard to the staining of the nucleolar pattern. For 24 samples that showed different ANA patterns between INOVA and MBL commercial kits, following confirmation by Euroimmun ANA slides, the numbers of nucleolar-positive samples were 6, 18 and 7 for INOVA, MBL and Euroimmun, respectively. Therefore, when encountering paradoxical IIF patterns according to anti-ENA profiles or existing discrepancies with past ANA patterns in other laboratories, different commercial ANA IIF kits may be utilized to confirm the pattern.
Comments (0)