暂无摘要。

The establishment of the International Consensus on ANA Patterns (ICAP) in 2014-2015 was welcomed by members of the medical community as a significant improvement in guiding harmonization of ANA test interpretation and reporting. In the subsequent years, several itinerant meetings and continuous interaction with the community contributed to disseminate the ICAP harmonization on the immunofluorescence patterns observed in the indirect immunofluorescence assay on HEp-2 cells (HEp-2 IFA) and to promote progressive improvement in the classification of HEp-2 IFA patterns. The 6th ICAP Workshop was held in person on September 6, 2021 as a satellite meeting of the 15th Dresden Symposium on Autoantibodies. This article summarizes the major discussions at the meeting as well as outlining the current plans for the ICAP committee.

The morphological patterns in indirect immunofluorescence assay on HEp-2 cells (HEp-2 IFA) reflect the autoantibodies in the sample. The International Consensus on ANA Patterns (ICAP) classifies 30 relevant patterns (AC-0 to AC-29). AC-4 (fine speckled nuclear pattern) is associated to anti-SS-A/Ro, anti-SS-B/La, and several autoantibodies. Anti-SS-A/Ro samples may contain antibodies to Ro60 and Ro52. A variation of AC-4 (herein designated AC-4a), characterized by myriad discrete nuclear speckles, was reported to be associated with anti-SS-A/Ro. The plain fine speckled pattern (herein designated AC-4b) seldom was associated with anti-SS-A/Ro. This study reports the experience of four expert laboratories on AC-4a and AC-4b.
Anti-Ro60 monoclonal antibody A7 was used to investigate the HEp-2 IFA pattern. Records containing concomitant HEp-2 IFA and SS-A/Ro tests from Durand Laboratory, Argentina (n = 383) and Fleury Laboratory, Brazil (n = 144,471) were analyzed for associations between HEp-2 IFA patterns and disease-associated autoantibodies (DAA): double-stranded DNA, Scl-70, nucleosome, SS-B/La, Sm, and U1-RNP. A total of 381 samples from Dresden Technical University (TU-Dresden), Germany, were assayed for HEp-2 IFA and DAA.
Monoclonal A7 recognized Ro60 in Western blot and immunoprecipitation, and yielded the AC-4a pattern on HEp-2 IFA. Analyses from Durand Laboratory and Fleury Laboratory yielded compatible results: AC-4a was less frequent (8.9% and 2.7%, respectively) than AC-4b (26.1% and 24.2%) in HEp-2 IFA-positive samples. Reactivity to SS-A/Ro occurred in 67.6% and 96.3% of AC-4a-pattern samples against 23% and 6.8% of AC-4b pattern samples. Reciprocally, AC-4a occurred in 24% and 47.1% of anti-SS-A/Ro-positive samples, and in 3.8% and 0.1% of anti-SS-A/Ro-negative samples. Data from TU-Dresden show that the AC-4a pattern occurred in 69% of 169 anti-SS-A/Ro-monospecific samples (62% of all anti-SS-A/Ro-positive samples) and in 4% of anti-SS-A/Ro-negative samples, whereas anti-SS-A/Ro occurred in 98.3% of AC-4a samples and in 47.9% of AC-4b samples. In all laboratories, coexistence of anti-SS-B/La, but not other DAA, in anti-SS-A/Ro-positive samples did not disturb the AC-4a pattern. AC-4a was predominantly associated with anti-Ro60 antibodies.
This study confirms the association of AC-4a pattern and anti-SS-A/Ro in opposition to the AC-4b pattern. The results of four international expert laboratories support the worldwide applicability of these AC-4 pattern variants and their incorporation into ICAP classification under codes AC-4a and AC-4b, respectively. The AC-4 pattern should be maintained as an umbrella pattern for cases in which one cannot discriminate AC-4a and AC-4b patterns. The acknowledgment of the AC-4a pattern should add value to HEp-2 IFA interpretation.

BACKGROUND: Antinuclear antibody pattern recognition is vital for autoimmune disease diagnosis but labor-intensive for manual interpretation. To develop an automated pattern recognition system, we established machine learning models based on the International Consensus on Antinuclear Antibody Patterns (ICAP) at a competent level, mixed patterns recognition, and evaluated their consistency with human reading.
METHODS: 51,694 human epithelial cells (HEp-2) cell images with patterns assigned by experienced medical technologists collected in a medical center were used to train six machine learning algorithms and were compared by their performance. Next, we choose the best performing model to test the consistency with five experienced readers and two beginners.
RESULTS: The mean F1 score in each classification of the best performing model was 0.86 evaluated by Testing Data 1. For the inter-observer agreement test on Testing Data 2, the average agreement was 0.849 (κ) among five experienced readers, 0.844 between the best performing model and experienced readers, 0.528 between experienced readers and beginners. The results indicate that the proposed model outperformed beginners and achieved an excellent agreement with experienced readers.
CONCLUSIONS: This study demonstrated that the developed model could reach an excellent agreement with experienced human readers using machine learning methods.