土井 洋輝

BACKGROUND: Donor-specific antibodies (DSAs) targeting human leukocyte antigens (HLAs) substantially reduce the longevity of transplanted organs. Desensitization of DSA-positive renal transplant recipients is achieved through intravenous administration of immunoglobulin (IVIg). However, the presence and detectability of anti-HLA antibodies in IVIg preparations following administration are not fully understood. We aimed to assess whether immunoglobulin preparations contain anti-HLA antibodies that can be detected as passive antibodies when administered into the body. METHODS: We evaluated 3 immunoglobulin preparations from different pharmaceutical companies, using anti-HLA class I and II antibody specificity tests and immunocomplex capture fluorescence analysis (ICFA). RESULTS: Direct testing for anti-HLA antibodies resulted in high background errors, particularly for Venoglobulin. Diluting Venoglobulin to physiological concentrations revealed the presence of anti-HLA class I antibodies; however, no common alleles were found between the specificity identification test and ICFA.For Glovenin and Venilon, anti-HLA class I and II antibodies were detected; however, variability was observed across different test reagent lots. Moreover, dilution of the globulin formulation revealed a prozone phenomenon. CONCLUSION: The administration of IVIg complicates the accurate detection of anti-HLA antibodies, underscoring the need for careful interpretation of test results post-IVIg administration.

BACKGROUND: Activated partial thromboplastin time (APTT) is susceptible to reagent composition. This study aimed to investigate a large number of specimens and determine the cause of discrepancies. METHOD: This study included 18,994 subjects who underwent coagulation tests at our hospital from May 2020 to December 2020. Measuring reagents included HemosIL SynthASil APTT (APTT-SS, Instrumentation Laboratory) and Coagpia APTT-N (APTT-N, Sekisui Medical). RESULTS: A total of 451 patients demonstrated APTT-N of >39 seconds and an APTT-N/SS ratio of >1.3. A C-reactive protein (CRP) level of ≥1.4 mg/L demonstrated a significant positive correlation, with a higher APTT-N/SS indicating higher CRP levels. All 28 subjects receiving no anticoagulants and who had remaining specimens underwent a cross-mixing test (CMT). Of them, 17 were suspected for lupus anticoagulant (LA) by both the waveform shape and the index of circulating anticoagulant (ICA) value, 6 by the ICA value, and 5 were difficult to determine. CONCLUSION: This study revealed that the APTT-N prolongation correlated with CRP degree and the transient involvement of LA in CMT results due to CRP. This study indicated various reactivities depending on the assay reagents used. Further testing is warranted if LA is suspected, considering the patient's background.

近年,大規模言語モデル(large language models;LLM)が世界的に様々な分野で注目を集めている。LLMとは,非常に巨大なデータセットとディープラーニング技術を用いて構築された言語モデルである。LLMは,人間に近い流暢な会話が可能であら,自然言語を用いたさまざまな処理を高精度で行えることから,世界中で注目を集めている。本研究では,LLMであるOpenAI社が開発したChatGPTの異なる2つのモデル(GPT-3.5,GPT-4)にて,過去3年間の臨床検査技師国家試験におけるChatGPTの正答率について評価を行った。GPT-3.5による正答率の平均は51.4%であった。一方,GPT-4では79.8%の正答率結果が得られた。本結果より,ChatGPTはこの先医療現場における有効なアドバイザーとして進化する可能性をもつことが示唆された。しかし,今回不正解となった20%の中には,患者を診断する際に誤診につながりかねない回答が含まれており,今後のChatGPTの精度向上は必須と考えられる。今回の検証は,LLMにおけるChatGPTの臨床検査領域での多様な応用の進展に寄与すると考えられ,この先の発展に期待したい。(著者抄録)