結核分枝桿菌核酸增幅檢驗在真實世界之診斷表現：北部某醫學中心的經驗

Abstract

背景：結核病為全球主要傳染病死因之一，2023年全球約有125萬人死於結核病，其中包含16.7萬名愛滋病毒感染者，結核病很可能已重新成為全球最致命的傳染病。臺灣屬結核病中度負擔國家，2024年新案發生率為每十萬人口26例，結核病防治，仍是疾病管制署之重點。近年來疾病管制署陸續推動多項落實第一線使用結核菌核酸增幅檢驗（Nucleic Acid Amplification Test，以下簡稱NAAT）的政策，希望能夠早期診斷結核病。本醫學中心收治個案多為免疫不全且疾病複雜度高，診斷開放性肺結核不易，近五年內歷經三種 NAAT 廠牌之更替，包括Cobas TaqMan MTB PCR assay（Roche Diagnostics, Basel, Switzerland；以下簡稱 Cobas）、2021年9月起改用Xpert MTB/RIF assay（Cepheid, Sunnyvale, CA, USA；以下簡稱 Xpert），2022年3月再轉為Xpert MTB/RIF Ultra assay（Cepheid, Sunnyvale, CA, USA；以下簡稱 Ultra）。本研究回顧2020至2024年間接受痰液檢體 NAAT 之個案，探討三種 NAAT 廠牌在診斷開放性肺結核之敏感性、特異性，並分析NAA test作為第一線診斷工具的使用比例是否隨政策與感染管制宣導而提升。此外，亦探討 Ultra 檢驗結果於結核病中度負擔國家之應用經驗，綜合分析其診斷效能與限制，以提供臨床實務及後續政策推動之參考。 研究目的：本研究欲探討：1）三種 NAAT 工具在傳染性結核病之診斷效能與 Ultra 檢驗的限制；2）NAAT 作為第一線結核病診斷的比例有沒有增加。 方法：本研究為一病例回溯性研究(retrospective study) ，資料由國立臺灣大學醫學院附設醫院之感染指標監測資料庫提供，撈取2020年至2024年間臺大總院及新竹分院的痰檢體，以檢測 NAAT 同一時期之細菌培養結果作為黃金標準，使用 SAS 9.4 進行 NAAT 相關統計分析。 結果：本研究共納入 6,549 份痰液樣本進行 NAAT 檢測，包括 Cobas（1,174 件） 、Xpert（1,177 件） 與 Ultra（4,198 件）。臺大總院資料顯示，Ultra 在各菌量層級中具良好診斷效能，於塗片陰性檢體中敏感性達 54.6%。相較下，新竹分院 Ultra 表現較差，偽陰性個案中外籍移工佔 35.7%，多無臨床症狀，可能因檢體品質不佳或病灶菌量偏低，影響檢測結果的敏感性。建議臨床影像高度懷疑肺結核，但無臨床症狀者，考慮重複採檢或安排支氣管鏡。此外，Ultra trace 結果偽陽性風險與結核病或糖尿病病史有關，臨床解讀應審慎評估。NAAT 作為第一線結核病診斷部分，自 2021 年 9 月起導入 Xpert 系列後，NAAT 使用比例持續上升，至 2024 年第 4 季達 24.1%，但仍有進步的空間。 結論：本研究回顧分析三種 NAAT 工具在診斷開放性肺結核的表現，整體結果顯示，目前使用的 Ultra 在高菌量檢體中表現出良好的敏感性與特異性，符合目前國際間對其作為第一線結核病診斷工具的期待。隨著政策推廣與感染管制宣導，NAAT 在臨床上的應用比例逐年提升，顯示政策與宣導對於改變臨床行為、加速診斷流程具有正面影響。然而，trace 結果的解讀是一項挑戰，臨床上遇到具有結核過去病史的病人，應該特別注意 Ultra 可能會有較高的偽陽性。臨床醫師在決策時應結合病史、影像、重複檢體等多面向資訊，以提升診斷精確度，且避免過度醫療產生。

Background: Tuberculosis (TB) remains one of the leading causes of death from infectious diseases worldwide. In 2023, an estimated 1.25 million people died from TB, including 167,000 individuals living with HIV, suggesting TB has likely reclaimed its status as the most lethal infectious disease globally. Taiwan is classified as a moderate TB burden country, with a reported incidence of 26 cases per 100,000 population in 2024. TB control remains a priority of the Taiwan Centers for Disease Control (CDC), which has implemented policies to promote the frontline use of nucleic acid amplification tests (NAATs) for early TB diagnosis. At our tertiary medical center, which serves a high proportion of immunocompromised and clinically complex patients, the diagnosis of pulmonary TB is particularly challenging. Over the past five years, our institution transitioned through three NAAT platforms: Cobas TaqMan MTB PCR assay (Roche Diagnostics, Basel, Switzerland), followed by Xpert MTB/RIF assay (Cepheid, Sunnyvale, CA, USA) in September 2021, and later Xpert MTB/RIF Ultra assay (Cepheid, Sunnyvale, CA, USA) in March 2022. This study retrospectively analyzed NAAT data from sputum specimens collected between 2020 and 2024, comparing the diagnostic performance of these three platforms for active pulmonary TB. Additionally, we evaluated whether the adoption of NAAT as a frontline diagnostic tool increased following national policy changes and infection control initiatives, and we further assessed the clinical implications and limitations of Ultra testing in a moderate TB burden setting.

Objectives: This study aimed to: (1) compare the diagnostic performance of three NAAT platforms (Cobas, Xpert, Ultra) in detecting infectious pulmonary TB and identify limitations of Ultra; and (2) assess trends in the frontline use of NAATs over time.

Methods: We conducted a retrospective study using data from the infection surveillance database of National Taiwan University Hospital (NTUH) and its Hsinchu Branch. All sputum specimens submitted for NAAT between 2020 and 2024 were included. Mycobacterial culture served as the reference standard. Statistical analyses were performed using SAS 9.4.

Results: A total of 6,549 sputum samples underwent NAAT, including Cobas (1,174), Xpert (1,177), and Ultra (4,198). At the NTUH main campus, Ultra showed good diagnostic performance, with 54.6% sensitivity in smear-negative specimens. In contrast, performance at the Hsinchu Branch was lower; 35.7% of false-negative cases were foreign migrant workers, most without symptoms, suggesting possible issues with specimen quality or low bacterial load. For patients with high radiographic suspicion but few symptoms, repeat sampling or bronchoscopy is advised. Ultra trace results were more likely to be false positives in those with prior TB or diabetes, highlighting the need for cautious interpretation. Since Xpert’s introduction in September 2021, NAAT use has increased to 24.1% by Q4 2024, though further uptake is needed

Conclusions: This study provides a comparative assessment of three NAAT platforms for the diagnosis of pulmonary TB in a real-world setting. Ultra demonstrated strong sensitivity and specificity in high-bacillary-load specimens and aligns with global expectations for frontline TB diagnostics. The increasing use of NAATs over the study period reflects the positive impact of national policy and infection control efforts on clinical practice and diagnostic efficiency. However, trace result interpretation remains a clinical challenge. In patients with prior TB history, Ultra is more prone to false positives. Clinicians are advised to interpret NAAT results in the context of clinical presentation, imaging, and, when necessary, repeat specimen testing to enhance diagnostic accuracy and avoid overtreatment.