中度結核盛行地區潛伏結核感染之社區性研究─應用丙型干擾素分泌檢測與結核菌素測驗

Abstract

研究背景：結核菌素皮膚測試(tuberculin skin test, TST)一直是主要偵測潛伏結核感染(latent TB infection, LTBI)的工具。 然而在過去十年間，以T細胞為檢驗基礎的商品化試劑，丙型干擾素分泌檢測(interferon-gamma-release assays, IGRAs)亦已發展用來診斷潛伏結核感染。IGRAs是利用結核菌特異性抗原刺激產生丙型干擾素的量做為評估是否有結核感染的方式，所以不受卡介苗接種與大部分的非結核分枝桿菌影響。相較於TST，IGRA的陽性反應和結核暴露程度相關性似乎較高，但是在結核低度盛行地區與中高度盛行地區的結論並不一致。迄今IGRA的研究對象大多是結核病病人與結核病接觸者或其高危險群，有關IGRA在一般族群的表現之研究仍然有限。 研究目的：依據IGRA與TST在中度結核盛行地區一般族群的表現，評估合併兩工具偵測LTBI的成效，並探討TST預測IGRA的可能性。 材料與方法：本研究為一橫斷面的社區性研究，研究期間為2011年5月，樣本來自參與彰化縣整合式健康篩檢之民眾，受檢者除身體檢查與臨床檢驗外，並同時施測丙型干擾素分泌檢測與結核菌素測驗。另外以問卷方式收集受檢者基本人口學特質與疾病情形、家族史、結核病史與結核暴露史。本研究之分析分三部分：第一部分為描述性分析，且根據TST在不同切點、不同年齡及卡介苗接種史的條件下，探討其與QFT-GIT之一致性。第二部分依據TST的工具特性，以陽性預測值(positive predictive value,PPV)與陰性預測值(negative predictive value,NPV)之事後勝算比(posterior odds)，推估其概似比(likelihood ratio,LR )，據以估計QFT-GIT的敏感度與特異度。並將所得之QFT-GIT之敏感度與特異度，計算潛伏結核感染盛行率，並與以台灣地區年結核感染率(annual risk of TB infection,ARTI)所推估之期望結核感染盛行率做比較。第三部分以QFT-GIT做為結核發病的替代指標(proxy indicatior)，估計以TST預測QFT-GIT的可能性。利用羅吉斯迴歸估計值計算臨床風險分數(clinical score)，並假定後者為常態分佈，再以貝氏統計方法估計QFT陽性與陰性的機率密度函數分佈。依據此函數可訂定臨床風險分數上限切點值及下限切點值，超過上限切點值或低於下限切點值，不進行QFT檢測，介於兩者之間則接受QFT檢測。另以概似度比(likelihood ratios)計算不同條件及臨床風險分數下的事後勝算。 結果：共有492人完成IGRA與TST，QFT-GIT(以(≥0.35IU/ml為切點)陽性率為10.6%；TST(≥10mm)陽性率為55.1%。QFT-GIT與TST(≥10mm)的一致性為50.2 %, κ=0.08。一致性在不同卡介苗接種世代分別為≤36歲：43.8%， κ=0.00；37-47歲：40.0%，κ=-0.03；48-67歲：55.2%，κ=0.13；>67歲：58.3%，κ=0.22。運用已知TST的工具特性推估QFT偵測潛伏結核感染的敏感度為5%-43%，特異度95%-99%。利用羅吉斯迴歸估計值得到臨床風險分數公式 Score=0.58×TST+0.20×Age +0.03×Sex。臨床分數上限切點值與下限切點值分別為3分及1.12分，其偽陽性與偽陰性分別為5.8%及5%。55歲以下且TST5mm以下，以及40歲以下且TST10mm以下者低於下限值；60歲以上且TST15mm以上者高於上限值。實際結果與預測結果相較之下，模式預測之敏感度為76.9%，偽陰性為23.1%；模式預測之特異度為70.3%，偽陽性為29.7%。 結論：TST與IGRA的一致性不受卡介苗接種史的影響，並隨年齡增加。IGRA偵測潛伏結核感染，具高特異度但低敏感度。運用臨床風險分數，依據TST結果、年齡及性別，區分出需進行QFT檢測者，可以減少21%之受試者。本研究發現可提供後續潛伏結核感染治療篩檢政策之參考。

Background: The tuberculin skin test (TST) was the mainstay for detecting latent TB infection (LTBI). Over the past decade, commercial tests based on T-cell assays and interferon-gamma-release assays (IGRAs) have been developed for diagnosing LTBI. IGRAs evaluate IFN-γ responses to the tuberculosis-specific antigens and would not be affected by BCG vaccination and most NTM exposure. A positive IGRA result was reported to be more strongly associated with degree of TB exposure than TST, but the findings in low prevalence countries are inconsistent with the studies in low-and intermediate prevalence countries. So far the majority of IGRAs studies have been restricted to TB patients, TB contacts or high risk groups for TB. Therefore, the performance of IGRAs in general population is still limited. Aims: Based on the performance of IGRA and TST in general population in a country with intermediate TB burden, we evaluated their operational characteristics, the yieled of the combination of TST and IGRA to detect LTBI and then explore the possibility of using TST result as a predictor for IGRA. Materials and methods: A community-based study was conducted in Changhua city in May 2011. The study subjects from the attendee of Changhua community-based integrated screen (CHCIS) were evaluated with a clinical and physical examination and underwent both QFT-GIT and TST. Subject’s information was collected by questionnaire including demographic characteristics, underlying disease, family disease history, prior TB history, and TB contact history. Our analysis consists of three major parts: we gave the results of descriptive analysis and then we used various cutoff of TST, age groups and BCG vaccination groups to compare agreement between TST and QFT-GIT. The second part was to use the operational characteristics of TST to estimate the sensitivity and specificity of QFT-GIT for LTBI. We compare expected prevalence of TB by annual risk of TB infection with prevalence of TB based on the estimated sensitivity and specificity of QFT-GIT. Given QFT-GIT as a proxy indicator of TB disease progression, we explore the possibility of using TST to predict the result of QFT-GIT. By assuming the clinical scores, as estimated from logistic regression, followed normal distribution, we applied Bayesian approach to estimate probability density function of the scores. Based on the function, we set up the upper limit and the lower limit of the scores.If the scores are within the area between the lower and the upper limit, QFT-GIT testings are in need, otherwise not. We also using likelihood ratio to estimate posterior probability by various condition and clinical scores. Results: Of 492 subjects who underwent QFT-GIT and TST, the positive rate of QFT-GIT (≥0.35IU/ml) was 10.6% (52/492), the positive rate of TST (cutoff ≥10mm.) was 55.1% (271/492). The overall level of agreement and Kappa(κ) between QFT-GIT and TST (cutoff≥10mm) was very low (50.2 %, κ=0.08).The agreement between two tests in various BCG vaccination groups was as follows: group 1 (aged ≤36 years), 43.8%,κ=0.00;group 2 (aged 37-47 years): 40.0%,κ=-0.03;group 3 (aged 48-67 years), 55.2%, κ=0.13;group 4(aged >67 years): 58.3%, κ=0.22. The result showed that agreement between two tests was affected by age, cutoff of TST and chronic disease history rather than BCG vaccination history. Sensitivity of the QFT-GIT using the operational characteristics of TST for detecting LTBI was only 5%-43%, but specificity was ranged from 95% to 99%.The score function resuling from logistic regression was:Score=0.58×TST+0.20×Age +0.03×Sex. The upper limit and the lower limit were set up at 3 and 1.12 with a false positive rate of 5.8% and a false negative rate of 5% respectively. Those younger than 55 with the induration size of TST <5mm and those younger than 40 with the induration<10mm were among the group lower than the lower limit.On the other hand, The people older than 60 with the induration size ≥15mm were among the group higher than the upper limit. The sensitivity, false negative rate, specificity and false positive rate of the model were 76.9%, 23.1%, 70.3%, 29.7%, respectively. Conclusion: Agreement between TST and IGRA was highly dependent on age and might not be affected by BCG vaccination history. IGRA was low sensitivity but high specificity tool for detecting LTBI. Using the score function, based on the age, sex and the induration sizes of TST, to identify those need QFT-GIT testing could help decrease the number needed to be testing by 21%. The above findings provide valuable information for the decision making of screening and treating LTBI.