Background: Rapid and reliable diagnosis of tuberculosis (TB) represents a diagnostic challenge in compartmentalized extrapulmonary TB infection because of the small number of mycobacteria (MTB) and the frequent lack of fresh samples to perform culture. Here, we estimate the performances of homemade droplet digital PCR (ddPCR)-based assays against culture in 89 biopsies, for those fresh and formalin-fixed and paraffin-embedded (FFPE) subsamples were available.

Methods: MTB diagnosis in fresh subsamples was performed by culture. Fresh subsamples were also analyzed for acid-fast bacilli smear-microscopy (AFB) and Xpert^® MTB/RIF (Xpert). MTB examination was repeated in blind in the 89 FFPE subsamples by in-house ddPCR assays targeting the IS6110 and rpoB. Analytical sensitivity of ddPCR assays was evaluated using serial dilution of H37Rv strain. Limit of detection (LOD) was calculated by probit analysis. Results were expressed in copies/10⁶ cells.

Results: IS6110 and rpoB ddPCR assays showed a good linear correlation between expected and observed values (R²: 0.9907 and 0.9743, respectively). Probit analyses predicted a LOD of 17 and 40 copies/10⁶ cells of MTB DNA for IS6110 and rpoB, respectively. Of the 89 biopsies, 68 were culture positive and 21 were culture negative. Considering mycobacterial culture as reference method, IS6110 assay yielded positive results in 67/68 culture-positive samples with a median interquartile range (IQR) of 1,680 (550–8,444) copies/10⁶ cells (sensitivity: 98.5%; accuracy: 98.9). These performances were superior to those reported by the rpoB assay in FFPE subsamples (sensitivity: 66.20%; accuracy: 74.1) and even superior to those reported by Xpert and AFB in fresh subsamples (sensitivity: 79.4 and 33.8%, respectively; accuracy: 84.3 and 49.4, respectively). When Xpert and AFB results were stratified according to mycobacterial load detected by rpoB and IS6110 ddPCR, bacterial load was lower in Xpert and AFB negative with respect to Xpert and AFB-positive samples (p = 0.003 and 0.01 for rpoB and p = 0.01 and 0.11 for IS6110), confirming the poor sensitivity of these methods in paucibacillary disease.

Conclusion: ddPCR provides highly sensitive, accurate, and rapid MTB diagnosis in FFPE samples, as defined by the high concordance between IS6110 assay and culture results. This approach can be safely introduced in clinical routine to accelerate MTB diagnosis mainly when culture results remain unavailable.

背景：由于分枝杆菌 (MTB) 数量少且经常缺乏新鲜样本进行培养，因此快速可靠地诊断结核病 (TB) 代表了肺外结核感染的诊断挑战。在这里，我们估计了基于自制液滴数字 PCR (ddPCR) 的测定在 89 次活检中对培养的性能，因为这些新鲜和福尔马林固定和石蜡包埋 (FFPE) 子样本可用。

方法：新鲜子样本中的 MTB 诊断通过培养进行。还对新鲜子样品进行了抗酸杆菌涂片显微镜检查 (AFB) 和 Xpert ^® MTB/RIF (Xpert) 分析。通过针对 IS6110 和 rpoB 的内部 ddPCR 分析，在 89 个 FFPE 子样本中盲重复了 MTB 检查。使用 H37Rv 菌株的连续稀释来评估 ddPCR 测定的分析灵敏度。通过概率分析计算检测限 (LOD)。结果以拷贝/10 ⁶细胞表示。

结果： IS6110 和 rpoB ddPCR 检测显示预期值和观察值之间存在良好的线性相关性（电阻²：分别为 0.9907 和 0.9743）。Probit 分析预测，IS6110 和 rpoB 的 MTB DNA的 LOD 分别为 17 和 40 拷贝/10 ^{6 个}细胞。在 89 次活检中，68 次为培养阳性，21 次为培养阴性。考虑分枝杆菌培养作为参考方法，IS6110 检测在 67/68 个培养阳性样本中产生阳性结果，中位数四分位距 (IQR) 为 1,680 (550–8,444) 拷贝/10 ⁶细胞（灵敏度：98.5%；准确度：98.9）。这些性能优于 rpoB 分析在 FFPE 子样本中报告的那些（灵敏度：66.20%；准确度：74.1），甚至优于 Xpert 和 AFB 在新鲜子样本中报告的那些（灵敏度：分别为 79.4 和 33.8%；准确度：84.3和 49.4，分别）。当根据 rpoB 和 IS6110 ddPCR 检测到的分枝杆菌载量对 Xpert 和 AFB 结果进行分层时，相对于 Xpert 和 AFB 阳性样品，Xpert 和 AFB 阴性的细菌载量较低。磷 = 0.003 和 0.01 对于 rpoB 和 磷 = 0.01 和 IS6110 的 0.11），证实了这些方法在少杆菌病中的敏感性较差。

结论：ddPCR 在 FFPE 样品中提供高度灵敏、准确和快速的 MTB 诊断，正如 IS6110 检测和培养结果之间的高度一致性所定义的那样。这种方法可以安全地引入临床常规以加速 MTB 诊断，主要是在培养结果仍然不可用时。