To face the new coronavirus disease 2019 (COVID‐19) pandemic, the need for early and accurate diagnosis of the disease among suspected cases quickly became obvious for effective management, and for better control of the spread of the disease in the population. Since the beginning of this disease epidemic caused by the severe acute respiratory syndrome coronavirus‐2 (SARS‐CoV‐2), reverse transcriptase‐polymerase chain reaction (RT‐PCR) has routinely been used to confirm the diagnosis. However, several authors have pointed out the poor performance of this technique, particularly in terms of sensitivity.1, 2 Indeed, according to some authors, sensitivity could be as low as 38%3 (ie, not better than chance). This made it necessary to find a more sensitive test, given the contagiousness of SARS‐CoV‐2. We, therefore, read with great interest the article published in your journal by Cassaniti et al.4 This article deals with the diagnosis of COVID‐19 by serology (immunoglobulin m/immunoglobulin G) as a complementary approach to RT‐PCR to improve its sensitivity. According to Cassaniti et al4 and Xiang et al,5 serology is faster to implement, less expensive, easier to use, and more accessible to staff with no specific laboratory training.5 The article describes the metrological performances of serology, and compare it with RT‐PCR as the gold standard. Using a test as the gold standard when its metrological properties are clearly perfectible raises questions from a methodological point of view. Indeed, when an existing test is considered as a reference, this suggests that the test in question is always correct and that all misclassifications (false negatives and false positives) are due to the new test. However, the new test (in this case, serology) might be better than the old test (in this case, RT‐PCR), but it would be impossible to demonstrate this. Consequently, the new test will never be able to achieve a sensitivity of 100%, since it is considered responsible for all misclassifications. The same mistake has also been made by other authors regarding the use of chest computed tomography scans as a diagnostic method.6, 7 In this situation, the best strategy would be to measure the degree of agreement (using the Kappa coefficient measures8) between the two tests, that is, neither of the two tests is considered to be the reference and, therefore, any discrepancies could be linked to either of the tests. Thus, the serology performances presented by Xiang et al5 are certainly better than those presented in their paper.

The difficulty of using a gold standard is an old debate,9, 10 but still relevant nonetheless. In the absence of an accurate reference test, alternative strategies could be to perform the test repeatedly over time, to use the patient's clinical course, or the combination of several tests as the gold standard.

The purpose of writing this contribution is not to discuss the best diagnostic strategy for COVID‐19, nor is it to question the results of the authors who used RT‐PCR as a reference. On the contrary, it purports that their results might actually be even better than those presented.

面对2019新型冠状病毒病（COVID-19）大流行，对疑似病例进行早期、准确诊断的必要性很快变得明显，以便有效管理，并更好地控制疾病在人群中的传播。自严重急性呼吸综合征冠状病毒-2（SARS-CoV-2）引起的这种疾病流行以来，逆转录酶聚合酶链反应（RT-PCR）已常规用于确诊。然而，一些作者指出该技术的性能较差，特别是在灵敏度方面。1, 2事实上，根据一些作者的说法，敏感性可能低至 38% 3（即，并不比偶然性更好）。鉴于 SARS-CoV-2 的传染性，因此有必要找到一种更灵敏的测试。因此，我们饶有兴趣地阅读了 Cassaniti 等人在贵刊上发表的文章。4本文讨论了通过血清学（免疫球蛋白 m/免疫球蛋白 G）诊断 COVID-19，作为 RT-PCR 的补充方法，以提高其敏感性。根据 Cassaniti 等人4和 Xiang 等人的说法，5血清学实施速度更快、成本更低、更容易使用，并且更容易为没有接受过特定实验室培训的工作人员所使用。5本文描述了血清学的计量性能，并将其与作为金标准的 RT-PCR 进行了比较。当计量特性明显可以完善时，使用测试作为黄金标准会从方法论的角度提出问题。事实上，当将现有测试视为参考时，这表明相关测试始终是正确的，并且所有错误分类（假阴性和假阳性）都是由于新测试造成的。然而，新测试（在本例中为血清学）可能比旧测试（在本例中为 RT-PCR）更好，但无法证明这一点。因此，新测试永远无法达到 100% 的灵敏度，因为它被认为是造成所有错误分类的原因。其他作者在使用胸部计算机断层扫描作为诊断方法时也犯了同样的错误。6, 7在这种情况下，最好的策略是测量两个检验之间的一致程度（使用 Kappa 系数测量8），即两个检验都不被视为参考，因此，任何检验都不会被视为参考。差异可能与任一测试有关。因此，Xiang 等人5提出的血清学性能肯定比他们论文中提出的要好。

使用金本位制的困难是一个古老的争论，9, 10，但仍然相关。在缺乏准确的参考测试的情况下，替代策略可能是随着时间的推移重复进行测试，使用患者的临床病程或几种测试的组合作为金标准。

撰写本文的目的不是讨论 COVID-19 的最佳诊断策略，也不是质疑使用 RT-PCR 作为参考的作者的结果。相反，它声称他们的结果实际上可能比所呈现的更好。