欧盟医疗器械协调小组（MDCG）于2019年10月11日发布了一份指南文件，旨在帮助医疗软件制造商了解新的欧盟医疗器械法规（MDR）和欧盟体外诊断医疗器械法规（IVDR）下的软件资格标准。

本指南讨论了根据MDR或IVDR分类的某些软件，例如可以直接控制（硬件）医疗器械的软件（例如放射治疗软件），可提供即时决策触发信息（如血糖仪软件）或为医护专业人员提供支持的软件（如心电图解读软件）。

例如，医疗器械软件可以是独立软件，并能够接收测量结果。针对这一点，该指南中提到了其中一个例子，即可以通过经直肠超声检查结果、年龄和体外诊断工具来计算患者患前列腺癌风险的软件。另一方面，软件也可能会驱动器械或影响器械使用。对此，该指南文件举例说明了连接到闭环胰岛素输送系统的软件。

“该软件可以，但不限于：a）通过接口（如软件、硬件）或通过器械操作员操作、修改器械的状态或控制器械，b）或提供与该器械的（硬件）功能相关的输出，”该指南文件中提到。软件还可以定性为“医疗器械软件”（无论其位置如何，例如，在云端、计算机、手机上运行，或者作为硬件医疗器械上的附加功能）。

这份28页的指南文件还包含一个涉及医疗器械软件资格认证五步骤的决策树，以及另一个涉及将医疗器械软件视为医疗器械或体外诊断医疗器械三步骤的决策树。

根据IVDR对医疗器械软件进行分类的情形举例如下：

“根据规则3(k)，预期安装在全自动酶联免疫吸附测定（ELISA）分析仪上的软件、预期通过人类HbA1c ELISA获得的结果确定血清中人类HbA1c的浓度的软件，以及预期用于筛查和诊断糖尿病并对糖尿病患者进行监测的软件，应属于C类。

根据规则3(h)，PAP染色自动宫颈细胞学筛查系统中的软件（预期用于判定PAP宫颈涂片为正常或可疑）应属于C类。

根据规则1，预期解释用于确认和测定人血清和血浆中HIV-1、HIV-1 O组和HIV-2抗体的线性免疫测定自动读数的软件，应属于D类。

根据规则3(l)，使用母体参数（如年龄、血清标记物浓度和通过胎儿超声检查获得的信息）评估21-三体综合征风险的软件，应属于C类。”

指南中关于执行规则部分进一步指出，根据所实现的预期用途，对既实现了自身预期用途、又驱动（硬件）器械或影响（硬件）器械使用的用于医疗用途的医疗器械软件进行了分类。然而，在这种情况下，其风险等级不得低于硬件医疗器械的风险等级。

对此，该指南举例说明了预期与近红外激光扫描仪同时使用的黑色素瘤图像分析软件（IIa类）。

“该软件用于驱动近红外激光扫描仪或影响近红外激光扫描仪的使用，因其旨在通过执行专有多重曝光程序来检测黑色素瘤来控制扫描仪。因此，执行规则3.3适用。然而，分类规则11也可能适用，这取决于软件的预期医疗用途（如癌症诊断）。根据分类规则11和附录VIII执行规则3.5，该医疗器械软件将被归类为III类。”

根据MDR的部分内容以及国际医疗器械监管者论坛（IMDRF）的国际指南，MDCG进一步解释了分类规则11，以及如何处理与有源器械提供的信息有关的风险，并对有源器械提供的信息对医疗决策（患者管理）的重要性（结合医疗状况（患者病情））进行了描述和分类。

该指南中还包含与IVDR相关的其他分类和执行规则。

英文原文

Classifying Software Under MDR, IVDR: New Guidance From MDCG

The European Commission’s Medical Device Coordination Group (MDCG) on Friday released guidance to help medical software manufacturers understand the criteria for the qualification of software under the new EU Medical Devices Regulation (MDR) and In Vitro Diagnostic Regulation (IVDR).

The guidance discusses certain types of software that would be classified under MDR or IVDR, such as software that can directly control a (hardware) medical device (e.g. radiotherapy treatment software), can provide immediate decision-triggering information (e.g. blood glucose meter software) or provide support for health professionals (e.g. electrocardiogram interpretation software).

For instance, medical device software (MDSW) may be independent and able to receive measurements. The guidance uses the example of software that can use transrectal ultrasound findings, age and in vitro diagnostic instruments to calculate a patient’s risk of developing prostate cancer. Or the software may “drive or influence” a medical device and the guidance points to software connected to a closed-loop insulin delivery system as an example.

“This software can, but is not limited to: a) operate, modify the state of, or control the device either through an interface (e.g.,software, hardware) or via the operator of this device b) or supply output related to the (hardware) functioning of that device,” the guidance notes. Software also may be qualified as MDSW “regardless of its location (e.g.operating in the cloud, on a computer, on a mobile phone, or as an additional functionality on a hardware medical device)."

The 28-page guidance also features a decision tree with five steps for the qualification of MDSW and another tree with three decisions for the qualification of MDSW as either a medical device or an IVD.

Examples for the classification of MDSW under the IVDR include:

"Software intended to be installed on a fully automated enzyme-linked immunosorbent assay (ELISA) analyser, and intended to determine the Human HbA1c concentration in serum from the results obtained with a Human HbA1c ELISA, intended to screen for and diagnose diabetes and monitor diabetic patients, should be in class C per Rule 3(k).

Software within a PAP stain automated cervical cytology screening system, intended to classify the PAP cervical smear as either normal or suspicious, should be in class C per Rule 3(h).

Software for the interpretation of automated readings of line immunoassay for the confirmation and determination of antibodies to HIV-1, HIV-1 group O and HIV-2 in human serum and plasma, should be in class D per Rule 1.

Software that uses maternal parameters such as age, concentration of serum markers and information obtained through foetal ultrasound examination for evaluating the risk of trisomy 21, should be in class C per Rule 3(l)."

The section of the guidance on implementing rules further notes that MDSW “that both achieves its own intended purpose and also drives or influences the use of a (hardware) device for a medical purpose is classified on its own, based on the intended purpose achieved. In such a case, however, the risk class shall not be lower than the risk class of the hardware medical device.”

For an example, the guidance points to melanoma image analysis software intended to be used with a near-infrared laser light scanner, which is considered class IIa.

“The software‘drives or influences the use of’the near-infrared laser light scanner as it is intended to take control of the scanner by letting it execute proprietary multi-exposure programs for the detection of melanoma. As such, implementing rule 3.3 applies. However, Rule 11 would also apply based on the intended medical purpose of the software e.g. cancer diagnosis. The MDSW would be classified as class III based on Rule 11 (see section Classification Rules) and per implementing rule 3.5 of Annex VIII,” the guidance says.

In line with part of MDR and international guidance from the International Medical Device Regulators Forum, the MDCG further explains Rule 11 and how it is intended to address the risks related to the information provided by an active device and that it describes and categorizes the “significance of the information provided by the active device to the healthcare decision (patient management) in combination with the healthcare situation (patient condition).”

Other classification and implementing rules related to IVDR are also included in the guidance.