Two recent decisions by the Food and Drug Administration will likely significantly impact testing for multiple myeloma (MM) minimal residual disease (MRD). First, on April 12, 2024, the FDA's Oncologic Drugs Advisory Committee (ODAC) voted to approve the use of MRD as an end point for accelerated approval of new treatments for patients with MM.¹ This was a result of near 10 years effort by multiple institutions, professional societies, and patient's advocacy groups,^2-6 and reflects the current state of MM treatment in which new therapeutic options have dramatically improved progression-free and overall survival (PFS and OS),⁷ making them impractical as the only clinical trial endpoints. Second, after many years of deliberations, FDA announced on April 29, 2024 that it will start overseeing laboratory developed tests (LDTs).⁸ LDTs are in vitro diagnostic products (IVDs) intended for clinical use; they are designed and validated for use within individual laboratories certified for performing high complexity testing under the Clinical Laboratory Improvement Amendments of 1988 (CLIA). The FDA's decision is currently being challenged by the American Clinical Laboratory Association.

The most commonly used assays for MRD detection in plasma cell neoplasms are next-generation sequencing (NGS) of the rearranged variable immunoglobulin heavy chain^{9, 10} and high sensitivity flow cytometry immunophenotyping (next-generation flow, NGF).^11-14 Each type of assay has its advantages and disadvantages. For example, NGS requires knowledge of the diagnostic specimen for sequence comparison, but that knowledge enables higher sensitivity (10⁻⁶). NGF, as developed by the Euroflow/Spanish flow cytometry experts and commercialized by Cytognos requires no prior diagnostic specimen and can provide information regarding quality of the specimen (hemodilution); the sensitivity of the NGF is between 10⁻⁵ and 2 × 10⁻⁶, depending on the number of cells (events) collected. In 2016, the International Myeloma Working Group established the minimum sensitivity of 10⁻⁵ for MM MRD testing.⁴

Currently the only FDA-approved test for MM MRD is NGS-based clonoSEQ® by Adaptive Biotechnologies.⁹ For the NGF method to be FDA-approved, it will require a comprehensive package submission to the FDA, either as Premarket Approval (PMA) application or a Premarket Notification 510(k). However, FDA has a category of investigational device exemption (IDE) which allows a test to be used in a clinical study to collect safety and effectiveness data. There is limited data available regarding the requirements for FDA IDE approval of a flow cytometry assay for MRD testing.

In our laboratory, we adopted and validated the NGF method as an LDT, and have, since 2017, tested over 13 000 unique specimens (unpublished data). Our test validation protocol had been certified by CLIA and New York State Department of Health. However, when we submitted our validation documents as a part of the application for the FDA IDE for the BIQSFP-Funded Study EAA171, FDA requested additional validation experiments due to the fact that the MRD result was going to be used for MM patient stratification. After several meetings to precisely define additional requirements, we agreed with the FDA on the plan of action. Here we summarize additional experiments required by the FDA in order to approve IDE for the use of NGF in a clinical trial.

Analyte, specimen, and/or matrix stability: Additional four bone marrow samples were stored in shipping containers at ambient temperature and tested after 24, 48, 72, and 96 h. Results from the accepted time points had to be qualitatively the same (MRD-positive or negative). The last acceptable time point was defined as the one in which the coefficient of variation (CV) ≤25%. The data showed that under ambient shipping condition, MRD detection can be confidently assessed up to 96 h post-draw (Table S1).

Precision/reproducibility: Additional three MM samples were spiked into normal bone marrow samples, for calculated tumor burden between 5 × 10⁻⁶ and 5 × 10⁻⁵. A total of eight replicates were performed for each sample. CVs were calculated for comparisons of intra-assay, inter-assay, operator-to-operator, and instrument-to-instrument precision. Each replicate had to be qualitatively the same, with all the CVs ≤25%. Standardized machine settings and standardized gating strategy showed a highly precise assay near the assay's limit of detection (LOD) (Tables S2a and S2b).

Accuracy: Twenty-one MM samples were used to compare MRD results by NGF with NGS-based clonoSEQ® assay, using split sample procedure. Three samples were tested with two different concentrations of spiked abnormal cells. Additional 13 diagnostic specimens were also a part of this comparison. The results showed 100% concordance of NGF with clonoSEQ® NGS testing for MRD ≥10⁻⁵. In addition, 10 out of 10 NGF-positive cases with the sensitivity between 2 × 10⁻⁶ and 10⁻⁵ were confirmed positive by NGS. Four NGF-negative cases showed very low level of positivity by NGS (3–4 × 10⁻⁶) (Figure 1 and Table S3). Overall Pearson correlation coefficient was 0.83.

Analytical sensitivity/limit of detection: Four MM bone marrow samples were spiked into normal bone marrow to achieve calculated tumor burden of 10⁻³ (baseline), 10⁻⁴, 10⁻⁵, and 5 × 10⁻⁶ (half-log below the assay LOD), in triplicates, with expected CVs ≤25%. Two samples showed precise MRD-positive LOD down to 2 × 10⁻⁶, and the other two down to 10⁻⁵ (Tables S4a and S4b).

In summary, we describe a successful approval of the MM MRD test by NGF as an FDA IDE, to be used as a decision-making tool for patient stratification in a clinical trial (packet content with submitted studies/files is listed in Table S5). Two main requirements by the FDA were (I) to show reliable MRD testing at a half-log higher sensitivity than the stated LOD of the assay; and (II) to show direct correlation of the MRD results with the FDA-approved NGS test (which had been approved in the time between our original CLIA-test validation and the time of IDE submission). We were able to achieve excellent correlation between NGF and NGS on a split sample cohort of specimens. Our experience may help other laboratories with IDE submissions and test validation, particularly in the new era of FDA regulation of LDTs. Furthermore, data presented here may be of interest to professional societies and companies seeking a higher level of FDA approval for the NGF in MM MRD testing (PMA or Premarket Notification 510(k)), as well as to individual clinical laboratories looking to validate MM MRD test as an LDT in the era of FDA oversight.

美国食品药品监督管理局 （Food and Drug Administration） 最近的两项决定可能会对多发性骨髓瘤 （MM） 微小残留病 （MRD） 的检测产生重大影响。首先，2024 年 4 月 12 日，FDA 肿瘤药物咨询委员会 （ODAC） 投票批准使用 MRD 作为加速批准 MM 患者新疗法的终点。这是多个机构、专业协会和患者权益团体近 10 年努力的结果，^2-6并反映了 MM 治疗的现状，其中新的治疗方案显着改善了无进展生存期和总生存期（PFS 和 OS）⁷，这使得它们作为唯一的临床试验终点不切实际。其次，经过多年的考虑，FDA 于 2024 年 4 月 29 日宣布将开始监督实验室开发的测试 （LDT）。⁸ LDT 是用于临床的体外诊断产品 （IVD）;它们经过设计和验证，可在根据 1988 年临床实验室改进修正案 （CLIA） 认证可用于执行高复杂性测试的单个实验室。FDA 的决定目前正受到美国临床实验室协会 （American Clinical Laboratory Association） 的质疑。

浆细胞肿瘤中 MRD 检测最常用的检测方法是重排可变免疫球蛋白重链^9、10 的下一代测序 （NGS） 和高灵敏度流式细胞术免疫表型（下一代流式，NGF）。^11-14 每种类型的测定都有其优点和缺点。例如，NGS 需要了解诊断标本才能进行序列比较，但该知识可实现更高的灵敏度 （10⁻⁶）。NGF 由 Euroflow/西班牙流式细胞术专家开发并由 Cytognos 商业化，不需要事先诊断标本，并且可以提供有关标本质量（血液稀释）的信息;NGF 的灵敏度在 ^10-5 到 2 × ^10-6 之间，具体取决于收集的细胞（事件）的数量。2016 年，国际骨髓瘤工作组确定了 MM MRD 检测的最低灵敏度为 ^10-5。⁴

目前，FDA 唯一批准的 MM MRD 检测是 Adaptive Biotechnologies 的基于 NGS 的 clonoSEQ®。⁹ 要使 NGF 方法获得 FDA 批准，需要向 FDA 提交全面的包装，要么是上市前批准 （PMA） 申请，要么是上市前通知 510（k）。但是，FDA 有一类研究器械豁免 （IDE），允许在临床研究中使用测试来收集安全性和有效性数据。关于 FDA IDE 批准用于 MRD 检测的流式细胞术检测的要求，可用的数据有限。

在我们的实验室中，我们采用并验证了 NGF 方法作为 LDT，自 2017 年以来，我们已经测试了 13000 多个独特的标本（未发表的数据）。我们的测试验证方案已获得 CLIA 和纽约州卫生部的认证。然而，当我们为 BIQSFP 资助的研究 EAA171 提交验证文件作为 FDA IDE 申请的一部分时，FDA 要求进行额外的验证实验，因为 MRD 结果将用于 MM 患者分层。经过几次会议以准确定义其他要求，我们与 FDA 就行动计划达成一致。在这里，我们总结了 FDA 为批准 IDE 在临床试验中使用 NGF 而要求的其他实验。

分析物、样本和/或基质稳定性：将另外 4 个骨髓样品在环境温度下储存在运输容器中，并在 24、48、72 和 96 小时后进行测试。来自接受时间点的结果必须在质量上相同 （MRD 阳性或阴性）。最后一个可接受的时间点定义为变异系数 （CV） ≤25% 的时间点。数据表明，在环境运输条件下，可以在绘制后 96 小时内可靠地评估 MRD 检测（表 S1）。

精度/重现性：将另外 3 个 MM 样品加标到正常骨髓样品中，计算出的肿瘤负荷在 5 × ^10-6 和 5 × ^10-5 之间。每个样品共进行 8 次重复。计算 CV 以比较分析内、分析间、操作员间和仪器间精密度。每个重复必须在质量上相同，所有 CV ≤25%。标准化的机器设置和标准化的门控策略显示，接近检测限 （LOD） 的高精度检测（表 S2a 和 S2b）。

准确性：使用 21 MM 样品比较 NGF 和基于 NGS 的 clonoSEQ® 检测的 MRD 结果，使用分流样品程序。用两种不同浓度的加标异常细胞测试了 3 个样品。另外 13 个诊断标本也是该比较的一部分。结果显示 NGF 与 clonoSEQ® NGS 检测对 MRD ≥^10-5 的 100% 一致性。此外，10 例敏感性在 2 × ^10-6 和 ^10-5 之间的 NGF 阳性病例中有 10 例被 NGS 确认为阳性。4 例 NGF 阴性病例的 NGS 阳性率非常低（3-4 × ^10-6）（图 1 和表 S3）。总体 Pearson 相关系数为 0.83。

分析灵敏度/检测限：将 4 个 MM 骨髓样本加标到正常骨髓中，以达到 ^10-3（基线）、^10-4、^10-5 和 5 × ^10-6（低于测定 LOD 的半对数）的计算肿瘤负荷，一式三份，预期 CV ≤25%。两个样品显示精确的 MRD 阳性 LOD 低至 2 × ^10-6，另外两个样品低至 ^10-5（表 S4a 和 S4b）。

总之，我们描述了 NGF 成功批准 MM MRD 测试作为 FDA IDE，用作临床试验中患者分层的决策工具（表 S5 中列出了包含提交研究/文件的数据包内容）。FDA 的两个主要要求是 （I） 显示可靠的 MRD 测试，灵敏度比测定规定的 LOD 高半个对数级;（II） 显示 MRD 结果与 FDA 批准的 NGS 检测（在我们最初的 CLIA 检测验证和 IDE 提交之间的时间内获得批准）的直接相关性。我们能够在分检样本队列上实现 NGF 和 NGS 之间的极好相关性。我们的经验可以帮助其他实验室进行 IDE 提交和测试验证，尤其是在 FDA 监管 LDT 的新时代。此外，此处提供的数据可能会引起寻求 FDA 在 MM MRD 检测中对 NGF 的更高级别批准（PMA 或上市前通知 510（k）的专业协会和公司的兴趣，以及希望在 FDA 监管时代将 MM MRD 检测验证为 LDT 的个体临床实验室。