IIT 之问：中国研究者发起试验的数据，在 FDA 那里到底能办成什么The IIT Question: What China's Investigator-Initiated Data Can Actually Do at the FDA

在本周发表的一篇中国行业媒体访谈里，RNA 干扰的先行者之一 Stuart Milstein 谈到，他正探索与中国医院合作开展研究者发起研究，将其作为一条快速且成本相对较低的取得概念验证之路。他不是个例。他描述的，是当今授权市场上的通行做法。按医药魔方（PharmCube）的统计，2025 年中国来源资产共成交 157 笔，总额 1,357 亿美元，其中相当一部分的底下，是研究者发起试验的早期数据。这批数据如今经常摆在美国交易桌的对面，而决定其价值的问题，是在太平洋这一侧提出的：它们到了美国食品药品监督管理局（FDA）手里，到底值多少？答案白纸黑字写在《联邦法规汇编》里，而且它奖励那些在签约之前先读过它的人。In an interview with Chinese trade media published this week, Stuart Milstein, one of the pioneers of RNA interference, described exploring investigator-initiated studies with Chinese hospitals as a fast and comparatively inexpensive route to proof of concept. He is not an outlier; he is describing standard practice in today's licensing market. China-origin assets changed hands in 157 transactions totaling $135.7 billion in 2025, by PharmCube's count, and early data from investigator-initiated trials sit underneath a substantial share of them. Those data now sit routinely across the table in U.S. deal rooms, and the question that decides their value is asked on this side of the Pacific: what are they worth once they reach the U.S. Food and Drug Administration? The answer is written down, in the Code of Federal Regulations, and it rewards those who read it before they sign.

这扇门，就是 21 CFR 312.120。依据这一条，FDA 可以接受一项未在 IND 下开展的境外临床研究，用来支持一项 IND 或一项上市申请。在 FDA 眼里，中国的研究者发起试验正是这样一项研究。接受有两个条件。第一，研究必须依照良好临床规范（GCP）开展，该条对 GCP 的界定包括：研究开始前经独立伦理委员会审查批准，此后持续接受伦理审查，以及自愿给予并取得、留档的知情同意。第二，该机构必须能够核实数据，必要时可以赴现场检查。配套条文 21 CFR 314.106 管的是仅凭境外数据获批的情形，它加了一道实质检验：数据必须适用于美国人群和美国医疗实践。这两条里，没有任何一句要求研究必须在美国 IND 下开展。障碍不在试验的护照。在试验的文件，以及这些文件经不经得起检查。The door is 21 CFR 312.120. Under it, the FDA will accept a foreign clinical study that was not conducted under an IND — which is what a Chinese investigator-initiated trial is, in the agency's eyes — in support of an IND or of a marketing application, provided two conditions are met. The study must have been conducted in accordance with good clinical practice, which the provision defines to include review and approval by an independent ethics committee before the study begins, continuing ethics review thereafter, and freely given informed consent, obtained and documented. And the agency must be able to validate the data, through an on-site inspection if it deems one necessary. A companion provision, 21 CFR 314.106, governs approval based solely on foreign data and adds a substantive test: the data must be applicable to the U.S. population and to U.S. medical practice. Nothing in either provision requires the study to have been run under a U.S. IND. The barrier is not the trial's passport. It is the trial's paperwork — and whether anyone can inspect it.

在这些规则之内使用，研究者发起数据能买到的是实打实的东西：支持在美国开启一项 IND，选定剂量，扎实到足以支撑一场 pre-IND 会议的概念验证，以及在许可谈判中的分量。最典型的案例是一款细胞疗法。LEGEND-2 是一项在中国开展的研究者发起的首次人体研究，它产生的数据促成了强生旗下的 Janssen 于 2017 年 12 月与传奇生物（Legend Biotech）签订许可与合作协议，其后美国试验所用的剂量也参考了这项研究。这里必须说精确：FDA 在 2022 年 2 月批准这款疗法时，即 cilta-cel，商品名 Carvykti，批准所依据的是申办方自行开展的 CARTITUDE-1，而不是 LEGEND-2。IIT 打开了门，并为剂量提供了参考。它没有扛起注册。这样的分工才是常态，任何把它说糊涂的叙述，都会误导买方。Used within those rules, investigator-initiated data buy things of substance: support for opening a U.S. IND, the selection of a dose, proof of concept strong enough to anchor a pre-IND meeting, and value in a licensing negotiation. The canonical case is a cell therapy. LEGEND-2, an investigator-initiated, first-in-human study conducted in China, produced the data that prompted Janssen's December 2017 license and collaboration agreement with Legend Biotech, and the dose carried into the U.S. trial that followed was informed by it. Precision matters here: when the FDA approved that therapy — cilta-cel, sold as Carvykti — in February 2022, the approval rested on the sponsor-run CARTITUDE-1 trial, not on LEGEND-2. The IIT opened the door and informed the dose; it did not carry the registration. That division of labor is the normal one, and any account that blurs it misleads a buyer.

一个更近的案例离这条线更近。2026 年 9 月，IASO Bio 宣布，FDA 已放行其 IASO208 的 IND 申请。IASO208 是一款体内 CAR-T 疗法，凭一项在中国开展的研究者发起试验的早期数据，该项目可直接进入美国 Ib 期研究，那项试验治疗了 11 名患者。这是公司自己的说法。FDA 不公布 IND 的放行情况，这一说法无法独立证实。但即使按面值接受，它也与法规相符，因为研究者发起数据最有分量的地方，恰恰在进入美国体系的入口处。它们几乎从不提供的，是批准所需的关键性证据。一项上市申请必须建立在充分且良好对照的研究之上，21 CFR 314.126 对此有明确界定：预先设定的目标和分析方法、与对照的有效比较，以及将偏倚降至最低的措施。多数研究者发起研究按其设计就是单臂的，并由研究者自行分析。它们生来是为了快速学习，不是为了注册产品，新闻稿如何措辞都改变不了这一点。A more recent case sits closer to the line. In September 2026, IASO Bio announced that the FDA had cleared its investigational new drug application for IASO208, an in vivo CAR-T therapy, allowing the program to proceed directly into a Phase Ib study in the United States, supported by early data from an investigator-initiated trial in China that had treated eleven patients. That is the company's account. The FDA does not publish IND clearances, and the claim cannot be independently confirmed; taken at face value, however, it is consistent with the regulation, because investigator-initiated data are at their strongest exactly here, at the entrance to the U.S. system. What they almost never supply is the pivotal evidence for an approval. A marketing application must rest on adequate and well-controlled investigations, as 21 CFR 314.126 defines them: prespecified objectives and methods of analysis, a valid comparison with a control, and measures that minimize bias. Most investigator-initiated studies are single-arm and investigator-analyzed, by design. They are built to learn quickly, not to register a product, and no drafting of a press release changes that.

这类数据最终能有多大分量，取决于适应症。近年的记录按类别来读，前后是一致的。在疾病于美国罕见、在亚洲多发且没有已批准疗法的地方，该机构曾凭中国数据批准药物：2023 年 10 月批准用于鼻咽癌的特瑞普利单抗（toripalimab），其试验被 FDA 本人描述为单区域、单国家的试验；2025 年 4 月批准用于同一种癌症的派安普利单抗（penpulimab），其两项支持性试验中，有一项全部在中国的中心开展。到了适应症庞大、竞争拥挤且已有良好治疗的地方，答案则是否定的。信迪利单抗（sintilimab）用于一线肺癌的申请，2022 年 2 月在顾问委员会以 14 比 1 的表决结果，被要求补做一项证明其适用于美国患者的试验，随后在 3 月收到完整回复函，函中建议开展多区域研究。索凡替尼（surufatinib）则在 2022 年 5 月收到完整回复函，尽管其申办方自述，两年前的一次 pre-NDA 会议上，它曾得到这样的理解：两项中国 III 期试验加一项美国桥接研究，可以支持一次申报。How much weight such data ultimately carry depends on the indication, and the recent record is consistent once it is read by category. Where a disease is rare in the United States, prevalent in Asia, and without an approved therapy, the agency has approved on China data: toripalimab for nasopharyngeal carcinoma in October 2023, on trials the FDA itself described as single-region and single-country; and penpulimab for the same cancer in April 2025, with one of its two supporting trials conducted entirely at sites in China. Where the indication is large, crowded, and already well served, the answer has been no. The application for sintilimab in first-line lung cancer drew a 14-to-1 advisory committee vote in February 2022 in favor of requiring an additional trial demonstrating applicability to U.S. patients, followed in March by a complete response letter recommending a multi-regional study. Surufatinib received a complete response letter in May 2022, even though its sponsor had understood from a pre-NDA meeting two years earlier that two China Phase III trials and a U.S. bridging study could support a submission.

任何认真的叙述，都应当带上两处纠正。ORIENT-11 是信迪利单抗那次决定背后的试验，它是申办方发起的注册性试验，不是研究者发起试验。这件事常被误归为 IIT 的失败，而它真正的教训比那个误传更严格：在试验开展之前，就设计、终点和对照征询该机构的意见。同样，并非每一次拒绝都关乎临床数据。卡瑞利珠单抗（camrelizumab）联合 rivoceranib 的多次完整回复函，按相关公司自己的说法，归因于生产场地的缺陷和未能完成的检查，其中并未发现涉及临床疗效或安全性的问题。拒绝和批准一样，都要按原因来读。Two corrections belong in any careful account. ORIENT-11, the trial behind the sintilimab decision, was a sponsor-initiated registrational trial, not an investigator-initiated one; the episode is routinely misfiled as an IIT failure, and its true lesson — consult the agency on design, endpoints, and control before the trial is run — is stricter than the myth. Nor is every refusal about clinical data. The repeated complete response letters for camrelizumab plus rivoceranib were attributed by the companies to deficiencies at manufacturing sites and to inspections that could not be completed, with no concern identified about clinical efficacy or safety. Refusals, like approvals, have to be read by cause.

2025 年以来的政策，同时朝着两个方向移动。2025 年 6 月，FDA 叫停了将美国患者的活细胞送往中国及某些其他国家进行基因工程处理的新试验。方向正好相反，底层却是同一个问题：知情同意、转移与核查。2025 年 9 月，该机构以最终版形式采用了 ICH E6(R3) 良好临床规范指南，境外研究向其申报时必须达到的文件标准随之提高。2026 年 5 月，FDA 细胞与基因治疗办公室的代理主任在一个研究会议上警告，在美国以外开展首次人体研究，对申办方可能「适得其反」。这是发言，不是规则，但它如实指示了如今审查的着力点：可检查性、数据完整性，以及对美国人群的适用性。今年，众议院一个拨款委员会的报告还写入了这样的措辞，将禁止该机构在支持 IND 申请时考虑在中国中心产生的数据。这是委员会报告的语言，不具法律效力，报道时也应当如实标明这一点。真正管事的条文，21 CFR 312.120 和 314.106，至今未作修改。门没有关。只是过门时的检查，比以前更细了。Policy since 2025 has moved in both directions at once. In June 2025 the FDA halted new trials sending U.S. patients' living cells to China and certain other countries for genetic engineering — the opposite direction of travel, but the same underlying question of consent, transfer, and verification. In September 2025 the agency adopted the ICH E6(R3) good clinical practice guideline in final form, raising the documentation standard that foreign studies must meet when they are presented to the agency. In May 2026 the acting director of the FDA's office for cell and gene therapies warned, at a research meeting, that running first-in-human studies outside the United States may prove "counterproductive" for sponsors — remarks, not a rule, but a fair indication of the scrutiny now applied to inspectability, data integrity, and applicability to the U.S. population. And language in a House appropriations committee report this year would bar the agency from considering data generated at sites in China in support of IND applications; it is committee report language, it carries no force of law, and it should be reported as exactly that. The provisions that govern, 21 CFR 312.120 and 314.106, stand unamended. The door has not closed. The inspection on the way through it has grown more thorough.

对于评估一项以这类数据为基础的资产的申办方、投资人或收购方，实务上的结论与其说是科学的，不如说是合同的。第一，从第一个患者起，就按 21 CFR 312.120(b) 的清单来建这套数据：研究方案和预先设定的统计分析计划；伦理委员会的身份与决定；知情同意文件以及同意的取得方式；研究者的资质、培训记录与书面承诺；监查与稽查记录；以及全程可追溯的源数据，从病例报告表到医院病历，再到分析数据集，并按法规要求的期限保存。第二，在需要之前先把权利拿到手：调取源数据的权利、研究者与医院配合 FDA 检查的安排、引用数据的权利。必须把这一切交出来的，是申请人，不是研究者。第三，请独立一方将结果重新推导一遍，并按美国当前的标准治疗校准终点与对照，而不是按研究设计时的标准。第四，中国这一侧要单独处理：向境外一方提供人类遗传资源信息，牵涉备案，以及特定情形下的安全审查，这套要求自 2024 年 5 月起由国家卫生健康委员会主管；一项只按境内研究备案的试验，未必覆盖外国被许可方将其数据用于 FDA 申报的情形。第五，把真正要紧的问题带到正式的 pre-IND 会议上问，并留存纪要。官员在会议场合的鼓励性表态不是监管政策。信迪利单抗的记录里，也有这一课。For the sponsor, investor, or acquirer evaluating an asset built on such data, the practical conclusions are contractual as much as scientific. First, build the dataset to the checklist at 21 CFR 312.120(b) from the first patient: the protocol and a prespecified statistical analysis plan; the ethics committee's identity and decisions; the consent documents and how consent was obtained; investigator qualifications, training records, and written commitments; monitoring and audit records; and source data traceable from case report form to hospital record to analysis dataset, retained for the periods the regulation requires. Second, secure the rights before they are needed — access to source data, the cooperation of investigator and hospital with an FDA inspection, and a right of reference — because it is the applicant, not the investigator, who must produce all of it. Third, have the results independently re-derived, and calibrate endpoints and comparators against the current U.S. standard of care, not the standard in place when the study was designed. Fourth, clear the China side separately: providing human genetic resources information to foreign parties engages filing and, in specified cases, security-review requirements administered since May 2024 by the National Health Commission, and a study filed as a purely domestic one may not cover a foreign licensee's use of its data in an FDA submission. Fifth, ask the agency the questions that matter in a formal pre-IND meeting, and keep the minutes. Encouraging remarks by officials at conferences are not regulatory policy; the sintilimab record contains that lesson as well.

Megs 于 2010 年创立，自 2016 年起从事跨境交易。我们标题所说的那项工作，连接中国与美国之间的项目方与投资人，越来越多地落在这样的记录上：它们在一个体系里产生，在另一个体系里受评判，在有人认真细读之前，从对岸几乎看不见。一份试验记录中要紧的东西，从太平洋对岸往往看不见，两个方向都是如此。我们的角色依然很小：帮助每一方弄清楚，对方到底在看什么。Megs was founded in 2010 and has worked on cross-border transactions since 2016. The work described by our headline — Connecting Project Sponsors and Investors Between China and the U.S. — increasingly turns on records of this kind: generated within one system, judged within another, and largely invisible from the other side until someone reads them closely. What matters in a trial record is often invisible from the other side of the Pacific, in both directions. Our role remains a modest one: to help each side understand what the other side is actually looking at.