跳到正文
小南AI实验室·AI 评分68

BioPharma APAC 汇总 2026 财年 FDA 亚太药品工厂 45 封 GMP 警告信并按检查量归一化

2026 财年 FDA 亚太药企警告信台账及四大缺陷

AI 导读

BioPharma APAC 汇总 2026 财年(2025 年 10 月 1 日至 2026 年 9 月 30 日)FDA CDER 发往亚太九个国家和地区的 45 封药品 GMP 警告信,其中印度 20 封、中国大陆 15 封。

正文

摘要

1. 2026 财年(2025 年 10 月 1 日至 2026 年 9 月 30 日),FDA 药品评价与研究中心(CDER)向亚太九个国家和地区发出 45 封药品 GMP 警告信。印度 20 封、中国大陆 15 封,合计 35 封。

2. 按每百次检查归一化后,印度 8.2 封、中国大陆 4.9 封,亚太其余地区至少 10.6 封。信件少的地区不是工厂更干净,而是 FDA 去得更少。

3. 45 封里 34 封来自现场检查,11 封来自记录审阅(704(a)(4) 条)或企业拒绝配合,中国占这 11 封中的 7 封。

4. 主要涉及四大缺陷:质量部门、数据完整性、放行检测、无菌。

FDA Warning Letters to APAC Drug Facilities in Fiscal 2026: The Full Ledger, Normalised

2026-10-20 | BioPharma APAC 分析 | 中英对照

一、谁都会引用的那组数字

The count everyone will quote

Every October, the same headline writes itself. India leads the count of US FDA warning letters to Asian drug plants, China follows, and the rest of the region barely registers. The headline is accurate. It is also close to useless, because a raw count of letters tells you more about where FDA sends its investigators than about how those plants behave once the investigators arrive.

每年十月,同样的标题会自动写出来:美国 FDA 发给亚洲药品工厂的警告信,印度最多,中国次之,亚太其余地区几乎不值一提。这个标题是准确的,也几乎没用。因为一封封数出来的信件总数,说明的是 FDA 把检查员派到了哪里,而不是检查员到了之后那些工厂表现如何。

This ledger is an attempt to fix that. It lists every drug good manufacturing practice (GMP) warning letter that FDA's Center for Drug Evaluation and Research (CDER) issued to a facility in Asia-Pacific between 1 October 2025 and 30 September 2026, the US government's fiscal year 2026. Each entry was read on FDA's own letter page. Then, where the letter came out of an inspection, we set it against the number of drug quality inspections FDA conducted in that country, so that a plant in Gujarat and a plant in Gyeonggi-do are measured on the same scale.

这份账本想解决的就是这个问题。它逐条列出 2025 年 10 月 1 日至 2026 年 9 月 30 日(美国联邦政府 2026 财年)期间,FDA 药品评价与研究中心(CDER)发给亚太地区设施的每一封药品 GMP 警告信。每一条都在 FDA 自己的警告信页面上读过。凡是从现场检查出来的信件,再把该国当年的药品质量检查次数放到分母上,让古吉拉特邦的工厂和京畿道的工厂用同一把尺子衡量。

The result is 45 letters across nine jurisdictions. The full table sits at the end of this article, and the method, including its limits, is published alongside it. This is the first edition of an annual franchise, and we intend the second to be better for the corrections the first attracts.

结果是九个国家和地区、45 封警告信。完整表格在文末,方法说明连同它的局限一起公开。这是年度系列的第一版,第二版会因为有第一版收到的更正而变得更好。

二、原始计数为什么会误导

Why the raw count misleads

A warning letter is the end of a chain. FDA first decides which plants to look at, sends investigators or requests records, classifies what it finds, and only then, for the most serious outcomes, writes a letter. A country that hosts more FDA-registered plants, and therefore receives more inspections, will generate more letters even if its plants are no worse than anyone else's. India and China host the largest pools of foreign sites supplying the US market, so they will always top a raw count.

警告信是一条链条的末端。FDA 先决定看哪些工厂,派人到场或索取记录,对发现的问题分类,只有最严重的结果才会写成信。注册工厂多、因而被检查得多的国家,自然会产生更多警告信,哪怕它的工厂并不比别处差。印度和中国拥有的对美供货境外工厂数量最多,所以在原始计数上永远排在最前面。

The fair question is not how many letters a country received but how many letters it received for every inspection FDA carried out there. That is the number this ledger reports, as letters per 100 inspections.

公平的问法不是某国收到多少封警告信,而是 FDA 在该国每做一次检查对应多少封信。账本给的就是这个数字:每百次检查的警告信数。

Two adjustments make the comparison honest. First, only letters that arose from an inspection go into the numerator. Nine of the 45 letters were written after FDA reviewed records it had demanded under section 704(a)(4) of the Food, Drug and Cosmetic Act, without anyone setting foot in the plant, and two more were issued because a firm refused to let FDA in or refused to hand over records. Dividing those letters by an inspection count would inflate the rate for countries where FDA leans on remote tools. We count them separately, and they carry their own lesson, discussed below.

两处调整让比较变得诚实。第一,只有来自现场检查的信件才进分子。45 封里有 9 封是 FDA 依据《食品、药品和化妆品法》第 704(a)(4) 条索取记录、审阅之后发出的,没有人踏进工厂;另有 2 封是因为企业拒绝 FDA 入场或拒绝交出记录。把这些信除以检查次数,会虚高那些 FDA 更依赖远程手段的国家的比率。我们把它们单独计数,这类信件本身也带着教训,后文再说。

Second, the denominator has to come from the right period. The median gap between the last day of an inspection and the date of the resulting letter in this ledger is 179 days, and the range runs from 123 days to 320. Of the 34 inspection-based letters, 14 stem from inspections that closed before the fiscal year began. A fiscal 2026 letter is, in large part, the product of fiscal 2025 fieldwork.

第二,分母必须来自正确的时期。本账本中,检查最后一天到警告信日期之间的中位间隔是 179 天,区间从 123 天到 320 天。34 封来自现场检查的信里,有 14 封对应的检查在本财年开始之前就已结束。一封 2026 财年的警告信,很大程度是 2025 财年现场工作的产物。

三、归一化之后

The normalised view

FDA has not yet published country-level inspection counts for fiscal 2026. The most recent official figures are in CDER's Report on the State of Pharmaceutical Quality for fiscal 2025, released on 29 June 2026, which counts 1,248 drug quality inspections worldwide: 702 in the United States and 546 abroad, including 207 in India and 163 in China. Given the lag described above, that fiscal 2025 cohort is a defensible denominator for this year's letters, and it is the one we use. When FDA's dashboard yields a clean fiscal 2026 count, we will publish the rerun.

FDA 尚未公布 2026 财年的国别检查次数。最新的官方数字在 CDER 2026 年 6 月 29 日发布的《2025 财年药品质量状况报告》里:全球药品质量检查 1,248 次,美国境内 702 次,境外 546 次,其中印度 207 次、中国 163 次。考虑到上文所说的滞后,2025 财年这组数据作为今年警告信的分母是站得住的,我们用的就是它。等 FDA 的数据看板给出干净的 2026 财年计数,我们会发布重算结果。

On that basis, India received 17 inspection-based letters against 207 inspections, or 8.2 letters per 100 inspections. China received eight inspection-based letters against 163 inspections, or 4.9 per 100. India's rate is higher, but by a factor of under two, not the gap the raw totals imply.

按这个口径,印度 207 次检查对应 17 封来自现场检查的警告信,每百次 8.2 封;中国大陆 163 次检查对应 8 封,每百次 4.9 封。印度的比率更高,但差距不到两倍,远小于原始总数给人的印象。

The more striking result is the rest of the region. FDA's report does not break out Japan, South Korea, Australia or South-East Asia individually; it folds them into a 'rest of world' group that recorded 85 inspections in fiscal 2025, a group that also includes sites outside Asia-Pacific. Nine inspection-based letters went to Asia-Pacific sites outside India and China. Even if every one of those 85 inspections had taken place in Asia-Pacific, which it did not, the rate would be 10.6 letters per 100 inspections. That is a floor, not an estimate, and it sits above both India and China.

更意外的结果在该区域的其余部分。FDA 的报告没有单列日本、韩国、澳大利亚或东南亚,而是把它们并进一个“世界其他地区”组,该组 2025 财年记录 85 次检查,其中也包含亚太以外的工厂。印度和中国以外的亚太工厂拿到 9 封来自现场检查的信。即使那 85 次检查全部发生在亚太(实际并非如此),比率也是每百次 10.6 封。这是下限而不是估计,并且高于印度和中国。

Put plainly, the parts of Asia-Pacific that rarely appear in warning letter headlines are not inspected often, but when FDA does go, it finds enough to write a letter at a rate at least as high as anywhere else in the region. Fewer letters reflect fewer visits, not cleaner plants.

那些很少出现在警告信标题里的亚太地区,检查次数确实不多。但 FDA 一旦到访,按比率写出的警告信并不比区域内任何地方少。信件少,说明到访少,不说明工厂更干净。

FDA's own inspection outcomes point the same way. In fiscal 2025, 17.9 per cent of Indian inspections were classified 'official action indicated', the agency's most serious outcome, against 8.0 per cent in China, 15.3 per cent in the rest of the world group and 21.6 per cent in the United States. The global figure was 18 per cent. Indian plants, in other words, were found wanting at a rate slightly below the global average and well below American plants on FDA's own scoring, even as they collected the largest single share of letters in this region.

FDA 自己的检查结果指向同一方向。2025 财年,印度受检工厂有 17.9% 被列为“需官方处置”(official action indicated),这是该机构最严重的一档;中国为 8.0%,世界其他地区组为 15.3%,美国为 21.6%,全球为 18%。按 FDA 自己的评分,印度工厂被判不合格的比率略低于全球平均,明显低于美国工厂,而它们同时拿走了本区域最大份额的警告信。

四、逐个国家与地区

Country by country

India's 20 letters were spread across ten states and union territories, which itself is a finding. Maharashtra led with four, from Chemspec Chemicals and Gopaldas Visram in Navi Mumbai to Shimoga Chemicals in Sangli and Reliance Life Sciences in Rabale. Telangana, Gujarat and Karnataka had three each, and the coastal union territory that holds Daman and Silvassa had two. The mix was broad as well: seven API makers, six OTC producers, three prescription finished dose sites, two sterile manufacturers and two contract testing laboratories. The two laboratories matter more than their number suggests, because FDA treats contract labs as extensions of every manufacturer that relies on them, and neither letter names those clients.

印度的 20 封分布在十个邦和联邦属地,这本身就是一项发现。马哈拉施特拉邦最多,四封,从新孟买的 Chemspec Chemicals、Gopaldas Visram,到桑格利的 Shimoga Chemicals 和拉巴莱的 Reliance Life Sciences。特伦甘纳、古吉拉特、卡纳塔克各三封,辖有达曼与锡尔瓦萨的沿海联邦属地两封。类型也很散:原料药厂七家、非处方药厂六家、处方成品药厂三家、无菌制剂厂两家、合同检测实验室两家。两家实验室的分量超过数量本身,因为 FDA 把合同实验室当作每一家依赖它的生产商的延伸,而这两封信都没有点名那些客户。

China's 15 tell a different story. Nine went to OTC makers and six to API producers; not one went to a prescription finished dose plant. Guangdong accounted for five and Jiangsu for four, largely consumer health manufacturers making sanitisers, wipes, sunscreens and talc products. Nearly half of China's letters came from paperwork or refusal rather than an inspection, which is why its normalised rate is the lowest in the region and also why that rate should be read with care. A plant that refuses an inspection never enters the denominator at all.

中国大陆的 15 封讲的是另一个故事。九封发给非处方药厂,六封发给原料药厂,没有一封发给处方成品药厂。广东五封、江苏四封,多是生产消毒液、湿巾、防晒品和滑石粉产品的消费健康类制造商。中国大陆近一半的警告信来自书面材料或拒绝配合,而不是现场检查。这既是它归一化比率全区域最低的原因,也是这个比率需要谨慎解读的原因:拒绝检查的工厂根本不会进入分母。

The rest of the region produced ten letters, and two of them deserve particular attention from partners. Huons, a listed Korean injectables manufacturer, received a letter describing some of the most detailed data integrity failures in the entire ledger. Sato's letter concerned aseptic filling, an area where Japanese manufacturers are usually assumed to be strong. Australia's two letters both went to Queensland skincare manufacturers making sunscreen-type products, and Malaysia's single letter, to Fulijaya Manufacturing in Kedah, is one of only four outside India in the ledger that centre on data integrity. Taiwan and Macau each recorded one OTC letter, and New Zealand's went to a homeopathic laboratory with a repeat finding on component testing.

亚太其余地区合计十封,其中两封值得合作方特别注意。韩国上市注射剂企业 Huons 收到的信,描述了整份账本中最详细的数据完整性失效之一。Sato 的信涉及无菌灌装,这通常是外界认为日本厂商最强的领域。澳大利亚两封都发给了昆士兰生产防晒类产品的护肤厂商;马来西亚唯一一封发给吉打州的 Fulijaya Manufacturing,是账本中印度以外仅四封以数据完整性为核心的信件之一。中国台湾和中国澳门各有一封非处方药警告信,新西兰那封发给一家顺势疗法实验室,涉及组分检测的重复缺陷。

五、没有检查员的警告信

Letters without an investigator

The 11 letters that did not come out of an inspection deserve their own attention, because they show where FDA's enforcement is heading.

那 11 封不是来自现场检查的警告信值得单独看,它们指示了 FDA 执法重心的走向。

China accounts for seven of them. Five Chinese firms received letters after FDA reviewed records sent under 704(a)(4): two OTC makers in Guangdong, a talc product maker in Shantou, and two API producers, Hangzhou Yiqi Biotechnology and Tianjin Kilo Pharmaceutical Sci-tech, the latter shipping to US compounding pharmacies. Two more were adulterated by refusal. Kellin Chemicals in Zhangjiagang declined a preannounced inspection in July 2025, and Yangzhou Hongshengding Chemical left two records requests unanswered. India had three records-based letters, all from the talc sweep, to Gopaldas Visram, Umendra Life Sciences and the Baddi site of Zydus Lifesciences. Japan's Asanuma Corporation was the fourth talc recipient in the region on records alone.

其中中国大陆占七封。五家中国企业是在 FDA 审阅 704(a)(4) 条下报送的记录之后收到信件的:广东两家非处方药厂、汕头一家滑石粉产品厂,以及两家原料药企业,杭州的 Yiqi Biotechnology 和天津的 Kilo Pharmaceutical Sci-tech,后者供货给美国的配制药房。另有两封因拒绝配合而被认定为掺假。张家港的 Kellin Chemicals 在 2025 年 7 月拒绝了一次预先通知的检查,扬州的 Hongshengding Chemical 对两次记录索取未作回应。印度有三封基于记录的信,全部出自滑石粉那一批,发给了 Gopaldas Visram、Umendra Life Sciences 和 Zydus Lifesciences 的 Baddi 工厂。日本的 Asanuma Corporation 是该区域第四家仅凭记录收到滑石粉相关警告信的企业。

For a quality head, the implication is uncomfortable. A records request arrives by email, and the response is judged on paper. Certificates of analysis that have never been verified, identical infrared readings on samples that should differ, and missing asbestos tests on talc were all visible to FDA reviewers who never left Maryland. A site that has prepared for the walk-through but not for the file review is preparing for the wrong exam.

对质量负责人来说,这个含义并不舒服。记录索取通过电子邮件到达,回应则按纸面判定。从未被核实过的检验报告、本该有差异却给出相同红外图谱的样品、滑石粉缺失的石棉检测,全都被从未离开马里兰的 FDA 审评员看在眼里。一个只准备了现场走查、没准备文件审阅的工厂,准备错了考试。

六、FDA 究竟引用了哪些条款

What FDA actually cited

Read the letters themselves and the citations cluster around a small number of failures that recur across borders.

逐封读下来,引用条款集中在少数几类缺陷上,且跨国重复出现。

The first is the quality unit. In letter after letter, FDA cites 21 CFR 211.22, the requirement that a site's quality function has the authority and the procedures to stop bad product. For API makers, the equivalent finding is framed under the ICH Q7 guideline. The failure takes many forms: Chemspec Chemicals in Navi Mumbai released APIs without batch records; Harbin Jixianglong Biotech's quality unit allowed relabelling of semaglutide from a source not on FDA's green list; Intas's Dehradun site let a vendor change electronic batch records without an audit trail.

第一类是质量部门。一封接一封,FDA 引用 21 CFR 211.22,即工厂的质量职能必须具备阻止不合格产品放行的权限与规程。对原料药企业,同等缺陷按 ICH Q7 指南表述。表现形式很多:新孟买的 Chemspec Chemicals 在没有批记录的情况下放行原料药;哈尔滨的 Jixianglong Biotech 质量部门允许把来自 FDA 绿色清单之外来源的司美格鲁肽重新贴标;Intas 的德拉敦工厂允许供应商在无审计追踪的情况下修改电子批记录。

The second is data integrity, and here the region's record is sobering. At least 16 of the 45 letters describe records that were altered, backfilled, discarded, falsified or never generated. Twelve of those are Indian. Tentamus India, a contract testing laboratory in Hyderabad, was found with torn records in garbage bags and was also cited for delaying the inspection. Dabur India's Silvassa plant kept a falsified equipment logbook. Wizcure Pharmaa swapped contaminated environmental monitoring plates. At Huons in Jecheon, FDA describes discarded bioburden plates, altered camera timestamps and 1,897 blank, uncontrolled forms. At Reliance Life Sciences in Navi Mumbai, microbiology samples were logged as collected when they had not been.

第二类是数据完整性,这一项上本区域的记录令人警醒。45 封里至少 16 封描述了被篡改、回填、丢弃、伪造或从未生成的记录,其中 12 封是印度企业。海得拉巴的合同检测实验室 Tentamus India 被发现把撕毁的记录装在垃圾袋里,还因拖延检查被另行列明。Dabur India 的锡尔瓦萨工厂保有一本伪造的设备日志。Wizcure Pharmaa 调换了被污染的环境监测培养皿。在堤川的 Huons,FDA 描述了被丢弃的生物负载培养皿、被改动的摄像头时间戳,以及 1,897 份空白且不受控的表单。在新孟买的 Reliance Life Sciences,微生物样品被登记为已采集,实际并未采集。

The third is testing. A large share of the OTC letters, which account for 24 of the 45, cite the same pair of failures: no identity or strength testing before release under 211.165, and reliance on supplier certificates of analysis without verifying components under 211.84. These are basic obligations, and their frequency says something about how many small consumer product makers in the region supply the US market without fully understanding that a sunscreen or an antiseptic wipe is a drug in American law.

第三类是检测。非处方药类警告信占 45 封中的 24 封,其中很大一部分引用了同一对缺陷:放行前未按 211.165 做鉴别或含量检测;按 211.84 应核实组分,却直接依赖供应商的检验报告。这些都是基本义务,它们出现的频率说明,本区域有相当多小型消费品制造商在对美供货,却没有完全理解在美国法律下防晒霜或消毒湿巾属于药品。

The fourth is sterility. Five of the OTC letters involved products labelled sterile or aseptically filled, and in two Chinese cases FDA found products labelled sterile that had not been sterilised at all. In the prescription space, three letters went to sterile manufacturers: Eugia's Unit 1 in Telangana, where FDA said Units II and III were also classified 'official action indicated'; Reliance Life Sciences; and Huons. Sato's Hachioji plant in Tokyo was cited for six failed media fills between November 2022 and February 2025.

第四类是无菌。五封非处方药警告信涉及标示无菌或无菌灌装的产品,其中两封中国案例里,FDA 发现产品标示无菌却根本没有经过灭菌。处方药领域有三封发给无菌制剂企业:Eugia 在特伦甘纳的一号厂,FDA 同时把其二号、三号厂列为“需官方处置”;Reliance Life Sciences;以及 Huons。Sato 在东京八王子市的工厂被指出,2022 年 11 月至 2025 年 2 月间有六次培养基灌装失败。

七、进口警示通常先落地

The import alert usually comes first

One pattern in the ledger changes how executives should read the warning letter calendar. In 26 of the 45 cases, FDA had already placed the site on Import Alert 66-40, which allows US border officers to detain its drugs without physical examination, before the warning letter was issued. A 27th, Yangzhou Hongshengding, was placed on the records-refusal alert 66-79 before its letter.

账本里的一个规律,会改变管理层读警告信日历的方式。45 个案例中有 26 个,在警告信发出之前,FDA 已把该工厂列入 66-40 号进口警示,该警示允许美国边境官员无须查验即扣留其药品。第 27 个,扬州的 Hongshengding,在被写信之前已列入针对拒绝提供记录的 66-79 号警示。

In practice, that means the commercial damage usually lands weeks or months before the public document. Reliance Life Sciences was on the import alert on 9 July 2026, six weeks before its letter. Dabur's Silvassa site was listed on 5 June and received its letter on 24 July. Patcos Cosmetics in Daman has been on the alert since November 2023, and its fiscal 2026 letter is its second. For investors and procurement teams, the warning letter is often the confirmation, not the signal. The import alert list is where the earlier signal sits.

实际效果是,商业损失通常在公开文件之前数周或数月就已经发生。Reliance Life Sciences 在 2026 年 7 月 9 日进入进口警示,比警告信早六周。Dabur 的锡尔瓦萨工厂 6 月 5 日上榜,7 月 24 日收到警告信。达曼的 Patcos Cosmetics 自 2023 年 11 月起就在警示名单上,2026 财年这封是它的第二封。对投资人和采购团队来说,警告信往往只是确认,不是信号。更早的信号在进口警示名单上。

Recalls followed in several cases. Shimoga Chemicals in Sangli recalled clomiphene citrate in May 2026; Reliance recalled all sterile products in June; Dabur and Shoolin Pharma Chem recalled in June and August respectively; Huons committed to a voluntary recall; and Wizcure halted US production.

多个案例随后发生召回。桑格利的 Shimoga Chemicals 在 2026 年 5 月召回枸橼酸氯米芬;Reliance 在 6 月召回全部无菌产品;Dabur 与 Shoolin Pharma Chem 分别在 6 月和 8 月召回;Huons 承诺自愿召回;Wizcure 停止了面向美国的生产。

八、配制用药的供应链

The compounding supply chain

A smaller thread, but one with policy weight, runs through the API letters. At least six of the 13 API letters went to firms supplying material destined for compounding or for peptide products in high US demand. Harbin Jixianglong made semaglutide and tirzepatide. Tianjin Kilo and Henan Lvyuan Pharmaceutical supplied APIs for compounding. Hangzhou Yiqi supplied APIs used in sterile products and compounding. Almon Healthcare in Ahmedabad was found to have knowingly accepted deliberately mislabelled material, and Shimoga's clomiphene was also for compounding. Shoolin's sildenafil and tadalafil add to the picture.

原料药警告信里还有一条更细的线,但有政策分量。13 封原料药警告信里至少 6 封发给了为配制用药或美国需求旺盛的多肽产品供货的企业。哈尔滨的 Jixianglong 生产司美格鲁肽和替尔泊肽;天津的 Kilo 与河南的 Lvyuan Pharmaceutical 为配制用药供应原料药;杭州的 Yiqi 供应用于无菌产品和配制的原料药。艾哈迈达巴德的 Almon Healthcare 被认定明知而接受了故意错误标示的物料,Shimoga 的氯米芬同样用于配制。Shoolin 的西地那非和他达拉非再添一例。

These are not the large generic exporters that dominate India's US business. They are smaller producers feeding a channel that sits outside the conventional approved-application pathway, and FDA's attention to them in fiscal 2026 tracks the agency's broader concern with compounded GLP-1 products and their inputs.

这些不是主导印度对美业务的大型仿制药出口商,而是规模更小的生产者,供应的是常规已批准申请路径之外的一条渠道。FDA 在 2026 财年对它们的关注,与该机构对配制 GLP-1 产品及其投入物的整体担忧一致。

九、没有一封结案

Nothing has closed

Not one of the 45 letters had been formally closed out by FDA as of 20 October 2026. That is not unusual, since close-out typically requires a follow-up inspection or a thorough file review, and the youngest letters in the ledger are only weeks old. But it does mean that every site in this table remains, in FDA's public record, an open case. FDA may withhold approval of new applications that name these sites as manufacturers until the violations are resolved, which turns an operational problem into a pipeline problem for any partner that relies on them.

截至 2026 年 10 月 20 日,45 封警告信没有一封被 FDA 正式结案。这并不异常,结案通常需要一次跟踪检查或一次彻底的文件审阅,而账本里最新的信件只有几周。但它说明表上每一家工厂在 FDA 的公开记录中都是未结案件。在缺陷解决之前,FDA 可以扣住把这些工厂列为生产商的新申请不予批准;对任何依赖它们的合作方,运营问题由此变成管线问题。

Close-outs did happen during the year for older letters. FDA closed out an earlier letter to Global Calcium in July 2026 and a records-refusal letter to Zenzi Pharmaceutical Industries in September, though Zenzi remains on the records import alert. They show the route out exists. They also show it is slow.

年内也有旧信结案。FDA 在 2026 年 7 月对 Global Calcium 早前的一封警告信结案,9 月对 Zenzi Pharmaceutical Industries 的拒绝提供记录警告信结案,但 Zenzi 仍在记录类进口警示名单上。这说明出路存在,也说明它很慢。

十、作为买方或合作方怎么读这份账本

Reading the ledger as a buyer or partner

For a sponsor, licensor or procurement head, three practical questions follow from the table. The first is exposure: does any site in the ledger appear in your supplier list, your regulatory filings or your contract testing arrangements? Because FDA can withhold approval of applications that name a cited site, a single API supplier or testing laboratory on this list can stall a filing that has nothing else wrong with it.

对申办方、许可方或采购负责人,这张表带出三个实际问题。第一是暴露面:账本里的工厂是否出现在你的供应商名单、注册申报资料或委托检测安排中?由于 FDA 可以扣住列有被点名工厂的申请,名单上一个原料药供应商或检测实验室,就足以让一份其他方面毫无问题的申报停摆。

The second is timing. Since the import alert preceded the letter in most cases, a supplier that appeared on Import Alert 66-40 during the year without a letter yet may simply be further back in the same queue. Watching the alert list is a better early warning than waiting for the letter to be posted, which, as Kellin Chemicals' eleven-month gap between letter date and publication shows, can take a long time.

第二是时间差。多数情况下进口警示先于警告信,所以年内出现在 66-40 号警示名单上、却还没收到信的供应商,可能只是排在同一条队列的更后面。盯警示名单比等警告信挂网更早。Kellin Chemicals 从信件日期到公开发布相隔十一个月,这个等待可以很长。

The third is the paper trail. Ask suppliers how they would answer a 704(a)(4) request tomorrow, and whether their certificates of analysis rest on testing they actually performed. In fiscal 2026, that question alone would have flagged a large share of the region's letters.

第三是纸面证据链。问供应商,如果明天收到 704(a)(4) 条记录索取会怎么答;再问他们的检验报告是否建立在自身实际完成的检测之上。在 2026 财年,仅这两个问题就能标出本区域相当一部分警告信。

十一、这份账本不主张什么

What this ledger does not argue

This is a ledger, not a thesis. It does not argue that Indian or Chinese manufacturing quality is improving or declining, and it should not be read as doing so. One fiscal year of letters is too small and too lagged a sample to carry that argument, and the argument about Indian quality in particular belongs to our Fault Lines series, which can cite this table as its evidence base.

这是一份账本,不是一篇论点。它不主张印度或中国的制造质量在改善或下滑,也不应被读成那个意思。一个财年的警告信样本太小、滞后太严重,撑不起这样的判断;关于印度质量的具体讨论放在原文的 Fault Lines 系列里,那一系列可以把这张表当作证据基础。

What the ledger does establish is narrower and, we think, more useful. Raw counts overstate India's position relative to its inspection volume. China's normalised rate is lower but its share of letters written without any inspection is far higher. And the smaller markets of the region, from Korea and Japan to Australia and Malaysia, are not quieter because they are better, but because FDA visits them less.

账本确立的东西更窄,我们认为也更有用。原始计数高估了印度相对于其检查量的位置;中国大陆的归一化比率更低,但完全没有现场检查的警告信占比远高于别处。区域里较小的市场,从韩国、日本到澳大利亚、马来西亚,之所以安静,不是因为它们更好,而是因为 FDA 去得更少。

十二、2027 财年看点

What to watch in fiscal 2027

The new fiscal year is already two days into its own ledger. On 2 October 2026, FDA issued a letter to Hanlim Pharm in Yongin, South Korea, over sterile manufacturing, repeating findings from a 2018 letter. It falls in fiscal 2027 and will open next year's table.

新财年已经有了两天的账。2026 年 10 月 2 日,FDA 就无菌生产向韩国龙仁的 Hanlim Pharm 发出警告信,重复了 2018 年一封警告信里的缺陷。它属于 2027 财年,将是明年那张表的第一行。

Three things will shape that table. The first is FDA's stated push to expand unannounced foreign inspections, which, if sustained, will raise the inspection count in India and China and test whether the rates reported here hold. The second is the talc sweep, which has not finished, and which falls hardest on OTC makers that never thought of themselves as pharmaceutical manufacturers. The third is the records request. If 704(a)(4) reviews keep growing as a share of enforcement, the denominator for the next edition will need a second column, letters per 100 records requests, and FDA will need to publish the numbers that make it possible.

三件事会塑造那张表。第一是 FDA 公开提出的扩大境外突击检查,如果持续下去,会抬高印度和中国的检查次数,并检验本文报告的比率是否站得住。第二是尚未结束的滑石粉清查,它打击最重的是那些从不把自己当作药品生产商的非处方药厂。第三是记录索取。如果 704(a)(4) 审阅在执法中的占比继续上升,下一版的分母就需要增加一列“每百次记录索取的警告信数”,而 FDA 需要公布让这一列可算的数字。

十三、账本一览

The ledger at a glance

辖区

警告信

FY2025 检查

每百次

印度
India

20
检查 17 · 记录 3

207

8.2

中国大陆
China, mainland

15
检查 8 · 记录 7

163

4.9

亚太其余地区*
Rest of Asia-Pacific

10
检查 9 · 记录 1

85 或更少

10.6 或更高

合计
Total

45
检查 34 · 记录 11

—

—

* 韩国 2、日本 2、澳大利亚 2、中国台湾 1、中国澳门 1、马来西亚 1、新西兰 1。FDA 将这七地并入“世界其他地区”组(85 次检查,含亚太以外工厂),故所示比率为下限。

原文配图给出的产品类型分布:非处方成品 24、原料药 13、处方成品 3、无菌制剂 3、检测实验室 2,合计 45。

表为原文数据,中文字段为本文补译。检查次数取自 CDER《2025 财年药品质量状况报告》,不是 2026 财年数据;比率是据此推算的归一化结果,不等于 FDA 官方指标。

附录、方法说明

Method

范围 Scope

Scope. Warning letters issued by CDER (including letters co-signed by CDER's Office of Manufacturing Quality) for violations of drug current good manufacturing practice, dated 1 October 2025 to 30 September 2026 inclusive, addressed to a manufacturing, API or contract testing site in Asia-Pacific. Asia-Pacific here means South Asia, East Asia, South-East Asia, Australia and New Zealand. Letters for compounding pharmacies, devices, food, tobacco, clinical research and marketing claims without GMP findings are excluded. Letters for refusal of inspection or records are included because they adulterate drugs under the same statute; they are flagged.

范围。CDER 发出的、针对药品现行 GMP 违规的警告信(含 CDER 生产质量办公室联署的信件),日期自 2025 年 10 月 1 日至 2026 年 9 月 30 日(含首尾),收件方为亚太地区的生产、原料药或委托检测工厂。此处的亚太指南亚、东亚、东南亚、澳大利亚和新西兰。配制药房、器械、食品、烟草、临床研究以及不含 GMP 缺陷的市场宣传类警告信不在统计内。拒绝检查或拒绝提供记录的信件计入,因为依据同一部法律它们使药品构成掺假,这类信件单独标记。

来源 Sources

Sources. Each letter was read on its fda.gov page. Letter date is the date on the letter, not the posting date, which can lag by weeks or months. Kellin Chemicals' letter, dated 6 November 2025, was not posted until 29 September 2026.

来源。每封信都在 fda.gov 的信件页面上读过。日期取信上标注的日期,不取挂网日期,后者可能滞后数周甚至数月。Kellin Chemicals 的信标注日期为 2025 年 11 月 6 日,直到 2026 年 9 月 29 日才挂出。

归一化 Normalisation

Normalisation. Numerator: inspection-based letters only (34 of 45). Denominator: drug quality inspections by country for fiscal 2025 from CDER's Report on the State of Pharmaceutical Quality FY2025 (India 207, China 163, rest of world 85). The fiscal 2025 cohort is used because the median lag from inspection close to letter is 179 days and FDA had not published fiscal 2026 country counts at the time of writing. The rest of Asia-Pacific figure is a lower bound, since FDA's rest of world group also contains non-Asia-Pacific sites. Records-based and refusal letters are reported separately. The regional inspection figures were read from a chart in FDA's report and cross-checked against its stated totals.

归一化。分子只取来自现场检查的警告信(45 封中的 34 封)。分母取 CDER《2025 财年药品质量状况报告》中的国别药品质量检查次数(印度 207、中国 163、世界其他地区 85)。采用 2025 财年数据,是因为从检查结束到警告信的中位滞后为 179 天,且写作时 FDA 尚未公布 2026 财年的国别计数。亚太其余地区的数字是下限,因为 FDA 的“世界其他地区”组里也包含亚太以外的工厂。基于记录和拒绝配合的信件单独报告。区域检查数字取自 FDA 报告中的图表,并与其公布的总数交叉核对。

结案状态与边界情形 Close-out and edge cases

Close-out status is as of 20 October 2026 and reflects whether FDA had posted a close-out letter. Henan Lvyuan Pharmaceutical's letter is listed by FDA under its Center for Veterinary Medicine but is co-signed by CDER's Office of Manufacturing Quality and covers APIs for human compounding; it is included and flagged. Removing it changes China's inspection-based rate from 4.9 to 4.3 per 100.

结案状态截至 2026 年 10 月 20 日,反映 FDA 是否已挂出结案信。河南 Lvyuan Pharmaceutical 的信在 FDA 网站上列于兽药中心名下,但由 CDER 生产质量办公室联署,且涉及人用配制用原料药,故计入并加注。若剔除该封,中国大陆基于现场检查的比率由每百次 4.9 变为 4.3。

完整性 Completeness

Completeness. FDA's warning letter search is not exhaustive and was rate-limited during compilation. We estimate the ledger captures roughly nine in ten qualifying letters. Corrections and additions are welcome and will be logged openly in the next edition.

完整性。FDA 的警告信检索并非穷尽,编纂期间还受到访问频率限制。原文估计账本覆盖了约九成符合条件的警告信。欢迎更正与补充,下一版会公开记录。

答辩权与编辑独立性 Right of reply

Right of reply. Every company named in this ledger is offered a right of reply of seven working days. Responses received are published in full as an addendum to this page, and any factual correction is made in the ledger with a dated note. Editorial independence. Inclusion in this ledger is determined solely by FDA's public record and the method above. No commercial relationship with BioPharma APAC has any bearing on whether, or how, a company appears.

答辩权。账本点名的每家公司均有七个工作日的答辩权。收到的回应会全文附在本页,任何事实性更正会带日期标注写入账本。编辑独立性。是否进入本账本仅由 FDA 公开记录及上述方法决定,与 BioPharma APAC 的商业关系对某家公司是否出现、以何种方式出现没有任何影响。

资料来源:BioPharma APAC,Analysis,2026 年 10 月 20 日,
原文:biopharmaapac.com/analysis/23/8583/fda-warning-letters-to-apac-drug-facilities-in-fiscal-2026-the-full-ledger-normalised.html
本文为非官方中文译本,仅供学习交流。译文如与英文原文存在出入,一律以英文原文为准。文中统计口径、数字与判断均属原文作者,非本文实测。

来源:小南AI实验室 · mp.weixin.qq.com