清洁验证的三代演进:从10 ppm阈值到HBEL矩阵与阶段3数据监控
清洁验证的三代演进:从 10 ppm 阈值到 HBEL 矩阵与阶段 3 数据监控
Nandkishor Adhao在ISPE iSpeak长文中梳理清洁验证三代演进:1993年Fourman与Mullen提出10 ppm、1/1000剂量与目视洁净三项标准;第二代以HBEL和最差情形矩阵替代一刀切10 ppm;第三代落在3P前置设计、FAT核黄素荧光覆盖与阶段3 CPV数据监控。经验基线还可支撑CIP配方瘦身,削减漂洗时间与WFI用量。
摘要
1. 作者 Nandkishor Adhao 把清洁验证放回精益框架:设备清洁属业务非增值活动,却直接占用换批停机时间,流程设计与验证框架一旦僵化就是对产能征税。
2. 第一代是 1993 年 Fourman 与 Mullen 的三项标准(10 ppm、1/1000 剂量、目视洁净),把主观判断换成可测量的标准作业;第二代用 HBEL 与最差情形矩阵替掉一刀切的 10 ppm,把验证资源压到高风险节点。
3. 第三代落在前置设计与生命周期数据:3P 阶段定喷淋装置与罐体几何、FAT 做核黄素荧光覆盖;阶段 3 CPV 用漂洗电导率、在线 TOC、温度与回流浊度加 SPC,区分批间波动与喷嘴堵塞、泵压下降这类机械漂移。
4. 经验基线反过来支撑 CIP 配方瘦身,从漂洗阶段削去多余分钟数与数百升 WFI,收益落在换批时间、设备可用性与 OEE 上。
ISPE iSpeak 长文:三项传统标准、最差情形矩阵、FAT 核黄素覆盖与 CPV 趋势如何压缩换批停机
2026-11-20 | 制药 AI 现场观察 | 主体材料:ISPE Pharmaceutical Engineering iSpeak(2026-11-20)
CIP 清洗撬装:不锈钢储罐、泵组与带触摸屏的控制柜,换批时的清洗液配制、加热与回流都在这一撬上完成。图源:sanimatic.com
In strict Lean manufacturing taxonomy, equipment cleaning is classified as a business non-value-added activity. It does not physically transform the product in a way the patient directly pays for, yet it is non-negotiable for ensuring patient safety, preventing cross-contamination, and maintaining regulatory compliance. Because cleaning sits squarely within equipment changeover downtime, poorly designed cleaning processes and overly rigid validation frameworks act as a direct tax on plant capacity.
在严格的精益制造分类中,设备清洁被归为业务非增值活动。它并不以患者直接付费的方式对产品进行物理转化,但对确保患者安全、防止交叉污染与维持法规合规又不可妥协。由于清洁正处在设备换批停机时间内,设计不良的清洁流程与过于僵化的验证框架会成为对工厂产能的直接征税。
However, when we trace the evolution of cleaning validation, from the foundational rules introduced by Gary Fourman and Michael Mullen in their seminal 1993 Pharmaceutical Technology article (“Cleaning Validation: An Industrial Perspective”) to today’s science- and risk-based lifecycle approaches, we see a clear blueprint for operational excellence.
不过,当我们追溯清洁验证的演进,从 Gary Fourman 与 Michael Mullen 在 1993 年 Pharmaceutical Technology 开创性文章《Cleaning Validation: An Industrial Perspective》中提出的基础规则,到今天的基于科学与风险的生命周期方法,可以看到一条通向卓越运营的清晰路线图。
By bridging historical standardization with modern health-based exposure limits (HBELs), upfront hygienic equipment design, single-minute exchange of die (SMED) changeover principles, and ongoing lifecycle monitoring, engineering teams can systematically compress non-production idle time while strengthening compliance.
通过把历史标准化与现代基于健康的暴露限值(HBEL)、前置卫生设备设计、快速换模(SMED)换批原则以及持续的生命周期监控衔接起来,工程团队可以在强化合规的同时系统性压缩非生产闲置时间。
一、1.0 代:从主观判断走向标准作业
Generation 1.0: Moving from Subjectivity to Standard Work
Prior to the early 1990s, cleaning validation across the pharmaceutical industry often suffered from subjective evaluations and inconsistent site-to-site practices. Fourman and Mullen transformed cleaning validation into a structured, measurable engineering discipline by introducing three traditional acceptance criteria that quickly became global industry benchmarks:
20 世纪 90 年代初之前,制药行业各处的清洁验证常常受困于主观评价与各工厂做法不一。Fourman 与 Mullen 引入三项传统接受标准,把清洁验证转变为结构化、可测量的工程学科,这三项标准很快成为全球行业基准:
· The 10 parts per million (ppm) Criterion: No more than 10 ppm of the residual active ingredient from the previous batch may appear in the subsequent product.
· The Dose Criterion: No more than 1/1000th of the minimum daily therapeutic dose of the previous product may carry over into the maximum daily dose of the following product.
· The Visual Cleanliness Criterion: All product-contact equipment surfaces must appear visibly clean to the naked eye under defined lighting and viewing conditions.
· 10 ppm 标准: 后续产品中来自上一批的残留活性成分不得超过 10 ppm。
· 剂量标准: 上一产品最小日治疗剂量的 1/1000 不得带入下一产品的最大日剂量。
· 目视洁净标准: 所有产品接触设备表面在规定的照明与观察条件下必须目视洁净。
Although not written in a Lean manufacturing context, these rules established the foundational prerequisite for any Lean system: standard work. As Lean pioneer Taiichi Ohno famously observed, where there is no standard, there can be no Kaizen (continuous improvement). By replacing subjective judgment with concrete numerical limits, the industry gained a reproducible baseline. Clear thresholds enabled engineering teams to benchmark automated clean-in-place (CIP) systems, apply error-proofing (poka-yoke) to process design, eliminate the waste of redundant justification paperwork, and reduce operator-to-operator variability across shifts. Most importantly, these standards anchored cleaning validation in the ultimate Lean definition of value: protecting the patient from cross-contamination.
虽然并非在精益制造语境下写就,这些规则确立了任何精益体系的基础前提:标准作业。正如精益先驱大野耐一的名言,没有标准就没有改善(Kaizen)。用具体数值限值替代主观判断,行业由此获得了可重复的基线。清晰的阈值使工程团队能够对在线清洗(CIP)系统做基准比较,在工艺设计中应用防错(poka-yoke),消除重复论证文件的浪费,并降低不同班次操作员之间的差异。最重要的是,这些标准把清洁验证锚定在精益对价值的终极定义上:保护患者免受交叉污染。
二、2.0 代:基于风险的优先级排序与 HBEL 科学
Generation 2.0: Risk-Based Prioritization and HBEL Science
部件清洗机(COP)机柜:罐体与管路之外的部件清洗同样在验证范围内,人工清洗与自动清洗的批间差异是本文比较的重点之一。图源:sanimatic.com
While the 1993 rules provided a vital historical baseline, applying arbitrary mathematical thresholds (such as a blanket 10 ppm limit) universally across diverse product portfolios can inadvertently generate operational waste. For highly soluble, low-hazard compounds, an artificially tight 10 ppm ceiling often triggers over-processing, forcing facilities into extended wash cycles, excessive water-for-injection (WFI) consumption, and costly investigations into scientifically negligible excursions. Conversely, for highly potent active pharmaceutical ingredients or compounds with specific toxicological hazards, a traditional 1/1000th dose limit may not provide an adequate margin of safety.
1993 年的规则提供了重要的历史基线,但把任意的数学阈值(例如一刀切的 10 ppm 限值)普遍套用于多样化的产品组合,可能无意中制造运营浪费。对高溶解性、低危害化合物,人为收紧的 10 ppm 上限往往触发过度加工,迫使工厂延长清洗周期、过量消耗注射用水(WFI),并对科学上可忽略的偏差做昂贵调查。反过来,对高活性原料药或有特定毒理危害的化合物,传统的 1/1000 剂量限值可能不足以提供足够的安全裕度。
Modern regulatory frameworks, including International Council for Harmonisation (ICH) Q9(R1), ICH Q10, European Medicines Agency guidelines, and the ISPE Baseline Guide on Risk-Based Manufacture of Pharmaceutical Products (Second Edition), resolve this inefficiency by integrating traditional thresholds with science-based HBELs, such as permitted daily exposure. Aligning cleaning acceptance limits with actual pharmacological and toxicological data embodies Lean thinking: it eliminates the waste of over-cleaning low-risk products while concentrating rigorous engineering controls where patient risk genuinely resides.
包括 ICH Q9(R1)、ICH Q10、EMA 指南以及 ISPE《基于风险的药品生产(Risk-MaPP)基准指南(第二版)》在内的现代法规框架,通过把传统阈值与基于科学的 HBEL(如允许日暴露量)相结合来化解这一低效。把清洁接受限值与实际药理毒理数据对齐,体现了精益思维:消除对低风险产品过度清洁的浪费,同时把严格的工程控制集中到患者风险真正所在之处。
At the portfolio level, regulators expect a risk-based worst-case validation strategy. In Lean terms, worst-case matrixing serves as a strategic prioritization filter, directing finite validation resources toward high-risk nodes rather than executing redundant validation runs across every product-equipment permutation:
在组合层面,监管方期望基于风险的最差情形验证策略。用精益语言,最差情形矩阵充当战略优先级过滤器,把有限的验证资源引向高风险节点,而不是对每个产品—设备组合都执行冗余验证:
· Worst-Case Product Selection: Prioritizing formulations with the lowest aqueous or detergent solubility (hardest physical residues to remove), the highest potency or toxicity (lowest HBEL and matrix carryover limits), the largest batch sizes or active concentrations, allergenic profiles, and viscous or sticky excipients that resist shear forces
· Worst-Case Equipment Selection: Targeting complex mechanical geometries (dead legs, elastomeric gaskets, instrument ports, and valve bodies), equipment trains with the largest shared product-contact surface areas, and manual cleaning operations, which inherently carry far greater human variability than automated CIP/steam-in-place (SIP) skids
· 最差情形产品选择: 优先考虑水溶性或洗涤剂溶解性最低(物理残留最难去除)、活性或毒性最高(HBEL 与矩阵携带限值最低)、批量或活性浓度最大、致敏特征,以及黏稠或抗剪切的辅料配方
· 最差情形设备选择: 针对复杂机械几何结构(死角、弹性垫圈、仪表口与阀体)、共享产品接触面积最大的设备链路,以及人工清洗操作,后者固有的人工差异远大于自动 CIP/在线蒸汽灭菌(SIP)撬装系统
When practitioners pair worst-case matrixing with Lean tools like value stream mapping, they can strip non-value-added sampling, redundant analytical testing, and repetitive protocol generation out of the validation lifecycle, accelerating new product introductions without compromising quality.
当实践者把最差情形矩阵与价值流图等精益工具结合使用时,可以从验证生命周期中剔除非增值取样、冗余分析测试与重复的方案编写,在不牺牲质量的前提下加快新产品引入。
三、前置精益设计(3P):用工程手段消除浪费
Upfront Lean Design (3P): Engineering Out the Waste
Risk matrices establish what constitutes an acceptable residue limit, but the physical cleaning process dictates how long a manufacturing line remains idle between campaigns. Cleaning cycle duration is governed by the interplay of Sinner’s Circle parameters, time, action (impingement/turbulence), concentration, and temperature (TACT), against product residue characteristics and vessel geometry. If engineering teams treat equipment design as fixed, they are forced to compensate for poor fluid mechanics by extending wash durations, increasing chemical concentrations, and consuming excess thermal energy.
风险矩阵确定可接受残留限值是什么,而物理清洁过程决定产线在两个生产周期之间闲置多久。清洗周期时长由 Sinner 圆参数(时间、作用(喷淋冲击/湍流)、浓度、温度,TACT)与产品残留特性、容器几何结构的相互作用决定。如果工程团队把设备设计当作固定不变,就只能靠延长清洗时长、提高化学品浓度、消耗多余热能来补偿不良的流体力学。
Applying Lean Production Preparation Process (3P) principles during detailed engineering attacks changeover waste before equipment ever reaches the plant floor:
在详细工程阶段应用精益生产准备过程(3P)原则,可以在设备进场之前就消解换批浪费:
1. Hygienic Design Enablers: Selecting rotary or static spray devices matched to the soil rheology, optimizing nozzle placement to eliminate mechanical shadowing behind agitator shafts and baffles, and minimizing tank inlets, outlets, and dip tubes to simplify fluid flow paths.
2. Early Verification at Factory Acceptance Testing (FAT): Executing rigorous spray coverage studies (such as riboflavin fluorescence testing) during FAT provides empirical proof of complete surface wetting early in the capital project lifecycle. Identifying and correcting a shadowed nozzle orientation or an un-swept port during FAT prevents costly field piping retrofits and permanently inflated CIP recipes during commercial manufacturing.
1. 卫生设计使能要素: 选择与污物流变特性匹配的旋转或静态喷淋装置,优化喷嘴布置以消除搅拌轴与挡板背后的机械阴影,尽量减少罐体进料口、出料口与浸没管以简化流体路径。
2. 在工厂验收测试(FAT)中早期验证: 在 FAT 阶段执行严格的喷淋覆盖研究(如核黄素荧光测试),可以在资本项目生命周期早期提供表面完全润湿的经验证据。在 FAT 中识别并纠正被遮挡的喷嘴朝向或未冲刷到的接口,可以避免商用生产阶段昂贵的现场管路改造与被永久抬高的 CIP 配方。
Consider a typical multi-product formulation vessel: when engineering teams eliminate redundant top-head nozzles and resolve agitator-hub shadowing during FAT, they remove the mass-transfer bottlenecks that normally require prolonged rinse phases to clear trapped detergent and product residue. Designing for cleanability upfront directly advances SMED (quick changeover) objectives on the shop floor:
以一台典型的多产品配液罐为例:当工程团队在 FAT 阶段消除多余的顶部喷嘴并解决搅拌器轮毂阴影问题时,就移除了通常需要延长漂洗阶段才能冲净截留洗涤剂与产品残留的传质瓶颈。前置的可清洁性设计直接推进车间现场的 SMED(快速换批)目标:
· Resource and Utility Conservation: Verified complete surface coverage minimizes the volume of purified water, WFI, chemical cleaning agents, and steam/electrical power required per CIP cycle.
· Idle Time Compression: Faster, more effective mass transfer shortens wash and rinse steps, directly compressing non-production equipment downtime.
· Latent Capacity Expansion: Reducing turnaround time between batches increases equipment availability, boosting overall equipment effectiveness (OEE) and unlocking additional commercial plant capacity without capital expansion.
· 资源与公用工程节约: 经验证的完全表面覆盖把每个 CIP 周期所需的纯化水、WFI、化学清洗剂以及蒸汽/电力降到最低。
· 闲置时间压缩: 更快、更有效的传质缩短清洗与漂洗步骤,直接压缩设备非生产停机时间。
· 潜在产能释放: 缩短批次间周转时间提高设备可用性,提升整体设备效率(OEE),在不追加资本开支的情况下释放商业工厂产能。
四、阶段 3 生命周期改善:数据驱动的周期优化
Stage 3 Lifecycle Kaizen: Data-Driven Cycle Optimization
清洗周期趋势记录界面:左侧是酸洗与碱洗的批次周期清单,右侧为温度实测值与设定值的曲线,阶段 3 CPV 的分析对象就是这类逐周期记录。图源:sanimatic.com
True continuous improvement (Kaizen) does not end once initial validation protocols (Stage 2) are signed off. When baseline design insights from FAT coverage studies are connected to Lifecycle Stage 3 Continued Process Verification (CPV), cleaning validation evolves from a static compliance snapshot into a dynamic operational feedback loop.
真正的持续改善(Kaizen)不会在初始验证方案(阶段 2)签署后结束。当 FAT 覆盖研究得到的基线设计洞见与生命周期阶段 3 的持续工艺确认(CPV)相连时,清洁验证就从静态合规快照演变为动态运营反馈回路。
By leveraging inline process analytical technology (PAT), such as final-rinse conductivity sensors, online total organic carbon (TOC) analyzers, temperature transmitters, and return-line flow/turbidity monitoring, engineering teams capture continuous performance data across hundreds of commercial cycles. Applying statistical process control (SPC) to Stage 3 CPV data allows scientists to distinguish normal batch-to-batch variability from early mechanical drift, such as a partially occluded spray nozzle or declining CIP supply pump pressure. Furthermore, empirical CPV baselines provide the regulatory justification needed to optimize conservatively padded CIP recipes, shaving unnecessary minutes and hundreds of liters of WFI off rinse phases while maintaining a robust, validated state of control.
通过利用在线过程分析技术(PAT),例如最终漂洗电导率传感器、在线总有机碳(TOC)分析仪、温度变送器以及回流管路流量/浊度监控,工程团队可以在数百个商业批次周期上捕获连续性能数据。对阶段 3 CPV 数据应用统计过程控制(SPC),使科学家能够区分正常的批间波动与早期机械漂移,例如部分堵塞的喷嘴或 CIP 供给泵压力下降。此外,经验性的 CPV 基线为优化保守放大的 CIP 配方提供了所需的法规依据,在维持稳健、经验证的受控状态的同时,从漂洗阶段削去不必要的分钟数与数百升 WFI。
五、精益清洁验证矩阵
The Lean Cleaning Validation Matrix
以下按原则逐条列出。原文为两列对照表,此处拆成条目以便窄屏阅读。
01 VALUE AND RISK PRIORITIZATION
价值与风险优先级
把携带限值与毒理 HBEL 及最差情形矩阵对齐,在保护患者安全的同时消除低风险设备链路上的过度加工与多余测试
Aligning carryover limits with toxicological HBELs and worst-case matrices protects patient safety while eliminating over-processing and excess testing on low-risk equipment trains
02 STANDARD WORK
标准作业
清晰的数值接受标准、经验证的 FAT 喷淋覆盖基线与自动 CIP 配方,建立起高度可重复、不依赖班次的换批周期
Clear numerical acceptance criteria, verified FAT spray coverage baselines, and automated CIP recipes establish highly reproducible, shift-independent changeover cycles
03 WASTE REDUCTION (SMED)
消除浪费(SMED)
卫生型容器几何结构(最小化进出口、消除死角、优化喷嘴布置)降低 WFI、洗涤剂、公用工程消耗与设备闲置时间
Hygienic vessel geometry (minimized inlets/outlets, eliminated dead legs, optimized nozzle placement) cuts WFI, detergent, utility consumption, and equipment idle time
04 ERROR PROOFING (POKA-YOKE)
防错(Poka-Yoke)
定义明确的分析限值、自动 CIP 参数联锁与在线 PAT 确认,消除人工清洗与主观放行固有的人工差异
Defined analytical limits, automated CIP parameter interlocks, and inline PAT verification eliminate the human variability inherent in manual cleaning and subjective release
05 CONTINUOUS IMPROVEMENT (KAIZEN)
持续改善(Kaizen)
把历史验证基线、FAT 覆盖洞见与阶段 3 CPV 趋势结合,实现配方持续精炼、更短周转时间与更高工厂 OEE
Combining historical validation baselines with FAT coverage insights and Stage 3 CPV trending enables ongoing recipe refinement, shorter turnaround times, and higher plant OEE
六、结论
Conclusion
Fourman and Mullen’s 1993 criteria pulled pharmaceutical cleaning validation out of ambiguity and established the industry’s first measurable standard work. Today, when we evolve that historical baseline using HBEL toxicology, ICH Q9 risk-based matrixing, upfront hygienic equipment design, and data-driven Stage 3 CPV, we bridge regulatory science and Lean engineering, transforming required changeover downtime into a strategic driver of manufacturing capacity and operational excellence.
Fourman 与 Mullen 的 1993 年标准把药品清洁验证从含混中拉出来,确立了行业第一套可测量的标准作业。今天,当我们用 HBEL 毒理学、ICH Q9 基于风险的矩阵方法、前置卫生设备设计以及数据驱动的阶段 3 CPV 来演进这一历史基线时,我们把法规科学与精益工程衔接起来,把必需的换批停机时间转变为制造产能与卓越运营的战略驱动因素。
资料来源:
Nandkishor Adhao,From Arbitrary Thresholds to Data-Driven Kaizen: Cleaning Validation as a Lean Enabler,ISPE Pharmaceutical Engineering iSpeak,2026-11-20。https://www.ispe.org/pharmaceutical-engineering/ispeak/arbitrary-thresholds-data-driven-kaizen-cleaning-validation-lean
文中设备照片与清洗数据界面取自 Sani-Matic 官网(sanimatic.com),为 CIP 撬装、部件清洗机与清洗周期记录界面的实拍图与软件截图,与本文作者及文中论述无对应关系。
来源:小南AI实验室 · mp.weixin.qq.com