Yang Liu

← Primer

HSC · Healing Source Code

一份 ex vivo HSC 疗法的成本拆解——钱到底花在采集、编辑还是放行 A Cost Breakdown of Ex Vivo HSC Therapy — Where the Money Actually Goes: Collection, Editing, or Release

2026-07-31 · Casgevy, ex vivo 基因治疗, 细胞采集, 放行检验, 镰状细胞病, 基因治疗定价

2023 年 12 月,Casgevy 的标价公布:220 万美元一次性治疗。多数人第一反应是去想象那个价格背后的场景——大概是某种极其昂贵的分子机器,在实验室里小心翼翼地剪开一段 DNA。可如果真去问一问这两百二十万美元花在哪儿,答案会让人意外:从病人手臂上抽出第一管血,到那一小瓶编辑好的细胞被送回治疗中心重新输回他的身体,中间要过去将近半年;而这半年里,真正”编辑基因”的那个动作,可能只占了流程里最便宜的一段。

一次治疗,其实是一场供应链

Ex vivo 基因治疗常被简化成一句话:取出细胞、改写基因、放回去。这个说法没错,却也几乎没有说出它真正的样子。取出的细胞要先在病人体内被动员出来、经过 apheresis 采集,再冷链运到一家往往远在另一个城市甚至另一个国家的制造工厂;在那里,细胞要被计数、纯化、编辑、扩增、检验、放行,再冷链运回治疗中心;病人这边则要先接受数天的清髓预处理,腾出骨髓里的位置,才能等来这瓶迟到数月的细胞。每一段路程都在花钱,而”编辑”只是这条链条中间一个相对短暂的工序。理解一款 ex vivo 疗法为什么贵,得沿着这条供应链走一遍。

第一公里已经开始掉队

链条的起点是采集,而采集远不是抽血那么简单。造血干细胞平时安静地待在骨髓里,治疗前必须先用药物把它们动员进外周血,再用 apheresis 机器把它们从血液里分离出来。对镰状细胞病患者而言,这一步天然比其他适应症更难:反复的血管闭塞事件会损伤骨髓微环境,长期使用的羟基脲可能影响造血干细胞的可动员性,而最常用的动员药物 G-CSF 本身又有诱发血管闭塞危象的风险,不能像在其他疾病里那样常规使用。于是这类患者只能依赖 CXCR4 拮抗剂 plerixafor 单药动员,但个体反应差异很大,不少人需要反复多轮采集才能凑够制造所需的细胞数——Leonard 和 Weiss 在 2024 年的一篇综述里把这一整套限制概括为镰状细胞病基因治疗的天然瓶颈:骨髓本身受损、常规动员方案不可用、个体反应又难以预测。

这不是纸面上的担忧。exa-cel(即 Casgevy)的关键临床试验 CLIMB SCD-121 按 ClinicalTrials.gov(NCT03745287)登记的实际入组人数为 63 名患者,而 Frangoul 等人 2024 年在 N Engl J Med 上报告,最终只有 44 名患者完成了清髓预处理并接受了 exa-cel 输注——中间那部分患者,有的止步于未能采集到足够进入下一步的造血干细胞,有的止步于其他环节,具体在哪一步、以多大比例退出,仍值得回到 NEJM 正文的试验流程图与补充材料逐字核对,但方向是清楚的:在整条链条里,最先掉队的病人,往往倒在编辑开始之前。这也是为什么”制造”这个词,在 ex vivo 基因治疗里其实要从病人被采血的那一刻算起,而不是从细胞进厂那一刻算起。

编辑,反而是相对便宜的一步

细胞真正进厂之后,“编辑”本身在成本上常常不是最重的一块——这背后有一段技术路线的分野。较早一代的 ex vivo 基因疗法,包括同样针对血红蛋白病的 Zynteglo 和 Lyfgenia,走的是基因添加路线:用慢病毒载体把一份功能性的珠蛋白基因整合进细胞基因组。慢病毒载体的生产是一套独立而昂贵的生物制造体系——UCL 团队 Comisel 等人 2021 年在 Biochemical Engineering Journal 上对慢病毒载体的工艺经济学做过系统建模,发现如果把生产工艺从传统的多层培养皿式平台(如 10-layer 平板)换成悬浮式或固定床生物反应器,大规模生产时载体本身的生产成本至少能降低九成——这从反面说明,慢病毒载体的生产工艺本身,原本就是一项分量不轻、也确实存在压缩空间的成本。

Casgevy 走的是另一条路:不整合外源基因,而是用 CRISPR-Cas9 直接在 BCL11A 的红系 enhancer 上做一次精准剪切。这类编辑通常依靠 Cas9 蛋白与 guide RNA 组成的 ribonucleoprotein 复合物,经由 electroporation 直接送进细胞——不需要生产病毒颗粒,也就绕开了慢病毒载体那部分沉重的成本。米兰 San Raffaele 团队 Ferrari 等人 2021 年在 Frontiers in Genome Editing 上的综述里,把 electroporation 称作递送可编程核酸酶进入体外培养造血干细胞的首选方法——给出的理由主要是安全性:通过 mRNA 或 RNP 递送核酸酶,能让编辑活性”高而短暂”,从而降低脱靶风险,而非依赖持续表达的整合型系统。这层安全性考量之外,不生产病毒颗粒也顺带省下了一整套载体生产线的成本,只是这一点 Ferrari 等人的综述本身并未展开论证。

但便宜是相对的。electroporation 本身需要专用设备和临床级 GMP 试剂,guide RNA、Cas9 蛋白都要按药品标准生产;真正让编辑这一步显得”便宜”,是因为它被拿去和采集、和后面即将出现的放行检验相比——这两端,才是这条供应链里真正吃掉预算的地方。

真正烧钱的是放行

细胞编辑完成,离能被输回病人体内,还差最费时间、也最费钱的一关:放行检验。根据 Casgevy 官方披露的流程,每一批产品在送回治疗中心之前,都要完成活力、纯度、含量、效力、无菌等一整套放行检验——其中效力检验尤其棘手,监管机构要求它必须是定量的、能反映产品作用机制的,而不是一个笼统的存活率数字。对于像 Zynteglo、Lyfgenia 这样依赖慢病毒载体整合的产品,还多出一项复制型慢病毒(RCL)检测——传统的细胞扩增法检测流程本身就可能需要数周才能出结果。把这些检验串起来,再加上运输和病人这边的清髓预处理窗口,Casgevy 官方给出的说法是,从采血到产品送回治疗中心,通常需要五到六个月。

这段时间不是被动等待,而是真金白银的成本。在一项针对自体细胞治疗生产成本结构的建模分析中,Lopes 等人 2018 年在 BioProcess International 上估算,一款自体细胞疗法的生产成本里,人工可以占到约五成,材料——其中很大一部分正是过程检验和放行检验所需的试剂与耗材——占到约四分之一;这个模型针对的是树突状细胞疗法而非造血干细胞产品,但它揭示的结构性事实具有普遍性:在自体细胞治疗里,真正吃掉预算的常常不是某一次分子操作,而是环绕着它的、必须逐批重复的质量证明工作。同属自体细胞治疗产业链的 CAR-T,也提供了一个可比照的参照系——英国国家 CAR T 专家组 2025 年在 Blood Cancer Journal 上报告,在近千名接受治疗的大 B 细胞淋巴瘤患者里,有 3.87% 遭遇了制造失败或产品不合格,无法按计划回输。这个数字是在多年工艺优化之后达到的相对低位,却依然提醒着一件事:自体细胞产品是逐个病人单独生产的一次性批次,任何一步不合格,损失的不是一批货,而是一个人本该等到的治疗窗口。

拼起来是多少钱

把这条链条上的每一段成本叠加起来,再看 Casgevy 220 万美元这个标价,会发现它甚至还不是病人实际要付出的全部。DeMartino 等人 2024 年在 Blood Advances 上更新的预算影响分析指出,在 220 万美元的批发采购价之外,还要加上约 9 万美元用于清髓预处理期间的住院费用,两项合计约 229 万美元;如果把不同产品和折扣情形都纳入敏感性分析,这个区间可以从约 178 万美元一路铺到 319 万美元。同一批 2023 年获批的疗法里,bluebird bio 将 Lyfgenia 定在 311 万美元,高于 Casgevy——两款药面对的是同一种疾病、相近的采集与放行流程,价格差异更多来自技术路线本身的成本结构与商业定价策略。独立评估机构 ICER 在 2023 年的最终证据报告中给出了另一个参照:按标准的成本效果阈值折算,这类基因疗法定价在 135 万到 205 万美元之间才算”物有所值”——而两款药的实际标价都超过了这个上限。换句话说,即便撇开厂商的定价策略不谈,单看采集、编辑、放行、清髓这条实打实的制造与治疗链条,它本身的重量就足以撑起一个百万美元级别的价格。

边界与未解

这份成本清单里,几乎没有一项是”编辑”本身造成的——真正沉重的,是采集端病人自身生物学条件的限制,和放行端逐批重复的质量证明。这也解释了为什么整个行业会不约而同地把下一程押在两个方向上:一是把制造流程搬进封闭、自动化的一体化设备,减少人工干预、缩短放行周期;二是干脆跳过采集与体外操作这整段旅程,转向 in vivo 编辑——让病人不必再经历动员失败的风险、不必再等待半年的制造窗口。但在电穿孔试剂和检验耗材可以被自动化压低成本的同时,采集端那道天然的生物学门槛、和放行端那套无法简化的安全证明,短期内都不会消失。这条供应链未来会变得更快、更便宜,但它大概率还会保留现在这个形状:两头重、中间轻。

参考文献

  1. Frangoul H, Locatelli F, Sharma A, et al; CLIMB SCD-121 Study Group. Exagamglogene Autotemcel for Severe Sickle Cell Disease. N Engl J Med. 2024;390(18):1649-1662. DOI
  2. Comisel RM, Kara B, Fiesser FH, Farid SS. Lentiviral vector bioprocess economics for cell and gene therapy commercialisation. Biochem Eng J. 2021;167:107868. DOI
  3. Leonard A, Weiss MJ. Hematopoietic stem cell collection for sickle cell disease gene therapy. Curr Opin Hematol. 2024;31(3):104-114. DOI
  4. Ferrari S, Vavassori V, Canarutto D, et al. Gene Editing of Hematopoietic Stem Cells: Hopes and Hurdles Toward Clinical Translation. Front Genome Ed. 2021;3:618378. DOI
  5. Lopes AG, Sinclair A, Frohlich B. Cost Analysis of Cell Therapy Manufacture: Autologous Cell Therapies, Part 1. BioProcess International. 2018;16(3):S3-S8.
  6. Dulobdas V, Kirkwood AA, Serpenti F, et al; UK National CAR T Panel. Risk factors for CAR T-cell manufacturing failure and patient outcomes in large B-cell lymphoma. Blood Cancer J. 2025;15:30. DOI
  7. DeMartino PC, Haag MB, Caughey AB, Roth JA. A budget impact analysis of gene therapy for sickle cell disease: an updated analysis. Blood Adv. 2024;8(17):4658-4661. DOI
  8. Vertex Pharmaceuticals / CRISPR Therapeutics. Vertex and CRISPR Therapeutics Announce US FDA Approval of CASGEVY™ (exagamglogene autotemcel) for the Treatment of Sickle Cell Disease. Press release, December 8, 2023. https://news.vrtx.com/news-releases/news-release-details/vertex-and-crispr-therapeutics-announce-us-fda-approval
  9. CASGEVY® 官方患者网站,治疗流程与放行检验说明(“The CASGEVY Treatment Journey”)。https://www.casgevy.com/sickle-cell-disease/treatment-journey
  10. Institute for Clinical and Economic Review (ICER). Sickle Cell Disease: Final Evidence Report, August 2023. https://icer.org/assessment/sickle-cell-disease-2023/
  11. ClinicalTrials.gov. A Safety and Efficacy Study Evaluating CTX001 in Subjects With Severe Sickle Cell Disease (CLIMB SCD-121). NCT03745287. https://clinicaltrials.gov/study/NCT03745287

In December 2023, Casgevy’s price was announced: $2.2 million for a one-time treatment. Most people’s first instinct is to picture what that price is buying — some extraordinarily expensive piece of molecular machinery, carefully snipping a strand of DNA somewhere in a lab. But ask where those $2.2 million actually go, and the answer is surprising: from the first tube of blood drawn from a patient’s arm to the moment that small vial of edited cells is infused back into his body, nearly half a year passes — and within that half-year, the act of actually “editing the gene” may be one of the cheapest steps in the entire process.

One treatment, actually a supply chain

Ex vivo gene therapy is often reduced to one sentence: take the cells out, rewrite the gene, put them back. That’s not wrong, but it barely describes what actually happens. The cells first have to be mobilized inside the patient’s body and collected by apheresis, then shipped cold-chain to a manufacturing facility that is often in another city or even another country; there, the cells are counted, purified, edited, expanded, tested, and released, then shipped cold-chain back to the treatment center; meanwhile the patient undergoes several days of conditioning to clear space in the bone marrow, so they’re ready when this vial of cells finally arrives, months later. Every leg of this journey costs money, and “editing” is just one relatively brief step in the middle of the chain. Understanding why an ex vivo therapy is expensive means walking the whole supply chain.

The first mile already sheds patients

The chain begins with collection, and collection is far from a simple blood draw. Hematopoietic stem cells normally sit quietly in the bone marrow; before treatment, drugs must first mobilize them into the peripheral blood, and then an apheresis machine separates them out of the blood. For sickle cell disease patients, this step is inherently harder than in other indications: repeated vaso-occlusive events damage the bone marrow microenvironment, long-term hydroxyurea use may impair how mobilizable the hematopoietic stem cells are, and the most commonly used mobilizing agent, G-CSF, itself carries a risk of triggering vaso-occlusive crisis, so it cannot be used routinely the way it is in other diseases. That leaves these patients dependent on the CXCR4 antagonist plerixafor alone for mobilization, with highly variable individual responses — many need repeated rounds of collection to accumulate enough cells for manufacturing. Leonard and Weiss, in a 2024 review, summed up this whole set of constraints as the natural bottleneck of sickle cell disease gene therapy: an already-damaged marrow, a mobilization regimen that can’t be used routinely, and individual responses that are hard to predict.

This isn’t a concern on paper only. The pivotal trial for exa-cel (Casgevy), CLIMB SCD-121, enrolled 63 patients according to its ClinicalTrials.gov registration (NCT03745287), while Frangoul and colleagues, reporting in 2024 in the New England Journal of Medicine, note that in the end only 44 patients completed conditioning and received the exa-cel infusion — some of the patients in between dropped out because they couldn’t collect enough hematopoietic stem cells to proceed, others for other reasons, and exactly where and in what proportions those exits happened is still worth cross-checking, word for word, against the trial’s profile figures and supplementary material in the NEJM paper itself. But the direction is clear: across the whole chain, the patients who fall out first tend to fall out before editing even begins. That’s also why, in ex vivo gene therapy, “manufacturing” really has to be counted from the moment a patient’s blood is drawn, not from the moment the cells arrive at the factory.

Editing turns out to be the relatively cheap step

Once the cells actually reach the factory, “editing” itself is often not the heaviest cost item — and behind that lies a split in technical approach. Earlier-generation ex vivo gene therapies, including Zynteglo and Lyfgenia, which also target hemoglobinopathies, take the gene-addition route: a lentiviral vector integrates a functional globin gene into the cell’s genome. Manufacturing lentiviral vectors is its own separate, expensive biomanufacturing system — Comisel and colleagues at UCL, in a 2021 paper in the Biochemical Engineering Journal, systematically modeled the process economics of lentiviral vector manufacturing and found that switching production from traditional multilayer flask-type platforms (such as 10-layer trays) to suspension or fixed-bed bioreactors could cut the vector’s own production cost by at least 90% at large scale — which, read in reverse, shows that lentiviral vector manufacturing itself was, to begin with, a substantial cost, and one with real room to compress.

Casgevy takes a different path: instead of integrating a foreign gene, it uses CRISPR-Cas9 to make one precise cut in the erythroid enhancer of BCL11A. This kind of editing typically relies on a ribonucleoprotein complex of Cas9 protein and guide RNA, delivered straight into the cell by electroporation — no viral particles need to be manufactured, which sidesteps the heavy cost of lentiviral vector production. Ferrari and colleagues at Milan’s San Raffaele institute, in a 2021 review in Frontiers in Genome Editing, write that “electroporation became the method of choice to efficiently deliver programmable nucleases in ex vivo cultured HSPCs” — for reasons that are mainly about safety: delivering the nuclease as mRNA or RNP has “become the gold standard to achieve a high but transient nuclease activity in HSPCs and other target cells,” lowering off-target risk, rather than relying on a continuously expressed, integrating system. Beyond that safety rationale, skipping viral-particle production also happens to save an entire vector production line’s worth of cost — a point the Ferrari review itself does not make.

But cheap is relative. Electroporation itself requires dedicated equipment and clinical-grade GMP reagents; guide RNA and Cas9 protein both have to be manufactured to drug-product standards. What actually makes editing look “cheap” is that it’s being compared to collection, and to the release testing that comes later — those two ends are where this supply chain really eats its budget.

What really burns money is release

Once editing is finished, there’s still the most time-consuming, most expensive gate before the cells can go back into the patient: release testing. According to the process Casgevy’s own materials describe, every batch has to pass a full panel of release tests — viability, purity, content, potency, and sterility — before it’s shipped back to the treatment center. Potency testing is particularly hard: regulators require it to be quantitative and reflective of the product’s mechanism of action, not just a generic viability number. Products like Zynteglo and Lyfgenia, which rely on integrating lentiviral vectors, need one more test on top of that: replication-competent lentivirus (RCL) testing — and the traditional cell-expansion-based assay for this can itself take weeks to return a result. String all these tests together, add shipping and the patient’s own conditioning window, and Casgevy’s own account is that it typically takes five to six months from blood draw to the product being shipped back to the treatment center.

That time is not passive waiting — it’s real money. In a modeling analysis of the cost structure of autologous cell therapy manufacturing, Lopes and colleagues, in 2018 in BioProcess International, estimated that labor accounts for roughly half the manufacturing cost of an autologous cell therapy, and materials — much of which are the reagents and consumables needed for in-process and release testing — account for roughly a quarter; that model was built around a dendritic cell therapy rather than a hematopoietic stem cell product, but the structural fact it reveals is general: in autologous cell therapy, what actually eats the budget is often not any single molecular operation, but the quality-proof work that surrounds it and has to be repeated batch by batch. CAR-T, which belongs to the same autologous cell therapy industry, offers a comparable reference point — the UK National CAR T Panel, reporting in 2025 in Blood Cancer Journal, found that among nearly a thousand large B-cell lymphoma patients treated, 3.87% experienced manufacturing failure or a non-conforming product and could not be infused as planned. That figure represents a relatively low point reached only after years of process optimization, and it still serves as a reminder: autologous cell products are manufactured as one-off batches for a single patient at a time, so a single failed step doesn’t just cost a batch of product — it costs a person the treatment window they were supposed to get.

Add it all up

Add up every segment of cost along this chain, and looking again at Casgevy’s $2.2 million price tag, it turns out that’s not even the whole amount a patient actually pays. DeMartino and colleagues, in an updated 2024 budget impact analysis in Blood Advances, note that on top of the $2.2 million wholesale acquisition price, there’s roughly $90,000 more for hospitalization during the conditioning period, bringing the combined total to about $2.29 million; factoring different products and discount scenarios into a sensitivity analysis stretches that range from roughly $1.78 million up to $3.19 million. Among the same 2023 cohort of approved therapies, bluebird bio priced Lyfgenia at $3.11 million, higher than Casgevy — the two drugs address the same disease and go through comparable collection and release processes, so the price gap comes down more to differences in the underlying technical approach’s cost structure and to commercial pricing strategy. The independent assessment body ICER offers another reference point in its 2023 final evidence report: translated into standard cost-effectiveness thresholds, these gene therapies would need to be priced between $1.35 million and $2.05 million to be considered good value — and both drugs’ actual list prices exceed that ceiling. In other words, even setting aside manufacturers’ pricing strategy entirely, the real weight of just the collection-editing-release-conditioning chain of manufacturing and treatment is, on its own, enough to support a price tag in the millions.

Limits and open questions

In this cost ledger, almost nothing traces back to “editing” itself — what’s actually heavy is the biological constraints on the patient’s own body at the collection end, and the batch-by-batch repeated quality proof at the release end. That also explains why the field, without any real coordination, has converged on betting the next stretch of progress on two directions: one is moving manufacturing into closed, automated, integrated equipment to cut down manual intervention and shorten the release cycle; the other is skipping the whole collection-and-ex-vivo-manipulation journey altogether and moving to in vivo editing — sparing the patient the risk of failed mobilization and the months-long wait for a manufacturing slot. But even as automation drives down the cost of electroporation reagents and testing consumables, the natural biological barrier at the collection end, and the safety proof at the release end that can’t really be simplified away, aren’t going anywhere in the near term. This supply chain will get faster and cheaper over time, but it will probably keep roughly the shape it has now: heavy at both ends, light in the middle.

References

  1. Frangoul H, Locatelli F, Sharma A, et al; CLIMB SCD-121 Study Group. Exagamglogene Autotemcel for Severe Sickle Cell Disease. N Engl J Med. 2024;390(18):1649-1662. DOI
  2. Comisel RM, Kara B, Fiesser FH, Farid SS. Lentiviral vector bioprocess economics for cell and gene therapy commercialisation. Biochem Eng J. 2021;167:107868. DOI
  3. Leonard A, Weiss MJ. Hematopoietic stem cell collection for sickle cell disease gene therapy. Curr Opin Hematol. 2024;31(3):104-114. DOI
  4. Ferrari S, Vavassori V, Canarutto D, et al. Gene Editing of Hematopoietic Stem Cells: Hopes and Hurdles Toward Clinical Translation. Front Genome Ed. 2021;3:618378. DOI
  5. Lopes AG, Sinclair A, Frohlich B. Cost Analysis of Cell Therapy Manufacture: Autologous Cell Therapies, Part 1. BioProcess International. 2018;16(3):S3-S8.
  6. Dulobdas V, Kirkwood AA, Serpenti F, et al; UK National CAR T Panel. Risk factors for CAR T-cell manufacturing failure and patient outcomes in large B-cell lymphoma. Blood Cancer J. 2025;15:30. DOI
  7. DeMartino PC, Haag MB, Caughey AB, Roth JA. A budget impact analysis of gene therapy for sickle cell disease: an updated analysis. Blood Adv. 2024;8(17):4658-4661. DOI
  8. Vertex Pharmaceuticals / CRISPR Therapeutics. Vertex and CRISPR Therapeutics Announce US FDA Approval of CASGEVY™ (exagamglogene autotemcel) for the Treatment of Sickle Cell Disease. Press release, December 8, 2023. https://news.vrtx.com/news-releases/news-release-details/vertex-and-crispr-therapeutics-announce-us-fda-approval
  9. CASGEVY® official patient website, treatment process and release testing description (“The CASGEVY Treatment Journey”). https://www.casgevy.com/sickle-cell-disease/treatment-journey
  10. Institute for Clinical and Economic Review (ICER). Sickle Cell Disease: Final Evidence Report, August 2023. https://icer.org/assessment/sickle-cell-disease-2023/
  11. ClinicalTrials.gov. A Safety and Efficacy Study Evaluating CTX001 in Subjects With Severe Sickle Cell Disease (CLIMB SCD-121). NCT03745287. https://clinicaltrials.gov/study/NCT03745287