Yang Liu

← Primer

HSC · Healing Source Code

Casgevy 与 Lyfgenia——同一种病、同一天获批,两套技术,两本风险账 Casgevy and Lyfgenia — one disease, one approval day, two technologies, two ledgers of risk

2026-10-09 · Casgevy, Lyfgenia, 镰状细胞病, 基因编辑, 慢病毒, 黑框警告, busulfan

2023 年 12 月 8 日,FDA 在同一份新闻稿里批准了两款治疗镰状细胞病的细胞基因疗法,适用人群同为 12 岁及以上。两款药都取自病人自己的造血干细胞,都在体外改造,都要在回输前先用高剂量化疗把骨髓腾空。对病人来说,这几乎是同一场手术。但把两张标签并排放在桌上,差别很快就显出来:其中一张在最前面印着黑框。

一个改写调控,一个加一个基因

Casgevy(exagamglogene autotemcel,exa-cel)不碰致病的 HBB 基因,而是去动 BCL11A 的红系 enhancer。CRISPR-Cas9 在体外切开这段 enhancer,让红系细胞里的 BCL11A 下调,胎儿血红蛋白(HbF)重新表达;CD34⁺ 细胞经电穿孔导入编辑组件(Frangoul 等人 2021 年在 N Engl J Med 上描述了这一做法),2024 年同刊的试验论文把它称为 nonviral 的细胞疗法。这条路线的基础工作之一是 Wu 等人 2019 年发表在 Nat Med 的研究:Cas9:sgRNA 核糖核蛋白复合物在 +58 BCL11A 红系 enhancer 的 GATA1 结合位点内切割,降低 BCL11A 表达、诱导 γ-globin。这个开关的来龙去脉见本连载第 41、53 篇。

Lyfgenia(lovotibeglogene autotemcel,lovo-cel)走的是更早一代的基因添加路线。FDA 处方信息写明,它用 BB305 慢病毒载体转导病人的 CD34⁺ 细胞,载体带着修饰过的 β-globin 基因 βA-T87Q,表达出的 HbAT87Q 设计上能阻碍 HbS 聚合;标签里还有一句对后文很要紧的话:基因通过不可逆的整合进入长期造血干细胞。2017 年,Ribeil 等人在 N Engl J Med 上报告的首例病人用的正是这一思路。所以两者的差别并不只是”剪”和”加”:Casgevy 把变化留在内源基因组的一个调控位点上;Lyfgenia 则把一段外源序列整合进基因组,整合在哪里,每个病人各不相同。

两项试验,两把尺子

支撑 Casgevy 的是 CLIMB SCD-121(NCT03745287),单组、开放标签,入组 12 到 35 岁、此前两年每年至少两次严重血管闭塞危象(VOC)的病人。Frangoul 等人 2024 年报告:44 人接受治疗,30 名随访足够的病人中 29 人(97%,95% CI 83–100)连续至少 12 个月没有严重 VOC。FDA 新闻稿给出的口径是 31 名可评估病人中 29 人(93.5%)。处方信息说明,30 人是随访至少 16 个月的主要疗效人群,另有 1 名随访不足 16 个月、被判定为无应答者的病人一并纳入,分母因此是 31。两个数字各有口径,本文不把它们互相换算。

支撑 Lyfgenia 的是 HGB-206(NCT02140554),一项 Phase 1/2 研究,治疗流程在试验中途被反复修改:最初的 Group A 共 7 人,HbAT87Q 水平不足以产生显著临床效果(Kanter 2023,Am J Hematol);随后建立的 Group C 才是关键评估队列,入组门槛收紧为此前 24 个月至少四次严重血管闭塞事件(Kanter 2022,N Engl J Med)。处方信息中的 Study 1-C:36 人回输;32 名纳入疗效分析的病人中,28 人(88%,95% CI 71–97)在回输后第 6 到 18 个月内血管闭塞事件(VOE)完全消失,严重 VOE 完全消失者为 30 人(94%)。

所以”93.5% 对 88%“并不是同一把尺子量出来的:88% 针对全部 VOE,包括不满足”严重”定义的事件,与 Casgevy 的”严重 VOC”更接近的是 94%,而两边对”严重”的定义仍不一样。2025 年 Adv Ther 上有一篇专门讨论这件事的论文:它比较了七项试验对血管闭塞的定义,发现 exa-cel 试验的 severe VOC 包括任何需要到医疗机构就诊的急性疼痛事件,而 lovo-cel 试验的 severe VOE 要求在医院或急诊观察单元停留至少 24 小时,或 72 小时内至少两次就诊;用 exa-cel 患者级数据套用后一种定义,30 人全部(100%)达到 12 个月无严重事件。这篇论文由 Vertex(Casgevy 的持有方)资助,作者含 exa-cel 试验的研究者 Frangoul 与 Locatelli 及三名 Vertex 员工;定义比较是描述性的,100% 来自事后(post hoc)分析,结论只是比较不同试验的终点时应当谨慎。两项试验均为单臂,观察窗口、入组门槛、事件定义和分母都不同,彼此之间没有头对头比较。

黑框警告写了什么

Lyfgenia 的黑框警告标题是 Hematologic Malignancy,首句为:“Hematologic malignancy has occurred in patients treated with LYFGENIA.”

警告背后是三个病例,写在处方信息 5.1 节里:获批时,两名接受早期版本(不同的生产工艺和移植流程,Group A)的病人发生了急性髓系白血病(AML);一名带 α-thalassemia trait 的 Group C 病人被诊断为骨髓增生异常综合征(MDS),确诊在第 24 个月之后。其中一例 AML 发表在 Goyal 等人 2022 年的 N Engl J Med 上:病人在回输后约 5.5 年发病,白血病细胞里检出了载体的一个整合位点,但作者根据整合位置、转基因表达极低以及对周围基因表达无影响,判断白血病不太可能与载体整合有关,并认为镰状细胞病患者移植后本就面临更高的血液恶性肿瘤风险。FDA 标签同样写了两层意思:镰状细胞病本身风险高于一般人群;动员、清髓和回输带来的额外造血压力也可能增加风险。标签另把 insertional oncogenesis 单列为潜在风险,并要求终身监测:至少 15 年内每 6 个月做一次全血细胞计数,第 6、12 个月及必要时做整合位点分析。

Casgevy 的处方信息(2026 年 7 月修订版)没有 boxed warning。它的 Warnings and Precautions 列的是中性粒细胞植入失败、血小板植入延迟、过敏反应,以及一条脱靶编辑风险:由于遗传变异,在个体 CD34⁺ 细胞中发生非预期脱靶编辑的风险不能排除,其临床意义未知。2024 年的试验论文报告没有发生癌症。

读这两段要先看随访长度:Casgevy 试验中位随访 19.3 个月(0.8–48.1),Lyfgenia 的 Study 1-C 中位 38 个月(12–61)。那例 AML 出现在另一批、另一种工艺的 Group A 病人身上,它提示这类事件可能在数年之后才出现,说明随访时间本身就限制着对任何一款药的结论;它并不是关于 Casgevy 的数据,不能据此推断 Casgevy 的风险高低。两款药的病人都将进入长期随访研究,Casgevy 标签写的是回输后共 15 年。

共同的一道门

两条路线在最贵的一步上重合:两张标签都要求回输前做清髓性预处理。Lyfgenia 的标签要求 busulfan;Casgevy 的标签写的是 busulfan 或其他清髓方案,试验中用的是 busulfan,12 岁及以上起始剂量为 3.2 mg/kg/day。这部分风险与载体、编辑方式无关:Lyfgenia 安全性人群中开始清髓的 45 人全部(45/45)至少经历一项归因于清髓的不良事件,≥3 级不良反应中 stomatitis 占 71%;Casgevy 标签则写明,SCD 与 TDT 病人 100% 出现 3 或 4 级的中性粒细胞减少和血小板减少。两份标签的表格都注明含与清髓相关的事件,读安全性数字时要分清哪些来自共同的预处理与自体移植。同时,Lyfgenia 标签认为造血压力可能增加血液肿瘤风险,所以黑框警告也不能读成与这一共同流程无关。清髓为什么必须、能不能免去,留到第 61、68 篇继续讲。

还没有答案的部分

载体整合与脱靶编辑,在两张标签里各占一条机制性的未知:Lyfgenia 的 insertional oncogenesis 被列为潜在风险,已发生的血液肿瘤其因果归属在发表文献里被判为不太可能与载体有关;Casgevy 的脱靶编辑尚无报告的事件,临床意义未知。这两条都要靠更长的随访来回答。

2023 年 12 月 8 日那天,病人拿到的是两个选项。它们像是同一场手术的两个版本,但标签最前面,多了一个黑框。

参考文献

  1. FDA. FDA Approves First Gene Therapies to Treat Patients with Sickle Cell Disease (news release, December 8, 2023). fda.gov
  2. FDA. LYFGENIA (lovotibeglogene autotemcel) Prescribing Information, revised 12/2023, BLA STN 125788. fda.gov
  3. FDA. CASGEVY (exagamglogene autotemcel) Prescribing Information, revised 07/2026, BLA STN 125787. fda.gov
  4. Frangoul H, et al. Exagamglogene Autotemcel for Severe Sickle Cell Disease. N Engl J Med. 2024;390(18):1649-1662. DOI (NCT03745287)
  5. Frangoul H, et al. CRISPR-Cas9 Gene Editing for Sickle Cell Disease and β-Thalassemia. N Engl J Med. 2021;384(3):252-260. DOI
  6. Wu Y, et al. Highly efficient therapeutic gene editing of human hematopoietic stem cells. Nat Med. 2019;25(5):776-783. DOI
  7. Kanter J, et al. Biologic and Clinical Efficacy of LentiGlobin for Sickle Cell Disease. N Engl J Med. 2022;386(7):617-628. DOI (NCT02140554)
  8. Kanter J, et al. Lovo-cel gene therapy for sickle cell disease: Treatment process evolution and outcomes in the initial groups of the HGB-206 study. Am J Hematol. 2023;98(1):11-22. DOI
  9. Goyal S, et al. Acute Myeloid Leukemia Case after Gene Therapy for Sickle Cell Disease. N Engl J Med. 2022;386(2):138-147. DOI
  10. Ribeil JA, et al. Gene Therapy in a Patient with Sickle Cell Disease. N Engl J Med. 2017;376(9):848-855. DOI (NCT02151526)
  11. Frangoul H, Locatelli F, et al. Impact of Different Definitions of Vaso-Occlusion on Efficacy Assessments in Sickle Cell Disease Clinical Trials. Adv Ther. 2025;42:2490-2499. DOI
  12. ClinicalTrials.gov. NCT03745287 (CLIMB SCD-121); NCT02140554 (HGB-206). clinicaltrials.gov

On December 8, 2023, the FDA approved two cell-based gene therapies for sickle cell disease in a single news release, both for patients 12 years and older. Both drugs are made from the patient’s own hematopoietic stem cells, both are engineered outside the body, and both require high-dose chemotherapy to clear the marrow before the cells go back in. For the patient, it is almost the same procedure. But lay the two labels side by side and a difference shows quickly: one of them opens with a black box.

One rewrites a regulatory switch, the other adds a gene

Casgevy (exagamglogene autotemcel, exa-cel) leaves the disease-causing HBB gene alone and instead works on the erythroid enhancer of BCL11A. CRISPR-Cas9 cuts this enhancer ex vivo, BCL11A is reduced in erythroid cells, and fetal hemoglobin (HbF) is re-expressed; the editing components are delivered into CD34⁺ cells by electroporation (Frangoul and colleagues described this approach in the N Engl J Med in 2021, and the 2024 trial paper in the same journal calls it a nonviral cell therapy). One of the foundational studies for this route is Wu and colleagues’ 2019 paper in Nat Med: a Cas9:sgRNA ribonucleoprotein complex cuts within the GATA1 binding site of the +58 BCL11A erythroid enhancer, lowering BCL11A expression and inducing γ-globin. The background of this switch is covered in articles 41 and 53 of this series.

Lyfgenia (lovotibeglogene autotemcel, lovo-cel) takes the older gene-addition route. The FDA prescribing information states that it transduces the patient’s CD34⁺ cells with a BB305 lentiviral vector carrying a modified β-globin gene, βA-T87Q; the HbAT87Q it expresses is designed to hinder HbS polymerization. The label also contains a sentence that matters a great deal later: the gene enters long-term hematopoietic stem cells through irreversible integration. In 2017, Ribeil and colleagues reported the first patient in the N Engl J Med using exactly this idea. So the difference is not merely “cut” versus “add”: Casgevy leaves its change at a regulatory site within the endogenous genome, whereas Lyfgenia integrates an exogenous sequence into the genome, and where it lands differs from patient to patient.

Two trials, two rulers

Casgevy rests on CLIMB SCD-121 (NCT03745287), a single-arm, open-label study enrolling patients aged 12 to 35 with at least two severe vaso-occlusive crises (VOCs) per year in the previous two years. Frangoul and colleagues reported in 2024 that 44 patients were treated, and that among 30 patients with sufficient follow-up, 29 (97%; 95% CI, 83–100) had no severe VOC for at least 12 consecutive months. The figure given in the FDA news release is 29 of 31 evaluable patients (93.5%). The prescribing information explains that 30 patients formed the primary efficacy population with at least 16 months of follow-up, and that one further patient, with less than 16 months of follow-up and judged a non-responder, was included as well, which makes the denominator 31. Each number has its own basis, and this article does not convert one into the other.

Lyfgenia rests on HGB-206 (NCT02140554), a Phase 1/2 study whose treatment process was revised repeatedly along the way. The initial Group A, seven patients, had HbAT87Q levels too low to produce a substantial clinical effect (Kanter 2023, Am J Hematol); Group C, established afterward, was the key evaluation cohort, with the entry threshold tightened to at least four severe vaso-occlusive events in the previous 24 months (Kanter 2022, N Engl J Med). In Study 1-C of the prescribing information, 36 patients were infused; of 32 patients included in the efficacy analysis, 28 (88%; 95% CI, 71–97) had complete resolution of vaso-occlusive events (VOEs) between months 6 and 18 after infusion, and 30 (94%) had complete resolution of severe VOEs.

So “93.5% versus 88%” is not measured with one ruler. The 88% concerns all VOEs, including events that do not meet the definition of “severe”; the figure closer to Casgevy’s “severe VOC” is the 94%, and even then the two sides define “severe” differently. A 2025 paper in Adv Ther addresses exactly this question: it compared how seven trials defined vaso-occlusion and found that in the exa-cel trial a severe VOC included any acute pain event requiring a visit to a medical facility, whereas in the lovo-cel trial a severe VOE required a stay of at least 24 hours in a hospital or emergency observation unit, or at least two visits within 72 hours; when the latter definition was applied to patient-level exa-cel data, all 30 patients (100%) were free of severe events for 12 months. The paper was funded by Vertex (the holder of Casgevy), and its authors include Frangoul and Locatelli, investigators in the exa-cel trial, and three Vertex employees; the definitional comparison is descriptive, the 100% comes from a post hoc analysis, and the conclusion is only that endpoints across different trials should be compared with caution. Both trials are single-arm, with different observation windows, entry thresholds, event definitions and denominators, and there is no head-to-head comparison between them.

What the boxed warning says

The title of Lyfgenia’s boxed warning is Hematologic Malignancy, and its first sentence reads: “Hematologic malignancy has occurred in patients treated with LYFGENIA.”

Behind the warning are three cases, described in section 5.1 of the prescribing information: at the time of approval, two patients who received an early version (a different manufacturing process and transplant procedure, Group A) had developed acute myeloid leukemia (AML), and one Group C patient with α-thalassemia trait had been diagnosed with myelodysplastic syndrome (MDS), the diagnosis made after month 24. One of the AML cases was published by Goyal and colleagues in the N Engl J Med in 2022: the patient fell ill about 5.5 years after infusion, and an integration site of the vector was detected in the leukemic cells, but based on the integration location, the very low transgene expression and the absence of an effect on expression of surrounding genes, the authors judged the leukemia unlikely to be related to vector integration, and noted that patients with sickle cell disease already face a higher risk of hematologic malignancy after transplantation. The FDA label likewise says two things: sickle cell disease itself carries a higher risk than the general population, and the additional hematopoietic stress from mobilization, myeloablation and infusion may also increase risk. The label separately lists insertional oncogenesis as a potential risk and requires lifelong monitoring: a complete blood count every 6 months for at least 15 years, and integration site analysis at months 6 and 12 and as needed.

Casgevy’s prescribing information (revised July 2026) carries no boxed warning. Its Warnings and Precautions cover neutrophil engraftment failure, delayed platelet engraftment and hypersensitivity reactions, plus one entry on off-target editing: because of genetic variation, the risk of unintended off-target editing in an individual’s CD34⁺ cells cannot be excluded, and its clinical significance is unknown. The 2024 trial paper reported that no cancers occurred.

Reading these two paragraphs starts with the length of follow-up: median follow-up was 19.3 months (0.8–48.1) in the Casgevy trial and 38 months (12–61) in Lyfgenia’s Study 1-C. The AML occurred in a different cohort of Group A patients treated with a different process; it suggests that events of this kind may appear years later, which means follow-up time itself limits what can be concluded about any of these drugs. It is not data about Casgevy and cannot be used to infer whether Casgevy’s risk is high or low. Patients on both drugs will enter long-term follow-up studies; the Casgevy label specifies 15 years in total after infusion.

A door they share

The two routes converge at the costliest step: both labels require myeloablative conditioning before infusion. The Lyfgenia label specifies busulfan; the Casgevy label says busulfan or another myeloablative regimen, with busulfan used in the trial, at a starting dose of 3.2 mg/kg/day from age 12 years. This risk has nothing to do with the vector or the editing method: of the 45 patients in the Lyfgenia safety population who started conditioning, all (45/45) had at least one adverse event attributed to conditioning, and stomatitis accounted for 71% of grade ≥3 adverse reactions; the Casgevy label states that 100% of SCD and TDT patients had grade 3 or 4 neutropenia and thrombocytopenia. The tables in both labels note that they include events related to myeloablation, so when reading safety numbers it matters to separate what comes from the shared conditioning and autologous transplant. At the same time, the Lyfgenia label considers that hematopoietic stress may increase the risk of hematologic malignancy, so the boxed warning cannot be read as unrelated to this shared procedure. Why myeloablation is necessary and whether it can be avoided is taken up in articles 61 and 68.

What remains unanswered

Vector integration and off-target editing each occupy one mechanistic unknown in the two labels: for Lyfgenia, insertional oncogenesis is listed as a potential risk, while the hematologic malignancies that have occurred were judged in the published literature unlikely to be related to the vector; for Casgevy, no off-target editing event has been reported, and its clinical significance is unknown. Both will have to be answered by longer follow-up.

On December 8, 2023, patients were handed two options. They look like two versions of the same procedure, but at the front of one label there is a black box.

References

  1. FDA. FDA Approves First Gene Therapies to Treat Patients with Sickle Cell Disease (news release, December 8, 2023). fda.gov
  2. FDA. LYFGENIA (lovotibeglogene autotemcel) Prescribing Information, revised 12/2023, BLA STN 125788. fda.gov
  3. FDA. CASGEVY (exagamglogene autotemcel) Prescribing Information, revised 07/2026, BLA STN 125787. fda.gov
  4. Frangoul H, et al. Exagamglogene Autotemcel for Severe Sickle Cell Disease. N Engl J Med. 2024;390(18):1649-1662. DOI (NCT03745287)
  5. Frangoul H, et al. CRISPR-Cas9 Gene Editing for Sickle Cell Disease and β-Thalassemia. N Engl J Med. 2021;384(3):252-260. DOI
  6. Wu Y, et al. Highly efficient therapeutic gene editing of human hematopoietic stem cells. Nat Med. 2019;25(5):776-783. DOI
  7. Kanter J, et al. Biologic and Clinical Efficacy of LentiGlobin for Sickle Cell Disease. N Engl J Med. 2022;386(7):617-628. DOI (NCT02140554)
  8. Kanter J, et al. Lovo-cel gene therapy for sickle cell disease: Treatment process evolution and outcomes in the initial groups of the HGB-206 study. Am J Hematol. 2023;98(1):11-22. DOI
  9. Goyal S, et al. Acute Myeloid Leukemia Case after Gene Therapy for Sickle Cell Disease. N Engl J Med. 2022;386(2):138-147. DOI
  10. Ribeil JA, et al. Gene Therapy in a Patient with Sickle Cell Disease. N Engl J Med. 2017;376(9):848-855. DOI (NCT02151526)
  11. Frangoul H, Locatelli F, et al. Impact of Different Definitions of Vaso-Occlusion on Efficacy Assessments in Sickle Cell Disease Clinical Trials. Adv Ther. 2025;42:2490-2499. DOI
  12. ClinicalTrials.gov. NCT03745287 (CLIMB SCD-121); NCT02140554 (HGB-206). clinicaltrials.gov