Luspatercept is the first erythropoiesis-modulating agent approved by the Food and Drug Administration (FDA) and the European Medicines Agency (EMA) for treating anemia in adult transfusion-dependent β-thalassemia (TDT) patients. As observed in clinical trials¹ and real-life experience,² response to luspatercept in TDT is heterogeneous. It can range from patients who do not respond to those who become transfusion-independent. So far, no predictors of response have been identified. However, the definition of the different profiles and predictors of response is necessary for optimizing treatment allocation, limiting costs, and increasing sustainability. The ELEMENT study is an observational prospective cohort study that enrolled adult TDT patients regularly followed at Fondazione IRCCS Ca' Granda Ospedale Maggiore Policlinico in Milan (Italy) treated with luspatercept. Luspatercept was administered, according to the indications of the Italian Regulatory Agency, subcutaneously at a starting dose of 1.0 mg/kg every 3 weeks and was increased to 1.25 mg/kg after dose 3 if a clinical response was not achieved.³ The drug was discontinued according to the Italian Regulatory Agency indications, namely, if the patient does not achieve transfusion reduction (amount not specified) after three doses at the maximum dosage or in the presence of unacceptable toxicity. Treatment response was assessed by comparing the transfusion burden (TB) during any 12-week treatment period with that in the 24 weeks before treatment. Responders (RSP) were defined as individuals who have a reduction of the TB ≥33% in any 12 weeks of treatment, while those with a TB reduction <33% were considered “non-responders” (NR). Moreover, as in the phase 3 trial, transfusion independence was defined as a transfusion-free period of at least 8 weeks.¹ In this study, TB was also expressed as the number of units of packed red blood cells (pRBC) per week (unit/week) by dividing the number of units transfused in a period by the number of weeks evaluated.

Between January 1, 2021, and May 31, 2024, 56 TDT patients received at least one dose of luspatercept after drug authorization. At the start of the study, treatment was offered to TDT patients with high TB, iron overload demonstrated by T2* magnetic resonance imaging (MRI), or other clinically relevant conditions that might benefit from TB reduction. Subsequently, the treatment was made available to all the patients who met the Italian regulatory agency criteria and were willing to receive the drug. The decision to initiate treatment was always discussed between the physician and the patient.

Seven patients discontinued the drug before completing at least 12 weeks. The reasons for discontinuation are detailed in Table S1 (early drug discontinuation group). At the time of the analysis, one patient had not yet completed 12 weeks of treatment.

Data from the 48 patients who received luspatercept for at least 12 weeks were included, and their characteristics at enrollment are presented in Figure 1A. Out of 48 patients, two paused their treatment temporarily, one due to a personal decision and the other due to an adverse event (AE). They were later re-challenged with the treatment and achieved a response similar to the first period, resulting in both patients attaining transfusion independence. The median age was 41, and 44% (21/48) were females. In this sample, 38% (18/48) were splenectomized. Sixteen out of 48 had a β0/β0 genotype, and one had HbE/β-thalassemia. Overall, 37/48 (77%) patients had a TB >15 units in the 24 weeks prior to treatment initiation. Among these, 12 patients had a TB ranging from 20 to 24 units. The median treatment period was 48 weeks (12–172), and 39 out of 48 (81%) received treatment for at least 24 weeks. Seventeen out of 48 (35%) were responders (Figure S1), of whom 10/48 (21%) showed a TB reduction ≥33% in weeks 13–24 (the primary endpoint of the phase 3 trial). Eleven out of 17 (65%) patients in the RSP group had a TB reduction of ≥50% for at least one 12-week interval. Four patients became transfusion-independent for at least 12 weeks, of whom three for at least 22 consecutive weeks. TB, expressed as unit/week, did not change in the NR group (0.7 unit/week in the 24 weeks before luspatercept vs. 0.7 unit/week during the treatment, p = 0.95), while it decreased in the RSP group (0.8 unit/week in the 24 weeks before luspatercept vs. 0.5 unit/week during the treatment, p = 0.001). Overall, in our sample, the mean pretreatment pre-transfusion Hb (pt-Hb) was similar in both groups (Figure 1A). The pt-Hb increased between pre- and during-treatment in both NR (9.1 ± 0.5 g/dL pretreatment vs. 9.3 ± 0.4 g/dL under treatment, p = 0.03) and RSP (from 9.2 ± 0.6 to 9.5 ± 0.5 g/dL, p = 0.025) groups.

Twenty-one out of 48 (44%) discontinued treatment after at least 12 weeks, with 16 being NR. The reasons for discontinuation are detailed in Table S1.⁴

The drug was generally well tolerated, and no deaths occurred. The frequency of AEs was similar to those reported in the phase 3 study, and a detailed list can be found in Table S2.

RSP and NR did not differ in TB during the 24 weeks before treatment initiation when evaluating the number of transfused units, even though iron intake was slightly higher in the NR (0.32 ± 0.07 vs. 0.27 ± 0.08 mg/kg/day, p = 0.09). Patients were overall well chelated. When comparing the two groups, RSP showed higher liver iron concentration (LIC), in a range considered borderline and in the lower range of optimal chelation therapy. As shown in Figure S2 after 24 weeks of treatment, we observed a greater increase in reticulocyte count in RSP than in NR (0.2, 0.0–1.1 × 10¹²/L vs. 0.1, 0.0–0.5 × 10¹²/L, p = 0.04).

Fetal hemoglobin (HbF) at baseline was the only clinically relevant hematological parameter significantly different between RSP and NR. Also, the baseline absolute value of HbF (g/dL) provided the best discrimination between RSP and NR after a receiver operating characteristic (ROC) analysis, with an area under the curve (AUC) of 0.82 (95% confidence interval [CI] 0.68–0.96; Figure 1B). Baseline values of reticulocytes did not increase the AUC. The statistical cut-off point of the ROC curve was identified for a baseline absolute value of HbF of 0.6 g/dL, corresponding to a negative predictive value (NPV) of 92% (95% CI: 74–99, 23 NR out of 25 with HbF <0.6 g/dL).

An increase in HbF was also observed in both groups; however, the RSP reached higher values than the NR at 24 weeks, 21.1% (4.2–59.1; corresponding to the median absolute HbF of 2.1 g/dL) versus 7.6% (3.2–37.1; corresponding to the median absolute value of HbF of 0.7 g/dL), respectively (p = 0.0001). In the random-intercept linear regression analysis, the best fit was achieved using a cubic model to describe the trend of HbF increase over time. Both RSP and NR exhibited a similar pattern, with an initial increase in levels that reached a plateau around week 16 (Figure 1C).

Currently, data on luspatercept safety, efficacy, and potential predictors of response primarily come from clinical trials.^{1, 2, 5} Here, we present findings from the largest published real-world cohort of TDT patients treated with luspatercept. Our study explores predictors of response, which are essential for clinicians to assess the risk–benefit profile when considering prescribing this agent. In our study, the response rate was similar to that of the phase 3 trial when evaluating weeks 13–24, namely the primary endpoint of the BELIEVE study. However, the overall response rate over any 12-week interval in our cohort was lower (35% vs. about 70% in the phase 3 trial). This difference could be attributed to the shorter median treatment duration in our study (48 vs. 64 weeks). Reasons for this include adherence to regulatory guidelines for drug prescription and the patients' involvement in treatment decisions in real life. This can affect, for example, the identification of late-responders.⁶ In our sample, most patients completed 24 weeks of treatment, which is a reasonable time for expecting a response.⁷ Other differences emerged when comparing our data to those of the trial. Our patients were more heavily transfused, with 77% having a TB >15 units of pRBC in the 24 weeks before the initiation of luspatercept, compared to 43.5% in the BELIEVE trial. Interestingly, we have identified baseline HbF levels as a predictor of response to treatment. A baseline HbF <0.6 g/dL was associated with a lower likelihood of response to luspatercept (8% of RSP having baseline HbF <0.6 g/dL). Conversely, TDT patients who have a baseline HbF of 0.6 g/dL or higher exhibit a high likelihood of responding to treatment (64% of patients with HbF >0.6 g/dL were RSP). Of note, those with a baseline HbF <0.6 g/dL did not have a higher TB before luspatercept (median TB in the HbF <0.6 g/dL group: 19 units vs. median TB in the HbF >0.6 g/dL group: 17 units; p = 0.11). We do not recommend excluding a priori patients with a baseline value of HbF <0.6 g/dL from treatment. Rather, we suggest carefully considering initiating luspatercept in these patients, especially in the presence of an uncertain risk–benefit ratio (e.g., a risk factor for an AE). Interestingly, regardless of the baseline level, all patients treated with luspatercept experienced an increase in HbF. However, responders showed a higher increase in absolute HbF values compared to NRs.

Additionally, the presence of severe genotypes did not prevent a response, as one-fourth of RSP had a β0/β0 genotype.

Some limitations in this study should be acknowledged. First, the sample size is limited given the rarity of the disease and the recent approval of luspatercept in Italy. Second, the observation period in this study was shorter than that of the Phase 3 BELIEVE trial.

In conclusion, this study confirms that the effectiveness of luspatercept in a real-world cohort of TDT patients is similar to the phase 3 clinical trial and identifies HbF at baseline and its trend in the first few weeks of treatment as a predictive marker of response. Further studies are needed to elucidate the differences in baseline HbF levels between responders and NRs.

Luspatercept 是美国食品药品监督管理局 （FDA） 和欧洲药品管理局 （EMA） 批准用于治疗成人输血依赖性β-地中海贫血 （TDT） 患者贫血的首个红细胞生成调节剂。正如临床试验¹ 和实际经验 ² 中观察到的那样，TDT 对 luspatercept 的反应是异质的。其范围从无反应的患者到不依赖输血的患者。到目前为止，尚未确定反应的预测因子。然而，定义不同的反应概况和预测因子对于优化治疗分配、限制成本和提高可持续性是必要的。ELEMENT 研究是一项观察性前瞻性队列研究，在米兰（意大利）的 Fondazione IRCCS Ca' Granda Ospedale Maggiore Policlinico 定期随访接受 luspatercept 治疗的成年 TDT 患者。根据意大利监管机构的适应症，Luspatercept 以每 3 周 1.0 mg/kg 的起始剂量皮下给药，如果未达到临床反应，则在第 3 剂后增加至 1.25 mg/kg。³ 根据意大利监管机构的适应症，即如果患者在以最大剂量服用 3 剂后或存在不可接受的毒性后未达到输血减少（未指定量）。通过比较任何 12 周治疗期间的输血负担 （TB） 与治疗前 24 周的输血负担 （TB） 来评估治疗反应。反应者 （RSP） 被定义为在任何 12 周的治疗中结核病减少 ≥33% 的个体，而结核病减少 <33% 的个体被认为是“无反应者”（NR）。 此外，与 3 期试验一样，不依赖输血被定义为至少 8 周的无输血期。¹ 在这项研究中，结核病也表示为每周（单位/周）的浓缩红细胞 （pRBC） 单位数，方法是将一段时间内输血的单位数除以评估的周数。

在 2021 年 1 月 1 日至 2024 年 5 月 31 日期间，56 名 TDT 患者在药物授权后接受了至少一剂 luspatercept。在研究开始时，为患有高结核病、T2* 磁共振成像 （MRI） 显示的铁过载或其他可能受益于结核病减少的临床相关病症的 TDT 患者提供治疗。随后，所有符合意大利监管机构标准并愿意接受该药物的患者都可以获得该治疗。开始治疗的决定总是由医生和患者讨论。

7 例患者在完成至少 12 周前停药。停药原因详见表 S1 （早期停药组）。在分析时，1 例患者尚未完成 12 周的治疗。

包括来自接受 luspatercept 至少 12 周的 48 名患者的数据，他们在入组时的特征如图 1A 所示。在 48 名患者中，两名暂时暂停治疗，一名是由于个人决定，另一名是由于不良事件 （AE）。他们后来再次接受治疗并获得了与第一阶段相似的反应，导致两名患者都实现了输血独立。中位年龄为 41 岁，其中 44% （21/48） 为女性。在该样本中，38% （18/48） 被脾切除。48 例中有 16 例具有 β0/β0 基因型，1 例患有 HbE/β-地中海贫血。总体而言，37/48 （77%） 患者在治疗开始前 24 周内患有 TB >15 单位。其中，12 名患者患有 20 至 24 个单位的结核病。中位治疗期为 48 周 （12-172），48 人中有 39 人 （81%） 接受了至少 24 周的治疗。48 人中有 17 人 （35%） 有反应（图 S1），其中 10/48 （21%） 显示第 13-24 周（3 期试验的主要终点）结核病减少 ≥33%）。RSP 组的 17 名患者中有 11 名 （65%） 的结核病减轻了 ≥50%，间隔至少有一个 12 周。4 例患者至少 12 周不依赖输血，其中 3 例至少连续 22 周。结核病以单位/周表示，NR 组没有变化 （luspatercept 前 24 周为 0.7 单位/周 vs. 治疗期间为 0.7 单位/周，p = 0.95），而 RSP 组下降 （luspatercept 前 24 周为 0.8 单位/周 vs. 治疗期间为 0.5 单位/周，p = 0.001）。总体而言，在我们的样本中，两组的平均治疗前输血前 Hb （pt-Hb） 相似（图 1A）。NR 在治疗前和治疗期间 pt-Hb 增加 （9.1 ± 0.5 g/dL 预处理与 9.3 ± 0.4 g/dL 治疗组，p = 0.03）和 RSP（从 9.2 ± 0.6 到 9.5 ± 0.5 g/dL，p = 0.025）组。

48 例中有 21 例 （44%） 在至少 12 周后停止治疗，其中 16 例为 NR。停药原因详见表 S1。⁴

该药物总体耐受性良好，无死亡发生。AE 的频率与 3 期研究中报告的频率相似，详细列表见表 S2。

在评估输血单位的数量时，在治疗开始前 24 周内，结核病的 RSP 和 NR 没有差异，尽管 NR 中的铁摄入量略高 （0.32 ± 0.07 vs. 0.27 ± 0.08 mg/kg/天，p = 0.09）。患者总体螯合良好。在比较两组时，RSP 显示较高的肝铁浓度 （LIC），处于被认为是临界的范围内，处于最佳螯合治疗的较低范围内。如图 S2 所示，治疗 24 周后，我们观察到 RSP 中网织红细胞计数的增加大于 NR 中的增加 （0.2， 0.0-1.1 × 10¹²/L vs. 0.1， 0.0-0.5 ×^{10 12}/L，p = 0.04）。

基线时的胎儿血红蛋白 （HbF） 是 RSP 和 NR 之间唯一显著差异的临床相关血液学参数。此外，在受试者工作特征 （ROC） 分析后，HbF 的基线绝对值 （g/dL） 提供了 RSP 和 NR 之间的最佳区分，曲线下面积 （AUC） 为 0.82 （95% 置信区间 [CI] 0.68-0.96;图 1B）。网织红细胞的基线值没有增加 AUC。当 HbF 的基线绝对值为 0.6 g/dL 时，确定了 ROC 曲线的统计截断点，对应于 92% 的阴性预测值 （NPV）（95% CI：74-99,25 例 NR 中有 23 例 HbF <0.6 g/dL）。

在两组中也观察到 HbF 增加;然而，RSP 在 24 周时达到高于 NR 的值，分别为 21.1%（4.2-59.1;对应于 2.1 g/dL 的中位绝对 HbF）和 7.6%（3.2-37.1;对应于 HbF 的绝对中位值 0.7 g/dL）（p = 0.0001）。在随机截距线性回归分析中，使用三次模型来描述 HbF 随时间增加的趋势，从而实现最佳拟合。RSP 和 NR 都表现出相似的模式，最初水平增加，在第 16 周左右达到平台期（图 1C）。

目前，有关 luspatercept 安全性、有效性和潜在反应预测因子的数据主要来自临床试验。^1、2、5在这里，我们展示了已发表的最大已发表的接受 luspatercept 治疗的 TDT 患者真实世界队列的结果。我们的研究探讨了反应的预测因素，这对于临床医生在考虑开具这种药物时评估风险-收益状况至关重要。在我们的研究中，在评估第 13-24 周（即 BELIEVE 研究的主要终点）时，反应率与 3 期试验的反应率相似。然而，在我们的队列中，任何 12 周间隔的总体缓解率较低（35% vs. 3 期试验中约为 70%）。这种差异可归因于我们研究中较短的中位治疗持续时间 （48 周 vs. 64 周）。原因包括遵守药物处方的监管指南以及患者在现实生活中参与治疗决策。例如，这可能会影响对迟到响应者的识别。⁶ 在我们的样本中，大多数患者完成了 24 周的治疗，这是预期反应的合理时间。⁷ 在将我们的数据与试验数据进行比较时，还出现了其他差异。我们的患者输血量更大，77% 的患者在开始使用 luspatercept 前 24 周内患有结核病 >15 单位的 pRBC，而 BELIEVE 试验中为 43.5%。有趣的是，我们已经确定基线 HbF 水平是对治疗反应的预测因子。基线 HbF <0.6 g/dL 与对 luspatercept 反应的可能性较低相关 （8% 的 RSP 具有基线 HbF <0.6 g/dL）。相反，基线 HbF 为 0.6 g/dL 或更高表现出对治疗有反应的可能性很高 （64% 的 HbF >0.6 g/dL 患者是 RSP）。值得注意的是，基线 HbF <0.6 g/dL 的患者在接受 luspatercept 之前没有更高的结核病（HbF <0.6 g/dL 组的中位结核病：19 个单位，而 HbF >0.6 g/dL 组的中位结核病：17 个单位;p = 0.11）。我们不建议将基线值为 HbF <0.6 g/dL 的先验患者排除在治疗之外。相反，我们建议谨慎考虑对这些患者开始使用luspatercept，尤其是在风险收益比不确定（例如，AE的危险因素）的情况下。有趣的是，无论基线水平如何，所有接受 luspatercept 治疗的患者都经历了 HbF 的增加。然而，与 NR 相比，反应者显示绝对 HbF 值增加更高。

此外，严重基因型的存在并不会阻止反应，因为四分之一的 RSP 具有 β0/β0 基因型。

应该承认本研究中的一些局限性。首先，鉴于这种疾病的罕见性和最近意大利批准了 luspatercept，样本量有限。其次，本研究的观察期短于 3 期 BELIEVE 试验。

总之，本研究证实 luspatercept 在真实世界的 TDT 患者队列中的有效性与 3 期临床试验相似，并确定基线时的 HbF 及其在治疗最初几周的趋势作为反应的预测标志物。需要进一步的研究来阐明反应者和 NRs 之间基线 HbF 水平的差异。