To the Editor:

Chronic myeloid leukemia (CML) is a myeloproliferative neoplasm mainly characterized by the excessive production of white blood cells. In 2018, the CML incidence in France was approximately 1.0–1.5/100 000 persons with a median age of 60–62 years, 55% of whom were men.¹ The first tyrosine kinase inhibitor (TKI) was imatinib, approved in 2001, which have improved the prognosis of this disease with patients reaching the life expectancy of the general population. Due to resistance or adverse drug reactions, other drugs have been developed for this indication: dasatinib and nilotinib, both recommended for first-line treatment (LT) by the European LeukemiaNet (ELN), bosutinib, and ponatinib.² Few data are available regarding the real-life use of TKIs for CML, especially on the entire population of a country, but these data can provide a description of care pathways, which are sometimes complex in real-life conditions.^3-5

The objective of this study was to describe TKI treatment courses in patients with CML who initiated treatment between 2014 and 2017 and who were followed for 3 years, using data from the French Health Data System (SNDS).

This study was based on the SNDS, which covers 99% of the French population (i.e., approximately 67 million inhabitants) and compiles several data from two sources: the French health insurance claims database and the French hospital and discharge database (online supplementary method). We included all adult patients (≥18 years old) newly treated with TKI for CML: imatinib, dasatinib, nilotinib, bosutinib, and ponatinib, between 01 January 2014 and 31 December 2017 in France. Patients with CML were identified using an algorithm already used within the SNDS⁶ (Table S1). Incident TKI users were defined as having at least one reimbursement for a TKI of interest between 01 January 2014 and 31 December 2017 and no TKI reimbursement during the 2 years prior to the first reimbursement. The index date is the date of the first reimbursement of TKI. Each patient was followed 3 years after the index date, that is, until 2020 for patients included in 2017, or until disease progression, defined as the blast phase (BP), or death, whichever came first. The five TKIs were identified in the database using the ATC code (Table S1). The first TKI was defined as the TKI dispensed at the index date. If the patient received ≥2 different TKIs, the 2-LT was defined as the second TKI prescribed chronologically and so on for the subsequent LT (third, fourth, fifth). Individuals who switched were defined as patients with at least two different TKIs during follow-up. The time frame was defined as the period between the index date and the date of the switch. A treatment interruption was defined as a gap of 90 days or more between two TKI dispensations. Medical comorbidities were identified at the index date, for the five previous years, using LTD and hospital diagnostic codes according to ICD-10 (Table S1). A Sankey diagram was designed to visualize the different LT for each patient using SankeyMatic. Characteristics of TKI utilization and time frames were described in relation to the first-line TKI. Cumulative incidence functions of the first switch according to the first TKI prescription and Gray's test (competitive risk, in our case death and BP, were taken into account) were used to estimate whether occurrence of the first switch was different. To determine whether the utilization pattern differed over time, we also described first-line and second-line prescription according to the year of prescription. Analyses were carried out using the SAS EG® 7.15 software.

Of the 8701 individuals with a first TKI reimbursement between 2014 and 2017, we identified 4162 CML adults newly treated with a TKI (Figure S1). The median age was 62 years and 56.3% of the patients were male (Table S1). More than 16.5% had cardiovascular comorbidities, 15.7% diabetes and 42.2% had antihypertensive drugs.1-L imatinib patients were older than 1-L nilotinib or dasatinib patients (median age, years): 65 versus 49 and 52 respectively, and had more comorbidities (cardiovascular comorbidities: 19.5% vs. 4.4% and 10.2%). During the 3-year follow-up, death occurred in 10.8% and progression to BP in 1.6% (median time of 526 and 454 days, respectively) of the patients. CML patients with death or disease progression presented more comorbidities (Table S1).

Approximately 70% of the overall population had only one TKI during the 3-year follow-up and among individuals alive and in the chronic phase at the end of follow-up, 88% were still under TKI treatment (Table S1). Individuals had approximately one dispensation per month with a median of 34 dispensations [IQR: 24–37] over the 36 months with a median time of 30 days [27–35] between dispensations. Treatment was interrupted for approximately 20% of the patients.

The main 1-LT were imatinib (78.4%) and nilotinib (18.2%). Dasatinib was the main 2-LT prescribed (approximately 70%) and nilotinib the main 3-LT (33%). Bosutinib and ponatinib were mainly prescribed in 2-LT or a subsequent line, unlike imatinib (Figure 1). More frequent courses were patients on imatinib alone (53%) followed by imatinib-dasatinib or nilotinib alone (both 12%).

Approximately 32% of the individuals switched at least once. Half of the population that switched did so the first time during the first year of treatment (median time [IQR]: 356 days [174–549]). During the first and second year of follow-up, a switch occurred in 17.0% and 14.4% in 1-LT imatinib patients, 13.2% and 15.3% in 1-LT nilotinib patients and 16.4% and 15.7%, respectively, in 1-LT dasatinib patients. TKI switch during follow-up was not significantly different in patients who started with imatinib, nilotinib, and dasatinib (Gray's test; p = .86, Figure S1).

The pattern of first-line prescription was similar across the study period (imatinib: 75%–80%, nilotinib: 16%–20%) (Figure S1). Second-line prescription differed over time, with bosutinib and ponatinib becoming more and more prescribed while dasatinib and nilotinib prescriptions decreased.

In this study, TKI LT initiation in new patients with CML is consistent with a previous SNDS study (2011–2014) and with ELN recommendations (2013).^{2, 5} Imatinib and nilotinib accounted for 95% of the initial treatments, while dasatinib was not legally reimbursed in 1-LT in France but was mainly used in 2-LT. While the rate of initial imatinib prescription is the same across studies (approximately 80% of the patients), the TKI used in 2-LT differs according to the country: either nilotinib was the main TKI used in 2-LT, or nilotinib and dasatinib were used at equal rates for 2-LT.^{3, 5}

Characteristics of the CML population described in our study are similar to the CML population in a French cancer registry in terms of age and gender ratio (median age of 62 years and 56% male) and is consistent with the incidence rate in France.¹ 1-LT imatinib patients were older and had more comorbidities than 1-LT nilotinib patients, which is consistent with other studies and recommendations since imatinib is a safer option for patients with comorbidities, although it may be less potent.

Initial TKI prescription took contraindications into account and potential TKI side effects. Dasatinib was less likely to be initiated in patients with respiratory comorbidities and those who started with nilotinib had fewer comorbidities, especially cardiovascular comorbidities, peripheral arterial diseases and stroke. The results of this study also showed chronic use of TKI treatments in CML patients (median of 34 dispensations during the 36-month follow-up). Switches rate in the first and second year were similar to that found in the SIMPLICITY study.⁴The SNDS covers the entire French population, which allowed us to exhaustively describe TKI LT at a national level. We used an identification algorithm previously used and similar to another which has been validated to identify patients with CML in this database.^{5, 6} Exposure to drugs was identified through reimbursement data prospectively collected, thereby avoiding recall and misclassification biases. We had no information on drug administration, but chronic use of this treatment with a dispensation every month substantially limits this type of bias.

Dispensation drugs within clinical trials are not identifiable in the SNDS, and some dispensations may not have been taken into account for the identification of LT or treatment characteristics. However, the number of dispensations in this setting is expected to be low as most clinical trials that included CML patients in France had already ended before the study period or were to end at the end of December 2014. We do not have access to clinical data such as the EUTOS score or results of biological tests, neither can we identify reasons for TKI switches. However, we had information for patients' comorbidities with diagnostic codes (CIM-10) in hospitalized patients and dispensation of all reimbursed drugs in an outpatient setting (antidiabetics, anti-hypertensive drugs…). We also did not have clinical information about disease progression, treatment intolerance or level of response. We identified CML patients who were hospitalized for BP and this rate might be underestimated.

This study illustrates the plurality of TKI treatment courses in real life, in newly diagnosed CML patients over a recent period and at a national level, and provides information about real-life practice.

致编辑：

慢性粒细胞白血病（CML）是一种骨髓增生性肿瘤，主要特征是白细胞过度产生。2018 年，法国 CML 发病率约为 1.0-1.5/100 000 人，中位年龄为 60-62 岁，其中 55% 为男性。¹第一个酪氨酸激酶抑制剂 (TKI) 是 2001 年获批的伊马替尼，它改善了这种疾病的预后，患者达到了一般人群的预期寿命。由于耐药或药物不良反应，已针对该适应症开发了其他药物：达沙替尼和尼罗替尼，均被欧洲白血病网（ELN）推荐用于一线治疗（LT）、博舒替尼和普纳替尼。²关于 CML TKI 在现实生活中的使用数据很少，特别是关于一个国家的全部人口，但这些数据可以提供对护理途径的描述，这些途径在现实生活中有时是复杂的。^3-5

本研究的目的是使用来自法国健康数据系统 (SNDS) 的数据，描述在 2014 年至 2017 年间开始治疗并随访 3 年的 CML 患者的 TKI 治疗过程。

本研究基于覆盖 99% 的法国人口（即约 6700 万居民）的 SNDS，并汇编了来自两个来源的多个数据：法国健康保险索赔数据库和法国医院和出院数据库（在线补充方法） . 我们纳入了 2014 年 1 月 1 日至 2017 年 12 月 31 日期间在法国新接受 TKI 治疗 CML 的所有成年患者（≥18 岁）：伊马替尼、达沙替尼、尼罗替尼、波舒替尼和波纳替尼。^{使用 SNDS 6}中已使用的算法识别 CML 患者（表 S1）。事件 TKI 用户被定义为在 2014 年 1 月 1 日至 2017 年 12 月 31 日期间至少有一次对感兴趣的 TKI 的报销，并且在第一次报销之前的 2 年内没有 TKI 报销。索引日期是 TKI 首次报销的日期。在索引日期后 3 年对每位患者进行随访，即 2017 年纳入的患者至 2020 年，或直至疾病进展，定义为爆发期 (BP) 或死亡，以先到者为准。使用 ATC 代码在数据库中识别了五个 TKI（表 S1）。第一个 TKI 被定义为在索引日期分发的 TKI。如果患者接受了≥2 种不同的 TKI，则 2-LT 被定义为按时间顺序开出的第二个 TKI，以此类推，用于随后的 LT（第三个、第四个、第五个）。转换的个体被定义为在随访期间至少有两种不同 TKI 的患者。时间范围被定义为索引日期和转换日期之间的时间段。治疗中断定义为两次 TKI 分配之间的间隔为 90 天或更长时间。根据 ICD-10（表 S1），使用 LTD 和医院诊断代码在索引日期确定了前五年的医学合并症。Sankey 图旨在使用 SankeyMatic 可视化每位患者的不同 LT。与一线 TKI 相关的 TKI 使用和时间框架的特征进行了描述。根据第一个 TKI 处方和 Gray 测试（竞争风险，在我们的案例中，死亡和 BP，被考虑）用于估计第一次切换的发生是否不同。为了确定使用模式是否随时间而不同，我们还根据处方年份描述了一线和二线处方。使用 SAS EG® 7.15 软件进行分析。

在 2014 年至 2017 年间首次获得 TKI 报销的 8701 名个体中，我们确定了 4162 名新接受 TKI 治疗的 CML 成人（图 S1）。中位年龄为 62 岁，56.3% 的患者为男性（表 S1）。超过 16.5% 有心血管合并症，15.7% 糖尿病和 42.2% 有抗高血压药物。1-L 伊马替尼患者年龄大于 1-L 尼罗替尼或达沙替尼患者（中位年龄，岁）：分别为 65 岁、49 岁和 52 岁，并且有更多合并症（心血管合并症：19.5% vs. 4.4% 和 10.2%）。在 3 年的随访中，10.8% 的患者死亡，1.6% 的患者进展为 BP（中位时间分别为 526 天和 454 天）。死亡或疾病进展的 CML 患者出现更多的合并症（表 S1）。

在 3 年的随访期间，大约 70% 的总人口只有一次 TKI，在随访结束时存活和处于慢性期的个体中，88% 仍在接受 TKI 治疗（表 S1）。在 36 个月中，每个人每月大约有一次分配，中位数为 34 次分配 [IQR：24-37]，分配之间的中位时间为 30 天 [27-35]。大约 20% 的患者中断了治疗。

主要的 1-LT 是伊马替尼 (78.4%) 和尼罗替尼 (18.2%)。达沙替尼是主要的 2-LT 处方（约 70%），尼罗替尼是主要的 3-LT（33%）。与伊马替尼不同，博舒替尼和波纳替尼主要用于 2-LT 或后续线（图 1）。更频繁的课程是单独使用伊马替尼（53%）的患者，然后是单独使用伊马替尼-达沙替尼或尼罗替尼（均为 12%）。

大约 32% 的人至少换过一次。有一半的人在治疗的第一年第一次这样做（中位时间 [IQR]：356 天 [174–549]）。在随访的第一年和第二年，1-LT 伊马替尼患者的转换发生率为 17.0% 和 14.4%，1-LT 尼罗替尼患者为 13.2% 和 15.3%，1-LT 患者分别为 16.4% 和 15.7%。 LT 达沙替尼患者。在开始使用伊马替尼、尼罗替尼和达沙替尼的患者中，随访期间的 TKI 转换没有显着差异（格雷检验；p  = .86，图 S1）。

在整个研究期间，一线处方的模式相似（伊马替尼：75%–80%，尼罗替尼：16%–20%）（图 S1）。二线处方随时间变化，博舒替尼和波纳替尼处方越来越多，而达沙替尼和尼罗替尼处方减少。

在这项研究中，新 CML 患者的 TKI LT 启动与之前的 SNDS 研究（2011-2014 年）和 ELN 建议（2013 年）一致。^2、5伊马替尼和尼罗替尼占初始治疗的95%，而达沙替尼在法国1-LT没有合法报销，主要用于2-LT。虽然各个研究的初始伊马替尼处方率相同（约 80% 的患者），但 2-LT 中使用的 TKI 因国家/地区而异：尼罗替尼是 2-LT 中使用的主要 TKI，或尼罗替尼和达沙替尼2-LT 的使用率相同。^3、5

我们研究中描述的 CML 人群的特征在年龄和性别比例（中位年龄 62 岁和 56% 男性）方面与法国癌症登记处的 CML 人群相似，并且与法国的发病率一致。¹ 1-LT 伊马替尼患者比 1-LT 尼罗替尼患者年龄更大且合并症更多，这与其他研究和建议一致，因为伊马替尼对于合并症患者来说是更安全的选择，尽管它可能不太有效。

最初的 TKI 处方考虑了禁忌症和潜在的 TKI 副作用。有呼吸道合并症的患者不太可能开始使用达沙替尼，而开始使用尼罗替尼的患者合并症较少，尤其是心血管合并症、外周动脉疾病和中风。该研究的结果还显示 CML 患者长期使用 TKI 治疗（在 36 个月的随访中，中位数为 34 次）。第一年和第二年的转换率与 SIMPLICITY 研究中发现的相似。⁴SNDS 涵盖了整个法国人口，这使我们能够在国家层面详尽地描述 TKI LT。我们使用了一种以前使用过的识别算法，并且与另一种算法相似，该算法已被验证用于识别该数据库中的 CML 患者。^{5, 6}通过前瞻性收集的报销数据确定药物暴露情况，从而避免召回和错误分类偏差。我们没有关于药物管理的信息，但长期使用这种治疗并每月分配一次，大大限制了这种类型的偏见。

临床试验中的配药药物在 SNDS 中无法识别，并且某些配药可能没有被考虑用于识别 LT 或治疗特征。然而，由于在法国包括 CML 患者在内的大多数临床试验在研究期间之前已经结束或将在 2014 年 12 月底结束，因此预计这种情况下的分配数量会很低。我们无法获得临床数据例如 EUTOS 分数或生物测试结果，我们也无法确定 TKI 开关的原因。然而，我们有住院患者的合并症诊断代码 (CIM-10) 和门诊所有报销药物（降糖药、抗高血压药……）的信息。我们也没有关于疾病进展的临床信息，治疗不耐受或反应水平。我们确定了因 BP 而住院的 CML 患者，这个比率可能被低估了。

这项研究说明了现实生活中的多个 TKI 治疗课程，在最近一段时间内和在国家一级新诊断的 CML 患者中，并提供了有关现实生活实践的信息。