To the Editor:

Heidenreich and colleagues described the insertion of central venous catheters (CVC) in the jugular vein (JV) as a main risk factor for CVC-related bloodstream infections (CRBSI) in patients with hematological malignancies.¹ This conclusion was based on a retrospective single-center study on 153 CVC in patients with high-dose chemotherapy for autologous hematopoietic stem cell transplantation (HSCT) or induction or consolidation chemotherapy for acute leukemia between January 2011 and June 2013. They compared CVC inserted in the JV (94; 61.4%) with CVC inserted in the subclavian vein (SV; 59; 38.6%). In total, 10 CRBSI cases were diagnosed, the CRBSI rate for JV-CVC was 9.6% (9/94) compared to 1.8% (1/59) for SV-CVC (p = .055). The corresponding CRBSI incidences were 5.7 CRBSI/1000 CVC days versus 1.2 CRBSI/1000 CVC days (no significance testing provided). In a univariate analysis, only the JV insertion site was identified as a risk factor for CRBSI (hazard ratio [HR] 5.405 [95% confidence interval {95% CI} 0.68–50.0]; p = .036); a multivariate analysis was not performed.¹

These results and, in particular, their strong interpretation raise some questions. Firstly, it is uncommon that, among other factors, not even neutropenia could be identified as a CRBSI risk factor. In stark contrast with the findings by Heidenreich et al.,¹ neutropenia is a well-established and important risk factor for CRBSI in patients with acute leukemia, as described by Bodey et al. many decades ago.² Furthermore, the advantage of SV-CVC regarding CRBSI risk is not clearly demonstrated so far despite several studies investigating CRBSI rates. A meta-analysis on 8501 CVC in ICU patients showed a comparable risk of CRBSI for JV and SV inserted CVC (risk ratio 1.16 [95% CI 0.57–2.36]).³ Accordingly, there is no preference in cancer patients for either jugular or subclavian as site of CVC insertion in evidence-based recommendations and guidelines, although femoral CVC insertions are discouraged.⁴ Notwithstanding the very limited sample size with a total of only 10 observed CRBSI, Heidenreich et al. draw a very far-reaching conclusions from their data when they state that the insertion site was the main risk factor for CVC-related complications in this patient population.¹

Since there seem to be different statements regarding CRBSI risk for JV inserted CVC, we aimed to confirm the results provided by Heidenreich et al.¹ in a different data set. We used data form the SECRECY registry (Study to Evaluate Central Venous Catheter-related Infections in Hematology and Oncology; German Clinical Trial Register [DRKS], no. DRKS00006551), an ongoing, clinical CRBSI registry starting in March 2013 in hematology and oncology departments, currently active in six German sites. Surveillance data are documented on CRBSI of non-selected, consecutive patients with short-term, non-tunneled JV, SV, or femoral vein CVC inserted for routine clinical use. All CVC were inserted according to local standard operating procedures. CRBSI were classified according to the 2012 Infectious Diseases Working Party (AGIHO) of the German Society of Hematology and Medical Oncology (DGHO) definition.⁴ The registry was approved by the central ethics committee (Magdeburg University Hospital, approval no. 84/14) as well as by respective local ethics committees. Registry data entered until February 2022 on JV-CVC and SV-CVC ≥1 day in situ of patients with hematological malignancy as underlying disease including patients undergoing autologous or allogeneic HSCT were used for this analysis. Only CRBSI cases classified as definite CRBSI⁴ were considered in order to apply the same diagnostic criteria as used by Heidenreich et al.¹ For direct comparison of both studies, we analyzed the same clinical and CVC parameters and endpoints respectively. Thus, we are able to directly compare our data with the Heidenreich study.¹ As time to CRBSI was one of the primary endpoints in the Heidenreich study, we compare that endpoint and additionally compared CRBSI rate, CRBSI incidence, and cumulative CRBSI incidence at day 15 after CVC insertion (CRBSI15). Statistical analysis was performed using OpenEpi, version 3.01 (Atlanta, GA, USA, https://www.openepi.com), and IBM® SPSS® Statistics, version 28 (Armonk, NY, USA). The statistical tests used are shown in the tables. Two-sided p values <.05 were considered statistically significant.

According to the inclusion criteria, we identified 3795 cases in the registry, including 220 SV-CVC (5.8%) (Table 1). Compared to the Heidenreich study,¹ our cohort included fewer cases with acute myeloid leukemia. Furthermore, in our cohort, 410 CVC (10.8%) were antimicrobial-coated, compared to none in the Heidenreich study.¹ On the other hand, only 1422 CVC (37.4%) had chlorhexidine gluconate (CHG)-coated dressings, while in the Heidenreich study all CVC were covered with CHG-coated dressings.¹

Our cohort included significantly more male patients in the JV-CVC group compared to the SV-CVC group (p = .016). In the JV-CVC group, the CVC time were significantly shorter (p < .001), and there were significantly fewer antimicrobial-coated CVC than in the SV-CVC cohort (p < .001).

Interestingly, we found a trend toward a lower CRBSI rate in the JV-CVC cohort (5.2% vs. 8.2%; p = .064), but there were no statistically significant differences regarding CRBSI incidence (p = .211) or time to CRBSI (p = .338) as well as in the CRBSI15 (p = .480) (Table 1).

To reduce the bias of co-variates, we additionally performed a matched-pair analysis, 1:1 matched for sex, diagnosis, CVC type, and CHG-coated dressing. Through this, 215 cases could be compared. We found no statistical difference neither for the CRBSI rate comparing JV-CVC with SV-CVC (11/215 [5.1%] vs. 18/215 [8.4%]; p = .248), nor for the CRBSI incidence (3.2 CRBSI/1000 CVC days vs. 4.5 CRBSI/1000 CVC days); p = .369), the time to CRBSI (HR 1.326 [95% CI 0.62–2.82]; p = .464) or the CRBSI15 (4.30% vs. 5.90%; p = .541) (Table 2).

Taken together, our results are in contrast with the Heidenreich study.¹ We found no significant difference in CRBSI rates between SV-CVC and JV-CVC and even observed a trend toward higher CRBSI rates in SV-CVC. These contraire results may be attributable to the low number of CVC cases or CRBSI events, respectively, in the Heidenreich study.¹ Furthermore, the results of the retrospective and observational Heidenreich study¹ are in contrast with results from a multicenter randomized controlled trial (RCT) on CRBSI regarding CHG-coated CVC dressings in neutropenic patients published 2016, the COAT trial.⁵ In this trial, CRBSI were significantly more frequent in patients with SV-CVC compared to JV-CVC (29/356 [8.1%] vs. 8/257 [3.1%]; p = .010). Importantly, the rate of CRBSI in the COAT trial were only cases of definite CRBSI,^{4, 5} which met the same diagnostic criteria as in our and in the study by Heidenreich et al.^{1, 4} Therefore, the results of these three discussed studies are comparable.

However, it must be critically noted, that in the mentioned COAT trial the majority of SV-CVC were recruited from only one center (319/366; 87.2%).⁵ A center effect seems very likely here that could have impacted the results of this trial. In our study, more than a half of the documented SV-CVC came from one center (Tables 1 and 2), therefore a center effect cannot be ruled out.

In the Heidenreich study, the great majority of patients suffered from acute leukemia (120/153 [78.4%]).¹ This proportion was lower in our study (1900/3795 [50.1%]; Table 1) and in the COAT trial (279/613 [45.6%].⁵ This could indicate a difference in the duration of neutropenia in these studies, which could limit comparability of the data. Unfortunately, in our registry, the duration of neutropenia is not documented. However, we can provide indirect data on the duration of neutropenia: In the whole cohort, 335 cases neutropenia was present at both time-points, CVC insertion and CVC removal. Assuming initial neutropenia was continuous in these cases, the median duration of neutropenia was 13 days (interquartile range 8–22), which is in the range of median values in patients with acute leukemia. According to this assumptions, the CRBSI incidence per 1000 neutropenic CVC days was statistically not different between JV-CVC and SV-CVC in the whole cohort (p = .766) as well in the meta-analysis (p = .403) (Table 2).

It is the nature of this non-controlled, real-world study that the procedures of CVC insertion and management of CVC as well as management of CRBSI were not uniform in the centers. Maybe this could have led to the different frequencies of CRBSI between the centers. In the whole cohort the CRBSI rate was 5.4% with a range of 1.5%–8.2% for all six centers; the CRBSI incidence was 3.3/1000 CVC days with a range of 0.9–6.8. Furthermore, the number of CVC documented in the registry varied between sites (159–1570, mean 633).

Epidemiological data on CRBSI derived from RCT may reliably be transferred to real-world settings.⁶ However, to the best of our knowledge, there are no published RCT available so far comparing CRBSI rates of JV-CVC and SV-CVC in patients with hematological malignancies. Therefore, any evidence addressed here has to derive from secondary endpoints of RCT or from real-world multicenter studies, or registries, with adequate sample sizes.

Based on the results of our large multicenter, registry-based (matched-pair) analysis presented here and the available studies so far, we think there is not enough evidence showing inferiority of jugular inserted CVC regarding CRBSI.

In conclusion, the jugular CVC insertion site cannot be considered as the main risk factor for CRBSI in patients with hematological malignancies. Further investigations, ideally including in form of RCT, seem warranted.

致编辑：

Heidenreich 及其同事将在颈静脉 (JV) 中插入中央静脉导管 (CVC) 描述为血液系统恶性肿瘤患者发生 CVC 相关血流感染 (CRBSI) 的主要危险因素。¹这一结论基于一项回顾性单中心研究，该研究纳入了 2011 年 1 月至 2013 年 6 月期间接受自体造血干细胞移植 (HSCT) 大剂量化疗或急性白血病诱导或巩固化疗的 153 例 CVC。他们比较了插入的 CVC在 JV (94; 61.4%) 中，CVC 插入锁骨下静脉 (SV; 59; 38.6%)。总共诊断出 10 例 CRBSI 病例，JV-CVC 的 CRBSI 率为 9.6% (9/94)，而 SV-CVC 为 1.8% (1/59)（p =.055）。相应的 CRBSI 发生率为 5.7 CRBSI/1000 CVC 天与 1.2 CRBSI/1000 CVC 天（未提供显着性检验）。在单变量分析中，只有 JV 插入部位被确定为 CRBSI 的危险因素（风险比 [HR] 5.405 [95% 置信区间 {95% CI} 0.68–50.0]；p = .036 ）  ；未进行多变量分析。^1个

这些结果，尤其是它们强有力的解释提出了一些问题。首先，很少见的是，除其他因素外，甚至中性粒细胞减少症也不能被确定为 CRBSI 风险因素。与 Heidenreich 等人的研究结果形成鲜明对比的是，正如 Bodey 等人所述，中性粒细胞减少症是急性白血病患者 CRBSI 的一个公认且重要的危险因素^。许多年前。²此外，尽管有几项研究调查了 CRBSI 发生率，但到目前为止，SV-CVC 在 CRBSI 风险方面的优势尚未得到明确证明。对 ICU 患者 8501 个 CVC 的荟萃分析显示，JV 和 SV 插入的 CVC 的 CRBSI 风险相当（风险比 1.16 [95% CI 0.57–2.36]）。^3个因此，尽管不鼓励股骨 CVC 插入，但在基于证据的建议和指南中，癌症患者并没有偏好颈静脉或锁骨下作为 CVC 插入的部位。⁴尽管样本量非常有限，总共只有 10 个观察到的 CRBSI，Heidenreich 等人。当他们声明插入部位是该患者人群中 CVC 相关并发症的主要危险因素时，他们从他们的数据中得出了非常深远的结论。^1个

由于关于 JV 插入 CVC 的 CRBSI 风险似乎有不同的说法，我们旨在确认 Heidenreich 等人提供的结果。^1个在不同的数据集中。我们使用的数据来自 SECRECY 注册表（评估血液学和肿瘤学中中心静脉导管相关感染的研究；德国临床试验注册库 [DRKS]，编号 DRKS00006551），这是一个正在进行的临床 CRBSI 注册表，从 2013 年 3 月开始在血液学和肿瘤科进行，目前活跃于六个德国站点。为常规临床使用插入短期、非隧道式 JV、SV 或股静脉 CVC 的非选定连续患者的 CRBSI 监测数据记录在案。所有 CVC 均根据当地标准操作程序插入。CRBSI 根据 2012 年德国血液学和肿瘤内科学会 (DGHO) 传染病工作组 (AGIHO) 的定义进行分类。^4个该登记处获得了中央伦理委员会（马格德堡大学医院，批准号 84/14）以及各自地方伦理委员会的批准。截至 2022 年 2 月输入的关于 JV-CVC 和 SV-CVC ≥ 1 天的原位血液恶性肿瘤患者（包括接受自体或同种异体 HSCT 的患者）的登记数据用于该分析。仅考虑分类为明确CRBSI ^{4 的}CRBSI 病例，以便应用与 Heidenreich 等人使用的相同诊断标准。¹为了直接比较这两项研究，我们分别分析了相同的临床和 CVC 参数和终点。因此，我们能够直接将我们的数据与 Heidenreich 研究进行比较。^1个由于 CRBSI 时间是 Heidenreich 研究的主要终点之一，我们比较了该终点并另外比较了 CRBSI 率、CRBSI 发生率和 CVC 插入后第 15 天的累积 CRBSI 发生率 (CRBSI15)。使用 OpenEpi 3.01 版（美国佐治亚州亚特兰大，https://www.openepi.com）和 IBM® SPSS® Statistics 28 版（美国纽约州阿蒙克）进行统计分析。使用的统计测试显示在表中。双侧p值 <.05 被认为具有统计学意义。

根据纳入标准，我们在登记处确定了 3795 例，包括 220 例 SV-CVC (5.8%)（表 1）。与 Heidenreich 研究¹相比，我们的队列包括较少的急性髓性白血病病例。此外，在我们的队列中，410 个 CVC (10.8%) 有抗菌涂层，而 Heidenreich 研究中没有。¹另一方面，只有 1422 个 CVC (37.4%) 有葡萄糖酸氯己定 (CHG) 涂层敷料，而在 Heidenreich 研究中，所有 CVC 都覆盖有 CHG 涂层敷料。^1个

与 SV-CVC 组相比，我们的队列包括 JV-CVC 组中显着更多的男性患者 ( p  = .016)。在 JV-CVC 组中，CVC 时间明显更短 ( p  < .001)，并且抗菌涂层 CVC 明显少于 SV-CVC 组 ( p  <.001)。

有趣的是，我们发现 JV-CVC 队列中 CRBSI 发生率较低的趋势（5.2% 对 8.2%；p = .064），但在 CRBSI 发生率 ( p  = .211) 或时间方面 没有统计学显着差异CRBSI ( p  = .338) 以及 CRBSI15 ( p  = .480)（表 1）。

为了减少协变量的偏差，我们还进行了配对分析，1:1 匹配性别、诊断、CVC 类型和 CHG 涂层敷料。通过这个，可以比较215个案例。我们发现 JV-CVC 与 SV-CVC 的 CRBSI 率（11/215 [5.1%] 对比 18/215 [8.4%]；p  = .248）和 CRBSI 发生率（3.2 CRBSI /1000 CVC 天数对比 4.5 CRBSI/1000 CVC 天数）；p  = .369），至 CRBSI 的时间（HR 1.326 [95% CI 0.62–2.82]；p  = .464）或 CRBSI15（4.30% 对 5.90%；p  = .541）（表 2）。

综上所述，我们的结果与 Heidenreich 的研究形成对比。¹我们发现 SV-CVC 和 JV-CVC 之间的 CRBSI 率没有显着差异，甚至观察到 SV-CVC 中 CRBSI 率更高的趋势。这些相反的结果可能分别归因于 Heidenreich 研究中 CVC 病例或 CRBSI 事件的数量较少。¹此外，回顾性和观察性 Heidenreich 研究¹的结果与 2016 年发表的一项关于 CRBSI 的多中心随机对照试验 (RCT) 的结果形成对比，该试验涉及中性粒细胞减少患者的 CHG 涂层 CVC 敷料，即 COAT 试验。⁵在此试验中，CRBSI 在 SV-CVC 患者中的发生率明显高于 JV-CVC（29/356 [8.1%] 对 8/257 [3.1%]；p  = .010）。重要的是，COAT 试验中的 CRBSI 发生率仅为确诊CRBSI病例^{4, 5}，其符合与我们和 Heidenreich 等人的研究相同的诊断标准。^{1, 4}因此，这三项讨论的研究结果具有可比性。

然而，必须严格指出，在上述 COAT 试验中，大多数 SV-CVC 仅从一个中心招募（319/366；87.2%）。⁵中心效应似乎很可能会影响本试验的结果。在我们的研究中，一半以上记录的 SV-CVC 来自一个中心（表 1 和 2），因此不能排除中心效应。

在 Heidenreich 研究中，绝大多数患者患有急性白血病 (120/153 [78.4%])。¹这一比例在我们的研究（1900/3795 [50.1%]；表 1）和 COAT 试验（279/613 [45.6%]）中较低^。5这可能表明这些研究中中性粒细胞减少持续时间的差异，这可能会限制数据的可比性。不幸的是，在我们的登记处，没有记录中性粒细胞减少症的持续时间。然而，我们可以提供关于中性粒细胞减少持续时间的间接数据：在整个队列中，335 例中性粒细胞减少在 CVC 插入和 CVC 移除这两个时间点均存在。假设在这些病例中初始中性粒细胞减少是持续性的，中性粒细胞减少的中位持续时间为 13 天（四分位间距 8-22），这在急性白血病患者的中位值范围内。根据这一假设，在整个队列 ( p  = .766) 以及荟萃分析中，JV-CVC 和 SV-CVC 每 1000 个中性粒细胞减少 CVC 天的 CRBSI 发生率在统计学上没有差异 ( p = .403)（表 2）。

由于这项非对照、真实世界研究的性质，CVC 插入和 CVC 管理以及 CRBSI 管理的程序在各中心并不统一。也许这可能导致中心之间的 CRBSI 频率不同。在整个队列中，所有六个中心的 CRBSI 率为 5.4%，范围为 1.5%–8.2%；CRBSI 发生率为 3.3/1000 CVC 天，范围为 0.9-6.8。此外，登记处记录的 CVC 数量因地点而异（159-1570，平均 633）。

来自 RCT 的 CRBSI 流行病学数据可以可靠地转移到现实环境中。⁶然而，据我们所知，目前还没有已发表的 RCT 比较 JV-CVC 和 SV-CVC 在血液系统恶性肿瘤患者中的 CRBSI 率。因此，这里提到的任何证据都必须来自 RCT 的次要终点或来自真实世界的多中心研究或注册，具有足够的样本量。

根据我们在此展示的大型多中心、基于登记（匹配对）的分析结果和迄今为止可用的研究，我们认为没有足够的证据表明颈静脉插入 CVC 在 CRBSI 方面的劣势。

总之，不能将颈静脉 CVC 插入部位视为血液系统恶性肿瘤患者 CRBSI 的主要危险因素。进一步的调查，最好包括随机对照试验的形式，似乎是有必要的。