Erythrocytosis is a common reason for referral to hematology, primarily to exclude polycythemia vera (PV), which has a high morbidity and mortality if untreated.¹ Recent changes to the World Health Organization (WHO) definition of PV reduced the hemoglobin thresholds required for diagnosis, leading to overlap with the normal range and resulting in more frequent testing.² In addition to a focused history for possible secondary causes, distinguishing PV from secondary erythrocytosis often requires laboratory investigation, including serum erythropoietin (EPO) measurement and/or molecular testing for JAK2 mutations. Although JAK2 mutations are found in other myeloproliferative neoplasms, in patients with erythrocytosis, presence of JAK2 mutations is highly sensitive and specific for PV.³ Nonetheless, molecular testing can be costly and relies on access to a specialized laboratory.

Various approaches to the investigation of erythrocytosis are found in the literature: some start with EPO measurement,⁴ while others advocate concurrent EPO and JAK2 testing.⁵ In a recent issue of this journal, Tefferi and Barbui³ recommend upfront JAK2 mutation screening in patients with suspected PV. The sequential approach starting with EPO measurement is premised on a normal or high EPO level having a high negative predictive value (NPV) to rule out PV. In contrast, the justification for concurrent or upfront JAK2 testing is one of clinical expediency in patients with a high pretest probability for PV.

Most evidence for EPO's utility in the investigation of erythrocytosis predates the advent of molecular testing for PV,⁶ and the added value of EPO measurement beyond JAK2 testing has been brought into question by more recent studies.⁷ This descriptive study revisited the utility of EPO measurement in the era of JAK2 testing by examining serum EPO distribution in a large real-world cohort of patients with erythrocytosis referred to our center, an academic healthcare organization, which serves a population of approximately 2 million in Ontario, Canada.

We reviewed all patients aged 18 years or older investigated for erythrocytosis (>160 g/L for women or >165 g/L for men) with JAK2 mutation testing between January 1, 2015, and May 12, 2021, extracting data on JAK2 mutation analysis, serum EPO levels, final diagnosis, and risk factors for secondary erythrocytosis. All patients with available EPO levels were included. JAK2 testing was performed by either quantitative polymerase chain reaction (qPCR) using the Roche 480 LightCycler (La Roche AG, Switzerland), single nucleotide polymorphism (SNP) allelotyping using the Agena MassARRAY system (Agena Biosciences, USA), or next generation sequencing (NGS) panel using the Oncomine Myeloid Research Assay (ThermoFisher Scientific, USA). qPCR and SNP allelotyping assays tested for JAK2V617F mutations; the NGS panel screened for any JAK2 mutations in exons 12–15, in addition to sequence variants in 40 other genes (including SH2B3) and 29 fusion driver genes associated with myeloid malignancies. Serum EPO levels were measured by chemiluminescent immunoassay (Unicel DXi 800; Beckman Coulter, USA) with a normal range of 2.6–18.5 mIU/mL. We performed receiver operating characteristic (ROC) analysis to evaluate the diagnostic accuracies of different EPO levels in predicting JAK2-positive PV. Given the high morbidity and mortality of untreated PV, we used a high NPV (>99%) to determine the threshold EPO level needed to exclude a diagnosis of PV.

Over the 5-year study period, a total of 883 patients referred for erythrocytosis underwent molecular testing. Of these patients, 696 (78.8%) had EPO levels measured; most were ordered simultaneously with JAK2 testing. Patient characteristics are shown in Table S1, stratified by JAK2 mutation status. The final diagnosis was PV in all 72 patients with JAK2 mutations (10.3%), and the remaining 624 patients (89.7%) were diagnosed with secondary causes of erythrocytosis, most commonly smoking (42.5%), obstructive sleep apnea (32.7%), chronic obstructive pulmonary disease (14.6%), and testosterone use (10.3%). Distribution of risk factors for secondary erythrocytosis is shown in Table S2. The median EPO levels for JAK2-positive and JAK2-negative patients were 1.8 mIU/mL (range 0–26.4 mIU/mL) and 9.2 mIU/mL (range 1.3–762 mIU/mL), respectively (p < .001). There was a considerable overlap in EPO levels between JAK2-positive and -negative patients, with 80.2% (n = 559) of our cohort falling within the normal range (Figure 1A). Of the JAK2-positive patients, 25 (34.7%) had EPO levels at or above the lower limit of normal (≥2.6 mIU/mL). One JAK2-positive patient had an EPO level above 18.5 mIU/mL, which may have been explained by concomitant heavy smoking. EPO testing was performed at the time of initial consultation in all patients; no patients were identified as being on active treatment for erythrocytosis, either in the form of phlebotomy or cytoreductive therapy, at the time of measurement. Of the JAK2-positive patients, 34.7% (n = 25) had a prior history of thrombosis at the time of initial presentation, with the majority (84%) being arterial events. Of the JAK2-negative patients with subnormal EPO levels (n = 15), potential confounding factors included acute kidney injury (n = 4), chronic kidney disease (n = 2), and liver disease (n = 5); none was identified as having EPO receptor mutations.

ROC analysis established 4.75 mIU/mL as the optimal cut point; EPO levels below this threshold had a sensitivity of 90.3% and specificity of 88.0% for predicting a diagnosis of PV (area under the curve [AUC] = 0.94, Figure 1B). For use as an initial screening test to exclude JAK2-positive PV, a higher sensitivity (95.8%) could be achieved by increasing the EPO threshold to 7.15 mIU/mL. An EPO level > 27 mIU/mL was required to achieve a sensitivity of 100% and a NPV of 100% to effectively rule out JAK2-positive PV. Applied to our cohort, only 4.3% (n = 30) were at or above this threshold, meaning that an EPO level could have been used to rule out JAK2-positive PV.

Given the significant number of PV patients with normal EPO levels (≥2.6 mIU/mL), we reviewed secondary causes that might potentially explain “inappropriately normal” EPO levels in these patients. We found that risk factors for secondary erythrocytosis were overrepresented in PV patients with normal/high vs. low EPO levels, including smoking (40.0% vs. 14.9%) and obstructive sleep apnea (16.0% vs. 4.3%), but only smoking reached statistical significance (p = .022).

In summary, these data on serum EPO distribution in a large real-world cohort of patients referred for suspected PV suggest that EPO measurement may have limited added value when used concurrently with JAK2 testing to rule out PV. We demonstrated that normal EPO levels are present in over a third of patients with JAK2-positive PV and therefore do not distinguish between PV and secondary erythrocytosis. A higher EPO threshold (>7.15 mIU/mL) showed improved sensitivity (95.8%), suggesting that EPO testing may still play an initial role if JAK2 testing is not available. Nonetheless, in our cohort, a very high EPO threshold (>27 mIU/mL) was required to rule out PV with sufficient confidence to avoid JAK2 testing, a level that would have excluded PV in only a small proportion of patients (4.3%).

These data question the value of simultaneous EPO and JAK2 testing in the diagnosis of PV. EPO may retain diagnostic utility in sequential testing; however, our findings challenge the premise implicit in many sequential algorithms that normal or high EPO levels are sufficient to exclude PV. Implementation of a diagnostic algorithm and selection of EPO thresholds must be based on local disease prevalence and test performance rather than normal reference ranges for EPO, which vary depending on population characteristics and laboratory methods, and are difficult to extrapolate between settings. The diagnostic utility of EPO is further confounded by preanalytical variables, which include diurnal variation, as well as risk factors for hypoxia such as smoking, obstructive sleep apnea, and place of residence. In our cohort, smoking was overrepresented in PV patients with normal or high EPO levels, in keeping with previous findings.⁷

Limitations include our study's retrospective design, making us unable to control for variables such as variation in timing of EPO measurement and method of JAK2 testing. Our study was conducted at a tertiary referral center with ready access to molecular diagnostics: it does not address the utility of EPO measurement in low resource settings or at centers with limited access to JAK2 testing. In such settings, EPO testing may be helpful provided that the user is aware of the performance characteristics of the assay in their patient population. Although our cohort is more representative of all patients referred to specialty hematology clinics for erythrocytosis compared to previous studies examining the utility of EPO testing,⁶ the prevalence of PV in our study remains higher than in population-based cohorts; in lower prevalence settings, initial EPO testing may have greater utility in excluding PV. Additionally, EPO measurement plays a diagnostic role in the rare subgroup of PV patients without JAK2V617F or exon 12 mutations.³ Given our objective to address the practical utility of EPO measurement in patients referred for erythrocytosis, we used elevated hemoglobin levels as an inclusion criteria for our study; we acknowledge this is an imperfect surrogate for red cell mass, which is no longer routinely measured in practice.

Despite these caveats, our study supplements existing literature on the limited utility of EPO measurement with data from a large cohort of patients investigated for erythrocytosis with JAK2 testing. Our findings are consistent with clinical experience that EPO levels often have limited value in discriminating between primary and secondary causes of erythrocytosis given that most results are in the normal range at values that can neither rule out nor rule in PV. Even for the limited number of patients with elevated EPO levels, these levels rarely reach a threshold sufficient to exclude a diagnosis of PV with a high NPV.

While EPO measurement is generally less costly than molecular diagnostics, the common practice of simultaneously ordering EPO levels and JAK2 mutation analysis negates any potential cost savings. In patients with a high pretest probability of PV, however, the sequential approach is likely not a cost-effective strategy for investigating erythrocytosis and may only add to costs in the form of unnecessary testing, repeat clinic visits and diagnostic delays, justifying the approach of upfront JAK2 mutation screening recently proposed by Tefferi and Barbui.³ With declining costs of molecular diagnostics, an updated cost–benefit analysis is needed to better define the role of EPO testing in the work up of PV. Given rising referrals for suspected PV in the wake of the 2016 WHO revision, further research should also examine whether alternate clinical or laboratory data aside from EPO levels, such as lactate dehydrogenase levels or complete blood count parameters,⁸ might be used to more effectively triage JAK2 testing in patients with erythrocytosis.

红细胞增多症是转诊至血液科的常见原因，主要是为了排除真性红细胞增多症 (PV)，如果未经治疗，其发病率和死亡率都很高。¹最近世界卫生组织 (WHO) 对 PV 定义的更改降低了诊断所需的血红蛋白阈值，导致与正常范围重叠并导致更频繁的检测。²除了可能的继发性原因的重点病史外，区分 PV 和继发性红细胞增多症通常需要实验室检查，包括血清促红细胞生成素 (EPO) 测量和/或JAK2突变的分子检测。尽管在其他骨髓增生性肿瘤中发现了JAK2突变，但在红细胞增多症患者中，存在JAK2突变对 PV 具有高度敏感性和特异性。³尽管如此，分子检测可能成本高昂，并且依赖于专业实验室。

文献中发现了各种红细胞增多症调查方法：一些从 EPO 测量开始，⁴而另一些则提倡同时进行 EPO 和JAK2检测。⁵在本期刊的最近一期中，Tefferi 和 Barbui ³建议对疑似 PV 的患者进行前期JAK2突变筛查。从 EPO 测量开始的顺序方法以正常或高 EPO 水平为前提，具有高阴性预测值 (NPV) 以排除 PV。相比之下，同时或预先进行JAK2测试的理由是对 PV 具有高预测概率的患者的临床权宜之计之一。

EPO 在红细胞增多症研究中的效用的大多数证据早于 PV 分子检测的出现，⁶并且 EPO 检测在JAK2检测之外的附加值已被最近的研究提出质疑。⁷这项描述性研究重新审视了 EPO 测量在JAK2检测时代的效用，方法是检查我们中心的大量红细胞增多症患者的血清 EPO 分布，该中心是一家学术医疗机构，为大约 200 万人口提供服务在加拿大安大略省。

我们回顾了2015 年 1 月 1 日至 2021 年 5 月 12 日期间所有接受JAK2突变检测的红细胞增多症（女性>160 g/L 或男性>165 g/L）患者，提取了有关JAK2突变的数据分析、血清 EPO 水平、最终诊断和继发性红细胞增多症的危险因素。包括所有具有可用 EPO 水平的患者。JAK2通过使用 Roche 480 LightCycler（La Roche AG，瑞士）的定量聚合酶链反应 (qPCR)、使用 Agena MassARRAY 系统（美国 Agena Biosciences）的单核苷酸多态性 (SNP) 等位基因分析或下一代测序（NGS ) 使用 Oncomine Myeloid Research Assay (ThermoFisher Scientific, USA) 的面板。qPCR 和 SNP 等位基因分析检测JAK2V617F突变；除了40 个其他基因（包括SH2B3) 和 29 个与髓系恶性肿瘤相关的融合驱动基因。通过化学发光免疫测定法（Unicel DXi 800；Beckman Coulter，USA）测量血清 EPO 水平，正常范围为 2.6-18.5 mIU/mL。我们进行了受试者工作特征 (ROC) 分析，以评估不同 EPO 水平在预测JAK2阳性 PV 中的诊断准确性。鉴于未经治疗的 PV 的高发病率和死亡率，我们使用高 NPV (>99%) 来确定排除 PV 诊断所需的阈值 EPO 水平。

在 5 年的研究期间，共有 883 名因红细胞增多症而转诊的患者接受了分子检测。在这些患者中，696 人 (78.8%) 测量了 EPO 水平；大多数是与JAK2测试同时订购的。患者特征如表 S1 所示，按JAK2突变状态分层。所有 72 例JAK2突变患者（10.3%）最终诊断为 PV，其余 624 例患者（89.7%）被诊断为继发性红细胞增多症，最常见的是吸烟（42.5%）、阻塞性睡眠呼吸暂停（32.7%）、慢性阻塞性肺病（14.6%）和睾酮使用（10.3%）。继发性红细胞增多症的危险因素分布见表 S2。JAK2阳性和JAK2阴性患者分别为 1.8 mIU/mL（范围 0–26.4 mIU/mL）和 9.2 mIU/mL（范围 1.3–762 mIU/mL）（p  < .001）。JAK2阳性和阴性患者的 EPO 水平有相当大的重叠，我们的队列中有 80.2% ( n  = 559) 在正常范围内 (图 1A)。在JAK2阳性患者中，25 名 (34.7%) 的 EPO 水平等于或高于正常下限 (≥2.6 mIU/mL)。一个JAK2-阳性患者的 EPO 水平高于 18.5 mIU/mL，这可能是由于同时大量吸烟所致。在所有患者初次会诊时进行 EPO 检测；在测量时，没有患者被确定为正在接受红细胞增多症的积极治疗，无论是静脉切开术还是细胞减灭治疗。在JAK2阳性患者中，34.7% ( n =  25) 在初次就诊时有血栓形成史，其中大多数 (84%) 是动脉事件。在 EPO 水平低于正常值的JAK2阴性患者中（n  = 15），潜在的混杂因素包括急性肾损伤（n  = 4）、慢性肾病（n = 2), 和肝病 ( n  = 5); 没有一个被确定为具有EPO受体突变。

ROC 分析确定 4.75 mIU/mL 为最佳切点；低于该阈值的 EPO 水平对预测 PV 诊断的敏感性为 90.3%，特异性为 88.0%（曲线下面积 [AUC] = 0.94，图 1B）。作为排除JAK2阳性 PV 的初始筛查测试，通过将 EPO 阈值增加到 7.15 mIU/mL 可以获得更高的灵敏度 (95.8%)。EPO 水平 > 27 mIU/mL 需要达到 100% 的灵敏度和 100% 的 NPV 才能有效排除JAK2阳性 PV。应用于我们的队列，只有 4.3% ( n  = 30) 达到或高于这个阈值，这意味着可以使用 EPO 水平来排除JAK2阳性 PV。

鉴于大量 PV 患者的 EPO 水平正常（≥2.6 mIU/mL），我们回顾了可能解释这些患者 EPO 水平“异常正常”的次要原因。我们发现继发性红细胞增多症的危险因素在 EPO 水平正常/高或低的 PV 患者中过多，包括吸烟（40.0% 对 14.9%）和阻塞性睡眠呼吸暂停（16.0% 对 4.3%），但只有吸烟达到统计显着性 ( p  = .022)。

总而言之，在大量真实世界中疑似 PV 患者队列中血清 EPO 分布的这些数据表明，当与JAK2检测同时使用以排除 PV 时，EPO 测量可能具有有限的附加价值。我们证明了超过三分之一的JAK2阳性 PV 患者的 EPO 水平正常，因此无法区分 PV 和继发性红细胞增多症。较高的 EPO 阈值 (>7.15 mIU/mL) 显示灵敏度提高 (95.8%)，这表明如果JAK2检测不可用，EPO 检测仍可能发挥初始作用。尽管如此，在我们的队列中，需要非常高的 EPO 阈值（>27 mIU/mL）来排除 PV，并有足够的信心避免JAK2测试，这一水平仅在一小部分患者（4.3%）中排除了 PV。

这些数据质疑同时 EPO 和JAK2的价值用于诊断 PV 的测试。EPO 可能在顺序测试中保留诊断效用；然而，我们的研究结果挑战了许多顺序算法中隐含的前提，即正常或高 EPO 水平足以排除 PV。诊断算法的实施和 EPO 阈值的选择必须基于当地疾病流行率和测试性能，而不是 EPO 的正常参考范围，这取决于人群特征和实验室方法，并且难以在不同环境之间进行推断。EPO 的诊断效用被分析前变量进一步混淆，这些变量包括昼夜变化，以及缺氧的危险因素，如吸烟、阻塞性睡眠呼吸暂停和居住地。在我们的队列中，在 EPO 水平正常或高的 PV 患者中吸烟的比例过高，⁷

局限性包括我们研究的回顾性设计，使我们无法控制变量，例如 EPO 测量时间和JAK2测试方法的变化。我们的研究是在一个可以随时获得分子诊断的三级转诊中心进行的：它没有解决 EPO 测量在资源匮乏的环境中或在JAK2测试访问受限的中心的效用。在这种情况下，只要用户了解检测在其患者群体中的性能特征，EPO 检测可能会有所帮助。尽管与之前检查 EPO 检测效用的研究相比，我们的队列更能代表所有因红细胞增多症而转诊到专科血液科诊所的患者，⁶我们研究中 PV 的患病率仍然高于基于人群的队列；在患病率较低的情况下，初始 EPO 检测在排除 PV 方面可能具有更大的效用。此外，EPO 测量在没有JAK2V617F或外显子 12 突变的罕见 PV 患者亚组中发挥诊断作用。³鉴于我们的目标是解决红细胞增多症患者中 EPO 测量的实际效用，我们使用升高的血红蛋白水平作为我们研究的纳入标准；我们承认这是红细胞质量的不完美替代物，在实践中不再常规测量。

尽管有这些警告，我们的研究补充了现有的关于 EPO 测量有限效用的文献，其中数据来自通过JAK2测试调查红细胞增多症的大量患者的数据。我们的研究结果与临床经验一致，即 EPO 水平通常在区分红细胞增多症的原发性和继发性原因方面的价值有限，因为大多数结果都在正常范围内，既不能排除也不能排除 PV。即使对于少数 EPO 水平升高的患者，这些水平也很少达到足以排除具有高 NPV 的 PV 诊断的阈值。

虽然 EPO 测量通常比分子诊断成本更低，但同时订购 EPO 水平和JAK2突变分析的常见做法否定了任何潜在的成本节约。然而，对于 PV 检测前概率较高的患者，序贯方法可能不是调查红细胞增多症的一种经济有效的策略，并且可能只会以不必要的检测、重复就诊和诊断延误的形式增加成本，证明采用序贯方法是合理的。Tefferi 和 Barbui 最近提出的前期JAK2突变筛选。³随着分子诊断成本的下降，需要更新的成本效益分析来更好地定义 EPO 测试在 PV 工作中的作用。鉴于在 2016 年 WHO 修订后疑似 PV 的转诊率上升，进一步的研究还应检查除 EPO 水平之外的其他临床或实验室数据，如乳酸脱氢酶水平或全血细胞计数参数⁸是否可用于更有效地进行分流在红细胞增多症患者中进行JAK2检测。