To The Editor:

Acute panmyelosis with myelofibrosis (APMF) is a rare subtype of acute myeloid leukemia (AML) characterized by acute panmyeloid proliferation with increased blasts, cytopenias, bone marrow fibrosis, and absence of splenomegaly. APMF is estimated to account for <1% of AML.¹ There is controversy regarding how to differentiate APMF from other myeloid malignancies such as AML with fibrosis, acute megakaryoblastic leukemia, or myelodysplastic syndrome with fibrosis.²

There is a paucity of studies describing the clinical features and outcomes of APMF. The prognosis of APMF is very poor, with a reported median survival of 1–9 months.^{1, 3} No consensus in treatment exists. No study has yet utilized the National Cancer Database (NCDB) or Surveillance, Epidemiology, and End Results (SEER) database to report risk factors, treatments received, and additional clinical features. Our study utilizes SEER and the NCDB to better describe the outcomes and survival trends of patients diagnosed with APMF from 2004 to 2015.^{4, 5}

We queried the United States SEER database and NCDB using the ICD-O-3 code 9931/3. The NCDB is a joint project of the Commission on Cancer of American College of Surgeons and the American Cancer Society that is a nationwide oncology outcomes database for >1500 cancer programs in the US and Puerto Rico, capturing 70% of all newly diagnosed cases of cancer in the US.⁴ SEER is a program of the National Cancer Institute that collects and publishes cancer incidence and survival data covering about 28% of the US population.⁵

Due to the challenges of diagnosing APMF, we only included patients in our cohort if the diagnostic confirmation included those with recorded positive histology or positive histology plus positive immunophenotyping and/or positive genetic studies. The SEER 17 registries (2004–2015) were used to find data on incidence, overall survival (OS), and relative survival (RS); the Alaska Native Tumor registry was excluded from this analysis. Incidence was age-adjusted to the U.S. 2000 standard population. Incidence ratios (IRs) were calculated with SEER*Stat software (version 8.3.6; NCI, Bethesda, MD). Descriptive statistics such as median, interquartile range (IQR), and OS were calculated by analyzing SEER data in BlueSky (Version 7.10). Relative survival (RS) was defined as the ratio of observed survivors in the group of APMF patients to the expected survivors in a comparable set of individuals without APMF. This was adjusted for the general survival of the US population by race, sex, age, and year of diagnosis. We utilized SEER*Stat to calculate standardized mortality ratios (SMRs) for all-cause mortality. SMR is the number of observed deaths in the study population divided by the number of expected deaths. Thus, SMR depicts the risk of death in APMF patients compared with the general US population.

The NCDB data included those diagnosed between 2004 and 2015. Comorbid disease burden was calculated using the Deyo adaptation (1992) of Charlson's comorbidity index, which considers as many reported ICD-9-CM or ICD-10 secondary diagnosis codes for the patient of interest and is reported as the Charlson Deyo Score (CDS). Overall survival (OS) was analyzed using the Kaplan–Meier method. Hazard ratios (HR) with confidence intervals (CI) were calculated using a Cox proportional hazards model. Variables significant in univariate analysis were included in a multivariate analysis. Statistical analyses for NCDB were performed using SAS version 9.4.

We identified 260 APMF patients using the SEER database. The median age at diagnosis was 67 years (range 20–93); 142 (55%) were male. 201 (77%) were non-Hispanic White (NHW), 26 (10%) were non-Hispanic Black (NHB), 20 (8%) were Hispanic, and 13 (5%) were non-Hispanic Asian Pacific Islander (NHAPI). The overall IR of APMF was 0.3 cases/million individuals and did not change significantly between the years 2004–2015. The IRs according to gender and racial groups were male 0.4, female 0.3, NHW 0.4, NHB 0.4, Hispanic 0.2, and NHAPI 0.2. With a median follow up of 6.9 years (95% CI 6.1–7.8), the median OS was 2.3 years (95% CI 1.7–2.8; Figure S1). In the general US population, the expected survival at 1, 3, and 5 years was 97%, 93%, and 89%. In contrast, RS for patients with APMF was 71%, 45%, and 34%, respectively (Figure S1). The standardized mortality ratio (SMR) for all-cause mortality was 10.15 and statistically significant (95% CI: 8.74–11.72).

We identified 530 APMF patients using the NCDB. The median age at diagnosis was 67 years (range 22–90) and 311 (59%) were male. With a median follow up of 5 years (95% CI 3.0–7.6), the median OS was 2.3 years (95% CI: 0.8–6.5, Figure 1A). OS was 69%, 31%, and 18% at 1-, 5-, and 10- years, respectively. Four hundred three patients (76%) had a CDS of 0, 83 had a CDS of 1 (16%), and 44 had a CDS of 2+ (8%). Patients were grouped by year of diagnosis, specifically 2004–2007, 2008–2011, and 2012–2015. Patients diagnosed in 2012–2015 had a significantly improved OS compared to those diagnosed in 2004–2007 (HR 0.65, 95% CI 0.49–0.85; p = .002).

Two hundred seventy-one patients (53%) received chemotherapy as first-line therapy. The median time from diagnosis to chemotherapy was 25 days (range 0–532 days). The OS for those that received chemotherapy was 70% at 1- and 30% at 5-years versus 70% and 32% at 1- and 5- years for those who did not (p = .99). Fifty-two patients (10%) underwent hematopoietic cell transplantation (HCT) as first-line therapy and the OS of those patients was 90% at 1 year and 45% at 5 years versus 67% and 29% at 1- and 5- years for those who did not (HR: 1.7 [95% CI: 1.2–2.6], p = .006). Among the 51 patients who underwent allogenic transplantation, 46 (90%) had a CDS of 0, 3 (6%) had a CDS of 1, and 2 (4%) had a CDS of 2+.

In univariable analysis, factors predicting inferior OS were age ≥65 years at diagnosis (HR 1.8, 95% CI 1.5–2.3; p < .001), male sex (HR 1.5, 95% CI 1.2–1.8; p < .001), CDS ≥1 (HR: 1.5 [95% CI: 1.2–2.0], p < .001), government insurance (HR 1.8, 95% CI 1.4–2.3; p < .001), diagnosis at a non-academic facility (HR 1.6, 95% CI 1.2–2.0, p < .001), and not receiving HCT as first-line therapy (HR 1.7, 95% CI 1.2–2.6; p = .006). In multivariate analysis, factors predicting inferior OS were age ≥65 years at diagnosis (HR 1.4, 95% CI 1.1–2.0; p = .02), male sex (HR 1.7, 95% CI 1.3–2.1; p < .001), CDS ≥1 (HR 1.4, 95% CI 1.1–1.8; p = .006), and diagnosis at a non-academic facility (HR 1.5, 95% CI 1.2–1.9; p < .001). Table S1 shows the univariate and multivariate analysis of factors predicting inferior OS. Figure 1B–E show the survival estimates based on age, insurance type, year of diagnosis, and facility type.

This is the largest retrospective study examining the incidence, OS trends, and factors predicting inferior OS of APMF at the population level. Prior series have reported widely variable survival ranging from 2 to 9 months.⁶ A series by Thiele et al., had a similar median age at diagnosis, but there was significant discrepancy between the median OS (2.3 years in our series vs. 9 months in Thiele et al.).³ The exact reasons for the discrepancy are unclear. One possibility is the misdiagnoses in our cohort given the challenge in diagnosing APMF. In addition, the median comorbidity in our cohort was low, whereas Thiele et al. did not report comorbidity. Finally, there was a significant male preponderance (74% male) in the series by Theiele et al., whereas our population-based cohort was more balanced (55% and 59% for SEER and NCDB, respectively). In a series of 40 younger patients who underwent HCT, median post-HCT survival was <1 year and male sex was associated with a significantly inferior survival.⁷

A potentially significant finding emerging from our study is that despite a low comorbidity burden, only 10% patients underwent HCT early in the course. Those undergoing HCT early had a significant survival advantage compared to those who did not (45% vs. 29% at 5 years). Therefore, HCT should be considered early in APMF.

Strengths of our study include the larger sample sizes found in the SEER and NCDB databases compared to the case series in the available published literature. We found significant similarities between the SEER and NCDB databases in terms of clinical characteristics and survival, reinforcing the consistency of our findings. Limitations include the possibility of diagnostic error due to our inability to independently review the histopathology to confirm the diagnosis. SEER and NCDB also do not provide specifics on the agents used in treatment. In addition, these databases do not include data on those that go on to receive chemotherapy or HCT following their first treatment regimen.

Our study confirms that APMF remains an exceedingly rare disease with a poor prognosis, with a 10-fold increased risk of death for those with APMF. The incidence of APMF has not changed between 2004 and 2015, but OS for 2012–2015 was improved compared to 2004–2007. While early utilization of chemotherapy did not improve outcomes, early utilization of HCT was associated with improved OS. Given the lack of standard of care and poor prognosis associated with APMF, further studies are needed to better understand potential therapeutic options.

致编辑：

急性全髓细胞增多症伴骨髓纤维化 (APMF) 是一种罕见的急性髓细胞白血病 (AML) 亚型，其特征是急性全髓细胞增殖伴原始细胞增多、血细胞减少、骨髓纤维化和无脾肿大。据估计，APMF 占 AML 的比例不到 1%。¹关于如何区分 APMF 与其他髓系恶性肿瘤（如伴有纤维化的 AML、急性巨核细胞白血病或伴有纤维化的骨髓增生异常综合征）存在争议。²

很少有研究描述 APMF 的临床特征和结果。APMF 的预后非常差，据报道中位生存期为 1-9 个月。^{1, 3}在治疗方面尚无共识。尚未有研究利用国家癌症数据库 (NCDB) 或监测、流行病学和最终结果 (SEER) 数据库来报告风险因素、接受的治疗和其他临床特征。我们的研究利用 SEER 和 NCDB 来更好地描述 2004 年至 2015 年诊断为 APMF 的患者的结果和生存趋势^。4、5

我们使用 ICD-O-3 代码 9931/3 查询了美国 SEER 数据库和 NCDB。NCDB 是美国外科学院癌症委员会和美国癌症协会的一个联合项目，它是一个全国性的肿瘤学结果数据库，用于美国和波多黎各超过 1500 个癌症项目，涵盖所有新诊断的癌症病例的 70%在美国。⁴ SEER 是美国国家癌症研究所的一个项目，它收集和发布覆盖美国约 28% 人口的癌症发病率和生存数据。⁵

由于诊断 APMF 的挑战，如果诊断确认包括记录为阳性组织学或阳性组织学加上阳性免疫表型和/或阳性遗传研究的患者，我们仅将患者纳入我们的队列。SEER 17 登记处 (2004-2015) 用于查找发病率、总生存期 (OS) 和相对生存期 (RS) 的数据；阿拉斯加原生肿瘤登记处被排除在该分析之外。发病率根据美国 2000 年标准人口进行年龄调整。使用 SEER*Stat 软件（版本 8.3.6；NCI，Bethesda，MD）计算发病率 (IR)。通过分析 BlueSky（版本 7.10）中的 SEER 数据计算描述性统计数据，例如中位数、四分位距 (IQR) 和 OS。相对存活率 (RS) 定义为在 APMF 患者组中观察到的幸存者与在没有 APMF 的一组可比个体中的预期幸存者的比率。根据种族、性别、年龄和诊断年份对美国人口的总体生存率进行了调整。我们利用 SEER*Stat 计算全因死亡率的标准化死亡率 (SMR)。SMR 是研究人群中观察到的死亡人数除以预期死亡人数。因此，SMR 描述了 APMF 患者与美国一般人群相比的死亡风险。SMR 是研究人群中观察到的死亡人数除以预期死亡人数。因此，SMR 描述了 APMF 患者与美国一般人群相比的死亡风险。SMR 是研究人群中观察到的死亡人数除以预期死亡人数。因此，SMR 描述了 APMF 患者与美国一般人群相比的死亡风险。

NCDB 数据包括 2004 年至 2015 年期间诊断出的患者。使用 Charlson 合并症指数的 Deyo 改编（1992 年）计算合并症负担，该指数考虑了许多报告的 ICD-9-CM 或 ICD-10 二级诊断代码为感兴趣的患者并报告为 Charlson Deyo 分数 (CDS)。使用 Kaplan-Meier 方法分析总生存期 (OS)。使用 Cox 比例风险模型计算具有置信区间 (CI) 的风险比 (HR)。在单变量分析中显着的变量被包括在多变量分析中。使用 SAS 9.4 版对 NCDB 进行统计分析。

我们使用 SEER 数据库确定了 260 名 APMF 患者。诊断时的中位年龄为 67 岁（范围 20-93）；142 (55%) 人是男性。201 人（77%）为非西班牙裔白人（NHW），26 人（10%）为非西班牙裔黑人（NHB），20 人（8%）为西班牙裔，13 人（5%）为非西班牙裔亚太岛民（ NHAPI)。APMF 的总体 IR 为 0.3 例/百万人，在 2004 年至 2015 年之间没有显着变化。根据性别和种族群体的 IR 为男性 0.4、女性 0.3、NHW 0.4、NHB 0.4、西班牙裔 0.2 和 NHAPI 0.2。中位随访时间为 6.9 年（95% CI 6.1-7.8），中位 OS 为 2.3 年（95% CI 1.7-2.8；图 S1）。在美国一般人群中，1、3 和 5 年的预期生存率为 97%、93% 和 89%。相比之下，APMF 患者的 RS 分别为 71%、45% 和 34%（图 S1）。

我们使用 NCDB 确定了 530 名 APMF 患者。诊断时的中位年龄为 67 岁（范围 22-90），311 人（59%）为男性。中位随访 5 年（95% CI 3.0–7.6），中位 OS 为 2.3 年（95% CI：0.8–6.5，图 1A）。1 年、5 年和 10 年的 OS 分别为 69%、31% 和 18%。403 名患者 (76%) 的 CDS 为 0，83 名的 CDS 为 1 (16%)，44 名的 CDS 为 2+ (8%)。患者按诊断年份分组，特别是 2004-2007 年、2008-2011 年和 2012-2015 年。与 2004-2007 年诊断的患者相比，2012-2015 年诊断的患者 OS 显着改善（HR 0.65，95% CI 0.49-0.85；p  = .002）。

271 名患者 (53%) 接受了化疗作为一线治疗。从诊断到化疗的中位时间为 25 天（范围 0-532 天）。接受化疗的患者在 1 年和 5 年的 OS 分别为 70% 和 30%，而未接受化疗的患者在 1 年和 5 年的 OS 分别为 70% 和 32% ( p  = .99)。52 名患者 (10%) 接受了造血细胞移植 (HCT) 作为一线治疗，这些患者的 OS 在 1 年和 5 年分别为 90% 和 45%，而在 1 年和 5 年分别为 67% 和 29%年（HR：1.7 [95% CI：1.2–2.6]，p  = .006）。在接受同种异体移植的 51 名患者中，46 名（90%）的 CDS 为 0，3 名（6%）的 CDS 为 1，2 名（4%）的 CDS 为 2+。

在单变量分析中，预测较差 OS 的因素是诊断时年龄 ≥ 65 岁（HR 1.8，95% CI 1.5-2.3；p  < .001），男性（HR 1.5，95% CI 1.2-1.8；p  < .001） , CDS ≥1 (HR: 1.5 [95% CI: 1.2–2.0], p  < .001), 政府保险 (HR 1.8, 95% CI 1.4–2.3; p  < .001), 在非学术机构诊断(HR 1.6, 95% CI 1.2–2.0, p  < .001)，并且未接受 HCT 作为一线治疗 (HR 1.7, 95% CI 1.2–2.6; p  = .006)。在多变量分析中，预测较差 OS 的因素是诊断时年龄≥65 岁（HR 1.4，95% CI 1.1-2.0；p  = .02），男性（HR 1.7，95% CI 1.3-2.1；p  < .001） , CDS ≥1 (HR 1.4, 95% CI 1.1–1.8;p  = .006)，在非学术机构进行诊断 (HR 1.5, 95% CI 1.2–1.9; p  < .001)。表 S1 显示了预测较差 OS 的因素的单变量和多变量分析。图 1B-E 显示了基于年龄、保险类型、诊断年份和设施类型的生存估计。

这是最大的回顾性研究，在人群水平上检查 APMF 的发生率、OS 趋势和预测较差 OS 的因素。先前的系列报道了从 2 到 9 个月的广泛变化的生存期。⁶ Thiele 等人的系列研究在诊断时具有相似的中位年龄，但中位 OS 之间存在显着差异（我们的系列中为 2.3 岁，而 Thiele 等人中为 9 个月）。³差异的确切原因尚不清楚。鉴于诊断 APMF 的挑战，一种可能性是我们队列中的误诊。此外，我们队列中的中位合并症较低，而 Thiele 等人。没有报告合并症。最后，在 Theiele 等人的系列中，男性占显着优势（74% 男性），而我们基于人群的队列更加平衡（SEER 和 NCDB 分别为 55% 和 59%）。在一系列接受 HCT 的 40 名年轻患者中，HCT 后中位生存期<1 年，男性与显着较差的生存率相关。⁷

我们的研究中出现的一个潜在的重要发现是，尽管合并症负担较低，但只有 10% 的患者在病程早期接受了 HCT。与未接受 HCT 的患者相比，早期接受 HCT 的患者具有显着的生存优势（5 年时分别为 45% 和 29%）。因此，在 APMF 早期应考虑 HCT。

与现有已发表文献中的案例系列相比，我们研究的优势包括在 SEER 和 NCDB 数据库中发现的更大样本量。我们发现 SEER 和 NCDB 数据库在临床特征和生存率方面存在显着相似性，从而加强了我们研究结果的一致性。限制包括由于我们无法独立审查组织病理学以确认诊断而导致诊断错误的可能性。SEER 和 NCDB 也没有提供治疗中使用的药物的细节。此外，这些数据库不包括那些在第一次治疗方案后继续接受化疗或 HCT 的数据。

我们的研究证实，APMF 仍然是一种极其罕见的疾病，预后不良，APMF 患者的死亡风险增加 10 倍。APMF 的发病率在 2004 年至 2015 年之间没有变化，但与 2004 年至 2007 年相比，2012 年至 2015 年的 OS 有所改善。虽然早期使用化疗并没有改善结果，但早期使用 HCT 与改善 OS 相关。鉴于缺乏与 APMF 相关的护理标准和不良预后，需要进一步研究以更好地了解潜在的治疗选择。