Assessing the Thrombotic Risk of Patients with Essential Thrombocythemia in the Genomic Era

Abstract

The molecular characterization of myeloproliferative neoplasms, including essential thrombocythemia (ET), has enabled deeper understanding of their pathogenesis. A driver lesion, namely, Janus kinase (JAK)2V617F, calreticulin (CALR) or myeloproliferative leukemia (MPL) gene mutation can be identified in the vast majority of patients. Each of these mutations is associated with distinct clinical features and may modulate the patients’ clinical course, risk of complications, including vascular events, and survival. JAK2V617F appears to be a risk-modifying mutation and has been shown to increase the likelihood of thrombotic events in patients with ET across studies. As such, it has been included in prognostic models and its presence may influence treatment decisions. The association of CALR and MPL mutations with the incidence of vascular events has been less clear. Even more limited information is available on the contribution of additional non-driver lesions to the thrombotic risk. In this review we discuss the available evidence on the role of recurrent mutations in the risk of thrombotic complications in patients with ET and how these mutations weigh into modern prognostic scores.

Introduction

Essential thrombocythemia (ET) is a Philadelphia-negative myeloproliferative neoplasm (MPN) characterized by clonal proliferation of the megakaryocytic lineage within the bone marrow and elevated platelet count in peripheral blood.¹ It is a rare disease, with an incidence rate of 0.2-2.5/100,000 people per year in western countries,^{2, 3} and it is associated with long overall survival. However, life expectancy of ET patients is shorter than in the general population,⁴ largely due to the occurrence of thrombotic events. Consequently, the treatment of this disorder is aimed at reducing the risk of vascular complications.⁵

The overall risk of thrombosis associated with ET is 1-3% per patient-year.⁶ The rates of non-fatal arterial and venous events are estimated at 1.2% patient-years and 0.6% patients-years, respectively,^{6, 7} but can fluctuate substantially as a multitude of additional factors variably weigh into each individual’s risk.⁸ Although age equal to, or greater than 60 years and prior history of thrombosis are firmly established risk factors for thrombosis in patients with MPN, including ET, several others have been proposed. In particular, the molecular characterization of MPN has enabled the definition of subgroups of patients exhibiting distinct clinical features and potentially, a different risk of thrombosis. In this review, we will discuss the potential role of driver and other recurrent somatic mutations in the modern-day thrombotic risk assessment of ET patients.

Molecular features of ET

Significant advances in understanding the pathobiology of MPN have been made over the last decade. In 2005 independent groups reported the discovery of a gain-of-function mutation of the Janus Kinase (JAK)2 gene in nearly all patients with polycythemia vera (PV) and 50–70% of those with ET or primary myelofibrosis (PMF).^9–12 JAK2 is an intracellular non-receptor tyrosine kinase that plays an essential role in normal hematopoiesis. The most frequent mutation of JAK2 in patients with MPN is a substitution of valine with phenylalanine at codon 617 (that is, V617F), which leads to constitutive activation of the protein and its signaling pathway.^9–12 JAK2 mutations involving exon 12 have been reported in <5% of patients and are usually not associated with ET.^13–15 Another driver mutation found in MPN patients affects the myeloproliferative leukemia (MPL) gene, which encodes the thrombopoietin (TPO) receptor. Collectively, point mutations at exon 10 in codon 515 of the MPL sequence (the most frequent) are found in 3-15% of ET patients. These alterations affect the intracellular domain of the protein and cause ligand-independent signaling through JAK2.^16–18 More recently, somatic mutations of the calreticulin (CALR) gene were discovered in 67-71% and 65-88% of patients with JAK2- and MPL-unmutated ET and PMF, respectively.^{19, 20} CALR is a highly conserved chaperone protein that plays an important role in cellular proliferation, differentiation, and apoptosis.²¹ Mutations occur at exon 9 and consist of either a 52-bp deletion (type 1 mutation), or a 5-bp insertion (type 2 mutation).^{19, 20} As a result of these alterations negatively charged amino acids are replaced with positively charged ones at the C-terminus of CALR,²⁰ causing loss of the endoplasmic reticulum retention sequence, binding of CALR to MPL associated with JAK2, and consequent overactivation of the JAK-STAT signaling pathway.^22–26 Patients who are wild type for the driver mutations JAK2, MPL and CALR are commonly referred to as “triple–negative”. Some of them carry variant gain-of-function MPL or JAK2 mutations.^{27, 28} For the remaining patients a molecular marker has not been identified.

Other recurring non-driver mutations can be found in patients with MPN, but tend to be less frequent in those with ET (28-46% vs. 96% in PMF).^29–31 These lesions involve ten-eleven-ten (TET)2, additional sex combs like (ASXL)1, enhancer of zeste homolog (EZH)2, isocitrate dehydrogenase 1 and 2 (IDH1/2), DNA methyltransferase (DNMT)3A, SH2B adaptor protein (SH2B)3, Casitas B-cell lymphoma (CBL), Ikaros family zinc-finger protein (IKZF)1, high-mobility group AT-hook (HMGA)2, TP53 and NRAS/KRAS (reviewed elsewhere³²). These mutations often accumulate in the same patient and can affect prognosis. For example ASXL1 mutations can co-exist with CALR mutations,^{33, 34} and DNMT3A alterations can occur in concomitance with those of JAK2, TET2 and ASXL1.³⁵ Some mutations (particularly TET2, IDH1 and 2, DNMT3A, LNK, TP53 and N/KRAS) are believed to preferentially occur at later stages of the disease course and/or play a role in leukemic transformation. In a recent study 50 patients with either ET (N = 22) or PV (N = 28) were tested through next-generation sequencing using an 18-gene panel. The group of patients, whose disease progressed within 3 years, was enriched in individuals with more than one mutation (including JAK2V617F) and an increase in the mutated allele burden. A multivariable model for the risk of progression was not reported, likely due to the small number of patients studied.³⁶

Patient- and disease-related risk factors for thrombosis

An accurate medical history can elicit key information on the vascular risk of a patient presenting with newly diagnosed ET. Age ≥60 years and a history of vascular events are well-established risk factors for thrombosis in patients with MPN. The finding of either one assigns an individual to the high-risk category, whereas if none is present patients are considered at low risk.³⁷ Not unexpectedly, general cardiovascular risk factors (for example, tobacco use, hypertension, and diabetes) further increase the likelihood of experiencing a thrombotic episode.^38–42 For example, younger patients who smoked had nearly 2-fold increased risk of thrombosis³⁸ and a shorter 10-year thrombosis-free survival compared to non-smokers.³⁹ Consequently, cardiovascular risk factors have recently been included as variables in prognostic scores for thrombosis in individuals with ET.³⁷

The presence of palpable splenomegaly has also been proposed as a risk factor for vascular events in patients with ET. In a large retrospective analysis (N = 1297) splenomegaly was independently associated with the occurrence of thrombotic events.⁴³ This finding, however, was not confirmed in a similar study (N = 560), perhaps due to differences in diagnostic criteria used and/or length of follow-up.⁴⁴

Finally, baseline and serial blood counts can provide additional important information on the thrombotic risk of ET patients. Indeed, leukocytosis was found to be correlated with the risk of vascular events and death in several retrospective studies.^45–47 In a study of 891 patients with ET, arterial, but not venous, thrombosis was associated with a leukocyte count > 11 × 10⁹/l.³⁸ Thrombocytosis has also been studied, but found to have a weaker correlation.⁴⁸ Interestingly, some studies revealed that, while the association between blood counts at the time of diagnosis and future vascular events may be weak, leukocytosis or abnormal platelet counts during follow-up were more strongly associated with thrombohemorrhagic events.^{47, 49}

Driver mutations and thrombotic risk in patients with ET

JAK2V617F

Mutated JAK2 predisposes to thrombosis

Both a potentially pro-thrombotic clinical phenotype⁵⁰ and pro-thrombotic alterations in blood cells^{51, 52} have been described in JAK2-mutated ET patients. The pathogenesis of such alterations is complex and involves multiple players. Flow cytometry studies revealed abnormalities particularly of platelets (increased expression of CD62P, or P-selectin, CD63, and tissue factor)⁵² and leukocytes (increased expression of CD11b, transferrin and plasminogen activator and increased production of proteases), as well as increased formation platelet–neutrophil aggregates.⁵³ Moreover, an increased production of pro-inflammatory cytokines and various adhesion molecules has been noted in patients with MPN. All these abnormalities were more prominent in patients with mutated allele burden >50%.⁵⁴

Recent studies on heterozygous JAK2-positive mouse knock-in models have shed light on the mechanism of JAK2 mutation-mediated thrombosis. In those animals platelets showed greater tendency to form thrombi, increased reactivity to collagen-related peptide and thrombin, increased aggregation, and increased spreading, regardless of the platelet number (Figure 1). Moreover, gene expression profiles revealed upregulation of genes with a known role in thrombosis.⁵⁵ On the other hand, the presence of JAK2 mutant liver endothelial cells in patients with Budd-Chiari syndrome suggests that the role of the mutation in promoting thrombosis extends beyond hematopoietic cells (Figure 1).⁵⁶ Despite these important laboratory observations, the relationship between JAK2 mutation and the increased risk of thrombosis has been more difficult to ascertain in the clinic.

Figure 1. Suggested role of mutated JAK2 in the pathogenesis of thrombosis in patients with ET.

The JAK2 mutation is present clonally in hematopoietic progenitors and affects all blood cell types in the circulation. Thrombosis occurs due to platelet activation and over-production of tissue factor, as well as the formation of platelet-neutrophil aggregates. Moreover, the increased red blood cell mass observed in JAK2-mutated individuals may further contribute to the genesis of thrombi (dashed arrow). JAK2V617F is also found in other non-blood cell types, such as the liver endothelial cells, where the likelihood of thrombosis may be higher than in the systemic circulation (i.e., Budd-Chiari syndrome, thick arrow). Abbreviations: TF, tissue factor; RBC, red blood cell.

JAK2 mutation is a risk factor for thrombosis

Multiple studies indicated that JAK2 mutation is associated with a distinct clinical phenotype in MPN patients, that is, older age, higher hemoglobin and hematocrit, lower platelet counts, more frequent need for, and greater sensitivity to cytoreductive treatment, and greater tendency to evolve into PV.^{50, 57} In fact, the JAK2 mutation may contribute to the disease phenotype independently of the World Health Organisation (WHO)-defined disease classification, as suggested by the fact that JAK2-positive ET patients have intermediate clinical features between JAK2-wild-type ET and PV (virtually always JAK2-positive).^{58, 59} A large number of studies analyzed the effect of JAK2 mutation on the thrombotic risk of ET patients (Table 1). Most analyses are retrospective in nature and, as such, contain a number of inherent biases. Variables with potential impact on the thrombotic risk that may not have been adequately accounted for include the mutated JAK2 gene dosage effect, dynamic changes in JAK2 allele burden, the presence of concomitant non-driver mutations, and the heterogeneous use of cytoreductive and/or antiplatelet therapy, among others.

Table 1.

Relationship between JAK2 mutation and the occurrence of thrombosis in patients with ET.

First Author, year, reference	N	Number of patients (%)		Median or mean age, years [range] or (±)		Median follow-up		Incidence of thrombosis after study entry, N (%)		P value or OR (95% CI), P-value
		Wild-type	Mutated	Wild-type	Mutated	Wild-type	Mutated	Wild-type	Mutated
Baxter, 200511	51	22 (43)	29 (57)	54 (17)	55 (16)	NR	NR	4 (18)	8 (27)	NS
Campbell, 200550	776	362 (47)	414 (53)	60 (39-77)	52 (32-75)	NR		Ven: 4 (1) Art: 21 (6)	Ven: 12 (3) Art: 25 (6)	Ven: 2.6 (1-6.9), 0.06 Art: 1.1 (0.6-2), 0.8
Ahn, 2007a	24	13 (54)	11 (46)	59 (47-70)	67 (62-72)	27.5 m	37 m	5 (38)	5 (45)	0.744
Alvarez-Larran, 200739	126	59 (47)	44 (53)	29 (7)	30 (7)	10 y		13 (22)	9 (20)	NS
Finazzi, 200758	179	76 (42)	103 (58)	45 (10-78)	50 (16-92)	4.7 y	6.1 y	13 (17)	34 (33)	2.39 (1.16-4.93)
Vannucchi, 20078	639	257 (40)	Hetero: 368 (58) Homo: 14 (2)	46 (17)	Hetero: 52 (17) Homo: 56 (21)	5 y		23 (9)	Hetero: 45 (12) Homo: 6 (43)	Homo vs. WT: 3.97 (1.34-11.7), .013
Heller, 2006a	50	26 (52)	24 (48)	30 (11-77)	48 (21-81)	74.5 m	82.5 m	5 (19)	19 (79)	<0.0001
Hsiao, 2007 a	53	18 (34)	35 (66)	55 (21)	62 (16)	34.5 m		2 (11)	15 (43)	0.029
Ohyashiki, 2008 a	54	21 (39)	33 (61)	57 (18)	63 (14)	NR	NR	0 (0)	6 (18)	P .038
Pemmaraju, 2007 a	80	42 (52)	38 (48)	38 (2-80)	59 (27-84)			Art: 13 (31) Ven: 3 (7)	Art: 10 (26) Ven: 0 (0)	NR
Rudzki, 2007 a	59	21 (36)	38 (64)	48 (17)	51 (13)	26.8 m	27.6 m	9 (43)	15 (39)	NS
Wong, 2008 a	95	35 (37)	60 (63)	54 (21-88)	66 (31-89)	68.4 m		3 (8)	23 (38)	NS
Carobbio, 200959	867	491 (57)	376 (43)	49 (8-88)	56 (12-93)	5 y	3.7 y	34 (9)	54 (11)	1.5 (1-2.31), NR
Giona, 201260	21	11 (52)	10 (48)	15 (5-19)	18 (6-19)	193 m	150 m	0 (0)	1 (10)	NS
Cho, 2009 a	108	47 (43)	61 (57)	61 (27-82)	55 (13-80)	35.7 m		14 (30)	21 (34)	NR
Wolankyj, 200562	150	77 (51)	73 (49)	42 (17-84)	53 (17-98)	11.1 y	11.6 y	22 (28.6)	24 (33)	NS
Antonioli, 200563	130	56 (43)	74 (57)	48 (16-81)	51 (19-84)	NR		13 (23)	17 (23)	NS
Cheung, 2006 a	60	31 (52)	29 (48)	41 (8-77)	46 (20-80)	NR		5 (16)	7 (24)	NS
Chim, 2010 a	141	61 (43)	80 (57)	50 (14)	62 (14)	70 m		8 (13)	27 (34)	0.006
Kittur, 2007 a	176	80 (45)	96 (55)	48 (15-91)	62 (16-88)	59 m		Art: 15 (19) Ven: 2 (3)	Art: 24 (25) Ven: 11 (11)	Art: NS Ven: 0.02
Patriarca, 2010 a	106	46 (43)	60 (57)	68 (25-91)	62 (25-92)	24 m		3 (6)	8 (13)	4.1 (1.7-10.4), <0.003b
Weston, 2011 a	61	20 (33)	41 (67)	59 (23-91)	67 (20-90)	33 m		1 (5)	13 (32)	NR
Pich, 2012 a	103	44 (43)	59 (57)	63 (16)	56 (16)	43.4 m		Ven: 7/37 (19) Art: 2/37 (5)	Ven: 11/49 (22) Art: 4/48 (8)	NS
Palandri, 200964	275	100 (36)	175 (64)	55 (19-80)	63 (16-88)	83 m	73 m	8 (8)	14 (8)	NS
Takata, 2014 a	51	20 (39)	31 (61)	51 (12-80)	69 (28-86)	65 mc		2 (10)	13 (42)
Boroczyk, 2015 a	186	60 (32)	126 (68)	54 (24-82)	58 (19-83)	34 m	28 m	15 (25)	39 (31)	5.15 (1.16-22.90), 0.024
Yonal, 2016 a	107	43 (40)	64 (60)	52 (16)	50 (15)	70 m	70 m	28 (65)	38 (59)	NS

Abbreviations: Art, arterial; CI, confidence interval; ET, essential Thrombocythemia; Hetero, heterozygous; Homo, homozygous; m, months; MPN, myeloproliferative neoplasms; NR, not reported; NS, non-significant; OR, odds ratio; Ven, venous; WT, wild type; y, years.

^aReferenced in Qin et al.⁶⁷, Ziakas⁶⁸ and Dahabreh et al.⁷⁰

^bFor all thrombosis events.

^cIncludes patients with all MPN.

Only studies that report data specifically for ET, and that report thrombosis after study entry are listed.

In a study of very young ET patients, the presence of a JAK2 mutation did not significantly alter the baseline thrombotic risk.⁶⁰ Other series in adults also showed that the presence of JAK2 mutation did not have a significant impact on the risk of thrombosis or lost its predictive value when other variables were controlled for in the analysis.^61–66 A glaring common denominator of all these studies is the small sample size, which weakens the strength of conclusions.

Most studies, on the other hand, suggest a positive correlation between JAK2 mutation and risk of vascular events. Campbell and colleagues analyzed the clinical and laboratory features of patients enrolled in the prospective randomized primary thrombocythemia-1 trial (which compared hydroxyurea plus aspirin with anagrelide plus aspirin) according to the presence of JAK2 mutation. Patients carrying the mutation had features that resembled those of PV, including a higher rate of thrombosis, and were more sensitive to hydroxyurea. Interestingly, while both arterial and venous thrombosis, both pre- and post-study enrollment, were more frequent in JAK2-positive patients, only the difference in rate of venous thrombosis prior to enrollment was statistically significant, suggesting that disease control may attenuate the thrombogenic potential of JAK2V617F.⁵⁰

Subsequent meta-analyses and systematic reviews of the literature showed that, among patients with ET, the risk of thrombosis is about twice as high in those with the JAK2 mutation compared to those without (odds ratio (OR) range 1.83-1.92). The risk was increased for both arterial (OR range 1.68-2.59) and venous thrombosis (OR range 2.09-2.5).^67–69 In the meta-analysis by Dahabreh and colleagues,⁷⁰ that examined 2,436 patients (1,375 of whom JAK2-mutated), and included blood counts in the meta-regression, the authors found that leukocytosis might in part mediate the impact of JAK2 mutation on thrombotic events. In addition, in line with results from the primary Thrombocythemia-1 cohort,⁵⁰ two meta-analyses revealed that the odds of thrombosis may be higher before vs. after ET diagnosis, confirming that control of the disease might reduce the risk of thrombotic events regardless of the JAK2 mutation status.^{67, 70} Due to an intrinsic lack of homogeneity and detail, these analyses did not clarify the contribution of JAK2 allelic status (homozygous vs. heterozygous) or burden to the thrombotic risk.

Mutated JAK2 gene dosage has an effect on thrombotic risk

Although rare in ET patients (2-4%), JAK2V617F homozygosity might confer an increased risk of thrombosis compared to a heterozygous or wild-type condition.⁸ Perhaps more importantly, a correlation was noted between the JAK2 allele burden and a number of parameters, including older age, platelet and leukocyte counts, palpable splenomegaly, and risk of venous thrombosis.⁷¹ Another study showed that an allele burden of 20-25% or higher independently predicted the risk of arterial and venous thrombosis.⁷² Finally, among patients with WHO-defined ET or early PMF a high (>50%) JAK2 allele burden was found to positively correlate with the thrombotic risk regardless of the WHO diagnosis.⁷³

JAK2 mutation has an impact on vascular events over time

In some ET patients, the pro-thrombotic potential of the JAK2 mutation may manifest itself over time, rather than early in the course of the disease. A study from Carobbio and colleagues⁵⁹ compared and contrasted the impact of the JAK2 mutation in patients with ET (N = 867) and PV (N = 415). Among patients with wild-type ET, JAK2-mutated ET and PV, the rate of arterial and venous thrombosis were similar in the first 5 years, but with additional follow-up the risk of vascular events in patients with JAK2-mutated ET and PV became similar, and higher than in patients with wild-type ET. Another analysis of the dynamics of the JAK2V617F burden suggested that patients with persistently high or unstable mutation load over time had a trend toward a higher incidence of thrombosis.⁷⁴ Finally, in a retrospective study of 143 individuals with ET and previous major thrombosis 30% experienced recurrence during follow-up. Homozygosity for JAK2V617F independently predicted an increased risk of recurrent thrombosis (hazard ratio 6.15), while the risk was similar between heterozygous and wild-type patients.⁷⁵

JAK2 mutation is associated with thrombosis at unusual sites

Venous thrombosis in patients with MPN may occur at unusual sites, such as the splanchnic and cerebral venous systems. However, the prevalence of JAK2 mutation, with or without “occult” MPN, in patients presenting with various types of venous thrombosis varies substantially in retrospective case series, potentially due to the significant heterogeneity of the populations studied.⁷⁶

For example, in a survey of 24 studies (N = 3,123) JAK2V617F was found on average in 32.7% of patients with splanchnic vein thrombosis (SVT) (OR 53.98). The estimated attributable risk of having a SVT for patients with a JAK2 mutation was 12.74%. In contrast, the mean prevalence of JAK2 mutation in patients with venous thrombosis at other sites ranged from 0.88% to 2.57%.⁷⁷ Other studies confirmed the low (0.9%) frequency of JAK2V617F in patients presenting with non-SVT and without history or signs of MPN, and showed that the median allele burden was low at presentation (around 5%).⁷⁸

Another meta-analysis showed that the prevalence of MPN and JAK2V617F in patients presenting with either Budd-Chiari syndrome (N = 1,062) or non-cirrhotic portal vein thrombosis (N = 855) were 40.9% and 41.1%, respectively, in the former and 31.5% and 27.7%, respectively, in the latter. The differences between the two patient groups were statistically significant. MPN was “occult” (that is, absence of symptoms or signs of disease and normal blood counts) in 17.1% and 15.4% of patients with Budd-Chiari syndrome and portal vein thrombosis, respectively.⁷⁹ Although other smaller studies suggested a lower prevalence of JAK2 mutation (<10%) in patients presenting with isolated SVT,⁸⁰ the current clinical practice contemplates screening for JAK2V617F and underlying MPN in patients presenting with isolated SVT of no obvious etiology. The value of such screening in patients with non-SVT remains to be defined. In addition, due to a high recurrence rate, SVT in patients with MPN is generally treated with long-term anticoagulation therapy.

CALR mutations

Studies of platelet activation markers revealed that, compared to JAK2-mutated patients, those harboring the CALR mutation display decreased platelet-neutrophil complexes and CD14+ expression on neutrophils, and longer platelet function analyzer-100 with collagen and epinephrine cartridge closure time. Leukocyte activation, however, was not different in CALR-mutated vs. triple-negative patients, suggesting that the absence of JAK2 mutation, rather than to the presence of CALR mutation, better explains the lesser pro-thrombotic tendency of these patients.⁸¹

Phenotypically, CALR-mutated ET patients are more frequently younger males with lower hemoglobin and white blood cell count, higher platelet count, greater marrow megakaryocytic predominance and lower rates of polycythemic transformation, compared to those with JAK2 mutation.^{19, 20, 82–84}

A synopsis of studies comparing the risk of thrombosis among patients with molecularly annotated ET is presented in Table 2. In an analysis of 576 ET patients, 15.5% of whom had CALR mutations, the 10-year cumulative incidence of thrombosis was 14.5% and 5% for JAK2-positive and CALR-positive patients, respectively. The thrombotic risk of CALR-mutated patients was similar to that of triple-negative ones and lower than JAK2- and MPL-mutated patients.⁸² A second study (N = 745), confirmed these findings, as well as the validity of the International Prognostic Score for Thrombosis in Essential Thrombocythemia (IPSET-thrombosis) model when CALR mutations were taken into consideration (see below).⁸¹ In a third series of 217 young patients with WHO-defined ET or early PMF, the lower incidence of thrombosis in patients with CALR mutation vs. JAK2 mutation persisted at 15 years (21.7% vs. 9.1%, respectively P = 0.04).⁷³

Table 2.

Relationship between driver mutation and occurrence of thrombosis in patients with molecularly annotated ET.

First Author, year	Number of patients (%)				Median or mean age, years				Median follow-up				Incidence of thrombosis, N (%)				P-value
	JAK2	CALR	MPL	TN	JAK2	CALR	MPL	TN	JAK2	CALR	MPL	TN	JAK2 (A)	CALR (B)	MPL (C)	TN (D)
Fu, 201485	240 (55)	99 (24)	6 (1)	89 (20)	56	50	50	50	37 m	37 m	55 m	38 m	28 (12)	4 (4)	0 (0)	3 (3)	A vs. B 0.029 A vs. D 0.022
Gangat, 201592	160 (53)	95 (32)	8 (3)	37 (12)	59	47	66	42	12.7 y				47 (29)	17 (18)	3 (37)	4 (11)	0.03
Rumi, 201484	466 (73)	176 (27)	-	-	50	45	-	-	5.2 y		-	-	25/1000 p-y	10/1000 p-y	-	-	0.001
Rotunno, 201482	369 (64)	89 (15)	25 (4)	93 (16)	61	55	54	53	71.9 m				111 (30)	12 (13)	10 (40)	15 (16)	A vs. B 0.011 B vs. C 0.012 B vs. D 0.894
Sun 2016104	149 (49)	92 (30)	65 (21)		62	55	45		102 m				2	1	5		NS
Tefferi, 201586	227 (57)	114 (28)	11 (3)	50 (12)	58	47	64	42	NR				55 (24)	20 (18)	3 (27)	4 (8)	0.03
Bertozzi, 2016105	114 (62)	44 (24)	3 (2)	22 (12)	56	49	60	38	9.6 y	11.9 y	13.3 y	9.7 y	26 (23)	6 (14)	1 (33)	0 (0)	0.027
Okabe, 2016106	179 (61)	47 (16)	10 (3)	57 (19)	63	60	68	52	NR				12	6	0	0	NS
Palandri, 201573	132 (71)	54 (29)	-	-	34	34	-	-	10.9 y	9.8 y	-	-	10-y: 11% 15-y: 21.7%	10-y: 3.5% 15-y: 9.1%	-	-	0.042
Torregrosa, 201681	40 (61)	13 (20)	12 (19)		66	51	72		10.4 y				12 (30)	0	1 (8)		0.023
Alvarez-Larran, 2016103	162 (28)	271 (62)	-	-	42	42	-	-	2215 p-y		-	-	ASA: 11.6/1000 p-y OBS: 21.6/1000 p-y	ASA: 9.7/1000 p-y OBS: 6.9/1000 p-y	-	-	NR
Vannucchi, 200889	546 (55)	418 a (42)	30 (3)		56	50a	56		59 m				Art: 20 (4) Ven: 12 (2)	Art: 6 (1) a Ven: 5 (1) a	Art: 4 (13) Ven: 2 (7)		Art: A vs. B: 0.04 A vs. C: 0.05 B vs. C: 0.002 Ven: NS
Beer, 200891	411 (53)	333 (43) a	32 (4)	-	60	52 a	67	-	36.5 m				Art: 25 Ven: 11	Art: 19 a Ven: 3 a	Art: 2 Ven: 2	-	Art: NS Ven: 0.02

Abbreviations: Art, arterial; ASA, low-dose aspirin; CALR, calreticulin; ET, essential Thrombocythemia; JAK2, Janus kinase; m, months; MPL, myeloproliferative leukemia; NR, not reported; NS, non-significant; OBS, observation; p-y, person-years; TN, triple negative; Ven, venous; y, years.

^aAssumed to include TN patients.

Only studies reporting mutation-specific data are listed. Only thrombosis after study entry is indicated.

The type of CALR mutation may also influence the risk of thrombosis in individuals with ET. Among 99 patients with CALR mutations, 3 of 38 with type 2 vs. 10 of 55 with type 1 mutations had a thrombotic event, although this difference did not reach statistical significance (P = 0.160).⁸⁵ Similarly, in large international retrospective data set (N = 1027), the thrombosis-free survival of type 1 and type 2–mutated patients were not different (P = 0.18).⁸⁶ A third study showed that patients with type 1-like mutations had an “intermediate” phenotype between those with type 2 mutations and those with JAK2 mutation. In addition, patients with type 1 mutations tended to have a PMF-like phenotype, whereas those with type 2-like mutations had an “ET-proper” phenotype with higher platelet counts and a more indolent clinical course.⁸⁷ We recommend caution in interpreting the results of these studies given the small number of thrombotic events recorded in each subgroup.

Of note, whereas the prevalence of JAK2 mutation is high (albeit variable across studies) in patients with SVT, CALR mutations are very rarely found in this setting (0-2%). Nonetheless, the population of patients with SVT, who are found to have an MPN and do not harbor JAK2 mutations may be enriched for these alterations (34.6%) and, thus, be candidate for CALR mutation analysis.⁸⁸

MPL mutations

Despite the low frequency, MPL mutations have also been associated with distinct clinical features. These individuals resemble JAK-mutated patients for age and gender distribution, and CALR-mutated ones for blood counts. Marrow erythroid and overall cellularity, however, are uniquely low.⁸⁹

Data on the thrombotic risk associated with MPL mutations are scant, primarily due to low overall frequency (Table 2). The 5-year cumulative incidence of thrombosis is estimated to be around 9%.⁹⁰ In the prospective primary thrombocythemia-1 cohort, among 776 patients with ET 32 (4.1%) had MPL mutations and these were not predictive of thrombosis, major hemorrhage, myelofibrotic transformation or survival in controlled analyses.⁹¹ In a larger series (N = 994) 3% of patients had the MPL mutation, which co-existed with JAK2 in eight cases. In this study the MPL mutation was associated with higher risk of thrombosis only when compared to JAK2 mutation or wild-type MPL (presumably mostly CALR-mutated).⁸⁹

“Triple-negative” state

Triple-negative patients constitute a small group, for which available clinical information is limited (Table 2). In an analysis of 300 cases (53% JAK2-, 32% CALR-, 3% MPL-mutated, and 12% triple-negative) 35% experienced arterial (n = 75) or venous (n = 43) thrombosis. In univariate analysis, triple-negative state and CALR mutation conferred a lower risk of thrombosis compared to JAK2 mutation. The triple-negative state remained significant when age and thrombosis history were controlled for.⁹² Another study (N = 217) confirmed the low risk of thrombosis in triple-negative patients, with no events recorded among 28 (13%) patients after a follow-up of 10.2 years.⁷³

Interactions between mutations

The risk of thrombosis may be modulated by the differential combination of driver and non-driver mutations but this hypothesis has been harder to demonstrate, given the exiguous number of such patients. Although the presence of additional mutations may modulate disease course and progression to acute leukemia, limited data exist on how this impacts the thrombotic risk.⁹³ For example, in a cohort of 167 ET patients JAK2 and CALR mutations coexisted in 7 (4.2%) cases. The JAK2+/CALR- group appeared to have the highest incidence of thrombosis, but the difference was only statically significant in comparison with the JAK2-/CALR- group.⁹⁴ In a study of 190 ET patients, 8 had concomitant CALR and ASXL1 mutation, and seemed to have more pronounced anemia, but no data on the thrombotic risk were reported.⁹⁵ In another experience, ASXL1 mutations (detected in 8.4% of ET patients, and co-existent with JAK2 mutations in two- thirds of them) appeared to confer a statistically insignificant increase in the risk of arterial, but not venous thrombosis.⁹⁶

In an elegant study of 24 patients with MPN, hematopoietic stem cells and progenitor cells were isolated and subjected to genotyping or next generation sequencing. Patients who acquired the JAK2 mutation before the TET2 mutation were younger, more likely to present with PV (rather than ET), had greater percentage of megakaryocytic and erythroid progenitors, exhibited a higher risk of arterial and venous thrombosis, and their progenitor myeloid cells exhibited greater sensitivity to ruxolitinib in vitro compared with those who had acquired a TET2 mutation first. In fact, in these cells, JAK2V617F-mediated up-regulation of genes associated with cell proliferation was prevented.⁹⁷

Current thrombotic risk assessment in patients with ET

Efforts have been made to improve the current dichotomous risk stratification of ET patients based on age and thrombosis history, particularly in light of data suggesting a pro-thrombotic role of JAK2 mutations. The IPSET-thrombosis incorporated cardiovascular risk factors and JAK2 mutation in the model and identified patients with low, intermediate or high thrombosis risk (1.03%, 2.35%, and 3.56% patients/year, respectively) (Figure 2).⁹⁸ An external validation study showed that only intermediate- and high-risk patients derived significant clinical benefit from cytoreductive therapy.⁹⁹ Unlike JAK2V617F, CALR mutations tend to segregate with other low-risk factors in this model and do not appear to modify a patient’s risk.¹⁰⁰ We propose that the IPSET-thrombosis score be followed to inform decisions on the use of antiplatelet and/or cytoreductive therapy in ET patients in routine clinical practice.

Figure 2. Evolution of the thrombotic risk assessment in patients with ET.

The figure illustrates how newer prognostic scores for thrombosis in patients with ET were derived from the conventional risk stratification valid for all MPN patients. The incorporation of new risk factors in the models and a different emphasis placed on age and thrombosis history are highlighted. Abbreviations: CV, cardiovascular; IPSET, International Prognostic Score for Thrombosis in Essential Thrombocythemia.

The IPSET-thrombosis score was recently revised based on new observations. For conventionally defined high-risk patients, JAK2 mutations appeared to have an added prognostic value only in patients >60 years without a history of thrombosis. On the other hand, for conventionally defined low-risk patients the thrombotic risk was very low if JAK2 was unmutated, and even lower in the absence of cardiovascular risk factors (1.05% and 0.44% patient-years, respectively). Patients previously adjudicated low-risk were divided into low risk and very low risk, based on the presence or absence, respectively, of JAK2 mutations. The revised IPSET-thrombosis, therefore, identified four categories: 1) high risk (thrombosis history or age > 60 years with JAK2 mutation); 2) intermediate risk (no thrombosis history, age > 60 years and JAK2-unmutated); 3) low risk (no thrombosis history, age ≤ 60 years and JAK2-mutated); and 4) very low risk (no thrombosis history, age ≤ 60 years and JAK2-unmutated) (Figure 2).¹⁰¹ The revised IPSET-thrombosis was also externally validated in a cohort of 585 ET patients, where only JAK2 mutations and a history of thrombosis were independently predictive of thrombotic events.¹⁰² No data are available yet on the role of other mutations in the setting of the revised IPSET-thrombosis score.

Concluding remarks

The molecular characterization of MPN is among the most comprehensive in hematologic malignancies, as a recognized driver lesion can be identified in nearly all cases and studies have begun to unravel the clonal architecture of these neoplasms. However, establishing clinical-molecular correlations in ET is a difficult exercise given the long life expectancy of patients, the small number of outcome-defining events, and the rarity of some mutations.

Nonetheless, some generalizations relevant for clinical practice can be made. First, among all (driver or non-driver) mutations, only JAK2V617F was found to modify the thrombotic risk in ET patients consistently across studies. This is likely because, unlike CALR mutations, JAK2V617F imparts a disease phenotype closer to that of PV, and thus more prone to thrombosis. In other words, the low-risk profile of CALR-positive ET may result from the lack of JAK2 mutation-associated high-risk features, rather than an effect of the CALR mutation per se.

Secondly, CALR mutations tend to segregate closely with other established low-risk factors, and do not appear to have an independent prognostic significance.^{84, 100} Although longer follow-up and more detailed studies will be necessary, in our estimation this mutation is unlikely to be included in prognostic models in the future.

Third, the contribution of individual mutations to ET patients’ thrombotic risk may inform specific clinical decision-making. For instance, given their extremely low thrombotic potential, patients who are classified as very low risk according to the revised IPSET-thrombosis model, are unlikely to benefit from primary prophylaxis with low-dose aspirin.¹⁰¹ Whether management of antiplatelet therapy should be diversified in patients with mutations other than JAK2V617F, is less clear. Among 271 patients with CALR mutations, not only low-dose aspirin did not decrease the risk of thrombosis, but it was also associated with higher incidence of bleeding. Notably, the opposite was true for JAK2-mutated patients. The results of this study are difficult to interpret because: i) the proportion of patients treated with antiplatelet therapy in CALR- vs JAK2-mutated patients is unknown, ii) a multivariable model for risk of bleeding was not reported, and iii) cytoreduction was started much earlier in CALR- vs. JAK2-mutated patients mostly to control the higher platelet count.¹⁰³

In conclusion, testing for somatic mutations has made its way into the thrombotic risk assessment of patients with ET. The presence of JAK2V617F independently increases such risk and may inform management at the time of presentation. On the other hand, CALR mutations typically cluster in patients who are low risk based on other clinical and laboratory features, and may not affect the decision-making process. Additional data on larger numbers of MPL-mutated or triple-negative patients will be necessary to determine their role, if any, in modulating the vascular risk of ET patients. At the present time, testing for additional non-driver mutations does not appear to have obvious clinical implications based on limited available evidence, and is not warranted in routine clinical practice.