Characteristics and risk factors for thrombosis in POEMS syndrome: a retrospective evaluation of 230 patients

Abstract

Thromboses are prevalent in POEMS syndrome, but few risk factors for POEMS-associated thrombosis have been identified. The objective of this study is to identify novel risk factors for POEMS-associated thrombosis. In this retrospective cohort of 230 POEMS patients, 27% developed thrombosis. Arterial events were slightly more common than venous. Stroke accounted for 26% of all thromboses and 53% of arterial events. There were differences in baseline features between the thrombosis group and the no thrombosis group, and these were driven by patients with arterial thrombosis. Risk factors for arterial thrombosis included thrombocytosis, elevated hemoglobin/hematocrit, extravascular volume overload and splenomegaly. Hyperprolactinemia appeared to be a risk factor for venous thrombosis. The risk of thrombosis was most striking among men with elevated hemoglobin (32% versus 5%, p <0.001) and hematocrit (42% versus 5%, p <0.001) compared to men without. Most thromboses occurred prior to POEMS directed therapy, and most that occurred during therapy happened within 3 months of diagnosis. Twenty-one percent of patients with thrombosis had recurrence. In recognition of high overall rates of thrombosis in this population, all patients with POEMS syndrome should receive prophylactic antiplatelet therapy, and clinicians should consider anticoagulation in patients with risk factors for POEMS-associated thrombosis.

Introduction

POEMS syndrome is a rare, multisystem paraneoplastic disorder driven by an aberrant plasma cell clone. The POEMS acronym, coined by Bardwick in 1980, summarizes five common associated signs and symptoms of this disease (polyneuropathy, organomegaly, endocrinopathy, monoclonal protein, and skin changes).¹ Additional features include papilledema, extravascular volume overload, sclerotic bone lesions, thrombocytosis (PEST), elevations in vascular endothelial growth factor (VEGF), and Castleman disease.

Although not included in formal diagnostic criteria, the association between POEMS syndrome and thrombosis is well established in case reports and larger observational studies.^2–8 Thrombosis occurs in 20% to 30% of POEMS patients, and ischemic stroke in 8% to 10%.^5–9 Studies on risk factors for thrombosis in POEMS syndrome, however, are limited and feature conflicting results. In one report, bone marrow plasmacytosis and thrombocytosis were associated with an increased risk of ischemic stroke.⁶ However, more recent investigations found no association between thrombocytosis and either ischemic stroke or thrombosis in general.^{7, 8} In one study, thrombotic risk was highest during the active disease phase, defined as a serum VEGF > 1000 pg/mL, but no additional risk factors were identified.⁷

The aims of this study are to provide a comprehensive overview of thrombosis in POEMS syndrome and to ascertain risk factors for POEMS-associated thrombosis. Identifying POEMS patients at high risk for thrombosis can assist clinicians in selecting appropriate prophylactic regimens for risk reduction.

Methods

Patient identification and inclusion criteria

The study cohort included all patients with a diagnosis of POEMS syndrome evaluated at Mayo Clinic, Rochester from June 2002 to March 2021. All cases satisfied diagnostic criteria established by Dispenzieri et al and endorsed by the International Myeloma Working Group.^{9, 10} Demographic, clinical, and laboratory data were collected through retrospective review of the electronic medical record. This study was approved by the Mayo Clinic Institutional Review Board, and all patients provided consent for study enrollment.

Baseline laboratory evaluation

Because patients with POEMS are often misdiagnosed with chronic inflammatory demyelinating polyneuropathy (CIDP) prior to their definitive diagnosis of POEMS syndrome, some receive intravenous gammaglobulin (IVIG), plasmapheresis, corticosteroids and myelosuppressive drugs such as azathioprine to treat CIDP. Patients who received myelosuppressive medications prior to a POEMS syndrome diagnosis were deemed to have missing baseline complete blood count (CBC). Similarly, if patients were on corticosteroids at or near the time of diagnosis, the baseline adrenal axis could not be assessed. The window permitted for baseline pulmonary function testing (PFT) and trans-thoracic echocardiograms (TTE) was within 90 days before or after instituting POEMS directed therapy.

Pulmonary hypertension was defined as right ventricular systolic pressure > 35 mmHg, and abnormal corrected diffusing capacity for carbon monoxide (DLCO) as < 70% of predicted, in accordance with previous studies on POEMS-associated pulmonary disease.¹¹ Restrictive patterns on PFTs were identified by the interpreting pulmonologist.

Because of relatively small sample sizes and differences between normal red cell indices for men and women, hemoglobin and hematocrit were “normalized” by dividing the measured value by the upper limit of normal (ULN) for sex so that all patients could be considered together. ULN values included hemoglobin of 16.6 g/dL for men and 15.0 g/dL for women, and hematocrit of 48.6% for men and 44.9% for women. For CBC parameters, we chose cut-off values with the highest corresponding Youden’s index derived from receiver operator characteristics curves.¹² High values were defined as platelet count > 437, normalized hemoglobin > 0.97, and normalized hematocrit > 0.99.

Statistical analysis

Differences in medians between groups were compared using the Wilcoxon rank-sum test, and proportions were compared using chi-square and Fisher’s exact tests. The log rank test was used to compare differences in time to event outcomes between groups. P values less than 0.05 were considered statistically significant. Statistical analysis was performed with BlueSky Statistics software v.7.2.

Results

Demographic and clinical characteristics

Two hundred and thirty patients with POEMS syndrome were seen between June 2002 and March 2021. The median age of the population was 52 (interquartile range 43–59), and 65% were male. Sixty-one patients (27%) had thrombosis. Rates of arterial, venous, and line-associated thromboses were 13%, 11%, and 2%, respectively. As shown in Table 1, there were only a few clinical features that differed between those with and without thrombosis. Patients were similar in terms of age, sex, race, sclerotic lesions, skin changes, papilledema, and pulmonary findings. Extravascular volume overload was more common in patients with thrombosis than in patients without thrombosis (95% versus 83%, respectively, p=0.02). These differences were driven by the arterial thrombosis subgroup. Patients with arterial thrombosis had significantly more effusions (53% versus 27%, p=0.004) and ascites (37% versus 20%, p=0.04) than the no thrombosis group. In addition, splenomegaly was more common in patients with arterial thrombosis than in patients with no thrombosis (60% versus 39%, respectively, p=0.03). As shown in Supplemental Table 1, there was no relationship between endocrine axis abnormalities and thrombosis, except for elevated prolactin, which was more common among patients with venous thrombosis (71% versus 44%, p=0.02).

Table 1:

Demographic and clinical characteristics at diagnosis for 230 POEMS patients with and without thrombosis, including arterial and venous thrombosis subgroups ^a

	No Thrombosis	Thrombosis
		Any	Arterial	Venous b
Patients	169 (73)	61 (27)	30 (13)	26 (11)
Male sex	108 (64)	42 (69)	22 (73)	19 (73)
White race	138 (82)	46 (75)	25 (83)	19 (73)
Median (IQR) age at diagnosis	53 (44, 59)	51 (41, 59)	56 (44, 61)	48 (40, 58)
Neuropathy	169 (100)	61 (100)	30 (100)	26 (100)
Organomegaly	107 (63)	44 (72)	24 (80)	17 (65)
Splenomegaly	66 (39)	30 (49)	18 (60) c	11 (42)
Hepatomegaly	41 (24)	13 (21)	8 (27)	5 (19)
Lymphadenopathy	72 (43)	29 (48)	14 (47)	13 (50)
Biopsy proven Castleman disease	17 (10)	6 (10)	4 (13)	0
Sclerotic lesions	139 (82)	47 (77)	23 (77)	21 (81)
Extravascular volume overload	140 (83)	58 (95) c	28 (93)	25 (96)
Peripheral edema	137 (81)	58 (95) d	28 (93)	25 (96)
Effusions	46 (27)	26 (43) c	16 (53) d	9 (35)
Ascites	34 (20)	17 (28)	11 (37) c	6 (23)
Skin changes	134 (79)	50 (82)	26 (87)	19 (73)
Papilledema	39 (23)	16 (26)	11 (37)	4 (15)
Pulmonary hypertension, n/N (%) e	34/75 (45)	14/27 (52)	7/13 (54)	7/12 (58)
Restrictive lung disease, n/N (%) f	29/99 (29)	13/41 (32)	7/22 (32)	5/15 (33)
Corrected DLCO
DLCO < 70%, n/N (%) g	42/86 (49)	20/33 (61)	10/18 (56)	8/11 (73)
Median (IQR) percent predicted corrected DLCO	71 (60, 78)	65 (52, 76)	67 (50, 76)	64 (61, 77)

Abbreviations: IQR, interquartile range; DLCO, diffusing capacity for carbon monoxide

^aSummary data are given as number (percent), unless otherwise noted

^b5 patients with line associated thrombosis not included

^cp<0.05 as compared to no thrombosis

^dp<0.01 as compared to no thrombosis

^e102 patients had baseline right ventricular systolic pressure available, including 2 with line associated thromboses.

^f140 patients had assessment for restrictive lung disease, including 4 with line associated thromboses.

^g119 had a corrected DLCO available, including 4 with line associated thromboses.

High platelet count was more common in patients with thrombosis compared to patients without thrombosis (66% versus 39%, p=0.004) (Table 2). Upon subgroup analysis, this effect held true for arterial events only and was most striking for men. Twenty-nine percent of men with high platelet counts had arterial thrombosis compared with 6% of men with normal platelet count (OR 6.67, 95% CI 1.68–26.41, p=0.003). Of note, only 5 women with arterial thrombosis had a baseline platelet count available.

Table 2:

Rates and risk for thrombosis based on platelet count and red cell indices

	No Thrombosis	Thrombosis	OR (95% CI)	Arterial Thrombosis	OR (95% CI)	Venous c Thrombosis	OR (95% CI)
Plt count available, n	117	38		18		16
Plt count (×109/L), median (IQR)	412 (308, 510)	497 (357, 576) a		521 (437, 618) b		411 (313, 516)
Plt count > 437 ×109/L, n (%)	46 (39)	25 (66) b	2.97 (1.38, 6.39)	13 (72) b	4.01 (1.34, 12.01)	8 (50)	1.54 (0.54, 4.40)
HgB available, n	114	37		17		16
NHgB e, median (IQR)	0.94 (0.83, 1.00)	0.99 (0.86, 1.05) a		1.01 (0.96, 1.07) a		0.96 (0.81, 1.02)
NHgB > 0.97, n (%)	36 (32)	22 (60) b	3.18 (1.48, 6.84)	11 (65) b	3.97 (1.36, 11.58)	7 (44)	1.69 (0.58, 4.88)
HCT available, n	104	33		15		14
NHCT f, median (IQR)	0.92 (0.84, 0.99)	1.00 (0.90, 1.05) b		1.03 (0.97, 1.07) a		0.99 (0.81, 1.03)
NHCT > 0.99, n (%)	28 (27)	19 (58) b	3.68 (1.63, 8.32)	10 (67) b	5.43 (1.71, 12.28)	6 (43)	2.04 (0.65, 6.40)

Abbreviations: OR, odds ratio; CI, confidence interval; Plt, platelet; IQR, interquartile range; NHgB, normalized hemoglobin = actual value divided by upper limit of normal for sex (hemoglobin 16.6 g/dL for men, 15.0 g/dL for women); NHCT, normalized hematocrit = actual value divided by upper limit of normal for sex (hematocrit 48.6% for men, 44.9% for women)

^ap<0.05 as compared to no thrombosis

^bp<0.01 as compared to no thrombosis

^c4 with line associated thrombosis not included

Elevated sex adjusted hemoglobin (60% versus 32%, OR 3.18, p=0.002) and hematocrit (58% versus 27%, OR 3.68, p=0.001) were more common in the thrombosis group, notably the arterial event group. Median hemoglobin in men with thrombosis was higher than in men without thrombosis (16.9 g/dL versus 15.2 g/dL, p <0.001) but similar in women (14.7 g/dL versus 14.3 g/dL, p=0.78). The interaction between male gender, arterial thrombosis, and baseline hemoglobin (or hematocrit) was most striking. Thirty-two percent of men with high hemoglobin had arterial thrombosis compared with 5% of men with normal hemoglobin (OR 8.84, 95% CI 2.17–36.09, p <0.001). Forty-two percent of men with high hematocrits had arterial thrombosis as compared to 5% without (OR 14, 95% CI 3.29–59.55, p <0.001). There were no significant associations between arterial thrombosis and red cell indices in women, however only 5 women had hemoglobin available and 3 had hematocrit.

Characterization of thrombotic events

Of the 61 initial thrombotic events, there were 16 ischemic strokes, which were the most common thrombotic event, accounting for 26% of all thromboses and 53% of arterial thromboses. Other arterial events included coronary artery (n=7), acute limb ischemia (n=2), splenic infarct (n=2), transient ischemic attack (TIA) (n=2), and chronic limb threatening ischemia (n=1). Deep vein thrombosis (DVT) was the most common venous thrombosis, occurring in 18 cases (69%). Six of these patients also had pulmonary embolism (PE), and another 4 had isolated PE. The other non-line associated venous thrombotic events included: superficial venous thrombosis (SVT) (n=2), venous sinus thrombosis (n=1), and ventricular thrombus (n=1). The line associated events included 3 DVTs, 1 DVT with associated PE, and 1 SVT.

Timing of thrombosis

Figures 1A and 1B and Supplemental Table 2 illustrate the timing of symptom onset and thrombosis in relation to diagnosis for patients with arterial and venous thromboses, respectively. Most arterial and venous thromboses occurred prior to instituting POEMS directed therapy (77% and 54%, respectively) (Table 3). For 7 patients (3 arterial and 4 venous), thrombosis antedated other POEMS symptoms. For the remaining 20 arterial events and 10 venous events, median time from POEMS symptom onset to thrombosis was similar (1.23 versus 1.21 years, p=0.16). Of these, thrombosis preceded POEMS diagnosis in 28 patients, and median time from thrombosis to diagnosis was similar for arterial and venous events, (0.34 versus 0.63 years, p=0.90). Two additional patients had thrombosis after diagnosis, but prior to therapy: 1 arterial and 1 venous thrombosis. Eleven patients had thrombosis during initial POEMS therapy; of these, median time from diagnosis to thrombosis was 0.33 years and 0.17 years for 4 arterial and 7 venous events, respectively (p=0.15).

Figure 1:

Temporal relationships between time from diagnosis and initial thrombosis

Time from diagnosis to first thrombosis stratified by A) arterial and B) venous thrombosis. Time zero corresponds to diagnosis and numbered rows represent individual patients. The ends of horizontal bars correspond to initial thrombotic events, such that patients with bars to the left of time zero had initial thromboses before a POEMS diagnosis, and patients with bars to the right of time zero had initial thromboses after diagnosis. Time of symptom onset is indicated by diamond symbols. Line associated thromboses are not included.

Table 3:

Characteristics of 61 initial thrombotic events in POEMS patients stratified by thrombosis type ^a

	Arterial	Venous	Line associated
Number of patients b	30 (49)	26 (43)	5 (8)
Timing of initial thrombosis
Prior to initial POEMS therapy	23 (77)	14 (54)	2 (40)
During initial POEMS therapy	4 (13)	7 (27)	2 (40)
Relapse	2 (7)	2 (8)	1 (20)
Remission and off POEMS therapy	1 (3)	3 (12)
On steroids	10 (33)	7 (27)	2 (40)
On IVIG	4 (13)		1 (20)
On IMiD	2 (7)	3 (12)
Hospitalized c	6 (20)	5 (20)	3 (60)
On anticoagulation	1 (3)	1 (4)
On antiplatelet	3 (10)	3 (12)

Abbreviations: IVIG, intravenous immunoglobulin; IMiD, immunomodulatory imide drug

^aSummary data are given as number (percent)

^bPercent of all 61 patients with thrombosis

^cHospitalized at the time of thrombosis

Approximately one-third of patients with arterial or venous events were taking corticosteroids at the time of thrombosis (Table 3). Four patients with arterial events were on IVIG for treatment of presumed CIDP prior to POEMS diagnosis. Five patients were taking an immunomodulatory imide drug (IMiD) at the time of the initial thrombosis (lenalidomide, 4; and thalidomide, 1). Overall, six patients had thrombotic events despite being on antiplatelet therapy, including 1 who was on IVIG and 2 who were on concurrent IMiD therapy. In total, 5 of 6 patients had a CBC at diagnosis; 3 had thrombocytosis and 3 had polycythemia. Two patients had breakthrough thromboses while on anticoagulation, both were taking IMiDs, and one had baseline polycythemia and thrombocytosis.

Treatment of initial thrombosis

Treatment regimens for the initial thrombotic events are outlined in Supplemental Table 3. The most common treatment strategy for arterial events was dual anti-platelet therapy (DAPT) (37%). In contrast, warfarin was the most common treatment for venous thromboses (50%). Eleven patients (18%) with thrombosis required invasive procedures including coronary artery stenting, thrombectomy, coronary artery bypass grafting, angioplasty, splenectomy, and catheter directed thrombolysis. Five patients did not receive therapy, including 1 with splenic thrombosis who was treated conservatively, 1 with TIA who was severely thrombocytopenic, 2 SVT, and 1 line-associated SVT. Characteristics of each patient’s initial thrombotic event, including use of concurrent antiplatelet, anticoagulant, or POEMS directed therapy, and subsequent treatment are listed in Supplemental Table 2.

Multiple thromboses

Of 61 patients with thrombosis, 13 (21%) had recurrent thrombotic events (Supplemental Table 4). Most patients with recurrent thrombosis (77%) had a total of 2 events. For the first recurrent thrombosis, venous events (n=7) were slightly more common than arterial (n=5). DVT (6, 46%) was the most common type of thrombosis overall. Median time from initial thrombosis to first recurrence was 2.0 years, and for 7 patients, both the initial and first recurrent events occurred prior to initiation of POEMS directed therapy. Notably, 3 patients developed recurrent thrombosis while on anticoagulation, including one on IMiD therapy, and six developed recurrent thromboses on antiplatelets.

Of the 7 patients with recurrent thrombosis evaluated for thrombophilia, 5 had abnormal testing. One patient had both protein C deficiency and anticardiolipin antibodies, 3 patients had anticardiolipin antibodies, and 1 patient was heterozygous for factor V Leiden mutation. Five patients with 1 thrombosis were tested for thrombophilia; none had anticardiolipin antibodies, 1 had protein C deficiency and 1 had a lupus anticoagulant.

There was no significant difference in median platelet count (557 versus 482, p=0.42), median normalized hemoglobin (1.03 versus 0.98, p=0.27) or median normalized hematocrit (1.03 versus 1.00, p=0.56) in patients with recurrent thrombosis compared with single events.

Discussion

This retrospective cohort study provides a comprehensive overview of the clinical characteristics, timing, and treatment of thrombosis in POEMS syndrome, and identifies novel risk factors for thrombosis in POEMS patients. We report a thrombosis rate of 27% in patients with POEMS, approximating previous estimates of 20% and 30%.^{7, 9} Ischemic stroke occurred in 7% of patients, compared with published rates of 8% and 9%.^{6, 8} Consistent with previous reports, most thromboses occurred prior to initiation of therapy, arterial events were more common than venous events, and several patients developed thrombosis despite anticoagulation.^6–8 We confirm that high platelet count is a risk factor for arterial thrombosis, ^{5, 6} but identify new risk factors for arterial thrombosis including elevated hemoglobin/hematocrit (especially in men), splenomegaly, and extravascular volume overload. An additional novel finding was that hyperprolactinemia appears to be a risk factor for venous thrombosis and there was a trend for thrombocytosis to be a risk factor for venous thrombosis.

The pathophysiologic relationship between thrombocytosis, polycythemia, male sex and thrombosis in POEMS syndrome also remains a mystery. In POEMS syndrome, high VEGF levels, produced in part by platelets, are associated with increased VEGF expression in endoneurial vessels and endothelial damage on light microscopy, but there is no evidence of a significant correlation between serum VEGF and platelet count.¹³ Of interest, thrombocytosis and thrombosis feature prominently in myeloproliferative neoplasms; however, among patients with essential thrombocytosis and polycythemia vera, there is no clear association between thrombocytosis and thrombotic risk.^14–16 Why these relationships were more striking in men is not explained, but could be due to the small sample size of women patients in our cohort.

Most thromboses occurred prior to POEMS therapy, and most events that occurred during initial therapy happened within 3 months of diagnosis. Thrombosis during relapse or remission was less common. Moreover, most first recurrent thromboses also occurred prior to instituting POEMS directed therapy. This reinforces previous findings that thrombosis in POEMS syndrome typically occurs during the active phase of disease.⁷ Unlike Feng and colleagues, ⁸ we found that anticardiolipin antibodies were present in patients with recurrent thrombosis, but not in patients with one event. We also could not confirm their observation that patients with recurrent stroke had higher baseline platelet counts.

Although this cross-sectional retrospective study cannot definitively assess the risk of IMiD-associated thrombosis in POEMS syndrome, several points should be noted. First, patients with POEMS syndrome are at high risk for thrombosis. Second, the association of IMiD therapy with thrombosis in MM is well documented in prospective clinical trials.¹⁹ Third, although 3 of the 5 patients who had their initial or first recurrent thrombosis while on anticoagulation were receiving IMiD therapy, and 2 of the 12 patients who developed their initial or first recurrent thrombosis on antiplatelet therapy were taking IMiDs, rates of thrombocytosis and polycythemia (newly identified risk factors) were high in these IMID treated patients. All this taken together, along with the knowledge that IMIDs are highly effective in patients with POEMS syndrome,¹⁸ suggests that IMiD therapy should not be discarded as a treatment option for most patients, but that alternatives be considered in those POEMS patients with elevated thrombotic risk, including those with baseline thrombocytosis, polycythemia, or a history of thrombosis.

In summary, we demonstrate that thrombocytosis, high hemoglobin/hematocrit, extravascular volume overload, and splenomegaly are all risk factors for arterial thrombosis in patients with POEMS syndrome. The rates of venous thrombosis are higher in patients with POEMS than expected in a population of normal individuals, and the only significant risk factor among that we identified was a high prolactin level.²⁰ Clinicians should consider antiplatelet prophylaxis at a minimum for all patients at diagnosis given high overall rates of thrombosis. Escalation to full anticoagulation should be weighed against fall and bleeding risk for patients with baseline polycythemia, thrombocytosis, splenomegaly, effusions and/or a prior history of thrombosis, particularly in patients who have not yet received POEMS therapy.