The natural history of intravascular lymphomatosis

Abstract

Intravascular lymphomatosis (IVL) is a rare and clinically devastating form of extranodal B-cell non-Hodgkin's lymphoma. We performed a comprehensive analysis of the literature on IVL's published between 1959 and 2011 and evaluated the natural history as well as identified prognostic and predictive factors in patients. Nonparametric two-tailed Mann-Whitney U-test and Mantel-Cox log rank test were used to evaluate the survival intervals and prognostic factors. Multivariate analysis of variance (MANOVA) and chi-squared statistics were carried out to examine treatment-related predictive factors. Of the 740 patients with IVL, 651 (88%) had a diagnosis of B-cell lymphoma, 45 (6%) with T-cell lymphoma, and 12 patients (2%) with NK cell lymphoma. Central nervous system (CNS) IVL had the highest proportion of postmortem diagnosis, 250 (60%) compared to 21 (8%) of skin, 28 (11%) of bone marrow (BM) and spleen, and 17 (7%) of lung IVL's. Age <70 years (P = 0.0073), non-CNS site of initial diagnosis (P = 0.0014), lactate dehydrogenase (LDH) <700 (P = 0.0112), and rituximab treatment (P < 0.0001) were favorable prognostic factors. Gender, ethnicity, hemoglobin, BM biopsy, and the type of imaging studies used were not significant. Rituximab and doxorubicin treatment worked significantly better in patients with age >71 and LDH >577 compared to nonrituximab, nondoxorubicin regimens (MANOVA 2 degrees of freedom, P = 0.0345), with a median time from treatment to death of 20.0 (95% confidence interval [CI] 14.0-N/A, n = 14) months versus 2.0 (95%CI 0.5-N/A, n = 5) (χ(2) = 4.7, P = 0.0304). Patients with CNS IVL relapsed primarily in the CNS (88%) while same-organ relapse occurred less frequently in skin (23%), BM and spleen (50%) and lung (20%) IVL's. Our results indicate that IVL is primarily a disease of B-lymphoma cells. Timely diagnosis and treatment with rituximab-based chemotherapy improve patient survival. The pattern of recurrence is different between CNS IVL and IVL's in other organs.

Keywords: Intravascular lymphomatosis; natural history; non-Hodgkin's lymphoma.

Figure 1

Flow diagram of inclusion criteria. A systematic search was performed using Medline/PubMed, Paperchase and Web of Science by applying keywords of intravascular lymphoma, intravascular lymphomatosis, angioendotheliomatosis, angiotropic/angiotrophic large cell lymphoma, endotheliomatosis, and systemic endotheliomatosis. After accounting for overlap publications between keyword strings, articles were further examined and excluded when they were (1) not written in English, (2) not concerning human subjects, or (3) unrelated to the clinical evaluation of IVL. Data were extracted from the remaining publications and only those containing sufficient information were included in the statistical analysis. IVL, intravascular lymphomatosis.

Figure 2

Distribution of surface markers in IVL (n = 740). Positive immunohistochemical staining include 523 (71%) CD20, 173 (23%) CD45, 172 (23%) CD79, 63 (9%) CD5, 42 (6%) CD3, 22 (3%) CD19, 17 (2%) CD22, 11 (1%) CD56, and 11 (1%) CD44. Only 40 (5%) had the T-cell marker CD45RO. IVL, intravascular lymphomatosis.

Figure 3

Patient distribution at initial diagnosis and at recurrence. (A) Among the newly diagnosed IVL patients, 308 (41%) presented in the CNS, 145 (20%) in skin, 124 (17%) in BM and spleen, 50 (7%) in lungs, 101 (13%) in other sites and 12 (2%) unknown. Among the 250 patients diagnosed at postmortem, IVL affecting the CNS was highest compared to other organ sites: 151 (60%) CNS, 21 (8%) skin, 28 (11%) BM and spleen, and 17 (7%) lung and 33 (13%) other organs. (B) Same-organ recurrence was highest in 14 of 16 (88%) CNS, as compared to 5 of 22 (23%) skin, 5 of 10 (50%) BM and spleen, and 1 of 5 (20%) lung IVL's. BM, bone marrow; CNS, central nervous system; IVL, intravascular lymphomatosis.

Figure 4

Kaplan–Meier survival curves according to age. Four epochs were examined according to time from (A) initial presentation of symptoms to death, (B) time from diagnosis to death, (C) time from treatment to death, and (D) time from treatment to first recurrence.

Figure 5

Kaplan–Meier survival curves for CNS versus non-CNS IVL patients. Patients with CNS IVL had significantly shorter survival than non-CNS patients in (A) time from initial presentation of symptoms to death, (C) time from diagnosis to death, and (F) time from treatment to death, but not (G) time from treatment to first recurrence. (B) The median survival of CNS IVL patients diagnosed postmortem was significantly shorter than those diagnosed premortem, 3.5 (range 0.1–30.0, n = 150) versus 17.0 (range 0.5–71.1, n = 98) months, respectively (P = 0.0001). (D) The difference between CNS and non-CNS IVL remains significant in each successive decade for median time from initial presentation of symptoms to death in the 1960–1980s (CNS versus non-CNS: 4.1 [range 1.0–24.0, n = 35] versus 12.0 [range 0.3–33.0, n = 23] months [P = 0.0005]), 1990s (CNS versus non-CNS: 6.0 [range 0.1–71.0, n = 91] versus 12.0 [range 0.4–87.0, n = 75] months [P = 0.0002]), and 2000s (CNS versus non-CNS: 9.0 [range 0.1–58.0, n = 122] versus 14.0 [range 0.1–110.0, n = 206] months [P = 0.0087]). For CNS IVL, the difference was also significant between decades 1960–1980s and 1990s (P = 0.0507) and 1960–1990s and 2000s (P = 0.0009). (E) For median time from diagnosis to death, the difference between CNS and non-CNS IVL remains significant in each successive decade in the 1960–1980s (CNS versus non-CNS: 6.0 [range 0.9–14.0, n = 6] versus 15.0 [range 0.5–42.0, n = 14] months [P = 0.0184]) and 1990s (CNS versus non-CNS: 8.0 [range 0.1–71.0, n = 43] versus 17.0 [range 0.1–96.0, n = 58] months [P = 0.0279]), and near significance in 2000s 18.0 [range 0.3–84.0, n = 98] versus 20.0 [range 0.1–165.0, n = 221] months [P = 0.0515]. For CNS IVL, the difference was only significant between decades 1960–1990s versus 2000s (P = 0.0462). (H) For time from treatment to death, only the decade 2000s was significant for CNS versus non-CNS IVL: 20.0 (range 0.3–84.0, n = 82) versus 30.0 (range 0.1–165.0, n = 171) months (P = 0.0196) but not significant for CNS IVL's in successive decades. (I) For time from treatment to first recurrence, no statistical difference was found in survival in CNS versus non-CNS, as well as CNS IVL's, in each successive decades. IVL, intravascular lymphomatosis; CNS, central nervous system.

Figure 6

Kaplan–Meier survival curves for CNS versus skin IVL patients. Compared to skin IVL, CNS IVL had a significantly shorter (A) median time from initial presentation of symptoms to death (18.0 [range 1.0–110.0, n = 112] versus 6.0 [range 0.1–71.1, n = 248] months, respectively [P = 0.0001]), (B) median time from diagnosis to death (23.0 [range 0.1–96.0, n = 110] versus 14.0 [range 0.1–84.0, n = 141] months, respectively [P = 0.0064]), and (C) median time from treatment to death (24.0 [range 0.7–96.0, n = 94] versus 16.0 [0.3–84.0, n = 117] months, respectively [P = 0.0238]). There was no difference between skin and CNS IVL patients with respect to (D) median time from treatment to first. recurrence (6.0 [range 1.0–33.0, n = 21] versus 8.0 [1.0–44.0, n = 16] months, respectively [P = 0.6375]). IVL, intravascular lymphomatosis; CNS, central nervous system.

Figure 7

Kaplan–Meier survival curves for patients with IVL in the CNS versus BM and spleen. Compared to IVL in the BM and spleen, CNS IVL also had a significantly shorter (A) median time from initial presentation of symptoms to death (11.0 [range 0.5–92.0, n = 80] versus 6.0 [range 0.1–71.1, n = 248] months, respectively [P = 0.0064]), (B) median time from diagnosis to death (17.0 [range 0.1–148.0, n = 92] versus 14.0 [range 0.1–84.0, n = 141] months, respectively [P = 0.0176]), (C) median time from treatment to death (24.7 [range 0.1–148.0, n = 65] versus 16.0 [0.3–84.0, n = 117] months, respectively [P = 0.0072]), and (D) median time from treatment to first recurrence (4.5 [range 0.3–18.0, n = 12] versus 8.0 [1.0–44.0, n = 16] months, respectively [P = 0.0430]). IVL, intravascular lymphomatosis; CNS, central nervous system; BM, bone marrow.

Figure 8

Kaplan–Meier survival curves for skin versus nonskin IVL patients. Patients with skin IVL had significantly better survival than nonskin patients in (A) time from initial presentation of symptoms to death but not (C) time from diagnosis to death, (E) time from treatment to death, or (G) time from treatment to first recurrence. (B) The difference between skin and nonskin IVL remains significant in each successive decade for median time from initial presentation of symptoms to death in the 1960–1980s (skin versus nonskin: 15.0 [range 5.0–31.7, n = 14] versus 4.0 [range 0.3–33.0, n = 44] months [P = 0.0001]), 1990s (skin versus nonskin: 12.0 [range 1.7–69.0, n = 32] versus 6.2 [range 0.1–87.0, n = 134] months [P = 0.0211]), and 2000s (skin versus nonskin: 20.0 [range 1.0-110.0, n = 66] versus 10.0 [range 0.1–92.0, n = 262] months [P = 0.0007]). For skin IVL, the difference was not significant between decades 1960–1980s and 1990s (P = 0.3848) or 1960–1990s and 2000s (P = 0.1158). (D) For median time from diagnosis to death in the 1960–1980s, there was significant difference between skin and nonskin IVL: 19.9 (range 2.1–42.0, n = 10) versus 10.0 (range 0.5–27.0, n = 10) months (P = 0.0445). No difference was found in 1990s (P = 0.3287) and 2000s (P = 0.2946). Comparing skin IVL between decades, the difference was not significant between decades 1960–1980s and 1990s (P = 0.9185) or 1960–1990s and 2000s (P = 0.6318). (F) For median time from treatment to death, there was no difference between skin and nonskin IVL in the 1960–1980s (P = 0.5492), 1990s (P = 0.3695), and 2000s (P = 0.5190). Comparing skin IVL between decades, there is also no difference between decades 1960-1980s and 1990s (P = 0.8617) or 1960–1990s and 2000s (P = 0.3601). (H) For median time from treatment to first recurrence, there is no difference between skin and nonskin IVL in the 1960-1980s (P = 0.5151), 1990s (P = 0.4654), and 2000s (P = 0.7491). Comparing skin IVL between decades, there is also no difference between decades 1960-1980s and 1990s (P = 0.6912) or 1960-1990s and 2000s (P = 0.3191). IVL, intravascular lymphomatosis.

Figure 9

Kaplan–Meier survival curves for BM and spleen versus non-BM and spleen IVL patients. No statistical difference in patient survival between BM and spleen IVL and non-BM and spleen IVL with respective to (A) time from initial presentation of symptoms to death, (C) time from diagnosis to death, (E) time from treatment to death or (G) time from treatment to first recurrence. The difference is also not significant between the two groups in the 1990s and 2000s, as well as BM and spleen IVL between 1990s and 2000s, for (B) median time from initial presentation of symptoms to death, (D) median time from diagnosis to death, (F) median time from treatment to death, and (H) median time from treatment to first recurrence. IVL, intravascular lymphomatosis; BM, bone marrow.

Kaplan–Meier survival curves for lung versus nonlung IVL patients. No statistical difference in patient survival between lung IVL and non-lung IVL with respective to (A) time from initial presentation of symptoms to death, (C) time from diagnosis to death, (E) time from treatment to death or (G) time from treatment to first recurrence. The difference is also not significant between the two groups in the 1990s and 2000s, as well as lung IVL between 1990s and 2000s, for (B) median time from initial presentation of symptoms to death, (D) median time from diagnosis to death, (E) median time from treatment to death, and (H) median time from treatment to first recurrence. IVL, intravascular lymphomatosis.

Figure 11

Predictive factors for response to rituximab and doxorubicin. Three-dimensional plot of the MANOVA analysis showing a cut off of age at 71 years and LDH at 577 had the greatest influence on the time from treatment to death (two degrees of freedom, P = 0.0345). LDH, lactate dehydrogenase; MANOVA, multivariate analysis of variance.

Figure 12

Kaplan–Meier plot of time from treatment to death. (A) Patients with LDH > 577 treated with rituxmab and doxorubicin had a significant prolongation of time from treatment to death, median not reached (95%CI 24.0–N/A, n = 14) months, as compared to nonrituximab nondoxorubicin regimens, median 4.9 (95%CI 2.0–N/A, n = 54) months (χ² = 21.4, P < 0.0001). (B) The benefit of rituximab remained significant when the age effect was added into the analysis. In patients with age > 71 and LDH > 577, there was still a significant difference between rituximab and doxorubicin and nonrituximab nondoxorubicin treatments, median 20.0 (95%CI 14.0–N/A, n = 14) months versus 2.0 (95%CI 0.5–N/A, n = 5) months (χ² = 4.7, P = 0.0304). LDH, lactate dehydrogenase.