J Neurooncol (2014) 120:597–605 DOI 10.1007/s11060-014-1592-1
CLINICAL STUDY
Leptomeningeal dissemination in glioblastoma; an inspection of risk factors, treatment, and outcomes at a single institution Jacob J. Mandel • Shlomit Yust-Katz • David Cachia • Jimin Wu • Diane Liu • John F. de Groot • Alfred W.K. Yung • Mark R. Gilbert
Received: 4 July 2014 / Accepted: 17 August 2014 / Published online: 29 August 2014 Ó Springer Science+Business Media New York 2014
Abstract There are few studies reporting the incidence of leptomeningeal dissemination (LMD) in patients with glioblastoma; only small case series have been published. Consequently, there are no established standards of care for these patients. Therefore, we undertook this retrospective review to evaluate a large series of patients with glioblastoma treated at MD Anderson Cancer Center to estimate the incidence of LMD and assess the impact of a variety of treatment modalities. Analysis was performed on 595 patients with glioblastoma treated on clinical trials from 2006 to 2012. The diagnosis of LMD was made by imaging or positive cerebrospinal fluid cytology in 24 patients. An additional 12 patients with known LMD diagnosed during this same period were included to evaluate the impact of treatment on outcome for a total of 36 patients. LMD developed in 4.0 % (24/595 patients) of the clinical trial
Jacob J. Mandel and Shlomit Yust-Katz have contributed equally to the manuscript. J. J. Mandel (&) Department of Neurology, Baylor College of Medicine, One Baylor Plaza, MS Nb302, Houston, TX 77030, USA e-mail: [email protected] S. Yust-Katz Rabin Medical Center, Jabotinski Street 39, 49100 Petah-Tikva, Israel D. Cachia J. F. de Groot A. W.K. Yung M. R. Gilbert Division of Neuro-Oncology, The University of Texas MD Anderson Cancer Center, 1515 Holcombe Blvd Unit 431, Houston, TX 77030-4009, USA J. Wu D. Liu Quantitative Sciences - Unit 1409, The University of Texas MD Anderson Cancer Center, P. O. Box 301402, Houston, TX 77230-1402, USA
cohort. Median survival from glioblastoma diagnosis was 16.0 months. Estimated median time of glioblastoma diagnosis to LMD was 11.9 months. Median overall survival from the time of LMD diagnosis was 3.5 months. Patients treated for LMD with chemotherapy/targeted therapy and radiation had a significantly prolonged survival (7.7 months) compared to chemotherapy/targeted therapy alone, radiation alone or palliative care. LMD remains an uncommon event in patients with glioblastoma. Patients treated aggressively with chemotherapy/targeted therapy and radiation had the longest median survival following diagnosis of LMD. However, patients receiving chemotherapy/targeted therapy and radiation were younger and this may have influenced survival. Given the overall poor outcomes, improved therapeutic approaches are needed for glioblastoma patients with LMD. Keywords Glioblastoma Leptomeningeal dissemination Leptomeningeal gliomatosis
Introduction Glioblastoma is the most common primary brain tumor in adults. Despite advances in surgical techniques, radiation treatments and chemotherapy, it continues to have a dismal prognosis [1]. Surprisingly, given the location of glioblastoma in the central nervous system and relatively high exposure of the ventricular system during tumor resection, the incidence of leptomeningeal dissemination (LMD) is thought to be uncommon. However, data regarding the true incidence of LMD in patients with glioblastoma is scarce with only a few case reports and small case series published in the past. These reports provide a wide range of incidence of LMD; up to 10–21 % on autopsy series [2],
123
598
but symptomatic dissemination is much lower at around 2 % [3]. Earlier studies have shown that survival is short after diagnosis of LMD, in the range of only 12–20 weeks [4]. No prospective clinical trials have been reported and there are no established standards of care for patients with LMD from glioblastoma. Many treatments have been used including radiation and a variety of chemotherapies including intrathecal chemotherapy with Methotrexate, araC, thio-TEPA, or ACNU with little success [5]. Recent advances in surgical techniques leading to more extensive tumor resections and the widespread use of agents such as temozolomide in conjunction with more refined radiotherapy techniques, have resulted in an overall increase in median survival with glioblastoma, and most significantly, an increase in the 5-year survival rate [6]. This increase in long-term survival has led to speculation that the incidence of LMD would increase, as has been reported in other cancers [3]. There is further speculation that, in addition to the recent increases in survival with the widespread use of temozolomide, other treatments such as anti-angiogenic agents may increase the incidence of LMD as a combined result of prolonged survival and changes in tumor biology to a more invasive and infiltrative phenotype [7, 8]. In addition to the potential impact on the incidence of LMD, these agents may alter outcomes once LMD has been diagnosed. The goal of this study was to estimate the incidence of LMD in glioblastoma and to examine the impact of treatments on survival in this patient group.
Methods Patients After obtaining institutional board review approval (protocol PA13-0216), we retrospectively identified patients with glioblastoma who developed LMD from our institutional database. Initially, we reviewed all 595 glioblastoma patients enrolled on clinical trials at MD Anderson during the years 2006–2012 because this group had the most extensive and complete follow up data. We later identified an additional 12 patients with known LMD diagnosed during this same period that were included to evaluate the impact of treatment on outcome. All of the patients had a pathologically-confirmed diagnosis of glioblastoma using WHO criteria. The diagnosis of LMD was determined by either radiographic findings on magnetic resonance imaging (MRI) or cerebrospinal fluid (CSF) cytology containing malignant cells. The radiographic criteria used to diagnose LMD were leptomeningeal contrast enhancement around the contours
123
J Neurooncol (2014) 120:597–605
of the gyri and sulci or multiple nodular deposits in the subarachnoid space, particularly focusing on the cerebellar folia, the cortical surface, and the basal cisterns. In the spinal cord, linear or nodular enhancement along the surface of the cord and linear or nodular enhancement of the cauda equina were deemed conclusive of leptomeningeal spread [9]. Leptomeningeal dissemination on imaging was confirmed by independent Radiology reports. Data collected included demographic characteristics (age and sex), tumor characteristics (primary or secondary glioblastoma, tumor location, distance from the ventricles), treatment (extent of resection as determined by imaging or operative reports if imaging was not available, whether the ventricles were opened during surgery, number of resections prior to LMD, time from last surgery to LMD, number of progressions prior to development of LMD, type of radiation and chemotherapy/targeted therapy received prior and after diagnosis of LMD), diagnostic results (MRI imaging and CSF studies including cytology), survival from glioblastoma diagnosis, time from glioblastoma diagnosis to LMD diagnosis and survival from the time of LMD diagnosis. Statistical analysis Data was first summarized using standard descriptive statistics for demographic and clinical variables. The survival, time from initial glioblastoma diagnosis to death, time from glioblastoma diagnosis to LMD diagnosis and time from LMD diagnosis to death, were evaluated using the Kaplan–Meier method and the comparison between or among patients’ characteristics was assessed using log-rank test [10]. Both univariate and multivariate Cox regression models were applied to assess the effect of covariates of interest on three time-to-event endpoints. All computations were carried out in SAS 9.3 (SAS Institute Inc., Cary, NC, USA).
Results A total of 24 patients out of the 595 patients from the clinical trial cohort were identified to have developed LMD; yielding a rate of 4.0 % of patients for this group. Targeted retrospective chart review of patients with known LMD added another 12 patients for a total of 36 patients with LMD. Patient’s characteristics before LMD diagnosis Among the 36 patients included in the analysis, 24 (66.6 %) were men and 12 were women, with a median age of 44 years (range 21–71). Primary tumor location was
J Neurooncol (2014) 120:597–605
599
Table 1 Clinical characteristics of the patients at time of glioblastoma diagnosis
Table 2 Patients characteristics at time of LMD diagnosis Variable
N (%)
Patient
36(100 %)
Variable
N (%)
Patient
36(100 %)
Age at LMD diagnosis: Median (Min–Max)
45(21–73)
Age at glioblastoma diagnosis: Median (Min–Max)
44(21-71)
Time from GB diagnosis to development of LMD (Months) Median (Min–Max))
11.9(0–62)
Male
24(66.7 %)
KPS at LMD diagnosis
Female
12(33.3 %)
Gender
KPS 70 or less 80 or greater
2(8.3 %) 22(91.7 %)
Unknown*
12
Tumor location
70 or less
11(34.4 %)
80 or higher
21(65.6 %)
Unknown* Number of resections prior to LMD 1 or less 2 or more
4 20(55.6 %) 16(44.4 %)
Parietal
7(19.4 %)
Frontal
11(30.5 %)
Time from last surgery to development of LMD (Months) Median (Min–Max)
Temporal
11(30.5 %)
Common presenting symptoms #
Midbrain
2(5.6 %)
None
6(16.7 %)
Cerebellar
1(2.8 %)
Headache
11(30.6 %)
Thalamus
1(2.8 %)
Ataxia
10(27.8 %)
Spine
2(5.6 %)
Pain
7(19.4 %)
1(2.8 %)
Visual changes
7(19.4 %)
Supratentorial
31(86.1 %)
Seizure Lower extremity weakness
6(16.7 %) 5(13.9 %)
Infratentorial
5(13.9 %)
Hypothalamus Glioblastoma location at diagnosis
Extent of resection of glioblastoma at diagnosis Biopsy
Brain MRI characteristics
8.04(0.12–61.56)
21(58.3 %)
Nodules 6(16.7 %)
Subarachnoid
11(30.5 %)
Subtotal
10(27.8 %)
Ventricular
10(27.8 %)
Gross total
20(55.6 %)
Communicating with the ventricle at time of glioblastoma diagnosis Yes No Spinal cord
17(50.0 %)
12(33.3 %)
Focal
7(19.4 %)
Diffuse
5(13.9 %)
17(50.0 %)
No evidence of LMD
2
Ependymal enhancement
23(63.9 %)
Sulci enhancement
3(8.3 %)
Folia enhancement
7(19.4 %)
Concurrent radiation ? chemo with temozolomide after glioblastoma diagnosis Yes
32(91.4 %)
No
1(2.9 %)
Radiation prior to glioblastoma diagnosis**
2(5.7 %)
Unknown* Adjuvant temozolomide for glioblastoma
Pial enhancement
1
Cranial nerve enhancement Spinal MRI characteristics
3(8.3 %)
4(11.1 %) 10(47.6 %)
Nodules Subarachnoid Ventricular
8(38.1 %) 2(9.5 %)
Pial enhancement
10(47.6 %)
Yes
32(91.4 %)
No
1(2.9 %)
Focal
5(23.9 %)
Temozolomide prior to glioblastoma diagnosis**
2(5.7 %)
Diffuse
5(23.8 %)
Unknown*
1
* Patients with ‘‘Unknown’’ status are excluded from proportion calculation ** for patients with secondary glioblastoma
predominantly supratentorial (31, 86 %) and the remainder (5) were infratentorial. Seventeen (50 %) of the tumors had direct communication with a ventricle of which 9 (26 %) had the ventricles opened during initial surgery.
No evidence of LMD
1(4.8 %)
Nerve root enhancement
12(57.1 %)
Spinal imaging showing LMD Positive
20(95.2 %)
Negative
1(4.8 %)
Unknown*
15
Spinal imaging showing LMD with conus/cauda equina affected No
13(61.9 %)
Yes
8(38.1 %)
123
600
J Neurooncol (2014) 120:597–605
Table 2 continued Variable Unknown*
Table 3 Descriptive summary N (%)
Variable
15(n/a %)
Treatment after LMD diagnosis
Cerebral spinal fluid cytology for malignant cells
N (%) 36
Hospice
5(14.7 %)
Positive Negative
5(45.5 %) 6(54.5 %)
Radiation alone
2(5.9 %)
Chemo alone
16(47.1 %)
Unknown*
25
Chemo ? radiation
10(29.4 %)
Resection
1(2.9 %)
Unknown
2
Bevacizumab use prior to LMD diagnosis No
24(66.7 %)
Yes
12(33.3 %)
Time from glioblastoma diagnosis to start of Bevacizumab
11.64(1.68–25.9)
(patients treated prior to LMD) (Months) Median (Min–Max) * Patients with ‘‘Unknown’’ status are excluded from proportion calculation # Patients may have more than one symptom
The patterns of treatment after glioblastoma diagnosis and patients characteristics before LMD diagnosis are detailed in Table 1. Patient’s characteristics at time of LMD diagnosis Patients presented with a variety of symptoms (Table 2). Spinal cord or posterior fossa symptoms such as ataxia, lower extremity weakness, and parasthesias were present in 47 %. The most common single presenting symptom of LMD was headaches, reported in 11 patients (31 %). 36 % (4/11) of patients with headaches were found to have hydrocephalus. Six patients (16 %) were asymptomatic at diagnosis that was made by imaging findings. At the time of LMD diagnosis, three patients (8.3 %) had stable primary disease. Patients had a median KPS of 80 (range 60–100) and a median of 1 recurrence (range 0–5). Sixteen patients (44 %) had 2 or more tumor resections prior to the development of LMD. Six out of these 16 patients with multiple surgeries had their ventricles opened on initial resection. 30 % (3/10) of patients without ventricular wall compromise on initial resection had a subsequent resection with opening of the ventricles. The median time from last surgery to diagnosis of LMD was 8.0 months (range, 0.1–61.6). Three patients (8.3 %) were diagnosed with LMD at the time of initial diagnosis of GBM before a surgical procedure and one patient (3 %) was diagnosed with LMD during the week following their resection. Of the three patients diagnosed with LMD prior to a surgical procedure, two had a brainstem location of their primary glioblastoma and the other had a hypothalamic lesion. All three underwent biopsies for diagnostic purpose that revealed the diagnosis of glioblastoma.
123
Type of radiation alone Focal palliative Type of chemo alone Intravenous chemo Bevacizumab containing regimen Type of radiation for Chemo ? radiation
2 2(100 %) 16 15(94 %) 13(81 %) 10
Craniospinal
4(40 %)
Focal brain and spine
3(30 %)
Focal brain
2(20 %)
Whole spine
1(10 %)
Spinal MR imaging was obtained in 21 patients (58 %) and leptomeningeal spread was observed in 20 patients (95 %). Leptomeningeal tumor deposits were found in the conus/cauda equina in eight patients (38 %) who had spinal imaging. However, only two of these patients displayed components of the classical symptoms of cauda equina syndrome (back pain, saddle anesthesia, lower extremity weakness/sensory loss, and loss of ankle reflexes). CSF analysis was performed in 11 patients (31 %) and cytologic examination was positive for malignant cells in five patients (45 %); four with the first CSF study. All of the patients with negative CSF studies underwent only a single lumbar puncture. Every patient with positive CSF also had leptomeningeal disease seen on spinal MR imaging, excluding one patient who did not obtain spinal imaging. IDH-1 status was obtained on seven patients (19 % of the total) and was mutated in two patients (29 % of those tested). P53 staining was obtained in seven patients (19 % of the total) and was positive in 4(57 % of those tested). Ki-67 % was recorded in four patients, three at time of glioblastoma diagnosis (19, 27, 35 %) and one at first tumor recurrence (14.7 %). Four tumors were noted to be pathologic variants with three gliosarcoma and one a giant cell variant. Treatment varied widely among the cohort of patients with LMD. As detailed in Table 3, 28 patients received either radiation, chemotherapy/targeted therapy or both chemotherapy/targeted therapy and radiation, one patient (3 %) had a resection of tumor in the cerebellar vermis that
J Neurooncol (2014) 120:597–605
601
was symptomatic. Five patients (14 %) were referred to hospice without further therapy and two patients (6 %) were lost to follow up.
Survival analysis At the time of this review, 30 patients (83 %) were deceased. The estimated overall median survival for all patients from glioblastoma diagnosis was 16.0 months (range 4–69). Median survival time from diagnosis of glioblastoma to LMD was 11.9 months (range 0–62). Median survival from the diagnosis of LMD was 3.5 months. Median survival from the diagnosis of LMD was 3.4 months for the clinical trial cohort patients and 4.2 months for non-trial patients (p = 0.38). Survival from glioblastoma diagnosis Univariate analysis found that longer overall survival from initial diagnosis of glioblastoma was found to be associated to prior treatment with bevacizumab (HR 2.63; 95 % CI 1.16 to 5.99; p = 0.0212) and tumors not communicating with the ventricles (HR 2.32; 95 % CI 1.05–5.13; p = 0.0371). Bevacizumab treatment (HR 3.40; 95 % CI 1.41–8.19; p = 0.0065) and tumors not communicating with the ventricles (HR 3.24; 95 % CI 1.36–7.75; p = 0.0082) remained significant on multivariate analysis. Patients with tumors directly communicating with the ventricles had a median overall survival of 14.8 months that compared unfavorably to non-communicating tumors who had an overall median survival of 23.2 months (p = 0.032). Patients who were diagnosed with LMD at time of diagnosis of glioblastoma had a worse prognosis compared to those who were diagnosed afterwards (estimated overall survival of 4.7 months compared to 16.0 months) although given the small numbers of patients, the difference was not statistically significant. Time to LMD diagnosis On univariate analysis the following factors were found to be statistically significantly associated with a shorter time from initial glioblastoma diagnosis to the finding of LMD: infratentorial tumor location at diagnosis, tumor communication with the ventricle at diagnosis and no bevacizumab use prior to LMD diagnosis. However, only tumor communication with the ventricle at diagnosis (HR 3.14; 95 % CI 1.35–7.26; p = 0.0076) and no bevacizumab use prior to LMD diagnosis (HR 2.97; 95 % CI 1.30–6.79; p = 0.0099) were significant on multivariate analysis.
Fig. 1 Kaplan Meier survival curve from LMD diagnosis according to treatment
Table 4 Univariate cox regression models-from LMD to death according to treatment Variable
Hazard ratio
95 % CI
p value
Lower CI
Upper CI
Treatment after LMD diagnosis \0.0001
Hospice
47.39
9.03
248.67
Radiation alone
14.27
2.22
91.84
0.0051
Chemo alone
3.87
1.25
12.05
0.0194
Chemo ? radiation
1.00
Infratentorial tumors were found to develop LMD at a shorter period of time compared to supratentorial tumors (4.9 months compared to 11.9 months, p = 0.003). Also, tumors communicating with the ventricles developed LMD quicker (11.9 months vs. 20.3 months, p = 0.037) than tumors not communicating with the ventricles. Survival from LMD diagnosis On univariate analysis, patients treated with a combination of chemotherapy/targeted therapy and radiation had a significantly prolonged survival compared to chemotherapy/ targeted therapy alone, radiation alone or palliative care (Fig. 1; Table 4). The patients that received radiation and chemotherapy/targeted therapy after LMD diagnosis did not have a better KPS prior to treatment compared to the other groups (p = 0.85). Patients receiving radiation and
123
602
J Neurooncol (2014) 120:597–605
Table 5 Comparison of patient’s characteristics Covariate
Levels
Total
LMD after glioblastoma diagnosis No LMD
In 2 years
198 (93 %)
15 (7 %)
p value
N (%) Patient
213 (100 %)
Gender
Female
89 (41.8 %)
86 (43.4 %)
KPS at glioblastoma diagnosis
Male 100
3 (20 %)
0.1034
124 (58.2 %) 83 (39 %)
112 (56.6 %) 78 (39.4 %)
12 (80 %) 5 (33.3 %)
– 0.0461
90
79 (37.1 %)
75 (37.9 %)
4 (26.7 %)
80
34 (16 %)
30 (15.2 %)
4 (26.7 %)
–
70
13 (6.1 %)
13 (6.6 %)
0 (0 %)
–
2 (13.3 %)
–
60 Location of glioblastoma
Extent of glioblastoma resection
Number of surgeries
Communicating with ventricle at diagnosis
Ventricle opened on initial resection
4 (1.9 %) 64 (30 %)
56 (28.3 %)
8 (53.3 %)
0.2427
Temporal
98 (46 %)
92 (46.5 %)
6 (40 %)
–
Parietal
38 (17.8 %)
37 (18.7 %)
1 (6.7 %)
–
Others
13 (6.1 %)
13 (6.6 %)
0 (0 %)
–
Unknown
18 (. %) 8 (53.4 %)
0.3533
–
GTR
112 (57.4 %)
104 (57.8 %)
STR
56 (28.7 %)
53 (29.4 %)
3 (20 %)
–
BX
27 (13.8 %)
23 (12.8 %)
4 (26.7 %)
–
1
91 (42.7 %)
84 (42.4 %)
7 (46.7 %)
0.2353
2 3
69 (32.4 %) 41 (19.2 %)
67 (33.8 %) 36 (18.2 %)
2 (13.3 %) 5 (33.3 %)
– –
4
12 (5.6 %)
11 (5.6 %)
1 (6.7 %)
–
Unknown
20 (. %)
–
No
135 (69.9 %)
125 (70.2 %)
10 (66.7 %)
Yes
58 (30.1 %)
53 (29.8 %)
5 (33.3 %)
Unknown
7 (. %) 184 (89.3 %)
172 (89.1 %)
12 (92.3 %)
Yes
22 (10.7 %)
21 (10.9 %)
1 (7.7 %)
Clinical cohort patients Of the 595 patients on clinical trials, seven patients (1.2 %) had infratentorial lesions (2 midbrain, 2 pons, 1 medullary and 2 cerebellar). One of the infratentorial (14 %) lesions in the clinical cohort developed LMD. 33 % (4/12) of the
0.7740 – –
No
chemotherapy/targeted therapy after LMD diagnosis were younger with a median age of 35.1 years compared to 47.3 years for patients sent to hospice, 52.6 years for patients who received radiation alone, and 45.3 years for patients receiving chemotherapy alone. Median survival for patients asymptomatic at the time of diagnosis of LMD was 4.6 months. Patients with large volume leptomeningeal disease (defined as nodules on MRI [ 5 9 10 mm in orthogonal diameter) did not have a significantly different overall survival than small volume disease (3.5 months vs. 3.8 months, p = 0.81).
123
2 (1 %)
Frontal
0.9999 –
non-clinical trial cases (2 medullary and 2 spinal cord) were infratentorial.
Comparison analysis An analysis was performed comparing the patients with glioblastoma who developed LMD to the patients that did not develop LMD in the clinical trials cohort group. Due to concern that overall survival could influence the development of LMD, a 2 year time cut off from diagnosis of glioblastoma was used to establish these groups. This allowed the majority of patients who developed LMD to be compared and the ratio of patients with LMD to those without it at this time was the most similar to the incidence of LMD among patients with glioblastoma found in our study. Fifteen of the 24 LMD patients in the clinical trial cohort were found to develop LMD within 2 years of the diagnosis of glioblastoma. In the clinical trial cohort, 198
J Neurooncol (2014) 120:597–605 Table 6 Comparison of continuous covariate
603
Covariate Age at glioblastoma diagnosis KPS at glioblastoma diagnosis Number of surgeries
out of the 593 patients who did not develop LMD survived more than 2 years after tumor diagnosis. Thus, the clinical and demographic characteristics of these 15 LMD patients were compared to the 198 patients that did not develop LMD over the first 2 years of survival (Tables 5, 6). Only KPS at initial diagnosis of glioblastoma was slightly lower in patients who developed LMD and was the only statistically significant difference between the two groups (p = 0.046). No comparisons of PFS and OS between the two groups were performed because the identification of the comparative group created a biased population as defined by the 2-year survival requirement.
Discussion This retrospective review of 36 patients with glioblastoma and LMD provides further insight into the prevalence, treatment patterns and outcome for this rare diagnosis. Our data verifies the low incidence (4.0 %) of LMD among patients with glioblastoma on clinical trials and identifies several novel findings influencing time to development of LMD. This study also corroborates the poor prognosis of patients with glioblastoma and LMD as previously described [4]. However, patients treated with radiation and chemotherapy/targeted therapy following LMD diagnosis had significantly prolonged survival compared to patients treated less aggressively, although treatment decisions may have been related to other prognostic factors. An analysis comparing the patients with glioblastoma who developed LMD to the patients that did not develop LMD in the clinical trials cohort was performed but did not reveal any clinical risk factors for the development of LMD. Previous studies have proposed that LMD occurs primarily in glioblastoma patients with a prolonged overall survival, thereby the consequence of improvements in treatment [3]. However, other studies have found the majority of patients developed LMD within 12 months of their glioblastoma diagnosis [4, 11]. Our study found that only 14/36 (36 %) of the patients were diagnosed with LMD more than a year after their glioblastoma diagnosis. This suggests that there may be innate tumor specific
LMD after glioblastoma diagnosis
N
Mean ± Std, Mean (Min, Max)
p value 0.1944
No LMD
198
49.84 ± 12.75, 52 (18, 84)
In 2 years
15
45.4 ± 13.52, 45 (24, 71)
–
No LMD
198
90.81 ± 9.47, 90 (60, 100)
0.2740
In 2 years
15
No LMD
198
In 2 years
15
86.67 ± 13.45, 90 (60, 100)
–
1.87 ± 0.9, 2 (1, 4)
0.7080
2 ± 1.07, 2 (1, 4)
molecular characteristics that make some tumors more successful in the leptomeninges and CSF. Although there has been no clearly established molecular profile associated with the development of LMD, prior studies have suggested that gains at the 1p36 chromosomal region are associated with symptomatic leptomeningeal spread of supratentorial glioblastomas [11]. Furthermore, mutations of the PTEN gene and a high KI-67 proliferative index have been associated with leptomeningeal dissemination of glioblastoma [12]. However, the comparison analysis of patients with glioblastoma who developed LMD to the patients that did not develop LMD in the clinical trials cohort did not reveal any clinical risk factors for the development of LMD. This suggests that the increase in tumor invasiveness and predilection to disseminate into the cerebrospinal fluid, causing LMD may be primarily related to molecular characteristics of the tumor. Although the overall propensity to develop LMD may be driven by molecular factors, this study determined additional factors that were associated with a shorter time frame from initial tumor diagnosis to LMD. Tumors communicating with a ventricle as well as tumors located in the cerebellum and spinal cord had an earlier development of LMD and had a poorer overall survival from initial glioblastoma diagnosis. There have also been concerns that the use of antiangiogenic agents such as bevacizumab may increase the rate of LMD because of reports that these agents may promote a more invasive tumor phenotype. Interestingly, patients who received bevacizumab prior to development of LMD were found to have a longer time to diagnosis of LMD and prolonged overall survival compared to patients who did not have bevacizumab before developing LMD. This would suggest that bevacizumab does not accelerate the development of LMD. Prior studies have shown a high specificity of markedly increased levels of CSF VEGF in LMD [13] . However, the overall improvement in outcome with bevacizumab may reflect either treatment efficacy or a patient selection bias. Of note, prior bevacizumab use did not appear to effect overall survival following LMD diagnosis. At present, there is no standard treatment for LMD secondary to glioblastoma. Despite new developments in
123
604
treatment, median survival from LMD diagnosis until death still remains dismal at 14 weeks, similar to previous studies [4, 14]. This is also consistent with survival from LMD secondary to other solid tumors which have median survival of 12 weeks [15]. A prior study suggested that patients with large volume leptomeningeal disease (defined as nodules on MRI [ 5 9 10 mm in orthogonal diameter) have a significantly worse overall survival than patients with small volume disease; however, this study excluded leptomeningeal disease from primary brain tumors [16]. Eighteen patients (50 %) in our study were found to have large volume disease, but survival was not significantly different. This contradictory finding may be due to the aggressive nature of brain tumors and the high percentage of patients with progressive primary disease at time of LMD diagnosis in our study. Patients treated most aggressively with chemotherapy/ targeted therapy and radiation had the longest median survival from LMD to death (30.7 weeks) compared to chemotherapy/targeted therapy alone (15.4 weeks) or radiation alone (9.2 weeks). Although selection bias cannot be conclusively excluded, overall the patients receiving chemotherapy/targeted therapy and radiation did not have a better KPS prior to treatment compared to the other groups. However, patients receiving radiation and chemotherapy/ targeted therapy after LMD diagnosis were younger and this likely influenced the overall survival. The role of intraCSF chemotherapy is unclear at this time. A prior trial using several intraventricular chemotherapies sequentially for LMD due to high grade glioma resulted in a median survival of only 3.5 months (similar to this review despite only two patients in this study receiving intraventricular chemotherapies) [14]. Moreover, only one patient was treated with craniospinal radiation and received intraventricular chemotherapy likely due to increased toxicity when using both therapies. Due to the heterogenous nature of the treatments and retrospective nature of this study, future randomized studies are necessary to determine if more aggressive treatment does in fact improve survival. Previous studies treating LMD have been challenging to interpret due to the absence of a uniform response criteria; however, a Response Assessment in Neuro-Oncology (RANO) group is reported to be working on establishing new response criteria to be validated in future clinical trials in LMD [17]. At this time treatment decisions should continue to be made on a case by case basis using sound clinical judgment. Additionally, clinical manifestations of LMD can overlie appreciably with those of primary brain tumor progression, metabolic disturbances, and treatment-related toxicities. It is important to note that this study may have excluded erroneously several patients with undiagnosed LMD who were asymptomatic or had symptoms from
123
J Neurooncol (2014) 120:597–605
LMD blamed on primary disease progression or other causes. Autopsy studies have demonstrated a higher rate of LMD and a spinal MRI should be obtained in patients with symptoms or clinical exam findings referring to the spinal cord [2]. Furthermore, patients with confirmed LMD should obtain complete neuraxis imaging for proper disease staging and treatment planning [18]. In conclusion, the overall incidence of LMD among glioblastoma patients remains low. Potential risk factors for earlier dissemination such as tumor location and proximity to the ventricular system have been identified in patients that develop LMD and there are preliminary suggestions that there are tumor molecular phenotypes that have a greater propensity to develop LMD. However, given the overall low incidence, early diagnosis will require increased awareness. Despite recent advances in the treatment of glioblastoma, there has been little improvement in the survival and treatment of patients after they develop LMD with survival remaining around 3–4 months. In the current study patients treated most aggressively with chemotherapy/targeted therapy and radiation had prolonged survival compared to chemotherapy/targeted therapy alone, radiation alone or palliative care. We did not discover any clinical factors predisposing patients to the development of LMD and hypothesize that the molecular profile of glioblastomas developing LMD may reveal a more invasive tumor phenotype. This underscores the need for further genetic research in this area to allow for earlier recognition of tumors likely to develop LMD and to help us design better therapies specifically targeting tumor in the CSF.
Conflict of interest
No conflict of interests to report.
References 1. Stupp R, Mason WP, van den Bent MJ et al (2005) Radiotherapy plus concomitant and adjuvant temozolomide for glioblastoma. N Engl J Med 352(10):987–996 2. Yung WA, Horten BC, Shapiro WR (1980) Meningeal gliomatosis: a review of 12 cases. Ann Neurol 8(6):605–608 3. Vertosick FT Jr, Selker RG (1990) Brain stem and spinal metastases of supratentorial glioblastoma multiforme: a clinical series. Neurosurgery 27(4):516–521 discussion 521–2 4. Arita N, Taneda M, Hayakawa T (1994) Leptomeningeal dissemination of malignant gliomas. Incidence, diagnosis and outcome. Acta Neurochir (Wien) 126(2–4):84–92 5. Engelhard HH, Corsten LA (2005) Leptomeningeal metastasis of primary central nervous system (CNS) neoplasms. Cancer Treat Res 125:71–85 6. Stupp R, Hegi ME, Mason WP et al (2009) Effects of radiotherapy with concomitant and adjuvant temozolomide versus radiotherapy alone on survival in glioblastoma in a randomised
J Neurooncol (2014) 120:597–605
7.
8.
9.
10. 11.
12.
phase III study: 5-year analysis of the EORTC-NCIC trial. Lancet Oncol 10(5):459–466. doi:10.1016/S1470-2045(09)70025-7 de Groot JF, Fuller G, Kumar AJ et al (2010) Tumor invasion after treatment of glioblastoma with bevacizumab: radiographic and pathologic correlation in humans and mice. Neuro-Oncology. 12(3):233–242 Lu KV, Chang JP, Parachoniak CA et al (2012) VEGF inhibits tumor cell invasion and mesenchymal transition through a MET/ VEGFR2 complex. Cancer Cell 22(1):21–35. doi:10.1016/j.ccr. 2012.05.037 Clarke JL (2012) Leptomeningeal metastasis from systemic cancer. Continuum (Minneap Minn). 18(2):328–342. doi:10. 1212/01 Kaplan EL, Meier P (1958) Nonparametric estimation from incomplete observations. J Am Stat Assoc 53:457–481 Korshunov A, Sycheva R, Golanov A, Pronin I (2007) Gains at the 1p36 chromosomal region are associated with symptomatic leptomeningeal dissemination of supratentorial glioblastomas. Am J Clin Pathol 127(4):585–590 Kato H, Fujimura M, Kumabe T, Ishioka C, Kanamaru R, Yoshimoto T (2004) PTEN gene mutation and high MIB-1 labeling index may contribute to dissemination in patients with glioblastoma. J Clin Neurosci 11(1):37–41
605 13. Herrlinger U, Wiendl H, Renninger M, Fo¨rschler H, Dichgans J, Weller M (2004) Vascular endothelial growth factor (VEGF) in leptomeningeal metastasis: diagnostic and prognostic value. Br J Cancer 91:219–224. doi:10.1038/sj.bjc.6601953www.bjcancer. com 14. Chamberlain MC (2003) Combined-modality treatment of leptomeningeal gliomatosis. Neurosurgery 52(2):324–329 15. Le Rhun E, Taillibert S, Chamberlain MC (2013) Carcinomatous meningitis: leptomeningeal metastases in solid tumors. Surg Neurol Int 2(4(Suppl 4)):S265–S288. doi:10.4103/2152-7806.111304 16. Chamberlain MC (2013) Prognostic utility of neuraxis imaging in leptomeningeal metastasis: a retrospective case series. Eur Assoc Neuro Oncol 3(1):6–10 17. Chamberlain MC, Soffietti R, Raizer J et al (2014) Leptomeningeal metastasis: a response assessment in neuro-oncology critical review of endpoints and response criteria of published randomized clinical trials. Neuro Oncol. doi:10.1093/neuonc/ nou089 18. Chamberlain MC (2013) Comprehensive neuraxis imaging in leptomeningeal metastasis: a retrospective case series. CNS Oncol 2(2):121–128. doi:10.2217/cns.12.45
123
CLINICAL STUDY
Leptomeningeal dissemination in glioblastoma; an inspection of risk factors, treatment, and outcomes at a single institution Jacob J. Mandel • Shlomit Yust-Katz • David Cachia • Jimin Wu • Diane Liu • John F. de Groot • Alfred W.K. Yung • Mark R. Gilbert
Received: 4 July 2014 / Accepted: 17 August 2014 / Published online: 29 August 2014 Ó Springer Science+Business Media New York 2014
Abstract There are few studies reporting the incidence of leptomeningeal dissemination (LMD) in patients with glioblastoma; only small case series have been published. Consequently, there are no established standards of care for these patients. Therefore, we undertook this retrospective review to evaluate a large series of patients with glioblastoma treated at MD Anderson Cancer Center to estimate the incidence of LMD and assess the impact of a variety of treatment modalities. Analysis was performed on 595 patients with glioblastoma treated on clinical trials from 2006 to 2012. The diagnosis of LMD was made by imaging or positive cerebrospinal fluid cytology in 24 patients. An additional 12 patients with known LMD diagnosed during this same period were included to evaluate the impact of treatment on outcome for a total of 36 patients. LMD developed in 4.0 % (24/595 patients) of the clinical trial
Jacob J. Mandel and Shlomit Yust-Katz have contributed equally to the manuscript. J. J. Mandel (&) Department of Neurology, Baylor College of Medicine, One Baylor Plaza, MS Nb302, Houston, TX 77030, USA e-mail: [email protected] S. Yust-Katz Rabin Medical Center, Jabotinski Street 39, 49100 Petah-Tikva, Israel D. Cachia J. F. de Groot A. W.K. Yung M. R. Gilbert Division of Neuro-Oncology, The University of Texas MD Anderson Cancer Center, 1515 Holcombe Blvd Unit 431, Houston, TX 77030-4009, USA J. Wu D. Liu Quantitative Sciences - Unit 1409, The University of Texas MD Anderson Cancer Center, P. O. Box 301402, Houston, TX 77230-1402, USA
cohort. Median survival from glioblastoma diagnosis was 16.0 months. Estimated median time of glioblastoma diagnosis to LMD was 11.9 months. Median overall survival from the time of LMD diagnosis was 3.5 months. Patients treated for LMD with chemotherapy/targeted therapy and radiation had a significantly prolonged survival (7.7 months) compared to chemotherapy/targeted therapy alone, radiation alone or palliative care. LMD remains an uncommon event in patients with glioblastoma. Patients treated aggressively with chemotherapy/targeted therapy and radiation had the longest median survival following diagnosis of LMD. However, patients receiving chemotherapy/targeted therapy and radiation were younger and this may have influenced survival. Given the overall poor outcomes, improved therapeutic approaches are needed for glioblastoma patients with LMD. Keywords Glioblastoma Leptomeningeal dissemination Leptomeningeal gliomatosis
Introduction Glioblastoma is the most common primary brain tumor in adults. Despite advances in surgical techniques, radiation treatments and chemotherapy, it continues to have a dismal prognosis [1]. Surprisingly, given the location of glioblastoma in the central nervous system and relatively high exposure of the ventricular system during tumor resection, the incidence of leptomeningeal dissemination (LMD) is thought to be uncommon. However, data regarding the true incidence of LMD in patients with glioblastoma is scarce with only a few case reports and small case series published in the past. These reports provide a wide range of incidence of LMD; up to 10–21 % on autopsy series [2],
123
598
but symptomatic dissemination is much lower at around 2 % [3]. Earlier studies have shown that survival is short after diagnosis of LMD, in the range of only 12–20 weeks [4]. No prospective clinical trials have been reported and there are no established standards of care for patients with LMD from glioblastoma. Many treatments have been used including radiation and a variety of chemotherapies including intrathecal chemotherapy with Methotrexate, araC, thio-TEPA, or ACNU with little success [5]. Recent advances in surgical techniques leading to more extensive tumor resections and the widespread use of agents such as temozolomide in conjunction with more refined radiotherapy techniques, have resulted in an overall increase in median survival with glioblastoma, and most significantly, an increase in the 5-year survival rate [6]. This increase in long-term survival has led to speculation that the incidence of LMD would increase, as has been reported in other cancers [3]. There is further speculation that, in addition to the recent increases in survival with the widespread use of temozolomide, other treatments such as anti-angiogenic agents may increase the incidence of LMD as a combined result of prolonged survival and changes in tumor biology to a more invasive and infiltrative phenotype [7, 8]. In addition to the potential impact on the incidence of LMD, these agents may alter outcomes once LMD has been diagnosed. The goal of this study was to estimate the incidence of LMD in glioblastoma and to examine the impact of treatments on survival in this patient group.
Methods Patients After obtaining institutional board review approval (protocol PA13-0216), we retrospectively identified patients with glioblastoma who developed LMD from our institutional database. Initially, we reviewed all 595 glioblastoma patients enrolled on clinical trials at MD Anderson during the years 2006–2012 because this group had the most extensive and complete follow up data. We later identified an additional 12 patients with known LMD diagnosed during this same period that were included to evaluate the impact of treatment on outcome. All of the patients had a pathologically-confirmed diagnosis of glioblastoma using WHO criteria. The diagnosis of LMD was determined by either radiographic findings on magnetic resonance imaging (MRI) or cerebrospinal fluid (CSF) cytology containing malignant cells. The radiographic criteria used to diagnose LMD were leptomeningeal contrast enhancement around the contours
123
J Neurooncol (2014) 120:597–605
of the gyri and sulci or multiple nodular deposits in the subarachnoid space, particularly focusing on the cerebellar folia, the cortical surface, and the basal cisterns. In the spinal cord, linear or nodular enhancement along the surface of the cord and linear or nodular enhancement of the cauda equina were deemed conclusive of leptomeningeal spread [9]. Leptomeningeal dissemination on imaging was confirmed by independent Radiology reports. Data collected included demographic characteristics (age and sex), tumor characteristics (primary or secondary glioblastoma, tumor location, distance from the ventricles), treatment (extent of resection as determined by imaging or operative reports if imaging was not available, whether the ventricles were opened during surgery, number of resections prior to LMD, time from last surgery to LMD, number of progressions prior to development of LMD, type of radiation and chemotherapy/targeted therapy received prior and after diagnosis of LMD), diagnostic results (MRI imaging and CSF studies including cytology), survival from glioblastoma diagnosis, time from glioblastoma diagnosis to LMD diagnosis and survival from the time of LMD diagnosis. Statistical analysis Data was first summarized using standard descriptive statistics for demographic and clinical variables. The survival, time from initial glioblastoma diagnosis to death, time from glioblastoma diagnosis to LMD diagnosis and time from LMD diagnosis to death, were evaluated using the Kaplan–Meier method and the comparison between or among patients’ characteristics was assessed using log-rank test [10]. Both univariate and multivariate Cox regression models were applied to assess the effect of covariates of interest on three time-to-event endpoints. All computations were carried out in SAS 9.3 (SAS Institute Inc., Cary, NC, USA).
Results A total of 24 patients out of the 595 patients from the clinical trial cohort were identified to have developed LMD; yielding a rate of 4.0 % of patients for this group. Targeted retrospective chart review of patients with known LMD added another 12 patients for a total of 36 patients with LMD. Patient’s characteristics before LMD diagnosis Among the 36 patients included in the analysis, 24 (66.6 %) were men and 12 were women, with a median age of 44 years (range 21–71). Primary tumor location was
J Neurooncol (2014) 120:597–605
599
Table 1 Clinical characteristics of the patients at time of glioblastoma diagnosis
Table 2 Patients characteristics at time of LMD diagnosis Variable
N (%)
Patient
36(100 %)
Variable
N (%)
Patient
36(100 %)
Age at LMD diagnosis: Median (Min–Max)
45(21–73)
Age at glioblastoma diagnosis: Median (Min–Max)
44(21-71)
Time from GB diagnosis to development of LMD (Months) Median (Min–Max))
11.9(0–62)
Male
24(66.7 %)
KPS at LMD diagnosis
Female
12(33.3 %)
Gender
KPS 70 or less 80 or greater
2(8.3 %) 22(91.7 %)
Unknown*
12
Tumor location
70 or less
11(34.4 %)
80 or higher
21(65.6 %)
Unknown* Number of resections prior to LMD 1 or less 2 or more
4 20(55.6 %) 16(44.4 %)
Parietal
7(19.4 %)
Frontal
11(30.5 %)
Time from last surgery to development of LMD (Months) Median (Min–Max)
Temporal
11(30.5 %)
Common presenting symptoms #
Midbrain
2(5.6 %)
None
6(16.7 %)
Cerebellar
1(2.8 %)
Headache
11(30.6 %)
Thalamus
1(2.8 %)
Ataxia
10(27.8 %)
Spine
2(5.6 %)
Pain
7(19.4 %)
1(2.8 %)
Visual changes
7(19.4 %)
Supratentorial
31(86.1 %)
Seizure Lower extremity weakness
6(16.7 %) 5(13.9 %)
Infratentorial
5(13.9 %)
Hypothalamus Glioblastoma location at diagnosis
Extent of resection of glioblastoma at diagnosis Biopsy
Brain MRI characteristics
8.04(0.12–61.56)
21(58.3 %)
Nodules 6(16.7 %)
Subarachnoid
11(30.5 %)
Subtotal
10(27.8 %)
Ventricular
10(27.8 %)
Gross total
20(55.6 %)
Communicating with the ventricle at time of glioblastoma diagnosis Yes No Spinal cord
17(50.0 %)
12(33.3 %)
Focal
7(19.4 %)
Diffuse
5(13.9 %)
17(50.0 %)
No evidence of LMD
2
Ependymal enhancement
23(63.9 %)
Sulci enhancement
3(8.3 %)
Folia enhancement
7(19.4 %)
Concurrent radiation ? chemo with temozolomide after glioblastoma diagnosis Yes
32(91.4 %)
No
1(2.9 %)
Radiation prior to glioblastoma diagnosis**
2(5.7 %)
Unknown* Adjuvant temozolomide for glioblastoma
Pial enhancement
1
Cranial nerve enhancement Spinal MRI characteristics
3(8.3 %)
4(11.1 %) 10(47.6 %)
Nodules Subarachnoid Ventricular
8(38.1 %) 2(9.5 %)
Pial enhancement
10(47.6 %)
Yes
32(91.4 %)
No
1(2.9 %)
Focal
5(23.9 %)
Temozolomide prior to glioblastoma diagnosis**
2(5.7 %)
Diffuse
5(23.8 %)
Unknown*
1
* Patients with ‘‘Unknown’’ status are excluded from proportion calculation ** for patients with secondary glioblastoma
predominantly supratentorial (31, 86 %) and the remainder (5) were infratentorial. Seventeen (50 %) of the tumors had direct communication with a ventricle of which 9 (26 %) had the ventricles opened during initial surgery.
No evidence of LMD
1(4.8 %)
Nerve root enhancement
12(57.1 %)
Spinal imaging showing LMD Positive
20(95.2 %)
Negative
1(4.8 %)
Unknown*
15
Spinal imaging showing LMD with conus/cauda equina affected No
13(61.9 %)
Yes
8(38.1 %)
123
600
J Neurooncol (2014) 120:597–605
Table 2 continued Variable Unknown*
Table 3 Descriptive summary N (%)
Variable
15(n/a %)
Treatment after LMD diagnosis
Cerebral spinal fluid cytology for malignant cells
N (%) 36
Hospice
5(14.7 %)
Positive Negative
5(45.5 %) 6(54.5 %)
Radiation alone
2(5.9 %)
Chemo alone
16(47.1 %)
Unknown*
25
Chemo ? radiation
10(29.4 %)
Resection
1(2.9 %)
Unknown
2
Bevacizumab use prior to LMD diagnosis No
24(66.7 %)
Yes
12(33.3 %)
Time from glioblastoma diagnosis to start of Bevacizumab
11.64(1.68–25.9)
(patients treated prior to LMD) (Months) Median (Min–Max) * Patients with ‘‘Unknown’’ status are excluded from proportion calculation # Patients may have more than one symptom
The patterns of treatment after glioblastoma diagnosis and patients characteristics before LMD diagnosis are detailed in Table 1. Patient’s characteristics at time of LMD diagnosis Patients presented with a variety of symptoms (Table 2). Spinal cord or posterior fossa symptoms such as ataxia, lower extremity weakness, and parasthesias were present in 47 %. The most common single presenting symptom of LMD was headaches, reported in 11 patients (31 %). 36 % (4/11) of patients with headaches were found to have hydrocephalus. Six patients (16 %) were asymptomatic at diagnosis that was made by imaging findings. At the time of LMD diagnosis, three patients (8.3 %) had stable primary disease. Patients had a median KPS of 80 (range 60–100) and a median of 1 recurrence (range 0–5). Sixteen patients (44 %) had 2 or more tumor resections prior to the development of LMD. Six out of these 16 patients with multiple surgeries had their ventricles opened on initial resection. 30 % (3/10) of patients without ventricular wall compromise on initial resection had a subsequent resection with opening of the ventricles. The median time from last surgery to diagnosis of LMD was 8.0 months (range, 0.1–61.6). Three patients (8.3 %) were diagnosed with LMD at the time of initial diagnosis of GBM before a surgical procedure and one patient (3 %) was diagnosed with LMD during the week following their resection. Of the three patients diagnosed with LMD prior to a surgical procedure, two had a brainstem location of their primary glioblastoma and the other had a hypothalamic lesion. All three underwent biopsies for diagnostic purpose that revealed the diagnosis of glioblastoma.
123
Type of radiation alone Focal palliative Type of chemo alone Intravenous chemo Bevacizumab containing regimen Type of radiation for Chemo ? radiation
2 2(100 %) 16 15(94 %) 13(81 %) 10
Craniospinal
4(40 %)
Focal brain and spine
3(30 %)
Focal brain
2(20 %)
Whole spine
1(10 %)
Spinal MR imaging was obtained in 21 patients (58 %) and leptomeningeal spread was observed in 20 patients (95 %). Leptomeningeal tumor deposits were found in the conus/cauda equina in eight patients (38 %) who had spinal imaging. However, only two of these patients displayed components of the classical symptoms of cauda equina syndrome (back pain, saddle anesthesia, lower extremity weakness/sensory loss, and loss of ankle reflexes). CSF analysis was performed in 11 patients (31 %) and cytologic examination was positive for malignant cells in five patients (45 %); four with the first CSF study. All of the patients with negative CSF studies underwent only a single lumbar puncture. Every patient with positive CSF also had leptomeningeal disease seen on spinal MR imaging, excluding one patient who did not obtain spinal imaging. IDH-1 status was obtained on seven patients (19 % of the total) and was mutated in two patients (29 % of those tested). P53 staining was obtained in seven patients (19 % of the total) and was positive in 4(57 % of those tested). Ki-67 % was recorded in four patients, three at time of glioblastoma diagnosis (19, 27, 35 %) and one at first tumor recurrence (14.7 %). Four tumors were noted to be pathologic variants with three gliosarcoma and one a giant cell variant. Treatment varied widely among the cohort of patients with LMD. As detailed in Table 3, 28 patients received either radiation, chemotherapy/targeted therapy or both chemotherapy/targeted therapy and radiation, one patient (3 %) had a resection of tumor in the cerebellar vermis that
J Neurooncol (2014) 120:597–605
601
was symptomatic. Five patients (14 %) were referred to hospice without further therapy and two patients (6 %) were lost to follow up.
Survival analysis At the time of this review, 30 patients (83 %) were deceased. The estimated overall median survival for all patients from glioblastoma diagnosis was 16.0 months (range 4–69). Median survival time from diagnosis of glioblastoma to LMD was 11.9 months (range 0–62). Median survival from the diagnosis of LMD was 3.5 months. Median survival from the diagnosis of LMD was 3.4 months for the clinical trial cohort patients and 4.2 months for non-trial patients (p = 0.38). Survival from glioblastoma diagnosis Univariate analysis found that longer overall survival from initial diagnosis of glioblastoma was found to be associated to prior treatment with bevacizumab (HR 2.63; 95 % CI 1.16 to 5.99; p = 0.0212) and tumors not communicating with the ventricles (HR 2.32; 95 % CI 1.05–5.13; p = 0.0371). Bevacizumab treatment (HR 3.40; 95 % CI 1.41–8.19; p = 0.0065) and tumors not communicating with the ventricles (HR 3.24; 95 % CI 1.36–7.75; p = 0.0082) remained significant on multivariate analysis. Patients with tumors directly communicating with the ventricles had a median overall survival of 14.8 months that compared unfavorably to non-communicating tumors who had an overall median survival of 23.2 months (p = 0.032). Patients who were diagnosed with LMD at time of diagnosis of glioblastoma had a worse prognosis compared to those who were diagnosed afterwards (estimated overall survival of 4.7 months compared to 16.0 months) although given the small numbers of patients, the difference was not statistically significant. Time to LMD diagnosis On univariate analysis the following factors were found to be statistically significantly associated with a shorter time from initial glioblastoma diagnosis to the finding of LMD: infratentorial tumor location at diagnosis, tumor communication with the ventricle at diagnosis and no bevacizumab use prior to LMD diagnosis. However, only tumor communication with the ventricle at diagnosis (HR 3.14; 95 % CI 1.35–7.26; p = 0.0076) and no bevacizumab use prior to LMD diagnosis (HR 2.97; 95 % CI 1.30–6.79; p = 0.0099) were significant on multivariate analysis.
Fig. 1 Kaplan Meier survival curve from LMD diagnosis according to treatment
Table 4 Univariate cox regression models-from LMD to death according to treatment Variable
Hazard ratio
95 % CI
p value
Lower CI
Upper CI
Treatment after LMD diagnosis \0.0001
Hospice
47.39
9.03
248.67
Radiation alone
14.27
2.22
91.84
0.0051
Chemo alone
3.87
1.25
12.05
0.0194
Chemo ? radiation
1.00
Infratentorial tumors were found to develop LMD at a shorter period of time compared to supratentorial tumors (4.9 months compared to 11.9 months, p = 0.003). Also, tumors communicating with the ventricles developed LMD quicker (11.9 months vs. 20.3 months, p = 0.037) than tumors not communicating with the ventricles. Survival from LMD diagnosis On univariate analysis, patients treated with a combination of chemotherapy/targeted therapy and radiation had a significantly prolonged survival compared to chemotherapy/ targeted therapy alone, radiation alone or palliative care (Fig. 1; Table 4). The patients that received radiation and chemotherapy/targeted therapy after LMD diagnosis did not have a better KPS prior to treatment compared to the other groups (p = 0.85). Patients receiving radiation and
123
602
J Neurooncol (2014) 120:597–605
Table 5 Comparison of patient’s characteristics Covariate
Levels
Total
LMD after glioblastoma diagnosis No LMD
In 2 years
198 (93 %)
15 (7 %)
p value
N (%) Patient
213 (100 %)
Gender
Female
89 (41.8 %)
86 (43.4 %)
KPS at glioblastoma diagnosis
Male 100
3 (20 %)
0.1034
124 (58.2 %) 83 (39 %)
112 (56.6 %) 78 (39.4 %)
12 (80 %) 5 (33.3 %)
– 0.0461
90
79 (37.1 %)
75 (37.9 %)
4 (26.7 %)
80
34 (16 %)
30 (15.2 %)
4 (26.7 %)
–
70
13 (6.1 %)
13 (6.6 %)
0 (0 %)
–
2 (13.3 %)
–
60 Location of glioblastoma
Extent of glioblastoma resection
Number of surgeries
Communicating with ventricle at diagnosis
Ventricle opened on initial resection
4 (1.9 %) 64 (30 %)
56 (28.3 %)
8 (53.3 %)
0.2427
Temporal
98 (46 %)
92 (46.5 %)
6 (40 %)
–
Parietal
38 (17.8 %)
37 (18.7 %)
1 (6.7 %)
–
Others
13 (6.1 %)
13 (6.6 %)
0 (0 %)
–
Unknown
18 (. %) 8 (53.4 %)
0.3533
–
GTR
112 (57.4 %)
104 (57.8 %)
STR
56 (28.7 %)
53 (29.4 %)
3 (20 %)
–
BX
27 (13.8 %)
23 (12.8 %)
4 (26.7 %)
–
1
91 (42.7 %)
84 (42.4 %)
7 (46.7 %)
0.2353
2 3
69 (32.4 %) 41 (19.2 %)
67 (33.8 %) 36 (18.2 %)
2 (13.3 %) 5 (33.3 %)
– –
4
12 (5.6 %)
11 (5.6 %)
1 (6.7 %)
–
Unknown
20 (. %)
–
No
135 (69.9 %)
125 (70.2 %)
10 (66.7 %)
Yes
58 (30.1 %)
53 (29.8 %)
5 (33.3 %)
Unknown
7 (. %) 184 (89.3 %)
172 (89.1 %)
12 (92.3 %)
Yes
22 (10.7 %)
21 (10.9 %)
1 (7.7 %)
Clinical cohort patients Of the 595 patients on clinical trials, seven patients (1.2 %) had infratentorial lesions (2 midbrain, 2 pons, 1 medullary and 2 cerebellar). One of the infratentorial (14 %) lesions in the clinical cohort developed LMD. 33 % (4/12) of the
0.7740 – –
No
chemotherapy/targeted therapy after LMD diagnosis were younger with a median age of 35.1 years compared to 47.3 years for patients sent to hospice, 52.6 years for patients who received radiation alone, and 45.3 years for patients receiving chemotherapy alone. Median survival for patients asymptomatic at the time of diagnosis of LMD was 4.6 months. Patients with large volume leptomeningeal disease (defined as nodules on MRI [ 5 9 10 mm in orthogonal diameter) did not have a significantly different overall survival than small volume disease (3.5 months vs. 3.8 months, p = 0.81).
123
2 (1 %)
Frontal
0.9999 –
non-clinical trial cases (2 medullary and 2 spinal cord) were infratentorial.
Comparison analysis An analysis was performed comparing the patients with glioblastoma who developed LMD to the patients that did not develop LMD in the clinical trials cohort group. Due to concern that overall survival could influence the development of LMD, a 2 year time cut off from diagnosis of glioblastoma was used to establish these groups. This allowed the majority of patients who developed LMD to be compared and the ratio of patients with LMD to those without it at this time was the most similar to the incidence of LMD among patients with glioblastoma found in our study. Fifteen of the 24 LMD patients in the clinical trial cohort were found to develop LMD within 2 years of the diagnosis of glioblastoma. In the clinical trial cohort, 198
J Neurooncol (2014) 120:597–605 Table 6 Comparison of continuous covariate
603
Covariate Age at glioblastoma diagnosis KPS at glioblastoma diagnosis Number of surgeries
out of the 593 patients who did not develop LMD survived more than 2 years after tumor diagnosis. Thus, the clinical and demographic characteristics of these 15 LMD patients were compared to the 198 patients that did not develop LMD over the first 2 years of survival (Tables 5, 6). Only KPS at initial diagnosis of glioblastoma was slightly lower in patients who developed LMD and was the only statistically significant difference between the two groups (p = 0.046). No comparisons of PFS and OS between the two groups were performed because the identification of the comparative group created a biased population as defined by the 2-year survival requirement.
Discussion This retrospective review of 36 patients with glioblastoma and LMD provides further insight into the prevalence, treatment patterns and outcome for this rare diagnosis. Our data verifies the low incidence (4.0 %) of LMD among patients with glioblastoma on clinical trials and identifies several novel findings influencing time to development of LMD. This study also corroborates the poor prognosis of patients with glioblastoma and LMD as previously described [4]. However, patients treated with radiation and chemotherapy/targeted therapy following LMD diagnosis had significantly prolonged survival compared to patients treated less aggressively, although treatment decisions may have been related to other prognostic factors. An analysis comparing the patients with glioblastoma who developed LMD to the patients that did not develop LMD in the clinical trials cohort was performed but did not reveal any clinical risk factors for the development of LMD. Previous studies have proposed that LMD occurs primarily in glioblastoma patients with a prolonged overall survival, thereby the consequence of improvements in treatment [3]. However, other studies have found the majority of patients developed LMD within 12 months of their glioblastoma diagnosis [4, 11]. Our study found that only 14/36 (36 %) of the patients were diagnosed with LMD more than a year after their glioblastoma diagnosis. This suggests that there may be innate tumor specific
LMD after glioblastoma diagnosis
N
Mean ± Std, Mean (Min, Max)
p value 0.1944
No LMD
198
49.84 ± 12.75, 52 (18, 84)
In 2 years
15
45.4 ± 13.52, 45 (24, 71)
–
No LMD
198
90.81 ± 9.47, 90 (60, 100)
0.2740
In 2 years
15
No LMD
198
In 2 years
15
86.67 ± 13.45, 90 (60, 100)
–
1.87 ± 0.9, 2 (1, 4)
0.7080
2 ± 1.07, 2 (1, 4)
molecular characteristics that make some tumors more successful in the leptomeninges and CSF. Although there has been no clearly established molecular profile associated with the development of LMD, prior studies have suggested that gains at the 1p36 chromosomal region are associated with symptomatic leptomeningeal spread of supratentorial glioblastomas [11]. Furthermore, mutations of the PTEN gene and a high KI-67 proliferative index have been associated with leptomeningeal dissemination of glioblastoma [12]. However, the comparison analysis of patients with glioblastoma who developed LMD to the patients that did not develop LMD in the clinical trials cohort did not reveal any clinical risk factors for the development of LMD. This suggests that the increase in tumor invasiveness and predilection to disseminate into the cerebrospinal fluid, causing LMD may be primarily related to molecular characteristics of the tumor. Although the overall propensity to develop LMD may be driven by molecular factors, this study determined additional factors that were associated with a shorter time frame from initial tumor diagnosis to LMD. Tumors communicating with a ventricle as well as tumors located in the cerebellum and spinal cord had an earlier development of LMD and had a poorer overall survival from initial glioblastoma diagnosis. There have also been concerns that the use of antiangiogenic agents such as bevacizumab may increase the rate of LMD because of reports that these agents may promote a more invasive tumor phenotype. Interestingly, patients who received bevacizumab prior to development of LMD were found to have a longer time to diagnosis of LMD and prolonged overall survival compared to patients who did not have bevacizumab before developing LMD. This would suggest that bevacizumab does not accelerate the development of LMD. Prior studies have shown a high specificity of markedly increased levels of CSF VEGF in LMD [13] . However, the overall improvement in outcome with bevacizumab may reflect either treatment efficacy or a patient selection bias. Of note, prior bevacizumab use did not appear to effect overall survival following LMD diagnosis. At present, there is no standard treatment for LMD secondary to glioblastoma. Despite new developments in
123
604
treatment, median survival from LMD diagnosis until death still remains dismal at 14 weeks, similar to previous studies [4, 14]. This is also consistent with survival from LMD secondary to other solid tumors which have median survival of 12 weeks [15]. A prior study suggested that patients with large volume leptomeningeal disease (defined as nodules on MRI [ 5 9 10 mm in orthogonal diameter) have a significantly worse overall survival than patients with small volume disease; however, this study excluded leptomeningeal disease from primary brain tumors [16]. Eighteen patients (50 %) in our study were found to have large volume disease, but survival was not significantly different. This contradictory finding may be due to the aggressive nature of brain tumors and the high percentage of patients with progressive primary disease at time of LMD diagnosis in our study. Patients treated most aggressively with chemotherapy/ targeted therapy and radiation had the longest median survival from LMD to death (30.7 weeks) compared to chemotherapy/targeted therapy alone (15.4 weeks) or radiation alone (9.2 weeks). Although selection bias cannot be conclusively excluded, overall the patients receiving chemotherapy/targeted therapy and radiation did not have a better KPS prior to treatment compared to the other groups. However, patients receiving radiation and chemotherapy/ targeted therapy after LMD diagnosis were younger and this likely influenced the overall survival. The role of intraCSF chemotherapy is unclear at this time. A prior trial using several intraventricular chemotherapies sequentially for LMD due to high grade glioma resulted in a median survival of only 3.5 months (similar to this review despite only two patients in this study receiving intraventricular chemotherapies) [14]. Moreover, only one patient was treated with craniospinal radiation and received intraventricular chemotherapy likely due to increased toxicity when using both therapies. Due to the heterogenous nature of the treatments and retrospective nature of this study, future randomized studies are necessary to determine if more aggressive treatment does in fact improve survival. Previous studies treating LMD have been challenging to interpret due to the absence of a uniform response criteria; however, a Response Assessment in Neuro-Oncology (RANO) group is reported to be working on establishing new response criteria to be validated in future clinical trials in LMD [17]. At this time treatment decisions should continue to be made on a case by case basis using sound clinical judgment. Additionally, clinical manifestations of LMD can overlie appreciably with those of primary brain tumor progression, metabolic disturbances, and treatment-related toxicities. It is important to note that this study may have excluded erroneously several patients with undiagnosed LMD who were asymptomatic or had symptoms from
123
J Neurooncol (2014) 120:597–605
LMD blamed on primary disease progression or other causes. Autopsy studies have demonstrated a higher rate of LMD and a spinal MRI should be obtained in patients with symptoms or clinical exam findings referring to the spinal cord [2]. Furthermore, patients with confirmed LMD should obtain complete neuraxis imaging for proper disease staging and treatment planning [18]. In conclusion, the overall incidence of LMD among glioblastoma patients remains low. Potential risk factors for earlier dissemination such as tumor location and proximity to the ventricular system have been identified in patients that develop LMD and there are preliminary suggestions that there are tumor molecular phenotypes that have a greater propensity to develop LMD. However, given the overall low incidence, early diagnosis will require increased awareness. Despite recent advances in the treatment of glioblastoma, there has been little improvement in the survival and treatment of patients after they develop LMD with survival remaining around 3–4 months. In the current study patients treated most aggressively with chemotherapy/targeted therapy and radiation had prolonged survival compared to chemotherapy/targeted therapy alone, radiation alone or palliative care. We did not discover any clinical factors predisposing patients to the development of LMD and hypothesize that the molecular profile of glioblastomas developing LMD may reveal a more invasive tumor phenotype. This underscores the need for further genetic research in this area to allow for earlier recognition of tumors likely to develop LMD and to help us design better therapies specifically targeting tumor in the CSF.
Conflict of interest
No conflict of interests to report.
References 1. Stupp R, Mason WP, van den Bent MJ et al (2005) Radiotherapy plus concomitant and adjuvant temozolomide for glioblastoma. N Engl J Med 352(10):987–996 2. Yung WA, Horten BC, Shapiro WR (1980) Meningeal gliomatosis: a review of 12 cases. Ann Neurol 8(6):605–608 3. Vertosick FT Jr, Selker RG (1990) Brain stem and spinal metastases of supratentorial glioblastoma multiforme: a clinical series. Neurosurgery 27(4):516–521 discussion 521–2 4. Arita N, Taneda M, Hayakawa T (1994) Leptomeningeal dissemination of malignant gliomas. Incidence, diagnosis and outcome. Acta Neurochir (Wien) 126(2–4):84–92 5. Engelhard HH, Corsten LA (2005) Leptomeningeal metastasis of primary central nervous system (CNS) neoplasms. Cancer Treat Res 125:71–85 6. Stupp R, Hegi ME, Mason WP et al (2009) Effects of radiotherapy with concomitant and adjuvant temozolomide versus radiotherapy alone on survival in glioblastoma in a randomised
J Neurooncol (2014) 120:597–605
7.
8.
9.
10. 11.
12.
phase III study: 5-year analysis of the EORTC-NCIC trial. Lancet Oncol 10(5):459–466. doi:10.1016/S1470-2045(09)70025-7 de Groot JF, Fuller G, Kumar AJ et al (2010) Tumor invasion after treatment of glioblastoma with bevacizumab: radiographic and pathologic correlation in humans and mice. Neuro-Oncology. 12(3):233–242 Lu KV, Chang JP, Parachoniak CA et al (2012) VEGF inhibits tumor cell invasion and mesenchymal transition through a MET/ VEGFR2 complex. Cancer Cell 22(1):21–35. doi:10.1016/j.ccr. 2012.05.037 Clarke JL (2012) Leptomeningeal metastasis from systemic cancer. Continuum (Minneap Minn). 18(2):328–342. doi:10. 1212/01 Kaplan EL, Meier P (1958) Nonparametric estimation from incomplete observations. J Am Stat Assoc 53:457–481 Korshunov A, Sycheva R, Golanov A, Pronin I (2007) Gains at the 1p36 chromosomal region are associated with symptomatic leptomeningeal dissemination of supratentorial glioblastomas. Am J Clin Pathol 127(4):585–590 Kato H, Fujimura M, Kumabe T, Ishioka C, Kanamaru R, Yoshimoto T (2004) PTEN gene mutation and high MIB-1 labeling index may contribute to dissemination in patients with glioblastoma. J Clin Neurosci 11(1):37–41
605 13. Herrlinger U, Wiendl H, Renninger M, Fo¨rschler H, Dichgans J, Weller M (2004) Vascular endothelial growth factor (VEGF) in leptomeningeal metastasis: diagnostic and prognostic value. Br J Cancer 91:219–224. doi:10.1038/sj.bjc.6601953www.bjcancer. com 14. Chamberlain MC (2003) Combined-modality treatment of leptomeningeal gliomatosis. Neurosurgery 52(2):324–329 15. Le Rhun E, Taillibert S, Chamberlain MC (2013) Carcinomatous meningitis: leptomeningeal metastases in solid tumors. Surg Neurol Int 2(4(Suppl 4)):S265–S288. doi:10.4103/2152-7806.111304 16. Chamberlain MC (2013) Prognostic utility of neuraxis imaging in leptomeningeal metastasis: a retrospective case series. Eur Assoc Neuro Oncol 3(1):6–10 17. Chamberlain MC, Soffietti R, Raizer J et al (2014) Leptomeningeal metastasis: a response assessment in neuro-oncology critical review of endpoints and response criteria of published randomized clinical trials. Neuro Oncol. doi:10.1093/neuonc/ nou089 18. Chamberlain MC (2013) Comprehensive neuraxis imaging in leptomeningeal metastasis: a retrospective case series. CNS Oncol 2(2):121–128. doi:10.2217/cns.12.45
123
