Treatment of dedifferentiated chordoma: a retrospective study from a large volume cancer center

Abstract

Objective

Dedifferentiated chordomas (DC) are genetically and clinically distinct from conventional chordomas (CC), exhibiting frequent SMARCB1 alterations and a more aggressive clinical course. We compared treatment and outcomes of DC and CC patients in a retrospective cohort study from a single, large-volume cancer center.

Methods

Overall, 11 DC patients were identified from 1994 to 2017 along with a cohort of 68 historical control patients with CC treated during the same time frame. Clinical variables and outcomes were collected from the medical record and Wilcoxon rank sum or Fisher exact tests were used to make comparisons between the two groups. Kaplan–Meier survival analysis and log-rank tests were used to compare DC and CC overall survival.

Results

DC demonstrated a bimodal age distribution at presentation (36% age 0–24; 64% age > 50). DC patients more commonly presented with metastatic disease than CC patients (36% vs. 3% p = 0.000). DC patients had significantly shorter time to local treatment failure after radiation therapy (11.1 months vs. 34.1 months, p = 0.000). The rate of distant metastasis following treatment was significantly higher in DC compared to CC (57% vs. 5%, p = 0.000). The median overall survival after diagnosis for DC was 20 months (95% CI 0–48 months) compared to 155 months (95% CI 94–216 months) for CC (p = 0.007).

Conclusion

DC patients exhibit significantly higher rates of both synchronous and metachronous metastases, as well as shorter overall survival rates compared to conventional chordoma. The relatively poor survival outcomes with conventional therapies indicate the need to study targeted therapies for the treatment of DC.

Introduction

Dedifferentiated chordoma (DC) is a clinically aggressive subtype of chordoma with a high grade sarcomatous component [1, 2]. DC accounts for 2–8% of all chordomas and can occur de novo or as malignant transformation of previously treated chordoma [3–5]. Dedifferentiated chordoma has a high metastatic potential, and can appear with distant lesions at first presentation [2, 6]. Its relative, conventional chordoma (CC), is a locally aggressive bone tumor derived from remnants of the embryonic notochord. Overall, chordoma accounts for 17.5% of primary spine bone tumors with an incidence of 0.08 per 100,000 persons [7, 8]. Chordomas can arise in the skull base (15%), sacrum (50%), or subaxial spine (35%) [4, 8, 9].

DC may be genetically distinct from conventional chordoma in that it exhibits a high frequency of mutations in the SWI/SNF complex member SMARCB1/INI1 [10–12]. SMARCB1/INI1 mutations are more commonly associated with atypical teratoid/rhabdoid tumor (AT/RT), a brain tumor which presents in young children [10, 11]. Similar to AT/RT, poorly differentiated chordomas display isolated losses affecting the SMARCB1 region in 22q11, but can be differentiated from AT/RT based on a distinct methylation profile [10]. While the mechanism of tumorigenesis based on this mutation is unclear, it is a reliable diagnostic parameter [11].

Despite these genetic differences, preoperative differentiation of DC from CC based on imaging and clinical presentation alone is difficult. Thus, treatment for DC is largely based on clinical algorithms developed for conventional chordoma. Surgery is considered the primary treatment modality, with adjuvant radiation therapy (RT) used to improve local control. Radiation therapy has been used as upfront treatment in cases with diffuse metastatic disease at presentation [13]. Chemotherapy is not currently effective for either form of chordoma [1, 14]. Unfortunately, even with radiation therapy, local recurrence is common [4, 14–17]. This may be due to regional invasion and microscopic disease beyond surgical margins.

Given its low incidence, much of the data on treatment of dedifferentiated chordoma is limited to published case reports [1–3, 6, 18]. In this study, we report clinical findings and management from 11 pathologically confirmed de novo DC patients treated at our single large-volume cancer center. We report clinical distinctions in presentation, observed radiation response, and overall survival in comparison to an institutional cohort of 68 patients with conventional chordoma.

Methods

Following Institutional Review Board approval (MSKCC #16–882), we identified a total of 11 dedifferentiated chordoma patients treated at Memorial Sloan Kettering Cancer Center between 1994 and 2017. We also identified a cohort of 68 historical control patients with conventional chordoma treated during the same time frame. There were no age restrictions and all patients included in analysis had biopsy or surgical pathology confirmed tissue diagnosis. Patient medical records were reviewed to obtain multiple variables including age, sex, tumor location, histology/pathology, surgical treatment, radiotherapy treatment, medical treatments, and clinical outcomes. Clinical outcomes included patient radiation response, tumor recurrence/ local control, and overall survival.

Statistical analysis

Clinical characteristics of DC and CC were compared using the Wilcoxon rank sum test for continuous variables and the Fisher exact test for categorical variables. Overall survival rates were displayed using a Kaplan–Meier curve, and comparisons between DC and CC were made using the log-rank test. P-values less than 0.05 were deemed statistically significant.

Volumetric analysis

For patients with available serial MRI images before and after radiation therapy, the volumes of gross disease in each MRI (GTV) were calculated using image analysis software (iNtuition, Terarecon, Foster City, CA). Gross disease was contoured in 3D by a radiation oncologist (EK) in conjunction with the MSKCC Radiology processing laboratory.

Results

Dedifferentiated chordoma patient characteristics and treatment details are presented as Table 1 and comparative data and statistics between DC and CC patients are presented as Table 2.

Table 1.

Individual dedifferentiated chordoma patient descriptions, treatment details, and clinical outcomes

Case no	Age at diagnosis	Sex	Location of primary tumor	Distant metastasis upon presentation	Surgery of primary (margin status)	RT to primary	Type of RT	RT dosage/fractions (Gy)	Local failure	Time to local failure (days)	Distant failure	Time to distant failure (days)	Current status (time post-treatment)
1	1	F	Clivus	Yes	No	Yes	Proton therapy	68/38	Yes	134	–	–	DOD
2	80	M	Sacrum	Yes	No	No	–	–	No	–	–	–	DOD
3	56	F	Sacrum	Yes	No	No	–	–	a	a	–	–	DOD
4	10	F	Clivus	Yes	No	No	–	–	a	a	–	–	AWD
5	80	M	Sacrum	No	Yes (+)	Yes	SBRT	21/1	Yes	285	Yes	75	DOD
6	80	M	Sacrum	No	Yes (−)	No	–	–	Yes	346	Yes	665	DOD
7	2	F	Clivus	No	No	Yes	Proton therapy	76/38	Yes	350	Yes	350	AWD
8	72	M	Sacrum	No	Yes (−)	Yes	SBRT	21/1	No	–	Yes	252	DOD
9	66	M	Sacrum	No	Yes (−)	Yes	Conventional fractionated	65/36	No	–	No	–	A/NED
10	64	M	Sacrum	No	Yes (−)	Yes	Conventional fractionated	63/35	No	–	No	–	D/NED
11	19	F	Mobile spine	No	Yes (a)	Yes	Conventional fractionated	a	a	a	No	–	DOD

RT radiation therapy, SBRT stereotactic body radiation therapy, AWD alive with disease, DOD died of disease or disease-related complications, A/NED alive with no evidence of disease, D/NED died with no evidence of disease.

^amissing data

Table 2.

Patient characteristics and univariate analysis of presenting factors, treatment course, and outcome for conventional chordoma and dedifferentiated chordoma

Factor	Conventional chordoma (n=68) No. of patients (%)	Dedifferentiated chordoma (n=11) No. of patients (%)	p value
Age at diagnosis
0–24	0 (0)	4 (36)	0.0001
25–49	17 (25)	0 (0)	0.0212
50–74	38 (56)	4 (36)	0.0968
75+	13 (19)	3 (28)	0.2483
Sex
M	48 (71)	6 (55)	0.1213
F	20 (29)	5 (45)	0.1214
Location of primary tumor
Clivus	8 (12)	3 (27)	0.0726
Mobile spine	24 (35)	1 (9)	0.0329
Sacrum	36 (53)	7 (64)	0.2404
Distant metastasis at presentation
Y	2 (3)	4 (36)	0.0000
N	66 (97)	7 (64)	0.0000
Initial surgery for primary
Y	50 (74)	6 (55)	0.7089
Surgical margins
Positive	20 (40)	2 (33)	0.5619
Negative	30 (60)	4 (67)	0.5619
N	18 (26)	5 (45)	0.2911
Initial RT for primary
Y	54 (79)	7 (64)	0.7722
N	14 (21)	4 (36)	0.2278
Mode of RT (mean, range)
Conventional (53, 27–72)	17 (31)	3 (42)	0.4171
SBRT (25, 20–36)	30 (56)	2 (29)	0.0228
Proton therapy (73, 70–76)	7 (13)	2 (29)	0.3501
RT dosage (Gy)
Mean (range)	40 (20–76)	50 (21–76)	0.0307
Method of primary treatment
Surgery alone	14 (21)	1 (9)	0.1579
Surgery + RT	35 (51)	4 (36)	0.1667
Surgery + chemo	0 (0)	1 (9)	0.0036
RT alone	17 (25)	0 (0)	0.0222
RT + chemo	1 (1)	1 (9)	0.0444
Chemo alone	0 (0)	2 (18)	0.0000
Surgery + RT + chemo	1 (1)	1 (9)	0.0444
No treatment	0 (0)	1 (9)	0.0036
Local failure
Y	32 (47)	4 (50)	0.2402
N	36 (53)	4 (50)	0.1358
Time to local failure (days)
Mean (range)	1022 (50–5358)	331 (134–557)	0.0000
Distant failure
Y	3 (5)	4 (57)	0.0022
N	63 (95)	3 (43)	0.0000

RT radiation therapy, SBRT stereotactic body radiation therapy

Clinical presentation

Of the 11 patients presenting with de novo dedifferentiated chordoma, there was a bimodal age distribution with 36% of patients diagnosed between ages 0–24 and 64% of patients diagnosed at age greater than 50. This was distinct from the age distribution for CC in which no patient presented under the age of 24, 25% were age 24–50, and 75% of patients were age greater than 50. The most common location of both DC and CC was the sacrum in 64% and 53%, respectively. Following the sacral location, clival tumors were more prevalent in DC (27%) compared to mobile spine tumors more commonly seen in CC (35%). Of the 11 DC patients, 4 patients (36%) presented with metastatic disease compared to only 2 patients (3%) of those with CC. Overall, 4 DC patients had confirmed loss of INI-1 on pathology. Median follow up was 14.5 months (range 2 to 274 months) (Fig. 1).

Fig. 1.

Kaplan–Meier curves for overall survival for conventional chordoma and DC. The number at risk indicates the number of DC and conventional chordoma patients alive at each interval time point. Median survival was 20 months post diagnosis for DC and was 155 months for conventional chordoma, p = 0.007

Treatment course

Of the patients with DC, 55% underwent surgical resection as first intervention compared to 74% of CC patients (including both en bloc and intralesional resection). In the case of DC, the rate of surgery as initial therapy was influenced by the presence of metastatic disease at presentation in 36% of patients. Rates of negative intraoperative margins were similar for DC and CC, 67% vs. 60%, respectively (confirmed on pathology).

For the 4 DC patients with metastatic disease at presentation, no patient had surgical resection of the primary mass, 1 had radiation therapy (RT) to the primary disease, and 3 received chemotherapy (including the patient who had RT) as upfront treatment. In addition, 3 patients had confirmed loss of INI-1 on pathology. For CC, only 1 patient received chemotherapy and radiation as the only upfront therapy, and none received chemotherapy alone.

Radiation therapy

Of patients with dedifferentiated chordoma, 7 patients (64%) underwent upfront adjuvant radiation therapy compared to 54 patients (79%) with conventional chordoma. Regarding specific RT for DC, 3 patients underwent conventional fractionated RT, 2 underwent stereotactic body radiation therapy (SBRT, 21 Gy in single fraction) and 2 underwent proton therapy. Mean cumulative dosage of RT for DC was 50 Gy (range 21–76 Gy). Details of combination therapy and subsequent response are provided in Tables 1 and 2.

For the 4 DC patients with available serial MRI imaging, volumetric tumor response to radiation therapy was analyzed and is presented as Fig. 2. These data were limited, but demonstrate treatment response in all 4 patients at 20 weeks post RT (one patient lost subsequent response).

Fig. 2.

Spider plot of DC patient tumor volume changes after targeted radiation therapy

Patterns of failure

Local failure was observed at similar rates in both the DC and CC cohorts with 50% and 47%, respectively (p = 0.240). However, the time to local failure in DC was significantly shorter than for CC (11.1 months vs. 34.1 months, p = 0.000), reflecting its more aggressive behavior and treatment resistance. Similarly, the rate of distant failure was significantly higher in DC compared to CC given its known higher metastatic potential (57% vs. 5%, p = 0.000). The most common sites of metastasis in the 11 patients with dedifferentiated chordoma were in order of decreasing frequency lung (55%), liver (18%), chest wall/mediastinum (18%), bowel (9%), osseous (9%), and oropharynx (9%).

Overall survival

The median overall survival (OS) after diagnosis for DC was 20 months (95% CI 0–48 months) compared to 155 months (95% CI 94–216 months) for CC. The difference in overall survival between DC and CC was statistically significant. (p = 0.007). Within the bimodal age distribution in DC patients, the median survival was 20 months for both age 0–24 years at diagnosis and for age 50 years and greater at diagnosis (range 3–43 months and range 4–274 months, respectively). There were 2 patients in the DC cohort with greater than 10-year survival, both of whom had negative margins at the time of initial surgical resection.

Subgroup analysis

Excluding the 2 patients with en bloc resection and negative margins, median overall survival was 17 months (range 3 to 53 months).

For the 4 DC patients with distant metastases at the time of initial presentation (stage IV), median overall survival was 5 months (range 3 to 35 months).

Discussion

Dedifferentiated chordoma is a clinically distinct entity from conventional chordoma with a reported incidence of 2–8% of all chordomas [4, 7, 8]. DC carries a poor prognosis with rapid local progression, distant metastasis, and treatment resistance [2, 19]. It has been historically defined by histopathologic criteria, which was the sole basis for inclusion into this series. DC and the new term ‘poorly differentiated’ chordoma has been more recently associated with loss of SMARCB1/INI1, a member of the SWI/SNF chromatin remodeling complex, marking a possible genomic distinction from conventional chordoma [10, 11, 20]. SMARCB1/INI1 loss was found in 4 of 11 of the patients in this series, which is provided as ancillary information, as the test was only available in patients treated more recently. Although most prior reports have characterized dedifferentiated chordoma as a malignancy of childhood or young-adulthood, the present study reveals a bimodal distribution, with approximately half of patients younger than 25 and half older than 50 [5, 10, 11].

The present study confirms what has been inferred from previous case reports. We found that DC patients are significantly more likely to have metastatic disease at the time of presentation, and more likely to develop distant metastasis after treatment [1, 2]. While both CC and DC patients are likely to develop local failure, our series found that time to local failure is significantly less for DC. Interestingly, patterns of metastatic disease for dedifferentiated chordoma were similar to other solid tumor cancers [21]. Finally, overall survival rates for DC are significantly lower than for conventional chordoma, with many DC cases complicated by local recurrence and metastatic disease [1].

Treatment for dedifferentiated chordoma is variable, but upfront surgical resection continues to be the preferred intervention in the absence of distant metastases at presentation, and with negative margins as the primary objective [1, 3]. This is reinforced by the fact that both DC long-term survivors in the present study had upfront surgical resection with negative margins on pathology. Surgery is often non-curative for DC patients, however, due to infiltrative disease and sometimes cryptic presence of distant metastatic disease [1]. Although there was some response to radiation therapy in our series, DC appears to have significant radioresistance given the demonstrated decreased time to local failure. While DC may exhibit initial response to RT, high levels of local recurrence and distant metastasis limit curative outcome [3, 19].

While chemotherapy might logically address the issue of metastatic disease from DC, there has been extremely limited experience. It is unclear even what agent to use. Fleming et al. reported 2 cases in which patients were effectively treated for dedifferentiated chordoma using ifosfamide in one case and a six-drug cocktail in another [22]. Given the genomic similarities of DC to AT/RT, studies have examined the use of the Head Start II protocol (vincristine, cisplatin, etoposide, cyclophosphamide, and high-dose methotrexate) for DC [23, 24]. For the 4 patients in our series receiving traditional chemotherapy for metastatic DC, however, it did not appear to be significantly effective. Ultimately, given the shared genomic profile, specific tumoral therapy (i.e. small molecule or immunotherapy) provides the most alluring therapeutic target. One potential such therapy is direct inhibition Enhancer of Zeste 2 (EZH2), which has been demonstrated to enhance radiosensitivity in AT/RT [25].

Although the overall survival for DC is relatively poor compared to CC, there were 2 long term survivors in our cohort, one of whom is still alive at the time of this report (Table 1: 13 and 22 years). The shared characteristics between these 2 patients were sacral location, absence of metastatic disease at presentation, and en bloc resection of the initial tumor with negative margins at the time of upfront surgery. In addition, following diagnosis of DC, both patients received adjuvant conventional fractionated radiation therapy. This implies that there is a subset of patients with dedifferentiated chordoma for whom surgery might be curative.

Limitations

There are several limitations to the present study. Dedifferentiated chordoma is a rare tumor with low prevalence, making its study difficult, even at a large volume cancer center. In 23 years, there were only 11 cases treated. This makes it difficult to draw conclusions from the present analysis, especially given the variability in treatment between patients. Tumors differed in location, and although en bloc, marginnegative resection was the goal, it was not feasible in all surgical cases. Some patients had incomplete follow up and not all patients who underwent radiation therapy had available serial imaging to monitor treatment response. Although limited, the current series of 11 patients at least confirms what has been inferred from previous case reports. Increased use of national surgical and oncologic registries may help provide enough cases to better elucidate the appropriate management paradigm for dedifferentiated chordoma.

Conclusions

Dedifferentiated chordoma is an aggressive tumor that is genetically and clinically distinct from conventional chordoma. Resection with negative margins and adjuvant RT has the potential for long term control. However, the probability of local recurrence and distant failure remains high and many patients die shortly after treatment due to distant failure, demonstrating a need for effective systemic therapy.