Osteosarcoma of the jaw/craniofacial region

Outcomes after multimodality treatment

Abstract

BACKGROUND:

The current study was performed to evaluate outcomes in patients with osteosarcoma of the head and neck (OHN) who were treated with surgery with or without radiotherapy (RT).

METHODS:

Between 1960 and 2007, 119 patients with OHN underwent macroscopic total resection with or without RT. The median age of the patients was 33 years (range, 7-77 years). Of these 119 patients 92 (77%) underwent surgery alone whereas 27 (23%) patients were treated with combined modality treatment (CMT) comprised of surgery and RT (median dose, 60 Gray [Gy]; range, 50-66 Gy).

RESULTS:

The median follow-up was 5.8 years. Overall survival (OS) rates at 5 years and 10 years were 63% and 55%, respectively. Corresponding disease‒specific survival (DSS) rates were 67% and 61%, respectively. Stratified analysis by resection margin status demonstrated that CMT compared with surgery alone improved OS (80% vs 31%; P = .02) and DSS (80% vs 35%; P = .02) for patients with positive/uncertain resection margins. Multivariate analysis indicated that CMT for patients with positive/uncertain resection margins improved OS (P < .0001). A total of 44 (37%) patients experienced local disease recurrence (LR) and 25 (21%) developed distant metastases (DM). There was no difference noted with regard to DSS if disease recurrence was isolated (LR vs DM: 26% vs 29%, respectively, at 5 years; P = .48) The use of CMT versus surgery alone improved local control (LC) (75% vs 24%; P = .006) for patients with positive/uncertain resection margins. The rate of surgical complications was 28% at 5 years. The rates of RT-associated complications were 40% and 47% at 5 years and 10 years, respectively.

CONCLUSIONS:

The results of the current study indicated that RT in addition to surgery improves OS, DSS, and LC for patients with OHN who have positive/uncertain resection margins. Cancer 2009. © 2009 American Cancer Society.

Osteosarcoma of the head and neck (OHN) is a rare presentation of osteosarcoma, with <10% of all cases of osteosarcoma presenting in this region.1 OHN exhibits a clinical behavior and natural history that are distinct from those of osteosarcoma of the trunk and extremity.1 Compared with the extremity/trunk presentation of this disease, OHN demonstrates less of a propensity to metastasize to pulmonary and other sites, but a high lethality associated with local disease that is difficult to control.1 Radical surgery is the mainstay of treatment in any attempt at curative management2, 3; however, to our knowledge, the role of radiotherapy (RT) has not been clearly defined to date. Data regarding the role of RT in OHN are limited to retrospective investigations. Prior studies examining treatment outcomes and prognostic factors for this rare presentation of osteosarcoma1 have involved relatively small cohorts of patients (median, 35 patients; range, 12-66 patients per study).3-15 In the current study, we have undertaken a review of our experience treating OHN patients who were managed definitively at the University of Texas M. D. Anderson Cancer Center (MDACC). We report on treatment outcomes as well as offer what to our knowledge is the first report concerning local therapy-related complications in these patients.

MATERIALS AND METHODS

Between 1960 and 2007, 127 consecutive patients with nonmetastatic OHN presented to and underwent surgery at MDACC. Of these, 7 patients received debulking surgery only, and 1 patient was lost to follow-up. The current study is comprised of the remaining 119 patients who underwent macroscopic total resection at MDACC with or without RT and/or chemotherapy. Review of the data for this investigation commenced after approval was obtained from our institution's investigation review board. Patients underwent a full history, complete physical examination, routine blood tests, and appropriate imaging before their treatment. A histologic diagnosis of osteosarcoma was confirmed in each case through a review of the slides by a pathologist at MDACC at the time of presentation.

Patient and Tumor Characteristics

Of the patients, 62 (52%) were male and 57 (48%) were female. Ages ranged from 7 to 77 years (median, 33 years). Sixteen patients (13%) presented with a radiation-associated OHN.

The distribution of primary lesions in the head and neck was as follows: 54 (45%) in the mandible, 48 (40%) in the maxilla, 6 (5%) in the calvarium, 2 (2%) in the hard palate, 2 (2%) in the mastoid, 1 (1%) in the skull base, 1 (1%)in the zygoma, 2 (2%) in the frontal or ethmoid sinuses, 1 (1%) in the infratemporal fossa, 1 (1%) spanning the paranasal sinus/sphenoid sinus/anterior cranial fossa, and 1 (1%) in the cervical soft tissues.

Tumor size was documented in 101 patients and ranged from 1.2 to 15 cm in maximal dimension. The tumor size was ≤5 cm in 48 patients (48%) and > 5 cm in 53 patients (52%), with a median size of 5.5 cm. Soft tissue extension of the primary tumor was documented in 81 patients (68%). Fourteen patients (12%) presented with disease recurrence after having undergone 1 (9 patients) or more (5 patients) previous attempts at definitive surgery. The histologic subtype of osteosarcoma was documented for 87 patients (73%) and was as follows: ethmoid sinuses in 21 (24%) patients, chondroblastic in 46 (53%) patients, fibroblastic in 15 (17%) patients, well-differentiated intraosseous osteosarcoma in 5 (6%) patients, small cell osteosarcoma in 2 (2%) patients, mixed osteoblastic and chondroblastic in 1 (1%) patient, mixed fibroblastic and osteoblastic in 1 (1%) patient, periosteal osteosarcoma in 1 (1%) patient, high-grade osteosarcoma with rhabdomyosarcomatous differentiation in 1 (1%) patient, and extraskeletal osteosarcoma in 1 (1%) patient. Tumor grade was specified in 60 of the 119 patients (50%) and was low in 13 (22%) patients, intermediate in 9 (15%) patients, and high in 38 (63%) patients.

Treatment

The surgeries performed were hemimandibulectomy in 33 (28%) patients, segmental mandibulectomy in 9 (8%) patients, marginal mandibulectomy in 2 (2%) patients, infrastructure maxillectomy (orbital floor preserved) in 25 (21%) patients, total maxillectomy (including the zygomatic body and orbital floor) in 12 (10%) patients, radical maxillectomy (including the eye) in 7 (6%) patients, palatectomy in 1 (1%) patient, and other procedures in 30 (25%) patients. Bone reconstruction was undertaken in 50 patients (42%), and soft tissue reconstruction in 68 (57%) patients. Resection margin status after surgery was as follows: 23 (19%) patients with positive margins, 81 (68%) patients with negative margins, and 15 (13%) patients with uncertain margins.

Ninety-two (77%) patients underwent surgery alone whereas 27 (23%) patients were treated with a combination of surgery and RT. RT was postoperative in 22 patients and preoperative in 5 patients. There was no significant difference noted in the proportion of patients receiving RT regardless of whether the resection margins were negative (16% of patients received RT) or positive (27% of patients received RT) (P = .20). RT was administered via megavoltage (cobalt-60 or higher) photons or electrons using techniques appropriate to each site. The median RT dose was 60 Gray (Gy) (range, 50-66 Gy), and the median dose per fraction was 2.0 Gy.

Of the 119 patients, 63 (53%) received chemotherapy as part of the primary management of their disease. Chemotherapy was doxorubicin‒ and/or cisplatin-based, and other agents used included ifosfamide, methotrexate, vincristine, cyclophosphamide, dactinomycin, and dacarbazine.

Follow-up and Statistical Analysis

Follow-up time was calculated from the date of surgery at MDACC. The median follow-up of patients who were still alive at the time of last follow‒up was 5.8 years (range, 3 months to 44.5 years).

The Kaplan-Meier16 method was used to calculate the actuarial curves for overall survival (OS), disease-specific survival (DSS), local control (LC), distant metastatic recurrence (DM), disease-free survival (DFS), and complication rates. The log-rank statistic was used to test for significance of differences between curves. Multivariate regression analysis was performed using the Cox proportional hazards model.17 Differences between proportions for categoric variables were analyzed using the chi-square statistic or Fisher exact test as appropriate. Surgical and RT-related surgical complications were scored as follows: mild (self-limited and requiring no treatment), moderate (requiring conservative medical management), and severe (requiring surgical intervention or hospitalization).

RESULTS

Survival

Thirty-nine patients (33%) died of their OHN. The actuarial OS rates at 5 years and 10 years were 63% and 55%, respectively. The DSS rates were 67% and 61%, respectively, at 5 years and 10 years. The DFS rates at 5 years and 10 years were 49% and 47%, respectively. The last disease recurrence occurred at 5.9 years. Figure 1 shows the curves for OS, DSS, and DFS. Of the 39 patients who died of their OHN, 33 (85%) had local disease recurrence (LR) with or without DM.

Figure 1. Overall survival, disease-free survival, and disease-specific survival for the entire study population.

Table 1 shows the univariate analyses for OS as well as DSS. Positive/unknown resection margin status was associated with poorer OS (P = .001). Patients with low‒grade or intermediate‒grade tumors were found to have a better OS than patients with high‒grade tumors (P = .004). Patients with RT-associated tumors were also found to have a worse OS (48% vs 65%; P = .006). When outcomes with the use of RT was stratified by resection margin status, patients with positive/uncertain margins who received RT were found to have better OS than those who did not (P = .02). Similarly, positive/uncertain resection margins and high‒grade pathology were found to be associated with inferior DSS. In a stratified analysis of treatment by resection margin status, the use of RT in patients with positive/uncertain resection margins also was found to improve DSS. The administration of chemotherapy did not appear to impact OS or DSS.

Table 1. Univariate Analysis for Overall Survival and Disease-specific Survival

Characteristic	No. (% of Total)	5-year OS	P Value	5-year DSS	P Value
OS indicates overall survival; DSS, disease-specific survival; RT, radiotherapy; CMT, combined modality local therapy (surgery + RT or RT + surgery). * Tumor size was not available for 18 patients. † Tumor grade was not available for 59 patients.
Entire cohort	119	63%		67%
Tumor size, cm
≤5	48 (48)*	69%	.09	74%	.21
>5	53 (52)	60%		64%
Prior attempts at definitive resection
No	105 (88)	66%	.57	71%	.11
Yes	14 (12)	37%		37%
Grade
Low/intermediate	22 (37)†	90%	.004	95%	.04
High	38 (63)	49%		52%
Soft tissue extension
No	38 (32)	73%	.20	75%	.29
Yes	81 (68)	58%		62%
RT-associated tumor
No	103 (87)	65%	.006	68%	.05
Yes	16 (13)	48%		60%
Surgical resection margin
Negative	81 (68)	71%	.001	74%	.001
Positive/unknown	38 (32)	45%		49%
Primary local management
Surgery	92 (77)	58%	.59	62%	.19
CMT	27 (23)	80%		83%
Chemotherapy
No	56 (47)	70%	.41	74%	.24
Yes	63 (53)	57%		60%
Primary local management stratified by margin status
Negative surgical margin
Surgery	65 (55)	69%	.51	72%	.45
CMT	16 (13)	82%		88%
Positive/unknown surgical margin
Surgery	27 (23)	31%	.02	35%	.02
CMT	11 (9)	80%		80%

On the multivariate analysis for OS, resection margin status was the only factor found to be predictive of OS (P = .002). However, when an interaction term was included in the model for the use of RT by resection margin status, margin status was no longer associated with OS and the interaction term of RT by margin status was found to be significantly associated with OS (P < .0001). This indicates that there was an effect modification on OS through the use of RT with respect to resection margin status; specifically, the use of RT for patients with positive resection margins resulted in improved OS.

Patterns of Recurrence

Of 119 patients, 55 (46%) developed disease recurrence (Table 2). The actuarial LC rates at 5 years and 10 years were 59% and 57%, respectively. The interval to the development of LR ranged from 1 to 65 months (median, 7.9 months) after the date of surgical resection, and 98% of LRs had occurred by 5 years. Twenty-five patients (21%) developed DM. Of these, 20 had DM to 1 site (15 to the lung, 4 to bone, and 1 to the brain) and 5 patients had DM to multiple sites (4 to the lung and bone and 1 to the lung, bone, and brain). The actuarial rate of DM development at 5 years and 10 years was 26% and 28%, respectively. The interval to the development of DM was 0.3 months to 14.4 years (median, 23 months), and 88% of DMs had occurred by 5 years. On univariate and multivariate analyses, the only factor that was found to be significantly associated with the development of DM was the presence of LR (P = .001).

Table 2. Patterns of Disease Recurrence

	No. of Patients (%)
Site	First Disease Recurrence	Subsequent Disease Recurrence
LR indicates local disease recurrence; LN, lymph node recurrence; DM, distant metastases.
LR	41 (34)	3 (3)
LN	0	1 (<1)
DM	14 (12)	11 (9)
LR + DM	0	—

LR and the Role of RT

Of the 27 patients who were treated with RT combined with surgery as their primary local management, 6 (22%) experienced LR, and all did so within the irradiation field.

On univariate analyses (Table 3), the following characteristics were found to be significantly associated with inferior LC rates: male sex, tumor size >5 cm, having undergone prior attempts at definitive resection, and positive/unknown resection margin status. In a stratified analysis of the use of combined modality local therapy (CMT; surgery followed by RT or RT followed by surgery) versus surgery alone by resection margin status, there was no significant difference in LC noted in patients with negative resection margins, regardless of whether local therapy was CMT or surgery alone (P = .48). However, for patients with positive resection margins, the addition of RT to the local management was found to be significantly associated with improved LC rates compared with surgery alone (P = .006) (Fig. 2).Stratified analysis of the effect of RT on tumor size revealed that patients with a tumor measuring >5 cm had superior LC when they were treated with CMT compared with surgery alone (67% vs 41%; P = .03). There was no significant difference in LC noted with respect to the addition of RT versus surgery alone for patients with extension into soft tissue.

Figure 2. (a) Local control for patients with negative resection margins. (b) Local control for patients with positive/uncertain resection margins. RT indicates radiotherapy.

Table 3. Local Control: Univariate Analysis for All 119 Patients

Characteristic	No. (% of Total)	5-year LC	P Value	5-year DMFS	P Value
LC indicates local control rate; DMFS, distant metastasis-free survival, NS, not significant (P >.05); RT, radiotherapy; CMT, combined modality local therapy (surgery + RT or RT + surgery). * Tumor size was not available for 18 patients. † Tumor grade was not available for 59 patients.
Entire cohort	119	59%		74%
Sex
Male	62 (52)	50%	.04	73%	.60
Female	57 (48)	69%		74%
Tumor size, cm
≤5	48 (48)*	72%	.02	78%	.26
>5	53 (52)	48%		68%
Site
Maxilla	48 (40)	62%	NS	81%	NS
Mandible	54 (45)	61%		68%
Other	17 (14)	42%		68%
Prior attempts at definitive resection
No	105 (88)	63%	.02	60%	.81
Yes	14 (12)	32%		75%
Soft tissue extension
No	38 (32)	64%	.40	86%	.08
Yes	81 (68)	57%		66%
RT-associated tumor
No	103 (87)	62%	.04	73%	.63
Yes	16 (13)	40%		83%
Grade
Low/intermediate	22 (37)†	74%	.08	100%	.07
High	38 (63)	56%		67%
Surgical resection margin
Negative	81 (68)	68%	.003	75%	.71
Positive/unknown	38 (32)	39%		70%
Primary local management
Surgery	92 (77)	55%	.08	71%	.98
CMT	27 (23)	73%		81%
Chemotherapy
No	56 (47)	70%	.03	70%	.35
Yes	63 (53)	50%		77%
Primary local management stratified by margin status
Negative surgical margin
Surgery	65 (55)	67%	.48	74%	.85
CMT	16 (13)	73%		80%
Positive/unknown surgical margin
Surgery	27 (23)	24%	.006	57%	.57
CMT	11 (9)	75%		80%

On the multivariate analysis, tumor size >5 cm (P = .01), positive/uncertain resection margins (P = .008), and local management with surgery alone (P = .008) were found to be significantly associated with inferior LC. However, when multivariate analysis was performed including an interaction term for the use of RT by resection margin status, only this interaction term was found to be significantly associated with LC outcome (P = .001). This indicated that the use of RT mitigated the influence of positive resection margins on risk of LR, but that RT was without significant effect on LR when resection margins were negative.

Outcomes After Recurrence

At 5 years and 10 years after recurrence (either LR or DM), OS was 27% and 17%, respectively. Similarly, DSS was also 27% and 17% at those respective time points. There was no significant difference in DSS noted regardless of whether the recurrence was isolated LR versus DM (26% vs 29% at 5 years, respectively; P = .48). Thus, isolated LR had a similarly poor prognosis when compared with DM.

Of the 41 patients with isolated LR, 38 received treatment and only 4 were successfully salvaged. No patient whose salvage therapy did not include surgery after LR was salvaged.

Complications From Treatment

Twenty-nine patients (24%) experience complications from surgery. The median time to surgical complications was 3.9 months (range, 0-22.5 months). The actuarial rate of surgical complications was 28% at ≥5 years. Surgical complications were as follows: flap failure (7 patients [5 complete and 2 partial]), fistula (7 patients), infection in reconstruction bed/flap/skin graft (4 patients, 3 of whom required debridement), nerve damage (3 patients), soft tissue necrosis (1 patient), exposure of mandibular hardware requiring surgical removal (1 patient), ill-fitting mandibular prosthesis requiring multiple plastics procedures and eventual removal (1 patient), exposure keratitis because of ectroprion requiring surgical closure of the eyelid (1 patient), hypertrophic scar requiring injections and excision (1 patient), mandibular plate infection requiring surgical removal (1 patient), insufficient mandibular vestibuloplasty with malocclusion ultimately requiring a bone graft (1 patient), and eustachian tube dysfunction with chronic otitis media (1 patient). Complications were considered mild in 1 patient, moderate in 8 patients, and severe in 20 patients.

Of the 27 patients who received RT as part of their local management, 8 (30%) experienced RT-induced complications. The median time to the development of RT-induced complications was 43 months (range, 4‒63 months). The actuarial rates of RT-induced complications at 5 years and 10 years were 40% and 47%, respectively. The RT-induced complications were as follows: soft tissue necrosis (2 patients), osteoradionecrosis (2 patients), xerostomia (1 patient), cataract (1 patient), hypopituitarism (1 patient), and chronic serous otitis media with hearing loss (1 patient). Complications were considered mild in 2 patients, moderate in 1 patient, and severe in 5 patients. The actuarial rates of moderate/severe RT-induced complications at 5 years and 10 years were 32% and 39%, respectively. There was a significant correlation found between the development of surgical complications and RT-induced complications (P = .04).

Of 116 assessable patients, 110 (95%) retained the ability to take solid foods orally >6 months from primary local management. Two (2%) patients required a liquid diet and 4 (3%) were dependent on a gastrostomy tube. Two patients (2%) were dependent on a tracheotomy tube. Speech function was oral in 117 (98%) patients and oral with a speech aid in 2 (2%) patients.

DISCUSSION

The results of the current analysis demonstrate that RT improves LC, DSS, and OS for patients with OHN who have positive or uncertain resection margins after surgery. Furthermore, it was found that although less than one‒quarter of patients (21%) developed DM, 40% experienced LR; and isolated LRs portended as poor a prognosis for DSS as did the development of DM. Only a small minority of patients with LR were successfully salvaged, thereby emphasizing the critical role primary local management plays in outcomes for this disease.

Other studies have demonstrated the importance of resection margin status in the outcomes of patients with OHN, but to our knowledge none have specifically addressed the role of RT. Investigators from the Canadian Society of Otolaryngology-Head and Neck Surgery Oncology Study Group reported that positive resection margins were significantly associated with poorer DFS and OS.9 Similarly, investigators from the Massachusetts General Hospital,10 Johns Hopkins University,12 and the Dutch Head and Neck Oncology Group7 reported inferior OS when the margins of resection were positive. A report on 44 patients treated at Memorial Sloan-Kettering Cancer Center also found positive resection margins to be predictive of worse DFS.8 To our knowledge, the Dutch study was the only other study to specifically address the role of RT for 18 patients with positive or uncertain resection margins and although they reported a trend toward improved LC with RT for those patients, the effect was not found to be statistically significant in their small study cohort.7 The other studies either did not specifically analyze the role of RT with respect to outcomes8, 10 or included patients who did not receive surgery (eg, RT was the primary LC modality) in their analyses, indicating no significant effect of RT.9, 12 Our unique finding of LC and survival benefit from RT for patients with positive resection margins may be attributable to the fact that we analyzed outcomes only for those patients who underwent macroscopic total resection with curative intent in an attempt to elucidate the role of RT as an adjuvant to surgery for patients with localized disease.

Our findings confirm that OHN exhibits clinical behavior that is distinct from osteosarcoma of the extremities and trunk. Although osteosarcomas of the extremity and trunk have a peak incidence in the second decade of life,1 OHN presents later, with average age at presentation of 35 years in the current series.4, 7-12, 18 Patients with osteosarcoma of the extremities and trunk have high incidence of micrometastases and a comparatively high incidence of overt metastases,19 and clearly benefit from the early administration of chemotherapy.20-23 Our finding of a relatively low rate of DM is consistent with reports of other investigations of OHN.3, 7, 8, 10 Although the use of chemotherapy did not appear to impact outcomes in the current series, its usefulness in patients with OHN is a matter of debate7, 8, 10 because other investigators have reported a benefit with adjuvant chemotherapy.9, 18 However, the data from the current study and others have shown that local failure is the major concern in patients with OHN and poses the greater threat to survival.4, 7, 9, 10

The results of the current study indicate that surgical and RT-related complication rates are high for patients regardless of whether they are treated with surgery or a combination of RT and surgery. However, these tumors are situated in anatomic locations that are in close proximity to critical normal structures in regions that are essential to swallowing, as well as auditory, visual, and respiratory functions. Moreover, studies have consistently demonstrated that radical surgery is the mainstay for the definitive treatment of OHN and that survival cannot be expected for those patients who do not undergo any attempt at complete resection.3, 4, 6, 7, 9-11 Therefore, the risk of surgical morbidity must be balanced against the need to eradicate disease. RT added to surgery compounds the risk of treatment-related complications, as shown in the data reported herein. However, newer techniques may improve the complication rates associated with RT because many of these patients were treated with RT in an era before 3-dimensional conformal or intensity‒modulated RT planning.

We acknowledge the limitations inherent in the retrospective nature of this review. The patients received heterogeneous management with respect to the addition of adjuvant therapies. Selection bias likely occurred such that patients with tumors that were more difficult to extirpate at the time of presentation were referred for neoadjuvant chemotherapy with the hope of affecting tumor shrinkage before resection. Thus, patients with the most adverse presentations of disease received chemotherapy and this may explain why patients who received chemotherapy had significantly inferior LC.

In conclusion, RT was found to improve LC, DSS, and OS for patients with osteosarcoma of the craniofacial bones who had positive/uncertain resection margins after surgical resection. A clinical implication from this analysis is that patients with negative resection margins do not benefit from the addition of RT to their primary local management. For patients who have positive margins after resection, postoperative RT is recommended at our institution. The RT dose is 64 to 68 Gy to the area of highest risk (resection margins), respecting tolerances of critical normal tissues, with lower dose levels (50 Gy and 60 Gy, respectively) being administered to the soft tissue flap and resection bed as deemed appropriate for each case. Overall treatment-related complication rates were high. However, LR is highly lethal in patients with this disease. The morbidity risks of aggressive local management must be balanced against the poor prognosis associated with failure to eradicate the primary tumor.

Conflict of Interest Disclosures

The authors made no disclosures