Key points
- •
Sinonasal and ventral skull base malignancies are rare and this has made it difficult to conduct randomized controlled trials. Much knowledge of the clinical outcomes for these malignancies is based on retrospective chart review studies.
- •
Overall survival for sinonasal and ventral skull base malignancies remains poor.
- •
For most histologies, primary treatment with surgical resection with or without adjuvant radiotherapy provides the best survival outcome.
| AC | Adenocarcinoma |
| ACC | Adenoid cystic carcinoma |
| DLBCL | Diffuse large B-cell lymphoma |
| DSS | Disease-specific survival |
| ENKTL | Extranodal natural killer/T-cell lymphoma |
| EP | Extramedullary plasmacytoma |
| LRC | Locoregional control |
| MM | Mucosal melanoma |
| NC | Neuroendocrine carcinoma |
| ON | Olfactory neuroblastoma |
| OS | Overall survival |
| PFS | Progression-free survival |
| QOL | Quality-of-life |
| RFS | Recurrence-free survival |
| RS | Relative survival |
| SCC | Squamous cell carcinoma |
Introduction
Sinonasal and ventral skull base malignancies are uncommon, and this has made it difficult to conduct randomized controlled trials. Much of what is known about the outcomes of these malignancies is based on retrospective, single-institution, or population-based database studies. Population-based databases, such as the Surveillance, Epidemiology, and End Results (SEER) and National Cancer Database (NCDB), allow researchers to pool cases from many institutions to study the behavior of these malignancies. They have significantly expanded the knowledge base on sinonasal and ventral skull base malignancies. However, with regard to outcomes research, these population-database studies have some inherent limitations that necessitate cautious interpretation of their findings.
SEER and NCDB capture approximately 26% and 70% of new cancer diagnoses in the United States, respectively. Therefore, a certain degree of selection bias may exist because cases reported in surveyed areas may not be representative of the entire population. For example, the SEER database collects information primarily from urban areas, where there may be a higher proportion of patients with lower socioeconomic status. Although single-institutional retrospective studies are also susceptible to this bias and many other types of selection bias, population-based studies have additional disadvantages that make it difficult to generalize some of their findings. The information in the databases is derived from the work of many different clinicians and pathologists, and the information is coded into the database by many different people, which may lead to inconsistencies in reporting. In particular, SEER lacks certain details of treatment, such as chemotherapy, the dose of radiotherapy, type of surgical treatment, tumor margins, and complications of treatment. Additionally, the databases do not contain information on the clinical reasoning that may be associated with treatment decisions. For example, in SEER, the intent of radiotherapy is not specified; radiotherapy with curative intent is indistinguishable from palliative radiotherapy. Retrospective chart reviews allow researchers to be able to take such nuances into consideration. Furthermore, death is not the only outcome of significance in oncology. For many of the sinonasal malignancies, recurrence is a key event, causing substantial morbidity even in the absence of mortality. In fact, morbidity is neglected altogether in the SEER database. This may result in studies underestimating the burden of disease for insidious malignancies with devastating local effects.
With regard to survival, the use of these databases presents additional challenges. At tertiary referral centers, the source of most studies not from databases, academic physicians are usually aware of the value of reporting results for rare malignancies and there may be a greater incentive to follow patients for a long period of time. In many population-based survival analyses, a large portion of patients are censored after a short time period. This has the potential to compromise the precision of the reported survival rates.
Another factor to consider is user error. There is a steep learning curve to understanding these databases. Accessing the right data and performing the appropriate analysis can be challenging because there are many subtleties to the use of software used to access the data. For example, SEERStat, the software used for SEER data, has hundreds of selections and settings. Many of these settings, although they may seem minor, can influence the results of a study. Furthermore, the databases undergo frequent changes, and information has to be recoded. This further complicates the coding system; several manuals and appendices may be required to make sense of the database.
Therefore, this article presents a dedicated discussion of outcomes for patients with sinonasal and ventral skull base malignancies organized by histology, without using population-based database study results. It focuses on the outcomes of death and recurrence, using studies in the literature that report survival rates and locoregional control (LRC) rates. Quality-of-life (QOL) outcomes and the side effects of certain treatments are also discussed. Factors that affect these outcome measures are highlighted.
Because there are many ways of calculating survival rates, it is not always possible to perform like-with-like comparisons. For example, some studies report overall survival (OS), whereas others report relative survival (RS) and/or disease-specific survival (DSS). RS takes into account the expected survival rate of people with demographic characteristics similar to the patient. There are several ways of calculating it, and researchers are not always familiar with the complexity of the underlying concepts. Moreover, the methods are not always fully reported in the literature. Articles sometimes fail to mention whether the actuarial or Kaplan-Meier model was used, or whether the Ederer II or Hakulinen method was used to calculate expected cumulative survival. It then becomes difficult to make an analogous comparison of survival outcomes.
In DSS, also known as cancer or carcinoma-specific survival, the outcome event is death from cancer. However, the cause of death is not always so clear. This introduces an interesting quandary, that is, the inability to agree on the definition of death. The term bias of cause-of-death interpretation has been used to describe this significant interobserver variability. There are plenty of conceivable scenarios in which the cause of death may not be clear-cut. Given that patients with head and neck cancer may have an increased risk of suicide compared with the general population, to what is the death of a patient who commits suicide attributed? If a patient with head and neck cancer dies of aspiration pneumonia, should it be assumed that this was a direct consequence of the cancer and, therefore, should be coded as a cancer-related death?
Despite the limitations of these survival metrics, studies can be meaningful when considered in the context of the larger body of literature. Therefore, survival outcomes should be thought of as estimates rather than precise measurements, and clinicians should maintain a healthy level of skepticism until multiple investigations confirm similar results.
Introduction
Sinonasal and ventral skull base malignancies are uncommon, and this has made it difficult to conduct randomized controlled trials. Much of what is known about the outcomes of these malignancies is based on retrospective, single-institution, or population-based database studies. Population-based databases, such as the Surveillance, Epidemiology, and End Results (SEER) and National Cancer Database (NCDB), allow researchers to pool cases from many institutions to study the behavior of these malignancies. They have significantly expanded the knowledge base on sinonasal and ventral skull base malignancies. However, with regard to outcomes research, these population-database studies have some inherent limitations that necessitate cautious interpretation of their findings.
SEER and NCDB capture approximately 26% and 70% of new cancer diagnoses in the United States, respectively. Therefore, a certain degree of selection bias may exist because cases reported in surveyed areas may not be representative of the entire population. For example, the SEER database collects information primarily from urban areas, where there may be a higher proportion of patients with lower socioeconomic status. Although single-institutional retrospective studies are also susceptible to this bias and many other types of selection bias, population-based studies have additional disadvantages that make it difficult to generalize some of their findings. The information in the databases is derived from the work of many different clinicians and pathologists, and the information is coded into the database by many different people, which may lead to inconsistencies in reporting. In particular, SEER lacks certain details of treatment, such as chemotherapy, the dose of radiotherapy, type of surgical treatment, tumor margins, and complications of treatment. Additionally, the databases do not contain information on the clinical reasoning that may be associated with treatment decisions. For example, in SEER, the intent of radiotherapy is not specified; radiotherapy with curative intent is indistinguishable from palliative radiotherapy. Retrospective chart reviews allow researchers to be able to take such nuances into consideration. Furthermore, death is not the only outcome of significance in oncology. For many of the sinonasal malignancies, recurrence is a key event, causing substantial morbidity even in the absence of mortality. In fact, morbidity is neglected altogether in the SEER database. This may result in studies underestimating the burden of disease for insidious malignancies with devastating local effects.
With regard to survival, the use of these databases presents additional challenges. At tertiary referral centers, the source of most studies not from databases, academic physicians are usually aware of the value of reporting results for rare malignancies and there may be a greater incentive to follow patients for a long period of time. In many population-based survival analyses, a large portion of patients are censored after a short time period. This has the potential to compromise the precision of the reported survival rates.
Another factor to consider is user error. There is a steep learning curve to understanding these databases. Accessing the right data and performing the appropriate analysis can be challenging because there are many subtleties to the use of software used to access the data. For example, SEERStat, the software used for SEER data, has hundreds of selections and settings. Many of these settings, although they may seem minor, can influence the results of a study. Furthermore, the databases undergo frequent changes, and information has to be recoded. This further complicates the coding system; several manuals and appendices may be required to make sense of the database.
Therefore, this article presents a dedicated discussion of outcomes for patients with sinonasal and ventral skull base malignancies organized by histology, without using population-based database study results. It focuses on the outcomes of death and recurrence, using studies in the literature that report survival rates and locoregional control (LRC) rates. Quality-of-life (QOL) outcomes and the side effects of certain treatments are also discussed. Factors that affect these outcome measures are highlighted.
Because there are many ways of calculating survival rates, it is not always possible to perform like-with-like comparisons. For example, some studies report overall survival (OS), whereas others report relative survival (RS) and/or disease-specific survival (DSS). RS takes into account the expected survival rate of people with demographic characteristics similar to the patient. There are several ways of calculating it, and researchers are not always familiar with the complexity of the underlying concepts. Moreover, the methods are not always fully reported in the literature. Articles sometimes fail to mention whether the actuarial or Kaplan-Meier model was used, or whether the Ederer II or Hakulinen method was used to calculate expected cumulative survival. It then becomes difficult to make an analogous comparison of survival outcomes.
In DSS, also known as cancer or carcinoma-specific survival, the outcome event is death from cancer. However, the cause of death is not always so clear. This introduces an interesting quandary, that is, the inability to agree on the definition of death. The term bias of cause-of-death interpretation has been used to describe this significant interobserver variability. There are plenty of conceivable scenarios in which the cause of death may not be clear-cut. Given that patients with head and neck cancer may have an increased risk of suicide compared with the general population, to what is the death of a patient who commits suicide attributed? If a patient with head and neck cancer dies of aspiration pneumonia, should it be assumed that this was a direct consequence of the cancer and, therefore, should be coded as a cancer-related death?
Despite the limitations of these survival metrics, studies can be meaningful when considered in the context of the larger body of literature. Therefore, survival outcomes should be thought of as estimates rather than precise measurements, and clinicians should maintain a healthy level of skepticism until multiple investigations confirm similar results.
Overall outcomes
As a whole, malignancy of the sinonasal cavity and ventral skull base portends a poor prognosis. In a report of 220 cases from 2 institutions and meta-analysis of 16,396 cases, a 41% 5-year OS was reported for the meta-analysis group. When cases were broken down by subsite, 5-year OS for the nasal cavity (1960s: 63%; 1970s; 54%; 1980s: 59%; 1990s: 66%) was higher than for the ethmoid (1960s: 27%; 1970s; 37%; 1980s: 56%; 1990s: 51%) and maxillary (1960s: 26%; 1970s; 31%; 1980s: 39%; 1990s: 45%) sinuses. Over these 4 decades, survival for the maxillary and ethmoid sinuses had improved but survival for the nasal cavity remained relatively unchanged.
In the same study, laterality of tumors was also found to affect survival. LRC and DSS at 5-years was highest for tumors with right-sided involvement (70% and 73%, respectively), followed by left-sided involvement (53% and 57%, respectively) and bilateral involvement (25% and 30%, respectively). Increasing tumor stage (T-stage) is associated with decreasing 5-year OS: T1 = 91%, T2 = 64%, T3 = 72%, and T4 = 49%. Demographic determinants of survival were not as clear. The study found that, in the 220 institutional cases, women had better 5-year LRC and DSS when compared with men (69% and 72% vs 53% and 57%, respectively).
Extension of malignancies to the pterygomaxillary fossa, frontal sinus, and sphenoid sinus; erosion of the cribriform plate; and invasion of the dura were factors associated with poor survival outcomes. Although orbital extension may lead to blindness and decrease QOL, it was not clearly found to be associated with decreased survival.
Although the purpose of this article is not to discuss treatment options in detail, treatment-specific survival outcomes are worth examining. In the meta-analysis, the surgery-alone group (70% 5-year DSS) had the highest survival, followed by the surgery and radiotherapy group (56% 5-year DSS), and the radiotherapy-alone group (33% 5-year DSS). Patients receiving chemotherapy, either alone or in combination with other treatment modalities, had 42% 5-year DSS. However, these patients were not randomized into treatment groups. Therefore, patients with localized or smaller lesions may have been more likely to be treated with surgery alone, accounting for the difference in survival seen between the treatment modalities. In the meta-analysis, 13% of patients developed metastasis to the neck lymph nodes and 5% of patients developed lymph node recurrence. Those with primary neck metastasis had a 32% 5-year OS, and those with metastasis to the neck after treatment had a 25% 5-year OS. Treatment of patients with local recurrence was successful 16% of the time. Several studies have expectedly reported decreased survival with increased tumor size and 1 study found that radiological tumor volume may also predict progression-free survival (PFS).
In patients treated with craniofacial surgery (CFS), high rates of postoperative complications have been reported. An international collaborative study of more than 1307 cases treated with CFS has reported any complications in 33%, wound complications in 18%, central nervous system complications in 15%, systemic complications in 4%, orbital complications in 1%, and mortality in 4% of cases. Endoscopic endonasal surgery (EES), which has been promoted as a potentially equally effective treatment option for selected cases, is associated with a shorter hospital stay, although several studies have found no statistically significant difference in complication rates.
Intensity-modulated radiation therapy (IMRT) and proton beam therapy (PBT) are frequently used to treat sinonasal malignancies, either as primary therapy or adjuvant postoperative therapy. In one study using IMRT, PFS was 47% and LRC was 51% at 5 years. Higher doses of radiation (≥60 Gy) were associated with improved PFS and LRC. One study looking at PBT found an LRC rate of 80%, and an OS rate of 47% at 5 years. In that study, smoking status was found to predict worse LRC; 5-year LRC for current smokers was significantly lower than for nonsmokers (23% vs 83%).
Research on QOL outcomes for patients with sinonasal malignancies has focused on the use of surveys such as the Anterior Skull Base Questionnaire, Sino-Nasal Outcome Test (SNOT), and Rhinosinusitis Outcome Measure. A systematic review has suggested that surgical resection is associated with an increase in QOL. Furthermore, the endoscopic approach seems to be associated with an earlier and greater increase in QOL when compared with open methods. A recent, prospective, SNOT-based study looking at patients with sinonasal cancer undergoing endoscopic resection has reported significant improvement in the psychological and sleep domains of the SNOT scores after 2 years but no significant improvement in rhinologic subdomain and overall SNOT scores.
Another systematic review, focusing on the sinonasal morbidity associated with EES, found that nasal crusting (51%), discharge (40%), airflow blockage, and disturbances in olfaction are the most commonly associated morbidities. Fortunately, most of these symptoms resolve in 3 to 4 months, although symptoms can persist for longer in more complex surgeries. Mucocele formation was seen in 8% of cases but this rate may have been between 14% and 50% in pediatric patients.
Histology-specific outcomes
Squamous Cell Carcinoma
Most of the available studies not from databases on sinonasal malignancies include multiple histologies. Although several studies focusing exclusively on sinonasal squamous cell carcinoma (SCC) have been published, some of these studies focused on a single subsite and most have small sample sizes. In the meta-analysis previously mentioned, 58% of cases were SCCs. Although survival for all other histologies remained stable from 1950 to 1990, it improved for SCC. Five-year OS for SCCs diagnosed in the 1960s, 1970s, 1980s, and 1990s was 25%, 34%, 45%, and 50%, respectively. In the same study, for the 126 institutional cases of SCC, LRC at 2, 5, and 10 years was 61%, 58%, and 56%, respectively; DSS at 2, 5, and 10 years was 73%, 60%, and 59%, respectively.
In a report of 33 sinonasal SCC cases, at 5-years, DSS and OS were 46% and 40%, respectively. Surgical resection followed by radiotherapy was associated with higher DSS when compared with surgery alone, regardless of stage, suggesting that postoperative radiotherapy may improve 5-year survival. Compared with other histologies, neck lymph node recurrence seems to be more common in SCC. In maxillary sinus SCC, 18.7% of cases have neck lymph node recurrence at 2 years and 20.7% of cases have neck lymph node recurrence at 5 years. In patients with maxillary sinus SCC, involvement of the alveolus and cheek may increase the risk of neck recurrence even further.
Histologic variants of sinonasal SCC may have different survival rates. Due to their rarity, few studies have reported specific survival rates for sinonasal SCC variants. However, results from studies on head and neck SCC variants may be used to estimate prognoses for the different variants. Verrucous SCC, a low-grade, human papillomavirus–associated variant, tends to have a more favorable prognosis, whereas basaloid SCC tends to have poorer prognosis. Papillary SCC and sarcomatoid SCC, despite having high locoregional recurrence rates, tend to have higher survival rates. Progression to sinonasal SCC from an inverted papilloma may be another negative prognosticator for SCC. One study of 87 such cases reported a 39.6% 5-year OS and 30.7% 10-year OS. For these tumors, advanced stage, metachronous tumors, and extension to the cranial base or orbit were found to be associated with decreased survival.
Adenocarcinoma
Sinonasal adenocarcinoma (AC), which is thought to be the second or third most common primary sinonasal malignancy, is relatively well-studied. One multicenter study from France has reported on 418 cases. Based on the cell of origin, it may be classified as salivary or nonsalivary type. The nonsalivary type may be classified as either nonintestinal type AC or as intestinal-type AC (ITAC), which has been heavily associated with occupational exposures. ITAC may further be classified into the papillary, colonic, mucinous, solid, or mixed subtypes.
Survival rates for sinonasal AC is thought to be higher than for SCC. According to 1 study, 5-year DSS is 78% for AC, and 60% for SCC. The LRC rate at 5 years is 69% for sinonasal AC. Another study, with 66 cases, reported similar values. Five-year OS and DSS for sinonasal AC were 66% and 79%, respectively. The large, multicenter study previously mentioned also reported a similar rate; 5-year OS was 64%. Several studies have found higher T-stage to predict poor prognosis and, as in SCC, tumor volume was found to be inversely related to survival. Local recurrence, positive surgical margins, female gender, and advanced age (especially when treated with craniofacial resection) have been associated with decreased survival. EES and open surgical methods were found to have comparable outcomes.
Histologic subtype also seems to be predictive of survival. Poorly differentiated subtypes, such as the solid and mucinous subtypes of ITAC, have been associated with lower survival compared with the papillary and mucinous subtypes. The cause may also affect survival because cases related to wood-dust exposure have been shown to have higher survival. Immunostaining positive for Ki-67 or CD31, which indicate high tumor growth fraction and microvessel density, respectively, have been associated with lower survival. A more extensive list of histopathologic and molecular or genetic factors that may be associated with decreased survival for sinonasal malignancies can be found in Table 1 .
