Marital status, treatment, and survival in patients with glioblastoma multiforme

A population-based study

Authors

  • Susan M. Chang M.D.,

    1. Department of Neurological Surgery, University of California, San Francisco, San Francisco, California
    Search for more papers by this author
  • Fred G. Barker II M.D.

    Corresponding author
    1. Brain Tumor Center, Neurosurgical Service, Massachusetts General Hospital, Boston, Massachusetts
    2. Division of Neurosurgery, Department of Surgery, Harvard Medical School, Boston, Massachusetts
    • Brain Tumor Center, Massachusetts General Hospital, Fruit Street, Boston, MA 02114
    Search for more papers by this author
    • Fax: (617) 724-8769


Abstract

BACKGROUND

Social factors influence cancer treatment choices, potentially affecting patient survival. In the current study, the authors studied the interrelations between marital status, treatment received, and survival in patients with glioblastoma multiforme (GM), using population-based data.

METHODS

The data source was the Surveillance, Epidemiology, and End Results (SEER) Public Use Database, 1988–2001, 2004 release, all registries. Multivariate logistic, ordinal, and Cox regression analyses adjusted for demographic and clinical variables were used.

RESULTS

Of 10,987 patients with GM, 67% were married, 31% were unmarried, and 2% were of unknown marital status. Tumors were slightly larger at the time of diagnosis in unmarried patients (49% of unmarried patients had tumors larger than 45 mm vs. 45% of married patients; P = 0.004, multivariate analysis). Unmarried patients were less likely to undergo surgical resection (vs. biopsy; 75% of unmarried patients vs. 78% of married patients) and were less likely to receive postoperative radiation therapy (RT) (70% of unmarried patients vs. 79% of married patients). On multivariate analysis, the odds ratio (OR) for resection (vs. biopsy) in unmarried patients was 0.88 (95% confidence interval [95% CI], 0.79–0.98; P = 0.02), and the OR for RT in unmarried patients was 0.69 (95% CI, 0.62–0.77; P < 0.001). Unmarried patients more often refused both surgical resection and RT. Unmarried patients who underwent surgical resection and RT were found to have a shorter survival than similarly treated married patients (hazard ratio for unmarried patients, 1.10; P = 0.003).

CONCLUSIONS

Unmarried patients with GM presented with larger tumors, were less likely to undergo both surgical resection and postoperative RT, and had a shorter survival after diagnosis when compared with married patients, even after adjustment for treatment and other prognostic factors. Cancer 2005. © 2005 American Cancer Society.

Glioblastoma multiforme (GM) is the most common primary brain tumor in adults and has the worst prognosis; the median survival is less than 1 year, with no known curable stage of the disease. Treatment for most patients consists of surgical resection (or biopsy if resection is considered unsafe) followed by radiation therapy.1, 2 Many patients also receive chemotherapy, with a modest benefit reported.3 Patient-related factors that are reported to be potentially prognostic for survival include age, tumor size, tumor site and histology, and Karnofsky performance status.1

The importance of the correlations between social factors and the diagnosis and treatment of cancer, as well as the survival of cancer patients, is being increasingly recognized. Social factors that have been studied to date include race, socioeconomic status or position, education, and marital status.4, 5 Goodwin et al.6 reported that, for some cancers, married persons presented with a less advanced stage of disease, were more likely to receive aggressive treatment, and lived longer than unmarried persons. These findings have been confirmed and extended for many types of cancer.7–13 In brain tumor management, social support networks have special importance because the patients themselves may not recognize subtle changes in function or may lose the ability to manage his or her own care early in the course of the disease, through cognitive decline or physical impairment. Age, gender, race, and time period of diagnosis have all been previously explored as possible prognostic factors in population-based studies.14, 15 However, to our knowledge, marital status has not previously been explored as a predictor of received treatment or survival in patients with primary brain tumors.

In the current study, we used a population-based data source, the Surveillance, Epidemiology, and End Results (SEER) database, to explore the correlations between marital status, treatment, and survival in patients in the U.S. with GM. Specifically, we compared tumor size at the time of diagnosis, rates of surgical resection and of receipt of postoperative radiation therapy, and survival after diagnosis between married and unmarried patients with GM.

MATERIALS AND METHODS

Data Source

The data source for this study was the SEER Public Use Database, maintained by the National Cancer Institute, Division of Cancer Control and Population Sciences, Surveillance Research Program, Cancer Statistics Branch (April 2004 release, based on November 2003 data submission). Data from all 11 SEER geographic registries plus the supplemental Alaska Native Tumor Registry were used. Each geographic SEER registry routinely collects information concerning cancer patients within a defined area in the U.S., including patient demographic data, primary tumor site, morphology, stage of disease at diagnosis, first course of treatment (within 4 months of diagnosis), and patients' status at follow-up. Because variables such as age, gender, race, socioeconomic status, geographic region, and year of diagnosis are expected to be correlated with marital status, and many of these have been shown to be correlated with patient survival,14, 15 these variables were included in all described multivariate analyses. An overview of the SEER Database is available at URL: http://seer.cancer.gov/about/ [accessed August 13, 2005].

Inclusion Criteria and Definition of Covariates and Endpoints

Patients included in this analysis had new diagnoses of GM (International Classification of Diseases-Oncology [ICD-O]-3rd edition; code 9440) or a subtype of GM (giant cell GM or gliosarcoma; ICD-O-3 codes 9441–9442) between 1988–2001, were age 18 years or older at the time of diagnosis, and had tumors coded as supratentorial (ICD-O-3 topography codes C71.0–71.5, C71.8–71.9). Except when specified, all patients had histologic confirmation of diagnosis and were actively followed for survival (autopsy-only and death certificate-only cases were excluded). In addition to the main analysis, a separate analysis of factors (including marital status) that could potentially affect histologic diagnosis during life was performed in all patients, including those with autopsy-only or death certificate-only diagnoses. For four persons included twice in the SEER data who had two primary GM tumors coded at the same topographic site, only one diagnosis was included in the analysis. Survival was measured as months after diagnosis.

Patient age at diagnosis, gender, and race (recoded by SEER as white, black, other, and unknown) were included in the SEER data,16 as was the year of diagnosis. ICD-O-3 codes were used to identify the site and histology of the primary tumor.17 As a proxy indicator of socioeconomic status, the 1993 U.S. Census estimate of the percentage of the population at or below the poverty level in the county of each patient's residence (for five patients from Alaska, the statewide estimate) was used.18 These were ranked by quartile as follows: fewer than 8.2%, 8.2–10.9%, 11–18%, or ≥ 18.1% of residents at or below the poverty level.

Marital status at the time of diagnosis was coded in the SEER data as single (never married), married (including common-law marriage), separated, divorced, widowed, and unknown. We recoded marital status as a binary variable by pooling the SEER classes single, separated, divorced, and widowed as “Unmarried.” Persons with unknown marital status (2%) were excluded from analyses.

For analysis of tumor size, the SEER category of Size of Primary Tumor (at the time of diagnosis; coded in mm) was used. Tumors coded as larger than 9 cm (43 patients, 0.4%) were recoded as 9 cm for these analyses. Tumor size data were highly skewed, and values suggested that they had been rounded to the nearest 5 mm. For multivariate analysis of tumor size as a dependent variable, proportional odds ordinal logistic regression was used, with tumor size recoded as quartiles. Because tumor size was missing for greater than 40% of the patients, models for surgical resection, radiation treatment, and survival were presented both for the whole group and for the subgroup with known tumor size. The extent of surgical resection was recoded from SEER data to a two-level scale (biopsy or resection) as follows: for patients diagnosed between 1988–1997, biopsy was SEER codes 2, 5, and 6, and resection was SEER codes 10, 20, 30, 35, 40, 50, 55, and 60; for those patients diagnosed between 1998–2001, biopsy was SEER code 0 and resection was SEER codes 10–60. Patients with an unknown extent of surgical resection were omitted from these analyses. The reason surgical resection was not performed was coded in SEER data as not recommended, contraindicated due to other conditions, refused, or unknown. Postoperative radiation therapy was recoded from SEER data as performed (included external beam radiation, with or without implants or isotopes) or not performed (none or refused). Patients for whom it was unknown whether radiation therapy was actually administered, patients who received nonstandard radiation therapy (implants or isotopes without external beam radiation), or patients for whom radiation therapy methods were not specified were omitted from these analyses. For surgery and radiation therapy analyses, logistic regression was used to allow multivariate adjustment for possible confounders.

Statistical Methods

Statistical methods included the chi-square, Kruskal–Wallis, and Wilcoxon rank sum tests; multivariate logistic regression; and multivariate proportional odds ordinal logistic regression.19 Survival analysis was performed using Kaplan–Meier product limit estimators and Cox univariate and multivariate proportional hazards models.20

Calculations were performed using SEER*Stat (version 5.2.2; Surveillance Research Program, National Cancer Institute, Bethesda, MD– available at URL: www.seer.cancer.gov/seerstat [accessed August 13, 2005]), SAS (version 8.2; SAS Institute Inc., Cary, NC), and S-plus (Version 3.3 for Windows; Insightful, Inc., Seattle, WA) software with the Hmisc and Design modeling function software libraries written by Harrell.20, 21 For curve fitting for graphic display, the Locfit local likelihood regression functions written by Loader22, 23 were used. The P values shown are two-tailed.

RESULTS

There were 10,987 adult patients in the SEER database with supratentorial glioblastomas diagnosed between 1988–2001 and for whom histologic confirmation was made during life. The clinical characteristics of the patients are shown in Table 1. The majority of the patients were white, with an age range of 53–72 years; there were more men than women. Two-thirds of the patients were married. The majority of the tumors were GM, rather than other WHO Grade 4 histologic subtypes. Two-third of the tumors were located in the frontal, temporal, parietal, or occipital lobes. Approximately 73% of the patients underwent surgical resection. Approximately 73% received external beam radiation therapy after surgery; 2% of these patients received additional radiation therapy using implants or isotopes. The median survival for the entire group was 7 months; the 1-year survival was 26%, the 2-year survival was 7.3%, and the 5-year survival was 2.4%.

Table 1. Clinical Characteristics of 10,987 Patients with Supratentorial Glioblastoma
  MarriedUnmarriedP valuea
  • NOS: not otherwise specified.

  • a

    P values are for comparison between married and unmarried groups and represent chi-squared or Kruskal–Wallis tests. Some column totals do not add up to 100% because of rounding.

Age    
Median64 yrs63 yrs66 yrs< 0.001
Interquartile range53–72 yrs53–71 yrs52–75 yrs 
Male gender6305 (57%)65%41%< 0.001
Race   
White10032 (91.3%)92.5%89.2%< 0.001
Black486 (4.4%)3.1%7.1% 
Other454 (4.1%)4.4%3.7% 
Unknown15 (0.1%)0.1%0.1% 
Marital status    
Married (including common law)7396 (67.3%)   
Single (never married)1076 (9.8%)   
Divorced888 (8.1%)   
Separated48 (0.4%)   
Widowed1339 (12.2%)   
Unknown240 (2.2%)   
Percentage at or below poverty level in county of residence    
Median11%11%11%< 0.001
Interquartile range8.2–18.1%7.8–18%8.6–18.1% 
Histology    
Glioblastoma10677 (97.2%)97%97%0.4
Giant cell glioblastoma81 (0.7%)0.7%0.7% 
Gliosarcoma229 (2.1%)2%2% 
Site    
Frontal2521 (23%)22%25%0.4
Temporal2538 (23%)24%21% 
Parietal1887 (17%)17%17% 
Occipital474 (4%)4%4% 
Other3567 (32%)32%32% 
Tumor size (n = 6349, [58%])    
Median45 mm44 mm45 mm0.02
Interquartile range30–55 mm30–54 mm33–60 mm 
Type of surgery    
Biopsy2510 (23%)21%24%0.001
Resection8071 (73%)75%71% 
Resection contraindicated71 (0.6%)0.6%0.6% 
Unknown406 (4%)4%3% 
Radiation therapy    
External beam radiation therapy7844 (71.4%)75%65%< 0.001
External beam radiation therapy with implants or isotopes200 (1.8%)2%2% 
Implants or isotopes alone32 (0.3%)0.3%0.3% 
Radiation therapy, NOS51 (0.5%)0.4%0.5% 
None2336 (21.3%)19%26% 
Refused203 (1.8%)2%3% 
Unknown321 (2.9%)3%4% 

Tumor Size at the Time of Diagnosis

Tumor size at the time of diagnosis was recorded for 6349 patients (58%); of these, the marital status was known for 6240 patients. Tumor size was not recorded in 43% of unmarried patients and in 42% of married patients. The median tumor size was 45 mm and the interquartile range was 30–55 mm. Unmarried patients were found to have larger tumors than married patients (49% were larger than the median size [45 mm], vs. 45% of married patients; P = 0.02, Wilcoxon rank sum test,). Figure 1 shows the cumulative distributions of tumor size for married and unmarried patients. The group was divided into quartiles by tumor size: ≤ 30 mm, 31–45 mm, 46–55 mm, and ≥ 56 mm. The percentage of married patients in each tumor size quartile (smallest to largest) was 26%, 29%, 23%, and 22%, respectively; for unmarried patients, the corresponding figures were 24%, 27%, 22%, and 26%, respectively. A multivariate proportionate odds ordinal regression model was used to relate clinical variables to tumor size quartile while adjusting for potentially confounding variables. The model included age, gender, race, poverty level quartile, tumor site and histology, year of diagnosis, SEER registry, and marital status. In this model, unmarried status was found to be a significant predictor of a larger tumor size at the time of diagnosis (odds ratio [OR] of 1.16; 95% confidence interval [95% CI], 1.05–1.29 [P = 0.004]). For subclasses of the unmarried group, the adjusted ORs from the multivariate model ranged from 1.08 (divorced) to 1.33 (single). The effect of marital status on tumor size was found to be larger in males than females, but an interaction term between marital status and gender was not found to be statistically significant (P = 0.11).

Figure 1.

Tumor size at the time of diagnosis in relation to marital status (cumulative size distribution plot). For any given tumor size (x-axis), fewer unmarried patients had tumors that were that size or smaller compared with married patients (P = 0.02, Wilcoxon rank sum test,).

Type of Surgical Procedure

For these analyses, we omitted 406 patients (3.7%) for whom the type of surgery was not contained in the database and 71 additional patients (0.6%) in whom surgical resection was contraindicated due to other conditions. Of the remaining 10,510 patients, 76.8% underwent surgical resection and 23.2% underwent biopsy.

Unmarried patients were less likely than married patients to undergo tumor resection rather than a biopsy; 75% of unmarried patients and 78% of married patients underwent surgical resection (P < 0.001, chi-square test). Figure 2 shows that unmarried patients were less likely than married patients to undergo surgical resection across a wide range of ages and tumor sizes. In a multivariate logistic regression model adjusted for age, gender, race, poverty level quartile, tumor site and histology, year of diagnosis, and SEER registry (n = 10,293), unmarried patients were found to be less likely to undergo resection (OR of 0.88; 95% CI, 0.79–0.98 [P = 0.02]). Adjusted ORs for subclasses of the unmarried group ranged from 0.83 (divorced) to 0.91 (widowed). The effect of marital status on type of surgery was found to be larger in males than females, but an interaction term between marital status and gender was not found to be statistically significant (P = 0.16). In a subset model also adjusted for tumor size (n = 6033), the OR for marital status in relation to type of surgery was nearly unchanged, but was no longer statistically significant (OR of 0.91; 95% CI, 0.78–1.06 [P = 0.2]).

Figure 2.

Plots using local-likelihood logistic regression to show type of surgical procedure (surgical resection or biopsy) in relation to marital status over a range of ages and tumor sizes. (A) Possibility of surgical resection in relation to age (n = 10,364). (B) Possibility of resection in relation to tumor size (n = 6079). Unmarried patients were less likely to undergo surgical resection for all but the oldest patients and those with the largest tumors.

For 46 patients in this group (0.4%), the reason surgical resection was not performed was coded as “refused” (0.6% of unmarried patients and 0.4% of married patients). In a logistic regression model adjusted for the factors listed earlier, refusal of surgical resection was found to be significantly associated with unmarried status (OR of 2.1; 95% CI, 1.1–4.0 [P = 0.02]).

Use of Postoperative Radiation Therapy

For these analyses, we omitted 404 patients (3.7%) for whom it was unknown whether radiation therapy was administered during initial treatment, who received nonstandard treatment (isotopes or implants without external beam radiation therapy), or for whom the type of radiation therapy administered was unknown.

Approximately 76% of patients received radiation therapy (70% of unmarried patients and 79% of married patients; P < 0.001, chi-square test). Figure 3 shows that unmarried patients were less likely than married patients to undergo radiation therapy across a wide range of ages and tumor sizes. In a multivariate logistic regression model adjusted for age, gender, race, poverty level quartile, tumor site and histology, year of diagnosis, SEER registry, and type of surgery (biopsy, surgical resection, or unknown) (n = 10,356), unmarried patients were found to be less likely to undergo radiation therapy (OR of 0.69; 95% CI, 0.62–0.77 [P < 0.001]). Adjusted ORs for subclasses of the unmarried group ranged from 0.50 (separated) to 0.81 (divorced). The effect of marital status on the receipt of radiation therapy was found to be slightly larger in males than females, but an interaction term between marital status and gender was not statistically significant (P = 0.4). In a subset model also adjusted for tumor size (n = 6018), the OR for unmarried status in relation to receipt of radiation therapy was nearly unchanged (OR of 0.69; 95% CI, 0.60–0.80 [P < 0.001]).

Figure 3.

Plots using local-likelihood logistic regression to show the likelihood of receiving radiation therapy during initial treatment in relation to marital status over a range of ages and tumor sizes. (A) Possibility of receiving radiation therapy in relation to age (n = 10,356). (B) Possibility of receiving radiation therapy in relation to tumor size (n = 6018). Unmarried patients were more likely to undergo radiation therapy for a wide range of ages and at all tumor sizes.

For 199 patients in this group (1.8%), the reason radiation therapy was not performed was coded as “refused” (1.6% of married patients and 2.8% of unmarried patients). In a logistic regression model adjusted for the factors listed earlier, refusal of radiation therapy was found to be significantly associated with unmarried status (OR of 2.0; 95% CI, 1.5–2.7 [P < 0.001]). The adjusted ORs for subclasses of the unmarried group ranged from 1.8 (divorced) to 5.5 (separated). Older age significantly predicted refusal of radiation therapy in this model (OR of 1.3 per decade; 95% CI, 1.2–1.5 [P < 0.001]); gender, race, and poverty level quartile were not found to be significant predictors of refusal.

Patient Survival

The median survival was 7 months for married patients and 6 months for unmarried patients (P < 0.001, log-rank test) (Fig. 4). At 6 months, the actuarial survival was 55% for married patients and 47% for unmarried patients; at 1 year, the corresponding actuarial survival was 28% and 23%, respectively. To study the possible correlation between marital status and survival in a cohort of patients treated in a relatively uniform and aggressive manner, we used a subgroup of patients who underwent both surgical resection and radiation therapy. Of the 6095 patients treated with surgical resection and radiation therapy, 5397 (89%) are known to have died. The median survival for both the unmarried and married patients in this group was 10 months; the 1-year survival was 35% for unmarried patients and 38% for unmarried patients (P = 0.4, univariate analysis). In a multivariate Cox proportional hazards regression model adjusted for age, gender, race, poverty level quartile, tumor site and histology, year of diagnosis, and SEER registry, unmarried patients were found to have a higher risk of mortality (OR of 1.10; 95% CI, 1.03–1.17 [P = 0.003]). The adjusted hazard ratios for subclasses of the unmarried group ranged from 1.05 (separated) to 1.17 (widowed). The effect of marital status on survival was found to be slightly larger in females than in males, but an interaction term between marital status and gender was not found to be statistically significant (P = 0.3). In a subset model also adjusted for tumor size (n = 3847), the hazard ratio for unmarried status was nearly unchanged (OR of 1.15; 95% CI, 1.06–1.25 [P < 0.001]).

Figure 4.

Kaplan–Meier plot showing survival after diagnosis in relation to marital status. Unmarried patients were found to have a higher risk of mortality (P < 0.001).

Histologic Diagnosis Not Made during Life

In addition to the main cohort of patients with histologically confirmed glioblastoma diagnosed during life reported earlier, 1020 additional adults in the SEER database had diagnoses of supratentorial glioblastoma made by clinical or radiographic criteria only or were reported only on the basis of autopsy or death certificate diagnoses between 1988–2001. These patients were older than the main cohort (median age of 77 years), had larger tumors at the time of presentation (median, 5.0 cm), were unlikely to receive radiation therapy (only 22% received therapy), and had a short survival (median survival of 2 months); 52% were married. All had glioblastoma histology (i.e., International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM] code 9440). In a multivariate logistic regression model adjusted for age, gender, race, poverty level quartile, SEER registry, and tumor site, unmarried status was found to be a significant predictor of a lack of histologic diagnosis being made during life (OR of 1.45; 95% CI, 1.23–1.69 [P < 0.001]). Adjusted ORs for subclasses of the unmarried group ranged from 1.3 (divorced) to 2.5 (separated). The effect of unmarried status was found to be larger in males than females, but an interaction term between marital status and gender was not statistically significant (P = 0.06).

DISCUSSION

We studied a population-based cohort of adult patients with supratentorial GM to determine the influence of marital status on diagnosis, treatment, and outcome. We found that, compared with married patients, unmarried patients presented with larger tumors, were less likely to receive histologic confirmation of their diagnosis during life, were less likely to receive aggressive treatment (including surgical resection and radiation therapy), and had shorter survival, even when similarly treated. There was no obvious difference in the effects of the unmarried state between the various subclasses of the group: single (never married), divorced, separated, or widowed. For many of the analyses, there was an interaction between gender and marital status, resulting in a stronger effect of marital status in males than in females, although none of the interactions reached statistical significance.

In Western societies, the married state has long been known to be associated with a longer lifespan and lower mortality from cancer and other common diseases.24, 25 This could be due to the selection of healthier individuals to enter into and remain in the married state, to a favorable influence of the married state in preventing disease and surviving its effects, or both.26, 27 In 1987, Goodwin et al. reported that married patients with cancer had longer survival than unmarried patients.6 They identified three ways in which the married state was associated with a favorable cancer prognosis: presentation at an earlier (and hence more treatable) stage of disease; choice or receipt of more aggressive therapy; and longer survival, even after adjustment for disease stage and treatment. Some subsequent studies have validated these findings in other cancer populations,7–13 but others have found no correlation between marital status and stage of disease at the time of presentation, treatment, or survival.28–30

To our knowledge, there currently is no accepted means of screening for glioblastoma and no means of identifying persons at increased risk for its occurrence. These factors might be expected to limit the influence of social factors on survival in glioblastoma. Prior studies of the possible relation between marital status, treatment, and survival in brain tumors have been limited. Cassileth et al. studied a small number of adult patients with glioma within a larger group of patients with advanced cancer and found no influence of marital status on prognosis in this group.28 Sehlen et al. reported that marital status was a marker of favorable prognosis in a small group of patients who had either primary or metastatic brain tumors.31

In contrast, we found a significant association between the married state and a relatively favorable course in patients with glioblastoma. Because glioblastoma rarely metastasizes or spreads outside the brain, we used larger tumor size as a measure of more advanced disease at the time of presentation. Larger tumor size at presentation has been associated with a poorer survival in patients with malignant glioma, which could reflect staging bias or the unfavorable natural history of larger tumors.32, 33 Although earlier presentation in certain other cancer types might be attributable to more regular medical care and screening in married patients,34, 35 our finding more likely may be the result of earlier diagnosis through the detection of subtle neurologic deficits by family members of married patients, or through earlier referral for imaging studies. Earlier diagnosis has been postulated to explain an apparent increased incidence of glioma and meningioma in married patients,36 as well as an observed longer survival in married patients with amyotrophic lateral sclerosis, a degenerative neurologic disease that lacks effective treatment.37 Because the difference in tumor size we found was relatively minor, and because the correlation between tumor size and clinical course is a weak one, the clinical significance of the association between tumor size and marital status most likely is small.

We also studied a group of patients who did not receive histologic confirmation of a clinical or radiographic diagnosis of GM during life. These patients were older and presented with larger tumors; the majority received no recorded treatment. Paszat et al. found that 12.7% of patients with GM recorded in Ontario between 1982–1994 lacked histologic confirmation; older age predicted membership in this group.38 In the current study, unmarried status was found to be strongly associated with membership in this group, which had very short survival. For other cancers, SEER data have been shown to be insensitive to the performance of a biopsy, as for other outpatient procedures.39 Brain tumor biopsies are typically performed only in the hospital, making it less likely that the SEER data significantly underreport such procedures, although we know of no published validation of SEER data for patients with brain tumors. It is likely that these patients' tumors were not biopsied because the perceived benefits did not outweigh the risk of the procedure.

Among patients with histologically confirmed diagnoses, both surgical resection and radiation therapy were more common for married patients, and unmarried patients were more likely to refuse both treatments. Older age has previously been shown to be associated with lower rates of use for both radiation therapy and surgical resection for the treatment of GM.38 We found lower rates of both therapies in unmarried patients across a wide range of patient ages and tumor sizes. The difference in resection rates between married and unmarried patients was approximately 1 additional resection per 25 married patients, and for radiation therapy the difference was approximately 1 additional treated person per 10 married patients (Table 1). The difference in resection rates was modest and of unclear clinical significance. The difference in rates associated with marital status was larger for radiation therapy, which typically requires patients to attend treatment sessions each weekday for 6 weeks. Several groups have suggested that abbreviated or hypofractionated treatment courses may be more appropriate for elderly patients with glioblastoma.40, 41 Such abbreviated treatment courses also might be appropriately offered to selected unmarried patients. Alternatively, increased attention to social support systems may be necessary to assist some unmarried patients with transportation and other needs during treatment.

Given the limitations of our data source, we cannot speculate whether lower rates of surgical resection and radiation therapy in unmarried patients represent differences in therapy offered by physicians or an active choice on the patients' part to refuse. Such refusal on the part of the patient might well represent a preference to avoid the perceived discomfort or risk of more aggressive treatment. Although to our knowledge no prior studies to date have demonstrated any correlation between marital status and depression, quality of life, or functional status in glioma patients,42, 43 a more detailed assessment of the correlations between these factors and marital status would be of interest.

Finally, we found that married patients had longer survival, even when compared with unmarried patients who were similarly treated. As in other studies with similar findings,6, 7, 13, 44 this survival difference remains unexplained. SEER data capture treatment information for 4 months after diagnosis, by which time many patients with glioblastoma would not have had adjuvant chemotherapy initiated. The increased use of chemotherapy3 or of surgical resection at the time of tumor recurrence45 among married patients could contribute to longer survival. These therapies are reported to have only modest efficacy, with hazard ratios for untreated patients roughly comparable to that we observed for the unmarried state; a substantial disproportion in their use would be necessary to explain our findings. The proportion of glioblastoma patients who receive adjuvant chemotherapy is unknown, but only approximately 15% of glioblastoma patients undergo second resections.45 Other demographic factors associated with the married state, such as higher socioeconomic status and education, also could contribute to the survival difference we observed. Information regarding these factors is not contained in SEER data.

The data source we used, the SEER Public Use Files, has certain limitations for a study such as ours. We could find no study that specifically validated SEER marital status data, and misclassification of this variable could have reduced the sensitivity of the current study. Marital status itself in this context most likely represents a proxy measure of the strength of the social support system. Although there was no difference in our findings with respect to the widowed, single, divorced, and separated states, it is possible that a better measure of social network strength might produce stronger findings by reducing misclassification. Performance status at diagnosis, a powerful prognostic factor in patients with malignant glioma, also is not contained in SEER data. A central review of pathology was not possible in the current study, although prior analyses have shown good agreement between community and central review neuropathologists for glioblastoma.46–48 Validation of SEER data compared with Medicare files generally has shown excellent agreement with regard to rates of surgery and radiation therapy use.39, 49, 50 For certain subpopulations within our data set that would be expected to receive the most aggressive therapy (such as young, married patients with tumors coded as lobar in location and limited in extent), we have found rates of surgical resection and radiation therapy receipt of 90% or higher (data not shown), suggesting that the underreporting of these therapies is minimal in at least some of the patient subsets we studied. Although the underreporting of surgical or radiation treatment might be correlated with patient age, we found that differences associated with marital status were consistently present across a wide range of patient ages (Figs. 2 and 3). Patients who receive surgical resection or radiation treatment outside the SEER geographic area in which they reside might have treatment information underreported in SEER files, but marital status has not been shown to predict travel to distant referral centers.51 Patients who do travel long distances for treatment at referral centers typically are young,51 which would bias the study in the opposite direction from our findings of higher rates of both surgical resection and radiation therapy in younger patients (Figs. 2 and 3).

Clinical trials addressing GM treatment are of great importance because this disease causes significant morbidity and mortality in a relatively young patient population. Patients who enter clinical trials for other types of cancer are typically highly educated, of middle-to-upper socioeconomic class, male, and married.52, 53 The opinion of “an important person (e.g., spouse)” is often cited as influencing the patient's decision to participate in a clinical trial.54, 55 Although efforts have been made to decrease barriers to enrollment in such trials for racial and ethnic minority populations, those of lower socioeconomic position, and the elderly,56–58 disparities based on marital status have to date attracted less attention.59, 60 Unmarried patients with GM receive even standard therapies such as surgical resection and radiation therapy less frequently than married patients, and we suggest that they might also be less likely to enter clinical trials. This question might be clarified by the examination of existing trial databases. Research to identify and, if present, to remedy such disparities will be important both to enhance overall trial participation and to ensure the generalizability of the results of completed trials to unmarried patients, who represent approximately one-third of those diagnosed with glioblastoma in the U.S.

Acknowledgements

The authors thank Sharon Reynolds, Department of Neurological Surgery, University of California San Francisco, for editorial support.

Ancillary