Trends in Brain Cancer Incidence and Survival in the United States: Surveillance, Epidemiology, and End Results Program, 1973 to 2001.

Sundeep Deorah, M.A.; Charles F. Lynch, M.D., Ph.D.; Zita A. Sibenaller, Ph.D.; Timothy C. Ryken, M.D.


Neurosurg Focus. 2006;20(4):E3 

In This Article

Clinical Material and Methods

There are two centralized population-based brain tumor databases in the US: the CBTRUS and the National Cancer Institute's SEER Program.[1] We selected the SEER database for our study because it provides population-based incidence data beginning in 1973 as well as patient survival data. The SEER Program collects data on malignant brain tumors through several population-based cancer registries across the US. Five states (Connecticut, Hawaii, Iowa, New Mexico, and Utah) and four metropolitan areas with diverse population subgroups (Atlanta, Detroit, San Francisco–Oakland, and Seattle–Puget Sound) began participating in the SEER Program as early as 1973. Together they represent 9.5% of the US population.[16] The data collected in the SEER database are known to be of high quality because the program has set a high standard for the collection of cancer data, which are used extensively for research. We requested and received permission to use the SEER data set.

For this study, SEER*Stat software (version 5.3.0) was used to calculate incidence rates, using the US population for the year 2000 as the standard. A total of 38,453 patients diagnosed with malignant brain tumor in the period between 1973 and 2001 were included. For patient survival analyses, only microscopically confirmed and actively followed cases were included. Patients with multiple primary tumors were excluded from these analyses because including them would have led to underestimation of actual survival rates. Relative survival rates were calculated using a table showing the expected rate of tumors for 1970, 1980, and 1990 for Caucasians, African-Americans, and members of other racial groups.

The study population was divided into children (< 20 years old at diagnosis), young/middle-aged adults (20–65 years old), and elderly adults (> 65 years old). Counties were classified into metropolitan and nonmetropolitan areas, based on a report by the US Office of Management and Budget, to act as surrogates for urban and rural counties, respectively. Determined from 1990 decennial census data, these definitions were announced by the Office of Management and Budget in 1996. A total of 32,659 cases occurred in people living in metropolitan counties, and the remaining 5794 cases occurred among those living in non-metropolitan counties. Brain tumors were classified by grouping ICD-O (second edition) four-digit histology codes into broad histology subgroups based on CBTRUS and CNS histology groupings (2002 revision): GBM, ICD-O codes 9440–9442, 9481; oligodendroglioma, ICD-O code 9450; anaplastic astrocytoma, ICD-O codes 9401, 9411; mixed glioma, ICD-O code 9382; medulloblastoma, ICD-O codes 8963, 9363, 9364, 9470–9473, 9501–9503; astrocytoma NOS, ICD-O code 9400; and malignant glioma, ICD-O code 9380.

We used SEER*Stat software to obtain age-standardized incidence rates adjusted to the 2000 US population by age, sex, race, and other user-defined variables. The relative risk, defined as the ratio of incidence rates of brain cancer for one value of a variable divided by another, was calculated for race, age, sex, and place of residence. The approximate Poisson method was used to calculate a 95% CI regarding the relative risk, and any interval not including the value zero was regarded as statistically significant. Age-standardized incidence rates were calculated for all years from 1973 to 2001, and time-trend analysis was performed using the Joinpoint regression analysis software (version 2.7) available through SEER*Stat. Joinpoint is statistical software for the analysis of trends by using joinpoint models, wherein several different lines are connected together at the "joinpoints." The software takes trend data (for example, cancer rates) and fits them to the simplest joinpoint model that the data allow. The user determines the minimum and maximum number of joinpoints. The program starts with the minimum number of joinpoints (for example, zero joinpoints, which is a straight line) and tests whether other joinpoints are statistically significant and must be added to the model (up to the maximum number). This enables the user to test whether an apparent change in trend is statistically significant. The software's tests of significance use a Monte Carlo permutation method.

Because GBM is the most frequently occurring brain cancer and is associated with a poor prognosis, we investigated its trend in greater detail, using the Joinpoint regression program. We also evaluated the survival rate of patients with GBM in three different decades to detect any significant changes in the survival rates over time. All comparisons were made using two-tailed tests, and results with a probability value less than 0.05 were reported as statistically significant.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.
Post as: