Volume 104, Issue 12 p. 2798-2806
Original Article
Free Access

A population-based description of glioblastoma multiforme in Los Angeles County, 1974–1999

Indro Chakrabarti M.D., M.P.H.

Corresponding Author

Indro Chakrabarti M.D., M.P.H.

Department of Neurosurgery, University of Southern California Keck School of Medicine, Los Angeles, California

Fax: (323) 226-7833

Department of Neurosurgery, University of Southern California Keck School of Medicine, 1200 North State Street, Suite 5046, Los Angeles, CA 90032===Search for more papers by this author
Myles Cockburn Ph.D.

Myles Cockburn Ph.D.

Department of Preventive Medicine, University of Southern California Keck School of Medicine, Los Angeles, California

Search for more papers by this author
Wendy Cozen D.O., M.P.H.

Wendy Cozen D.O., M.P.H.

Department of Preventive Medicine, University of Southern California Keck School of Medicine, Los Angeles, California

Department of Pathology, University of Southern California Keck School of Medicine, Los Angeles, California

Search for more papers by this author
Ya-Ping Wang M.S., M.D.

Ya-Ping Wang M.S., M.D.

Department of Preventive Medicine, University of Southern California Keck School of Medicine, Los Angeles, California

Search for more papers by this author
Susan Preston-Martin Ph.D.

Susan Preston-Martin Ph.D.

Department of Preventive Medicine, University of Southern California Keck School of Medicine, Los Angeles, California

Search for more papers by this author
First published: 15 November 2005
Citations: 116



There have been reports that the incidence rates of brain tumors have increased over the past few decades, but most have considered all brain tumors together. The authors analyzed the pattern of glioblastoma multiforme (GBM) occurrence in Los Angeles County, California to shed light on the incidence and descriptive epidemiology of this type of brain tumor.


Data were obtained from the Los Angeles County Cancer Surveillance Program. Incidence rates were analyzed by gender, race, age at diagnosis, period of diagnosis (1974–1981, 1982–1988, or 1989–1999), and socioeconomic status (SES). In addition, data were stratified according to anatomic subsite. A multivariate model describing changes in rates by each of these variables was constructed.


Age-specific incidence rates (ASIR) rose sharply after age 30 years. The peak ASIR was at age 70–74 years in males and at age 75–79 years in females. The age-adjusted incidence rate (AAIR) of GBM increased from 1974 to 1999 by an estimated 2.4% per year among males and 2.8% per year among females. Overall, males had a 60% increased risk of brain tumors compared with females. Males had a higher incidence of GBM compared with females at each anatomic subsite except the posterior fossa. The largest male:female ratio occurred in the occipital lobes. Non-Latino whites had the highest incidence rates (2.5 per 100,000) followed by Latino whites (1.8 per 100,000), and blacks (1.5 per 100,000). After 1989, compared with the period before magnetic resonance imaging (MRI) was available, there was an increase in GBM incidence rates among those with of higher SES that was most pronounced in females. The incidence of GBM was highest for frontal lobe tumors and for tumors that involved two or more lobes (overlapping tumors), followed by tumors in the temporal and parietal lobes. In the multivariate analysis, year of diagnosis, SES, gender, race (Latino but not black), site, and age at diagnosis all were important predictors of incidence rate.


GBM incidence increased in Los Angeles County over the last 30 years and especially after 1989, suggesting that the introduction of MRI may have contributed to the increase. Individuals older than age 65 years experienced the greatest increase in incidence over time. Older age, male gender, higher SES, and non-Latino white race increased the risk of GBM. Previously unreported incidence rates for GBM among Latino whites were significantly lower than among non-Latino whites but were intermediate between non-Latino whites and blacks. Cancer 2005. © 2005 American Cancer Society.

Glioblastoma multiforme (GBM) is the most common primary brain tumor.1 Optimal management of these tumors requires accurate documentation of their natural history. Recent population-based analysis is lacking. Some cooperative multiinstitutional studies have been performed, but these likely contain referral and selection biases associated with populations entered in clinical trials.2, 3 The majority of clinical knowledge comes from single-institution series with relatively small numbers.4-9 Population-based data bases allow for assessment of the overall impact of the tumor in a community and also can confirm facts or open new questions to be answered. Unchanged low survival rates and reports of apparent increases in GBM over time demand continued epidemiologic study of this disease.10

Epidemiologic reports to date most often have identified malignant gliomas together without adequately defining the anatomic and histologic subsets. The inconsistent application of the label “glioma” may account for some conflicting results. For example, in some studies, glioma may include GBM, which is a highly aggressive tumor, and anaplastic astrocytoma, which is of lower grade.11 In other studies, GBM is combined with several other glioma subtypes.10, 12 Focused analysis of GBM alone is lacking.

Some clinical studies note a more favorable prognosis for frontal tumors or for tumors that come to the surface.3, 13 Other studies note gender differences in survival in anaplastic astrocytomas (World Health Organization [WHO] Grade 3) but not necessarily GBM (WHO Grade 4).14 To date, there has been no report on GBM in the Latino population. Finally, the advent of magnetic resonance imaging (MRI) and its role in the diagnosis of GBM has not been described. We examined the descriptive epidemiology of GBM and overall trends by anatomic subsite. To perform this study, we used incidence data from the population-based Cancer Surveillance Program (CSP) of Los Angeles County.


Source of Patients

GBM of the brain (International Classification of Diseases for Oncology [ICD-O] codes C71.0–C71.9) that were diagnosed between 1974 and 1999 were identified using the Los Angeles County CSP, which has been the population-based cancer registry for Los Angeles County since 1972 and has participated in the National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER) Registry Program since 1992.15 The CSP captures 99% of cancer diagnoses in Los Angeles County through mandated reporting and additional screening of diagnostic facilities.

We classified GBMs further according to subsite location within the brain. The subsite distribution that was used for this study is outlined in Tables 1 and 2. However, some subsite terms require clarification. The morphology code 9440/3 corresponds to GBM. The topography site C71.0 corresponds to the cerebrum and includes all deep structures of the supratentorial brain, such as the basal ganglia, central white matter, insula, and thalamus. The ventricle (C71.5) indicates tumor of the ventricular system, including the lateral, third, and fourth ventricles. The cerebellum (C71.6) includes the cerebellar hemispheres, vermis, and cerebellopontine angle. The brainstem (C71.7) includes any infratentorial brain excluding the cerebellum. For this study, the absolute numbers of patients with GBM in the ventricular system, cerebellum, and brainstem were small. Therefore, these three categories were combined and were named “posterior fossa” GBM. Overlapping GBM (C71.8) indicates tumors that cross into ≥ 2 lobes or the so-called “butterfly glioma,” which grow through the corpus callossum. Finally, general locations, such as the anterior or middle fossa, are designated as brain not otherwise specified (NOS) (C71.9). In the current study, we assessed microscopically confirmed GBM by period and anatomic subsite to determine the impact of changing methods of diagnosis, introduced by MRI, on time trends.

Table 1. Age-Adjusted Incidence Rates of Glioblastoma Multiforme by Age and Period of Diagnosis: Males and Femalesa
Period Ages 20–49 yrs Ages 50–64 yrs Age > 65 yrs All ages ≥ 20 yrs
 AAIR 0.80 4.48 5.07 1.66
 95% CI 0.74–0.85 4.41–4.54 5.00–5.15 1.55–1.77
 Absolute no. 171 393 311 875
 AAIR 0.81 4.00 7.22 1.87
 95% CI 0.76–0.87 3.93–4.07 7.13–7.30 1.74–1.99
 Absolute no. 185 308 416 909
 AAIR 0.83 5.29 10.62 2.50
 95% CI 0.79–0.87 5.23–5.35 10.50–10.70 2.39–2.61
 Absolute no. 365 639 1044 2048
All yrs
 AAIR 0.82 4.70 8.25 2.11
 95% CI 0.80–0.85 4.66–4.74 8.20–8.30 2.0–2.17
 Absolute no. 721 1340 1771 3832
  • AAIR: age-adjusted incidence rate per 100,000; 95% CI: 95% confidence interval.
  • a Source data: Los Angeles County, 1974–1999.
Table 2. Age-Adjusted Incidence Rates of Glioblastoma Multiforme by Brain Subsite, Gender, and Agea
Brain subsiteb Ages 20–49 yrs Ages 50–64 yrs Age > 65 yrs All ages ≥ 20 yrs Gender ratio
Males Females Males Females Males Females Males Females
Cerebrum (C71.0) 1.63c
 AAIR 0.06 0.04 0.35 0.21 0.45 0.24 0.13 0.08
 95% CI 0.05–0.06 0.03–0.05 0.33–0.36 0.20–0.22 0.44–0.47 0.23–0.25 0.11–0.16 0.06–0.09
 Absolute no. 26 17 47 31 40 31 113 79
Frontal (C71.1) 1.38c
 AAIR 0.27 0.15 1.16 0.93 2.03 1.56 0.55 0.4
 95% CI 0.25–0.29 0.13–0.17 1.13–1.19 0.90–0.95 1.99–2.07 1.53–1.58 0.50–0.60 0.36–0.44
 Absolute no. 118 65 158 138 176 200 452 403
Temporal (C71.2) 1.75c
 AAIR 0.13 0.11 1.06 0.63 2.19 1.07 0.49 0.28
 95% CI 0.12–0.15 0.09–0.12 1.03–1.08 0.61–0.65 2.15–2.23 1.05–1.09 0.44–0.54 0.24–0.31
 Absolute no. 55 47 143 93 189 136 387 276
Parietal (C71.3) 1.55c
 AAIR 0.14 0.1 1.00 0.71 2.14 1.29 0.48 0.31
 95% CI 0.13–0.16 0.09–0.11 0.98–1.03 0.69–0.73 2.10–2.18 1.26–1.31 0.43–0.53 0.28–0.35
 Absolute no. 62 43 136 106 185 163 383 312
Occipital (C71.4) 2.6c
 AAIR 0.04 0.02 0.27 0.09 0.55 0.25 0.13 0.05
 95% CI 0.03–0.05 0.01–0.02 0.26–0.28 0.09–0.10 0.53–0.57 0.24–0.26 0.10–0.15 0.04–0.07
 Absolute no. 19 7 37 14 48 32 104 53
Posterior fossa (C71.5–C71.7) 1.17
 AAIR 0.05 0.06 0.12 0.11 0.20 0.11 0.07 0.06
 95% CI 0.04–0.06 0.05–0.07 0.11–0.13 0.10–0.12 0.19–0.21 0.10–0.11 0.05–0.08 0.04–0.07
 Absolute no. 28 31 16 16 17 14 61 61
Overlapping (C71.8) 1.73c
 AAIR 0.23 0.13 1.41 0.77 2.61 1.56 0.64 0.37
 95% CI 0.21–0.25 0.12–0.15 1.38–1.44 0.75–0.79 2.57–2.66 1.53–1.58 0.58–0.7 0.33–0.41
 Absolute no. 96 56 194 115 221 199 511 370
NOS (C71.9)
 AAIR 0.07 0.04 0.41 0.27 0.74 0.45 0.19 0.12 1.58c
 95% CI 0.06–0.08 0.04–0.05 0.39–0.42 0.26–0.29 0.72–0.77 0.43–0.46 0.15–0.22 0.10–0.14
 Absolute no. 31 20 55 41 63 57 149 118
All sites
 AAIR 0.99 0.65 5.77 3.72 10.92 6.51 2.68 1.67
 95% CI 0.95–1.03 0.62–0.69 5.71–5.83 3.67–3.77 10.83–11.01 6.46–6.57 2.56–2.80 1.59–1.75
 Absolute no. 435 286 786 554 939 832 2160 1672
 Gender ratio 1.52c 1.55c 1.68c 1.6c
  • AAIR: age-adjusted incidence rate per 100,000; 95% CI: 95% confidence interval; NOS: not otherwise specified.
  • a Source data: Los Angeles County, 1974–1999.
  • b International Classification of Diseases for Oncology (ICD-O) codes are in parentheses.
  • c AAIR was significantly greater in males (P < 0.05).

Race/Ethnicity and Socioeconomic Status Measures

We analyzed data for all racial/ethnic groups, but we analyzed current data only for non-Latino whites, Latino whites, and blacks, because the sample size for Asians and others was too small to provide meaningful information. Racial/ethnic groups were defined on the basis of the CSP report of race (obtained from the hospital or clinic record), and we used the United States Bureau of the Census 1980 Spanish surname list to distinguish Latino whites from non-Latino whites. Only those who had surnames that appeared on the list of Spanish surnames were classified as Latino.16 For Los Angeles County incidence data, we assigned socioeconomic status (SES) classification to each individual in the CSP based on their census tract of residence at the time of diagnosis, as described elsewhere.17 Briefly, to create SES assignments for each census tract, we calculated quintiles of the combined percentage distributions of educational attainment and the median household income reported in the 1970, 1980, 1990 Censuses, and we interpolated linearly through these values for all other years. SES trends were evaluated in tertiles, collapsing the top two and bottom two categories to maintain sufficient numbers of patients in each cell for analysis. SES trends were analyzed only through 1999, because the required Census 2000 data were not available to generate SES measures at the time of the analysis. Because of sparse data, we analyzed and compared the incidence rates by SES for two periods: before and after the introduction of MRI.

Population Estimates

We obtained annual estimates of the Los Angeles County population classified by gender, age, race/ethnicity, and SES for the intercensal years between 1973 and 1989 by linear interpolation using the 1970, 1980, and 1990 Census data. We based our population estimates for 1991–1999 on the 1990 Census results and the age-specific, gender-specific, and race-specific annual population figures estimated by the Department of Finance of the State of California.18

Statistical Methods for Incidence Analysis

Using the year 2000 United States Census population as the standard for direct age adjustment, we calculated gender-specific and race-specific, age-adjusted incidence rates for 5-year periods between 1974 and 1999. We separately assessed incidence rates by anatomic subsite.

To determine whether any observed increase in incidence was related to the introduction of MRI, we divided the study period into 3 periods: 1974–1981, 1982–1988, and 1989–1999. Dates prior to 1981 represent the period prior to MRI. During 1982–1988, MRI was becoming widespread, and dates after 1989 represent the MRI era.

Age was considered as a continuous variable for some analyses; for other analyses, age was categorized into 3 groups: ages 20–49 years, ages 50–64 years, and older than age 65 years. Pvalues for trends in incidence rates were assessed using the Mantel–Haenszel test.19 Slopes for trend lines are presented as estimated annual percent change (EAPC) with the appropriate confidence intervals (95% CI).

We conducted a multivariate analysis of incidence rates using Poisson regression with year of diagnosis (1974–1988 vs. 1989–1999, to allow for potential changes in incidence due to the introduction of MRI), SES (high, middle, or low), gender, race (Spanish surnamed whites or Latino whites, blacks, non-Spanish surnamed whites or non-Latino whites), and tumor site (frontal lobe, overlapping, other excluding NOS, collapsed to ensure adequate sample size in each category), and age (ages 20–64 yrs vs. age 65 yrs or older, the age at which the most striking increases in incidence occurred in univariate analyses) as predictors. These categories were chosen after considering the results of the univariate analyses and to maintain meaningful sample sizes in each of the combined categories.


The rate of microscopic confirmation of GBM for all subsites taken together was 82.0%. However, as expected, the rate was higher in more surgically accessible areas of the brain. For example, in lobar tumors, the rate of microscopic confirmation was 87.8% for frontal lobe tumors, 87.7% for temporal lobe tumors, 86.9% for parietal lobe tumors, 89.6% for occipital lobe tumors, and 83.8% for overlapping lobar tumors. However, with tumors located deeper in the cerebrum, the microscopic confirmation rate dropped to 62.4%; and, in the brainstem, the rate was 16.2%. The rate of microscopic confirmation increased over time: During 1973–1985, the rate was 72.3%, and it climbed to 85.4% after 1985.

Trends in GBM Incidence

GBM incidence peaked at ages 70–74 years among males and at ages 75–79 years among females (Fig. 1). Prior to 1985, the peak age for both males and females was 5–10 years younger (data not shown). In addition, the increase in incidence noted with increasing age was maintained when nonwhite and lower socioeconomic classes were studied (data not shown).

Details are in the caption following the image

Age-specific incidence rates of glioblastoma multiforme (source data: Los Angeles County, 1974–1999).

There was a clear increase in incidence rates for both males and females over time (Fig. 2). The EAPC was 2.4 (95% CI, 1.6–3.2) for males and 2.8 (95% CI, 1.6–3.2) for females. The increase in incidence was most notable in individuals older age 65 years, in whom the incidence has more than doubled (Table 1, Fig. 1). In individuals younger than age 65 years, the incidence has been roughly stable.

Details are in the caption following the image

Secular trends for glioblastoma multiforme (source data: Los Angeles County, 1974–1999).

Trends in Incidence by Anatomic Subsite

The male:female ratio for GBM in all sites combined was 1.6 (P < 0.05) (Table 2) This relation was true for all ages, but was greatest in patients older than age 65 years (1.68). For each brain subsite except the posterior fossa, the gender ratio similarly was elevated. The ratio was greatest with GBM located in the occipital lobe.

Incidence rates were highest in non-Latino whites and were significantly lower among Latino whites and blacks (Table 3). The gender ratio was similar and was statistically significant in non-Latino whites and blacks, but it was slightly lower among Latino whites.

Table 3. Age-Adjusted Incidence Rates of Glioblastoma Multiforme by Brain Subsite and Race, Ages 20 Years and Older: Males and Femalesa
Location Ages ≥ 20 yrs
Non-Latino whites Latino whites Blacks
 AAIR 0.13 0.06 0.09
 95% CI 0.11–0.15 0.03–0.09 0.04–0.13
 Absolute no. 148 20 17
Frontal lobe
 AAIR 0.56 0.41 0.33
 95% CI 0.52–0.60 0.33–0.50 0.25–0.41
 Absolute no. 644 112 63
Temporal lobe
 AAIR 0.45 0.29 0.24
 95% CI 0.41–0.49 0.22–0.36 0.17–0.32
 Absolute no. 522 77 43
Parietal lobe
 AAIR 0.48 0.28 0.2
 95% CI 0.44–0.52 0.21–0.35 0.14–0.26
 Absolute no. 555 73 38
Occipital lobe
 AAIR 0.1 0.06 0.06
 95% CI 0.09–0.12 0.03–0.09 0.02–0.09
 Absolute no. 123 17 11
Posterior fossa
 AAIR 0.07 0.06 0.08
 95% CI 0.05–0.08 0.03–0.10 0.04–0.12
 Absolute no. 76 23 16
 AAIR 0.58 0.47 0.32
 95% CI 0.54–0.62 0.37–0.56 0.23–0.40
 Absolute no. 669 120 56
 AAIR 0.16 0.19 0.14
 95% CI 0.14–0.18 0.13–0.25 0.08–0.19
 Absolute no. 183 50 25
All sites
 AAIR 2.53 1.83b 1.45b
 95% CI 2.44–2.62 1.65–2.01 1.27–1.62
 Absolute no. 2920 492 269
 Gender ratio 1.66c 1.32 1.6c
  • AAIR: age-adjusted incidence rate per 100,000; 95% CI: 95% confidence interval.
  • a Source data: Los Angeles County, 1974–1999.
  • b P < 0.05 with non-Latino whites as the reference group.
  • c The AAIR was significantly greater in males at the P < 0.05 level.


The highest incidence of GBM occurred in individuals with high SES (Tables 4, 5). The incidence in this group increased more dramatically after 1989 compared with the increase in the middle and low SES groups (Table 5). The higher SES groups were up to 70% more likely to be diagnosed with GBM in the frontal lobes. In for all other tumor sites, the higher SES groups also had higher incidence rates, except for GBM of the posterior fossa, which had a ratio < 1.0.

Table 4. Age-Adjusted Incidence Rates of Glioblastoma Multiforme by Brain Subsite and Socioeconomic Status Among Males and Females, Age 20 Years and Older: Mean Age at Diagnosis by Subsitea
Brain subsite Socioeconomic status High/low ratio Mean age at diagnosis (yrs)
High Middle Low
Cerebrum 1.63 58.4
 AAIR 0.13 0.09 0.08
 95% CI 0.1–0.15 0.06–0.12 0.06–0.1
 Absolute no. 99 36 57
Frontal lobe 1.70 60.6
 AAIR 0.56 0.53 0.33
 95% CI 0.51–0.62 0.46–0.60 0.28,–0.37
 Absolute no. 425 205 225
Temporal lobe 1.64 62.4
 AAIR 0.46 0.36 0.28
 95% CI 0.41–0.51 0.3–0.42 0.24–0.32
 Absolute no. 341 137 185
Parietal lobe 1.86 62.8
 AAIR 0.52 0.33 0.28
 95% CI 0.46–0.57 0.27–0.39 0.24–0.32
 Absolute no. 383 126 186
Occipital lobe 1.57 62.8
 AAIR 0.11 0.09 0.07
 95% CI 0.08–0.13 0.06–0.12 0.05–0.08
 Absolute no. 79 34 44
Posterior fossa 0.85 49.7
 AAIR 0.06 0.05 0.07
 95% CI 0.04–0.08 0.03–0.07 0.05–0.09
 Absolute no. 49 18 55
Overlapping 1.43 62.4
 AAIR 0.57 0.47 0.4
 95% CI 0.52–0.63 0.4–0.54 0.36–0.45
 Absolute no. 430 177 274
NOS 1.30 61.2
 AAIR 0.17 0.14 0.13
 95% CI 0.14–0.2 0.1–0.17 0.1–0.16
 Absolute no. 125 51 91
All sites 1.58
 AAIR 2.57 2.05 1.63
 95% CI 2.45–2.69 1.91–2.2 1.54–1.73
 Absolute no. 1931 784 1117
  • AAIR: age-adjusted incidence rate per 100,000; 95% CI: 95% confidence interval; NOS: not otherwise specified.
  • a Source data: Los Angeles County, 1974–1999.
Table 5. Age-Adjusted Incidence Rates of Glioblastoma Multiforme by Socioeconomic Status and Period of Diagnosis: Males and Femalesa
Period High SES Middle SES Low SES High/low ratio
1974–1981 1.35
 AAIR 1.88 1.74 1.39
 95% CI 1.69–2.07 1.5–1.99 1.22–1.55
 Absolute no. 403 199 273
1982–1988 1.45
 AAIR 2.2 1.83 1.52
 95% CI 1.99–2.41 1.56–2.09 1.33–1.70
 Absolute no. 448 188 273
1989–1999 1.70
 AAIR 3.14 2.37 1.85
2.95–3.33 2.13,–2.60 1.69–2.00
 Absolute no. 1080 397 571
All yrs 1.58
 AAIR 2.57 2.05 1.63
 95% CI 2.45–2.69 1.91–2.2 1.54–1.73
 Absolute no. 1931 784 1117
  • SES: socioeconomic status; AAIR: age-adjusted incidence rate per 100,000; 95% CI: 95% confidence interval.
  • a Source data: Los Angeles County, 1974–1999.

Multivariate Analyses

Each of the variables studied (yr of diagnosis, SES, gender, race, tumor site, and age) in the multivariate analysis was a highly significant predictor of the GBM incidence rate (Table 6). After adjustment for all other variables, there was a 46% increase in incidence of GBM during 1989–1999 compared with 1974–1988, and patients age 65 years and older had an almost a 5-fold higher rate of GBM compared with patients ages 20–64 years. Non-Latino whites had twice the rate of GBM experienced by Latino whites, and the Latino rate was 43% higher than the rate among blacks.

Table 6. Multivariate Analysis of Glioblastoma Multiforme Risk Factors
Risk factor RR 95% CI P value
Yr of diagnosis
 1974–1988 1.00
 1989–1999 1.46 1.36–1.56 < 0.0001
 High 1.00
 Middle 0.83 0.76–0.91 < 0.0001
 Low 0.75 0.69–0.82 < 0.0001
 Male 1.00
 Female 0.64 0.60–0.69 < 0.0001
 Non-Latino white 1.00
 Latino white 0.50 0.45–0.56 < 0.0001
 Black 0.57 0.49–0.65 < 0.0001
Tumor site
 Other except NOS 1.00
 Frontal lobe 0.47 0.43–0.51 < 0.0001
 Overlapping 0.48 0.44–0.52 < 0.0001
 20–64 yrs 1.00
 ≥ 65 yrs 4.70 4.38–5.03 < 0.0001
  • RR: relative risk; 95% CI: 95% confidence interval; SES: socioeconomic status; NOS: not otherwise specified.


A limitation of previous descriptive epidemiologic studies of gliomas has been their lack of specificity regarding the histologic subtypes of GBM investigated. More recent studies have presented separate analyses by tumor grade, although they nonetheless combined various histologic types.10, 12, 20 Large surgical series, including multiinstitutional studies, have reported data that may reflect a referral bias or a practice bias and, thus, may not be representative of all GBMs. In the current study, we overcame those two limitations with a population-based study of GBM. In addition, we believe this is the first report of incidence rates for GBM among Latinos, a group that will become an increasingly important segment of the United States population. It is noteworthy that, in Los Angeles County, 45% of the population self-reported Latino ethnicity in the 2000 Census, representing 12% of the total United States Latino population. The clinical management of any brain tumor is affected by its location. Therefore, we also defined the epidemiology of GBM by anatomic subsite.

The age-specific incidence rates among Los Angeles County residents are similar to those reported by others.1, 11 In one study, it was noted that the slope of increase was similar for males and females with all primary brain tumors, but the increased incidence with age was most marked for those diagnosed with GBM.21 This suggests that GBM may have an etiology distinct from that of other primary gliomas. Our current results certainly reiterate the finding that GBM is largely a disease of older age groups.

A male predominance in Los Angeles County is consistent with the findings of other reports.11 Higher brain tumor rates in males are reported in registries throughout the world.22 In our study, we found that the gender ratio showed significantly higher incidence rates among males in all age groups, periods, anatomic subsites, and races, with two exceptions: Latino whites (a similar pattern, but not statistically significant) and patients with GBM of the posterior fossa.

The increasing secular trends observed in Los Angeles County have been noted in larger population-based studies (including all SEER cancer registries)10 and in other studies.23-25 In our current study, this increase did not occur among patients ages 20–49 years, and it was strongest for patients older than age 65 years (as noted by others24). This increased incidence rate with age also persisted when we considered minorities and individuals with low SES.

For decades, it has been shown that increasing social class has been linked to an increase in the risk of glioma among males.26, 27 We observed a similar association, but we also found that high SES is a risk factor for females. The incidence rates after 1989 show higher incidences in males and females, with the greatest increases seen among those of higher SES.

Some investigators have theorized that the increase in brain tumors is been real and is not merely a result of better detection due to the availability of MRI.23 This argument is plausible, because, in general, if they are left untreated, brain tumors will result in death and neurologic deficit prior to death and, thus, should not escape detection. This argument also is logical, because GBM of the posterior fossa, which includes the brainstem, is associated with the youngest age at diagnosis and is the least likely to be found in males or in individuals with high SES. This is unlike all of the other brain subsites, in which the incidence of GBM was significantly higher among males and among individuals with higher SES. The brainstem is where many vital neurologic pathways converge and will cause symptoms relatively rapidly. Therefore, it seems that detection is not delayed. In the current study, we evaluated this hypothesis by examining incidence rates in 3 periods, prior to the introduction of MRI (1974–1981), during the introduction of MRI (1982–1988), and the after widespread and routine use of MRI for diagnostic confirmation (from 1989 onward). We observed a larger increase in incidence after the period of MRI introduction. This suggests that MRI may have influenced the increase. Our findings of increased incidence and stronger SES gradient after 1988 may imply that MRI is having an impact on diagnosis in the elderly, either by detection of previously unnoticed tumors or by increasing the willingness of physicians to pursue a more aggressive work-up in this age group. It is likely that MRI is providing earlier diagnosis of GBM and is detecting GBM in individuals whose GBM may not be diagnosed otherwise. Our speculation is that access to MRI is detecting GBM at an earlier stage, before neurologic deficit. Prior to the introduction of MRI, individuals who developed a neurologic deficit may not have been diagnosed with GBM, perhaps because the physician may have attributed the deficit to a stroke.

One focus of this report was the description of GBM by anatomic subsite. There have been some reports of better survival rates among patients with GBM of the frontal lobe or cerebellum.3, 13, 20 Variations in survival by anatomic subsite have been observed in patients with other types of brain tumors, such as astrocytoma and anaplastic astrocytoma.20 Treatment decisions, such as surgery with macroscopic total resection versus subtotal resection versus biopsy, also are influenced by tumor location. This report provides a population-based description based on anatomic subsite to aid clinical decisions. In the multivariate analysis, year of diagnosis, SES, gender, race (Latino but not black), tumor site, and age at diagnosis all were important predictors of incidence rate.

In conclusion, males are at a significantly greater risk of developing GBM in each anatomic subsite, age group, race, and socioeconomic class. We also observed increasing GBM incidence rates over the last 30 years, especially in the elderly. All of these effects are independent of one another. The changes in incidence both in peak age and SES after 1989 may be driven in part by MRI technology. The information gained from this study may help clinicians counsel their patients with this disease.