Trends in Incidence of Primary Malignant Brain Tumors in USA,

Transcription

1 International Journal of Epidemiology International Epidemiological Association 1995 Vol. 24, No. 6 Printed in Great Britain Trends in Incidence of Primary Malignant Brain Tumors in USA, HABIBUL AHSAN,* ALFRED I NEUGUT** AND JEFFREY N BRUCE* Ahsan H (School of Public Health, Columbia University, 630 West 168th Street, New York, NY 10032, USA), Neugut A I and Bruce J N. Trends in incidence of primary malignant brain tumors in USA, International Journal of Epidemiology 1995; 24: Background. There has been considerable controversy regarding whether a recent observed rise in brain tumor incidence is real and so suggestive of increasing exposure to environmental carcinogens, or whether it is largely explainable by changes in diagnostic technology, particularly the introduction of computerized axial tomography (CT) scans in the 1970s. We analysed data from the US Surveillance, Epidemiology, and End Results (SEER) Program to investigate whether there was a rise of brain tumor incidence during , the period after CT scans became extensively available, and if so, the rates of which specific histologic subtypes have risen. Methods. Age- and sex-specific, as well as age-adjusted incidence rates were calculated for each brain tumor histologic subtype. Regression analysis was used to estimate age-adjusted incidence rate ratios (RR) for different periods and to examine time trends. Results. For all histologic subtypes, the incidence rates increased with age. Regarding time trends, lymphomas in men increased in all age groups during the study period with the age-adjusted rate ratio reaching 5.6 (95% Cl : ) for as compared to Lymphoma in women and glioblastoma multiforme in both sexes also appear to have increased, particularly in the elderly. Other histologies did not show any time trends in the age-specific or ageadjusted analyses. Conclusion. Our histology-specific analysis is not suggestive of any significant increase in the incidence of brain tumors during , except for lymphomas in men. Keywords: brain tumor, CNS lymphoma, time trend, glioblastoma multiforme, epidemiology Primary malignant brain tumors in adults (henceforth referred to as brain tumors) are of increasing concern because, while survival remains poor, a number of studies have shown a rise in its incidence over recent decades. 1 ~ A While some have considered this rise to be real, others have attributed it to the introduction of improved diagnostic methods, particularly the computerized axial tomography (CT) scans which, in the US, were first introduced in the early 1970s. 5 Changes in specific environmental risk factors that have been linked to brain tumors have been unable to explain this changing incidence rate. 5 ' 6 Two important issues regarding this rise in brain tumor incidence have not yet been investigated: 1) whether the * School of Public Health and f the Departments of Medicine and * Neurological Surgery, College of Physicians and Surgeons, Columbia University, 630 West 168th Street, New York, NY 10032, USA. Reprint requests to: Dr Alfred I Neugut, Division of Oncology, College of Physicians and Surgeons, Columbia University, 630 West 168th Street, New York, NY 10032, USA. incidence rates continued to rise during the period after CT scans became extensively available, and 2) whether this changing pattern involves all or specific histologic subtypes of brain tumor. The former would predict a slowing or disappearance in the trends with less dramatic changes in diagnostic methods. The identification of specific histologic subtypes which are rising could suggest possible aetioiogical risk factors which would need to be investigated. METHODS The US Surveillance, Epidemiology, and End Results (SEER) Program of the National Cancer Institute (NCI) has been collecting information on tumors through several population-based registries since SEER collects information on the demographics, diagnosis, pathology, treatment and survival of each patient. The data collected by SEER are known to be of high quality with about 95% of cases diagnosed microscopically. This study used the dataset 7 with information 1078

2 TRENDS IN PRIMARY MALIGNANT BRAIN TUMORS IN USA 1079 TABLE 1 Distribution of primary malignant brain tumors in adults" by sex and histology, SEER ( ) Histologic type ICD-0 Codes' Men Women No. % No. Astrocytoma Glioblastoma multiforme Oligodendroglioma Medulloblastoma Ependymoma Other gliomas Meningioma Lymphoma Nerve sheath tumors Other brain cancers Total ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) "For people aged 2=30 years. b International Classification of Diseases for Oncology.'' on about 1.6 million cases, covering approximately 10% of the US population. 8 Since CT scans became extensively available in the US by the end of the 1970s, 9 we analysed brain tumor cases, aged 30 or older, who were diagnosed after 1980 to minimize the effect of CT scans. For each histologic subtype, we calculated age- and sex-specific incidence rates for every 2-year period from 1981 to 1990 using the category specific cases and the population in each registry. The populations used as denominators are estimates for intracensal years provided by the US Census Bureau and computed by the NCI. We defined brain tumors as having International Classification of Diseases code no to and which include tumors of the brain, cranial nerves and cranial meninges. We excluded non-malignant cases, in-situ cancers and cases whose age or year of diagnosis were not known. Tumors diagnosed from autopsy reports or death certificates were also excluded in order to eliminate the effect of changes in the use of these procedures during the study period. Incidence rates were plotted against calendar years to illustrate the trend. Age-specific curves for rare histologies were not presented because of the difficulty in interpreting the unstable age-specific rates due to low numbers in several strata. However, age-adjusted analysis for these histologies was presented. We performed regression analysis to assess the statistical significance of the trend over time and also to estimate age-adjusted incidence rate ratios (RR) for each successive 2-year time period, using the rates as baseline. Since brain tumors have a low probability of occurrence we applied logistic modelling to the event frequencies relative to the population (rather than odds of disease) using 'Proc Logist' in SAS to approximate Poisson regression. 10 In situations where a rare event occurs in a large population (as in this study), the Poisson distribution approximates the binomial distribution 11 and odds ratios obtained through logistic regression modelling with count data are essentially the same as rate ratios for corresponding Poisson regression. 12 The general model used for each sex-histology category was as follows: In (n/n) = a + P, (age group) + P 2 (period) with n as the number of cases in a category, N as the estimated population in that category, age group as 1 = '30-44',... 4 = '2=75', and period as 0 = the reference period ( ), 1 = otherwise. Four dummy variables were constructed to designate five time periods. The model for a particular age group within a sex-histology category was simplified to one with only period as independent variable. We also examined the interaction between age and time period by including an age group*period term in the model, but since this interaction was not significant, the final model did not include any interaction term. RESULTS Although men had a higher total number of cancers, the distribution of different histologic subtypes was similar in both sexes (Table 1) with astrocytic glial tumors (astrocytoma and glioblastoma multiforme)

3 1080 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY comprising about 75% of the total. The main exception was malignant meningioma, which was higher in women, and lymphoma, which was higher in men. Figures 1-3 show the age-specific incidence rates of different histologic subtypes over the study period for men and women separately. We used a narrower Y-axis to depict the curves for the more uncommon histologies. Overall, incidence rates rose significantly with age for all subtypes. This effect of age on incidence was more pronounced for gliomas (glioblastoma multiforme and other gliomas). Regarding the changes over time as shown in Figure 1, incidence rates for the most common cancers (astrocytoma and glioblastoma multiforme) remained more or less the same over the study period. However, there was a non-significant rise (/' = 0.39 in men and P = 0.15 in women) of glioblastoma multiforme in the oldest age group (3=75 years). While malignant meningiomas remained stable in incidence over the study period (Figure 2), there was a dramatic rise in incidence of lymphoma, particularly in men, probably reflecting the rise in the prevalence of human immunodeficiency virus (HIV) infection and its association with central nervous system (CNS) lymphomas. 13 " 14 For lymphoma, although there was a rising trend in all age groups in men, only the rises in the and year age groups were statistically significant (P = and P = respectively). The observed rise in older women was also not statistically significant. Similarly, despite some variations in the curves of other gliomas and other brain cancers (Figure 3), particularly in the oldest age group, none of these changes attained statistical significance. The incidence rates of oligodendroglioma and ependymoma also did not change significantly (curves not shown). For each histologic subtype, the age-adjusted incidence rate ratios for each successive 2-year period compared to are shown for men and women separately (Table 2). As seen from the rate ratios, only lymphoma in men showed a clear upward trend over time (P = ) with the RR reaching 5.6 (95% CI: ) for the last 2-year period ( ). The rate ratios for glioblastoma multiforme and lymphoma in women, although higher in the late 1980s, did not show a significant trend. There was a sudden drop in the age-adjusted rate of astrocytoma in the last 2-year period ( ), for which the explanation is unknown. We also examined the age-adjusted incidence rate overall (all histologies combined) and found that the rate remained more or less unchanged over the study duration, both in men and women (not presented). DISCUSSION Several studies have demonstrated a rise in the overall incidence of brain tumors in the elderly over the last two decades both in the US and abroad. 1 " 4 Studies showing a rise have mostly examined brain tumors as a whole, although a few did look at histologic subtypes and found that the observed rise mainly affected the glioblastomas and astrocytomas. 3 ' 15 One recent study using SEER data examined histology-specific (excluding CNS lymphoma) incidence for all age groups together and compared rates between different time periods ( , , ) and found basically no differences between periods. 16 Our histologic subgroup analyses for revealed a statistically significant increase only for lymphoma in men. This increase in men may mainly be attributed to HIV infection and its association with CNS lymphomas which were also rising during the same period. In older age groups this rise cannot easily be explained by HIV and it appears that an HIV-independent rising trend of CNS lymphoma also exists in parallel. Increases in non-hodgkin's lymphoma in non-hiv populations have recently been reported. 17 We also found a non-significant rise in the incidence of CNS lymphoma in women. Although not statistically significant, there was a rising trend of glioblastoma multiforme in the elderly. Age-adjusted analysis also showed an elevated rate ratio in the late 1980s. If this rise is real it would suggest a different aetiological pathway related to this specific histologic subtype. The sudden fall in age-adjusted incidence rate of astrocytoma in the time period is unusual and difficult to explain. The findings related to rare histologic subtypes in our paper are less stable because of the small number of cases in each stratum. However, none of these histologies showed a significant trend. While restriction of our analysis to the 1980s probably eliminated the effect of changes in the use of CT scans, this study could not eliminate the effect of magnetic resonance imaging (MRI) scans which were introduced in the 1980s. MRI scans are considered more sensitive 18 in detecting brain tumors and may explain the rise in the incidence of glioblastoma multiforme seen in the late 1980s. Although this study conducted histology-specific analysis, the difficulty in histopathologic diagnosis of brain tumors, particularly of rare subtypes and distinction of continuum of astrocytoma to glioblastoma multiforme, may have resulted in some misdiagnosis. This would cause an artificially large proportion in the 'others' category and should not affect the time trends unless there was a rapid change of the histopathologic techniques in a short period. The

4 TRENDS IN PRIMARY MALIGNANT BRAIN TUMORS IN USA S-8S >-7S < >-75 >-76 FIGURE 1 Age-specific incidence ( to ) of glioblastoma multiforme in (a) men and (b) women, and of astrocyloma in (c) men and (d) women

5 1082 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY Z Age X end Above <- -«60-74» > o a >-75 >-75 FIGURE 2 Age-specific incidence ( to ) of lymphoma in (a) men and (b) women, and of malignant melanoma in (c) men and (d) women

6 TRENDS IN PRIMARY MALIGNANT BRAIN TUMORS IN USA o-»-o S9 > > > * » *-« >-75 > FIGURE 3 Age-specific incidence ( to ) of other gliomas in (a) men and (b) women, and of other brain cancers in (c) men and (d) women

7 1084 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY TABLE 2 Age-adjusted incidence rate ratios" by sex and histology Men Women Year Rate ratio 95% CI Rate ratio 95% CI Astrocytoma Glioblastoma Multifonne Oligodendroglioma Medulloblastoma Ependymoma Other Gliomas Malignant Meningioma Lymphoma Nerve Sheath Cancers Other Brain Cancers ^ " Using the rates of as reference.

8 TRENDS IN PRIMARY MALIGNANT BRAIN TUMORS IN USA 1085 relatively limited span ( ) of our study, while sufficient to examine a trend, actually makes the study less liable to be affected by factors which may change over time including histopathologic diagnostic techniques, and other factors such as attitude towards the elderly, perception of specific illnesses, and population migration. Our exclusion of cases diagnosed through autopsy reports or death certificates and the use of uniform ICD-0 diagnostic codes also eliminated effects on our results due to changes in these. Although malignant brain tumors are not very common, a rise in their incidence would be of great concern, as their prognosis is poor and our understanding about the causes of brain tumors remains unclear. A rise in the incidence of brain tumors would imply that people are being increasingly exposed to one or more aetiological risk factors. Our study, showing no significant time trends except for lymphoma in men, is mainly suggestive of the association between the rising prevalence of HIV infection and the increase in incidence of CNS lymphoma in men. A rise of CNS lymphoma in non-hiv populations and of glioblastoma multiforme in the elderly in the late 1980s is also evident in this study and this needs further research. REFERENCES 1 Preston-Martin S, Lewis S, Winkelmann R, Borman B, Auld J, Pierce N. Descriptive epidemiology of primary cancer of brain, cranial nerves, and cranial meninges in New Zealand, Cancer Causes Control 1993; 4: Garfinkel L, Sarokhan B. Trends in brain cancer tumor mortality and morbidity in the United States. Ann N Y AcadSci 1982; 381: Greig N H, Ries L G, Yancik R, Rapoport S I. Increasing annual incidence of primary malignant brain tumors in the elderly. JNatl Cancer Inst 1990; 82: Schoenberg B S, Christine B W, Whisnant J P. The descriptive epidemiology of primary intracranial neoplasms: The Connecticut experience. Am J Epidemiol 1976; 104: Boyle P, Maisonneuve P, Saracci R, Muir C S. Is the increased incidence of primary malignant brain tumors in the elderly real? J Nail Cancer Inst 1990; 82: Larsen N S. Brain tumor incidence rising; Researchers ask why. J Natl Cancer Inst 1993; 85: Surveillance, Epidemiology, and End Resuts (SEER) Program special public use tape ( ). National Cancer Institute, DCPC, Surveillance Program, Cancer Statistics Branch, November Miller B A, Ries LAG, Hankey B F et al. (eds). SEER Cancer statistics review: National Cancer Institute. NIH Pub. No , 'Modan B, Wagener D K, Feldman J J, Rosenberg H M, Feinleib M. Increased mortality from brain tumors: A combined outcome of diagnostic technology and change of attitude toward the elderly. Am J Epidemiol 1992; 135: SAS Institute Inc. The Logistic Procedure. In: SAS/STAT User's Guide, Version 6, Fourth Edition, Volume 2, Cary, North Carolina: SAS Institute Inc., 1990, pp Kleinbaum D G, Kupper L L, Muller K E. Applied regression analysis and other multivariable methods. Second edition. PWS-Kent, 1988, pp Vine M F, Schoenbach V J, Hulka B S, Koch G G, Samsa G. Atypical metaplasia and incidence of bronchogenic carcinoma. Am J Epidemiol 1990; 131: Payan M J, Gambarelli D, Routy J P et al. Primary lymphoma of the brain associated with AIDS. Ada Neuropathol 1984; 64: Rosenblum M L, Levy R M, Ziegler J L. Primary central nervous system lymphomas in patients with AIDS. Ann Weuro/1988;23:S13. l5 Velema J P, Percy C L. Age curves of central nervous system tumor incidence in adults: Variation of shape by histologic type. J Natl Cancer Inst 1987; 79: Polednak A P, Flannery J T. Brain, other central nervous system, and eye cancer. Cancer 1995; 75 (Suppl.): l7 Carli PM, Boutron M C, Maynadie M, Baily F, Caillot D, Petrella T. Increase in the incidence of non-hodgkin's lymphomas: evidence for a recent sharp increase in France independent of AIDS. Br J Cancer 1994; 70: Deck M D, Henschke C, Lee BC et al. Computed tomography versus magnetic resonance imaging of the brain. A collaborative inter-institutional study. Clin Imaging 1989; 13: "Berg i, Maguin P, Muir C et al. In: Percy C, Holten V V, eds. International Classification of Diseases for Oncology, Morphology. Field Trial Edition, Lyon: International Agency for Research on Cancer, (Revised version received May 1995)